Datasets:

kilian-group
/

arxiv-classifier

Dataset card Data Studio Files Files and versions Community

Subset (4)

default · ~99.6k rows (showing the first 120k)

Split (2)

train · ~72.4k rows (showing the first 93.1k)

paper_id stringlengths 9 16	version stringclasses 26 values	yymm stringclasses 311 values	created timestamp[s]	title stringlengths 6 335	secondary_subfield sequencelengths 1 8	abstract stringlengths 25 3.93k	primary_subfield stringclasses 124 values	field stringclasses 20 values	fulltext stringlengths 0 2.84M
1607.07861	1	1607	2016-07-26T19:50:00	Cometary ices in forming protoplanetary disc midplanes	[ "astro-ph.EP", "astro-ph.SR" ]	Low-mass protostars are the extrasolar analogues of the natal Solar System. Sophisticated physicochemical models are used to simulate the formation of two protoplanetary discs from the initial prestellar phase, one dominated by viscous spreading and the other by pure infall. The results show that the volatile prestellar fingerprint is modified by the chemistry en route into the disc. This holds relatively independent of initial abundances and chemical parameters: physical conditions are more important. The amount of CO2 increases via the grain-surface reaction of OH with CO, which is enhanced by photodissociation of H2O ice. Complex organic molecules are produced during transport through the envelope at the expense of CH3OH ice. Their abundances can be comparable to that of methanol ice (few % of water ice) at large disc radii (R > 30 AU). Current Class II disc models may be underestimating the complex organic content. Planet population synthesis models may underestimate the amount of CO2 and overestimate CH3OH ices in planetesimals by disregarding chemical processing between the cloud and disc phases. The overall C/O and C/N ratios differ between the gas and solid phases. The two ice ratios show little variation beyond the inner 10 AU and both are nearly solar in the case of pure infall, but both are sub-solar when viscous spreading dominates. Chemistry in the protostellar envelope en route to the protoplanetary disc sets the initial volatile and prebiotically-significant content of icy planetesimals and cometary bodies. Comets are thus potentially reflecting the provenances of the midplane ices in the Solar Nebula.	astro-ph.EP	astro-ph	Mon. Not. R. Astron. Soc. 000, 1–18 (2016) Printed 27 July 2016 (MN LATEX style file v2.2) Cometary ices in forming protoplanetary disc midplanes Maria N. Drozdovskaya1⋆, Catherine Walsh1, Ewine F. van Dishoeck1,2, Kenji Furuya1, Ulysse Marboeuf3,4, Amaury Thiabaud3,4, Daniel Harsono5 and Ruud Visser6 1 Leiden Observatory, Leiden University, P.O. Box 9513, 2300 RA, Leiden, The Netherlands 2 Max-Planck-Institut fur Extraterrestrische Physik, Giessenbachstrasse 1, 85748 Garching, Germany 3 Center for Space and Habitability, Universitat Bern, 3012 Bern, Switzerland 4 NCCR PlanetS - Universitat Bern, Physikalisches Institut, Universitat Bern, 3012 Bern, Switzerland 5 Heidelberg University, Center for Astronomy, Institute for Theoretical Astrophysics, Albert-Ueberle-Strasse 2, 69120 Heidelberg, Germany 6 European Southern Observatory, Karl-Schwarzschild-Strasse 2, 85748 Garching, Germany Accepted xxx. Received xxx; in original form xxx ABSTRACT Low-mass protostars are the extrasolar analogues of the natal Solar System. Sophisticated physicochemical models are used to simulate the formation of two protoplanetary discs from the initial prestellar phase, one dominated by viscous spreading and the other by pure infall. The results show that the volatile prestellar fingerprint is modified by the chemistry en route into the disc. This holds relatively independent of initial abundances and chemical parameters: physical conditions are more important. The amount of CO2 increases via the grain-surface reaction of OH with CO, which is enhanced by photodissociation of H2O ice. Complex or- ganic molecules are produced during transport through the envelope at the expense of CH3OH ice. Their abundances can be comparable to that of methanol ice (few % of water ice) at large disc radii (R > 30 AU). Current Class II disc models may be underestimating the complex or- ganic content. Planet population synthesis models may underestimate the amount of CO2 and overestimate CH3OH ices in planetesimals by disregarding chemical processing between the cloud and disc phases. The overall C/O and C/N ratios differ between the gas and solid phases. The two ice ratios show little variation beyond the inner 10 AU and both are nearly solar in the case of pure infall, but both are sub-solar when viscous spreading dominates. Chemistry in the protostellar envelope en route to the protoplanetary disc sets the initial volatile and prebiotically-significant content of icy planetesimals and cometary bodies. Comets are thus potentially reflecting the provenances of the midplane ices in the Solar Nebula. Key words: astrochemistry – stars: protostars – protoplanetary discs – comets: general. 1 INTRODUCTION Protoplanetary discs encircling young protostars are the birth loca- tions of future mature planetary systems (see Johansen et al. 2014 for a review). Past observations have suggested that dust growth from micron sizes may begin as early as the prestellar core stage and dust grains may reach millimetre dimensions during infall in the envelopes of protostars (Pagani et al. 2010; Miotello et al. 2014; Jones et al. 2016; Ysard et al. 2016, see Testi et al. 2014 for a review). Recent ALMA data have shown strongly defined ring structures in the disc around the Class I–II protostar HL Tau, which is thought to be younger than 6 1 − 2 Myr and still embedded in a large envelope (ALMA Partnership et al. 2015). Models have sug- gested that these rings may be caused by several planets of at least 0.2 MJ in mass clearing gaps in the dust distribution (Dipierro et al. 2015; Pinte et al. 2016; Dong, Zhu & Whitney 2015). An alterna- ⋆ E-mail: [email protected] tive hypothesis is that the contrast reflects enhanced dust growth to centimeter sizes behind various snowlines, leading to a change in opacity in the emitting dust (Zhang, Blake & Bergin 2015). This is potentially aided by sintering close to the snowlines of the volatiles (Okuzumi et al. 2016). These findings are pushing the onset of planetesimal formation much earlier along the star-disc evolution- ary sequence than previously thought, perhaps even as early as the embedded phase. At the same time, the exoplanet community has demon- formation (see, e.g., strated the diverse outcomes of planet www.exoplanet.eu, Schneider et al. 2011). Planet population syn- thesis modellers have carried out pioneering work in linking pro- toplanetary disc theory with the final architecture of planetary systems. Such models take the key physical processes across all these evolutionary stages into account, and with large sets of initial conditions, to make statistical predictions on the ex- oplanet population (see Benz et al. 2014 for a review). It has been postulated that planetary atmospheres form initially via pebble accretion and heating from this assembly prior to run- 2 Maria N. Drozdovskaya et al. away gas accretion (Inaba & Ikoma 2003; Ormel & Klahr 2010; Bitsch, Lambrechts & Johansen 2015). Based on the newest results from the star formation community, it may be necessary to use the initial conditions from the earlier embedded phase rather than the classical T Tauri (Class II) stage. Typically, the constituents of planetesimals that feed proto- planets are grouped into volatiles (H, O, C, N and S containing molecules) and refractories/rocks (minerals/inorganic refractories and complex organic refractory components such as PAHs and macromolecular complex organic matter). For regions of a pro- toplanetary disc where most volatiles are frozen out as solids, an ice/rock mass ratio of ∼ 2 − 4 is suggested (Pontoppidan et al. 2014). Various instruments aboard the Rosetta mission are at- tempting to constrain the the icy and dusty contents of comet 67P/Churyumov-Gerasimenko to an unprecedented precision. The RSI experiment indicates a ice/dust mass of ratio of ∼ 4 based on the gravity field of the comet (Patzold et al. 2016). CONSERT measurements of an ice/dust volumetric ratio of ∼ 0.4 − 2.5 (Kofman et al. 2015) imply a ice/dust mass ratio of ∼ 0.1 − 0.9, assuming an average dust density of 2 600 kg m−3 and an ice den- sity of 940 kg m−3 as in Patzold et al. (2016). These ratios mean that the solid-phase chemical composition is a pivotal parameter for the subsequent protoplanet makeup. The density is highest in the midplane of a protoplanetary disc, thus harbouring the bulk of the mass and implying that predominantly the ices in that region will likely shape the chemical composition of the atmospheres of the exoplanets that are observed today. The chemical composition of discs in the embedded phase has been probed by models of Visser et al. (2009), Visser, Doty & van Dishoeck (2011), Drozdovskaya et al. (2014), Harsono, Bruderer & van Dishoeck (2015). Observational evi- dence for them is scarce, because high spatial resolution (. 30 AU) is necessary to get spatial information for a small (∼ 100 AU) target,whose emission is easily overwhelmed by that of the more massive envelope, making it hard to constrain their physical parameters. Only recently, observations have shown em- bedded Keplerian discs of ∼ 50 − 300 AU in radius in embed- ded protostars (Tobin et al. 2012, 2013, 2015; Brinch & Jørgensen 2013; Murillo et al. 2013; Sakai et al. 2014; Harsono et al. 2014; Chou et al. 2014). Instead, significant modelling effort has focused on discs in the Class II/III stages of star formation (see, e.g., Henning & Semenov (2013) and table 1 in Woitke et al. 2016 and the references therein). The chemistry in such Class II discs has been modelled by means of gas-grain models with grain-surface reactions, e.g., Willacy et al. (2006); Semenov, Wiebe & Henning (2006); Hersant et al. (2009); Semenov et al. (2010); Henning et al. (2010); Walsh, Millar & Nomura (2010); Walsh et al. (2012); Vasyunin et al. (2013). Only recently complex organics (loosely defined in both chemistry and as- tronomy as large, > 6 atoms, carbon-containing species, Herbst & van Dishoeck 2009) have also been considered by Semenov & Wiebe (2011), Furuya & Aikawa (2014), Walsh et al. (2014b), Walsh et al. see also Henning & Semenov (2013) for a review. Meanwhile, such molecules have been fre- quently observed towards protostars and in Solar System bodies (Herbst & van Dishoeck 2009). (2011); Akimkin et al. (2014a), The aim of this work is to compute the chemical composi- tion of planetesimals and cometary materials in the midplanes of protoplanetary discs in the embedded phase of star formation. This is done using state-of-the-art physicochemical models that are dy- namic in nature and include chemical kinetics, building upon the work in Drozdovskaya et al. (2014), which studied the discs as a whole. Main volatiles that are expected to dominate the icy com- ponent of plantesimals in the midplanes of protoplanetary discs are considered first. Trace complex organic species are also analysed in this work, since they are potentially the earliest precursors to prebiotic molecules. The dynamic nature of the model allows the study of the chemical effects stemming from the transport from the prestellar cloud through the envelope and into the disc. The key model features are highlighted in Section 2. The results are shown in Section 3 and are discussed in Section 4, including all the derived implications for the population synthesis models and the Solar Sys- tem community. The conclusions are presented in Section 5. 2 MODELS Models simulating the formation of protostars typically start from the so-called prestellar core/cloud phase, which undergoes spher- ical collapse. Cloud rotation leads to the formation of a proto- planetary disc and associated bipolar outflows due to the conserva- tion of angular momentum. This work and the preceding publica- tion by Drozdovskaya et al. (2014) use the axisymmetric, 2D semi- analytic physical model developed by Visser et al. (2009) (and Visser & Dullemond 2010; Visser, Doty & van Dishoeck 2011; Harsono et al. 2013), which describes the density and velocity of the collapsing material. Wavelength-dependent radiative trans- fer calculations are performed with RADMC-3D2 to compute the dust temperature and far-ultraviolet radiation field (FUV; 912 − 2066 A; 6.0 − 13.6 eV). Thereafter, the physical pa- rameters are coupled with a large chemical network in the framework of a two-phase (gas and solid) model solved using the rate-equation method, including grain-surface chemistry and the chemistry of several families of complex organic molecules (McElroy et al. 2013; Garrod & Herbst 2006; Walsh et al. 2014a,b; Walsh, Nomura & van Dishoeck 2015). Initial atomic abundances are evolved under constant prestellar cloud conditions (nH = 4 × 104 cm−3, Tdust = 10 K and cosmic ray-interaction with H2 as the only source of FUV photons; assuming a cosmic ray ionisation rate ζ0 = 5 × 10−17 s−1) for 3 × 105 yr (table 2 of Drozdovskaya et al. 2014) to obtain initial molecular abundances at the onset of col- lapse (Table 1). Chemical abundances are computed along evolu- tionary infall trajectories that are obtained from the physical model with path-dependent physical conditions. The collapse of the sys- tem takes an additional 2.46 × 105 yr, which corresponds to the time it takes for the envelope to accrete on to the star-disc system for the initial physical conditions chosen for this work. In this paper, as in Drozdovskaya et al. (2014), two discs are studied. One is the so-called spread-dominated case, with a lower initial cloud rotation rate, in which the disc primarily grows via viscous spreading and has a final outer radius of ∼ 50 AU. The second is the infall-dominated case, in which the disc is assem- bled predominantly via pure infall of material and has a final outer radius of ∼ 300 AU (table 1 of Drozdovskaya et al. 2014). The rea- son for studying both discs is that their unique large-scale dynamics result in different dust temperature and FUV flux profiles along the trajectories, i.e., they have different physical histories for the pro- toplanetary disc material (this is pictorially illustrated in Fig. 1). At ∼ 40 K, the smaller spread-dominated disc is warmer than the more massive infall-dominated disc, which is ∼ 20 K close to the mid- plane (middle panels of figs. 4 and 5 of Drozdovskaya et al. 2014). 2 http://www.ita.uni-heidelberg.de/dullemond/software/radmc-3d/ Cometary ices in forming disc midplanes 3 Table 1. Molecular abundances at the end of the prestellar phaseaand their binding energiesb. Species Name H2O CO CO2 NH3 CH4 CH3OH H2S N2 H2CO CH2CO HCOOH HCOOCH3 CH3CHO CH3OCH3 C2H5OH CH3COOH HOCH2CHO glycolaldehyde water carbon monoxide carbon dioxide ammonia methane methanol hydrogen sulfide nitrogen formaldehyde ketene formic acid methyl formate acetaldehyde dimethyl ether ethanol acetic acid n (Xgas) /nH 7.7 × 10−8 1.4 × 10−5 7.9 × 10−8 2.1 × 10−7 1.2 × 10−7 2.0 × 10−10 5.3 × 10−10 1.3 × 10−5 9.7 × 10−9 4.6 × 10−10 2.4 × 10−11 1.1 × 10−10 1.0 × 10−9 2.4 × 10−11 2.2 × 10−11 2.2 × 10−17 2.2 × 10−12 n (Xice) /nH 1.9 × 10−4 5.9 × 10−5 1.9 × 10−5 4.8 × 10−6 1.5 × 10−5 3.7 × 10−6 8.1 × 10−9 1.1 × 10−5 2.3 × 10−6 8.0 × 10−8 4.9 × 10−11 6.4 × 10−14 1.5 × 10−6 9.2 × 10−8 4.7 × 10−8 9.7 × 10−14 5.3 × 10−9 n (Xice) /n (H2Oice) Edes(X) (K) 1.0 × 100 3.1 × 10−1 1.0 × 10−1 2.5 × 10−2 7.9 × 10−2 1.9 × 10−2 4.3 × 10−5 5.8 × 10−2 1.2 × 10−2 4.2 × 10−4 2.6 × 10−7 3.4 × 10−10 7.9 × 10−3 4.8 × 10−4 2.5 × 10−4 5.1 × 10−10 2.8 × 10−5 5773c 1150d 2990e 5534d 1090f 4930g 2743d 790h 2050i 2200i 5000j 4000j 3800j 3300j 5200j 6300j 6680k a The constant physical conditions of the prestellar phase are: nH = 4 × 104 cm−3, Tdust = 10 K and only cosmic ray-induced FUV photons, ζ = 5 × 10−17 s−1. The duration is taken to be 3 × 105 yr. b The differences between this table and table 2 of Drozdovskaya et al. 2015 are due to this work including the glycolaldehyde formation via glyoxal, reaction-diffusion competition, a coverage factor for dust grains (limiting the number of reactive layers) and a higher barrier thickness between grain surface sites. c Fraser et al. (2001) - measurement for pure water ice in the thin films (≪ 50 monolayers) regime (within errors of measurements of Collings et al. (2015) for pure water ice in the multilayer regime) d Collings et al. (2004); Garrod & Herbst (2006) - measurement for CO on H2O ice (to account for the deep trapping in the inner layers seen with multilayer models of (Taquet, Ceccarelli & Kahane 2012) and for the trapping in pores of amorphous water ice seen in the experiments of Collings et al. (2003b)); and measurements for NH3 and H2S on H2O ice e Edridge (2010) - measured average for pure carbon dioxide ice in the multilayer regime f Herrero et al. (2010) - measurement for CH4 on H2O ice (to account for the diffusion of methane molecules into the micropores of amorphous solid water (ASW) seen in these experiments) g Brown & Bolina (2007) - measurement for pure methanol ice in the multilayer regime h Oberg et al. (2005) - measurement for pure N2 ice in the multilayer regime (within errors of measurements of Bisschop et al. (2006), Collings et al. (2015), and Fayolle et al. (2016) for 15N2, for pure ices in the multilayer regime) i Garrod & Herbst (2006) j Oberg et al. (2009) - measurements for pure ices in the multilayer regime k Garrod (2013) A larger disc mass hinders passive heating by reprocessed stellar ra- diation. These models exclude viscous heating and its effects, such as those discussed in (Harsono, Bruderer & van Dishoeck 2015); however, this is most important only very close to the star. These models do not consider magnetic field effects and dust growth, as discussed in, e.g., (Zhao et al. 2016) and (Pauly & Garrod 2016). Once collapse is initiated, the central protostar is the dom- inant source of FUV photons. Its luminosity, which varies with time and peaks at ∼ 11 L⊙ for the spread-dominated case (fig. 2 of Drozdovskaya et al. 2014), is given by the sum of the accre- tion luminosity (Adams & Shu 1986) and the photospheric lumi- nosity (due to gravitational contraction and deuterium burning; D'Antona & Mazzitelli 1994; Visser et al. 2009). Excess UV ob- served for classical T Tauri stars (effective temperatures of ∼ 10 000 K) as a result of boundary layer accretion from the disk onto the star (Bertout, Basri & Bouvier 1988) is not accounted for in this work. External sources of FUV are excluded, since such young ob- jects are expected to be deeply embedded in molecular clouds and hence highly shielded. The maximal FUV flux experienced by a parcel is set by the 1/r2 dependence of the flux and the corre- Figure 1. An illustration of the key parameters (such as the midplane dust temperature, outer radius and predominant parcel motion) of the two discs that are central to this work and Drozdovskaya et al. (2014). 4 Maria N. Drozdovskaya et al. sponding decrease in the attenuating column as the parcel moves closer to the star. Once the parcel enters the disk, the attenuating column increases substantially, leading to an almost instantaneous reduction in FUV flux. The material entering the spread-dominated disc experiences higher peak FUV irradiation, because many of the parcels pass closer to the protostar prior to disc entry than in the infall-dominated case. The smaller disc also exerts less attenua- tion of cosmic rays, thus there are less internally produced cosmic ray-induced FUV photons in the infall-dominated disc. Both, the dust temperature and FUV flux, are crucial parameters for chemi- cal grain-surface reactions controlling the mobility and availability of radicals. All further details on the physical and chemical model and the radiative transfer setup can be found in Drozdovskaya et al. (2014, 2015). In this work, a different set of parcels is consid- ered - more than 100 for each disc, all with final positions at z = 0.01 AU and R in the ∼ 1 − 50 AU range for the spread- dominated disc (∼ 1 − 300 AU range for the infall-dominated disc) to better sample the two midplanes. For the spread-dominated case at R = 5.2 AU, ∼ 90 per cent of the mass lies below 1 AU in height, while the disc surface extends up to ∼ 2 AU. This work includes the experimentally suggested glycolaldehyde formation via glyoxal (Fedoseev et al. 2015, as described in sec- tion 3.4.3 of Drozdovskaya et al. 2015). Furthermore, diffusion- reaction competition is now accounted for via the formulation of Garrod & Pauly (2011) (equation 6). This enhances the probability of reactions with activation barriers (EA) when thermal diffusion rates are low. A larger 1.5 A barrier thickness (a) between grain surface sites is assumed to decrease the efficiency of quantum tun- nelling reactions involving atomic and molecular hydrogen. The ef- fects of varying the barrier to quantum tunneling are investigated in Section 3.2.1. Furthermore, a 'coverage factor' of 2 ice monolayers has been imposed. This parameter implies that grain-surface associ- ations can occur only in the two upper most monolayers. However, in a two-phase model the surface is not distinguished from the bulk in terms of composition; therefore, the composition of these two chemically active layers is averaged together with that of the non- chemically active bulk. Finally, since gas column densities at po- sitions along the midplane mostly exceed 100 g cm−2, cosmic ray attenuation is now accounted for according to ζ = ζ0 exp(cid:18) −Σ(R,z) 96 g cm−2(cid:19) , (1) z that µmuR surface where Σ(R,z) is the column density between the position of time (Σ(R,z) = interest (R, z) and the disc surface at ngasdz, where µ = 2.3 is the mean molecular mass of the gas, mu is the atomic mass unit, ngas is the gas parti- cle number density; fully molecular gas is assumed for simplic- ity, i.e., ngas = nH2 and nH = 2nH2). A lower limit of 7.3 × 10−19 s−1 is imposed, which stems from 26Al radionuclide decay (Umebayashi & Nakano 1981; Cleeves, Adams & Bergin 2013). 2.1 Initial conditions: 'hot' disc start versus 'cold' cloud start scenarios The typical initial assumption of the solar system community and population synthesis models is that the initial conditions of the So- lar Nebula are hot, i.e., all disc material (whether volatile or refrac- tory) is in the gas phase. The disc is thought to be small, dense and hot (heated by viscous dissipation, initially), and subsequently to expand in radius and cool. Typically, to bypass this evolution, the initial temperature is simply assumed high (≫ 300 K) out to at least 30 AU. For planet population synthesis statistics, disc mod- els provide radial pressure, temperature and surface density profiles (Alibert et al. 2005, 2013; Marboeuf et al. 2014b,a; Thiabaud et al. 2014, 2015). Individual volatiles are then allowed to condense in- stantaneously at their respective temperature-pressure of condensa- tion (and also be trapped in clathrates, so-called water cages that form from crystalline water ice, the importance of which is still under debate, e.g., Lunine & Stevenson 1985; Mousis et al. 2012; Thurmer et al. 2015), assuming thermodynamic equilibrium. This forms the ice and rock components of small grains as a function of disc radius (R). It is assumed that the composition of 300 m plan- etesimals at a certain R in the disc is identical to that of the small grain composition. The population synthesis model then describes how 10 planetary Lunar-sized embryos (of the same composition as the planetesimals at a certain R) grow through planetesimal accre- tion during migration, while the disc evolves and disperses by pho- toevaporation. By running a very large number of disc models with varying initial conditions, general trends in planet populations are studied (Marboeuf et al. 2014b,a; Thiabaud et al. 2014). No chem- istry occurs in these models; variations in the ice composition are set by the locations of snowlines (condensation fronts) at the start (t = 0). In contrast, the chemical models of the star formation com- munity include a kinetic chemical network and many, if not all, of the possible chemical processes at play. The initially hot Solar Nebula ('hot' disc start scenario) ver- sus the cold prestellar phase and warm protoplanetary disc ('cold' cloud start scenario) is one of the major differences between the approaches of the Solar System and star plus disc formation com- munities. Under the 'cold' start scenario, a large icy reservoir is initially in place, which is not necessarily mimicked by the gas. Meanwhile, for the 'hot' start, only gases are available at the start. It remains unclear if potentially both scenarios are valid in dif- ferent regions of the disc (see the review of Davis et al. 2014). Calcium-aluminium-rich inclusions (CAIs) are typically presented as evidence for the 'hot' scenario, since their production occurs at temperatures > 1300 K. CAIs are found in meteorites, which mostly originate from the asteroids in the inner Solar System, but have also been detected in comet 81P/Wild 2 with the Star- dust mission (Brownlee et al. 2006). High temperature-processing is also suggested by the presence of crystalline silicates in proto- planetary discs and comets, which can only form by annealing at temperatures of 800 − 1000 K or via condensation upon cooling below ∼ 1200 − 1500 K. However, all observations of prestel- lar and protostellar sources point towards the 'cold' and 'warm' set of conditions. Localized temperature enhancement could occur via accretion shocks at the disc surface or via episodic accretion with subsequent spreading of crystalline silicates by radial or ver- tical large-range mixing throughout the protoplanetary disc (e.g., Visser & Dullemond 2010; Ciesla 2011). It is still debated whether either of these processes can reach high enough temperatures, but if so, strongly heated materials would be available at the largest disc radii for incorporation into distant larger bodies. The models of Marboeuf et al. (2014a,b) and Thiabaud et al. (2014, 2015) impose initial gas-phase abundances based on obser- vations of prestellar cores and protostellar sources, assuming com- plete inheritance without modification of the prestellar icy compo- sition by discs upon formation. In other words, they are a hybrid of the two scenarios just discussed. This is visually summarized in Fig. 2. This work tests the simplified approach of phase dia- grams under the assumption of complete prestellar inheritance to obtain the volatile composition of discs by comparing with the out- put obtained from a full kinetic chemical model, which also takes Cometary ices in forming disc midplanes 5 Figure 2. An illustration of the 'hot' and 'cold' start perspectives. The results obtained with this physical model in various publications, including this work, are also summarized. spread−dominated infall−dominated 10000 10000 1000 ) 2 − m c g ( Σ Σ Tdust nH P 100 10 1016 1014 1012 1010 108 106 104 0 0 10 10 20 20 30 30 40 40 R (AU) ) 3 − m c ( H n T d u s t ( K ) P ( b a r ) 1000 100 10 10−2 10−4 10−6 10−8 10−10 50 50 Figure 3. The midplane (z ≈ 0.01 AU) physical conditions of the spread- (solid lines) and infall-dominated (dashed lines) discs at the end of the simulation (2.46 × 105 yr after the precollapse phase): dust temperature (Tdust), H nuclei number density (nH), surface density (Σ) and pressure (P ) as a function of disc radius (R). The infall-dominated disc extends out to 300 AU, only the inner 50 AU is shown. dynamic collapse into account. The focus lies on the disk beyond the inner few AU. 3 RESULTS In Fig. 3, the midplane conditions for both the spread- and infall- dominated discs are shown (disc slice at z ≈ 0.01 AU). These are the physical conditions that the parcels reach at the final timestep of the evolution, depending on their final radial coordinate. density (Σ = 2µmuR surface The inner few AU of each midplane, close to the protostar, has the highest final dust temperature (Tdust > 200 K) and den- sity (nH > 1014 cm−3). The smaller spread-dominated disc tends to be warmer (∼ 40 K for R > 25 AU) than the more massive infall-dominated disc (∼ 20 K for R > 30 AU) due to more efficient passive heating. For R & 25 AU, the density, surface 0.01 AU ngasdz, where the initial factor of 2 accounts for the full vertical extent of the disc) and pressure (P = kBngasTgas, where kB is the Boltzmann constant and assum- ing Tgas = Tdust) of the midplane of the infall-dominated disc are higher than that of the spread-dominated disc. The surface den- sity typically exceeds 100 g cm−2 (except for R > 40 AU for the spread-dominated case) justifying the need for cosmic ray attenua- tion. The midplane pressure of each disc never exceeds 0.5 millibar (beyond 5 AU: P < 10−7 bar). 3.1 Dominant simple ices The left panels of Figs. 4a and 4b show the radial distribution of simple ices and gases in the inner 50 AU in an embedded disc for the spread-dominated and infall-dominated cases, respectively, at the final timestep of the evolution (2.46 × 105 yr since the on- set of collapse). A number of snowlines/icelines are seen in both discs, e.g., a steep ice-to-gas transition for water (H2O) and a more gradual transition for methane (CH4). The sharpness of the transi- tions is set by the temperature gradient along the midplane of each disc, which is steeper closer to the protostar (Fig. 3). The order of the snowlines is set by the binding energies used in the chemical 6 Maria N. Drozdovskaya et al. H2O CO CH2CO solid lines: ices dashed lines: gases H2CO HCOOH 10−4 10−4 10−6 10−6 10−8 10−8 10−10 10−10 10−4 10−6 10−8 H n / ) X ( n N2 10−10 10−4 10−6 10−8 10−10 10−4 10−4 10−6 10−6 10−8 10−8 10−10 10−10 10−4 10−6 10−8 H n / ) X ( n H2O CO CO2 CH4 NH3 CH2CO solid lines: ices dashed lines: gases H2CO HCOOH CH3CHO HCOOCH3 CH3OCH3 10−10 10−4 10−6 10−8 10−10 N2 CH3OH H2S HOCH2CHO CH3COOH C2H5OH CO2 NH3 CH4 HCOOCH3 CH3OCH3 CH3CHO CH3COOH C2H5OH HOCH2CHO CH3OH H2S 0 10 20 30 40 50 R (AU) 10 20 30 40 50 0 10 20 30 40 50 R (AU) 10 20 30 40 50 (a) Spread-dominated disc (b) Infall-dominated disc Figure 4. The midplane abundances of volatiles (left column of each subfigure) and complex organics (right column of each subfigure) relative to the number of H nuclei (nH) as a function of disc radius (R) in the gaseous (dashed lines) and solid (solid lines) phases at the end of the simulation (2.46 × 105 yr after the onset of collapse). Ice abundances beyond the respective snowlines (Table 2) are tabulated in Appendix A. Table 2. Snowlinesl in the midplanes of the spread- dominated and infall-dominated discs at the end of the simulation (2.46 × 105 yr after the onset of col- lapse). Species Rsnow (AU) spread-dom. infall-dom. 3 21 7 3 23 3 8 36 11 8 3 4 5 6 3 3 3 H2O CO CO2 NH3 CH4 CH3OH H2S N2 H2CO CH2CO HCOOH HCOOCH3 CH3CHO CH3OCH3 C2H5OH CH3COOH HOCH2CHO 3 > 50 7 3 > 50 4 7 > 50 20 9 4 5 7 7 3 3 3 l Defined as the point when n (Xice) /n (Xgas) = 0.5 and rounded to the nearest AU. Note that these do not always correspond to a snowline in the traditional sense, for example, see the curve of HCOOH in the spread-dominated disc in Fig. 4a, which is set by chemistry. network (right column of Table 1). For both discs, the water snow- line lies around ∼ 3 AU. The small spread-dominated disc lacks CH4 and nitrogen (N2) ices, and is low on carbon monoxide (CO) ice; while the infall-dominated disc is rich in all simple ices for R & 25 AU due to its overall lower temperature. Table 2 tabulates all the snowlines in both discs. Figs. 5a and 5b show pie charts for three radial ranges, illus- trating the percentage contribution of each species to the ice mantle. The values are number averages and are calculated according to n (Xice) PX n (Xice) (%) = 100% × PRmax n (Xice)(R,z) dR PRmax Rmin PX n (Xice)(R,z) dR Rmin , (2) which accounts for radial ranges dR between individual parcels, and where Rmax and Rmin are the maximum and minimum posi- tions, respectively, of the radial range considered. The pie charts show that typically midplane ice is ∼ 70 − 80% H2O, ∼ 10 − 20% carbon dioxide (CO2), 1 − 10% ammonia (NH3) and 1 − 4% methanol (CH3OH). As mentioned above, CO, CH4 and N2 ices are only available in the cooler infall-dominated disc and are no more than a few per cent of the total ice content. However, this only holds as long as these species are not mixed with H2O ice (see the discussion on trapping in Section 3.2.2), which would allow these molecules to be retained in the ice at a level of a few per cent. The contributions of various molecules show some radial dependencies, which are a reflection of certain species thermally desorbing at the respective snowlines along the midplane. Other ices, like complex organics and radicals make up no more than ∼ 10% of the total. For the infall-dominated disc, in Figs. 4b and 5b it can be seen that for R > 30 AU, CO2 becomes more abundant than H2O ice. This comes from a delicate balance of rehydrogenation of the OH radical (leading to H2O) and its reaction with CO on the Cometary ices in forming disc midplanes 7 (a) Spread-dominated disc (b) Infall-dominated disc (c) Infall-dominated disc beyond 30 AU, as ob- tained with the three-phase chemical model of Furuya et al. (2015), for bulk and surface com- bined with and without the swapping mecha- nism Figure 5. The percentage contribution of volatiles to the total ice content averaged over various radial ranges of the disc midplanes at the end of the simulation (2.46 × 105 yr after the precollapse phase). Only contributions of at least ∼ 1% are shown. The 'other' category includes all ices that are not the following eight volatiles: H2O, CO, CO2, CH3OH, CH4, N2, NH3 and H2S. grain surface (producing CO2 with an activation barrier of 400 K, Noble et al. 2011) according to the following scheme. E A = 4 0 0 K + CO H2Oice + hν EA = 0 OH + H CO2 + H This effect comes into play only in the outer regions of the infall- dominated disc, where CO ice is available. The initial photodisso- ciation of H2O ice, producing the OH radical, stems from the expo- sure to stellar FUV photons in the envelope en route to the disc and partially from cosmic ray-induced FUV photons. It appears that the route towards CO2 is favoured; and it is efficiently produced at the expense of H2O and CO. This is clearly demonstrated when look- ing at the individual trajectories in detail, as in Fig. 6, which shows the physical parameters and chemical abundances as a function of time for a parcel that ends up at R ≈ 47 AU along the midplane. The increase of CO2 ice follows the profile of the FUV flux and is maximized once the FUV flux spikes at ∼ G0 (where G0 is the in- terstellar FUV radiation field of 1.6 × 10−3 erg cm−2 s−1, Habing 1968) immediately prior to disc entry. This effect persists under a static assumption (Section 3.4.2), but to a lesser degree, thus dis- playing the subdominant importance of cosmic ray-induced FUV photons towards the H2O ice photodissociation process. Such ef- ficient CO2 production was already noticed in the molecular layer at intermediate heights of Class II protoplanetary disc models of Walsh et al. (2014b) (which also includes a discussion on the un- certainties in this reaction) and in the relatively cool disc around an M dwarf star in Walsh, Nomura & van Dishoeck (2015). 3.2 Sensitivity of results to chemical parameters 3.2.1 Diffusion and quantum tunneling barriers To verify the robustness of the predicted CO2 ice-rich zone in the infall-dominated disc, additional parameter tests were run to test Table 3. Abundances ratios of selected solid species at the end of the simulation (2.46 × 105 yr after the precollapse phase) for a parcel in the infall-dominated disc with final R ≈ 50 AU with three different sets of parameters in the chemical model. Species Fiducialg Ediff = 0.5Edes a = 1 A CO2/H2O 1.04 CO/H2O 8.3 × 10−8 1.28 1.3 × 10−3 2.76 4.5 × 10−3 g In the fiducial setup, Ediff = 0.3Edes and a = 1.5 A. the sensitivity of our results to the assumed grain-surface parame- ters. In the fiducial setup of our models, the diffusion barrier (Ediff) is related to the binding energy (Edes, also called the desorption en- ergy) via Ediff = 0.3Edes (Hasegawa, Herbst & Leung 1992). The ratio between the two energies is uncertain. Off-lattice Monte Carlo simulations predict a value of ∼ 0.3−0.4 for CO and CO2 on amor- phous water ice (Karssemeijer & Cuppen 2014; Karssemeijer et al. 2014), while experiments suggest a range of ∼ 0.5 − 0.9 for O and N on amorphous water ice (Minissale, Congiu & Dulieu 2016); however, it is not excluded that this ratio may differ between molecules and atoms. Astrochemical models typically employ val- ues in the ∼ 0.3 − 0.8 range (e.g., Vasyunin & Herbst (2013). For the first test, the factor between the two energies is set to 0.5 to in- crease the diffusion barrier, thereby impeding radical-radical asso- ciations on grain surfaces and limiting the efficiency of the CO+OH surface reaction. For the second test, the fiducial barrier thickness of 1.5 A is decreased, leading to a thinner barrier for quantum tun- neling of H and H2 between grain surface sites and, thus, the pro- motion of the OH+H surface reaction. Both tests were computed for the same parcel with a final ra- dial coordinate R ≈ 50 AU at the end of the simulation (in the infall-dominated disc; this is not the same parcel as that shown in Fig. 6). The same initial molecular abundances as computed with fiducial parameters were used for all tests. Table 3 displays the fi- nal abundance ratios of the key species in relation to grain-surface chemistry of CO2. For a higher diffusion barrier, the abundances of CO and radicals such as OH and HCO are higher in the solid phase, since radical mobility is reduced. However, an increase by 8 Maria N. Drozdovskaya et al. Class 0 cloud interstellar UV radiation field 1012 1010 ) 3 - m c ( H n 108 FUV spike FUV H2O CO CO2 106 10-3 10-4 10-5 10-6 10-7 H n / ) X ( n 10-5 10-6 10-7 Class I evolved disk 50 102 100 40 ) ( K t s u d T ) 0 10-2 G ( V U F 10-4 10-6 20 10 nH Tdust solid lines: ices dashed lines: gas HCOOH 10-8 HOCH2CHO 10-9 10-10 HCOOCH3 2.0×105 1.5×105 1.0×105 tacc-t (yr) 5.0×104 Figure 6. The physical parameters and chemical abundances as a function of time (t) for one trajectory (with a final radial position of ∼ 47 AU) in the case of the infall-dominated disc. The upper panel shows the H nuclei num- ber density (nH), the dust temperature (Tdust) and the FUV flux (FUV, in units of G0, which is the interstellar FUV radiation field of 1.6 × 10−3 erg cm−2 s−1, Habing 1968). The middle panel shows the abundances rel- ative to nH of water, carbon monoxode and carbon dioxide in the solid and gasesous phases (solid and dashed lines, respectively). The lower panel displays the abundances of formic acid, methyl formate and glycolalde- hyde. tacc is the accretion time of 2.46 × 105 yr in this model run (see Drozdovskaya et al. 2014 for details). Evolution is from left to right. Note the transformation of CO to CO2 ice and the build up of complex organics in the envelope upon the increase of FUV flux. ∼ 2 orders of magnitude in HCO abundance leads to more CO2 ice formation via the atom-radical O+HCO reaction. For a thinner barrier for quantum tunneling, less atomic H is available on the grain, since it is easier for it to react with other species. This in turn reduces the availability of OH, reducing the amount of CO2 ice produced via OH+CO. However, atomic H is not only used to re- form H2O, but to also hydrogenate other species, thereby reducing the amount of H2O ice as well. The net result is that the ice re- mains CO2-rich. Thus such a CO2 ice-rich zone appears character- istic of the outer parts of the infall-dominated disc independent of the exact parameters used in the chemical model within the frame- work of the two-phase model considered here (although they do change the abundances quantatively). In fact, two additional path- ways to CO2 have recently been experimentally verified, namely CO+O with EA = 1000 K and H2CO+O with EA = 335 K (Minissale et al. 2013, 2015). Currently, they are not included in the chemical network, but are expected to positively contribute to the CO2 production. 3.2.2 Binding energies The availability of specific ices in midplanes is set by their ini- tial abundances, the temperature structure of the disc, and also by the binding energies of these species (Table 1). This is why in the warmer spread-dominated disc, CH4 and N2 ices are ab- sent, and CO ice is low (Section 3.1, Fig. 4a). The adopted set of binding energies also sets the exact locations of the snowlines. In this work, the binding energies for molecules on amorphous wa- ter ice have been adopted for CO, NH3, H2S and CH4. For the larger complex organic species and methanol, the binding energies for pure ices have been used. The binding energies of pure ices tend to be lower than those on H2O ice, which is the dominant ice constituent (Fig. 5a). For example, for pure CO ice in the multi- layer regime Edes = 878 K (Collings et al. 2015, which is within Oberg et al. (2005), Bisschop et al. errors of measurements of (2006), Cleeves et al. (2014) and Fayolle et al. (2016) for pure ices in the multilayer regime), while for CO on H2O ice Edes = 1150 K (Collings et al. 2004; Garrod & Herbst 2006 or 1320 K, Cleeves et al. 2014). In the laboratory, thermal desorption of CO ice deposited on top of H2O ice is a multi-step process (Sandford et al. 1988; Ayotte et al. 2001; Collings et al. 2003b,a; Bisschop et al. 2006). At the cold (10 − 15 K) temperatures of prestellar cores, CO ice forms on top of the H2O ice mantle (Pontoppidan 2006; Oberg et al. 2011); and the layers are frozen in place, since the mobility of such large species is minimal at these temperatures. Upon heating, as much as 20% of CO ice can get trapped in amor- phous H2O ice as it undergoes phase transitions (Collings et al. 2003a). This implies that a single desorption energy is an oversim- plification of the process of thermal desorption of CO. Visser et al. (2009) (based on the approach of Viti et al. 2004) attempted to account for this effect by considering four flavours of CO with four different desorption energies, which allowed the authors to preserve ∼ 15% of CO ice in a disc at the end of the simula- tion. Unfortunately, such an approach is not feasible in this work due to a much larger size of the chemical network and the inclu- sion of grain-surface reactions (thus making reaction-dependent CO-'flavouring' non-trivial). Deep trapping of CO in the inner layers is seen as a natural consequence in the multilayer models (Taquet, Ceccarelli & Kahane 2012). Similar trapping mechanisms have also been seen ex- perimentally for other volatiles, such as N2, CH4, CO2, O2, Cometary ices in forming disc midplanes 9 ment (see the spike in Fig. 6). Since the two-phase model assumes two chemically active monolayers of the same composition as the entire mantle, reactants which may not be available in the three- phase model are available in the two-phase model. Thus, the results tend towards each other when swapping is included. The two-phase model produces even more CO2 (an additional 7%) than under the assumption of a marginally reactive bulk in the swapping-on case. Currently, there is no consensus on the chemical reactivity of the bulk ice mantle and further experimental verification is necessary. 3.3 Trace complex organic ices Although simple ices dominate the ice composition, there is also a variety of complex organic ices present. Table 4 quantifies the abundance of such ices relative to methanol ice for both discs and in three radial ranges. Some complex organics may become close in abundance to methanol ice, i.e., at a level of up to a few per cent of the total ice mantle. Such trace species could potentially have prebiotic implications for the synthesis of amino acids by the Strecker mechanism (e.g., Fresneau et al. 2014, 2015). The right panels of Figs. 4a and 4b show the radial distribution of com- plex ices and gases in the inner 50 AU around a protostar for the spread-dominated and infall-dominated discs, respectively, at the final timestep of the evolution. Similar to the situation with sim- ple ices, snowlines of complex organics are also recovered in the midplanes of both discs. Most are clustered near the water and methanol snowlines, as complex organics are tightly bound species (i.e., all those shown in the right panels of Figs. 4a and 4b except formaldehyde (H2CO) and ketene (CH2CO)). For both discs, complex organic ices are most plentiful in the outer regions (R > 30 AU). In the infall-dominated case, complex organic ices survive closer to the protostar and as far in as their re- spective snowline. In the spread-dominated case, complex organic ices start decreasing around R ∼ 25 AU (except for ethanol and acetic acid). Complex organic ice formation is facilitated by weak FUV irradiation (no more than a few orders of magnitude higher than the typical cosmic ray-induced FUV flux ∼ 10−7 erg cm−2 s−1, Prasad & Tarafdar 1983, or ∼ 104 photons cm−2 s−1), such as that found for trajectories leading into more distant disc regions. Too strong FUV irradiation (∼ G0) can quench complexity by photodissociation of complex organic ices. At the same time, luke warm temperatures are needed to enable radical mobility on grain surfaces; however, too high temperatures will result in the loss of radicals to the gas phase via thermal desorption. The warmer tem- perature profile of the spread-dominated disk is the main reason for the differences in complex organic abundances between the two disks. Higher temperatures explain also the larger abundance and sole survival of species like ethanol and acetic acid in the spread- dominated disc, which are considered to only form via radical- radical associations (e.g., Walsh et al. 2014a,b). Meanwhile in the infall-dominated disc, the complex organic species which are effi- ciently formed via hydrogenation of solid CO and derivatives are also prevalent. 2008; Shi, Teolis & Baragiola NH3; and noble gases, such as Ar, Kr, Xe (Ayotte et al. 2001; 2003; Bar-Nun, Notesco & Owen Notesco, Bar-Nun & Owen 2007; G´alvez et al. 2009; Fayolle et al. 2011; Yokochi et al. 2012; Bar-Nun & Laufer 2013; Fresneau et al. 2014). It has also been suggested that irradiation of H2O ice may collapse pores in the amorphous structure, thereby further enhancing trapping of volatiles (e.g., Shi, Teolis & Baragiola 2009). Such experimental findings imply that weakly bound volatiles can be preserved in the modelled discs at temperatures that are warmer than their respective thermal desorption regimes. However, this is expected to not exceed the level of a few per cent; thus the qualitative conclusions will remain unaltered. Additional observations of interstellar ices are necessary to verify the layering of ices and to quantify the degrees of mixing and segregation, which could imply a change of the trapping medium from H2O ice to CO2 or CH3OH ices, for example. 3.2.3 Comparison with a three-phase chemical model To further ensure that the production of a CO2 ice-rich zone is not a feature of the two-phase chemical model, additional mod- elling has been carried out with the three-phase chemical model of Furuya et al. (2015). In the three-phase model, the ices are par- titioned between a chemically active surface and an inert bulk at the expense of CPU time. For these comparative model runs, the surface is considered to consist of the uppermost 4 monolayers. The same binding energies and initial prestellar molecular abun- dances (for species common to both networks), as used in the re- sults presented thus far in this work, are assumed. For the sur- face layers, Ediff = 0.3Edes is used, similarly to the assumption made in the two-phase model. Further model details can be found in Furuya et al. (2015). Fig. 5c shows the percentage contribution of volatiles to the total ice content averaged for the midplane beyond 30 AU of the infall-dominated disc at the end of the simulation (2.46 × 105 yr after the precollapse phase), as obtained with the three-phase model of Furuya et al. (2015). This figure is analogous to the right piechart of Fig. 5b. Model results with and without the swapping mecha- nism are shown, which allows the exchange of molecules between the surface layers and the inert bulk ice (according to the formal- ism of Garrod 2013). In the swapping-on model, bulk ice chem- istry (two-body grain-surface reactions and photochemistry) is also included with Ediff = Edes. Thus, the swapping-off model corre- sponds to a fully inert bulk; while the swapping-on setup imposes marginal chemical reactivity in the bulk ice. Upon the inclusion of swapping, weakly bound species such as CH4 and N2 preferen- tially diffuse to the surface layers and are subsequently lost to the gas phase via thermal desorption. When swapping is excluded, the quantities are consistent between the two- and three-phase models for these weakly bound species. The contributions of CH3OH and other minor species are also in agreement. Swapping enhances the supply of CO molecules available for reactions with OH in the surface, thus generating more CO2 over- all via the reaction with OH, as discussed earlier. This demonstrates that the same scheme that was responsible for CO2 ice-dominance in the framework of a two-phase model persists with a three-phase model as well, once the assumption of total bulk inactivity is broken by allowing the bulk-surface exchange and marginal bulk reactiv- ity. This implies that this is not a feature of the two-phase model, but rather a chemical effect consistent with the physical evolution of the system. The stellar FUV irradiation in the envelope during infall towards the disc is the driving force behind this CO2 enhance- Table 4. Average abundances in disc midplanes of complex organic ices relative to methanol ice (n (Xice) /n (CH3OHice)) for three radial ranges at the end of the simulation (2.46 × 105 yr after the precollapse phase) for the spread- and infall-dominated discs. The values are given for the case of the fiducial dynamic simulation and under a static assumption (i.e., constant final disc physical parameters)m. Additionally, in the final line, the average abundance of methanol ice relative to the number of H nuclei is given for the respective radial ranges and models. R (AU) 1 − 10 10 − 30 > 30 disc spread-dom. infall-dom. spread-dom. infall-dom. spread-dom. infall-dom. Species H2CO CH2CO HCOOH HCOOCH3 CH3CHO CH3OCH3 C2H5OH CH3COOH HOCH2CHO n (CH3OHice) /nH dyn. stat. dyn. stat. dyn. stat. dyn. stat. dyn. stat. dyn. stat. 4.7 (−7) 1.7 (−6) 1.6 (−7) 2.7 (−4) 1.7 (−9) 1.4 (−6) 3.7 (−1) 9.8 (−1) 3.2 (−4) 3.2 (−7) 1.1 (−6) 3.9 (−7) 1.6 (−5) 2.7 (−9) 1.4 (−2) 1.8 (−2) 3.3 (−1) 1.0 (−3) 3.4 (−4) 9.8 (−5) 1.7 (−5) 2.4 (−2) 1.7 (−8) 1.6 (−3) 2.7 (−1) 4.4 (−1) 1.5 (−1) 4.6 (−6) 2.1 (−5) 2.2 (−7) 3.7 (−5) 1.6 (−7) 1.0 (−2) 1.8 (−2) 3.5 (−1) 1.1 (−2) 1.4 (−2) 5.7 (−3) 1.1 (−3) 3.9 (−2) 1.6 (−4) 1.1 (−3) 2.0 (−1) 2.1 (−1) 1.0 ( 0) 1.7 (−3) 1.4 (−2) 3.7 (−5) 2.5 (−4) 1.5 (−6) 2.6 (−2) 3.1 (−2) 2.0 (−1) 2.3 (−1) 7.1 (−5) 4.6 (−3) 2.6 (−3) 5.1 (−3) 1.1 (−3) 6.8 (−3) 3.8 (−2) 1.7 (−2) 2.3 (−1) 3.3 (−1) 3.9 (−2) 3.0 (−3) 1.9 (−2) 1.9 (−1) 2.9 (−2) 2.2 (−2) 2.6 (−2) 8.6 (−1) 1.4 (−2) 2.9 (−2) 2.0 (−4) 1.2 (−2) 1.3 (−2) 1.5 (−2) 1.7 (−2) 5.7 (−3) 4.2 (−1) 1.5 (−2) 7.7 (−2) 4.8 (−3) 8.3 (−4) 1.6 (−3) 7.1 (−3) 2.2 (−2) 2.4 (−3) 4.6 (−1) 2.4 (−1) 1.3 (−2) 4.1 (−1) 2.4 (−2) 7.7 (−3) 1.5 (−2) 2.6 (−2) 4.4 (−1) 2.1 (−1) 1.1 (−3) 2.0 (−2) 7.0 (−4) 3.0 (−3) 8.7 (−4) 1.4 (−2) 1.3 (−2) 2.1 (−3) 5.8 (−2) 2.8 (−6) 1.9 (−6) 5.1 (−6) 2.0 (−6) 5.2 (−6) 2.5 (−6) 8.5 (−6) 3.0 (−6) 5.4 (−6) 5.8 (−6) 1.7 (−6) 3.0 (−5) m The notation a (b) stands for a × 10b. 1 0 M a r i a N . D r o z d o v s k a y a e t a l . For both discs an inner reservoir rich in gas-phase complex or- ganics is recovered in these models. This zone appears to be the in- ner few AU of the disc midplanes, where the temperatures are high enough to thermally desorb all tightly bound species; and shielding by dust is sufficiently strong to preserve these molecules from the protostar's photodissociating FUV photons. The abundances in Table 4 show that complex organic ices are plentiful in protoplanetary disks, especially in the outer regions. The current model is, however, tailored to the most favourable con- ditions for complex organic molecule production. This is achieved by assuming Ediff = 0.3Edes (Hasegawa, Herbst & Leung 1992), which allows efficient radical-radical chemistry on grain surfaces. A two-phase model, additionally, assumes the same composition for the two chemically active monolayers as for the entire mantle, thus increasing the availability of species for radical-radical chem- istry without accounting for, as an example, lower radical mobility deeper in the mantle. Hence, the presented abundances may be seen as upper limits on the quantities of complex organic ices in Class 0/I protoplanetary disk midplanes. These aspects are explored in a detailed parameter study in Drozdovskaya et al. (2015). On the other hand, the abundances of complex organic molecules may be enhanced by additional low (∼ 15 K) temperature formation routes that are seen in recent laboratory studies even in the absence of FUV radiation (e.g., Fedoseev et al. 2015; Chuang et al. 2016). −10−2, 10−2 The predicted ice reservoirs can potentially be traced indi- rectly via cometary studies, which are showing that comets are rich in large complex organic species. For example, the highly complex molecule glycine is detected at a ratio of 0.5 relative to methanol on 67P (Altwegg et al. 2016; Le Roy et al. 2015). This implies that the modelled abundances are not very different from such observational evidence. A more detailed comparison with comets is presented in Section 4.4. Observations of hot inner re- gions (so-called hot cores or corinos) place the gas-phase abun- dances of methyl formate, dimethyl ether, ethanol and glycolalde- hyde at ∼ 1−10−2, 1−10−2, 10−1 −10−3 relative to gaseous methanol, respectively (fig. 8 of Taquet et al. 2015). These values are comparable to those in Table 4 with the exception of gly- colaldehyde, which is most likely caused by the lack of destructive chemical pathways in the network, while already including addi- tional formation pathways via glyoxal (Drozdovskaya et al. 2015). Comparisons of the modelled complex organic ice abundances in disk midplanes with gaseous abundances in hot cores/corinos should, however, be made with caution since it remains unclear what the origin of the hot core emission is. In hot cores, ther- mal desorption leaves desorbed molecules intact, but the observed abundances are potentially modified by gas-phase reactions, since the temperatures and densities are high. Typically, ratios of isomers vary between hot core observations and models, which could in- dicate either missing chemical pathways in the network or the im- portance of gas-phase reactions in hot cores. Meanwhile, observed gases in (outer) disk regions are released via non-thermal desorp- tion processes, which may break molecules apart (Cruz-Diaz et al. 2016; Bertin et al. 2016; Walsh et al. 2016). 3.4 Comparison to planet population synthesis models Marboeuf et al. (2014a) studied the volatiles of planetesimals in a static cooling protoplanetary disc. The cooling occurs, because the disc surface density decreases with time, as a consequence of mate- rial accreting on to the star. As explained in Section 2.1, the models of Marboeuf et al. (2014a) assume a hybrid of the 'hot' and 'cold' start scenarios. The volatiles are assumed to consist of H2O, CO, Cometary ices in forming disc midplanes 11 CO2, CH3OH, CH4, N2, NH3 and H2S and are the sole poten- tial ice constituents (if disc conditions allow the solid phase), be- cause they are observed to be dominant interstellar and cometary volatiles. All other species are ignored, which means that as much as ∼ 10% of the total ice is possibly missed (Figs. 5a and 5b). Taking these assumptions about volatiles, the hybrid start sce- nario, the use of phase diagrams and the differences in disc mod- els into account, there are still some similarities in the findings of Marboeuf et al. (2014a) and the dynamic models presented in this work. Here, the inner discs in the embedded phase are some- what more massive than those of Marboeuf et al. (2014a), where Σ . 500 g cm−2 at R = 5.2 AU, while here Σ ∼ 630 g cm−2 and Σ ∼ 540 g cm−2 for the spread-dominated and infall-dominated discs at R = 5.2 AU, respectively (Σ ∼ 110 g cm−2 and Σ ∼ 300 g cm−2 for the spread-dominated and infall-dominated discs at R = 30 AU; Fig. 3). First, similar trends in snowlines are observed. Most volatiles thermally desorb in the inner few AU (. 10 AU). The most volatile species, CH4, CO and N2, freeze out further out in the midplanes, if at all, such as in the hotter irradiated disc model of Marboeuf et al. (2014a) and the spread- dominated disc of this work. This consequently leads to the agree- ment in water being the dominant ice constituent in planetesimals at ∼ 55 − 75% of the total ice content, with its contribution in- creasing with decreasing radius, as molecules that are more weakly bound thermally desorb in accordance with the midplane temper- ature profile (table 9 and figs. 15, 16 of Marboeuf et al. 2014a). The denominator used to obtain the values given in Marboeuf et al. (2014a) accounts only for 8 molecules, while in this work all avail- able solids are summed to obtain the denominator. Further quanti- tative comparison of the planetesimal ice composition, as obtained via full prestellar inheritance by discs and subsequent disc cooling (Marboeuf et al. 2014a), versus that via full kinetic chemical mod- elling and dynamic collapse (this work) is not possible within the framework of our models. Instead, the assumption of complete prestellar inheritance by discs can be tested by contrasting the assumed initial volatile abundances of Marboeuf et al. (2014a) with the midplane abun- dances obtained in this work. Table 5 shows the modelled midplane ice composition for three sets of initial abundances: the model- consistent fiducial prestellar phase abundances; and the CO-poor (CO:CO2 = 1 : 1) and CO-rich (CO:CO2 = 5 : 1) sets, which are based on observations of prestellar cores and protostellar sources as justified in Marboeuf et al. (2014a). Table 5 contains the results for both the spread- and infall-dominated disc midplanes. All values are number averages and focus on the most ice-rich outer regions of the midplanes beyond 30 AU. By comparing the tabulated initial and final disc values, it can be seen that the largest differences lie in the amount of CO2 produced. In this work, more CO2 is produced via the reaction of OH and CO mentioned earlier, which in turn eats away at the water and CO ice reservoirs. For the infall-dominated case with an excess of CO (CO:CO2 = 5 : 1), this effect is at its most extreme as water ice is almost entirely converted to CO2 ice. Another signficant change is seen for CH3OH ice. If more than the fiducial prestellar quantity is injected into the system initially, as under the CO-poor and CO-rich conditions, then the final disc con- tribution is still . 3 per cent, as with the fiducial input. Methanol ice is chemical processed en route to the disc via photon-induced chemistry and used to build larger, more complex molecules, lead- ing to a disc abundance of a few per cent that is independent of the initial abundance. 12 Maria N. Drozdovskaya et al. Table 5. Icy volatiles in outer (R > 30 AU) disc midplanes for different initial abundances at the end of the simulation (2.46 × 105 yr after the precollapse phase) for the spread- and infall-dominated discs for models including dynamic collapse and those for which static conditions were assumed. Initial assumption Species Initial abundancen n (Xgas) /nH n (Xice) /nH initial n (Xice) /P n (Xice) (%)o spread-dom. dyn. stat. infall-dom. dyn. stat. Prestellar CO:CO2 = 1 : 1 CO:CO2 = 5 : 1 7.7 × 10−8 H2O 1.4 × 10−5 CO 7.9 × 10−8 CO2 2.1 × 10−7 NH3 1.2 × 10−7 CH4 CH3OH 2.0 × 10−10 5.3 × 10−10 H2S 1.3 × 10−5 N2 1.2 × 10−3 H2O 2.4 × 10−4 CO 2.4 × 10−4 CO2 8.3 × 10−5 NH3 7.1 × 10−5 CH4 CH3OH 1.8 × 10−4 2.4 × 10−5 H2S 8.3 × 10−5 N2 8.3 × 10−4 H2O 8.2 × 10−4 CO 1.6 × 10−4 CO2 5.7 × 10−5 NH3 4.9 × 10−5 CH4 CH3OH 1.2 × 10−4 1.6 × 10−5 H2S 5.7 × 10−5 N2 1.9 × 10−4 5.9 × 10−5 1.9 × 10−5 4.8 × 10−6 1.5 × 10−5 3.7 × 10−6 8.1 × 10−9 1.1 × 10−5 - - - - - - - - - - - - - - - - 51 16 5 1 4 1 0q 3 57p 11p 11p 4p 3p 8p 1p 4p 39p 39p 8p 3p 2p 6p 1p 3p 69 0q 14 4 0q 2 0q 0q 63 0q 26 3 0q 3 0q 0q 62 0q 27 3 0q 3 1 0q 57 0q 27 3 0q 2 0q 0q - - - - - - - - - - - - - - - - 35 0q 46 1 4 1 0q 3 27 0q 48 0q 4 0q 0q 5 1 16 73 0q 0q 0q 0q 3 51 1 21 4 3 10 0q 2 - - - - - - - - - - - - - - - - n Under the precollapse phase scenario, initial abundances are split over the gas and solid phases. Under the two sets of initial conditions adopted from Marboeuf et al. (2014a), the abundances are set for the gas-phase only in line with the 'hot' disc assumption. o Number averages for R > 30 AU are given, as that region is the richest in ices. Note that the radial range considered for the infall-dominated disc is larger since the average is computed from R = 30 AU and out to the outer radius, which is much larger than for the spread-dominated disc. p If all the gases deposit as ices. q A few per cent can be trapped in H2O ice (Section 3.2.2). 3.4.1 Dependence on initial conditions Table 5 also gives insight into how the modelled disc abundances depend on the initial chemical conditions used. If all the gaseous volatiles in Marboeuf et al. (2014a) are assumed to deposit as ices, then, overall, more volatiles are frozen out than when a model- consistent prestellar phase is calculated. For the CO-poor abun- dances, the initial volatile quantities are increased by factors of 6.3, 4.3, 13, 17, 4.7, 49, 3.0 × 103, 7.5 for H2O, CO, CO2, NH3, CH4, CH3OH, H2S and N2, respectively, relative to the fiducial presteller ice abundances. The largest increases, thus, are for CH3OH and H2S. The CO-rich abundances are extreme with CO ice being avail- able at the level of H2O ice. For the spread-dominated disc, it is then seen that if more CO2 is present initially with the CO-poor abundances, then that quantity is preserved and more CO2 is made en route as also seen with the model-consistent prestellar phase abundances. If more CO is pro- vided initially with the CO-rich abundances, then it is efficiently converted to CO2 via the reaction with OH. In both situations, this leads to CO2 occupying a more significant portion of the ice man- tle than under those of the prestellar phase. This also holds for the infall-dominated disc. Since CO2 was already abundant in the outer regions of that disc, under the CO-poor and CO-rich initial condi- tions favoring larger quantities of CO2, the contribution of CO2 to the ice mantle is even more extreme. 3.4.2 Static scenario To further test the importance of dynamic motions in forming pro- tostellar systems, a static scenario has also been computed. Ma- terial is assumed to be at the constant final physical conditions of the midplane for the entire duration of the evolution, i.e., for 2.46 × 105 yr, as is tradionally assumed for Class II disc mod- els. For both discs, the largest impact is seen for complex organics (Table 4 for R > 30 AU). Fewer of the largest complex organic molecules are produced in both discs under a static assumption. By taking dynamic infall into account, protoplanetary disc materials are exposed to elevated temperature and increased FUV fluxes en route to the disc, which facilitates radical production and mobil- ity on the grain surfaces, both of which are key to forming larger species in these models. This is clearly illustrated in Fig. 6, since the abundances of complex organics increase along with the FUV flux increase in the envelope. Under the static assumption for the infall-dominated disc in the 10 − 30 AU range, the most favorable conditions for complex organic formation appear to be shifted from the outer disc to a more inner region. This in turn affects the avail- ability of methanol ice, which is underproduced by a factor of 10 for this zone in comparison to the dynamic model runs (not shown). The static assumption also shifts the balance of H2O and CO2 ices in the outer regions of the two discs (Table 5). For the spread- dominated disc, the static scenario means cooler temperatures for the duration of the simulation, leading to a higher availability of CO on the grains (if only transiently), which leads to more CO2 ice at the expense of H2O ice. For the infall-dominated disc, the static scenario means lower FUV irradiation, leading to less fre- quent H2O ice photodissociation and thus a reduction of OH, on the contrary reducing the amount of CO2 ice contribution in the outer disc. Other volatiles show variations of no more than a few per cent between the static and dynamic cases. 4 DISCUSSION 4.1 Dynamics, chemistry and inheritance In this work, the importance of dynamic collapse and kinetic chem- istry en route to protoplanetary disc midplanes has been assessed by considering two physical models and by running various test cases. The two modelled discs predominantly grow by different mech- anisms, either via viscous spreading or by pure infall, leading to material being transported along different trajectories. The results have shown that there is a delicate balance between H2O and CO2 ices. Under cool (∼ 20 K) temperatures, allowing the presence of CO ice, and weak FUV irradiation, producing the OH radical from the photodissociation of H2O ice, a significant water-carbon diox- ide imbalance can occur leading to large quantities of CO2 ice at the expense of H2O (and CO) ice(s). Such conditions are encountered at large radii (R > 30 AU) for the infall-dominated disc (∼ 46% of the ice mantle is CO2). Under an assumption of a static disc, this also occurs for the outer region of the spread-dominated disc, while the outer region of the infall-dominated disc becomes less imbal- anced (∼ 25% of the ice mantle is CO2). If either the chemistry or the dynamic transport of volatiles was neglected, then these shifts in the ice mantle composition would be missed. This effect has been shown to be insensitive to chemical model parameters such as the diffusion and quantum tunnelling barriers, and binding energies. It also manifests when three-phase models are used. If the FUV field during the physical evolution would be modified, then the qualita- tive results would still hold; and the amount of CO2 ice produced would only change quantitatively. In the case of a weaker FUV field, the amount of CO2 produced would decrease (by at most ∼ 25 and ∼ 13 per cent for the infall- and spread-dominated cases, based on the static case without any FUV exposure in the envelope). In the case of a stronger FUV field, the amount of CO2 is expected to increase, since more FUV photons are available to dissociate H2O ice. The FUV field can also be harder or softer, and photodis- sociation cross-sections are wavelength-dependent, thus potentially shifting regional enhancements. Given the importance of the FUV flux demonstrated here, this should be explored in future work. The trace complex organic species also show a strong depen- dence on dynamics. Their production is a delicate process balanced by dust temperatures that are warm enough to allow radical mobil- Cometary ices in forming disc midplanes 13 ity on grain surfaces, but low enough to make radical thermal des- orption inefficient; and by FUV irradiation that is sufficiently strong to produce radicals in the ice, but sufficiently weak to not efficiently photodissociate the complex organics back into radicals. As a re- sult, dynamic infall leading to elevated temperatures and increased FUV exposure, facilitates complex organic molecule production, which would otherwise be impeded by the midplane physical con- ditions. This also influences the availability of methanol ice, which is a key parent species for many complex organics, making it sen- sitive to dynamic transport and the chemistry en route. The outer zones (R > 30 AU) of both the spread- and infall-dominated discs contain the most complex organic ices (at abundances as high as that of methanol ice, i.e., at a level of up to a few per cent of the total ice mantle), while the inner discs (inner few AU) are rich in gaseous complex organics. These molecules can only be modelled by using full kinetic chemical calculations. Other volatiles are found to be insensitive to the dynamics and chemistry between the cloud and disc phases. The findings made with these physicochemical models are summarized in Fig. 2. The implication is that not all prestellar ices are simply inherited by the midplanes of protoplanetary discs. The balance between dom- inating volatiles may change depending on the route taken by the material to get to the planetesimal- and comet-forming zones and the chemistry that occurs during that time. The variations seen in H2O, CO2 and CH3OH ices are consistently obtained under dif- ferent initial volatile abundances. The complex organic molecule production is facilitated by the favorable conditions in the enve- lope, thus determining the prebiotically-significant composition of planetesimals outside of the protoplanetary disc. 4.2 Comparison to other disc models A large number of models exist for the Class II protoplanetary discs of varying physical and chemical complexity. Therefore, in this sec- tion, the models of young embedded discs will only be compared to the more evolved discs from studies including chemistry of com- plex organic molecules. The series of papers Walsh et al. (2014a,b), Walsh, Nomura & van Dishoeck (2015) have studied a full range of molecules in protoplanetary discs under static conditions, along ac- cretion flows within the disc, and for various central irradiating pro- tostars, using the same prescription for the chemistry as that used in this work. Comparing with the midplane abundances of complex organ- ics for the static Class II disc of Walsh et al. (2014b), it appears that the spead-dominated disc contains less acetaldehyde (∼ 2 orders of magnitude) and more confined to the outer region (R & 30 AU) rather than up to the thermal desorption front; less dimethyl ether (∼ 1 order of magnitude), but more acetic acid (∼ 2 orders of mag- nitude). The infall-dominated disc seems to be richer in complex organic molecules than the static Class II disc (more formic acid, methyl formate and acetic acid, all by ∼ 1 order of magnitude). These chemical variations stem from different physical conditions and time-scales of static Class II discs in comparison to these dy- namically formed embedded Class I discs. Comparing with the z = 0 AU accretion flow through the disc of Walsh et al. (2014a), it seems that the spead-dominated disc abundances lie somewhere in- between the isolated and extremely irradiated disc models (formic acid abundances are closer to the extremely irradiated disc, while methyl formate abundances are closer to the isolated disc, and ac- etaldehyde abundances lie inbetween the two models). The infall- dominated disc abundances are very similar to those in the mid- plane accretion flow of the isolated disc with only variations of 14 Maria N. Drozdovskaya et al. factors of a few. Due to the inclusion of the additional pathway to glycolaldehyde via glyoxal in this work, more glycolaldehyde is made here by ∼ 2 orders of magnitude. Comparing the spread- and infall-dominated discs to the re- sults from turbulent disc models of Furuya & Aikawa (2014), it appears that formaldehyde, methanol and ammonia agree within a factor of a few with the non-turbulent (αz = 0) model. How- ever, there are subtle differences for complex organics such as methyl formate and dimethyl ether. Here, the outer disc in its en- tirety is rich in large complex species, however in the models of Furuya & Aikawa (2014) the distribution of such species is highly localized. The abundances peak and match those obtained in this work only in a narrow radial range. This is most likely because Furuya & Aikawa (2014) assume Ediff = 0.5Edes, which inhibits efficient complex organic molecule formation (see appendix A of Walsh et al. 2014b). These comparisons imply that current Class II disc models may be underestimating the complex organic content, since the initial chemical conditions set by the preceeding embed- ded phase could be much higher than currently assumed. 4.3 Implications for population synthesis models The results of these models on midplane volatiles have been com- pared to the initial volatile composition of planetesimals used in so- phisticated planet population synthesis models of Marboeuf et al. (2014a,b), Thiabaud et al. (2014, 2015). The two sets of models and assumptions are put in perspective in Fig. 2. The main point is that complete inheritance of cloud volatiles by discs is not a valid assumption for species such as H2O, CO2 and CH3OH for all re- gions of the disc, which have been shown to be affected by the dynamics and chemistry during transport. Thus, under current as- sumptions, planet population synthesis models are potentially over- estimating the availability of CH3OH by a factor of ∼ 2 − 4 and could be missing large regional CO2 reservoirs. Complete exclu- sion of complex organic ices means that as much as ∼ 10% of the total ice may be currently omitted. If the results of these models were to be used as the initial volatile abundances for population synthesis models, then the ini- tial C/O and C/N ratios of planetesimals would be altered (the ratios denote the overall volatile carbon, oxygen, and nitrogen budgets). The midplane gaseous and solid C/O and C/N ratios are shown as a function of disc radius for the spread-dominated disc in Fig. 7a and for the infall-dominated disc in Fig. 7b. The first major point is that the two ratios in ices do not match those in gases, i.e., the two phases are decoupled in terms of these tracers. These findings are fairly consistent with those in fig. 1 Oberg, Murray-Clay & Bergin (2011), which were based on of a much simpler adsorption-thermal desorption prescription for a static disc temperature profile, with the exception of the large gas- phase C/O jump seen in the outer regions of the infall-dominated disc. For the spread-dominated disc, the gas-phase C/O ratio ∼ 1. The ratio decreases to ∼ 0.8 only in the inner 10 AU, once CO2 and H2O ices sequentially thermally desorb and increase the oxygen budget in the gas. The solid-phase C/O ratio in this warmer (∼ 40 K) disc is subsolar at a value ∼ 0.3, which reflects the ices being dominated by water and carbon dioxide setting C/O ∼ 1/3. The ratio reaches towards the solar value of 0.5 in the inner few AU once CO2 and H2O are lost as gases. For the infall-dominated disc, the gaseous ratio is comparable to that of the spread-dominated disc for the inner 30 AU (with a value in the ∼ 0.8 −2 range). However, beyond that radius, the ratio increases by an order of magnitude as a result of CO adsorbing and locking away almost all oxygen carriers. The solid-phase ratio nearly matches the solar C/O ratio with values in the 0.4 − 0.45 range along the entire extent of the disc, which stems from the larger CO2 reservoir seen for in this disc, which drives the ratio towards 1/2. The C/N ratio in the spread-dominated disc varies roughly be- tween 0.8 and 1.8 in the gas and between 2.5 and 4.5 in the ice, thus below the solar value of 4.9 in both phases. The biggest variations are primarily driven by the thermal desorption events of the main icy carbon carrier CO2 and the main icy nitrogen carrier NH3 un- der the warm ∼ 40 K conditions of this midplane. Beyond 30 AU, where complex organic ices are plentiful, the fraction of carbon stored in the solid phases increases and so does the C/N ratio in the ice. In the infall-dominated disc, the C/N ratio in the ice is approxi- mately solar for all radii. The gas-phase ratio rapidly decreases be- yond 20 AU, since the majority of both C and N carriers are frozen out at the cold ∼ 20 K conditions of this disc. The solid phase C/N ratio in this disc is in agreement with the observed cometary value of ∼ 7, however it does not explain the ratios on the order of ∼ 50 for the Earth and chondrites in the inner few AU (see fig. 1 of Bergin et al. 2015). In Figs. 7a and 7b, gray lines depict the C/O and C/N ratios as calculated based on the abundances of the main volatiles only, in contrast to the colored lines, which are calculated based on the full set of species in the network and model-consistent prestellar initial abundances. For the gas phase, the differences between the two calculations are minor (except the very inner 1 −2 AU). On the other hand, for the solid phase, variations by a factor of 2 − 3 are seen. These differences reflect the quantities of C, O and N locked away in complex organic ices and demonstrate the importance of the ∼ 10 per cent of the icy mantle labelled as 'other' in the earlier discussions. The implications of the volatile budget derived with these models for the planet populations are the subject of future col- laborative work, which will account for drift beyond the traditional inner 30 AU planet-forming zone. Such work will show if chemi- cal processing during the earlier dynamic phases of protoplanetary disc formation may help explain CO2-rich exoplanet atmospheres, which cannot be replicated with chemical equilibrium models at atmospheric temperatures and pressures (Heng & Lyons 2016), al- though chemical processing within the atmospheres themselves is non-negligible. The differences in the C/O and C/N ratios between the two phases could potentially be used to trace the contribution of ices from planetesimals to planetary atmospheres versus that of ambient disc gas. 4.4 Comparison with comets Comets are thought to spend the dominant portion of their lifetime at large radii, far away from the heat and irradiation from the Sun. As a result, they are considered to be the most pristine probes of the composition of the young protosolar disc. Observed chemical simi- larities between comets and protostellar environments are puzzling (see Mumma & Charnley 2011 and Caselli & Ceccarelli 2012 for reviews). It remains unclear whether the agreements signify comets inheriting materials from earlier phases of star and planet forma- tion, or if it simply is a general coincidence caused by chemistry proceeding along similar pathways under astrophysical conditions. The modelled midplane volatile contributions relative to water ice generally agree with the observed ranges for cometary volatiles within a factor of 2 (comparing the values quoted in Figs. 5a and 5b with those in fig. 4 of Mumma & Charnley 2011). The dominating ice constituents are found to be H2O, CO2, NH3, CH4 and CH3OH Cometary ices in forming disc midplanes 15 10.0 H2O CO2 For R<10AU: 0.83 C/Ogas = 0.46 C/Oice = For R:10-30AU: C/Ogas = C/Oice = 1.1 0.29 For R>30AU: 1.0 C/Ogas = 0.35 C/Oice = O C / 1.0 0.1 10.0 gas ice ice H2O CO2 CO 10.0 For R<10AU: 0.81 C/Ogas = 0.45 C/Oice = For R:10-30AU: C/Ogas = C/Oice = 1.9 0.40 O C / 1.0 gas 0.1 10.0 N C / 1.0 N C / 1.0 gas For R<10AU: 1.8 C/Ngas = 2.6 C/Nice = 0.1 NH3 0 10 For R:10-30AU: C/Ngas = C/Nice = 1.2 3.3 For R>30AU: 0.81 C/Ngas = 4.5 C/Nice = 20 30 40 50 R (AU) For R<10AU: 1.5 C/Ngas = 4.9 C/Nice = 0.1 NH3 0 10 For R:10-30AU: 0.34 C/Ngas = 5.3 C/Nice = gas 20 CH4 R (AU) 30 For R>30AU: 43 C/Ogas = 0.44 C/Oice = ice ice For R>30AU: C/Ngas = C/Nice = 9.7x10-4 5.4 N2 40 50 (a) Spread-dominated disc (b) Infall-dominated disc Figure 7. The midplane C/O and C/N ratios as a function of disc radius (R) in the gaseous (coral colored lines) and solid (torquoise colored lines) phases at the end of the simulation (2.46 × 105 yr after the onset of collapse). The average ratios over three radial ranges (separated by grey dashed lines) are provided; and the locations of some snowlines are marked (the full list is in Table 2). The grey curves with dots and stars are the ratios in the gasesous and solid phases, respectively, as calculated with abundances of simple volatiles only (namely H2O, CO, CO2, CH3OH, CH4, N2 and NH3.) consistently. Only CO is underproduced in our models, but as men- tioned earlier, this could be resolved by trapping in amorphous wa- ter ice (Section 3.2.2). The CO2 ice-rich zone seen in this work can help explain CO2-rich comets (fig. 4 of Mumma & Charnley 2011), and the CO2-rich and H2O-poor jets of comet 103P/Hartley 2 (A'Hearn et al. 2011). Comets are rich in many complex organics; and the recent measurements for 67P/Churyumov-Gerasimenko with the Rosetta mission are revealing even more complexity than suggested by ear- lier observations (Capaccioni et al. 2015; Goesmann et al. 2015; Le Roy et al. 2015). Molecules as large as ethanol and glyco- laldehyde have also been recently detected for comet Lovejoy (Biver et al. 2015), so chemical complexity seems to be charac- teristic to comets. The detected abundances are fractions of a per cent relative to water, however relative to methanol the abun- dances are as high as few per cent. This is consistent with the modelling outcomes of this work, where complex organics are shown to be efficiently formed relative to methanol (Table 4). This may imply that cometary complex organics formed en route in the envelope in the earliest embedded phases of star forma- tion and were incorporated into planetesimals and cometary bodies in disc midplanes. Potentially, there are also contributions stem- ming from the prestellar phase, where complex organics have also been observed (Requena-Torres et al. 2007; Bacmann et al. 2012; Vastel et al. 2014). However, gas-phase processes have been pro- posed as the origin. Meanwhile, recent laboratory experiments indicate additional low-temperature pathways to complex organ- ics (Fedoseev et al. 2015; Chuang et al. 2016). More detailed and quantitative comparisons between modelled midplane and ob- served cometary volatiles are the subject of future publications. In particular, the authors will aim to relate data on 67P obtained with the ROSINA instrument, which are free from temporal peculiarities (such as those discussed in Qi et al. 2015) that could potential affect remote sensing observations, with the results of these models. 5 CONCLUSIONS This paper builds on previous work by Visser et al. (2009); Visser, Doty & van Dishoeck (2011) and Drozdovskaya et al. (2014) and is focused on the midplane composition of protoplane- tary discs in the embedded phase. State-of-the-art physicochemical models are employed to simulate the formation of discs from the initial prestellar phase for 2.46 × 105 yr. Two discs are studied that vary in their respective dominant disc-growth mechanism, either viscous spreading or pure infall. Subsequently, the path of material towards the midplanes of these discs differs, predominantly in-out or simply inwards, respectively. These routes in turn set the temperatures and UV fluxes during the transport of parcels from cloud to disc. More than 100 trajectories into each disc are computed to sample the disc midplane, and the icy content in the framework of such a dynamical model is analysed thereafter. The main conclusions are as follows: • The typical main ice constituents in the midplanes are ∼ 70 − 80% H2O, ∼ 10 − 20% CO2, 1 − 10% NH3 and 1 − 4% CH3OH; however, CO2 may dominate (∼ 40%) in the outer disc when grown via pure infall. • Trace complex organic ices are most abundant in the outer disc (R & 30 AU). Some may be as plentiful as methanol ice in the midplane, at a level of ∼ 1% of the total ice content, similar to that seen in comet observations. The inner disc is rich in complex organic gases. • The positions of snowlines of volatiles and complex organ- ics in the midplanes of protoplanetary discs, according to relative 16 Maria N. Drozdovskaya et al. volatilities, are retained even when dynamics and chemistry during disc formation are taken into account. • Dynamic infall and the chemistry en route to the midplane may enhance the amount of CO2 and diminish CH3OH ices in com- parison to the prestellar phase. Not all volatiles are simply inherited by the midplane from the cloud. Icy planetesimals and cometary bodies reflect the provenances of the midplane ices. • The elevated temperatures and additional FUV photons in the envelope facilitate the formation of prebiotically-significant molecules, which may consitute as much as ∼ 10% of the icy mantles. Current Class II disc models may be underestimating the complex organic content, since the initial abundances set in the em- bedded phase could be much higher than currently assumed. • The C/O and C/N ratios differ between the gas and solid phases. The two ratios in the ice show little variation beyond the inner 10 AU and both are nearly solar in the case of pure infall. The latest theories and observations are suggesting much earlier planetesimal formation than previously thought. The results pre- sented in this paper for the midplanes of protoplanetary discs around low-mass protostars in the embedded phase may probe the volatile and prebiotically-significant content of the pebbles that go on to feed exoplanet atmospheres and form comets. Future work will focus on deeper understanding of the consequences of these model results on cometary composition and planetary populations. 6 ACKNOWLEDGEMENTS This work is supported by a Huygens fellowship from Lei- den University, by the European Union A-ERC grant 291141 CHEMPLAN, by the Netherlands Research School for Astronomy (NOVA) and by a Royal Netherlands Academy of Arts and Sci- ences (KNAW) professor prize. C.W. acknowledges support from the Netherlands Organisation for Scientific Research (NWO, pro- gram number 639.041.335). U.M. and A.T. acknowledge the Na- tional Centre for Competence in Research PlanetS supported by the Swiss National Science Foundation. REFERENCES Adams F. C., Shu F. H., 1986, ApJ, 308, 836 A'Hearn M. F. et al., 2011, Science, 332, 1396 Akimkin V., Zhukovska S., Wiebe D., Semenov D., Pavlyuchenkov Y., Vasyunin A., Birnstiel T., Henning T., 2013, ApJ, 766, 8 Alibert Y., Carron F., Fortier A., Pfyffer S., Benz W., Mordasini C., Swoboda D., 2013, A&A, 558, A109 Alibert Y., Mordasini C., Benz W., Winisdoerffer C., 2005, A&A, 434, 343 ALMA Partnership et al., 2015, ApJ, 808, L3 Altwegg K. et al., 2016, Science Advances, 2 Ayotte P., Smith R. S., Stevenson K. P., Dohn´alek Z., Kimmel G. A., Kay B. D., 2001, J. Geophys. Res., 106, 33387 Bacmann A., Taquet V., Faure A., Kahane C., Ceccarelli C., 2012, A&A, 541, L12 Bar-Nun A., Laufer D., 2013, Planet. Space Sci., 86, 160 Bar-Nun A., Notesco G., Owen T., 2007, Icarus, 190, 655 Benz W., Ida S., Alibert Y., Lin D., Mordasini C., 2014, in Hen- rik B., Klessen R. S., Dullemond C. P., Henning T. eds, Proto- stars and Planets VI, Planet Population Synthesis. Univ. Arizona Press, Tucson, AZ, pp. 691–713 Bergin E. A., Blake G. A., Ciesla F., Hirschmann M. M., Li J., 2015, Proceedings of the National Academy of Science, 112, 8965 Bertin M. et al., 2016, ApJ, 817, L12 Bertout C., Basri G., Bouvier J., 1988, ApJ, 330, 350 Bisschop S. E., Fraser H. J., Oberg K. I., van Dishoeck E. F., Schlemmer S., 2006, A&A, 449, 1297 Bitsch B., Lambrechts M., Johansen A., 2015, A&A, 582, A112 Biver N. et al., 2015, Science Advances, 1, 1500863 Brinch C., Jørgensen J. K., 2013, A&A, 559, A82 Brown W. A., Bolina A. S., 2007, MNRAS, 374, 1006 Brownlee D. et al., 2006, Science, 314, 1711 Capaccioni F. et al., 2015, Science, 347, aaa0628 Caselli P., Ceccarelli C., 2012, A&A Rev., 20, 56 Chou T.-L., Takakuwa S., Yen H.-W., Ohashi N., Ho P. T. P., 2014, ApJ, 796, 70 Chuang K.-J., Fedoseev G., Ioppolo S., van Dishoeck E. F., Lin- nartz H., 2016, MNRAS, 455, 1702 Ciesla F. J., 2011, ApJ, 740, 9 Cleeves L. I., Adams F. C., Bergin E. A., 2013, ApJ, 772, 5 Cleeves L. I., Bergin E. A., Alexander C. M. O. ., Du F., Graninger D., Oberg K. I., Harries T. J., 2014, Science, 345, 1590 Collings M. P., Anderson M. A., Chen R., Dever J. W., Viti S., Williams D. A., McCoustra M. R. S., 2004, MNRAS, 354, 1133 Collings M. P., Dever J. W., Fraser H. J., McCoustra M. R. S., 2003a, Ap&SS, 285, 633 Collings M. P., Dever J. W., Fraser H. J., McCoustra M. R. S., Williams D. A., 2003b, ApJ, 583, 1058 Collings M. P., Frankland V. L., Lasne J., Marchione D., Rosu- Finsen A., McCoustra M. R. S., 2015, MNRAS, 449, 1826 Cruz-Diaz G. A., Mart´ın-Dom´enech R., Munoz Caro G. M., Chen Y.-J., 2016, preprint (arXiv:1605.01767) D'Antona F., Mazzitelli I., 1994, ApJS, 90, 467 Davis A. M., Alexander C. M. O. ., Ciesla F. J., Gounelle M., Krot A. N., Petaev M. I., Stephan T., 2014, in Henrik B., Klessen R. S., Dullemond C. P., Henning T. eds, Protostars and Planets VI, Samples of the Solar System: Recent Developments. Univ. Arizona Press, Tucson, AZ, pp. 809–831 Dipierro G., Price D., Laibe G., Hirsh K., Cerioli A., Lodato G., 2015, MNRAS, 453, L73 Dong R., Zhu Z., Whitney B., 2015, ApJ, 809, 93 Drozdovskaya M. N., Walsh C., Visser R., Harsono D., van Dishoeck E. F., 2014, MNRAS, 445, 913 Drozdovskaya M. N., Walsh C., Visser R., Harsono D., van Dishoeck E. F., 2015, MNRAS, 451, 3836 Edridge J. L., 2010, PhD thesis, University of London, University College London (United Kingdom) Fayolle E. C., Balfe J., Loomis R., Bergner J., Graninger D., Ra- jappan M., Oberg K. I., 2016, ApJ, 816, L28 Fayolle E. C., Oberg K. I., Cuppen H. M., Visser R., Linnartz H., 2011, A&A, 529, A74 Fedoseev G., Cuppen H. M., Ioppolo S., Lamberts T., Linnartz H., 2015, MNRAS, 448, 1288 Fraser H. J., Collings M. P., McCoustra M. R. S., Williams D. A., 2001, MNRAS, 327, 1165 Fresneau A., Danger G., Rimola A., Duvernay F., Theul´e P., Chi- avassa T., 2015, MNRAS, 451, 1649 Fresneau A., Danger G., Rimola A., Theule P., Duvernay F., Chi- avassa T., 2014, MNRAS, 443, 2991 Furuya K., Aikawa Y., 2014, ApJ, 790, 97 Furuya K., Aikawa Y., Hincelin U., Hassel G. E., Bergin E. A., Vasyunin A. I., Herbst E., 2015, A&A, 584, A124 G´alvez ´O., Mat´e B., Herrero V. J., Escribano R., 2008, Icarus, 197, 599 Garrod R. T., 2013, ApJ, 765, 60 Garrod R. T., Herbst E., 2006, A&A, 457, 927 Garrod R. T., Pauly T., 2011, ApJ, 735, 15 Goesmann F. et al., 2015, Science, 349, 020689 Habing H. J., 1968, Bull. Astr. Inst. Neth., 19, 421 Harsono D., Bruderer S., van Dishoeck E. F., 2015, A&A, 582, A41 Harsono D., Jørgensen J. K., van Dishoeck E. F., Hogerheijde M. R., Bruderer S., Persson M. V., Mottram J. C., 2014, A&A, 562, A77 Harsono D., Visser R., Bruderer S., van Dishoeck E. F., Kristensen L. E., 2013, A&A, 555, A45 Hasegawa T. I., Herbst E., Leung C. M., 1992, ApJS, 82, 167 Heng K., Lyons J. R., 2016, ApJ, 817, 149 Henning T., Semenov D., 2013, Chemical Reviews, 113, 9016 Henning T. et al., 2010, ApJ, 714, 1511 Herbst E., van Dishoeck E. F., 2009, ARA&A, 47, 427 Herrero V. J., G´alvez ´O., Mat´e B., Escribano R., 2010, Physi- cal Chemistry Chemical Physics (Incorporating Faraday Trans- actions), 12, 3164 Hersant F., Wakelam V., Dutrey A., Guilloteau S., Herbst E., 2009, A&A, 493, L49 Inaba S., Ikoma M., 2003, A&A, 410, 711 Johansen A., Blum J., Tanaka H., Ormel C., Bizzarro M., Rickman H., 2014, in Henrik B., Klessen R. S., Dullemond C. P., Henning T. eds, Protostars and Planets VI, The Multifaceted Planetesimal Formation Process. Univ. Arizona Press, Tucson, AZ, pp. 547– 570 Jones A. P., Kohler M., Ysard N., Dartois E., Godard M., Gavilan L., 2016, A&A, 588, A43 Karssemeijer L. J., Cuppen H. M., 2014, A&A, 569, A107 Karssemeijer L. J., Ioppolo S., van Hemert M. C., van der Avoird A., Allodi M. A., Blake G. A., Cuppen H. M., 2014, ApJ, 781, 16 Kofman W. et al., 2015, Science, 349 Le Roy L. et al., 2015, A&A, 583, A1 Lunine J. I., Stevenson D. J., 1985, ApJS, 58, 493 Marboeuf U., Thiabaud A., Alibert Y., Cabral N., Benz W., 2014a, A&A, 570, A36 Marboeuf U., Thiabaud A., Alibert Y., Cabral N., Benz W., 2014b, A&A, 570, A35 McElroy D., Walsh C., Markwick A. J., Cordiner M. A., Smith K., Millar T. J., 2013, A&A, 550, A36 Minissale M., Congiu E., Dulieu F., 2016, A&A, 585, A146 Minissale M., Congiu E., Manic`o G., Pirronello V., Dulieu F., 2013, A&A, 559, A49 Minissale M., Loison J.-C., Baouche S., Chaabouni H., Congiu E., Dulieu F., 2015, A&A, 577, A2 Miotello A., Testi L., Lodato G., Ricci L., Rosotti G., Brooks K., Maury A., Natta A., 2014, A&A, 567, A32 Mousis O., Lunine J. I., Madhusudhan N., Johnson T. V., 2012, ApJ, 751, L7 Mumma M. J., Charnley S. B., 2011, ARA&A, 49, 471 Murillo N. M., Lai S.-P., Bruderer S., Harsono D., van Dishoeck E. F., 2013, A&A, 560, A103 Noble J. A., Dulieu F., Congiu E., Fraser H. J., 2011, ApJ, 735, 121 Notesco G., Bar-Nun A., Owen T., 2003, Icarus, 162, 183 Oberg K. I., Boogert A. C. A., Pontoppidan K. M., van den Broek S., van Dishoeck E. F., Bottinelli S., Blake G. A., Evans, II N. J., Cometary ices in forming disc midplanes 17 2011, ApJ, 740, 109 Oberg K. I., Garrod R. T., van Dishoeck E. F., Linnartz H., 2009, A&A, 504, 891 Oberg K. I., Murray-Clay R., Bergin E. A., 2011, ApJ, 743, L16 Oberg K. I., van Broekhuizen F., Fraser H. J., Bisschop S. E., van Dishoeck E. F., Schlemmer S., 2005, ApJ, 621, L33 Okuzumi S., Momose M., Sirono S.-i., Kobayashi H., Tanaka H., 2016, ApJ, 821, 82 Ormel C. W., Klahr H. H., 2010, A&A, 520, A43 Pagani L., Steinacker J., Bacmann A., Stutz A., Henning T., 2010, Science, 329, 1622 Patzold M. et al., 2016, Nature, 530, 63 Pauly T., Garrod R. T., 2016, ApJ, 817, 146 Pinte C., Dent W. R. F., M´enard F., Hales A., Hill T., Cortes P., de Gregorio-Monsalvo I., 2016, ApJ, 816, 25 Pontoppidan K. M., 2006, A&A, 453, L47 Pontoppidan K. M., Salyk C., Bergin E. A., Brittain S., Marty B., Mousis O., Oberg K. I., 2014, in Henrik B., Klessen R. S., Dulle- mond C. P., Henning T. eds, Protostars and Planets VI, Volatiles in Protoplanetary Disks. Univ. Arizona Press, Tucson, AZ, pp. 363–385 Prasad S. S., Tarafdar S. P., 1983, ApJ, 267, 603 Qi C., Hogerheijde M. R., Jewitt D., Gurwell M. A., Wilner D. J., 2015, ApJ, 799, 110 Requena-Torres M. A., Marcelino N., Jim´enez-Serra I., Mart´ın- Pintado J., Mart´ın S., Mauersberger R., 2007, ApJ, 655, L37 Sakai N. et al., 2014, Nature, 507, 78 Sandford S. A., Allamandola L. J., Tielens A. G. G. M., Valero G. J., 1988, ApJ, 329, 498 Schneider J., Dedieu C., Le Sidaner P., Savalle R., Zolotukhin I., 2011, A&A, 532, A79 Semenov D. et al., 2010, A&A, 522, A42 Semenov D., Wiebe D., 2011, ApJS, 196, 25 Semenov D., Wiebe D., Henning T., 2006, ApJ, 647, L57 Shi J., Teolis B. D., Baragiola R. A., 2009, Phys. Rev. B, 79, 235422 Taquet V., Ceccarelli C., Kahane C., 2012, A&A, 538, A42 Taquet V., L´opez-Sepulcre A., Ceccarelli C., Neri R., Kahane C., Charnley S. B., 2015, ApJ, 804, 81 Testi L. et al., 2014, in Henrik B., Klessen R. S., Dullemond C. P., Henning T. eds, Protostars and Planets VI, Dust Evolution in Protoplanetary Disks. Univ. Arizona Press, Tucson, AZ, pp. 339–361 Thiabaud A., Marboeuf U., Alibert Y., Cabral N., Leya I., Mezger K., 2014, A&A, 562, A27 Thiabaud A., Marboeuf U., Alibert Y., Leya I., Mezger K., 2015, A&A, 574, A138 Thurmer K., Yuan C., Kimmel G. A., Kay B. D., Scott Smith R., 2015, Surface Science, 641, 216 Tobin J. J., Hartmann L., Chiang H.-F., Wilner D. J., Looney L. W., Loinard L., Calvet N., D'Alessio P., 2012, Nature, 492, 83 Tobin J. J., Hartmann L., Chiang H.-F., Wilner D. J., Looney L. W., Loinard L., Calvet N., D'Alessio P., 2013, ApJ, 771, 48 Tobin J. J. et al., 2015, ApJ, 805, 125 Umebayashi T., Nakano T., 1981, PASJ, 33, 617 Vastel C., Ceccarelli C., Lefloch B., Bachiller R., 2014, ApJ, 795, L2 Vasyunin A. I., Herbst E., 2013, ApJ, 762, 86 Vasyunin A. I., Wiebe D. S., Birnstiel T., Zhukovska S., Henning T., Dullemond C. P., 2011, ApJ, 727, 76 Visser R., Doty S. D., van Dishoeck E. F., 2011, A&A, 534, A132 18 Maria N. Drozdovskaya et al. Visser R., Dullemond C. P., 2010, A&A, 519, A28 Visser R., van Dishoeck E. F., Doty S. D., Dullemond C. P., 2009, A&A, 495, 881 Viti S., Collings M. P., Dever J. W., McCoustra M. R. S., Williams D. A., 2004, MNRAS, 354, 1141 Walsh C., Herbst E., Nomura H., Millar T. J., Widicus Weaver S., 2014a, Faraday Discuss., 168, 389 Walsh C. et al., 2016, ApJ, 823, L10 Walsh C., Millar T. J., Nomura H., 2010, ApJ, 722, 1607 Walsh C., Millar T. J., Nomura H., Herbst E., Widicus Weaver S., Aikawa Y., Laas J. C., Vasyunin A. I., 2014b, A&A, 563, A33 Walsh C., Nomura H., Millar T. J., Aikawa Y., 2012, ApJ, 747, 114 Walsh C., Nomura H., van Dishoeck E., 2015, A&A, 582, A88 Willacy K., Langer W., Allen M., Bryden G., 2006, ApJ, 644, 1202 Woitke P. et al., 2016, A&A, 586, A103 Yokochi R., Marboeuf U., Quirico E., Schmitt B., 2012, Icarus, 218, 760 Ysard N., Kohler M., Jones A., Dartois E., Godard M., Gavilan L., 2016, A&A, 588, A44 Zhang K., Blake G. A., Bergin E. A., 2015, ApJ, 806, L7 Zhao B., Caselli P., Li Z.-Y., Krasnopolsky R., Shang H., Naka- mura F., 2016, MNRAS, 460, 2050 APPENDIX A: ICE ABUNDANCES This paper has been typeset from a TEX/ LATEX file prepared by the author. Cometary ices in forming disc midplanes 19 Table A1. Mean, minimum and maximum abundances of ices beyond their respective snowlines in the spread-dominated disc at the end of the simulation (2.46 × 105 yr after the onset of collapse). Species Rsnow (AU) n (Xice) /nH n (Xice) /n (H2Oice) mean min max mean min max H2O CO CO2 NH3 CH4 CH3OH H2S N2 H2CO CH2CO HCOOH HCOOCH3 CH3CHO CH3OCH3 C2H5OH CH3COOH HOCH2CHO 3 > 50 7 3 > 50 4 7 > 50 20 9 4 5 7 7 3 3 3 1.8 (−4) 1.6 (−5) 1.9 (−4) 1.0 ( 0) 1.0 ( 0) 1.0 ( 0) 3.6 (−5) 1.3 (−5) 1.0 (−6) 1.1 (−5) 3.9 (−5) 2.1 (−5) 2.0 (−1) 7.2 (−2) 5.9 (−3) 6.0 (−2) 2.3 (−1) 1.2 (−1) 5.3 (−6) 3.2 (−9) 1.1 (−6) 1.1 (−11) 6.2 (−6) 2.1 (−8) 2.9 (−2) 1.8 (−5) 6.4 (−3) 6.5 (−8) 3.3 (−2) 1.2 (−4) 4.9 (−8) 1.6 (−8) 3.0 (−9) 1.8 (−7) 4.3 (−10) 6.7 (−9) 1.3 (−6) 1.9 (−6) 3.1 (−6) 4.7 (−11) 1.9 (−13) 1.2 (−19) 3.5 (−19) 2.5 (−17) 1.5 (−12) 1.0 (−8) 9.5 (−8) 5.6 (−17) 1.3 (−7) 4.4 (−8) 1.0 (−8) 4.0 (−7) 2.6 (−9) 2.6 (−8) 2.6 (−6) 5.5 (−6) 5.8 (−6) 2.7 (−4) 9.0 (−5) 1.6 (−5) 9.6 (−4) 2.4 (−6) 3.7 (−5) 7.2 (−3) 1.5 (−2) 1.7 (−2) 2.6 (−7) 1.1 (−9) 6.9 (−16) 2.1 (−15) 1.4 (−13) 8.5 (−9) 6.6 (−4) 5.9 (−3) 3.3 (−13) 6.9 (−4) 2.4 (−4) 5.6 (−5) 2.1 (−3) 1.4 (−5) 1.4 (−4) 1.4 (−2) 4.8 (−1) 3.2 (−2) Table A2. Same as Table A2, but for the infall-dominated disc. Species Rsnow (AU) n (Xice) /nH n (Xice) /n (H2Oice) mean min max mean min max H2O CO CO2 NH3 CH4 CH3OH H2S N2 H2CO CH2CO HCOOH HCOOCH3 CH3CHO CH3OCH3 C2H5OH CH3COOH HOCH2CHO 3 21 7 3 23 3 8 36 11 8 3 4 5 6 3 3 3 1.2 (−4) 1.8 (−6) 8.8 (−5) 3.6 (−6) 8.4 (−6) 4.6 (−6) 1.0 (−8) 6.2 (−6) 3.3 (−7) 8.6 (−8) 5.8 (−7) 8.0 (−8) 5.2 (−8) 5.7 (−8) 1.0 (−7) 1.1 (−6) 1.3 (−6) 9.0 (−7) 1.9 (−17) 2.7 (−6) 5.0 (−9) 1.7 (−8) 2.2 (−10) 5.6 (−15) 6.5 (−11) 6.6 (−12) 9.6 (−10) 7.9 (−12) 5.5 (−13) 1.3 (−13) 3.9 (−11) 6.7 (−11) 4.8 (−11) 3.3 (−12) 1.6 (−4) 1.9 (−5) 1.1 (−4) 6.6 (−6) 1.7 (−5) 1.2 (−5) 2.1 (−8) 1.5 (−5) 1.5 (−6) 2.9 (−7) 2.8 (−6) 3.0 (−7) 6.3 (−17) 3.6 (−7) 3.7 (−7) 5.4 (−6) 6.3 (−6) 1.0 ( 0) 1.2 (−2) 9.7 (−1) 2.7 (−2) 1.1 (−1) 3.3 (−2) 9.7 (−5) 7.4 (−2) 4.2 (−3) 5.9 (−4) 5.8 (−3) 6.3 (−4) 3.8 (−4) 4.2 (−4) 7.4 (−4) 1.1 (−2) 1.1 (−2) 1.0 ( 0) 3.0 (−13) 1.7 (−2) 3.4 (−4) 1.7 (−4) 2.4 (−4) 4.7 (−11) 1.0 (−6) 5.6 (−8) 9.2 (−6) 1.5 (−7) 4.7 (−9) 7.8 (−10) 1.0 (−6) 1.7 (−6) 9.9 (−7) 2.8 (−8) 1.0 ( 0) 1.2 (−1) 3.6 ( 0) 4.2 (−2) 4.0 (−1) 7.8 (−2) 3.8 (−4) 3.8 (−1) 2.3 (−2) 1.8 (−3) 2.8 (−2) 2.3 (−3) 4.4 (−3) 3.0 (−3) 2.3 (−3) 6.9 (−2) 2.1 (−1)
1311.6750	2	1311	2013-12-03T13:37:15	Spin-orbit coupling and chaotic rotation for coorbital bodies in quasi-circular orbits	[ "astro-ph.EP" ]	Coorbital bodies are observed around the Sun sharing their orbits with the planets, but also in some pairs of satellites around Saturn. The existence of coorbital planets around other stars has also been proposed. For close-in planets and satellites, the rotation slowly evolves due to dissipative tidal effects until some kind of equilibrium is reached. When the orbits are nearly circular, the rotation period is believed to always end synchronous with the orbital period. Here we demonstrate that for coorbital bodies in quasi-circular orbits, stable non-synchronous rotation is possible for a wide range of mass ratios and body shapes. We show the existence of an entirely new family of spin-orbit resonances at the frequencies $n\pm k\nu/2$, where $n$ is the orbital mean motion, $\nu$ the orbital libration frequency, and $k$ an integer. In addition, when the natural rotational libration frequency due to the axial asymmetry, $\sigma$, has the same magnitude as $\nu$, the rotation becomes chaotic. Saturn coorbital satellites are synchronous since $\nu\ll\sigma$, but coorbital exoplanets may present non-synchronous or chaotic rotation. Our results prove that the spin dynamics of a body cannot be dissociated from its orbital environment. We further anticipate that a similar mechanism may affect the rotation of bodies in any mean-motion resonance.	astro-ph.EP	astro-ph	Draft version March 25, 2021 Preprint typeset using LATEX style emulateapj v. 5/2/11 SPIN-ORBIT COUPLING AND CHAOTIC ROTATION FOR COORBITAL BODIES IN QUASI-CIRCULAR ORBITS Alexandre C.M. Correia Departamento de F´ısica, I3N, Universidade de Aveiro, Campus de Santiago, 3810-193 Aveiro, Portugal; and Astronomie et Syst`emes Dynamiques, IMCCE-CNRS UMR8028, 77 Av. Denfert-Rochereau, 75014 Paris, France Astronomie et Syst`emes Dynamiques, IMCCE-CNRS UMR8028, 77 Av. Denfert-Rochereau, 75014 Paris, France Philippe Robutel (Dated: 2013 November 22) Draft version March 25, 2021 ABSTRACT Coorbital bodies are observed around the Sun sharing their orbits with the planets, but also in some pairs of satellites around Saturn. The existence of coorbital planets around other stars has also been proposed. For close-in planets and satellites, the rotation slowly evolves due to dissipative tidal effects until some kind of equilibrium is reached. When the orbits are nearly circular, the rotation period is believed to always end synchronous with the orbital period. Here we demonstrate that for coorbital bodies in quasi-circular orbits, stable non-synchronous rotation is possible for a wide range of mass ratios and body shapes. We show the existence of an entirely new family of spin-orbit resonances at the frequencies n ± kν/2, where n is the orbital mean motion, ν the orbital libration frequency, and k an integer. In addition, when the natural rotational libration frequency due to the axial asymmetry, σ, has the same magnitude as ν, the rotation becomes chaotic. Saturn coorbital satellites are synchronous since ν (cid:28) σ, but coorbital exoplanets may present non-synchronous or chaotic rotation. Our results prove that the spin dynamics of a body cannot be dissociated from its orbital environment. We further anticipate that a similar mechanism may affect the rotation of bodies in any mean-motion resonance. Subject headings: celestial mechanics -- planetary systems -- planets and satellites: general 1. INTRODUCTION Coorbital bodies have fascinated astronomers and mathematicians since Lagrange (1772) found an equi- librium configuration where three bodies are located at the vertices of an equilateral triangle. Gascheau (1843) proved that for a circular motion of the three bodies, the Lagrange equilibrium points were stable under spe- cific conditions fulfilled by the three masses. In 1906, the first object of this kind was observed (Wolf 1906), the asteroid Achilles, that shares its orbit with Jupiter around the Sun, leading on average by 60◦. At present, more than 4000 coorbital bodies are known in the So- lar System1, sharing their orbits with the planets. More interestingly, pairs of tidally evolved coorbital satellites were also observed around Saturn in a wide variety of orbital configurations (e.g. Robutel et al. 2012). These objects present very low eccentricities (less than 0.01) and their rotations appear to be synchronous, although there is no confirmation yet (Tiscareno et al. 2009). Tidal dissipation slowly modifies the rotation rate of close-in planets and satellites (e.g MacDonald 1964; Cor- reia 2009). For rigid bodies, when the rotation rate and the mean motion have the same magnitude, the dissipa- tive tidal torque may be counterbalanced by the conser- vative torque due to the axial asymmetry of the inertia ellipsoid. For eccentric orbits, this conservative torque allows for capture of the spin rate in a half-integer com- mensurability with the mean motion, usually called spin- 1 http://www.minorplanetcenter.net/ orbit resonance (Colombo 1965; Goldreich & Peale 1966; Correia & Laskar 2009). In addition, for very eccentric orbits or large axial asymmetries, the rotational libration width of the individual resonances may overlap, and the rotation becomes chaotic (Wisdom et al. 1984; Wisdom 1987). However, for nearly circular orbits, the only pos- sibility for the spin is the synchronous resonance (e.g. Goldreich & Peale 1966; Correia & Laskar 2009). Since tidal dissipation simultaneously damps the eccentricity to zero (e.g. Hut 1980; Correia 2009), all the main satel- lites in the Solar System are observed in quasi-circular orbits and synchronous rotation. Contrarily to the classical two-body problem, where circular orbits are unperturbed, in the case of coor- bital bodies, the orbits often present long-term librations around the Lagrange equilibrium points. As a conse- quence, there is a permanent misalignment of the ro- tating body long inertia axis from the radius vector to the central body (Fig. 1). The resulting torque on the rotating body's figure induces some rotational libration (Tiscareno et al. 2009; Robutel et al. 2011, 2012). The combination of both libration motions (orbital and rota- tional) may give rise to some unexpected behaviors for the rotation rate. In this, paper we investigate all the possibilities for the final rotation of coorbital bodies in quasi-circular orbits. 2. MODEL Let us denote m0 the mass of the central body, m the mass of the rotating body, and mc the mass of the coorbital companion (Fig. 1). We adopt here the the- 2 A.C.M. Correia & P. Robutel Fig. 1. -- Reference angles for the coorbital system. m0 is the mass of the central body, m the mass of the rotating body, and mc the mass of the coorbital companion. θ is the rotation angle of m, r its distance to the central body, and f its true anomaly. ζ is the angle between the directions of mc and m. ory developed by ´Erdi (1977) adapted to the planetary problem (Robutel et al. 2012), and limited to the first order in µ = (m + mc)/(m0 + m + mc). We addition- ally assume quasi-circular orbits (negligible eccentricity) with average radius r0 for both coorbital bodies. The polar coordinates (r, f ) for m centered on m0 are given by (Robutel et al. 2012): r = r0 (cid:19) 1 − 2δ 3n (cid:18) 1 − (2 − 2 cos ζ)−3/2(cid:105) f = δζ + nt + f0 , ζ , ζ = −3µn2(cid:104) (1) (2) (3) sin ζ , where n is the orbital mean motion, δ = mc/(m + mc), and f0 is a constant. The system can be stable as long as µ < 0.03812 (Gascheau 1843; Siegel & Moser 1971). √ Equation governs the relative angular position of the Its solutions, plotted in figure 2, are peri- coorbitals. µn odic with an associated frequency ν of the order of (´Erdi 1977). There are stable equilibria, L4 and L5, for (ζ0, ζ0) = (±π/3, 0), and unstable equilibrium, L3, for (ζ0, ζ0) = (π, 0). The trajectories starting with initial conditions ζ0 = ±π/3, ζ 0 < n 6µ describe tadpole or- bits around L4 or L5, and those starting from ζ0 = ±π/3, √ ζ 0 > n 6µ evolve on horseshoe orbits (Fig. 2). The separatrix between these two types of orbits is given by ζmin ≈ 24◦ (for more details see Robutel et al. 2012). The amplitude of the radial variations is usually very small, so that the orbit remains nearly circular (Eq. 1), but the longitudinal half-maximal libration amplitude, √ α ≡ ζmax − ζmin/2 Fig. 2. -- Longitudinal variations of coorbital bodies (Eq. 14) in the plane (ζ, dζ/dτ ) (top), and orbital libration frequency taken over the dashed line (ν versus ζ with dζ/dτ = 0) (bottom). The black curve is the separatrix between the tadpole and the horse- shoes orbits (ζmin ≈ 24◦). The two tadpole orbits surrounding L4 correspond to α = 10◦ and α = 50◦, while the two horseshoe orbits are associated to α = 160◦ and α = 166◦. angle, θ, is then given by (e.g. Murray & Dermott 1999): (cid:16) r0 (cid:17)3 where and sin 2(γ − δ(ζ − ζ0)) , γ = − σ2 2 γ = θ − nt − f0 − δζ0 , r (cid:114) σ = n 3 B − A C , (4) (5) (6) (8) which is approximately the amplitude rotational librations. frequency for small- For the tidal dissipation we adopt a viscous linear model, whose contribution to the rotation is given by (Mignard 1979; Correia & Laskar 2004): ( γ − δ ζ) , γ = −K (7) (cid:16) r0 (cid:17)6 (cid:18) R r k2 ξQ (cid:19)3(cid:16) m0 (cid:17) r0 m where K = 3n can be very large, depending on the initial conditions (Fig. 2). Let us denote A < B < C the moments of inertia of the rotating body. The equation of motion for the rotation is a dissipation constant, with k2 the second Love num- ber, Q−1 ≡ n∆t the dissipation factor, ∆t the dissipa- tion time lag, R the radius of the rotating body, and ξ its normalized moment of inertia. reference axisL4L5L3 1 1.5 2 2.50π/32π/3π4π/35π/32πν / (μ1/2n) 1 1.5 2 2.50π/32π/3π4π/35π/32πν / (μ1/2n)ζ-4-3-2-1 0 1 2 3 4dζ / d τ0π/32π/3π4π/35π/32π0π/32π/3π4π/35π/32πSpin-orbit coupling for quasi-circular coorbital bodies 3 3. DYNAMICAL ANALYSIS For small-amplitude orbital librations (α (cid:28) 1), a sim- ple linear theory can be used to understand the diversity of rotational behaviors. In this case, the solution of equa- tion (3) is given by (´Erdi 1977) (9) ζ = ζ0 + α sin(νt) , (cid:104) (cid:105) t) − α− sin 2(γ + 27µ/2. At first order in α, expression (4) sin 2γ + α+ sin 2(γ − ν 2 √ where ν ≈ n simplifies as: γ = − σ2 ν t) 2 2 (10) where α± = α(1 ± ν/n)δ. We then have three main islands of rotational libration, γ = θ − n = 0,±ν/2, with α±, respectively. For µ (cid:28) 1 and half-widths σ and σ m (cid:28) mc, we get α± = α. Therefore, together with the classical synchronous equilibrium at θ = n, there exists two additional possibilities for the spin at the super- and sub-synchronous resonances θ = n ± ν/2. √ More generally, since ζ is a periodic function with fre- quency ν, we can write (Robutel et al. 2011): e−i2δζ = ρkei(kνt+φk) , (11) (cid:16) r0 (cid:17)3 r (cid:88) k∈Z where ρk and φk are the amplitude and the phase shift of each harmonic, that depend on α. Thus, expression (4) becomes: ρk sin(2γ + kνt + φk) , (12) (cid:88) k∈Z γ = − σ2 2 where the main resonances can be found for θ = n±kν/2, k being an integer. The general problem for the spin-orbit evolution of quasi-circular coorbital bodies is reduced to the analy- sis of the two frequencies, ν and σ, and the amplitude α. µnt, we can However, by rescaling the time using τ = rewrite equations (1), (3) and (4) as √ µ 2δ 3 r = r0 (cid:18) (cid:19) 1 − √ 1 − (2 − 2 cos ζ)−3/2(cid:105) (cid:104) dτ 2 = −3 (cid:16) r0 (cid:17)3 dτ 2 = − β2 dζ dτ d2ζ d2γ 2 (cid:39) r0 , r sin 2(γ − δ(ζ − ζ0)) , √ β ≡ σ/( µn) ∼ σ/ν . sin ζ , (14) (13) (15) with We see that the orbital motion is almost independent of µ, since µ < 0.038, and that the rotational motion only depends on β. The global dynamics of the spin is then approximately controlled by only two parameters: α and β. We can perform a stability analysis of γ in the plane (α, β) to quickly identify the rotational regime for any system of coorbital bodies that is near the synchronous equilibrium. isolated, the half-width of the syn- chronous resonant island in the direction of dγ/dτ is If , Fig. 3. -- Stability analysis of the rotation rate close to the syn- chronization for (α, log10 β) ∈ [0◦ : 70◦] × [−1.5 : 2] (Tadpole orbits). The color index indicates the proportion of chaotic or- bits inside the studied domain: from dark blue for fully regular to red for entirely chaotic. Above the chaotic region (log10 β > 0) only synchronous rotation is possible, while below it (log10 β < 0) several spin-orbit resonances are possible. In this plot we fixed µ (cid:39) mc/m0 = 10−3 and m/mc = 10−3, but these values do not significantly affect the results as long as m/mc < 0.1 (Fig. 6). The black circles correspond to the locations of the six examples presented in Figures 4 and 5. equal to β. Thus, for a given (α, β) we select 400 equi- spaced values of dγ/dτ in the interval [−2β : 2β] and fix the initial value of γ at the synchronization libration center. The corresponding solutions are integrated using the equations (13)-(15) and their dynamical nature (sta- ble/unstable) is deduced from frequency analysis (Laskar 1990, 1993), which gives the fraction of chaotic trajecto- ries. In Figure 3 we show the results for tadpole orbits with µ (cid:39) mc/m0 = 10−3 and m/mc = 10−3 (for instance, an Earth-like planet around a Sun-like star with a Jupiter- like coorbital). The color index indicates the proportion of chaotic orbits inside the studied domain: from dark blue for fully regular to red for entirely chaotic. Depend- ing on the α and β values, the rotation can present a wide variety of behaviors, ranging from non-synchronous equilibria to chaotic motion. A common way of visual- izing and understanding the different regimes is to use Poincar´e surface sections (e.g. Wisdom 1987; Morbidelli 2002). Therefore, we selected a few representative pairs (α, β) and plotted the corresponding diagrams in Fig- ure 4. For α = 0, the coorbital is at equilibrium at a La- grangian point. In this particular case, the orbital mo- -1.5-1-0.5 0 0.5 1 1.5 2 10 20 30 40 50 60 70-1.5-1-0.5 0 0.5 1 1.5 2 10 20 30 40 50 60 70-1.5-1-0.5 0 0.5 1 1.5 2 10 20 30 40 50 60 70-1.5-1-0.5 0 0.5 1 1.5 2 10 20 30 40 50 60 70 0 0.2 0.4 0.6 0.8 1blogβ10α (º)aecfd4 A.C.M. Correia & P. Robutel Fig. 4. -- Poincar´e surface sections in the plane (γ, θ/n) deduced from the time-2π/ν map of the flow (Eq. 4). The angle γ is reduced modulo π. In the top graphs we fix β = 1 and increase the amplitude α: (a) α = 0◦. In this case, the dynamics is the same as a simple pendulum for all β; (b) α = 10◦. Although the super and sub-synchronous islands are small (Eq. 10), the partial overlap of the resonances already generate a significant chaotic region; (c) α = 50◦. For large amplitudes, the islands' overlap give rise to a huge unstable region surrounding the synchronization. In the bottom graphs we fix α = 50◦ and vary β: (d) β = 101.3. The three islands merged generating a chaotic layer in the neighborhood of the synchronization separatrix, as for the modulated pendulum; (e) β = 10−0.4. The three main resonant islands are isolated, narrow chaotic regions are present along the associated separatrices; (f) β = 100.4. Deep inside the chaotic zone, in this case no stable motion is possible. tion is the same as on an unperturbed circular orbit, so the only possibility for the spin is the synchronous rotation (Fig. 4a). For non-zero amplitude orbital li- brations α, the equation for the rotation of the body (Eq. 4) can be decomposed in a series of individual reso- nant terms with frequencies n ± kν/2 (at first order), where k is an integer (Eq. 12). Each term individu- ally behaves like a pendulum, where the rotation can be trapped. The amplitude of each term increases with α. For small-amplitude orbital librations (α (cid:28) 1) only the first three terms k = 0,±1 are important (Eq. 10). As α increases, additional spin-orbit equilibria appear and extended chaotic regions are possible for the spin. For β (cid:28) 1, the resonant islands are well separated apart (Fig. 4e). Thus, the rotation can be captured in individual spin-orbit resonances and stay there. In the linear approximation (Eq. 10), for rotation rates de- creasing from higher values, the super-synchronous reso- nance θ = n + ν/2 is the most likely possibility (Fig. 5e). Synchronous rotation is also possible, if the rotation es- capes capture in the previous resonance. For rotation rates increasing from lower values, the sub-synchronous resonance θ = n − ν/2 is encountered first. For large- amplitude orbital librations, capture in stable higher or- der spin-orbit resonances is also possible. When β ∼ 1, some individual resonant islands over- lap. Analyzed separately, rotational libration could be expected in neighboring resonances, but such behavior is not possible and the result is chaotic rotation (Chirikov 1979; Morbidelli 2002) (Fig. 4f). This means that the rotation exhibits random variations in short periods of time. For moderate chaos, the chaotic region may pro- vide a path into the sub-synchronous resonance (Fig. 5b), or to initially escaped higher-order resonances (Fig. 5c). In the linear approximation (Eq. 10), the chaos is con- √ fined within the super- and sub-synchronous resonances ( θ− n (cid:46) σ(1 + 2 α)) (Fig. 4b), but for larger orbital li- bration amplitudes, the chaotic regions can be extended (Fig. 4c and Fig. 6). For β (cid:29) 1, the resonances come closer, and all individ- ual rotational libration widths embrace the synchronous equilibrium. At this stage, chaotic rotation can still be observed around the separatrix of the synchronous resonance, but the motion is regular nearer the center (Fig. 4d). Therefore, when the rotation of a body is decreasing by tidal effect, the rotation will be shortly chaotic during the separatrix transition, but it is ulti- mately stabilized in the synchronous resonance (Fig. 5d), as if the body evolved in a unperturbed circular orbit (Fig. 5a). 4. DISCUSSION In order to test the reliability of the dynamical pic- ture described in the previous section, we have numeri- cal integrated the equations (1)−(4) together with tidal dissipation (Eq. 7) using K = 250 yr−1. Dissipation is set to a high value so that we can speed-up the simula- (a)(f)(e)(d)(c)(b)Spin-orbit coupling for quasi-circular coorbital bodies 5 Fig. 5. -- Examples of the final evolution of the rotation of a coorbital body. We numerical integrate the equations (1−4) with n = 17.78 yr−1 (corresponding to m0 = 1 M(cid:12) and r0 = 0.5 AU), together with tidal dissipation using K = 250 yr−1 (Eq. 7). We show an example for each pair (α, β) taken from Fig. 4. The initial rotation rate is θ/n = 2.5 in (d) and θ/n = 1.6 in all the other plots. The green dotted lines give the position of the super- and sub-synchronous resonances n ± ν/2 (Eq. 10). tions. However, lower K values have no impact in the capture scenario (e.g. Henrard 1982), and the evolution time-scale is roughly proportional to K−1. In Figure 5 we show an example for each pair (α, β) taken from Fig- ure 4. The initial rotation rate is θ/n = 2.5 in Fig. 5d and θ/n = 1.6 in all the other plots. The green dotted lines give the position of the super- and sub-synchronous resonances n ± ν/2 (Eq. 10). It is interesting to observe that the sub-synchronous resonance can be reached af- ter some wandering in the chaotic zone (Fig. 5b). Cap- ture in higher order spin-orbit resonances can also occur (Fig. 5c). Slightly different initial conditions may lead to totally different final equilibrium configurations. Different tidal models can also change the individual capture probabil- ities in resonance (Goldreich & Peale 1966), but not the global picture described in this paper. Instead of using the simplified equations (1)−(3), we also run some simu- lations integrating direct n-body equations for the orbital motion, but no differences were observed. We can also test the robustness of the stability analy- sis diagram shown in Figure 3 for different mass ratios. In Figure 6 we then fix mc/m0 = 10−6 and vary m/mc, instead of using m/mc = mc/m0 = 10−3 (Fig. 3). We confirm that the impact of µ (cid:39) mc/m0 in the global analysis of the spin is imperceptible (Fig. 6a,b). How- ever, for comparable coorbital masses (m/mc > 0.1) the chaotic region shrinks (Fig. 6c), because the resonance width is proportional to δ = mc/(m + mc) (Eq. 10). For identical coorbital masses (m = mc) the orbital libration amplitude is reduced by one-half, so there is less overlap between neighbor spin-orbit resonances. In Figure 6 we simultaneous plot the stability analy- sis for tapole-type orbits (top) and horseshoe-type orbits (bottom). We observe that for horseshoe orbits the same rotational regimes as for tadpole orbits are still present. However, the chaotic regions are more extended, since the orbital libration amplitude is larger α ∈ [156◦ : 166◦]. Therefore, more harmonics to the expansion of the ro- tational torque must be taken into account (Eq. 12), in- creasing the number of relevant spin-orbit resonances and the superposition between them. In the stability analysis diagrams, β (cid:29) 1 corresponds to large rotational libration widths and/or small mass- ratios (Eq. 15). This is exactly the situation for all Sat- urn's coorbital satellites, due to their prominent ellip- soidal figures (σ/n ∼ 1) and tiny mass-ratios (µ < 10−6; Robutel et al. 2012). As a consequence, the only pos- sibility for these satellites is the synchronous resonance. However, for close-in exoplanets one can expect smaller axial asymmetries and larger mass-ratios (Laughlin & Chambers 2002; Beaug´e et al. 2007; Cresswell & Nelson 2009; Giuppone et al. 2010), that is, smaller β values. For Earth-like planets (m/m0 ∼ 10−6), we have σ/n ∼ 10−3 (Yoder 1995) which gives β ∼ 10−1 for a Jupiter-like coorbital (µ ∼ 10−3), and β ∼ 1 for another Earth- like coorbital (µ ∼ 10−6), respectively. Thus, Earth-like planets may present non-synchronous or chaotic rotation (Fig. 5), an important point to take into account in fu- ture orbital evolution studies (e.g. Rodr´ıguez et al. 2013), and habitability studies (e.g. Selsis et al. 2007). The present work should apply more generally to coor- bital bodies in eccentric orbits, for which the classic spin- orbit resonances (Colombo 1965; Goldreich & Peale 1966; Correia & Laskar 2009) will split into several components 0.00.51.01.52.02.5 0 1 2 3 4 50.81.01.21.41.6 0 1 2 3 4 50.81.01.21.41.6 0 1 2 3 4 50.81.01.21.41.6 0.81.01.21.41.6 0.81.01.21.41.6 (a)(f)(e)(d)(c)(b)time (Myr)time (Myr)time (Myr)6 A.C.M. Correia & P. Robutel Fig. 6. -- Same plots as in Figure 1 (m/mc = mc/m0 = 10−3), but for different mass ratios. Top pictures correspond to tadpole-type orbits (α, log10 β) ∈ [0◦ : 70◦]× [−1.5 : 2], while bottom pictures correspond to horseshoe type orbits (α, log10 β) ∈ [156◦ : 166◦]× [−1 : 1.5]. We fix mc/m0 = 10−6 and vary m/mc: (a) m/mc = 10−3 (δ = 0.999); (b) m/mc = 0.1 (δ = 10/11); (c) m/mc = 1 (δ = 1/2). (k1n ± k2ν)/2, where k1 and k2 are integers. It should also apply to other orbital resonant configurations, in particular for low order mean motion resonances. We acknowledge support by PICS05998 France- Portugal program, and Funda¸cao para a Ciencia e a Tec- nologia, Portugal (PEst-C/CTM/LA0025/2011). REFERENCES Beaug´e, C., S´andor, Z., ´Erdi, B., & Suli, ´A. 2007, Astron. Murray, C. D., & Dermott, S. F. 1999, Solar System Dynamics Astrophys. , 463, 359 Chirikov, B. V. 1979, Physics Reports, 52, 263 Colombo, G. 1965, Nature , 208, 575 Correia, A. C. M. 2009, Astrophys. J. , 704, L1 Correia, A. C. M., & Laskar, J. 2004, Nature , 429, 848 Correia, A. C. M., & Laskar, J. 2009, Icarus, 201, 1 Cresswell, P., & Nelson, R. P. 2009, Astron. Astrophys. , 493, 1141 ´Erdi, B. 1977, Celestial Mechanics, 15, 367 Gascheau, G. 1843, C. R. Acad. Sci. Paris, 16, 393 Giuppone, C. A., Beaug´e, C., Michtchenko, T. A., & (Cambridge University Press) Robutel, P., Rambaux, N., & Castillo-Rogez, J. 2011, Icarus, 211, 758 Robutel, P., Rambaux, N., & El Moutamid, M. 2012, Celestial Mechanics and Dynamical Astronomy, 113, 1 Rodr´ıguez, A., Giuppone, C. A., & Michtchenko, T. A. 2013, Celestial Mechanics and Dynamical Astronomy, 117, 59 Selsis, F., Kasting, J., Levrard, B., Paillet, J., Ribas, I., & Delfosse, X. 2007, Astron. Astrophys. , 476, 1373 Siegel, C. L., & Moser, J. 1971, Lectures on celestial mechanics (Berlin: Springer), 13 Ferraz-Mello, S. 2010, Mon. Not. R. Astron. Soc. , 407, 390 Tiscareno, M. S., Thomas, P. C., & Burns, J. A. 2009, Icarus, Goldreich, P., & Peale, S. 1966, Astron. J. , 71, 425 Henrard, J. 1982, Celestial Mechanics, 27, 3 Hut, P. 1980, Astron. Astrophys. , 92, 167 Lagrange, J. J. 1772, OEuvres Compl`etes VI, 272 (Paris: Gauthier-Villars), (1869) Laskar, J. 1990, Icarus, 88, 266 Laskar, J. 1993, Physica D, 67, 257 Laughlin, G., & Chambers, J. E. 2002, Astron. J. , 124, 592 MacDonald, G. J. F. 1964, Revs. Geophys., 2, 467 Mignard, F. 1979, Moon and Planets, 20, 301 Morbidelli, A. 2002, Modern celestial mechanics : aspects of solar system dynamics (London: Taylor & Francis) 204, 254 Wisdom, J. 1987, Astron. J. , 94, 1350 Wisdom, J., Peale, S. J., & Mignard, F. 1984, Icarus, 58, 137 Wolf, M. 1906, Astronomische Nachrichten, 170, 353 Yoder, C. F. 1995, in Global Earth Physics: A Handbook of Physical Constants, ed. T. J. Ahrens (Washington, DC: American Geophysical Union), 1 logβ-1.5-1-0.5 0 0.5 1 1.5 2 10 20 30 40 50 60 70-1.5-1-0.5 0 0.5 1 1.5 2 10 20 30 40 50 60 70-1.5-1-0.5 0 0.5 1 1.5 2 10 20 30 40 50 60 70-1.5-1-0.5 0 0.5 1 1.5 2 10 20 30 40 50 60 7010(a)-1.5-1-0.5 0 0.5 1 1.5 2 10 20 30 40 50 60 70-1.5-1-0.5 0 0.5 1 1.5 2 10 20 30 40 50 60 70(c)(b)-1-0.5 0 0.5 1 1.5 157 158 159 160 161 162 163 164 165 166-1-0.5 0 0.5 1 1.5 157 158 159 160 161 162 163 164 165 166logβ10α-1-0.5 0 0.5 1 1.5 157 158 159 160 161 162 163 164 165 166-1-0.5 0 0.5 1 1.5 157 158 159 160 161 162 163 164 165 166(a)(c)-1-0.5 0 0.5 1 1.5 157 158 159 160 161 162 163 164 165 166-1-0.5 0 0.5 1 1.5 157 158 159 160 161 162 163 164 165 166(b)αα
1801.07320	2	1801	2018-11-09T22:16:27	Discovery of a Transiting Adolescent Sub-Neptune Exoplanet with K2	[ "astro-ph.EP" ]	The role of stellar age in the measured properties and occurrence rates of exoplanets is not well understood. This is in part due to a paucity of known young planets and the uncertainties in age-dating for most exoplanet host stars. Exoplanets with well-constrained ages, particularly those which are young, are useful as benchmarks for studies aiming to constrain the evolutionary timescales relevant for planets. Such timescales may concern orbital migration, gravitational contraction, or atmospheric photo-evaporation, among other mechanisms. Here we report the discovery of an adolescent transiting sub-Neptune from K2 photometry of the low-mass star K2-284. From multiple age indicators we estimate the age of the star to be 120 Myr, with a 68% confidence interval of 100-760 Myr. The size of K2-284 b ($R_P$ = 2.8 $\pm$ 0.1 $R_\oplus$) combined with its youth make it an intriguing case study for photo-evaporation models, which predict enhanced atmospheric mass loss during early evolutionary stages.	astro-ph.EP	astro-ph	Draft version November 13, 2018 Typeset using LATEX preprint2 style in AASTeX61 8 1 0 2 v o N 9 . ] P E h p - o r t s a [ 2 v 0 2 3 7 0 . 1 0 8 1 : v i X r a DISCOVERY OF A TRANSITING ADOLESCENT SUB-NEPTUNE EXOPLANET WITH K2 Trevor J. David,1 Eric E. Mamajek,1, 2 Andrew Vanderburg,3, ∗ Joshua E. Schlieder,4 Makennah Bristow,5 Erik A. Petigura,6, † David R. Ciardi,7 Ian J. M. Crossfield,8 Howard T. Isaacson,9 Ann Marie Cody,10 John R. Stauffer,11 Lynne A. Hillenbrand,6 Allyson Bieryla,12 David W. Latham,12 Benjamin J. Fulton,7 Luisa M. Rebull,13, 11 Chas Beichman,14 Erica J. Gonzales,15, ‡ Lea A. Hirsch,9 Andrew W. Howard,6 Gautam Vasisht,1 and Marie Ygouf16 1Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Drive, Pasadena, CA 91109, USA 2Department of Physics & Astronomy, University of Rochester, Rochester, NY 14627, USA 3Department of Astronomy, The University of Texas at Austin, Austin, TX 78712, USA 4Exoplanets and Stellar Astrophysics Laboratory, Code 667, NASA Goddard Space Flight Center, Greenbelt, MD 20771, USA 5Department of Physics, University of North Carolina at Asheville, Asheville, NC 28804, USA 6Department of Astronomy, California Institute of Technology, Pasadena, CA 91125, USA 7Caltech/IPAC-NASA Exoplanet Science Institute, Pasadena, CA 91125, USA 8Department of Physics, Massachusetts Institute of Technology, Cambridge, MA, USA 9Astronomy Department, University of California, Berkeley, CA 94720, USA 10NASA Ames Research Center, Moffet Field, CA 94035, USA 11Spitzer Science Center (SSC), Infrared Processing and Analysis Center (IPAC), California Institute of Technology, Pasadena, CA 91125, USA 12Harvard -- Smithsonian Center for Astrophysics, Cambridge, MA 02138, USA 13Infrared Science Archive (IRSA), Infrared Processing and Analysis Center (IPAC), California Institute of Technology, Pasadena, CA 91125, USA 14NASA Exoplanet Science Institute, California Institute of Technology, Jet Propulsion Laboratory, Pasadena, CA 91125, USA 15Astronomy and Astrophysics Department, University of California, Santa Cruz, CA, USA 16Infrared Processing and Analysis Center (IPAC), California Institute of Technology, Pasadena, CA 91125, USA (Received January 19, 2018; Revised October 27, 2018; Accepted November 4, 2018) Submitted to AAS Journals ABSTRACT The role of stellar age in the measured properties and occurrence rates of exoplanets is not well un- derstood. This is in part due to a paucity of known young planets and the uncertainties in age-dating for most exoplanet host stars. Exoplanets with well-constrained ages, particularly those which are young, are useful as benchmarks for studies aiming to constrain the evolutionary timescales relevant for planets. Such timescales may concern orbital migration, gravitational contraction, or atmospheric photo-evaporation, among other mechanisms. Here we report the discovery of an adolescent tran- siting sub-Neptune from K2 photometry of the low-mass star K2-284. From multiple age indicators Corresponding author: Trevor J. David [email protected] 2 David et al. we estimate the age of the star to be 120 Myr, with a 68% confidence interval of 100 -- 760 Myr. The size of K2-284 b (RP = 2.8 ± 0.1 R⊕) combined with its youth make it an intriguing case study for photo-evaporation models, which predict enhanced atmospheric mass loss during early evolutionary stages. Keywords: planets and satellites: physical evolution -- planets and satellites: gaseous low-mass -- stars: planetary systems -- Galaxy: open planets -- stars: clusters and associations: individual (Cas-Tau) ∗ NASA Sagan Fellow † NASA Hubble Fellow ‡ NSF Graduate Research Fellow The Adolescent Sub-Neptune K2-284 b 3 1. INTRODUCTION Exoplanet properties are intrinsically linked to the properties of their host stars. The pri- mary parameters governing stellar structure are mass, metallicity, and age. Planet occurrence is known to correlate with stellar mass (Cumming et al. 2008; Howard et al. 2012) and metallicity (Fischer & Valenti 2005). The degree to which planet demographics are time-dependent, how- ever, remains under-explored. This is due to both the scarcity of known young planets as well as the large uncertainties in the ages of typical exoplanet hosts. Compiling a sample of planetary systems with well-constrained ages is a critical step on the path towards statistical comparisons of the frequencies and properties of planets across time. There is a long history of planet searches within clusters and other coeval stellar popu- lations. Early wide-field transit searches for hot Jupiters targeted globular clusters for the large sample sizes afforded by these popula- tions (Gilliland et al. 2000; Weldrake et al. 2005, 2008). These searches resulted in no detec- tions, leading to a claim of lower occurrence rates within older populations. However, Ma- suda & Winn (2017) revisited that claim and concluded the globular cluster null results were consistent with Kepler hot Jupiter statistics af- ter accounting for frequency trends with stellar mass. Within open clusters of intermediate (∼1 -- 7 Gyr) and young ages (<1 Gyr), numerous surveys have searched for planets across a wide range of mass and separation, using the transit, radial velocity (RV) and direct imaging methods (see Bowler 2016, for a review of young exoplan- ets detected through imaging). In the ∼3.5 Gyr- old M67 cluster, there is a claimed excess of hot Jupiters around solar-mass stars, while the rate of giant planets at wider separations seems to be in agreement with field statistics (Bru- calassi et al. 2014, 2016, 2017). At intermediate ages, RV surveys searching for hot Jupiters in the nearby Hyades (∼750 Myr) and Praesepe (∼790 Myr) clusters have resulted in varying degrees of success (Cochran et al. 2002; Quinn et al. 2012, 2014). More recently, RV monitor- ing has revealed a number of hot Jupiters or- biting T Tauri and post-T Tauri stars (Donati et al. 2016; Johns-Krull et al. 2016; Yu et al. 2017). As far as transit searches in clusters go, the majority of prior surveys were sensitive only to hot Jupiters yet still lacked the combination of sensitivity and sample size needed to distinguish differences in planet populations in clusters and the field (see Janes & Kim 2009, for a review of early cluster surveys). A meta-analysis of early transit searches within open clusters showed that the null results from those surveys were consistent with expectations from field statis- tics (van Saders & Gaudi 2011). To date, only a single survey has compared the cluster and field occurrence rates of planets smaller than Nep- tune. That work used Kepler observations of the ∼1 Gyr-old cluster NGC 6811 to find agree- ment between field and cluster rates, from two transiting planets around G-type stars (Meibom et al. 2013). Compared with the Kepler mission, K2 (How- ell et al. 2014) has targeted a much more di- verse set of astrophysical sources, enabling a wide range of Solar System, planetary, stellar, galactic, and extragalactic investigations. Since early 2014, K2 has steadily assembled a legacy archive of precision photometry for more than 300,000 stars, including thousands of members of young clusters and associations. From these data, the first secure transiting planets in young (< 1 Gyr) clusters have been established. For each of the clusters surveyed, the K2 data are unprecedented in precision, cadence, baseline, and number of members surveyed. Recently, Rizzuto et al. (2017) presented a uniform search for transits in the K2 cluster data. Our group is 4 David et al. Figure 1. K2 light curve of K2-284. In the top panel, the stellar variability pattern due to rotational modulation of starspots is apparent, as are the transits of K2-284 b. In the middle panel, the stellar variability has been removed. Missing transits are due to sections of the light curve that were removed in the detrending procedure. In the bottom panel, phase-folded model fits to the transits of K2-284 b. The red curves show 200 randomly selected models from the MCMC chain. 2990300030103020303030403050306030700.981.001.02299030003010302030303040305030603070BJD - 24548330.9981.000Relative brightness21012Time from mid-transit [hours]0.9970.9980.9991.0001.0011.002The Adolescent Sub-Neptune K2-284 b 5 also involved in a parallel effort to measure the completeness of those data, laying the founda- tion for comparative planet occurrence at young ages. A handful of the young transiting planets found with K2 seem anomalously large com- pared to close-in planets around field-age stars of a similar mass, a possible hint for ongoing radius evolution (Mann et al. 2017a). How- ever, most of the cluster planets transit low- mass (mid-K and later type) stars where our knowledge of planet populations is more incom- plete relative to the solar-type (FGK) stars tar- geted by Kepler. Thus, the question which must be answered is whether these planets are large because they are young, or whether we are only finding them because they are easier to detect. An important step in answering this question is to compare the densities between young and old planets, but to date none of the known young exoplanets have both radius and mass measure- ments. Close-in sub-Neptunes with ages (cid:46)100 Myr are particularly interesting, given theoretical predictions that their cores may continue to be cooling (Vazan et al. 2017) and the at- mospheres of such planets should experience enhanced photo-evaporative mass-loss at early times (Owen & Wu 2013; Lopez & Fortney 2013; Chen & Rogers 2016). The bimodal radius distribution of close-in sub-Neptunes has been interpreted as evidence of photo-evaporative sculpting of this planet population (Fulton et al. 2017). Here we report the discovery and charac- terization of a sub-Neptune-sized planet tran- siting a young star (τ = 120+640−20 Myr). The star's kinematics prior to Gaia DR2 were sug- gestive of membership with the poorly-studied Cas-Tau association. However, the Gaia DR2 data weaken the case for membership and a de- tailed study of the existence, membership, and substructure of the association is left to a fu- ture work. Nevertheless, K2-284 b is one of the younger known transiting exoplanets and thus a useful benchmark for studying the evolution of close-in sub-Neptunes. 2. OBSERVATIONS 2.1. K2 Photometry The Kepler space telescope observed EPIC 247267267 (KP =12.811 mag) between UT 2017 March 8 and 2017 May 27 during Campaign 13 of the K2 mission. Due to roll angle vari- ations and non-uniform intra-pixel sensitivity, photometry from the K2 mission contains sys- tematic artifacts, which are often much larger in amplitude than planet transit signals or even the intrinsic stellar variability. We corrected for these systematic effects using the k2sc pack- age (Aigrain et al. 2016), which simultaneously models time- and position-dependent flux vari- ations using Gaussian process regression. From these data we discovered a periodic signal in a systematic search for transiting planets among the K2 C13 targets. We also extracted photom- etry from a small square aperture (Figure 2) and circular apertures of different radii to mit- igate the impact of nearby stars. The transits of K2-284 b are recovered at a consistent depth within apertures between 4(cid:48)(cid:48) and 16(cid:48)(cid:48) in radius. This argues against the transit signal being due to a diluted eclipsing binary at a projected sep- aration larger than 4(cid:48)(cid:48). We also constructed a separate light curve, initially correcting for systematics using the k2sff routine (Vander- burg & Johnson 2014), and then using that pre- liminary correction as a starting point to pro- duce a light curve by performing a simultaneous least-squares minimization (prior to the transit model-fitting stage) to the transits, stellar activ- ity, and systematics after removing flares (fol- lowing Vanderburg et al. 2016). We flattened the light curve by dividing away the best-fit stel- lar variability pattern from the light curve. This 6 David et al. Table 1. Astrometry and Photometry of K2- 284 Parameter Value Source Astrometry α R.A. (hh:mm:ss) 05:16:33.76 EPIC δ Dec. (dd:mm:ss) µα (mas yr−1) µδ (mas yr−1) (mas) Photometry N U V (mag) B (mag) V (mag) G (mag) g(cid:48) (mag) r(cid:48) (mag) i(cid:48) (mag) J (mag) H (mag) Ks (mag) W 1 (mag) W 2 (mag) W 3 (mag) W 4 (mag) 20:15:18.39 EPIC 25.000 ± 0.082 Gaia DR2 -45.938 ± 0.059 Gaia DR2 9.2935 ± 0.0431 Gaia DR2 GALEX DR5 APASS DR9 APASS DR9 21.688 ± 0.364 14.713 ± 0.006 13.322 ± 0.015 12.8598 ± 0.0011 Gaia DR2 14.089 ± 0.034 12.758 ± 0.038 12.230 ± 0.011 10.868 ± 0.024 10.206 ± 0.025 10.058 ± 0.018 9.975 ± 0.023 10.007 ± 0.020 9.902 ± 0.060 AllWISE AllWISE AllWISE 2MASS 2MASS 2MASS APASS DR9 APASS DR9 APASS DR9 >8.961 AllWISE 2.3. Adaptive optics imaging Adaptive optics imaging of K2-284 at Ks fil- ter (λ◦ = 2.159; ∆λ = 0.011 µm) was ac- quired with the ShARCS infrared camera be- hind the ShaneAO adaptive optics system on the Lick 3-m telescope on 31 August 2017 UT. The ShARCS camera has an unvignetted field of view approximately 20(cid:48)(cid:48) and has a pixel scale of 0.033(cid:48)(cid:48) pixel−1. The AO data were obtained in a 9-point dither pattern with dither point sep- arated by 5(cid:48)(cid:48) and a 60 s integration time per frame for a total of 540 s. We used the dithered images to remove sky background and dark cur- rent, and then align, flat-field, and stack the individual images. The resolution of the Lick imaging was 0.25(cid:48)(cid:48) (FWHM) with a detection Figure 2. Pan-STARRS r-band image centered on K2-284 showing the adopted K2 aperture in red and a smaller aperture in orange, from which the transits were also recovered at a consistent depth. We also inspected photometry from 4(cid:48)(cid:48) wide square apertures centered on the neighboring stars to the south and to the east to confirm that neither are eclipsing binaries. light curve proved to be of higher precision and we adopted it for the remaining analysis. From Box-fitting Least Squares periodogram analyses (Kov´acs et al. 2002) of light curves both includ- ing and excluding the transits of K2-284 b we find no evidence for other periodic signals cor- responding to additional transiting planets. 2.2. Literature data To aid our stellar characterization process, we gathered astrometric and photometric data from the literature. These data included a par- allax, proper motions, and broadband photom- etry from Gaia DR2 (Gaia Collaboration et al. 2018a), as well as photometry from the GALEX DR5 (Martin et al. 2005), APASS DR9 (Henden et al. 2016), 2MASS (Cutri et al. 2003), and All- WISE (Cutri & et al. 2013) catalogs. The pho- tometric and astrometric properties of K2-284 are summarized in Table 1. 79°08'45"30"15"00"20°15'45"30"15"00"Right Ascension (J2000)Declination (J2000)The Adolescent Sub-Neptune K2-284 b 7 contrast of 2.8 magnitudes at one FWHM sep- aration from the target. To obtain a higher resolution and deeper im- age, we also observed K2-284 with infrared high-resolution adaptive optics (AO) imaging, both at Keck Observatory and Lick Observa- tory. The Keck Observatory observations were made with the NIRC2 instrument on Keck-II behind the natural guide star AO system. The observations were made on 2017 Oct 31 in the narrow-band Br − γ filter (λo = 2.1686µm, ∆λ = 0.0326µm) in the standard 3-point dither pattern that is used with NIRC2 to avoid the left lower quadrant of the detector which is typ- ically noisier than the other three quadrants. The dither pattern step size was 3(cid:48)(cid:48) and was re- peated three times, with each dither offset from the previous dither by 0.5(cid:48)(cid:48). The observations utilized an integration time of 10 seconds with one coadd per frame for a total of 90 seconds. The camera was in the narrow-angle mode with a full field of view of 10(cid:48)(cid:48) and a pixel scale of approximately 0.1(cid:48)(cid:48) per pixel. The resolution of the Keck imaging was 0.06(cid:48)(cid:48) (FWHM) with a detection contrast of 3.5 magnitudes at one FWHM separation from the target. The sensitivity of the final combined AO im- ages were determined by injecting simulated sources separated from the primary target in integer multiples of the central source FWHM. The brightness of each injected source was scaled until standard aperture photometry de- tected the injected source with 5σ significance. The resulting brightness of the injected sources relative to the primary target set the 5σ con- trast limits (see Figure 3). We find no evidence for nearby stars brighter than ∆Ks ≈ 4 mag outside of 0.5(cid:48)(cid:48), which corresponds to a Kp limit of ≈ 6 mag, using the Kp−Ks empirical relation for dwarf stars (Howell et al. 2012), and is used to set the limits on the dilution of the observed transit (Ciardi et al. 2015) for the false-positive assessment (§ 3.2). Figure 3. Contrast sensitivity and inset image of K2-284 in Ks as observed with the Lick Observa- tory 3m Shane adaptive optics system (above) and in Br-γ from the NIRC2 camera on the Keck-II tele- scope (below). In each case the 5σ contrast limit in ∆-magnitude is plotted against angular separation in arcseconds. 2.4. Keck-I/HIRES High-dispersion spectra of K2-284 were ac- quired on UT 2017 Aug 29 and Nov 8 using the HIRES spectrograph (Vogt et al. 1994) on the Keck-I telescope. The spectra were ob- tained with the C2 decker, providing a spec- tral resolution of R ≈ 50000 in the range of ∼3640 -- 7990A. The achieved SNR was 32/pixel 8 David et al. at the peak of the blaze function near 5500A. The star's radial velocity (RV) was measured from the HIRES spectra using the telluric A and B absorption bands as a wavelength reference (Chubak et al. 2012). These RVs are accurate at the ∼200 m s−1 level, which we adopt as the uncertainty on each telluric RV measurement. From the HIRES spectra we also derived stellar parameters which we adopted for the remain- ing analysis. Our stellar characterization pro- cedures are described in § 3.4 and summarized in Table 4. The RV measurements from HIRES and TRES (described below) are reported in Ta- ble 2. 2.5. TRES Using the Tillinghast Reflector Echelle Spec- trograph (TRES) on the 1.5m telescope at Fred L. Whipple Observatory, we observed K2-284 on UT 2017 Sep 29. The resolution of this spectrum is R ≈ 44000 between 3850 -- 9096 A. From a 2600s integration, the achieved SNR is 18.9 per pixel at 5110 A. We measured spec- troscopic parameters and the absolute RV for K2-284 from the TRES spectrum using the Stel- lar Parameter Classification (SPC) tool (Buch- have et al. 2012, 2014). SPC measures RV from cross-correlating Kurucz (1992) synthetic tem- plate spectra with the target spectrum, allowing for rotational line broadening. We adopt an er- ror of 0.2 km s−1 in the TRES RV, which is mainly due to the uncertainty in transforming the RV onto the IAU absolute velocity scale. The spectroscopic parameters found with SPC are broadly consistent with those found from the HIRES spectrum (see § 3.4). 3. ANALYSIS 3.1. Transit model fitting We used the pytransit package (Parviainen 2015), based on the Mandel & Agol (2002) for- malism, to generate transit models and fit these Table 2. Radial velocities of K2-284 UT Date BJD RV (km s−1) Instrument 2017 Aug 29 2457995.120599 2017 Sep 29 2458025.897972 2017 Nov 08 2458066.060714 16.85 ± 0.20 17.23 ± 0.20 16.80 ± 0.20 HIRES TRES HIRES to the K2 photometry. Parameter uncertainties were estimated through Markov chain Monte Carlo (MCMC) analysis using the emcee pack- age (Foreman-Mackey et al. 2013). The free pa- rameters in the transit fits are the orbital pe- riod (Porb), the time of mid-transit (T0), the fractional stellar radius (R∗/a), the planet-star radius ratio (Rp/R∗), cosine of the inclination (cos i), eccentricity (e) and the longitude of pe- riastron (ω). We first performed a fit assuming a circular orbit, then relaxed this assumption and allowed eccentricity and the longitude of perias- tron to be free parameters. Transit models were numerically integrated to match the Kepler long cadence (1766 s) prior to fitting. For both fits we initialized 50 walkers with 50000 steps each. The autocorrelation length of each free param- eter was estimated every 1000 steps and once the chain length exceeded N times the auto- correlation length for each parameter and the fractional change in the autocorrelation length estimates was less than n% the chain was con- sidered to be converged and the MCMC sampler was halted. In the circular fit we used N = 100 and n = 1%, while for the eccentric fit we used N = 50 and n = 2%. From the final chains, we determined the burn-in as 10 times the maxi- mum autocorrelation length (1390 steps for the circular fit and 101945 steps for the eccentric fit) and discarded these values. The median parameters of the transit fits determined from the truncated MCMC chains and the uncertain- ties, determined from the 16% and 84% quan- tiles, are reported in Table 3. For the eccentric The Adolescent Sub-Neptune K2-284 b 9 fit, we assumed a Gaussian prior on the mean stellar density centered at 3.97 g cm−3 with width 0.47 g cm−3. The mean stellar density prior originates from the stellar mass and ra- dius we ultimately adopt, as described in § 3.4. In both fits we assumed quadratic limb dark- ening parameters with Gaussian priors centered on aLD=0.7129 and bLD=0.0229 with widths of 0.11 and 0.036 respectively. The choice of limb- darkening values was based on our atmospheric parameters and interpolating between the ta- bles of Claret et al. (2012). We found our model fitting, and hence overall conclusions, to be rel- atively insensitive to the precise choice of limb- darkening parameters. From the directly fit- ted parameters in the MCMC analysis, we de- rived the transit duration and mean stellar den- sity using equations (3) and (19) from Seager & Mall´en-Ornelas (2003), respectively. The mean stellar density in the eccentric case was calcu- lated from equation (39) in Kipping (2010). The mean stellar density clearly indicates the planet is orbiting a dwarf star and not a giant, but we can not rule out that the star is at the end of the pre-main-sequence phase of contraction. We note the equation for mean stellar density assumes a circular orbit, but the general con- clusion remains unchanged given the vast dif- ference in stellar densities for dwarfs and giant stars. Transit model fits to the K2 light curve are shown in Figure 1. 3.2. False positive assessment Two nearby stars within 15(cid:48)(cid:48) of K2-284 are contained within our photometric aperture. The Pan-STARRS survey (Chambers et al. 2016; Flewelling et al. 2016) measured these sources, PSO J051634.085+201504.266 and PSO J051634.329+201522.312, to be approx- imately 4.42 mag and 5.52 mag fainter than K2-284 at r band, respectively. From equation (7) of Ciardi et al. (2015) we calculated that the Figure 4. Contrast versus projected angular sepa- ration. The grey shaded regions show the excluded areas of parameter space in which a putative false positive could reside. The black points show nearby sources detected by Pan-STARRS. Note the sec- ondary line search is blind to companions with ve- locity separations <15 km s−1 from the primary. flux dilution from these nearby stars affects the inferred planet radius at a level of ≈1.2%, such that the true planet radius is negligibly larger than quoted. Here we are not concerned with this dilution, but with the possibility that this source or any other background source might be a contaminating eclipsing binary (EB) that is being diluted by K2-284. The transit signa- ture can be recovered with a consistent depth from photometry extracted using a 4(cid:48)(cid:48) radius aperture, though at lower signal-to-noise due to the difficulties of detrending in the face of in- creased aperture losses. This effectively argues against the possibility of the nearby star being a background EB, since its light should not con- taminate the photometry extracted from the smaller aperture. Other nearby stars within 16(cid:48)(cid:48) either reside outside our aperture or are too faint to explain the observed transit depth (Figure 4). In principle, an EB can dim by a maximum of 100% (although such systems are rare). The observed transit depth thus sets a limit on the faintness of a diluted EB of approximately 0246810121416Separation [arcsec]02468Contrast [ mag]10 David et al. Table 3. Results of K2-284 b transit fits Parameter Prior (Fit 1) Value (Fit 1) Prior (Fit 2) Value (Fit 2) Directly sampled parameters U (4.785, 4.805) Orbital period, Porb (days) Time of mid-transit, T0 (BJD-2450000) U (7859.01726, 7859.20906) Radius ratio, RP /R∗ Scaled semi-major axis, a/R∗ Cosine of inclination, cos i U (cos 90◦, cos 50◦) U (-1, 1) U (0, ∞) U (4.785, 4.805) 7859.11316+0.00057 4.79507+0.00012 −0.00012 −0.00058 U (7859.01726, 7859.20906) 0.0418+0.0011 −0.0010 17.03+0.52−0.66 0.0166+0.0083 −0.0098 U (cos 90◦, cos 50◦) U (-1, 1) U (0, ∞) 4.795069+0.000086 −0.000086 7859.11316+0.00043 −0.00042 0.0420+0.0013 −0.0011 16.84+0.62−0.70 0.017+0.011 −0.011 0.078+0.108 −0.055 180.2+126.4 −129.6 0.684+0.094 −0.092 0.034+0.031 −0.023 2.78+0.14−0.12 0.04771+0.00025 −0.00025 42.6+1.8−1.8 653+16−14 0.28+0.19−0.19 89.00+0.65−0.62 2.147+0.050 −0.045 1.950+0.051 −0.053 3.91+1.07−0.88 U (0, 1) U (0, 360) G(0.7129, 0.11) G(0.0229, 0.036) Eccentricity, e Longitude of periastron, ω (degrees) Limb darkening coefficient, aLD Limb darkening coefficient, bLD Derived parameters Planet radius, RP (R⊕)a Semi-major axis, a (AU) Insolation flux, S (S⊕) Equilibrium temperature, Teq (K)b Impact parameter, b Inclination, i (degrees) Total duration, T14 (hours) Full duration, T23 (hours) Mean stellar density, ρ∗ (g cm−3) G(0.7129, 0.11) G(0.0229, 0.036) G(3.97, 0.47) 0.0 (fixed) 0.0 (fixed) 0.697+0.093 −0.092 0.034+0.030 −0.023 2.77+0.12−0.12 0.04771+0.00025 −0.00025 42.6+1.8−1.8 649+15−13 0.28+0.13−0.16 89.05+0.56−0.48 2.152+0.045 −0.043 1.963+0.050 −0.052 4.06+0.39−0.46 U : Uniform distribution (left bound, right bound). G: Gaussian distribution (center, width). (a) The planet radius does not account for dilution from nearby stars within the photometric aperture, and may be negligibly larger by ≈1.2%. (b) The equilibrium temperature is calculated assuming an albedo of 0.3. ∆Kp (cid:46) 6.9 mag. In the simplified case of a tar- get star with constant flux and a contaminating EB contained in the same photometric aperture, the observed depth of a diluted eclipse neglect- ing sky background is δobs = δecl∆F/(1 + ∆F ), where ∆F is the flux ratio between the tar- get and the contaminating EB in the observed bandpass, and δecl is the intrinsic eclipse depth of the EB. In this case, if the nearby star is in fact an EB, only eclipses with depths greater than ≈ 28% depth are capable of producing the observed transit depth. We extracted photome- try from small apertures centered on the neigh- boring stars to the south and to the east, find- ing no evidence for dimmings of a depth greater than the observed transit depth and at the pe- riod of K2-284 b. We have thus ruled out the possibility that either of the neighboring stars are EBs with periods comparable to the period of K2-284 b. Using the TRILEGAL galactic model (Gi- rardi et al. 2005), we simulated a 1-deg2 field in the direction of K2-284. From the simu- lated field, we calculated the expected colors and surface density of background stars bright enough to produce the observed transit depth (i.e. V (cid:46) 20.2 mag). We then scaled the re- sulting surface density by the size of the K2 aperture to estimate the total number of ex- pected contaminants. We found that < 0.4 pu- The Adolescent Sub-Neptune K2-284 b 11 tative contaminants are expected within a 12(cid:48)(cid:48) aperture or < 0.2 within an 8(cid:48)(cid:48) aperture. The number of expected contaminants that would be EBs is approximately two orders of magnitude smaller based on the statistical frequency of EBs in the Kepler field (Kirk et al. 2016). The mean near-IR colors of putative contaminants in the simulated field are (J − H)=0.49 mag and (H − K)=0.08 mag, suggestive of a K-type dwarf or giant. As noted earlier, the mean stel- lar density from the transit fit effectively rules out the possibility of a planet transiting a giant star. We searched for secondary spectral lines in the HIRES spectrum from 2017 Aug 29 using the procedure described in Kolbl et al. (2015). We found no evidence for a nearby star down to 3% the brightness of the primary and within 0.8(cid:48)(cid:48). Notably, this method is blind to compan- ions with velocity separations <15 km s−1. We show the excluded regions of parameter space for hypothetical false positive scenarios in Fig- ure 4. We also quantified the false positive proba- bility (FPP) using the vespa software package (Morton 2015). From the input K2 photom- etry, the star's spectroscopic parameters and photometry, and high resolution imaging con- straints (the ShaneAO K-band contrast curve, in this case), vespa evaluated the relative likeli- hoods of transiting planet scenarios versus vari- ous diluted eclipsing binary scenarios. The soft- ware accounts for binary population statistics and the ambient surface density of stars using the TRILEGAL galactic model. We found an overall false positive probability of 1/153, with the primary contributor to the FPP being an EB at twice the inferred period. In this case, one might expect differences in the depths of "odd" and "even" transits, so long as the hypo- thetical background EB has different primary and secondary eclipse depths. Figure 5. Radial velocities phased to the orbital ephemeris of K2-284 b. We find no evidence for orbital motion and from these measurements place an upper limit to the planet's mass of <3 MJup at 95% confidence. The expected RV curves for planet masses corresponding to Neptune, Jupiter, and three times the mass of Jupiter are shown by the colored curves. As with any transiting planet candidate lack- ing a mass measurement, it is difficult to rule out all hypothetical false positive scenarios. Nevertheless, from the qualitative arguments presented above and the quantitative vespa light curve analysis, we conclude that a tran- siting planet around K2-284 is the most secure interpretation for the K2 signal. 3.3. Upper limit to the planet mass From three RV measurements we find no evidence for orbital motion corresponding to Doppler semi-amplitudes greater than ∼200 m s−1 at the period of the planet (Figure 5). All three measurements are also consistent with being equal at the ≈1σ level. From these three measurements we performed a one parameter MCMC fit to determine an upper limit to the Doppler semi-amplitude and thus the planet's mass. We performed these fits using the radvel package (Fulton et al. 2018)1, fixing the planet's ephemeris to that determined from the transit 1 https://github.com/ California-Planet-Search/radvel --/20+/2+Phase16.616.817.017.217.4Radial velocity [km s1]3 MJup1 MJup1 MNepHIRESTRES12 David et al. fits and assuming a circular orbit. We did not allow for RV jitter nor did we allow for any sys- tematic offset between the HIRES and TRES RVs, as no such offset should exist. We fixed the systemic velocity to the value reported in Table 4. From this fit we determined an upper limit to the mass of K2-284 b of <3 MJup at 95% confidence, which rules out the possibil- ity that a stellar or brown dwarf companion is responsible for the transits. 3.4. Stellar characterization Below we discuss the various procedures used to characterize the host star. Unless other- wise noted, our quoted uncertainties in the non- spectroscopic parameters were derived through Monte Carlo simulations assuming normally distributed errors in the input parameters. Our spectroscopic analysis points to a dwarf-like gravity suggesting that the star is on or very nearly on the ZAMS. The theoretical pre-main- sequence lifetime of a 0.65 M(cid:12) star (correspond- ing to our adopted mass) is ∼110 Myr (see Fig- ure 6). If K2-284 is in fact at the very end of its pre-main-sequence contraction the true radius would still be encompassed by our radius un- certainties. Thus, our stellar characterization procedures are valid in employing spectral tem- plates of field-aged stars as well as empirical relations based on field star properties. The stellar parameters resulting from our character- ization are reported in Table 4. Spectroscopic characterization. From the HIRES spectrum, we determined the stellar Teff (4108 ± 70 K), radius (0.64 ± 0.10 R(cid:12)), and [Fe/H] (-0.06 ± 0.09 dex) using the SpecMatch- Emp pipeline (Yee et al. 2017). SpecMatch- Emp uses a library of HIRES spectra for bench- mark stars with securely measured parame- ters (via interferometry, asteroseismology, LTE spectral synthesis, and spectrophotometry) to find the optimal linear combination of these templates that matches a target spectrum. The parameters of the target star are determined via Figure 6. Theoretical predictions from the MIST models (Choi et al. 2016) of the evolution in ra- dius (upper panel) and mean stellar density (lower panel) for low-mass stars. The grey lines and shaded regions show the adopted stellar radius and the mean stellar density measured from the transit fit. interpolation between the parameters for the templates in the optimal linear combination. The spectroscopic temperature and particu- larly the metallicity from the TRES spectrum and the SPC analysis (Teff,SPC = 4288 ± 50 K, SPC = -0.382 ± 0.08 dex) are in tension with the values found from SpecMatch-Emp. We do not have a satisfactory explanation for the metallicity discrepancy, but it may be related to the fact that the SpecMatch-Emp library of empirical template stars in this temperature range do not sample an evenly-spaced range of metallicities. Notably, the effective temperature inferred from the star's photometric colors and empirical relations (Pecaut & Mamajek 2013; Mann et al. 2015) is closer to the value from the SpecMatch-Emp analysis. Spectral type and extinction. The best- matching template star from the SpecMatch- Emp analysis is GJ 3494, which has been as- signed spectral types of M0 and K5 (Skiff 2014). From the spectroscopically determined Teff and the empirical spectral-type-Teff relations pre- sented in Pecaut & Mamajek (2013), hereafter PM13, we find the Teff to be consistent with a 0.60.81.0R [R]0.55 M0.65 M0.75 M050100150200250300Age [Myr]0.02.55.07.5 [g cm3]The Adolescent Sub-Neptune K2-284 b 13 Table 4. Parameters of EPIC 247267267 Parameter Value Source Kinematics and position Barycentric RV (km s−1) U (km s−1) V (km s−1) W (km s−1) Distance (pc) Adopted parameters M∗ (M(cid:12)) R∗ (R(cid:12)) L∗ (L(cid:12)) Teff (K) log g (dex) [Fe/H] (dex) AV (mag) Rotation period (days) v sin i∗ (km s−1) log R(cid:48) S-index HK (dex) Estimated age τisoc,1 (Myr) τisoc,2 (Myr) τgyro,1 (Myr) HK (Myr) τgyro,2 (Myr) τR(cid:48) τNUV (Myr) τ∗ (Myr) Gaia DR2 + RV 16.96 ± 0.19 HIRES, TRES -14.5 ± 0.2 -27.6 ± 0.1 Gaia DR2 + RV -5.56 ± 0.05 Gaia DR2 + RV 107.6 ± 0.5 Gaia DR2 0.63 ± 0.01 0.607 ± 0.022 0.097 ± 0.004 4140 ± 50 4.67 ± 0.01 0.00 ± 0.08 0.27 ± 0.05 8.88 ± 0.40 3.54 ± 0.50 -3.9 ± 0.5 5 ± 1 isoclassify isoclassify isoclassify isoclassify isoclassify isoclassify Teff , B − V , PM13 K2 TRES+SPC HIRES HIRES 113+703−25 133+573−69 124+13−15 262+35−41 139+1353−119 111+160−65 120+640−20 Teff , L∗ Teff , ρ∗ Prot, (B − V )0, B07 Prot, (B − V )0, MH08 log R(cid:48) (N U V − J)0, (J − K)0, F11 HK , MH08 PM13: Pecaut & Mamajek (2013); B12: Boyajian et al. (2012); M15: Mann et al. (2015); B07: Barnes (2007); MH08: Mamajek & Hillenbrand (2008); F11: Findeisen et al. (2011). spectral type of K6.5. Given the stellar effective temperature, we interpolated between the em- pirical Teff-(B − V )0 relation of PM13 to deter- mine an expected intrinsic color of (B − V )0 = 1.305 mag, corresponding to a color excess of E(B − V ) = 0.086 ± 0.016 mag. We then as- sumed the Cardelli et al. (1989) extinction curve to derive AV . We used the (B − V ) color excess above and the extinction coefficients derived by Yuan et al. (2013) for the GALEX and 2MASS passbands to derive near-UV and near-IR col- ors, which we later use to estimate the stellar age, as described in § 3.5. Mass, radius, and luminosity. We derived the luminosity using the spectroscopically de- termined Teff, radius and the Stefan-Boltzmann Law. We derived a separate luminosity estimate from an empirical Teff-luminosity relation based on interferometry of low-mass stars (Boyajian et al. 2012). This second estimate is not en- 14 David et al. tirely independent of the first estimate, since the spectroscopic parameter pipeline is calibrated to the same interferometric standards, among other benchmark stars. We derived a model- dependent mass from a theoretical H-R diagram using the solar-metallicity (Z=0.0152) PAR- SECv1.2S models (Bressan et al. 2012; Chen et al. 2014), our spectroscopically determined Teff, and the Stefan-Boltzmann determined lu- minosity. We also derived a distance-dependent mass from the kinematic distance, the apparent Ks magnitude, and a semi-empirical mass-MKs relation (Mann et al. 2015). Notably, this mass is 2σ lower than the model-dependent mass we adopt. We assume the discrepancy is due to the uncertainty in the distance. If the mass esti- mate from this empirical relation is correct, the mean stellar density from the transit fit would seem to reinforce the notion that the star is still on the pre-main-sequence. However, as a sanity check we compared our stellar parame- ters with those of the nearly equal-mass bench- mark eclipsing binary GU Boo (L´opez-Morales & Ribas 2005), which agree reasonably well with our adopted mass, radius, and temperature. We used the isoclassify2 package (Huber et al. 2017) in Python for our final determina- tion of the stellar mass, radius, and luminos- ity. The package has two operational modes, both of which take input observables (in our case spectroscopic constraints, photometry, and parallax) in order to derive stellar parameters. In the "grid" mode, isoclassify takes the in- put observables and interpolates between the MIST isochrones (Choi et al. 2016; Dotter 2016) to derive posterior probability densities for Teff, log g, [Fe/H], R∗, M∗, ρ∗, L∗, age, distance and AV . In the "direct" mode, the software can take the same input parameters and use bolomet- ric corrections (taken from the MIST models) and extinction maps to determine Teff, R∗, L∗, 2 https://github.com/danxhuber/isoclassify distance, and AV directly from physical rela- tions. We classified K2-284 in both modes using the HIRES spectroscopic Teff and [Fe/H] con- straints, the log g constraint from TRES, the Gaia DR2 parallax, and JHK+gri photome- try. Both modes predicted stellar radii that were consistent within 1σ, and we ultimately adopted the mass, radius, and luminosity from the grid method, though with the more conser- vative radius uncertainties derived from the di- rect method. We also checked that the parame- ters did not change substantially when only in- cluding the JHK photometry or only the K magnitude. Rotation period and projected rota- tional velocity. A period of 8.88 ± 0.40 d, which we attribute to surface rotation of the star, was measured from a Lomb-Scargle pe- riodogram analysis (Lomb 1976; Scargle 1982) of the K2 light curve (Figure 7). The uncer- tainty in the rotation period was estimated from the standard deviation of a Gaussian fit to the oversampled periodogram peak. This uncer- tainty is likely overestimated, but encompasses the more difficult to quantify uncertainty in the rotation period due to e.g. differential rotation. The formal uncertainty, estimated by the peri- odogram peak width divided by the peak signal- to-noise, is 0.0085 d. A second peak in the pe- riodogram at 4.41 ± 0.11 d is a harmonic of the true rotation period. The projected rotational velocity, v∗ sin i∗ = 3.54 ± 0.50 km s−1, was measured from the TRES spectrum by broad- ening synthetic template spectra to match the observations. An independent and consistent v∗ sin i∗ estimate of 3 -- 4 km s−1 was found from the HIRES spectrum and SpecMatch-Emp by broadening empirical template spectra, assum- ing the template stars were not rotating. Using the TRES value and the K2 rotation period we estimated the minimum stellar radius, R∗ sin i∗ = 0.621 ± 0.092 R(cid:12). This value is within the uncertainty of our adopted radius, suggesting The Adolescent Sub-Neptune K2-284 b 15 were compared to the proper motions and radial velocities of members of these groups from the literature. K2-284's proper motion is clearly in- consistent with the nearby 118 Tau group (µα, µδ = 4, -39 mas yr−1). Although K2-284's ra- dial velocity (16.96 ± 0.19 km s−1) is similar to that of the Tau-Aur association (+16 km s−1, Luhman et al. 2009), its proper motion is very different compared to the mean proper motion for the group (µα, µδ = 6, -21 mas yr−1, or any of the subgroups (Luhman et al. 2009). The only group which provides a near match of proper motion and radial velocity is the Cas- Tau association. Prior to the determination of a parallax from Gaia DR2, we used the methodology of Mamajek (2005), the UCAC5 (Zacharias et al. 2017) proper motion for K2- 284 and the "spaghetti" velocity solution from de Zeeuw et al. (1999), to find the bulk of K2-284's proper motion appeared to be mov- ing towards the Cas-Tau convergent point (µυ = 51.3± 1.2 mas yr−1) with negligible perpen- dicular motion (µτ = 4.7± 1.2 mas yr−1). The predicted kinematic distance from this analy- sis was 79± 10 pc (kinematic parallax = 12.7± 1.6 mas), with predicted radial velocity vr = 15.4 km s−1 (compared to our measured value of 16.96± 0.19 km s−1), and predicted pe- culiar motion 1.7± 0.5 km s−1. However, con- tradicting the spaghetti velocity solution, the true distance to the system now provided by Gaia DR2 is d=107.6 ± 0.5 pc. de Zeeuw et al. (1999) estimated the space velocity of Cas-Tau using the spaghetti method with their Hipparcos membership to be U, V, W = -13.24, -19.69, -6.38 km s−1 (U positive to- wards Galactic Center). As a check, and to provide a modern estimate, we cross-referenced de Zeeuw's membership of Cas-Tau mem- bers with the revised Hipparcos catalog (van Leeuwen 2007), Gaia DR1 (preferred, when available), and the radial velocity compilation of Gontcharov (2006). This provided UVW Figure 7. Lomb-Scargle periodogram from K2 photometry of K2-284 (top) and the light curve phased to the rotation period of 8.88 d (bottom). the stellar spin-axis is nearly edge-on. Put an- other way, for our adopted radius, the measured photometric rotation period, and assuming a uniform distribution in cos i∗, the median and 68% confidence interval predicted for v∗ sin i∗ is 3.6 ± 0.6 km s−1, in good agreement with our measurements. Kinematics, Membership & Distance. The EPIC catalog contains a preliminary pho- tometric distance estimate of 84+18−11 pc, assum- ing the star is on the main sequence (Huber et al. 2016). Are there any nearby young stel- lar populations that K2-284 might be a kine- matic member of which could help in constrain- ing its age? K2-284 occupies a busy region of sky with regard to nearby young stellar popu- lations. Within 200 pc and within 30◦ of K2- 284's position are three open clusters (Hyades, 32 Ori & Pleiades), the Tau-Aur association, the Cas-Tau association, and the recently iden- tified 118 Tau group. The Gaia DR2 proper motions for K2-284 are µα, µδ = 25.000, -45.938 (±0.082, ±0.059) mas yr−1. The proper motions 0.111040Period [d]0.00.20.40.6L-S Power1.00.50.00.51.0Phase0.020.010.000.010.02Kp [mag]16 David et al. velocity estimates for 48 candidate Cas-Tau members. These are plotted in Fig. 8 along with the mean velocities for the Cas-Tau group from de Zeeuw et al. (1999) and the α Per- sei cluster, along with the values for K2-284 given the Gaia DR2 kinematics and the RV we determined here. The median U V W for the 48 members is U, V, W = -14.7±0.9, -21.3±0.8, -7.1±0.4 km s−1. Using the probit method, which is resilient to the effects of extreme val- ues, the 1σ scatters reflecting the core of the velocity distributions are estimated as 3.9, 3.7, 2.7 km s−1). Accounting for the mean U V W velocity component uncertainties (2.4, 1.9, 1.4 km s−1), this suggests the intrinsic U, V, W ve- locity dispersions among the de Zeeuw et al. Cas-Tau membership to be approximately 3.0, 3.1, and 2.2 km s−1. This is likely reflecting the adopted 3 km s−1 velocity dispersion used by de Zeeuw et al. in their original kinematic membership selection. Further work is needed to clarify the membership of Cas-Tau with Gaia astrometry, and to search for kinematic and age substructure, however this is beyond the scope of this study. In Appendix A, we discuss the history of the Cas-Tau association, examine the main sequence turnoff for proposed members, and derive a new estimate of the association age. We also used the BANYAN Σ tool (Gagn´e et al. 2018) to estimate the membership proba- bility of K2-284 to various young moving groups and clusters within 150 pc. We note that the proposed Cas-Tau association is not included in BANYAN. We calculated membership probabil- ities both including and excluding the XYZ po- sition of the star. The latter scenario is useful for identifying putative moving group or cluster members that are widely separated on the sky from the core population. The closest kinematic match amongst the young associations included in BANYAN was Tau-Aur, although the most likely hypothesis found in both cases is that K2- 284 is a field star with 99.9% probability. 3.5. Youth indicators Rotation: The photometric rotation period provides evidence of youth, as shown in Fig- ure 9. For a star of its mass or color, K2- 284 has a rotation period consistent with the slowly-rotating sequence of Pleiades members (Rebull et al. 2016), but about 3-4 d shorter than expected for members of Praesepe (Rebull et al. 2017) or the Hyades (Douglas et al. 2016). Given the star's intrinsic (B − V ) color and its rotation period, we calculated the age of the star using the gyrochronology relations of both Barnes (2007), hereafter B07, and Mamajek & Hillenbrand (2008), hereafter MH08. Our gy- rochronology ages take into account the uncer- tainties in the rotation period, (B− V ) color, as well as the published errors on the coefficients in the age-rotation relations (see Table 4). The B07 calibration produces an age that is roughly a factor of two younger than the age predicted from the MH08 relations (τgyro,B07 = 124 Myr, compared to τgyro,MH08 = 262 Myr). For com- pleteness, we also investigated the Angus et al. (2015), hereafter A15, gyrochronology calibra- tion and found that it closely reflects the MH08 predictions in the age and color range of in- terest here. To further investigate the differ- ences and potential systematics in existing gy- rochronology calibrations, we compared the re- lations to the intrinsic (B − V ) colors and rota- tion periods for members of the Pleiades and Praesepe clusters. The Pleiades photometry were gathered from Stauffer & Hartmann (1987) and Kamai et al. (2014), the Praesepe photom- etry from Upgren et al. (1979); Weis (1981); Stauffer (1982) and Mermilliod et al. (1990), and the rotation periods originate from Rebull et al. (2016, 2017). For this exercise we as- sumed E(B − V ) = 0.04 for the Pleiades, and no reddening for Praesepe. Figure 9 shows that the B07 calibration most closely matches The Adolescent Sub-Neptune K2-284 b 17 Figure 8. XYZ positions and UVW space motions for proposed Cas-Tau members (open circles; de Zeeuw et al. 1999), the α Per cluster (filled triangle; based on the Gaia DR2 astrometry and radial velocity from Gaia Collaboration et al. 2018b), and K2-284 (red star). The errorbars reflect 1σ uncertainties for de Zeeuw et al. (1999) Cas-Tau members using Gaia DR2 parallaxes, combined with RVs from DR2 when available or de Bruijne & Eilers (2012) otherwise. In the bottom panels, the filled circle indicates the median velocity of the proposed Cas-Tau members, without outlier rejection. The cluster of points in the lower right of the first two panels is due to the newly identified µ Tau group, which will be the subject of a future work. the Pleiades slowly-rotating sequence at the accepted age of the cluster, while the MH08 and A15 relations overpredict the age of the Pleiades. It is worth noting that all existing gy- rochronology calibrations predict a younger age for Praesepe (∼500-600 Myr), that is more in line with recent color-magnitude diagram esti- mates (Gossage et al. 2018), but in tension with the older estimate of ∼790 Myr from Brandt & Huang (2015). A complete reassessment of gy- rochronology calibrations using the voluminous rotation data now provided by K2 is in order but outside the scope of this paper. We ten- tatively conclude that the younger gyrochronol- ogy age of K2-284 predicted by the B07 relations is likely to be more accurate given the ability of that calibration to reproduce the Pleiades data, but also note that gyrochronology is fundamen- tally a statistical age-dating method, only appli- cable to main-sequence stars, and assumes the star is on the slowly-rotating sequence. In this case, our stellar characterization suggests K2- 284 has indeed arrived on the main sequence and other youth indicators discussed below are consistent with an age similar to that of the Pleiades. Chromospheric activity: K2-284 shows significant emission in the Ca II H&K lines (Fig. 12). The precise H&K values in our spec- tra are ambiguous due to the low SNR of ∼4 per pixel in the H&K orders. Nevertheless, we report our measured log R(cid:48) HK and the S-index 400350300250200150100500X [pc]3002001000100200300400500Y [pc]400350300250200150100500X [pc]15010050050100Z [pc]2001000100200300Y [pc]15010050050100Z [pc]302520151050U [km/s]4035302520151050V [km/s]302520151050U [km/s]14121086420W [km/s]4035302520151050V [km/s]14121086420W [km/s]18 David et al. with large uncertainties in Table 4. Our best estimate of log R(cid:48) HK is just barely outside the high-activity range where the activity-age re- lations of Mamajek & Hillenbrand (2008) were calibrated. Regardless, we estimated an ac- tivity age by modeling log R(cid:48) HK as a normal distribution with the values specified in Table 4 and imposing a cutoff upwards of -4.0. From this analysis we estimated the activity age to be <435 Myr at 68% confidence. Near-UV emission: While there is no X- ray detection of K2-284, the star was detected at near-UV wavelengths with GALEX. Young, low-mass stars have been shown to exhibit sig- nificant emission above photospheric levels in the near-UV (NUV, 1750 -- 2750 A) and far-UV (FUV, 1350 -- 1750 A) GALEX passbands (Find- eisen & Hillenbrand 2010; Shkolnik et al. 2011; Rodriguez et al. 2011, 2013; Kraus et al. 2014). Specifically, Shkolnik et al. (2011) found that young (<300 Myr) late-K and M-dwarfs gener- ally show fractional flux densities of FNUV/FJ > 10−4 while older stars tend to fall below this threshold. K2-284 has a fractional flux density of FNUV/FJ = 1.1 × 10−4. Using near-UV pho- tometry from GALEX and near-IR photome- try from 2MASS, we estimated the stellar age based on the (NUV-J) and (J − K) colors and the empirical relations presented in Findeisen et al. (2011). Empirical isochrones from that work are shown in Figure 10, along with com- parisons to other known young stellar popula- tions. While there is a large amount of scatter in this color-color diagram, particularly for later- type stars, there is a clear qualitative trend of declining NUV flux for older stars. Proposed Upper Sco and Sco-Cen members were selected from the Young Stellar Object Corral (YSOC), Tuc-Hor members from Kraus et al. (2014), and Hyades members from Perryman et al. (1998). The photometry were dereddened using the ex- tinction coefficients of Yuan et al. (2013) and Figure 9. Top: Gyrochrones in the period ver- sus (B − V ) plane. The solid, dashed, and dot- ted lines show gyrochrones predicted from Barnes (2007), Mamajek & Hillenbrand (2008), and An- gus et al. (2015), respectively. Bottom: Rotation periods versus (V − Ks) color for Praesepe (red) and Pleiades (orange) members. In both figures the cluster rotation periods are taken from Rebull et al. (2016, 2017) and the white star indicates K2-284. assuming A(V ) = 0.7 mag for Upper Sco and A(V ) = 0.16 mag for Sco-Cen. Spectroscopic indicators: K2-284 exhibits a weak Hα absorption feature with emission fill- ing in the wings of the line (Fig. 12). This is 0.40.60.81.01.21.4B-V [mag]15102040Rotation period [d]50 Myr125 Myr500 MyrPraesepePleiades12345V-Ks [mag]0.11102040Rotation period [d]Praesepe (790 Myr)Pleiades (125 Myr)The Adolescent Sub-Neptune K2-284 b 19 Figure 10. NUV and NIR color-color diagram showing empirical isochrones of Findeisen et al. (2011) and members of young stellar populations. K2-284 is indicated by the white star. Figure 11. Li I 6708 A equivalent width versus Teff for members of Pleiades (Soderblom et al. 1993), IC 2391/2602 (Randich et al. 2001), and young moving groups (Mentuch et al. 2008; Kraus et al. 2014). K2-284 is indicated by the white star. Figure 12. Sections of the HIRES spectra used as age diagnostics. Chromospheric emission in the Ca II H&K lines is clearly detected (top panels) as well as H emission. The Hα profile shows ab- sorption with emission filling in the wings of the line (middle panel), reminiscent of slowly-rotating late-type stars in the Pleiades and some G-type stars in α Per. The Li I 6708 A absorption line is clearly not present, which is broadly consistent with slowly-rotating mid-K dwarfs in the Pleiades and stars of similar Teff in moving groups with ages >20 -- 50 Myr. consistent with the model line profiles produced for weakly active dwarf stars in Cram & Mullan (1979). Hα profiles of this type have been ob- served for some of the most slowly-rotating late- type stars in the Pleiades, e.g. the M0 member SK 17 (Stauffer et al. 2016), some G-type mem- bers of α Per (Stauffer et al. 1989), as well as the M-type Praesepe planet host K2-95 (Obermeier et al. 2016). At the age of the Pleiades, there is a transition at mid-K spectral types where nearly all earlier type stars show Hα in absorp- tion and at later types nearly all show the line in emission (Stauffer & Hartmann 1987). In α Per, this transition occurs approximately at a spectral type of K6 (Prosser 1992). Thus, the 0.20.40.60.81.0J-Ks [mag]4681012NUV-J [mag]Upper Sco (10 Myr)Sco-Cen (15-25 Myr)Tuc-Hor (45 Myr)Hyades (750 Myr)30004000500060007000Teff [K]0100200300400500600700EW Li 6708 [mÅ]Pleiades (125 Myr)AB Dor (150 Myr)IC 2391/2602 (50 Myr)Tuc-Hor (45 Myr) Pic (23 Myr) Cha (11 Myr)TW Hya (10 Myr)3932393439360100200300Ca II K3966396839700100200300Ca II HH6550.06552.56555.06557.56560.06562.56565.06567.56570.050075010001250CountsH6702670467066708671067126714671667186720Wavelength [Å]5001000Li I 6707.8Ca I 6717.720 David et al. lack of strong Hα emission in K2-284 is at least consistent with expectations of other stars of a similar mass and age, and in fact some mem- bers of Sco-Cen ((cid:46)20 Myr) with a similar ef- fective temperature also show Hα in absorption (Pecaut & Mamajek 2016). Similar to other late-type stars in young moving groups, K2-284 exhibits weak emission in other Balmer lines, including H (seen in Fig. 12), Hζ, and Hη. We do not detect Li I 6708 A within the spectrum of K2-284. From the HIRES spec- trum, we estimated an upper limit to EW(Li) of <20 mA. This is not unexpected given that some late-K dwarfs with ages (cid:38)20 Myr are ob- served to show significant lithium depletion (see Fig. 11). Depletion of lithium below detectable levels has been observed in mid- to late-K mem- bers of IC 2391 and 2602 (∼50 Myr Barrado y Navascu´es et al. 2004; Dobbie et al. 2010), AB Dor (149+51−19 Myr, Bell et al. 2015), and Tuc-Hor (45± 4 Myr, Bell et al. 2015). In the 125 Myr-old Pleiades, Soderblom et al. (1993) found that mid- to late-K stars exhibit a wide range of Li I 6708 A equivalent widths, of ap- proximately 20 -- 300 mA. Furthermore, at the age of the Pleiades, some stars of a similar mass or color to K2-284 have yet to spin down. Bou- vier et al. (2017) have found that more slowly rotating Pleiads in a given mass range also tend to have weaker lithium absorption. Considered together, the rotation and lithium properties of K2-284 are consistent with Pleiades-aged or younger mid- to late-K dwarfs. In Figure 11 we show the distribution of Li I 6708 A equiva- lent width measurements as a function of Teff for members of young moving groups and clusters. H-R diagram and stellar density: Since the star is on or very nearly on the main se- quence, where evolution is slow for these low- mass stars, isochronal age estimates carry large uncertainties. Nevertheless, as we estimated the mass from interpolation between the PAR- SECv1.2S models (Bressan et al. 2012; Chen et al. 2014), we also estimated the age in the theoretical H-R diagram using the spectroscopic Teff and the Stefan-Boltzmann luminosity. Be- cause K2-284 is expected to be near or on the main sequence, the mean stellar density from the transit fits is also not particularly useful for constraining the stellar age, in part due to the fact that the impact parameter is not tightly constrained by the K2 data. Regard- less, we also estimated the stellar age from the directly-determined stellar density distribution (from the eccentric orbit transit fit discussed in § 3.1), a normal distribution in Teff, and the PARSECv1.2S models. Though not very pre- cise, the isochronal age estimates (through the H-R diagram or the mean stellar density) do provide a consistent lower limit of 30 -- 70 Myr. From the lack of lithium, it is very unlikely the star is as young as the β Pic moving group (23 ± 3 Myr, Mamajek & Bell 2014). With respect to the lithium levels in other young low-mass stars, ages corresponding to the moving groups Tuc- Hor (Kraus et al. 2014), AB Dor (Mentuch et al. 2008) or the clusters IC 2391/2602 (Randich et al. 2001) would seem plausible. However, a color-absolute magnitude diagram analysis pre- sented below suggests such young ages are un- likely. Isochronal age estimates are notoriously uncertain for main-sequence stars, and the age distributions resulting from both our H-R di- agram and stellar density analyses are highly skewed with long tails to old ages but very clear peaks around ∼100 Myr. To account for this, the isochronal ages we quote in Table 4 are the modes of the distributions resulting from the Monte Carlo error analysis, with the lower and upper bounds given by the 1% and 67% per- centiles. We found this choice more adequately describes the bulk of the probability density around the peaks of each distribution. For com- parison, the median, 16th, and 84th percentiles of the age distributions are 650+280−460 Myr and The Adolescent Sub-Neptune K2-284 b 21 Figure 13. Violin plot demonstrating the kernel density estimates for stellar age distributions re- sulting from different age-dating methods discussed in § 3.5. 430+1000−260 Myr, for the H-R diagram and stellar density analyses, respectively. Color-absolute magnitude diagram: We placed K2-284 in a color-absolute magnitude diagram using the Gaia photometry and par- allax, and compared it with the positions of young cluster members from Gaia Collabora- tion et al. (2018b). For comparison, we also included "field" stars observed by the K2 mis- sion (Fig. 14). The field star data were col- lected from the Gaia-K2 cross-match compiled by Megan Bedell.3 From this empirical analysis we conclude that K2-284 is likely older than α Per (∼70 Myr). In Table 4, we report several determinations of the host star age derived through the differ- ent methods described above. We also show the resulting age distributions from these methods and Monte Carlo error propagation in the var- ious input parameters in Figure 13. While the age indicators discussed above are statistical in nature, they present a consistent picture of a star that is (1) on or very nearly on the ZAMS, (2) unlikely to be as young as the youngest mov- ing groups in the solar neighborhood, and (3) almost certainly younger than the Hyades or Praesepe. The H-R diagram and stellar density 3 https://gaia-kepler.fun Figure 14. Color-absolute magnitude diagram for K2-284 as well as young cluster members and field stars, for comparison. K2-284 is apparently on the main sequence, with an age that is likely older than α Per, IC 2602, or IC2391, but consistent with the locations of Pleiades members and field stars. analyses are not precise age indicators in this case, but they at least present consistent lower limits to the age of >30 -- 40 Myr (at 68% con- fidence) or >10 Myr (at 95% confidence). Due to the large uncertainty in log R(cid:48) HK, our chro- mospheric activity age distributions also have long tails to unrealistically old ages, but we can still derive lower limits of >20 Myr (at 68% confidence) or >10 Myr (at 95% confidence). The NUV emission levels suggest an age of 45 -- 270 Myr at 68% confidence or 20 -- 640 Myr at 95% confidence, though we note the Findeisen et al. (2011) study calibrated the NUV/NIR age relations using cluster ages that have since been revised. Our tightest age constraints re- sult from the gyrochronology relations, which suggest 95% confidence intervals in age of 100 -- 160 Myr or 200 -- 350 Myr depending on the pre- ferred calibration. Considered collectively, these independent age estimates are consistent with a stellar age of τ∗ = 120+640−20 Myr (corresponding to the mode and 68% confidence interval of the age distribu- tion resulting from combining each of the differ- ent methods and weighting them equally). The 101001000Age [Myr]Stellar densityH-R diagramGyrochronology (B07)Gyrochronology (MH08)NUV emissionChromospheric activity1.01.52.02.53.0(GBP-GRP)0 [mag]567891011MG [mag]FieldPleiades PerIC2602/IC2391EPIC 24726726722 David et al. long tail towards older ages is due to the H-R di- agram and stellar density analyses as well as the uncertain log R(cid:48) HK value. Prior to the release of Gaia DR2, the kinematics of K2-284 were suggestive of membership to the Cas-Tau asso- ciation. However, the newly available parallax suggests this interpretation is unlikely and we leave a detailed investigation on the existence, membership, and substructure of the Cas-Tau association to a future work. Nevertheless, K2- 284 is clearly young, with an age that is likely close to that of the Pleiades. 4. DISCUSSION At first glance, K2-284 b appears fairly typ- ical when compared with other transiting sub- Neptunes receiving similar incident flux. That is, K2-284 b does not reside in a region of par- ticularly low occurrence in the plane of planet radius and insolation flux (see Figure 10 of Ful- ton et al. 2017). Thus, at least some young (<1 Gyr) sub-Neptunes superficially resemble the statistically older population uncovered by Kepler. This much was known for slightly more mature planets in the (cid:39)600 -- 800 Myr-old Hyades and Praesepe clusters, and we can now extend this conclusion to younger ages. However, the stars in the California-Kepler Survey are all more massive than K2-284 (Pe- tigura et al. 2017). When compared to other small transiting planets around low-mass stars, K2-284 b does appear to reside in the large- radius tail of the size distribution for close-in sub-Neptunes. This is apparent in both the planet radius versus period and planet radius versus insolation flux planes for low-mass hosts (Figures 15 and 16). In this case, it seems clear the K2 photometry of K2-284 are sensitive to planets much smaller than K2-284 b, though injection and recovery tests would be needed to quantify how sensitive the data are. In any event, while we can not be sure that the rela- tively large size of K2-284 b is due to its young age, it at least does not appear to be merely a consequence of observational bias. Other transiting planets around young, low- mass stars also appear to be uncharacteristically large (Fig. 16), which has now been pointed out numerous times (e.g. Mann et al. 2016; David et al. 2016a; Obermeier et al. 2016) due to the discovery of over a dozen transiting planets around stars in clusters and associations from K2 photometry. However, most transiting plan- ets found around young cluster or field stars of earlier spectral types do not appear to be clear outliers in the period-radius diagram (e.g. Cia- rdi et al. 2017; Mann et al. 2017b; Livingston et al. 2018a; David et al. 2018), with the no- table exception of the apparently single planet K2-100 b (Mann et al. 2017a). Why might young planets around low-mass stars appear as outliers in the period-radius di- agram, while planets of the same age around earlier-type stars seem to reflect the field planet population? One possible explanation for this observed behavior is provided by the theory of photo-evaporation. In the photo-evaporation framework, atmospheric escape is driven by X- ray and EUV radiation from the host star. A relevant quantity for interpreting the photo- evaporation history of any given planet is thus the time-integrated X-ray exposure, moreso than the current bolometric insolation, as pointed out in Owen & Wu (2013). A star's X-ray luminosity is highest when it is young and the X-ray emission is in the so-called "sat- urated" regime (LX/Lbol ∼ 10−3, Gudel 2004). After about 100 Myr, corresponding to a typ- ical pre-main-sequence lifetime, a star's X-ray luminosity declines steeply with age (Jackson et al. 2012; Tu et al. 2015). Relative to solar- type stars, low-mass stars are observed to satu- rate at higher values of LX/Lbol (Jackson et al. 2012), and thus they are expected to be more efficient at eroding planetary atmospheres, with an efficiency that scales as M−3∗ at a fixed Fbol The Adolescent Sub-Neptune K2-284 b 23 Figure 15. The distribution of small transiting planets in the plane of planet radius and orbital period for the full California-Kepler Survey sample, at left, and only "low-mass" hosts (<0.97 M(cid:12)) at right. K2-284 b is indicated by the white star. Overlaid are contours of completeness-corrected occurrence rates (Fulton et al. 2017; Fulton & Petigura 2018). (Lopez & Rice 2016, and references therein). As a result, the maximum planet radius at a given bolometric exposure varies substantially across different spectral types, while the maximum ra- dius at a given X-ray exposure appears to be less sensitive to the host-star type, as shown by Owen & Wu (2013) and discussed further in Hirano et al. (2018). However, it is also important to keep in mind that important degeneracies likely influence ob- served exoplanet populations. For example, it has been shown for solar-type stars that the occurrence of warm sub-Neptunes is higher for metal-rich stars (Petigura et al. 2018). Notably, the Hyades and Praesepe clusters, where some of the anomalously large, young transiting plan- ets have been found, are significantly metal- rich (Pace et al. 2008; Cummings et al. 2017). Thus, in order to separate age-dependent and metallicity-dependent trends in e.g. planetary radii, one must compare the planet populations in these clusters to field stars of a similar metal- licity. Additionally, Fulton & Petigura (2018) have recently examined the stellar-mass depen- dence of the radius gap using high-precision stel- lar radii enabled by Gaia parallaxes. Those au- thors find evidence that the bimodal distribu- tion of planet sizes shifts to smaller sizes around later-type stars, which might indicate that low- mass stars produce smaller planet cores. Thus, differences in the sizes of planets around low- mass and solar-type stars may not only reflect scalings in the photo-evaporation efficiency, but also in the initial core-mass function. The best way to bring clarity to these issues is through the characterization of larger samples of exo- planets around stars that exhibit a wide range of diversity in mass, metallicity, and age. It is also notable that the young planets that appear most clearly as outliers in the period- radius plane are all apparently single planet sys- tems (K2-25 b, K2-33 b, K2-95 b, K2-100 b, K2- 284 b), while those that appear more similar to the field planet population occur in multi-planet systems (K2-136, K2-233, K2-264). However, the statistics are simply too small to make a meaningful conclusion about the differences be- tween young single- and multi-planet systems at this point. Ultimately, a comparison between the typi- cal densities of young and old planets may be more elucidating than simply comparing radii. This requires a determination of the planet's mass. From the planet radius distribution, 131030100Orbital period [days]1.01.52.43.5Planet Size [Earth radii]typicaluncert.0.0000.0050.0100.0150.0200.0250.0300.0350.040Relative Occurrence131030100Orbital period [days]1.01.52.43.5Planet Size [Earth radii]typicaluncert.0.0000.0050.0100.0150.0200.025Relative Occurrence24 David et al. we calculated a predicted mass for K2-284 b of 8.5+6.4−3.8M⊕ using the forecaster4 tool in python, which is based on the Chen & Kip- ping (2017a) mass-radius relations for exoplan- ets. For this range of plausible planet masses and the stellar mass we adopt, we calculated an expected Doppler semi-amplitude of 2.4 -- 7.7 m s−1. Notably, existing exoplanet mass- radius relations are calibrated using field-aged planets. If sub-Neptunes as young as K2-284 b are less dense at early times, then the true Doppler amplitude may be on the lower end of the range quoted. While the expected Doppler amplitude is within reach of current precision RV instruments, the relatively high stellar ac- tivity will likely present challenges. Given the measured chromospheric activity level for K2- 284, it is likely the RV jitter is greater than 30 m s−1 and possibly larger than 100 m s−1 (Hillenbrand et al. 2015). The RV jitter may also be approximated from the amplitude of photometric variability and v sin i∗, from the equation σRV = rmsK2 × v sin i∗, which yields 33 m s−1, considerably larger than the ex- pected signal from the planet. Since the star is brighter and activity should be lower at in- frared wavelengths, it would be advantageous to measure the planet's mass with an IR pre- cision spectrograph such as the PARVI instru- ment planned for Palomar Observatory or one of many other spectrographs in operation or de- velopment (Plavchan et al. 2015). Interestingly, no transiting planets have yet been confirmed in the Pleiades, despite system- atic searches within the K2 data of ∼1000 mem- bers (Rizzuto et al. 2017; Gaidos et al. 2017). A single candidate was reported by Rizzuto et al. (2017), but the planet has not yet been val- idated and the probability of Pleiades mem- bership was estimated to be 62%. By com- parison, eight confirmed transiting planets and 4 https://github.com/davidkipping/forecaster Figure 16. The distribution of confirmed, small transiting planets from the NASA Exoplanet Archive (Akeson et al. 2013) in the plane of planet radius and insolation flux. Planets transiting stars in clusters or associations are circled in red. K2- 284 b is indicated by the red star. Each panel cor- responds to a different range in host star spectral type (annotated at top right). K2-284 b is on the larger end of known, close-in sub-Neptunes around stars of a similar spectral type. A number of other planets orbiting cool young cluster stars also appear to be anomalously large. The blue region in the top panel indicates the hot planet desert described in Lundkvist et al. (2016). 0246810Radius [R]SaturnNeptuneF5-K0MeVEMaK2-100 b0246810Radius [R]SaturnNeptuneK0-K5Insolation flux [S]0246810Radius [R]SaturnNeptuneK5-M0EPIC 247267267 bK2-127 bK2-43 bK2-45 bKepler-1314 bKepler-1321 b10-210-1100101102103104Insolation flux [S]0246810Radius [R]SaturnNeptuneM0-M5K2-104 bK2-264 cGJ 3470 bGJ 436 bK2-137 bK2-14 bK2-22 bK2-25 bK2-33 bK2-95 bKepler-1624 bKepler-1628 bThe Adolescent Sub-Neptune K2-284 b 25 one candidate have been reported in Praesepe (Mann et al. 2017a; Libralato et al. 2016; Ober- meier et al. 2016; Pepper et al. 2017; Rizzuto et al. 2018; Livingston et al. 2018b), a cluster with a distance and metallicity not much differ- ent from the Pleiades and for which a similar number of members were observed by K2. In the Hyades, four transiting planets around two hosts have been found in a search of <200 mem- bers (Mann et al. 2016, 2017b; Ciardi et al. 2017; Livingston et al. 2018a; David et al. 2016b), in addition to a single-transit planet candidate (Vanderburg et al. 2018). An important dif- ference between the clusters is that at the age of the Pleiades most members are spinning as rapidly as they ever will, while Praesepe and Hyades stars have spun down considerably and are thus more amenable to transit searches. It is also possible that the Pleiades members show enhanced photometric activity (in the form of larger and more frequent flares, larger variabil- ity amplitudes, and/or more rapidly evolving spot patterns), making the removal of these trends more difficult. It may be tempting to ascribe the lack of planets in the Pleiades (to this point) to some physical mechanism such as ongoing orbital migration or differences in the cluster environments. However, with an age un- likely to be much older than the Pleiades, the case of K2-284 b highlights the importance of taking a holistic approach towards the compar- ison of planet occurrence rates at young and old ages. 5. CONCLUSIONS We report the discovery of K2-284 b, a transit- ing sub-Neptune orbiting a young (τ = 120+640−20 Myr), low-mass star. The kinematics of K2- 284 prior to Gaia DR2 were suggestive of mem- bership to the poorly-studied Cas-Tau associ- ation, which we examined here. However, the Gaia parallax places the star at a distance that seems to be incompatible with that interpre- tation. Nevertheless, through a detailed stel- lar age analysis using multiple indicators of youth we were able to find evidence for a self- consistent Pleiades-like age that suggests the planet host may be a zero-age main-sequence star. The collection of young transiting planets are important benchmarks for photo-evaporation models, which predict the mass-loss evolution of close-in planets. The majority of photo- evaporation driven mass-loss is expected to oc- cur within the first ∼100 Myr of a star's life, when stellar XUV fluxes are highest and when the planet's surface gravity is expected to be lower due to ongoing contraction (Owen & Wu 2013; Lopez & Fortney 2013). Observing photo- evaporation in action requires a sample of young transiting planets around relatively bright stars and an effective probe of atmospheric escape. As discussed in § 4, it will be necessary to use one of the new generation NIR spectrographs to measure the mass of K2-284 b. Finally, young exoplanets are useful for con- straining migration scenarios and timescales. Presently, it is unclear when the population of close-in planets assembled. By refining the ages of known exoplanet host stars and surveying young stellar populations with greater intensity, it may be possible to observe temporal evolution in the orbital properties (periods, eccentricities, obliquities) of exoplanets. Any such evolution- ary trends could be important clues about the dynamical histories and formation scenarios of close-in exoplanets. c(cid:13) 2018. All rights reserved. This research was carried out at the Jet Propulsion Labora- tory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration. We thank the anony- mous referee for comments which improved this manuscript. TJD and EEM acknowledge support from the Jet Propulsion Laboratory Exoplanetary Science Initiative. MB acknowl- 26 David et al. edges support from the North Carolina Space Grant Consortium. This work was performed in part under contract with the California In- stitute of Technology (Caltech)/Jet Propulsion Laboratory (JPL) funded by NASA through the Sagan Fellowship Program executed by the NASA Exoplanet Science Institute. This paper includes data collected by the Kepler mission. Funding for the Kepler mission is provided by the NASA Science Mission directorate. Some of the data presented in this paper were ob- tained from the Mikulski Archive for Space Telescopes (MAST). STScI is operated by the Association of Universities for Research in As- tronomy, Inc., under NASA contract NAS5- 26555. Support for MAST for non-HST data is provided by the NASA Office of Space Sci- ence via grant NNX09AF08G and by other grants and contracts. Some of the data pre- sented herein were obtained at the W. M. Keck Observatory, which is operated as a scientific partnership among the California Institute of Technology, the University of California and the National Aeronautics and Space Adminis- tration. The Observatory was made possible by the generous financial support of the W. M. Keck Foundation. The authors wish to rec- ognize and acknowledge the very significant cultural role and reverence that the summit of Maunakea has always had within the in- digenous Hawaiian community. We are most fortunate to have the opportunity to conduct observations from this mountain. This research has made use of the VizieR catalogue access tool, CDS, Strasbourg, France. The original description of the VizieR service was published in A&AS 143, 23. The Pan-STARRS1 Surveys (PS1) and the PS1 public science archive have been made possible through contributions by the Institute for Astronomy, the University of Hawaii, the Pan-STARRS Project Office, the Max-Planck Society and its participating insti- tutes, the Max Planck Institute for Astronomy, Heidelberg and the Max Planck Institute for Extraterrestrial Physics, Garching, The Johns Hopkins University, Durham University, the University of Edinburgh, the Queen's Univer- sity Belfast, the Harvard-Smithsonian Center for Astrophysics, the Las Cumbres Observa- tory Global Telescope Network Incorporated, the National Central University of Taiwan, the Space Telescope Science Institute, the National Aeronautics and Space Administration under Grant No. NNX08AR22G issued through the Planetary Science Division of the NASA Sci- ence Mission Directorate, the National Science Foundation Grant No. AST-1238877, the Uni- versity of Maryland, Eotvos Lorand Univer- sity (ELTE), the Los Alamos National Labora- tory, and the Gordon and Betty Moore Founda- tion. This work has made use of data from the European Space Agency (ESA) mission Gaia (https://www.cosmos.esa.int/gaia), pro- cessed by the Gaia Data Processing and Analy- sis Consortium (DPAC, https://www.cosmos. esa.int/web/gaia/dpac/consortium). Fund- ing for the DPAC has been provided by national institutions, in particular the institutions par- ticipating in the Gaia Multilateral Agreement. Facilities: FLWO:1.5m (TRES), Keck:I (HIRES), Keck:II (NIRC2), Kepler, PS1, Shane (ShARCS) Software: emcee (Foreman-Mackey et al. 2013), forecaster (Chen & Kipping 2017b), iso- classify (Huber et al. 2017), k2sc (Aigrain et al. 2016), k2sff (Vanderburg & Johnson 2014), py- transit (Parviainen 2015), radvel (Fulton et al. 2018), vespa (Morton 2015) The Adolescent Sub-Neptune K2-284 b 27 APPENDIX A. THE CAS-TAU ASSOCIATION AND ITS TURNOFF AGE Cas-Tau was first formally proposed as an association by Blaauw (1956), based on the common motions of 49 B-stars covering a remarkably large patch of sky of about 100◦ × 140◦. The association shares motions with and spatially surrounds the α Persei (Per OB3) cluster, which led Blaauw to suggest a common origin for the two groups. Indeed, Rasmuson (1921) had already noted the kinematic group was not limited to the central α Per cluster, but that several other B- and A-stars formed a co-moving stream extending well beyond the cluster core. In the years following Blaauw's work, the status of Cas-Tau as a bona fide moving group was debated in the literature on the basis of radial velocities (Petrie 1958) and large scatter in the color-Hβ relation (Crawford 1963). However, based on Hipparcos parallaxes, de Zeeuw et al. (1999) concluded that Cas-Tau is indeed a physical association that likely shares a common origin with α Per, though only a third of Blaauw's original sample were finally regarded as members. Today, the low-mass membership of Cas-Tau remains essentially unknown. An X-ray survey in the direction of Taurus found evidence for a population of stars that are older and more widely distributed than the CTTS in the Taurus-Auriga star-formation complex (Walter et al. 1988). Those authors found that this distributed older population outnumbers the CTTS population by a factor of 10:1, and there are suggestions that this older population includes members of the Cas-Tau association (Hartmann et al. 1991; Walter & Boyd 1991). Assuming that all of Blaauw's original B-stars are indeed Cas-Tau members, Hartmann et al. (1991) argued based on expectations from the initial mass function that the projected surface density of members with masses (cid:38) 0.8M(cid:12) should be about 0.2 per square degree. In hindsight, that may be an overestimate given that the Hipparcos study found many of Blaauw's original sample are not likely to be members. Nevertheless, within the K2 Campaign 13 field one might expect a couple dozen members in this mass range and an even larger number of lower-mass members. Since the area of Cas-Tau is so large on the sky, additional members might have plausibly been observed during other K2 campaigns. The precise age of Cas-Tau is not well known, in part due to our incomplete knowledge of the low-mass members. From the kinematics of the originally proposed members, Blaauw (1956) derived an expansion age for Cas-Tau in the range of 50 -- 70 Myr. Due to the common kinematics between the associations, it is generally believed that Cas-Tau is younger than or coeval with α Per. Early examinations of the main sequence turnoff for α Per found ages of 50 Myr using models with no convective overshoot (Mermilliod 1981; Meynet et al. 1993). The age of α Per has since been refined using the lithium depletion boundary (LDB) technique, with estimates of 90 ± 10 Myr (Stauffer et al. 1999), 85 ± 10 Myr (Barrado y Navascu´es et al. 2004), and most recently 80 ± 11 ± 4 Myr (Soderblom et al. 2014). The LDB ages are broadly consistent with age estimates of 80 Myr from a color-magnitude diagram (CMD) of the lower main sequence (Prosser 1992), 80 Myr from an upper main sequence CMD age using models with moderate convective overshoot (Ventura et al. 1998), and 70 Myr from an H-R diagram of the upper main sequence (David & Hillenbrand 2015). To our knowledge, the only determination of a turnoff age for Cas-Tau is the estimate of 20-30 Myr from de Zeeuw & Brand (1985). Motivated by our suggestion that the planet host K2-284 belongs to the association, we derive a new turnoff age here. We began with the list of 83 B- and A-type members proposed by de Zeeuw et al. (1999). For each of the proposed members, we gathered trigonometric 28 David et al. parallaxes from the Gaia TGAS catalog (Gaia Collaboration et al. 2016a,b) when available and from the Extended Hipparcos compilation otherwise (XHIP, Anderson & Francis 2012). For each star we then gathered U BV photometry from Mermilliod (2006) and uvbyβ photometry from Paunzen (2015). Of the 83 proposed members, 19 stars were missing both U BV and uvbyβ photometry from the aforementioned compilations, while 5 stars lacked only the U BV data and 11 stars lacked only the uvbyβ data. Nearly all of the stars missing photometry have spectral types of B8 or later, and given that we determine the main-sequence turnoff to be around spectral type B2 for Cas-Tau, these stars contribute little information to the turnoff age anyhow. Our motivation for including both U BV and uvbyβ photometry was for the purposes of consistency checks. We ultimately derived the turnoff age from U BV photometry, so stars missing those data were excluded from our analysis, and any star that lacked both U BV and uvbyβ photometry was not included in our various consistency checks described below. To guide our analysis, we additionally gathered spectral types from Skiff (2014), v sin i measurements and multiplicity information from Abt et al. (2002). For each star we also performed literature searches for further information on multiplicity and to vet for eclipsing binaries (EBs). Many of the proposed members are reddened. We determined the amount of reddening for each star using the U BV photometry and the revised Q-method presented in Pecaut & Mamajek (2013). For those stars with uvbyβ photometry we used the iterative dereddening scheme of Shobbrook (1983) to determine an independent value for the extinction. Among the stars with both sets of photometry, we found the A(V ) values derived from the Q-method and the uvbyβ iterative method to be well- described by a one-to-one relation with a scatter of 0.066 mag. From an empirical relation between (b− y)0 and (B− V )0 for B-type stars (Crawford 1978), we also compared the intrinsic (B− V ) colors from the two different dereddening methods and found these to be in good agreement with a scatter of 0.01 mag. We ultimately used the intrinsic colors and A(V ) values from the U BV photometry, but we adopted 0.01 mag as the uncertainty in (B − V )0 for our turnoff age analysis to account for the different estimates provided by the uvbyβ photometry. Using the intrinsic (B−V )0 colors and MV magnitudes calculated from the V -band photometry and trigonometric parallaxes, we then proceeded to estimate the turnoff age from comparison with the PARSECv1.2S evolutionary models (Bressan et al. 2012). The uncertainties in the MV magnitudes were determined from Monte Carlo error estimation, accounting for the uncertainties in V magnitudes and the parallaxes. For high-mass stars such as those considered here, the PARSECv1.2S models are transformed into the observational system through the use of Castelli & Kurucz (2004) model atmospheres, Bessell (1990) U BV RI passbands, and the zero-points presented in Ma´ız Apell´aniz (2006). We used models with a solar metallicity of Z=0.0152 (Caffau et al. 2011) for this analysis. From the color-magnitude diagram, it is apparent that there is a significant amount of scatter around the turnoff. It is possible that there are interlopers in the de Zeeuw et al. (1999) sample, so in an attempt to address this issue we considered only stars with membership probabilities ≥90%, where the probability values originate from those authors. We additionally excluded two high-probability members since these are emission line stars. These stars are HD 9709 (HIP 7457), a B7IV/Vne shell star, and φ Per (HIP 8068), a B1.5V:e shell star and double-lined spectroscopic binary. Furthermore, several of the proposed members are eclipsing binaries. These stars are 1 Per (HIP 8704), τ Ari The Adolescent Sub-Neptune K2-284 b 29 Figure 17. Left: Color-magnitude diagram for proposed members of Cas-Tau. Black points are the high probability members we used to determine the turnoff age, while the grey points show members that were excluded for the reasons described in the text. Isochrones from the PARSECv1.2S models are indicated by the colored curves. The grey points indicate stars rejected for reasons explained in the text. Right: Histogram of turnoff ages resulting from 104 Monte Carlo simulations. (HIP 15627), 17 Aur (HIP 24740), and 15 Cam (HIP 24836).5 We ultimately excluded 1 Per and 17 Aur on the basis of large eclipse depths (>0.3 mag) and included the other two systems given their more moderate eclipse depths (Avvakumova et al. 2013). In the course of our analysis, we also found that two of the proposed members are surrounded by reflection nebulae (HD 26676 and HD 17443). Despite the additional extinction, these stars do not appear to be obvious outliers in the CMD and were included in the age analysis. Using a fine grid of isochrones (∆ log τ = 0.0025 dex) with ages between 106 and 109 yr we fit an isochrone of each age to the data and evaluated χ2. We determined the uncertainty on the turnoff age from 104 Monte Carlo simulations in which a new χ2 min age was calculated from perturbed MV and (B− V )0 values for each star. In this analysis the perturbed MV and (B− V )0 values were drawn from normal distributions in accordance with that star's individual errors. For those stars missing V magnitude error estimates, we assumed an error of 0.01 mag. Ultimately, we found a turnoff age of τCas−Tau = 46 ± 8 Myr, where the value and uncertainty are the median and standard deviation, respectively, of the distribution of ages from the Monte Carlo simulations. This age is in good agreement with the original kinematic estimate of 50 -- 70 Myr from Blaauw (1956), and somewhat younger than the lithium depletion boundary age (80 ± 11 ± 4 Myr, Soderblom et al. 2014) and turnoff ages derived for the α Per cluster (80 Myr, Ventura et al. 1998). We note that the analysis above has not made use of the more precise Gaia DR2 parallaxes. A preliminary analysis utilizing the new parallaxes and eliminating the stars HIP 2377, HIP 8387, HIP 20171 (which appear as outliers in the parallax distribution of proposed members), suggest a slightly older turnoff age of τ = 59+14−8 Myr and reveal a potentially bimodal age distribution. We leave a more detailed study of the age and substructure of this proposed association to a further study. 5 µ Eri (HIP 22109) is also listed as an EB in SIMBAD, but we did not find published evidence in support of this interpretation in the literature. This star was excluded from the age analysis anyhow on the basis of its low membership probability. 0.30.20.10.00.1(B-V)0 [mag]321012MV [mag]10 Myr50 Myr100 Myr200 Myr203040506070Age [Myr]0.000.010.020.030.04Probability density30 David et al. . s r e b m e m u a T - s a C d e s o p o r P . 5 e l b a T V A ) g a m ( ) V − B ( E 0 ) B − U ( 0 ) V − B ( ) g a m ( ) g a m ( ) g a m ( B − U ) g a m ( V − B ) g a m ( V M ) g a m ( ) g a m ( ) s a m ( ) % ( V T p S . b o r P P I H e m a N 5 7 0 . 0 ± 9 9 2 . 0 3 2 0 . 0 ± 2 9 0 . 0 · · · · · · 0 0 0 . 0 · · · · · · 0 0 0 . 0 · · · · · · · · · · · · · · · · · · 4 7 0 . 0 ± 2 4 6 . 0 - 8 5 1 . 0 ± 5 3 1 . 0 - 0 2 0 . 0 ± 3 7 1 . 0 - 4 0 0 . 0 ± 6 6 0 . 0 - 1 6 0 . 0 ± 9 7 5 . 0 - 8 5 1 . 0 ± 5 3 1 . 0 - · · · · · · · · · 8 3 0 . 0 ± 9 5 1 . 0 7 2 0 . 0 ± 1 3 4 . 0 9 3 0 . 0 ± 7 2 1 . 0 2 1 0 . 0 ± 9 4 0 . 0 8 0 0 . 0 ± 2 3 1 . 0 2 1 0 . 0 ± 9 3 0 . 0 1 2 0 . 0 ± 1 5 2 . 0 7 0 0 . 0 ± 7 7 0 . 0 0 0 0 . 0 0 0 0 . 0 · · · · · · 5 1 0 . 0 ± 7 3 6 . 0 - 7 0 0 . 0 ± 3 5 4 . 0 - 5 1 0 . 0 ± 6 0 5 . 0 - 0 2 0 . 0 ± 0 4 3 . 0 - 3 1 0 . 0 ± 6 6 2 . 0 - · · · 4 0 0 . 0 ± 1 7 1 . 0 - 2 0 0 . 0 ± 6 2 1 . 0 - 3 0 0 . 0 ± 8 3 1 . 0 - 4 0 0 . 0 ± 1 0 1 . 0 - 3 0 0 . 0 ± 4 8 0 . 0 - · · · 0 2 0 . 0 ± 0 4 3 . 0 - 0 8 4 . 0 - 0 1 2 . 0 - · · · 9 3 0 . 0 ± 5 7 2 . 0 2 1 0 . 0 ± 4 8 0 . 0 1 2 0 . 0 ± 1 7 5 . 0 - 5 0 0 . 0 ± 4 5 1 . 0 - 6 1 0 . 0 ± 2 1 5 . 0 - 0 1 0 . 0 ± 3 0 6 . 0 - 3 0 0 . 0 ± 9 5 3 . 0 - 9 0 0 . 0 ± 2 2 1 . 0 - 7 0 0 . 0 ± 6 0 0 . 0 · · · · · · 0 3 4 . 0 - · · · 4 0 0 . 0 ± 1 0 1 . 0 - 5 0 0 . 0 ± 7 0 0 . 0 - 0 0 1 . 0 - 9 0 0 . 0 ± 9 6 0 . 0 - 2 2 0 . 0 ± 8 3 0 . 0 0 1 0 . 0 ± 1 8 0 . 0 - 4 0 0 . 0 ± 6 6 0 . 0 - · · · · · · · · · 0 5 0 . 0 - · · · 0 4 0 . 0 ± 6 7 2 . 0 2 1 0 . 0 ± 4 8 0 . 0 5 1 0 . 0 ± 0 9 4 . 0 - 3 0 0 . 0 ± 4 3 1 . 0 - · · · · · · · · · · · · 0 3 0 . 0 ± 1 9 8 . 0 0 4 0 . 0 ± 7 5 3 . 0 3 2 0 . 0 ± 1 5 0 . 0 9 0 0 . 0 ± 6 7 2 . 0 2 1 0 . 0 ± 9 0 1 . 0 7 0 0 . 0 ± 6 1 0 . 0 5 1 0 . 0 ± 1 3 1 . 1 - 5 1 0 . 0 ± 0 9 2 . 0 - 2 1 0 . 0 ± 4 4 1 . 0 - 4 0 0 . 0 ± 8 1 3 . 0 - 3 0 0 . 0 ± 9 8 0 . 0 - 3 0 0 . 0 ± 3 5 0 . 0 - 4 3 0 . 0 ± 4 0 2 . 0 1 3 0 . 0 ± 4 1 0 . 0 0 4 0 . 0 ± 7 4 2 . 0 1 1 0 . 0 ± 3 6 0 . 0 9 0 0 . 0 ± 4 0 0 . 0 2 1 0 . 0 ± 6 7 0 . 0 4 1 0 . 0 ± 0 8 8 . 0 - 8 1 0 . 0 ± 2 9 5 . 0 - 5 1 0 . 0 ± 5 6 3 . 0 - 4 0 0 . 0 ± 2 4 2 . 0 - 5 0 0 . 0 ± 9 5 1 . 0 - 3 0 0 . 0 ± 6 0 1 . 0 - · · · · · · · · · · · · · · · · · · · · · · · · 0 4 0 . 0 ± 3 0 2 . 0 3 1 1 . 0 ± 9 7 0 . 0 2 1 0 . 0 ± 2 6 0 . 0 5 3 0 . 0 ± 4 2 0 . 0 4 1 0 . 0 ± 4 2 5 . 0 - 1 3 0 . 0 ± 8 1 4 . 0 - 3 0 0 . 0 ± 2 4 1 . 0 - 7 0 0 . 0 ± 8 1 1 . 0 - · · · · · · · · · · · · 0 4 0 . 0 ± 3 5 4 . 0 9 3 0 . 0 ± 8 4 0 . 0 2 1 0 . 0 ± 9 3 1 . 0 2 1 0 . 0 ± 5 1 0 . 0 5 1 0 . 0 ± 9 0 5 . 0 - 0 1 0 . 0 ± 3 9 4 . 0 - 3 0 0 . 0 ± 9 3 1 . 0 - 2 0 0 . 0 ± 5 3 1 . 0 - · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 2 1 0 . 0 ± 5 3 9 . 0 - 7 0 0 . 0 ± 2 4 0 . 0 - 9 0 0 . 0 ± 3 3 1 . 0 - 5 0 0 . 0 ± 7 3 0 . 0 - 0 1 2 . 0 - 0 2 0 . 0 9 0 0 . 0 ± 4 3 8 . 0 - 4 1 0 . 0 ± 1 9 5 . 0 - 8 0 0 . 0 ± 9 7 1 . 0 - 7 0 0 . 0 ± 5 5 1 . 0 - · · · · · · 7 1 0 . 0 ± 1 0 4 . 0 - 9 2 0 . 0 ± 4 9 0 . 0 - · · · 0 1 4 . 0 - · · · 0 0 0 . 0 5 0 0 . 0 ± 3 8 4 . 0 - 0 0 0 . 0 ± 0 2 1 . 0 - 0 1 3 . 0 - · · · 0 8 4 . 0 - 0 3 0 . 0 - · · · 0 8 0 . 0 - · · · · · · · · · · · · · · · · · · 3 7 4 . 0 ± 2 1 9 . 1 - 1 6 2 . 0 ± 3 1 9 . 0 - 6 4 1 . 0 ± 6 7 6 . 0 - 8 1 1 . 0 ± 2 7 5 . 0 - 7 1 2 . 0 ± 3 2 9 . 0 7 8 1 . 0 ± 7 0 1 . 2 - 8 9 3 . 0 ± 6 2 3 . 0 9 1 4 . 0 ± 4 4 5 . 1 - 7 8 0 . 0 ± 3 8 0 . 0 · · · · · · · · · 6 1 4 . 0 ± 3 1 5 . 0 - · · · 2 9 0 . 0 ± 5 5 6 . 2 - 9 4 3 . 0 ± 3 3 1 . 0 - 6 4 0 . 0 ± 8 2 2 . 1 · · · 8 0 3 . 0 ± 2 1 5 . 2 - 7 1 1 . 0 ± 4 3 1 . 2 - 5 4 1 . 0 ± 0 1 9 . 0 · · · 4 9 1 . 0 ± 6 3 8 . 0 - 2 9 0 . 0 ± 5 4 1 . 0 5 4 1 . 0 ± 4 4 0 . 1 · · · · · · · · · · · · · · · 5 1 0 . 0 ± 1 7 5 . 5 4 0 0 . 0 ± 2 4 9 . 5 4 5 0 . 0 ± 9 4 7 . 4 9 0 0 . 0 ± 7 9 8 . 7 0 7 1 . 6 6 4 0 . 0 ± 3 7 5 . 4 5 1 0 . 0 ± 2 5 8 . 7 3 1 0 . 0 ± 8 2 5 . 6 1 1 0 . 0 ± 1 5 5 . 5 · · · 2 1 0 . 0 ± 2 6 0 . 4 3 7 0 . 0 ± 8 0 5 . 5 9 0 0 . 0 ± 0 7 3 . 3 0 2 0 . 7 0 6 8 . 5 · · · · · · · · · 0 7 0 . 7 · · · 0 6 6 . 6 · · · 0 0 1 . 6 0 5 5 . 5 · · · 0 8 6 . 7 · · · · · · · · · · · · 3 6 2 . 0 ± 2 1 0 . 0 3 1 0 . 0 ± 9 2 1 . 0 0 5 0 . 0 ± 6 2 1 . 1 1 4 0 . 0 ± 3 5 2 . 0 · · · 9 7 0 . 0 ± 4 0 0 . 0 4 0 0 . 0 ± 0 4 0 . 0 5 1 0 . 0 ± 1 4 3 . 0 3 1 0 . 0 ± 8 7 0 . 0 · · · 3 8 1 . 0 ± 4 8 0 . 0 4 1 0 . 0 ± 7 6 4 . 0 - 8 2 0 . 0 ± 8 9 0 . 0 - 5 1 0 . 0 ± 4 2 6 . 0 - · · · 7 7 0 . 0 ± 3 3 0 . 0 3 0 0 . 0 ± 9 2 1 . 0 - 8 0 0 . 0 ± 9 3 0 . 0 - 4 0 0 . 0 ± 8 6 1 . 0 - · · · 9 5 0 . 0 ± 5 3 2 . 0 2 4 0 . 0 ± 1 1 4 . 0 8 1 0 . 0 ± 3 7 0 . 0 3 1 0 . 0 ± 5 2 1 . 0 0 2 0 . 0 ± 3 2 8 . 0 - 5 1 0 . 0 ± 1 3 2 . 0 - 6 0 0 . 0 ± 5 2 2 . 0 - 4 0 0 . 0 ± 5 7 0 . 0 - 6 0 1 . 0 ± 5 6 0 . 0 2 1 0 . 0 ± 9 3 4 . 0 - 9 1 0 . 0 ± 5 4 1 . 0 2 1 0 . 0 ± 7 3 0 . 0 2 0 0 . 0 ± 9 8 0 . 0 - 0 1 0 . 0 ± 2 0 3 . 0 1 2 3 . 0 ± 5 5 1 . 1 7 0 1 . 0 ± 7 2 4 . 0 - 8 5 1 . 0 ± 8 4 4 . 1 2 0 0 . 0 ± 3 6 5 . 7 0 0 0 . 0 ± 0 4 7 . 8 0 8 4 . 5 3 1 0 . 0 ± 1 7 7 . 0 - 4 1 0 . 0 ± 2 5 1 . 0 - 0 7 5 . 0 - · · · 0 9 0 . 0 - · · · 0 4 1 . 0 - 0 5 0 . 0 7 4 2 . 0 ± 7 2 0 . 2 - 2 4 1 . 0 ± 5 8 8 . 0 8 2 0 . 0 ± 5 3 8 . 4 0 5 0 . 7 · · · · · · · · · · · · d e u n i t n o c 5 e l b a T 3 6 . 0 ± 6 2 . 3 8 2 . 0 ± 9 2 . 4 3 4 . 0 ± 4 6 . 8 5 4 . 0 ± 4 6 . 3 8 3 . 0 ± 4 6 . 4 4 . 0 ± 4 0 . 4 5 . 0 ± 4 2 . 4 5 4 . 0 ± 8 4 . 2 2 3 . 0 ± 6 0 . 8 4 5 . 0 ± 2 . 3 4 6 . 0 ± 5 . 3 9 3 . 0 ± 2 . 7 8 5 . 0 ± 3 1 . 3 2 5 . 0 ± 4 5 . 3 2 . 0 ± 4 5 . 4 8 5 . 0 ± 6 7 . 3 5 2 . 0 ± 5 8 . 1 1 1 6 . 0 ± 3 2 . 2 3 3 . 0 ± 2 5 . 2 3 4 . 0 ± 2 9 . 7 8 4 . 0 ± 9 0 . 7 1 4 . 0 ± 7 . 4 7 3 . 0 ± 1 . 4 4 3 . 0 ± 1 3 . 8 6 2 . 0 ± 2 2 . 4 2 3 . 0 ± 3 7 . 4 5 . 0 ± 5 2 . 4 9 3 . 0 ± 5 3 . 3 9 3 . 0 ± 1 4 . 3 5 2 . 0 ± 2 4 . 5 7 7 . 0 ± 9 2 . 5 5 2 . 0 ± 9 4 . 3 2 3 . 0 ± 6 . 6 4 . 0 ± 9 9 . 4 8 . 0 ± 6 8 . 5 8 4 . 0 ± 8 2 . 4 8 3 . 0 ± 5 8 . 5 n I I I 9 B V I 5 B V 6 B n n V 8 B I I I 5 B V 9 B V 3 B V 5 . 9 B : 8 B 9 B a e n V / V I 7 B V 0 A a e : V 5 . 1 B V I 8 B V 5 . 9 B V 9 B V / V I 8 B V 2 B V 3 B 0 0 1 0 0 1 8 9 0 8 7 9 7 9 9 9 9 9 0 0 1 1 5 7 9 9 9 8 9 8 8 0 0 1 0 0 1 0 0 1 4 5 8 9 3 9 V I 9 B 0 0 1 8 9 9 9 9 8 5 9 9 9 1 2 9 1 7 7 3 2 4 7 4 2 5 0 5 2 7 4 6 2 6 6 8 2 4 0 5 3 7 3 4 4 2 6 0 5 6 6 5 5 3 1 8 5 0 8 4 6 7 5 4 7 8 8 9 7 8 6 0 8 8 0 1 8 7 8 3 8 1 5 5 8 4 0 7 8 6 8 8 8 6 5 6 9 0 9 8 9 4 2 9 0 1 4 4 9 0 1 6 7 9 1 D H 6 2 6 2 D H s a C 3 1 s a C λ 4 7 9 2 D H 1 9 2 3 D H s a C i m o 9 0 4 5 D H 2 0 3 R H 2 4 3 R H 9 4 3 7 D H 6 4 3 8 D H 9 0 7 9 D H 4 0 4 0 1 D H r e P φ 7 7 5 0 1 D H i r A 4 4 0 1 1 1 D H r e P 1 s a C 8 1 5 2 1 D H 4 4 8 2 1 D H 9 7 6 R H d n A 3 6 4 7 9 0 1 5 9 1 7 6 + D B 5 9 2 1 1 5 9 7 4 1 D H i S p V 9 B V 5 B V 5 B V 5 B V 9 B V 9 B 8 B 2 B s V 0 A V 7 B V 9 B V 4 B e : V I 2 B V I 9 B V 9 B 0 0 1 8 1 2 2 1 0 6 7 R H 9 9 3 5 4 2 1 5 8 4 6 1 D H 0 0 1 7 7 4 2 1 9 4 4 6 1 D H 4 7 3 9 6 8 0 0 1 0 0 1 5 7 7 8 8 9 3 0 0 3 1 1 9 2 5 6 + D B 4 2 1 3 1 7 2 3 3 1 0 3 3 3 1 7 8 8 4 1 5 6 0 5 1 0 2 5 5 1 1 3 5 5 1 9 5 3 7 1 D H 3 4 4 7 1 D H 1 8 9 9 1 D H 6 3 3 0 2 D H 0 1 5 0 2 D H 0 5 9 R H i r A g i s The Adolescent Sub-Neptune K2-284 b 31 ) d e u n i t n o c ( 5 e l b a T V A ) g a m ( ) V − B ( E 0 ) B − U ( 0 ) V − B ( ) g a m ( ) g a m ( ) g a m ( B − U ) g a m ( V − B ) g a m ( V M ) g a m ( ) g a m ( ) s a m ( ) % ( V T p S . b o r P P I H e m a N 0 6 0 . 0 ± 6 0 3 . 0 8 1 0 . 0 ± 6 1 2 . 0 6 4 0 . 0 ± 4 3 1 . 0 6 4 0 . 0 ± 4 4 1 . 0 · · · 8 1 0 . 0 ± 4 9 0 . 0 6 0 0 . 0 ± 7 6 0 . 0 4 1 0 . 0 ± 1 4 0 . 0 4 1 0 . 0 ± 4 4 0 . 0 · · · 1 6 0 . 0 ± 7 9 5 . 0 - 1 1 0 . 0 ± 5 6 6 . 0 - 8 1 0 . 0 ± 1 7 0 . 0 - 7 1 0 . 0 ± 3 9 1 . 0 - · · · 6 1 0 . 0 ± 1 6 1 . 0 - 3 0 0 . 0 ± 9 7 1 . 0 - 6 0 0 . 0 ± 1 3 0 . 0 - 4 0 0 . 0 ± 6 6 0 . 0 - · · · 8 4 0 . 0 ± 1 3 5 . 0 - 8 0 0 . 0 ± 8 1 6 . 0 - 8 0 0 . 0 ± 7 6 0 . 0 - 4 0 0 . 0 ± 2 1 1 . 0 - 6 4 0 . 0 - 0 6 1 . 0 - · · · 1 1 0 . 0 ± 2 2 0 . 0 - 0 1 0 . 0 · · · 1 5 2 . 0 ± 8 0 7 . 0 - 3 0 1 . 0 ± 0 3 7 . 0 - 3 1 1 . 0 ± 7 6 3 . 1 5 2 1 . 0 ± 5 8 8 . 0 · · · 8 4 0 . 0 ± 2 7 2 . 0 5 1 0 . 0 ± 3 8 0 . 0 9 1 0 . 0 ± 8 3 5 . 0 - 5 0 0 . 0 ± 6 4 1 . 0 - 3 1 0 . 0 ± 9 7 4 . 0 - 2 1 0 . 0 ± 2 6 0 . 0 - 4 9 1 . 0 ± 3 6 3 . 0 0 0 0 . 0 · · · · · · 0 0 0 . 0 · · · · · · 0 4 2 . 0 - · · · · · · 0 8 0 . 0 - · · · · · · 0 4 2 . 0 - · · · · · · 0 8 0 . 0 - · · · · · · · · · · · · · · · 2 1 0 . 0 ± 1 7 2 . 5 0 2 0 . 0 ± 1 4 3 . 5 6 6 0 . 7 0 0 1 . 6 · · · 0 1 3 . 6 · · · · · · · · · 0 2 0 . 0 ± 4 7 5 . 0 1 4 0 . 0 ± 7 2 3 . 0 8 2 0 . 0 ± 9 9 0 . 0 6 0 0 . 0 ± 5 7 1 . 0 3 1 0 . 0 ± 0 0 1 . 0 9 0 0 . 0 ± 0 3 0 . 0 0 0 0 . 0 · · · 0 0 0 . 0 · · · 7 1 0 . 0 ± 0 6 4 . 0 - 3 2 0 . 0 ± 4 9 5 . 0 - 4 1 0 . 0 ± 8 3 3 . 0 - 5 1 0 . 0 ± 4 7 2 . 0 - · · · 4 0 0 . 0 ± 7 2 1 . 0 - 6 0 0 . 0 ± 0 6 1 . 0 - 3 0 0 . 0 ± 0 0 1 . 0 - 9 1 0 . 0 ± 2 9 0 . 0 - · · · 4 1 0 . 0 ± 5 3 3 . 0 - 7 1 0 . 0 ± 2 2 5 . 0 - 1 1 0 . 0 ± 5 1 3 . 0 - 5 1 0 . 0 ± 4 7 2 . 0 - · · · 4 0 0 . 0 ± 8 4 0 . 0 9 0 0 . 0 ± 0 6 0 . 0 - 7 0 0 . 0 ± 0 7 0 . 0 - 9 1 0 . 0 ± 2 9 0 . 0 - 0 3 0 . 0 - 1 4 0 . 0 ± 4 5 4 . 0 3 1 0 . 0 ± 9 3 1 . 0 7 1 0 . 0 ± 9 1 6 . 0 - 4 0 0 . 0 ± 6 6 1 . 0 - 2 1 0 . 0 ± 0 2 5 . 0 - 0 1 0 . 0 ± 7 2 0 . 0 - · · · · · · · · · · · · · · · · · · 9 7 3 . 0 ± 2 6 3 . 0 - 4 0 1 . 0 ± 3 5 4 . 1 - 9 5 3 . 0 ± 7 0 3 . 0 - 6 6 0 . 0 ± 9 5 7 . 0 7 5 1 . 0 ± 9 0 6 . 0 7 9 0 . 0 ± 6 0 1 . 1 - · · · 4 8 0 . 0 ± 5 2 2 . 6 0 1 0 . 0 ± 0 8 2 . 4 0 1 0 . 0 ± 9 4 8 . 4 · · · 0 1 2 . 6 0 5 3 . 5 0 0 8 . 6 6 1 0 . 0 ± 6 7 1 . 0 5 0 0 . 0 ± 4 5 0 . 0 7 0 0 . 0 ± 0 8 5 . 0 - 2 0 0 . 0 ± 6 5 1 . 0 - 5 0 0 . 0 ± 3 4 5 . 0 - 4 0 0 . 0 ± 2 0 1 . 0 - 8 3 1 . 0 ± 2 9 1 . 0 9 0 1 . 0 ± 5 4 5 . 5 7 4 0 . 0 ± 7 1 0 . 0 4 1 0 . 0 ± 5 0 0 . 0 2 2 0 . 0 ± 0 5 5 . 0 - 5 0 0 . 0 ± 9 4 1 . 0 - 6 1 0 . 0 ± 8 4 5 . 0 - 1 1 0 . 0 ± 4 4 1 . 0 - 9 0 1 . 0 ± 0 2 0 . 0 - 0 2 0 . 0 ± 9 9 7 . 5 · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 4 7 0 . 0 ± 8 7 5 . 0 2 2 0 . 0 ± 9 1 0 . 0 2 2 0 . 0 ± 0 5 7 . 0 6 4 0 . 0 ± 6 9 3 . 0 3 2 0 . 0 ± 7 7 1 . 0 7 0 0 . 0 ± 6 0 0 . 0 7 0 0 . 0 ± 0 3 2 . 0 4 1 0 . 0 ± 0 2 1 . 0 5 2 0 . 0 ± 0 2 4 . 0 - 4 1 0 . 0 ± 6 7 5 . 0 - 3 1 0 . 0 ± 4 1 6 . 0 - 7 1 0 . 0 ± 6 3 1 . 0 - 6 0 0 . 0 ± 8 1 1 . 0 - 4 0 0 . 0 ± 5 5 1 . 0 - 4 0 0 . 0 ± 5 6 1 . 0 - 5 0 0 . 0 ± 0 5 0 . 0 - · · · · · · · · · · · · · · · · · · · · · · · · 8 3 0 . 0 ± 4 0 0 . 0 2 1 0 . 0 ± 1 0 0 . 0 · · · · · · · · · · · · 3 2 0 . 0 ± 3 2 4 . 0 0 4 0 . 0 ± 6 0 5 . 0 1 1 0 . 0 ± 1 8 1 . 0 4 4 0 . 0 ± 2 4 1 . 0 0 0 0 . 0 3 4 0 . 0 ± 3 7 2 . 0 3 4 0 . 0 ± 2 6 0 . 0 3 1 1 . 0 ± 3 9 0 . 0 9 0 0 . 0 ± 4 3 1 . 0 · · · 7 0 0 . 0 ± 0 3 1 . 0 2 1 0 . 0 ± 5 5 1 . 0 3 0 0 . 0 ± 6 5 0 . 0 3 1 0 . 0 ± 3 4 0 . 0 0 0 0 . 0 · · · 3 1 0 . 0 ± 3 8 0 . 0 3 1 0 . 0 ± 9 1 0 . 0 5 3 0 . 0 ± 8 2 0 . 0 3 0 0 . 0 ± 1 4 0 . 0 1 1 0 . 0 ± 9 6 6 . 0 - 7 8 0 . 0 ± 6 3 1 . 0 - 5 0 0 . 0 ± 1 7 5 . 0 - 5 1 0 . 0 ± 1 5 4 . 0 - 2 1 0 . 0 ± 8 7 5 . 0 - 7 1 0 . 0 ± 7 7 0 . 0 - 6 1 0 . 0 ± 0 8 1 . 0 - 4 4 0 . 0 ± 7 8 1 . 0 - 8 2 0 . 0 ± 1 1 2 . 0 - 4 0 0 . 0 ± 8 1 7 . 0 - · · · · · · · · · 3 0 0 . 0 ± 0 8 1 . 0 - 1 0 0 . 0 ± 9 5 0 . 0 - 1 0 0 . 0 ± 4 5 1 . 0 - 3 0 0 . 0 ± 5 2 1 . 0 - 3 0 0 . 0 ± 6 5 1 . 0 - 5 0 0 . 0 ± 3 3 0 . 0 - · · · 4 0 0 . 0 ± 2 6 0 . 0 - 2 1 0 . 0 ± 4 6 0 . 0 - 7 0 0 . 0 ± 0 7 0 . 0 - 1 0 0 . 0 ± 3 9 1 . 0 - · · · · · · 6 1 0 . 0 ± 3 9 2 . 0 - 1 1 0 . 0 ± 3 7 5 . 0 - 0 0 0 . 0 ± 0 5 4 . 0 - 8 1 0 . 0 ± 7 5 0 . 0 5 0 0 . 0 ± 9 4 1 . 0 - 5 0 0 . 0 ± 5 6 0 . 0 · · · · · · · · · · · · 0 5 0 . 0 - · · · · · · 0 7 0 . 0 · · · · · · 7 0 2 . 0 ± 3 3 3 . 0 - 6 6 0 . 0 ± 9 0 0 . 2 - 3 2 0 . 0 ± 4 6 0 . 6 8 0 0 . 0 ± 2 1 0 . 4 · · · · · · · · · · · · 8 5 1 . 0 ± 1 9 2 . 1 · · · · · · · · · · · · 0 1 3 . 8 · · · · · · 5 0 0 . 0 ± 9 6 6 . 0 - 7 8 0 . 0 ± 6 3 1 . 0 - 1 0 0 . 0 ± 9 7 4 . 0 - 9 0 0 . 0 ± 8 7 1 . 0 - 1 0 0 . 0 ± 9 5 0 . 0 - 6 0 0 . 0 ± 4 2 0 . 0 - 0 0 0 . 0 ± 0 4 5 . 0 - 1 0 0 . 0 ± 0 0 1 . 0 - 0 4 3 . 0 - 0 3 0 . 0 4 3 0 . 0 ± 3 9 1 . 1 - 3 9 1 . 0 ± 8 6 3 . 0 0 0 4 . 0 ± 7 6 9 . 0 - 0 1 2 . 0 ± 7 4 6 . 0 3 2 1 . 0 ± 7 2 7 . 0 - 8 0 0 . 0 ± 2 7 1 . 3 3 1 0 . 0 ± 3 4 1 . 6 0 0 0 . 0 ± 0 2 1 . 6 8 0 0 . 0 ± 6 1 4 . 5 0 2 0 . 8 5 3 0 . 0 ± 4 7 1 . 0 - 3 0 0 . 0 ± 9 8 6 . 0 - 0 9 1 . 0 - 5 0 0 . 0 ± 5 4 0 . 0 - 8 2 0 . 0 ± 0 4 0 . 0 - 2 0 0 . 0 ± 2 5 1 . 0 - 0 5 0 . 0 - · · · 0 2 1 . 0 - 0 1 0 . 0 · · · 0 2 0 . 0 8 6 2 . 0 ± 8 3 4 . 1 8 9 1 . 0 ± 1 2 1 . 0 - 7 0 2 . 0 ± 0 5 3 . 0 - 5 9 0 . 0 ± 1 2 4 . 0 - · · · · · · 3 3 0 . 0 ± 7 0 6 . 6 8 0 0 . 0 ± 8 6 1 . 5 · · · · · · 0 5 6 . 7 0 7 4 . 5 3 7 . 0 ± 1 4 . 6 9 2 . 0 ± 1 1 . 6 9 3 . 0 ± 6 2 . 7 1 5 . 0 ± 6 1 . 7 5 5 . 0 ± 8 4 . 6 5 . 0 ± 7 0 . 9 8 4 . 0 ± 2 . 5 7 3 . 0 ± 2 . 7 4 3 . 0 ± 2 7 . 6 7 7 . 0 ± 7 8 . 4 4 3 . 0 ± 6 1 . 7 8 . 0 ± 6 0 . 5 6 3 . 0 ± 8 0 . 2 1 1 4 . 0 ± 9 7 . 5 8 2 . 0 ± 3 4 . 6 3 3 . 0 ± 4 4 . 6 5 3 . 0 ± 2 5 . 8 4 5 . 0 ± 2 3 . 6 5 5 . 0 ± 4 0 . 5 5 3 . 0 ± 6 8 . 6 9 4 . 0 ± 3 3 . 6 4 7 . 0 ± 1 0 . 6 9 1 . 0 ± 5 2 . 6 1 5 . 0 ± 1 7 . 4 9 2 . 0 ± 5 9 . 3 3 . 0 ± 1 5 . 5 6 7 . 0 ± 2 3 . 5 2 . 0 ± 4 . 3 1 1 6 . 0 ± 5 0 . 7 2 3 . 0 ± 8 3 . 3 4 3 . 0 ± 4 9 . 5 3 3 . 0 ± 9 6 . 7 7 . 0 ± 1 9 . 3 5 . 0 ± 8 2 . 5 4 . 0 ± 4 6 . 4 9 6 . 0 ± 8 . 5 3 7 . 0 ± 7 6 . 7 7 3 . 0 ± 7 0 . 4 3 3 . 0 ± 3 6 . 7 n n V 9 B V 5 . 9 B V 3 B 0 8 8 9 3 9 5 9 3 6 5 7 1 1 8 6 7 1 7 0 7 7 1 7 0 9 7 1 I I I 5 B 0 0 1 7 2 6 5 1 7 7 4 3 2 D H 7 4 1 1 R H 0 9 9 3 2 D H i r A τ u a T u i S p 9 B 0 0 1 3 3 0 8 1 u a T 6 6 7 V V 5 . 9 B V 8 B n V 8 B V I 3 B V 2 A 7 9 8 9 2 8 3 7 6 9 7 6 0 8 1 0 9 1 8 1 6 6 4 9 1 0 2 7 9 1 0 6 8 9 1 2 6 6 3 2 D H 6 5 4 4 2 D H 3 2 3 6 2 D H 6 7 6 6 2 D H u a T µ V 9 B 0 0 1 3 6 0 0 2 8 2 3 1 R H p s V 9 B n V 0 A V I 4 B V 5 B g H p I I I 9 B V 8 B V I 4 B I I I 5 B V 9 B 8 B V 5 . 9 B V 3 B V 5 B V 8 B V 5 B n V 5 . 9 B V 8 B V 9 B 7 9 6 9 1 7 1 0 2 9 2 2 0 2 8 2 5 7 2 D H u a T 3 5 0 0 1 4 5 3 0 2 r e P d 6 9 4 9 0 0 1 0 0 1 7 9 6 8 1 8 7 9 3 7 0 0 1 0 0 1 8 9 0 9 9 9 0 8 9 9 8 5 7 9 2 6 9 9 7 9 4 8 4 2 4 0 2 4 8 8 0 2 5 3 1 1 2 7 7 1 1 2 2 9 1 1 2 0 4 6 1 2 7 0 7 7 2 D H 5 1 4 1 R H 5 1 7 8 2 D H 6 9 7 8 2 D H 4 5 5 9 2 D H i r E Z D 3 7 9 1 2 7 8 9 6 8 2 D H 4 3 0 2 2 9 0 1 2 2 8 2 1 2 2 5 1 4 2 2 0 3 1 3 2 5 4 7 3 2 7 6 7 3 2 0 4 7 4 2 6 3 8 4 2 7 5 1 5 2 9 9 4 5 2 5 5 5 5 2 1 6 5 5 2 7 5 6 5 2 5 9 6 5 2 6 6 8 9 2 D H i r E µ 2 2 1 0 3 D H 9 0 4 0 3 D H 9 9 7 1 3 D H 4 8 8 2 3 D H r u A η r u A 7 1 m a C 5 1 4 3 0 5 3 D H u a T 5 1 1 u a T 6 1 1 5 8 7 5 3 D H 5 4 9 5 3 D H u a T 8 1 1 0 0 1 4 3 0 6 2 3 5 4 6 3 D H V I 3 B 4 9 0 4 6 6 2 u a T 5 2 1 d e u n i t n o c 5 e l b a T 32 David et al. ) d e u n i t n o c ( 5 e l b a T V A ) g a m ( ) V − B ( E 0 ) B − U ( 0 ) V − B ( ) g a m ( ) g a m ( ) g a m ( B − U ) g a m ( V − B ) g a m ( V M ) g a m ( ) g a m ( ) s a m ( ) % ( V T p S . b o r P P I H e m a N 2 4 0 . 0 ± 1 8 0 . 0 3 1 0 . 0 ± 5 2 0 . 0 2 3 0 . 0 ± 8 0 4 . 0 - 7 0 0 . 0 ± 5 1 1 . 0 - 5 2 0 . 0 ± 9 8 3 . 0 - 9 0 0 . 0 ± 1 9 0 . 0 - 3 7 2 . 0 ± 8 5 0 . 0 - 0 4 0 . 0 ± 0 1 4 . 0 7 5 0 . 0 ± 1 6 0 . 0 0 4 0 . 0 ± 2 5 1 . 0 3 3 0 . 0 ± 2 2 0 . 0 2 1 0 . 0 ± 5 2 1 . 0 8 1 0 . 0 ± 9 1 0 . 0 2 1 0 . 0 ± 7 4 0 . 0 0 1 0 . 0 ± 7 0 0 . 0 5 1 0 . 0 ± 0 5 4 . 0 - 0 3 0 . 0 ± 8 6 6 . 0 - 5 1 0 . 0 ± 3 4 5 . 0 - 0 1 0 . 0 ± 2 6 5 . 0 - 3 0 0 . 0 ± 5 2 1 . 0 - 8 0 0 . 0 ± 9 7 1 . 0 - 4 0 0 . 0 ± 7 4 1 . 0 - 3 0 0 . 0 ± 2 5 1 . 0 - 2 2 0 . 0 ± 6 5 6 . 0 - 0 1 0 . 0 ± 1 1 5 . 0 - 6 0 0 . 0 ± 9 5 5 . 0 - 0 6 3 . 0 - 3 1 0 . 0 ± 1 6 1 . 0 - 0 0 0 . 0 ± 0 0 1 . 0 - 8 0 0 . 0 ± 5 4 1 . 0 - 0 0 0 . 0 · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 7 0 3 . 0 ± 4 2 1 . 0 - 3 1 1 . 0 ± 3 8 5 . 1 - 7 0 5 . 0 ± 4 9 2 . 0 - 9 8 0 . 0 ± 0 6 0 . 1 - · · · · · · 7 0 0 . 0 ± 7 1 4 . 4 0 0 0 . 0 ± 0 4 3 . 7 4 0 0 . 0 ± 2 9 0 . 5 0 7 0 . 6 · · · · · · 0 0 8 . 6 3 7 . 0 ± 7 9 . 5 4 0 . 1 ± 2 7 . 4 4 0 . 1 ± 8 . 4 7 5 . 0 ± 8 1 . 4 3 3 . 0 ± 2 3 . 6 6 6 . 0 ± 5 0 . 3 4 2 . 0 ± 9 8 . 5 V I 5 . 9 B 0 0 1 3 2 7 7 2 4 1 1 9 3 D H n V 9 B 7 8 1 2 4 7 2 0 7 6 8 3 D H I I I 0 A V I 3 B 8 B n V 5 B 3 9 2 6 6 4 7 7 2 2 6 9 7 2 5 8 2 9 3 D H 3 7 7 9 3 D H 0 0 1 8 3 0 9 2 i r O u n 5 9 8 9 0 8 1 0 3 8 7 2 1 3 2 7 1 4 4 D H 5 9 3 2 R H r a t s l l e h S . a The Adolescent Sub-Neptune K2-284 b 33 REFERENCES Abt, H. A., Levato, H., & Grosso, M. 2002, ApJ, -- . 2017b, Forecaster: Mass and radii of planets 573, 359 Aigrain, S., Parviainen, H., & Pope, B. J. S. 2016, MNRAS, 459, 2408 predictor, Astrophysics Source Code Library, , , ascl:1701.007 Chen, Y., Girardi, L., Bressan, A., et al. 2014, Akeson, R. L., Chen, X., Ciardi, D., et al. 2013, MNRAS, 444, 2525 PASP, 125, 989 Choi, J., Dotter, A., Conroy, C., et al. 2016, ApJ, Anderson, E., & Francis, C. 2012, Astronomy 823, 102 Letters, 38, 331 Chubak, C., Marcy, G., Fischer, D. A., et al. 2012, Angus, R., Aigrain, S., Foreman-Mackey, D., & ArXiv e-prints, arXiv:1207.6212 McQuillan, A. 2015, MNRAS, 450, 1787 Ciardi, D. R., Beichman, C. A., Horch, E. P., & Avvakumova, E. A., Malkov, O. Y., & Kniazev, A. Y. 2013, Astronomische Nachrichten, 334, 860 Barnes, S. A. 2007, ApJ, 669, 1167 Barrado y Navascu´es, D., Stauffer, J. R., & Jayawardhana, R. 2004, ApJ, 614, 386 Howell, S. B. 2015, ApJ, 805, 16 Ciardi, D. R., Crossfield, I. J. M., Feinstein, A. D., et al. 2017, ArXiv e-prints, arXiv:1709.10398 Claret, A., Hauschildt, P. H., & Witte, S. 2012, A&A, 546, A14 Cochran, W. D., Hatzes, A. P., & Paulson, D. B. Bell, C. P. M., Mamajek, E. E., & Naylor, T. 2002, AJ, 124, 565 2015, MNRAS, 454, 593 Bessell, M. S. 1990, PASP, 102, 1181 Blaauw, A. 1956, ApJ, 123, 408 Bouvier, J., Barrado, D., Moraux, E., et al. 2017, ArXiv e-prints, arXiv:1712.06525 Bowler, B. P. 2016, PASP, 128, 102001 Boyajian, T. S., von Braun, K., van Belle, G., et al. 2012, ApJ, 757, 112 Brandt, T. D., & Huang, C. X. 2015, ApJ, 807, 24 Bressan, A., Marigo, P., Girardi, L., et al. 2012, MNRAS, 427, 127 Brucalassi, A., Pasquini, L., Saglia, R., et al. 2014, Cram, L. E., & Mullan, D. J. 1979, ApJ, 234, 579 Crawford, D. L. 1963, ApJ, 137, 523 -- . 1978, AJ, 83, 48 Cumming, A., Butler, R. P., Marcy, G. W., et al. 2008, PASP, 120, 531 Cummings, J. D., Deliyannis, C. P., Maderak, R. M., & Steinhauer, A. 2017, AJ, 153, 128 Cutri, R. M., & et al. 2013, VizieR Online Data Catalog, 2328 Cutri, R. M., Skrutskie, M. F., van Dyk, S., et al. 2003, 2MASS All Sky Catalog of point sources. David, T. J., & Hillenbrand, L. A. 2015, ApJ, 804, A&A, 561, L9 146 -- . 2016, A&A, 592, L1 Brucalassi, A., Koppenhoefer, J., Saglia, R., et al. 2017, A&A, 603, A85 David, T. J., Hillenbrand, L. A., Petigura, E. A., et al. 2016a, Nature, 534, 658 David, T. J., Conroy, K. E., Hillenbrand, L. A., Buchhave, L. A., Latham, D. W., Johansen, A., et al. 2016b, AJ, 151, 112 et al. 2012, Nature, 486, 375 David, T. J., Crossfield, I. J. M., Benneke, B., Buchhave, L. A., Bizzarro, M., Latham, D. W., et al. 2018, AJ, 155, 222 et al. 2014, Nature, 509, 593 de Bruijne, J. H. J., & Eilers, A.-C. 2012, A&A, Caffau, E., Ludwig, H.-G., Steffen, M., Freytag, 546, A61 B., & Bonifacio, P. 2011, SoPh, 268, 255 Cardelli, J. A., Clayton, G. C., & Mathis, J. S. 1989, ApJ, 345, 245 Castelli, F., & Kurucz, R. L. 2004, ArXiv Astrophysics e-prints, astro-ph/0405087 Chambers, K. C., Magnier, E. A., Metcalfe, N., et al. 2016, ArXiv e-prints, arXiv:1612.05560 Chen, H., & Rogers, L. A. 2016, ApJ, 831, 180 Chen, J., & Kipping, D. 2017a, ApJ, 834, 17 de Zeeuw, P. T., Hoogerwerf, R., de Bruijne, J. H. J., Brown, A. G. A., & Blaauw, A. 1999, AJ, 117, 354 de Zeeuw, T., & Brand, J. 1985, in Astrophysics and Space Science Library, Vol. 120, Birth and Evolution of Massive Stars and Stellar Groups, ed. W. Boland & H. van Woerden, 95 -- 101 Dobbie, P. D., Lodieu, N., & Sharp, R. G. 2010, MNRAS, 409, 1002 34 David et al. Donati, J. F., Moutou, C., Malo, L., et al. 2016, Nature, 534, 662 Dotter, A. 2016, ApJS, 222, 8 Douglas, S. T., Agueros, M. A., Covey, K. R., et al. 2016, ApJ, 822, 47 Findeisen, K., & Hillenbrand, L. 2010, AJ, 139, 1338 Hillenbrand, L., Isaacson, H., Marcy, G., et al. 2015, in Cambridge Workshop on Cool Stars, Stellar Systems, and the Sun, Vol. 18, 18th Cambridge Workshop on Cool Stars, Stellar Systems, and the Sun, ed. G. T. van Belle & H. C. Harris, 759 -- 766 Hirano, T., Dai, F., Gandolfi, D., et al. 2018, AJ, Findeisen, K., Hillenbrand, L., & Soderblom, D. 155, 127 2011, AJ, 142, 23 Fischer, D. A., & Valenti, J. 2005, ApJ, 622, 1102 Flewelling, H. A., Magnier, E. A., Chambers, K. C., et al. 2016, ArXiv e-prints, arXiv:1612.05243 Foreman-Mackey, D., Hogg, D. W., Lang, D., & Goodman, J. 2013, PASP, 125, 306 Fulton, B. J., & Petigura, E. A. 2018, ArXiv e-prints, arXiv:1805.01453 Fulton, B. J., Petigura, E. A., Blunt, S., & Sinukoff, E. 2018, ArXiv e-prints, arXiv:1801.01947 Fulton, B. J., Petigura, E. A., Howard, A. W., et al. 2017, AJ, 154, 109 Gagn´e, J., Mamajek, E. E., Malo, L., et al. 2018, ApJ, 856, 23 Gaia Collaboration, Brown, A. G. A., Vallenari, A., et al. 2018a, ArXiv e-prints, arXiv:1804.09365 Gaia Collaboration, Prusti, T., de Bruijne, J. H. J., et al. 2016a, A&A, 595, A1 Gaia Collaboration, Brown, A. G. A., Vallenari, A., et al. 2016b, A&A, 595, A2 Gaia Collaboration, Babusiaux, C., van Leeuwen, F., et al. 2018b, A&A, 616, A10 Gaidos, E., Mann, A. W., Rizzuto, A., et al. 2017, MNRAS, 464, 850 Gilliland, R. L., Brown, T. M., Guhathakurta, P., et al. 2000, ApJL, 545, L47 Girardi, L., Groenewegen, M. A. T., Hatziminaoglou, E., & da Costa, L. 2005, A&A, 436, 895 Gontcharov, G. A. 2006, Astronomy Letters, 32, 759 Gossage, S., Conroy, C., Dotter, A., et al. 2018, ApJ, 863, 67 Gudel, M. 2004, A&A Rv, 12, 71 Hartmann, L., Stauffer, J. R., Kenyon, S. J., & Jones, B. F. 1991, AJ, 101, 1050 Henden, A. A., Templeton, M., Terrell, D., et al. 2016, VizieR Online Data Catalog, 2336 Howard, A. W., Marcy, G. W., Bryson, S. T., et al. 2012, ApJS, 201, 15 Howell, S. B., Rowe, J. F., Bryson, S. T., et al. 2012, ApJ, 746, 123 Howell, S. B., Sobeck, C., Haas, M., et al. 2014, PASP, 126, 398 Huber, D., Bryson, S. T., Haas, M. R., et al. 2016, ApJS, 224, 2 Huber, D., Zinn, J., Bojsen-Hansen, M., et al. 2017, ApJ, 844, 102 Jackson, A. P., Davis, T. A., & Wheatley, P. J. 2012, MNRAS, 422, 2024 Janes, K., & Kim, J.-H. 2009, in IAU Symposium, Vol. 253, Transiting Planets, ed. F. Pont, D. Sasselov, & M. J. Holman, 548 -- 551 Johns-Krull, C. M., McLane, J. N., Prato, L., et al. 2016, ApJ, 826, 206 Kamai, B. L., Vrba, F. J., Stauffer, J. R., & Stassun, K. G. 2014, AJ, 148, 30 Kipping, D. M. 2010, MNRAS, 407, 301 Kirk, B., Conroy, K., Prsa, A., et al. 2016, AJ, 151, 68 Kolbl, R., Marcy, G. W., Isaacson, H., & Howard, A. W. 2015, AJ, 149, 18 Kov´acs, G., Zucker, S., & Mazeh, T. 2002, A&A, 391, 369 Kraus, A. L., Shkolnik, E. L., Allers, K. N., & Liu, M. C. 2014, AJ, 147, 146 Kurucz, R. L. 1992, in IAU Symposium, Vol. 149, The Stellar Populations of Galaxies, ed. B. Barbuy & A. Renzini, 225 Libralato, M., Nardiello, D., Bedin, L. R., et al. 2016, MNRAS, 463, 1780 Livingston, J. H., Dai, F., Hirano, T., et al. 2018a, AJ, 155, 115 -- . 2018b, ArXiv e-prints, arXiv:1809.01968 Lomb, N. R. 1976, Ap&SS, 39, 447 Lopez, E. D., & Fortney, J. J. 2013, ApJ, 776, 2 Lopez, E. D., & Rice, K. 2016, ArXiv e-prints, arXiv:1610.09390 L´opez-Morales, M., & Ribas, I. 2005, ApJ, 631, 1120 The Adolescent Sub-Neptune K2-284 b 35 Luhman, K. L., Mamajek, E. E., Allen, P. R., & Petigura, E. A., Howard, A. W., Marcy, G. W., Cruz, K. L. 2009, ApJ, 703, 399 et al. 2017, AJ, 154, 107 Lundkvist, M. S., Kjeldsen, H., Albrecht, S., et al. Petigura, E. A., Marcy, G. W., Winn, J. N., et al. 2016, Nature Communications, 7, 11201 Ma´ız Apell´aniz, J. 2006, AJ, 131, 1184 Mamajek, E. E. 2005, ApJ, 634, 1385 Mamajek, E. E., & Bell, C. P. M. 2014, MNRAS, 445, 2169 Mamajek, E. E., & Hillenbrand, L. A. 2008, ApJ, 687, 1264 Mandel, K., & Agol, E. 2002, ApJL, 580, L171 Mann, A. W., Feiden, G. A., Gaidos, E., Boyajian, T., & von Braun, K. 2015, ApJ, 804, 64 Mann, A. W., Gaidos, E., Mace, G. N., et al. 2016, ApJ, 818, 46 Mann, A. W., Gaidos, E., Vanderburg, A., et al. 2017a, AJ, 153, 64 Mann, A. W., Vanderburg, A., Rizzuto, A. C., et al. 2017b, ArXiv e-prints, arXiv:1709.10328 Martin, D. C., Fanson, J., Schiminovich, D., et al. 2005, ApJL, 619, L1 Masuda, K., & Winn, J. N. 2017, AJ, 153, 187 Meibom, S., Torres, G., Fressin, F., et al. 2013, Nature, 499, 55 Mentuch, E., Brandeker, A., van Kerkwijk, M. H., Jayawardhana, R., & Hauschildt, P. H. 2008, ApJ, 689, 1127 Mermilliod, J. C. 1981, A&A, 97, 235 -- . 2006, VizieR Online Data Catalog, 2168 Mermilliod, J.-C., Weis, E. W., Duquennoy, A., & Mayor, M. 1990, A&A, 235, 114 Meynet, G., Mermilliod, J.-C., & Maeder, A. 1993, A&AS, 98, 477 Morton, T. D. 2015, VESPA: False positive probabilities calculator, Astrophysics Source Code Library, , , ascl:1503.011 Obermeier, C., Henning, T., Schlieder, J. E., et al. 2016, AJ, 152, 223 Owen, J. E., & Wu, Y. 2013, ApJ, 775, 105 Pace, G., Pasquini, L., & Fran¸cois, P. 2008, A&A, 489, 403 Parviainen, H. 2015, MNRAS, 450, 3233 Paunzen, E. 2015, A&A, 580, A23 Pecaut, M. J., & Mamajek, E. E. 2013, ApJS, 208, 9 -- . 2016, MNRAS, 461, 794 Pepper, J., Gillen, E., Parviainen, H., et al. 2017, AJ, 153, 177 2018, AJ, 155, 89 Petrie, R. M. 1958, MNRAS, 118, 80 Plavchan, P., Latham, D., Gaudi, S., et al. 2015, ArXiv e-prints, arXiv:1503.01770 Prosser, C. F. 1992, AJ, 103, 488 Quinn, S. N., White, R. J., Latham, D. W., et al. 2012, ApJL, 756, L33 -- . 2014, ApJ, 787, 27 Randich, S., Pallavicini, R., Meola, G., Stauffer, J. R., & Balachandran, S. C. 2001, A&A, 372, 862 Rasmuson, N. H. 1921, Meddelanden fran Lunds Astronomiska Observatorium Serie II, 26, 3 Rebull, L. M., Stauffer, J. R., Hillenbrand, L. A., et al. 2017, ApJ, 839, 92 Rebull, L. M., Stauffer, J. R., Bouvier, J., et al. 2016, AJ, 152, 113 Rizzuto, A. C., Mann, A. W., Vanderburg, A., Kraus, A. L., & Covey, K. R. 2017, AJ, 154, 224 Rizzuto, A. C., Vanderburg, A., Mann, A. W., et al. 2018, ArXiv e-prints, arXiv:1808.07068 Rodriguez, D. R., Bessell, M. S., Zuckerman, B., & Kastner, J. H. 2011, ApJ, 727, 62 Rodriguez, D. R., Zuckerman, B., Kastner, J. H., et al. 2013, ApJ, 774, 101 Scargle, J. D. 1982, ApJ, 263, 835 Seager, S., & Mall´en-Ornelas, G. 2003, ApJ, 585, 1038 Shkolnik, E. L., Liu, M. C., Reid, I. N., Dupuy, T., & Weinberger, A. J. 2011, ApJ, 727, 6 Shobbrook, R. R. 1983, MNRAS, 205, 1215 Skiff, B. A. 2014, VizieR Online Data Catalog, 1 Soderblom, D. R., Hillenbrand, L. A., Jeffries, R. D., Mamajek, E. E., & Naylor, T. 2014, Protostars and Planets VI, 219 Soderblom, D. R., Jones, B. F., Balachandran, S., et al. 1993, AJ, 106, 1059 Stauffer, J. 1982, PASP, 94, 678 Stauffer, J., Rebull, L., Bouvier, J., et al. 2016, AJ, 152, 115 Stauffer, J. R., & Hartmann, L. W. 1987, ApJ, 318, 337 Stauffer, J. R., Hartmann, L. W., & Jones, B. F. 1989, ApJ, 346, 160 Perryman, M. A. C., Brown, A. G. A., Lebreton, Stauffer, J. R., Barrado y Navascu´es, D., Bouvier, Y., et al. 1998, A&A, 331, 81 J., et al. 1999, ApJ, 527, 219 36 David et al. Tu, L., Johnstone, C. P., Gudel, M., & Lammer, Vogt, S. S., Allen, S. L., Bigelow, B. C., et al. H. 2015, A&A, 577, L3 Upgren, A. R., Weis, E. W., & Deluca, E. E. 1979, AJ, 84, 1586 van Leeuwen, F. 2007, A&A, 474, 653 van Saders, J. L., & Gaudi, B. S. 2011, ApJ, 729, 63 Vanderburg, A., & Johnson, J. A. 2014, PASP, 126, 948 Vanderburg, A., Latham, D. W., Buchhave, L. A., et al. 2016, ApJS, 222, 14 Vanderburg, A., Mann, A. W., Rizzuto, A., et al. 2018, AJ, 156, 46 1994, in Proc. SPIE, Vol. 2198, Instrumentation in Astronomy VIII, ed. D. L. Crawford & E. R. Craine, 362 Walter, F. M., & Boyd, W. T. 1991, ApJ, 370, 318 Walter, F. M., Brown, A., Mathieu, R. D., Myers, P. C., & Vrba, F. J. 1988, AJ, 96, 297 Weis, E. W. 1981, PASP, 93, 437 Weldrake, D. T. F., Sackett, P. D., & Bridges, T. J. 2008, ApJ, 674, 1117 Weldrake, D. T. F., Sackett, P. D., Bridges, T. J., & Freeman, K. C. 2005, ApJ, 620, 1043 Yee, S. W., Petigura, E. A., & von Braun, K. 2017, ApJ, 836, 77 Yu, L., Donati, J.-F., H´ebrard, E. M., et al. 2017, MNRAS, 467, 1342 Vazan, A., Ormel, C. W., & Dominik, C. 2017, Yuan, H. B., Liu, X. W., & Xiang, M. S. 2013, ArXiv e-prints, arXiv:1712.06158 Ventura, P., Zeppieri, A., Mazzitelli, I., & D'Antona, F. 1998, A&A, 334, 953 MNRAS, 430, 2188 Zacharias, N., Finch, C., & Frouard, J. 2017, AJ, 153, 166
1711.00647	1	1711	2017-11-02T08:26:13	Dome C UltraCarbonaceous Antarctic MicroMeteorites Infrared and Raman fingerprints	[ "astro-ph.EP" ]	UltraCarbonaceous Antarctic MicroMeteorites (UCAMMs) represent a small fraction of interplanetary dust particles reaching the Earth's surface and contain large amounts of an organic component not found elsewhere. They are most probably sampling a contribution from the outer regions of the solar system to the local interplanetary dust particle flux. We characterize UCAMMs composition focusing on the organic matter, and compare the results to the insoluble organic matter (IOM) from primitive meteorites, IDPs, and the Earth.We acquired synchrotron infrared microspectroscopy and micro-Raman spectra of eight UCAMMs from the Concordia/CSNSM collection, as well as N/C atomic ratios determined with an electron microprobe. The spectra are dominated by an organic component with a low aliphatic CH versus aromatic C=C ratio, and a higher nitrogen fraction and lower oxygen fraction compared to carbonaceous chondrites and IDPs. The UCAMMs carbonyl absorption band is in agreement with a ketone or aldehyde functional group. Some of the IR and Raman spectra show a C$\equiv$N band corresponding to a nitrile. The absorption band profile from 1400 to 1100 cm-1 is compatible with the presence of C-N bondings in the carbonaceous network, and is spectrally different from that reported in meteorite IOM. We confirm that the silicate-to-carbon content in UCAMMs is well below that reported in IDPs and meteorites. Together with the high nitrogen abundance relative to carbon building the organic matter matrix, the most likely scenario for the formation of UCAMMs occurs via physicochemical mechanisms taking place in a cold nitrogen rich environment, like the surface of icy parent bodies in the outer solar system. The composition of UCAMMs provides an additional hint of the presence of a heliocentric positive gradient in the C/Si and N/C abundance ratios in the solar system protoplanetary disc evolution.	astro-ph.EP	astro-ph	F-91405 Orsay, France Received January 10, 2017 ABSTRACT Infrared and Raman fingerprints 3 Synchrotron SOLEIL, L'Orme des Merisiers, BP48 Saint Aubin, 91192 Gif-sur-Yvette Cedex, France E. Dartois1,(cid:63) (cid:63)(cid:63), C. Engrand2, J. Duprat2, M. Godard2, E. Charon2, L. Delauche2, C. Sandt3, and F. Borondics3 Dome C UltraCarbonaceous Antarctic MicroMeteorites 1 Institut d'Astrophysique Spatiale (IAS), CNRS, Univ. Paris Sud, Université Paris-Saclay, F-91405 Orsay, France 2 Centre de Sciences Nucléaires et de Sciences de la Matière (CSNSM), CNRS/IN2P3, Univ. Paris Sud, Université Paris-Saclay, Context. UltraCarbonaceous Antarctic MicroMeteorites (UCAMMs) represent a small fraction of interplanetary dust particles reach- ing the Earth's surface and contain large amounts of an organic component not found elsewhere. They are most probably sampling a contribution from the outer regions of the solar system to the local interplanetary dust particle (IDP) flux. Aims. We characterize UCAMMs composition focusing on the organic matter, and compare the results to the insoluble organic matter (IOM) from primitive meteorites, IDPs, and the Earth. Methods. We acquired synchrotron infrared microspectroscopy (µFTIR) and µRaman spectra of eight UCAMMs from the Concor- dia/CSNSM collection, as well as N/C atomic ratios determined with an electron microprobe. Results. The spectra are dominated by an organic component with a low aliphatic CH versus aromatic C=C ratio, and a higher nitrogen fraction and lower oxygen fraction compared to carbonaceous chondrites and IDPs. The UCAMMs carbonyl absorption band is in agreement with a ketone or aldehyde functional group. Some of the IR and Raman spectra show a C≡N band corresponding to a nitrile. The absorption band profile from 1400 to 1100 cm−1 is compatible with the presence of C-N bondings in the carbonaceous network, and is spectrally different from that reported in meteorite IOM. We confirm that the silicate-to-carbon content in UCAMMs is well below that reported in IDPs and meteorites. Together with the high nitrogen abundance relative to carbon building the organic matter matrix, the most likely scenario for the formation of UCAMMs occurs via physicochemical mechanisms taking place in a cold nitrogen rich environment, like the surface of icy parent bodies in the outer solar system. The composition of UCAMMs provides an additional hint of the presence of a heliocentric positive gradient in the C/Si and N/C abundance ratios in the solar system protoplanetary disc evolution. Key words. Meteorites, Interplanetary medium, Protoplanetary disks, Abundances, Methods: laboratory: solid state Acceptedin A & A 1. Introduction Micrometeorites represent the largest mass flux of extraterres- trial material falling on Earth (e.g. Love & Brownlee 1993; Tay- lor et al. 1996; Duprat et al. 2006; Zolensky et al. 2006; Prasad et al. 2013; Engrand et al. 2017). They sample the interplane- tary dust present in the inner solar system, a large fraction com- ing from Jupiter-family comets together with a contribution at- tributed to asteroids, and Kuiper Belt and Oort Cloud comets (e.g. Janches et al. 2006; Nesvorný et al. 2010, 2011; Poppe et al. 2011; Poppe 2016). The search for and collection of microm- eteorites performed on the Antarctic continent has allowed the recovery of particles preserved from significant atmospheric en- try heating and which still hold information from interplanetary dust. This information otherwise could only be obtained via sam- pling by space probes. Several countries are involved in these collecting expeditions (e.g. Nakamura et al. 2005; Duprat et al. 2007; Taylor et al. 2008; van Ginneken et al. 2012; Yabuta et al. 2012, 2017). The Concordia/CSNSM collection contains thou- sands of Antarctic micrometeorites collected near the French- Italian Concordia station on Dome C during several campaigns performed since 1999 with the help and support of the French Polar Institute Paul-Émile Victor (IPEV). Details of the collect- ing methods can be found in Duprat et al. (2007). A particu- larly interesting subset in the collections is the fraction of mi- crometeorites arising from the outer parts of the solar system, known as Kuiper Belt and/or Oort Cloud comet contributions, which are largely undersampled by space missions. These dust particles contain the record of past formation and evolution in the outer solar system. UltraCarbonaceous Antarctic MicroMe- teorites (UCAMMs, Duprat et al. 2010) and a subset of chon- dritic porous interplanetary dust particles (IDPs) are most prob- ably of this class (e.g. Dobrica et al. 2010). The UCAMMs are exceptionally organic-rich, in particular containing organic matter that is significantly larger than a mi- cron in size (Dartois et al. 2013; Bardin et al. 2014; Engrand et al. 2015; Charon et al. 2017) that is not found in other types of extraterrestrial matter. The organic content of UCAMMs is high enough to avoid the requirement of prior chemical treatments to be studied in the infrared, and can be compared with the in- soluble organic matter (IOM) extracted from meteorites for an insight into their similarities and differences. In this work, we report a µFTIR and µRaman study of UCAMM samples, and discuss the implications for their origin and significance. (cid:63) send offprint requests to [email protected] (cid:63)(cid:63) Part of the equipment used in this work has been financed by the French INSU-CNRS program "Physique et Chimie du Milieu Interstel- laire" (PCMI). 7 1 0 2 v o N 2 . ] P E h p - o r t s a [ 1 v 7 4 6 0 0 . 1 1 7 1 : v i X r a A&A proofs: manuscript no. ucamms_edartois_arXiv Fig. 1. Backscattered electron images recorded at 15kV of fragments of the UCAMMs samples analysed in this study: (a) DC060919, (b) DC060594, (c) DC021119, (d) DC060443, (e) DC060718, (f) DC060741, (g) DC0609119, (h) DC060565. Scale bars correspond to 5µm for each image. The unfragmented UCAMMS initial sizes are respectively: (a) 79 x 50 µm, (b) 66 x 53 µm, (c) 110 x 87 µm (unfragmented), (d) 28 x 24 µm, (e) 87 x 53 µm, (f) 200 x 80 µm, (g) 275 x 108 µm, (h) 44 x 33 µm. Acceptedin A & A We analysed eight UCAMMs from the Concordia/CSNSM col- lection. Fragments of each UCAMM was micro-manipulated and transferred into a diamond compression cell. The fragments were then flattened to provide an optimal thickness for infrared spectroscopic transmission measurements. The infrared mea- surements were performed on the SMIS beam line at the syn- chrotron SOLEIL during several runs in the period from 2010 to 2016. The synchrotron beam was coupled to a Nicolet Nic- Plan infrared microscope. For the measurements presented in this analysis, the infrared spot size was optimized close to the diffraction limit, with 5x5µm to 15x15µm sampling windows, adapted to the geometry of each UCAMM fragment. A Thermo Fisher DXR spectrometer was used to record a Raman spectrum for each of the samples, using a laser source at 532 nm at min- imum power (100 µW on sample) to prevent sample alteration. Ten scans with integration times of 20s each were co-added for each spectrum. The spot size of the Raman spectra is lower than that of the IR spectra, of the order of a micron. For five of the UCAMMS, light element analysis (C, N, O) was conducted at Université Paris 6 Jussieu using a Cameca SX100 or SXFive electron microprobe at 10 keV and 40 nA or 20 nA. Before analysis, the samples and standards were coated with a 9 nm layer of gold using a sputter-coating technique equipped with a quartz thickness monitoring. Care was taken to gold-coat the sample and the standard mounts at the same time whenever possible. The C, N, and O Kα emission lines were recorded us- ing a PC1 crystal (W-Si multilayer crystal), and were calibrated using graphite for C, BN for N, and Fe2O3 for O. As UCAMMs contain variable proportions of minerals intimately mixed with the organic matter; the Na, Mg, Al, Si, P, S, K, Ca, Ti, Cr, Mn, and Fe Kα lines were recorded as well, using adequate analysing crystals (LTAP, LPET, or LLIF) and calibrated on usual mineral standards (albite for Na; Olivine for Si and Mg; orthoclase for Al and K; apatite for P and Ca; pyrophanite (MnTiO3) for Mn and Ti; chromite (Cr2O3) for Cr; Fe2O3 for Fe). Totals different from 2. Experiments 100% can be due to porosity and matrix effects slightly differ- ing between the standard and crushed samples or to a high con- centration of an element not measured in the standard (e.g. Ni). Since these analyses were performed on crushed grains, only analyses with reasonable totals in wt% (85% < totals < 115%) were selected to calculate the N/ C ratios. We note that the ra- tios between C, N, and O measured in a given sample are robust even in the cases of very anomalous totals as they are measured on the same detector. The typical statistical uncertainties for the determination of C, N, and O in the UCAMMs are on average 4 wt%, 1 wt%, and 0.5 wt%, respectively. The external uncertainty (reproducibility) on the C and N standards are on average 4 wt% and 2.5 wt%. The intrinsic variability of a UCAMM sample N/C ratio is usually larger than the analytical uncertainty. 3. Analysis The measured UCAMM infrared and Raman spectra are dis- played in Fig.2. The IR spectra are baseline corrected, in a sim- ilar way to previously reported in Dartois et al. (2013), and de- composed into individual contributions corresponding to spe- cific vibrational modes. We also indicate the global fingerprint absorption band in the 1500-800 cm−1 spectral region, together with the silicates stretching complex contribution. We represent by different colours the various contributions to the spectra in Fig. 2. Except for the ∼1100-900 cm−1 spectral region, where silicate absorptions prevail if present, the spectra are dominated by the organic matter contribution. The broad band extending from about 3600 to 3000 cm−1 includes contributions from OH and NH stretching modes. The asymmetric profile of the band in this range, by comparison with laboratory analogues (Dartois et al. 2013), indicates a major contribution of the NH stretching mode. We measured the amount of nitrogen in DC060594 and DC060565 independently (N/C ratios of 0.05 and 0.12, locally exceeding 0.15; Dartois et al. 2013). The intrinsic intensity of the OH mode is about ten times higher than the NH stretching mode. It is probable that the broad and structureless feature be- a!c!d!b!e!f!g!h!: Acceptedin A & A tween 3300 and 3500 cm−1 in DC060919 contains an OH com- ponent. A few spectra possess a series of CH stretching mode absorp- tions, shown in yellow in the left panel of Fig. 2, with a close- up shown in Fig. 3, left panel. The main bands are due to aliphatic CH stretching modes (∼2960 cm−1, asymmetric CH3; ∼2920 cm−1, asymmetric CH2; ∼2870-2850 cm−1, symmetric CH3 and CH2). A small contribution at ∼3050 cm−1 from the aromatic CH stretching mode is observed for the first time in UCAMMs. We also observed a potential small contribution of aldehyde CH at low wavenumber (2820 cm−1) in DC060919. Three out of the eight UCAMM fragment analyses show an in- frared C≡N absorption feature at 2200 cm−1, which have also been measured with the Raman spectrometer for five UCAMMs (Fig.3, centre panel) and previously observed by Dobrica et al. (2011) for DC060919. The main vibrational bands associated with the observed IR absorption and Raman emission bands are summarized in Table 1. Fig. 2. Left: UCAMM samples µFTIR optical depth spectra. The spectra have been vertically shifted for clarity (left horizontal dashed lines indicate the shift level). They are ordered from top to bottom by decreasing amount of silicate band absorption. The main band contributions are labelled above the upper spectrum. The spectra are deconvolved into several contributions shown with distinct colours, used in the analysis. A zoom on the nitrile region (∼2200 cm−1) is provided when a band was measured. Right: UCAMM samples Raman spectra, normalized to the G- band maximum. The spectra are analysed using a classical Raman bands fitting procedure (Sadezky et al. 2005; Kouketsu et al. 2014) contributing to the D (green) and G (red) bands. The D/G band peak intensity ratio is given on the right, as defined in Busemann et al. (2007). A zoom on the nitrile region (∼2200 cm−1) is provided when a band was observed. Spectra are vertically shifted for clarity. frared spectrum is therefore affected, mainly in the carbonyl spectral region, and the carbonyl absorption for this UCAMM will thus not be considered in the following analysis. The baseline corrected silicate optical depth contribution is shown in the right panel of Fig.3, displayed together with several model absorption spectra of olivine and pyroxene magnesium- rich end members (forsterite and enstatite). The models were calculated with the continuous distribution of ellipsoids (CDE), spheroids (CDS), and hollow spheres (HS) to take into account the variabilities expected due to shape effects in scattering and absorption by randomly oriented particles that are small com- pared to the wavelength (Min et al. 2003). This statistical ap- proach allows us to compare the amplitude of deformation in- duced by shapes in the spectra recorded and to infer the sili- cates composition (one pyroxene and two olivine bands are indi- cated by vertical dotted lines). It shows notably that the pyroxene band complex still absorbs significantly at higher wavenumbers (above about 1070-1080 cm−1) than the olivine band. UCAMMs contain a mixture of silicates on a much smaller scale than that probed by classical IR wavelength microscopy. DC060919 and DC060594 display typical forsterite-like absorption bands, Unlike the other fragment, the DC021119 fragment was re- trieved from a previous preparation where it was stuck in a carbon-tape for SEM analysis. The DC021119 UCAMM in- 4000350030002500200015001000 Wavenumber (cm-1)0.0 1.0 2.502.863.334.005.006.6710.00Wavelength (µm)Optical depthDC 06 09 19DC 06 05 94DC 06 04 43X5DC 02 11 19DC 06 07 18X10DC 06 07 41DC 06 09 119X5DC 06 05 65-C=N-HN-H/O-HC-HC(cid:62)NC=OArom.C=C/C=NC-N/C-CForsteriteEnstatite350030002500200015001000 Wavenumber (cm-1) Normalized Raman Intensity2.863.334.005.006.6710.00Wavelength (µm)01DC 06 09 19ID/IG1.078DC 06 05 940.828DC 06 04 430.818DC 02 11 19x100.995DC 06 07 41x101.216DC 06 07 18x101.393DC 06 09 119x100.869DC 06 05 65x101.090D2GDD3D4A&A proofs: manuscript no. ucamms_edartois_arXiv Table 1. Infrared and Raman band list Fig. 3. Measured UCAMMs spectra. Left: CH stretching mode infrared optical depth. Centre panel: µFTIR (red) and Raman (black) close-up spectra in the nitriles region, normalized to the C=C (IR) and G band (Raman) measured intensities. Right: Baseline corrected silicate infrared optical depth, along with different absorption models for several size distributions of olivine (forsterite) and pyroxene (enstatite) magnesium rich end members (see text for details). Acceptedin A & A whereas DC021119 absorbs significantly above 1070 cm−1, in- dicating an enstatite-like contribution. To provide a systematic comparison with the organic matter ex- tracted from meteorite IOM, the published infrared spectra from Kebukawa et al. (2011); Orthous-Daunay et al. (2013); Quirico et al. (2014) were analysed in the exact same way and are dis- cussed below. We also performed infrared measurements at the synchrotron Soleil SMIS beam line on IOM extracted from the Paris meteorite, kindly provided by V. Vinogradoff and L. Re- musat (IMPMC/ MNHN)1. The analysis of this spectra was per- formed in the same way and added to the spectra displayed in Fig. 4. It is difficult to obtain IR spectra covering the full 4000- 600 cm−1 range for the insoluble organic matter fraction of IDPs. A few exceptions with Hydrofluoric acid (HF) attack of IDPs to eliminate the inorganic fraction, followed by IR measurements, were recorded by e.g. Matrajt et al. (2005), and are added for comparison to the analysis presented here. The UCAMMs Raman spectra are analysed using a classical Ra- man bands fitting procedure decomposition contributing to the D and G bands, consisting of the deconvolution of five sub-bands (e.g. Sadezky et al. 2005; Kouketsu et al. 2014) grouped into two main contributions ('D'=D1+D2+D4 and 'G'=D3+G), shown in the right panel of Fig. 2. 4. Discussion The UCAMMs Raman spectra can be compared to previous measurements for meteoritic IOM (Busemann et al. 2007), Star- dust (Rotundi et al. 2008), and previous UCAMM measure- ments (Dobrica et al. 2011). The broad and relatively intense Raman D band and intensity ratio of the D and G bands un- derscore the disordered nature of the carbonaceous network for UCAMMs organic matter. In the D-band full width at half max- imum (FWHM) versus D-band position diagram and D-FWHM versus G-FWHM shown in Fig. 4, UCAMMs lie in the region as- sociated with the most primitive meteorites and Stardust grains. Busemann et al. (2007) found and discussed a potential correla- tion of the atomic N/C element abundance ratios with the D band FWHM (Fig. 8, Busemann et al. 2007) and a correlation of the atomic H/C element abundance ratios with the G band FWHM (Fig. 9, Busemann et al. 2007). The UCAMMs measured val- ues are clearly not in line with what would be extrapolated from these correlations, showing that UCAMMs, with their high N/C and low H/C abundances ratios, evolve following another trend. The integrated optical depth of the carbonyl divided by the C=C value is reported in Fig. 5 as a function of the carbonyl ab- IR Position (cm−1) 3600-3000 OH/NH stretch (br) CH aro. stretch 3050 CH3 asym. stretch 2960 CH2 asym. stretch 2920 CH3 sym. stretch 2870 CH2 sym. stretch 2860 C≡N 2220 1750-1650 C=O 1600-1580 C=C / C=N 1475 1400-1100 C-N/C-C (br) 1375 1370-1340 1100-800 CH2,3 deformation CH2,3 deformation 1 The full spectrum will be published by these authors. SiO stretch (Silicates, br) C sp2 'Defect' band Mode Raman C≡N C sp2 br: broad 320031003000290028002700Wavenumber (cm-1)0.000.050.100.15Optical depthWavelength (µm)3.123.233.333.453.573.70 DC 06 09 19DC 06 05 94DC 06 04 43DC 02 11 19DC 06 07 18DC 06 07 41DC 06 09 119DC 06 05 65CHaroCH3CH2CH3CH224002300220021002000Wavenumber (cm-1)0.000.050.100.150.200.25Normalized IR/Raman IntensityWavelength (µm)4.174.354.554.765.00 DC 06 09 119DC 06 07 41DC 06 07 18DC 06 05 65DC 02 11 19Nitrile Poly. aromaticsUnsaturated nitrilesAryl nitriles>N-CH=C-C(cid:62)NSat. RingRCO-C(cid:62)NIso-nitrilesNitriles120011001000900800Wavenumber (cm-1)0.00.20.40.60.81.01.2Optical depthDC 02 11 19Wavelength (µm)8.339.0910.0011.1112.50 DC 06 05 65DC 06 09 119DC 06 07 41DC 06 07 18DC 06 04 43DC 06 05 94DC 06 09 19ForsteriteHSCDECDSEnstatiteHSCDECDS: Fig. 4. Raman D band full width at half maximum (FWHM) versus position (left) and G band FWHM (right) diagrams, combining meteoritic IOM (C2,CI,CM: circles, CR: diamonds, triangles: CO,CV & EH,LL: half circles Busemann et al. 2007), Stardust (squares, Rotundi et al. 2008), previous UCAMM measurements (green stars, Dobrica et al. 2011), and UCAMMs from this work (red stars). The fitted trend is associated with the organic matter primitivity; the most primitive meteorites have D band width ∼> 250 cm−1. Acceptedin A & A haviour of neighbouring atoms has an influence on the bond and thus can alter its position in the spectrum, it is not possible to infer an exact attribution of the carbonyl by comparison with the expected classical positions. However, as they appear in the 1680-1705 cm−1 range, it seems to favour an aldehyde or ketone carbonyl function over carboxylic acids and ester. They fall close to the positions of the primitive classes A and B from the spectra from Kebukawa et al. (2011), Orthous-Daunay et al. (2013), and Quirico et al. (2014). The integrated optical depth of the carbonyl absorption band di- vided by the C=C value is reported in Fig. 6 (left panel) as a function of the integrated optical depth of the CH stretching mode to C=C ratio. We did not include DC021119 in this plot because the tape contamination induces a spectral overlapping contamination in the C=O region. The UCAMMs position in this diagram is specific as they clearly lie in the lower left corner with respect to the IOM from meteorites and IDPs. The low ratio of C=O to C=C of the UCAMMs confirms that the O/C ratio in the organic matter of UCAMMs is lower than that of CR and CM meteorites and IDPs. The low CH to C=C ratio Fig. 6 (right panel) also indicates a low hydrogen content. In this diagram, UCAMMs for which atomic H/C ratio were not available are set at a zero H/C value. How- ever, given the general trend observed with the measured infrared CH/C=C band ratio, these UCAMMs most probably are below H/C(cid:46)0.5. This value is further confirmed for the two UCAMMs for which a H/C ratio was measured with the NanoSIMS (Duprat et al. 2010). The H/C in UCAMMs is substantially lower than those measured for meteorites from the infrared spectra or inde- pendently by other means by Alexander et al. (2007). The physicochemical composition modifications undergone by meteorites, IDPs, and micrometeorites during their atmospheric entry is a long-standing question (e.g. Greshake et al. 1998; Top- Fig. 5. Comparison of UCAMM infrared measurements for car- bonyl, displayed with the corresponding infrared features in mete- orites. Carbonyl-to-aromatic-carbon ratio as a function of the position of carbonyl stretching vibrations, and corresponding chemical functional groups. sorption central position. A chemical function information arises from the position of the carbonyl. Because the electrophile be- 133013401350136013701380D band position (cm-1)50100150200250300350400D band FWHM (cm-1)UCAMMs(This work)81P/Wild 2UCAMMsC2,CI,CMMeteoritesCRCO, CVEH, LL406080100120140160180G band FWHM (cm-1)100200300400D band FWHM (cm-1)17401720170016801660C=O position (cm-1)0.00.20.40.60.81.0Integrated Area C=O/C=COrthous-DaunayVinogradoffQuiricoKebukawa(A)Kebukawa(B)Kebukawa(C)MatrajtUCAMMsCM2C2-ungCI1CM2CI1CR2CM2LL3CI1CM2CM2CM2CM2CM2CM2CR2CR2CR2CM2CM2CR1CI1LL3L3.05C2-ungCO3.0CV3LL3CV3CV3LL3C3-ungCV3CM-heatedIDPIDPIDPIDPAldehydeKetoneEsterCarboxylicAmideA&A proofs: manuscript no. ucamms_edartois_arXiv Fig. 6. Comparison of UCAMM infrared measurements for carbonyl, aromatic carbon, and hydrogen bonded to carbon, displayed with the corresponding infrared features in meteorite (Orthous-Daunay et al. 2013; Quirico et al. 2014; Kebukawa et al. 2011, separating the first three infrared subclasses defined in the Kebukawa spectral analysis) and IDP (Matrajt et al. 2005) insoluble organic matter. Left: Carbonyl-to-aromatic- carbon ratio as a function of the CH to C=C areas Right: Integrated optical depth area of the CH versus C=C bands as a function of H/C ion probe (NanoSIMS) independent measurements. The UCAMMs without available H/C measurements are set at a zero H/C value. Labels correspond to meteorite groups, as reported in these articles Acceptedin A & A pani et al. 2001; Matrajt et al. 2005; Suttle et al. 2017). In the Concordia micrometeorite collection, a large fraction (∼35%) of the particles exhibit textures suggesting that they were not substantially altered during atmospheric entry (e.g. Dobrica et al. 2010). Several indicators point toward a moderate heating for most UCAMMs. Chemical and structural modifications ex- pected on minerals during thermal processing give a strong up- per limit on the temperature suffered by the UCAMMs. First of all, there is no significant magnetite conversion from pyrrhotite in UCAMMs, estimated to occur in the 500-900oC range (e.g. Rietmeijer 2004; Craig 1974). The annealing of tracks in the minerals occurs at a temperature below 600oC. Some of the min- erals in UCAMMs display these tracks (Charon et al. 2017). The high organic content of UCAMMs is also an indicator of at most moderate heating. The UCAMM organic matter Raman G and D band FWHMs are large; for some UCAMMs they lie above 140 cm−1 for the G band, the largest among extraterrestrial or- ganic matter, which points toward a low atmospheric entry heat- ing by comparison to experimental simulations (e.g. Bonnet et al. 2015). The high amount of nitrogen in UCAMMs organic matter and the presence of NH stretching modes confirm the low heating; above about 500oC these NH bonds should be re- duced drastically, and this implies that the N/C would otherwise be even higher before entry. The greatest unknown is probably the initial C-H content in UCAMMs, which may be reduced dur- ing atmospheric entry. IDPs often preserve higher aliphatic C- H content in their organic matter than measured for UCAMMs. Because IDPs are usually smaller in size, they may have suffered less heating during the atmospheric entry (e.g. Love & Brownlee 1991). Undoubtedly ices, including nitrogen ices, are lost long before the atmospheric entry, in the evolving parent body, and during the journey of the dust grains toward the inner solar sys- tem. This evolution most probably helped to form their original macromolecular organic content as discussed in the UCAMMs formation scenarios (Dartois et al. 2013; Augé et al. 2016). The nitrogen budget of UCAMMs before atmospheric entry may be higher than measured in this study if they suffered some heating during atmospheric entry. The nitrogen-to-carbon atomic ratio is indeed expected to decrease with the heating of the particles (e.g. Bonnet et al. 2015), but should remain relatively stable up to about 500oC. Above such temperatures, the polyaromatic net- work is also strongly modified and the nitrile infrared signature disappears, whereas it is observed in some of the UCAMMs in this study. If the particles were heated to much higher temper- atures, the measured N/C ratios should be considered as lower limits of the initial N/C ratios as heating would result in nitrogen loss. The chemical element the most prone to substantial loss from the organic content of UCAMMs is hydrogen as it is the most labile. The CH content in UCAMMs is variable, as mea- sured in the infrared spectra, and in some cases relatively low. It is therefore not possible to definitely rule on the pre-atmospheric entry CH content of UCAMMs. It may be higher before atmo- spheric entry than measured in this study and reduced upon at- mospheric entry. It could also be the initial and preserved CH content, with the hydrogen being mainly bound as NH, as cur- rently observed. However, taken together, the above-mentioned arguments and measurements provide a substantial body of ev- idence that UCAMMs probably did not suffer a temperature of more than about 100-500oC during atmospheric entry. Nitrogen contribution and the N/C solar system gradient The position of the C≡N band measured in the IR and in the Ra- man spectra, shown in the central panel of Fig. 3, corresponds 0.010.101.00Integrated Area C-H/C=C0.00.20.40.60.8Integrated Area C=O/C=COrthous DaunayVinogradoffQuiricoKebukawa(A)Kebukawa(B)Kebukawa(C)IDPIDPIDPIDPMatrajtUCAMMsCM2C2-ungCI1CM2CI1CR2CM2LL3CI1CM2CM2CM2CM2CM2CR2CR2CR2CM2CM2CR1CI1LL3L3.05C2-ungCO3.0CV3LL3CV3CV3LL3C3-ungCV3CM-heated0.00.20.40.60.8H/C0.00.20.40.60.8Integrated Area C-H/C=COrthous DaunayQuiricoKebukawa(A)Kebukawa(B)Kebukawa(C)UCAMMsC2-ungCM2CI1CM2LL3CI1CM2CM2CM2CM2CR2CR2CR2CM2CM2CR1CI1LL3L3.05C2-ungCO3.0CV3CV3CV3C3-ungCV3CM-heated: Acceptedin A & A to a nitrile function; an isonitrile group would absorb at lower frequencies (e.g. Bonnet et al. 2015). The absorption profile in the fingerprint region (1500-800 cm−1) is significantly differ- ent in UCAMMs compared to meteorites' IOM. It is particu- larly evident when the organic content is the highest and the silicates represent a negligible fraction of the absorption pro- file (lower spectra of left panel in Fig.2). The absorption bands peak around 1380 cm−1, whereas in IOM spectra the maximum of the broad absorptions peaks around 1200 cm−1 (e.g. Fig. 1 Dartois et al. 2014; Quirico et al. 2014; Orthous-Daunay et al. 2013; Kebukawa et al. 2011), with variable methyl and methy- lene deformation mode contributions on top of the profile at higher wavenumbers. This UCAMM fingerprint region profile is reminiscent of the nitrogen-rich carbonaceous network, as ob- served in the spectra of laboratory a-CNH analogues (e.g. Augé et al. 2016; Quirico et al. 2008; Lazar et al. 2008; Gerakines et al. 2004; Fanchini et al. 2002; Rodil et al. 2001; Hammer et al. 2000; Ong 1996), in agreement with the higher nitrogen fraction in the organic phase of UCAMMs (Dobrica et al. 2011; Dartois et al. 2013). If the main nitrogen contribution to the UCAMMs is not the C≡N, but C=N and C-N bonds contributing to the in- frared spectra at lower frequencies, what makes the C≡N feature unique is that it falls in a clean spectral region, and when mea- sured in organic matter, it is in most cases the sign of a high nitrogen content. UCAMMs are probably one of the poles in the nitrogen history of the solar system solids, probably sharing commonalities with some of the nitrogen-rich hot spots found at much smaller scales (i.e. lower organic matter fraction) in some IDPs (e.g. Aléon 2010). The nitrogen-to-carbon abundance ratio in UCAMMs, measured with an electron microprobe for five of them, are shown in Fig. 7 (left panel) together with their nitro- gen absolute abundances (in weight fraction), and are compared to other solar system solids. The N/C is one of the indicators used to discriminate between different reservoirs for the organic matter in solar system solids. Although difficult to evaluate, the Bulk Silicate Earth (BSE) N/C value lies about an order of mag- nitude lower (Halliday 2013; Marty 2012; Bergin et al. 2015, and references therein). The interstellar medium N/C is some- what more difficult to constrain as-in contrast to measurements on interplanetary carbonaceous dust-nitrogen in diffuse ISM solids is spectroscopically elusive. Nevertheless, an upper limit can be determined assuming that the depletion of nitrogen and carbon observed in the diffuse medium is locked into carbona- ceous dust. Considering the missing fraction of nitrogen (δN) one can form the elemental ratio N/C (cid:46) (δN × [N])/(δC × [C]) ≈ (0.2 × 80ppm)/(0.4 × 290ppm) ≈ 0.14 (e.g. Verstraete 2011, and references therein). This ratio represents a stringent upper limit for the N/C ISM value that is in agreement with the fact that spectroscopic observation of both the aromatic infrared bands emission carriers, also called 'astrophysical PAHs' (polycyclic aromatic hydrocarbons), and hydrogenated amorphous carbon solids in the ISM do not show a large incorporation of nitro- gen as an heteroatom. Abundances of N/C in the organic matter from measurements for meteorites from the asteroid belt (Ker- ridge 1985) are also shown in Fig.7. Their N/C ratio and abso- lute N abundance are both lower than in UCAMMs, with a slight overlap for the extremes in each distribution. The bulk N abun- dance in meteorites is lower than that of UCAMM by up to an order of magnitude. UCAMMs are placed in a N/C versus esti- mated heliocentric distance diagram in the right panel of Fig.7. As discussed in Bergin et al. (2015); Millar (2015); Lee et al. (2010), the N/C ratio is sensitive to the disc chemistry and the radial transport, with a greater retention of C over N for thermal or impact events. Disc models (e.g. Piso et al. 2016) endeavor to demonstrate to what extent the C/O and N/C ratio can be re- lated to the abundance of specific carriers, the composition of the ice (dominated or not by water ice, which controls the bind- ing energy of more volatile ices), and the effect on the temporal evolution of disc temperatures on the position of the snow lines of the volatile species. Up to the planetesimals formation phase, observations show outer regions dense and cold enough to con- dense even the most volatile species. This includes the nitrogen molecule that is unfortunately weakly infrared active via ice in- teraction induced transitions, and thus escaping to date a direct abundance determination in ice mantles in this phase. These con- densation phases are indirectly traced by the observation of re- lated species in the gas phase, such as the N2H+ radical. Following this early phase, and for up to 4 Gyrs, larger icy bod- ies at large heliocentric distances can retain volatiles such as CH4 and N2 on their surfaces. An emblematic object of this type is Pluto, with a detailed spatial record of CH4- and N2-rich ices at its surface, recently mapped by the New Horizons mission (e.g. Protopapa et al. 2016). Different terrains and ice composi- tions were observed, and the dynamical history of Pluto seems to favour ices segregations, with regions stable for very long peri- ods of time, possibly over the age of Pluto itself (Hamilton et al. 2016). Pluto is among the largest bodies with (sub)surface CH4 and N2 volatile ices; however, in the Kuiper belt and the Oort cloud regions many smaller icy bodies should have (sub)surface N2 and CH4 dominated ices (e.g. Brown et al. 2011). These sur- faces are exposed to Galactic cosmic-ray irradiation, and macro- molecular precursors can be synthesized by irradiation of such ices (see the scenarios and experiments detailed in Dartois et al. 2013; Augé et al. 2016). In UCAMMs we see little evidence of a pristine, directly incorporated and unmodified, interstellar dust organic component. If the UCAMM organic matter was pro- duced closer to the Sun by an alternative scenario, it would be difficult to reconcile with the high N/C and D/H and lower O/C ratios than found in meteorite IOM. The overall results indicate that the UCAMM bulk organic matter is not a direct heritage from and incorporation of the presolar cloud. The C/Si solar system gradient The silicon-to-carbon abundance ratio in UCAMMs can be eval- uated by forming the ratio of the integrated silicates band around 1000 cm−1 to the contribution from the C=C band around 1590 cm−1, which largely dominates the carbon network. Us- ing the integrated absorption cross sections of the silicates band of 1.8 ± 0.2 × 10−16 cm/Si (Matrajt et al. 2005; Bowey & Adamson 2002; Dartois 1998) and that of the C=C band of 1.5 ± 0.5 × 10−18 cm/C evaluated in Dartois et al. (2013), this ratio can be converted into Si/C values. They are reported in the right ordinate of Fig. 8, showing that the UCAMM silicates con- tent is low (Si/C ∼< 0.03). This Si/C ratio measured in the in- frared confirms the ultracarbonaceous nature of the fragments analysed. In this figure are displayed IDPs measurements of the Si/C by Thomas et al. (1993) and values retrieved from the in- frared spectra from Matrajt et al. (2005). For the latter, the C/Si value reported for each of the seven IDPs analysed, is the mean of the C/Si given in Table 4 of Matrajt et al. (2005), including only aliphatic carbons, and a reevaluation based on the infrared spectra of the maximum C possibly hidden in the C=C band absorption region. The vertical lines indicate the full range be- tween the two calculated values. The major flux of such IDPs is attributed to parent bodies coming from Jupiter-family comets and asteroids, with heliocentric distances in the 3-10 AU range (Poppe et al. 2011; Poppe 2016). Measurements for meteorites A&A proofs: manuscript no. ucamms_edartois_arXiv Fig. 7. UCAMM nitrogen abundances relative to carbon (atomic ratio) are compared to solar system solids: meteorites (Kerridge 1985; Alexander et al. 2007, green and purple circles, respectively; labels correspond to meteorites groups as reported in these articles), the Bulk Silicate Earth (BSE, Bergin et al. 2015), and the interstellar medium (see text for details). Left: Reported in function of bulk nitrogen weight fraction; Right: Reported in function of heliocentric distance. Figure A.1 summarizes the distance, N/C, and N weight fraction information in a single plot. UCAMMs values are reported in Table B.1. Acceptedin A & A from the asteroid belt (Kerridge 1985) are also shown. The bulk silicate Earth (BSE, Bergin et al. 2015, and references therein) C/Si value lies orders of magnitude below. The C/Si can also be evaluated from interstellar medium measurements. An upper limit can be determined by simply forming the cosmic abun- dance ratio for carbon and silicon, which is around 10, based on recent measurements (Jenkins 2014; Jenkins & Tripp 2011; Jenkins 2009; Przybilla et al. 2013). The lower bound can be es- timated for carbon by looking either at the depletions, attributed mainly to solids formation, or also by quantifying the spectro- scopically identified carbonaceous dust (e.g. Dartois et al. 2015, and references therein) or the carbon fraction required in dust models (Jones 2015; Draine 2015; Wang et al. 2014). Most of the silicon is locked into the formation of inorganic compounds (mainly amorphous silicates in the diffuse interstellar medium; Kemper et al. 2004). The lower bound is therefore around 40% of the cosmic abundance. We report the C/Si ratio as a func- tion of the estimated heliocentric distance for the different ob- jects considered (Fig. 8, right panel), and compare it to the inter- stellar medium. The UCAMMs display the highest ratios above the meteorites and, importantly also above the interstellar value, which is considered the maximum direct C/Si abundance her- itage value. As stated in Millar (2015), the carbon abundance ap- pears to be closer to normal value (i.e. C/Si cosmic abundance) in the cooler materials that reside in the outer solar system, but globally the incorporation of carbon into refractory bodies is dif- ficult. The preferential destruction of carbon grains over silicates in the inner protosolar disc (Lee et al. 2010) would naturally ex- plain a carbon deficiency gradient. Recent models propose that a large fraction of carbon in the early solar system was removed from the dust component, especially in the inner regions of the solar system. These models thus pre- dict a radial dependence of the abundance of carbon dust and could explain the depletion observed for the carbon abundance in planetesimals in the asteroid belt and in the terrestrial planet region (e.g. Gail & Trieloff 2017) These variations may have observational consequences for protoplanetary discs. In the re- mote observations of protoplanetary discs, PAH emission lines and carbonaceous dust carriers in general, are conspicuous by their absence, whereas silicates are clearly observed in emis- sion in the zones approximately corresponding to the sizes of the actual inner solar system. In addition to the inherent emis- sivity lower contrast in bands and/or lines for many carbona- ceous solids compared to silicates, such a C/Si radial variation could play an important role in the difficulty encountered to de- tect them. Finding a high C/Si value in an extraterrestrial dust grain-as is found in UCAMMs (C/Si ∼> 10) and some IDPs-associated with a high N/C ratio (Duprat et al. 2010; Dartois et al. 2013), can thus most probably be assigned to an incorporation via a mechanism occurring in the outer solar nebula, such as the pro- posed irradiation scenario for UCAMMs; this is suggested by the C/Si gradient and also because the direct incorporation from the diffuse ISM would apparently lead to a lower C/Si value. 5. Conclusion Infrared and Raman µspectroscopy spectra of eight UCAMMs fragments provide a more comprehensive picture on the physic- ochemical composition of UCAMMs. The UCAMMs clearly contain a large amount, usually well over a micron in size, of N-rich organic matter that is not found in other types of ex- traterrestrial matter. The high carbon abundance relative to sil- icon in UCAMMs is well above that of most solar system prim- itive samples (IDPs, meteorites). It is also, for some of them, significantly above the mere interstellar C/Si value which is 100101102103104105N (µg/g)0.010.10N/C (atomic ratio)UCAMMsUGCICICMHCMCMCMCMCMCMHCMCMCRCMCMHUGUGCRCVOCVRCVRCVOCVOCMHCOCRCRCOCOUGOCCVOCRELOCELEHOCOCOCOCOCOCOCCREHEHOCOCOCCVOEHOCCOOCOCCOOCCVOCOOCCRWCOCRWCRWUGCICICICMCICMCMCMCMCMCMCRCMCMCMCMCMCMCMCMCMCVUGCVCVCVCVCVCOCOCVCOCOCVCVCVUGCOUGMeteoritesBSE110100Distance (au)0.010.10N/C (atomic ratio)/ // /UCAMMsUGCICICICMCICMCMCMCMCMCMCRCMCMCMCMCMCMCMCMCMCVUGCVCVCVCVCVCOCOCVCOCOCVCVCVUGCOUGMeteoritesUGCICICMHCMCMCMCMCMCMHCMCMCRCMCMHUGUGCRCVOCVRCVRCVOCVOCMHCOCRCRCOCOUGOCCVOCRELOCELOCOCOCOCOCOCOCCREHEHOCOCOCCVOEHOCCOOCOCCOOCCVOCOOCCRWCOCRWCRWBSEISM: Fig. 8. Left: Silicates-to-aromatic C=C integrated area ratio and corresponding silicon-to-carbon atomic ratio (right axis, see text for details) in UCAMMs. Right: UCAMMs (stars) carbon abundances relative to silicon (atomic ratio) are compared to solar system solids: IDPs (Matrajt et al. 2005; Thomas et al. 1993, triangles and diamonds, respectively), meteorites (Kerridge 1985, circles), the bulk silicate Earth (BSE, Bergin et al. 2015, cross), and the interstellar medium (square, see text for details). Acceptedin A & A probably not compatible with the hypothesis of a direct incor- poration of an interstellar precursor. This favours scenarios in which the N enrichment is acquired as a secondary physico- chemical process occurring during the protoplanetary phase or later, and not a heritage from pristine incorporated interstellar matter. The UCAMMs micrometeorites reveal the existence of nitrogen-rich precursors formed beyond the nitrogen snow line, and a chemistry regime occurring in the solar system in outer regions. UCAMMs are additional evidence for the presence of a positive gradient of the C/Si and N/C abundance ratios with heliocentric distance, as expected from protoplanetary disc evo- lution models and progressively revealed by astrophysical disc observations. Acknowledgements. We acknowledge SOLEIL for the use of the synchrotron ra- diation facilities. The authors warmly thank Paul Dumas for the discussions and support over many years and for the successful realization of these experiments. The authors are grateful to Bruno Crane, Nicolas Szwec, and Silvin Hervé for their help in the mechanical designs used to perform the experiments. We ac- knowledge Y. Kebukawa for kindly providing us with her original data. The authors would like to cordially thank Martin David Suttle and Tristan Guillot for their constructive comments and the language editor Helenka Kinnan for suggestions, improving the scientific content quality and readability of this ar- ticle. Part of the equipment used in this work has been financed by the French INSU-CNRS program "Physique et Chimie du Milieu Interstellaire" (PCMI). The measurements of the N/C with the electron microprobe were performed thanks to the CAMPARIS team at Jussieu. This work was supported by the ANR projects COSMISME (Grant ANR- 2010-BLAN-0502) and OGRESSE (Grant ANR2011-BS56-026-01) of the French Agence Nationale de la Recherche as well as INSU, IN2P3, CNES, DIM-ACAV (Région Ile de France), CNRS, and Université Paris-Sud. This work is part of the JWST emblematic project from the LABEX-P2IO. We are also grateful to the French and Italian polar institutes IPEV and PNRA, for their financial and logistic support of the micrometeorites Concordia collection. Bonnet, J.-Y., Quirico, E., Buch, A., et al. 2015, Icarus, 250, 53 Bowey, J. E., & Adamson, A. J. 2002, MNRAS, 334, 94 Brown, M. E., Burgasser, A. J., & Fraser, W. C. 2011, ApJ, 738, L26 Busemann, H., Spring, N. H., Alexander, C. M. O., Nittler, L. R. 2011. Raman Spectroscopy on Cometary and Meteoritic Organic Matter. Spectroscopy Let- ters 44, 554-559. Busemann, H., Nguyen, A. N., Cody, G. D., Hoppe, P., Kilcoyne, A. L. D., Stroud, R. M., Zega, T. J., Nittler, L. R. 2009. Ultra-primitive interplanetary dust particles from the comet 26P/Grigg-Skjellerup dust stream collection. Earth and Planetary Science Letters 288, 44-57. Busemann, H., Alexander, M. O., Nittler, L. R. 2007. Characterization of in- soluble organic matter in primitive meteorites by microRaman spectroscopy. Meteoritics and Planetary Science 42, 1387-1416. References Alexander, C. M. O. '., Cody, G. D., De Gregorio, B. T., Nittler, L. R., & Stroud, Survey of data sources on sulfide phase equilibria. In Sulfide Mineralogy (ed. P.H. Ribbe), pp. CS 3-23. Reviews in Mineralogy I , Mineralogical Society of America, Washington, D.C., USA. Dartois, E. 1998, Ph.D. Thesis, Dartois, E., and 16 colleagues 2013. UltraCarbonaceous Antarctic micromete- orites, probing the Solar System beyond the nitrogen snow-line. Icarus 224, 243-252. Dartois, E., and 16 colleagues 2014, Geochemical Journal, 48 Dartois, E., Alata, I., Engrand, C., et al. 2015, Bulletin de la Societe Royale des Bergin, E. A., Blake, G. A., Ciesla, F., Hirschmann, M. M., & Li, J. 2015, Pro- ceedings of the National Academy of Science, 112, 8965 R. M. 2017, Chemie der Erde / Geochemistry, 77, 227 Aléon, J. 2010, ApJ, 722, 1342 Alexander, C. M. O. '., Fogel, M., Yabuta, H., & Cody, G. D. 2007, Geochim. Cosmochim. Acta, 71, 4380 Augé, B., Dartois, E., Engrand, C., et al. 2016, A&A, 592, A99 Bardin, N., Slodzian, G., Wu, T.-D., et al. 2014, Lunar and Planetary Science Charon, E., Engrand, C., Benzerara, K., et al. 2017, Lunar and Planetary Science Conference, 45, 2647 Conference, 48, 2085 Sciences de Liege, 84, 7 Dobrica, E., Engrand, C., Duprat, J., & Gounelle, M. 2010, Meteoritics and Plan- etary Science Supplement, 73, 5213 Dobrica, E., Engrand, C., Quirico, E., Montagnac, G., Duprat, J. 2011. Raman characterization of carbonaceous matter in CONCORDIA Antarctic microm- eteorites. Meteoritics and Planetary Science 46, 1363-1375. 1020304050Integrated Area C=C012345Integrated Area Silicates/C=C0.000.010.020.03N(Si)/N(C)C=C110100Distance (au)10-410-2100102104C/Si (atomic ratio)/ // /UCAMMsIDPsBSEMeteoritesISM dustA&A proofs: manuscript no. ucamms_edartois_arXiv Draine, B. T. 2015, IAU General Assembly, 22, 2253136 Duprat, J., Engrand, C., Maurette, M., et al. 2006, Meteoritics and Planetary Supplement, 31, Taylor, S., Lever, J. H., & Harvey, R. P. 1996, Meteoritics and Planetary Science Taylor, S., Alexander, C. M. O., & Wengert, S. 2008, Lunar and Planetary Sci- Duprat, J., Engrand, C., Maurette, M., et al. 2007, Advances in Space Research, ence Conference, 39, 1628 Science Supplement, 41, 5239 39, 605 Conference, 46, 1902 81P/Wild 2 by the Stardust Spacecraft. Science 314, 1720. Suttle, M. D., Genge, M. J., Folco, L., & Russell, S. S. 2017, Geochim. Cos- mochim. Acta, 206, 112 Physics, 104, 073534 Planetary Science, 36, 1377 66, 195415 ence, 33, 267 97 Zolensky, M., Bland, P., Brown, P., & Halliday, I. 2006, Meteorites and the Early Hammer, P., Lacerda, R. G., Droppa, R., Jr., & Alvarez, F. 2000, Diamond and Related Materials, 9, 577 Yabuta, H., Noguchi, T., Itoh, S., et al. 2017, Geochim. Cosmochim. Acta, 214, Planetary Science, 47, 228 Greshake, A., Kloeck, W., Arndt, P., et al. 1998, Meteoritics and Planetary Sci- ference, 43, 2239 172 Solar System II, 869 Thomas, K. L., Blanford, G. E., Keller, L. P., Klock, W., & McKay, D. S. 1993, Geochim. Cosmochim. Acta, 57, 1551 Toppani, A., Libourel, G., Engrand, C., & Maurette, M. 2001, Meteoritics and Halliday, A. N. 2013, Geochimica Et Cosmochimica Acta 105, 146 Hamilton, D. P., Stern, S. A., Moore, J. M., & Young, L. A. 2016, Nature, 540, Duprat, J., Dobrica, E., Engrand, C., et al. 2010, Science, 328, 742 Engrand, C., Benzerara, K., Leroux, H., et al. 2015, Lunar and Planetary Science Engrand, C., Duprat, J., Dartois, E., Godard, M., & Delauche, L. 2017, EGU General Assembly Conference Abstracts, 19, 9979 Fanchini, G., Tagliaferro, A., Conway, N. M., & Godet, C. 2002, Phys. Rev. B, Verstraete, L. 2011, European Physical Journal Web of Conferences, 18, 01001 Wang, S., Li, A., & Jiang, B. W. 2014, Planet. Space Sci., 100, 32 Yabuta, H., Itoh, S., Noguchi, T., et al. 2012, Lunar and Planetary Science Con- Janches, D., Heinselman, C. J., Chau, J. L., Chandran, A., & Woodman, R. 2006, Journal of Geophysical Research (Space Physics), 111, A07317 Gail, H.-P., & Trieloff, M. 2017, A&A, 606, A16 Gerakines, P. A., Moore, M. H., & Hudson, R. L. 2004, Icarus, 170, 202 van Ginneken, M., Folco, L., Cordier, C., & Rochette, P. 2012, Meteoritics and Kemper, F., Vriend, W. J., & Tielens, A. G. G. M. 2004, ApJ, 609, 826 Kerridge, J. F. 1985, Geochim. Cosmochim. Acta, 49, 1707 Kouketsu, Y., 2014. Island Arc 23, 33-50. Lazar, G., Bouchet-Fabre, B., Zellama, K., et al. 2008, Journal of Applied Jenkins, E. B. 2014, arXiv:1402.4765 Jenkins, E. B., & Tripp, T. M. 2011, ApJ, 734, 65 Jenkins, E. B. 2009, ApJ, 700, 1299 Jones, A. 2015, arXiv:1511.07988 Kebukawa, Y., Alexander, C. M. O. '., Cody, G. D. 2011. Compositional diver- sity in insoluble organic matter in type 1, 2 and 3 chondrites as detected by infrared spectroscopy. Geochimica et Cosmochimica Acta 75, 3530-3541. Acceptedin A & A Nesvorný, D., Vokrouhlický, D., Pokorný, P., & Janches, D. 2011, ApJ, 743, 37 Nesvorný, D., Jenniskens, P., Levison, H. F., et al. 2010, ApJ, 713, 816 Ong, C. 1996, Thin Solid Films, 280, 1 Orthous-Daunay, F.-R., Quirico, E., Beck, P., Brissaud, O., Dartois, E., Pino, T., Schmitt, B. 2013. Mid-infrared study of the molecular structure variability of insoluble organic matter from primitive chondrites. Icarus 223, 534-543. Lee, J.-E., Bergin, E. A., & Nomura, H. 2010, ApJ, 710, L21 Love, S. G., & Brownlee, D. E. 1991, Icarus, 89, 26 Love, S. G., & Brownlee, D. E. 1993, Science, 262, 550 Marty, B. 2012 Earth and Planetary Science Letters 313, 56 Matrajt, G., Muñoz Caro, G. M., Dartois, E., et al. 2005, A&A, 433, 979 Matrajt, G., Brownlee, D. E., Joswiak, D. J., & Taylor, S. 2005, 36th Annual Rotundi, A., and 19 colleagues 2008. Combined micro-Raman, micro-infrared, and field emission scanning electron microscope analyses of comet 81P/Wild 2 particles collected by Stardust. Meteoritics and Planetary Science 43, 367- 397. Sadezky, A., Muckenhuber, H., Grothe, H., Niessner, R., Poschl, U. 2005. Ra- man microspectroscopy of soot and related carbonaceous materials: Spectral analysis and structural information. Carbon 43, 1731-42. Sandford, S. A., and 54 colleagues 2006. Organics Captured from Comet Piso, A.-M. A., Pegues, J., & Öberg, K. I. 2016, ApJ, 833, 203 Poppe, A., James, D., & Horányi, M. 2011, Planet. Space Sci., 59, 319 Poppe, A. R. 2016, Icarus, 264, 369 Prasad, M. S., Rudraswami, N. G., & Panda, D. K. 2013, Journal of Geophysical Millar, T. J. 2015, Plasma Sources Science Technology, 24, 043001 Min, M., Hovenier, J. W., & de Koter, A. 2003, A&A, 404, 35 Nakamura, T., Noguchi, T., Ozono, Y., Osawa, T., & Nagao, K. 2005, Meteoritics Rietmeijer, F. J. M. 2004, Meteoritics and Planetary Science, 39, 1869 Rodil, S. E., Ferrari, A. C., Robertson, J., & Milne, W. I. 2001, Journal of Applied Quirico, E., Montagnac, G., Lees, V., et al. 2008, Icarus, 198, 218 Quirico, E., Orthous-Daunay, F.-R., Beck, P., et al. 2014, Geochim. Cos- Protopapa, S., Grundy, W. M., Reuter, D. C., et al. 2016, arXiv:1604.08468 Przybilla, N., Nieva, M. F., Firnstein, M., & Butler, K. 2013, EAS Publications mochim. Acta, 136, 80 Physics, 89, 5425 Lunar and Planetary Science Conference, 36. Research (Planets), 118, 2381 Series, 64, 37 and Planetary Science Supplement, 40, 5046 : Fig. A.1. UCAMM nitrogen abundances relative to carbon (atomic ra- tio) are compared to solar system solids: meteorites (Alexander et al. 2007; Kerridge 1985), the bulk silicate Earth (BSE, Bergin et al. 2015), and the interstellar medium (see text for details). The sizes of the sym- bols for UCAMMs and meteorites give the absolute nitrogen content (in wt %). UCAMM values are reported in Table B.1. Acceptedin A & A Appendix A: N/C extended version Appendix B: Measurements summary B.1 110100Distance (au)0.010.10N/C (atomic ratio)/ // / 2 wt% N abs. scaleUCAMMs 0.2 wt%MeteoritesBSEISM dustA&A proofs: manuscript no. ucamms_edartois_arXiv 7 . . 1 . 3 3 . . . 0 7 1 0 . 5 . . 6 3 0 . 5 . . . 0 . 3 - - - - 5 . . 3 4 . . . . 3 1 5 1 9 6 . . 2 . . . 5 1 . 8 . 2 0 3 . 2 2 . 8 . 3 3 . 9 . 2 1 . 5 4 . 5 3 . 3 . 5 5 . e h t n i . 8 4 7 2 . 3 5 3 2 . 4 7 . 0 2 2 3 . 5 2 1 . 5 7 4 1 . 7 3 4 1 6 . 9 7 2 5 . 0 2 1 s e n o z 6 . 2 4 3 1 0 ± 0 0 ± 3 5 ± 1 4 ± 7 6 ± 7 8 . 8 7 5 1 0 ± 3 0 ± 6 . 0 0 5 3 1 . 9 5 5 3 1 1 . 2 6 3 1 . 2 6 8 5 1 . 4 5 8 5 1 8 . 4 8 5 1 t n e c r e p . d e b o r p 0 ± 6 6 t n e r e ff i d . 8 3 ± 1 . 0 ± 3 . 2 1 4 7 0 6 0 C D 8 1 7 0 6 0 C D 5 6 5 0 6 0 C D 2 ± 9 9 0 . 2 ± 1 . 2 1 . 6 3 1 ± 0 9 1 1 9 0 6 0 C D . 8 1 1 ± 5 9 . 8 1 1 ± 5 9 . 1 ± 2 . 2 . 5 8 6 1 8 2 ± 0 1 1 . . 6 0 0 0 ± 2 0 . 6 0 2 ± 5 8 . 6 . 5 1 1 ± 7 . 6 1 . 0 4 0 0 ± 7 4 1 8 1 ± 9 5 0 7 1 2 1 ± 6 0 0 7 1 c i m o t a C & N e h t 4 7 ± 8 8 9 6 1 3 3 ± 2 7 8 0 ± 8 5 0 0 1 ± 8 0 5 9 3 ± 7 . 1 7 0 ± 0 7 1 . 0 Acceptedin A & A m c 5 . 2 ± n o i t i s o p ∼ y t n i a t r e c n u s n o i t a i r a v M M A C U c i s n i r t n i 5 8 . 2 ± 2 . 0 1 5 3 . 5 ± 0 . 6 7 4 5 . 1 ± 1 4 . 2 0 5 . 2 ± 3 3 . 8 m c 0 1 ± M H W F s t n e m e r u s a e m d e r a r f n I 7 . 7 ± 7 . 3 4 2 . 0 ± 5 . 9 2 1 . 5 ± 0 . 7 8 6 1 5 1 . 0 ± 5 4 . 1 8 . 7 ± 6 . 3 9 6 1 9 0 . 1 ± 4 8 . 0 9 2 0 . 0 ± 7 5 0 . 0 1 . 4 1 ± 5 . 8 5 5 . 2 1 ± 1 . 5 8 6 1 3 3 . 5 ± 5 7 . 7 1 2 1 . 0 ± 3 7 . 0 e b o r p o r c i m n o r t c e l e 9 3 0 . 0 ± 5 3 1 . 0 8 7 . 0 1 ± 3 9 . 5 3 5 1 . 0 ± 2 7 . 7 ± 0 7 1 0 . 7 ± 7 . 5 8 6 1 1 2 . 1 ± 2 6 9 . 0 1 ± 1 . 7 2 5 . 0 ± 5 . 7 s t n e m e r u s a e m e h t s t n e m e r u s a e m # 4 8 . 5 1 ± 7 7 9 . 4 ± 0 . 3 1 n o i t i s o p O = C w d e n i m r e t e d 5 . 1 ± 3 . 3 ) a σ 1 ± ( c i m o t a C N / n o i t i s o P M H W F n o i t i s o P M H W F n a m a R d n a b D 9 1 9 0 6 0 C D 9 1 1 1 2 0 C D 4 9 5 0 6 0 C D 3 4 4 0 6 0 C D m c ( s e t a c i l i s r e t e m a r a p M M A C U % t a ( % t a ( s e d u l c n i y r a m m u S ) 1 − ) 1 − m c ( H C d n a b G (cid:82) (cid:82) (cid:82) (cid:82) 9 . 5 9 5 1 9 . 9 8 5 1 5 . 6 9 5 1 0 . 9 4 3 1 6 . 6 5 3 1 8 . 6 9 5 1 4 . 8 5 3 1 5 . 2 6 3 1 d e t a m s r o r r e 4 . 0 5 1 3 . 0 0 1 C = C O = C m c ( m c ( 2 . 6 3 2 4 . 5 8 2 e h t h t i . 2 5 . 3 ) 1 − m c ( 0 . 7 6 2 ) 1 − s o i t a r . . 2 / C N N C 6 . 2 5 2 7 . 9 8 6 . 0 9 1 3 3 4 ) 1 − S ) 1 − , 1 − 0 2 f o i t s e ( d n a n o i t a i v e d d r a d n a t s e h T a - - - - - - - - ) ) . 1 . B e l b a T
1702.02542	1	1702	2017-02-08T17:52:53	Probabilistic Constraints on the Mass and Composition of Proxima b	[ "astro-ph.EP" ]	Recent studies regarding the habitability, observability, and possible orbital evolution of the indirectly detected exoplanet Proxima b have mostly assumed a planet with $M \sim 1.3$ $M_\oplus$, a rocky composition, and an Earth-like atmosphere or none at all. In order to assess these assumptions, we use previous studies of the radii, masses, and compositions of super-Earth exoplanets to probabilistically constrain the mass and radius of Proxima b, assuming an isotropic inclination probability distribution. We find it is ~90% likely that the planet's density is consistent with a rocky composition; conversely, it is at least 10% likely that the planet has a significant amount of ice or an H/He envelope. If the planet does have a rocky composition, then we find expectation values and 95% confidence intervals of $\left<M\right>_\text{rocky} = 1.63_{-0.72}^{+1.66}$ $M_\oplus$ for its mass and $\left<R\right>_\text{rocky} = 1.07_{-0.31}^{+0.38}$ $R_\oplus$ for its radius.	astro-ph.EP	astro-ph	Draft version May 17, 2018 Preprint typeset using LATEX style AASTeX6 v. 1.0 PROBABILISTIC CONSTRAINTS ON THE MASS AND COMPOSITION OF PROXIMA B Alex Bixel2 and D´aniel Apai1,2 Steward Observatory 933 North Cherry Avenue Tucson, AZ 85721, USA 7 1 0 2 b e F 8 . ] P E h p - o r t s a [ 1 v 2 4 5 2 0 . 2 0 7 1 : v i X r a 1Department of Planetary Science/Lunar and Planetary Laboratory, The University of Arizona, 1640 E. University Blvd., Tucson, AZ 85718, USA 2Earths in Other Solar Systems Team, NASA Nexus for Exoplanet System Science ABSTRACT Recent studies regarding the habitability, observability, and possible orbital evolution of the indirectly detected exoplanet Proxima b have mostly assumed a planet with M ∼ 1.3 M⊕, a rocky composition, and an Earth-like atmosphere or none at all. In order to assess these assumptions, we use previous studies of the radii, masses, and compositions of super-Earth exoplanets to probabilistically constrain the mass and radius of Proxima b, assuming an isotropic inclination probability distribution. We find it is ∼ 90% likely that the planet's density is consistent with a rocky composition; conversely, it is at least 10% likely that the planet has a significant amount of ice or an H/He envelope. If the planet does have a rocky composition, then we find expectation values and 95% confidence intervals of (cid:104)M(cid:105)rocky = 1.63+1.66−0.72 M⊕ for its mass and (cid:104)R(cid:105)rocky = 1.07+0.38−0.31 R⊕ for its radius. 1. INTRODUCTION The recent radial velocity detection of a planet in the habitable zone of the nearby M dwarf Proxima Centauri (hereafter 'Proxima b' and 'Proxima') (Anglada-Escud´e et al. 2016) has spurred over a dozen theoretical papers speculating on the planet's atmosphere (e.g., Brugger et al. 2016; Goldblatt 2016), habitability (e.g., Ribas et al. 2016; Turbet et al. 2016), and orbital and forma- tion histories (e.g., Barnes et al. 2016; Coleman et al. 2017) as well as prospects for a direct detection or at- mospheric characterization (e.g., Lovis et al. 2016; Luger et al. 2016). As Proxima is the nearest neighbor to the solar system, it has been suggested as a target for future space missions, including those hoping to characterize its atmosphere and search for life (e.g., Belikov et al. 2015; Schwieterman et al. 2016). In many of these studies, authors have assumed a rocky planet with a thin atmosphere or no atmosphere at all, and some have assumed a mass near or equal to the projected mass of M sin(i) = 1.27+0.20−0.17 M⊕, but little has been done to assign a degree of certainty to these as- sumptions. Most notably, previous studies have revealed two distinct populations of exoplanets with super-Earth radii: 'rocky' planets composed almost entirely of rock, iron, and silicates with at most a thin atmosphere, and 'sub-Neptune' planets which must contain a significant amount of ice or a H/He envelope (e.g., Rogers 2015; Weiss & Marcy 2014). If there is a significant proba- bility that Proxima b is of the latter composition, then this should be taken into account when assessing its po- tential habitability or observability. In this letter, we generate posterior distributions for the mass of Proxima b using Monte Carlo simulations of exoplanets with an isotropic distribution of inclinations, where the radii, masses, and compositions of the simu- lated planets are constrained by results from combined transit and radial velocity measurements of previously detected exoplanets. By comparing the posterior mass distribution to the composition of planets as a function of mass, we determine the likelihood that Proxima b is, in fact, a rocky world with a thin (if any) atmosphere. 2. PRIOR ASSUMPTIONS Radial velocity and transit studies of exoplanets have yielded mass and radius measurements for a statistically significant number of targets, thereby enabling the study of how the occurrence and composition of exoplanets varies with planet radii, orbital periods, and host star type. In this section, we review previous results which we will use to place stronger constraints on the mass and composition of Proxima b. 2.1. sin(i) distribution It can be shown (e.g., Ho & Turner 2011) that the probability distribution of sin(i) corresponding to an 2 isotropic inclination distribution is (cid:113) P (sin(i)) = sin(i)/ 1 − sin2(i) (1) Since this distribution peaks at sin(i) = 1, the mass distribution of an RV-detected planet - assuming no prior constraints on the mass - peaks at the minimum mass M0. In their models of the possible orbital histories of Proxima b, Barnes et al. (2016) find that galactic tides could have inflated the eccentricity of the host star's (at the time unconfirmed) orbit around the α Cen bi- nary, leading to encounters within a few hundred AU and the possible disruption of Proxima's planetary sys- tem. If so, this could affect the likely inclination of the planet in a non-isotropic way. However, Kervella et al. (2017) have presented radial velocity measurements showing that Proxima is gravitationally bound to the α Cen system with an orbital period of 550,000 years, an eccentricity of ∼ 0.5, and a periapsis distance of 4,200 AU. At this distance, the ratio of Proxima's gravita- tional field to that of α Cen at the planet's orbit (∼ 0.05 AU) is greater than 108; unless Proxima's orbit was sig- nificantly more eccentric in the past, it seems unlikely that α Cen would have disrupted the system. 2.2. Occurrence rates for M dwarfs Mulders et al. (2015) provide up-to-date occurrence rates of planets around M dwarf stars from the Kepler mission. The sample is limited to 2 < P < 50 days, over which they find the occurrence rates to be mostly inde- pendent of the period. The binned rates and a regres- sion curve, as well as their uncertainties, are presented in Figure 1. Kepler statistics for M dwarfs remain incomplete be- low 1 R⊕, but complete statistics for earlier-type stars suggest a flat distribution for 0.7 < R < 1 R⊕ (Mulders et al. 2015). Since mass-radius relationships typically find a strong dependence of mass on radius (M ∝ R3−4) (e.g. Weiss & Marcy 2014; Wolfgang et al. 2016), we as- sume a priori that Proxima b (M (cid:38) 1.3 M⊕) is larger than 0.7 R⊕. Therefore, for this letter we adopt the regression curve fitted to the binned data, but set the occurrence rates to be flat for R < 1 R⊕. 2.3. Compositions Multiple works (e.g. Weiss & Marcy 2014; Rogers 2015) have determined the existence of two distinct pop- ulations of exoplanets smaller than Neptune (R (cid:46) 4 R⊕): a small radius population with densities consistent with an entirely iron and silicate composition (hereafter 'rocky'), and a large radius population with lower den- sity planets which must have significant amounts of ice or a thick H/He atmosphere (hereafter 'sub-Neptunes'). Occurrence Rates for M Dwarf Planets (with 2 < P < 50 days) Mass-Radius Relationships for R < 4 R⊕ Figure 1. Top: Occurrence rates from Mulders et al. (2015), fitted by a regression curve. We adjust the rates below 1 R⊕ (dotted) to be flat, since the sample is incomplete in this range. Bottom: Mass-radius relationships for the rocky (blue) and sub-Neptune (red) populations. The plotted rela- tionships are from Wolfgang et al. (2016) (solid) and Weiss & Marcy (2014) (dashed). Rogers (2015) studies the abundance of planets of each composition as a function of radius. They define fα(R) as the likelihood that a planet of radius R is dense enough to be consistent with a rocky composition, and determine fα for a sample of planets with known masses and radii. They suggest fitting the data with a two- parameter linear model:  = fα (RP , Rthresh, ∆R) Rthresh − RP ∆R 1 0.5 + 0 RP < Rthresh − 1 RP − Rthresh < 1 RP > Rthresh + 1 2 ∆R 2 ∆R 2 ∆R (2) They find a step function to best describe the data, with ∆R fixed at zero and Rthresh ≈ 1.5 R⊕. For the purposes of this letter, we prefer this fit, but will also vary Rthresh and ∆R to see how they affect our results. We stress that a planet for which fα = 1 is only suf- ficiently dense to be rocky; we still cannot necessarily exclude an ice or volatile component. Here, we will as- sume that all planets for which fα = 1 follow the low- radius M-R relationships given in the following section, which were empirically fitted without prior knowledge of the planets' compositions. For simplicity, we refer to these as 'rocky' planets, and the other population as 'sub-Neptunes', but we will revisit this distinction later on. Since Proxima b is in the habitable zone, it receives an amount of stellar flux comparable to that received by Earth, so we should bear in mind the possibil- ity that the volatile envelope of a sub-Neptune could be lost due to photoevaporation. Owen & Mohanty (2016) model rocky planets with thick H/He envelopes in the habitable zones of M dwarfs, finding that plan- ets with M > 0.8 M⊕ maintain their envelopes over Gyr timescales and are therefore uninhabitable. The 2σ lower limit on the minimum mass of Proxima b is 0.93 M⊕, so it is unlikely that any H/He envelope on the planet would evaporate under this rule. However, we note that this study focuses on planets with a primarily rocky composition, so it may not be directly applicable to habitable zone sub-Neptunes. Additionally, Zahnle & Catling (2013) empirically de- fine boundaries for atmospheric evaporation as a func- tion of stellar heating, escape velocity, and atmospheric composition. In particular, a planet receiving an Earth- like flux must have an escape velocity above ∼ 8 km/s in order to maintain an H2 atmosphere for 5 Gyr. We will revisit this requirement in Section 4.2. 2.4. Mass-radius relationships Empirically derived relationships between exoplanet masses and radii rely on radial velocity (RV) or transit- timing variation (TTV) measurements of transiting ex- oplanet masses. Weiss & Marcy (2014) fit a mass-radius (hereafter M-R) relationship to a sample of 65 transiting exoplanets, in which they find evidence for the two pop- ulations discussed in Section 2.3. Through least-squares regression, they find the densities of the rocky planets (cid:18) RP (cid:19) RE (cid:19)0.63 (cid:18) RP RE MP M⊕ = 4.87 (cid:19)2.3 (cid:19)1.3 (cid:18) RP (cid:18) RP RE RE MP M⊕ = 1.4 MP M⊕ = 2.7 3 (3) (4) (5) (6) to increase linearly with planet radius: ρP = 2.43 + 3.39 g cm−3 while the RV-measured masses of sub-Neptunes increase nearly linearly with planet radius: Wolfgang et al. (2016) use an expanded version of this data set to fit power law M-R relationships using a more statistically robust Bayesian method. For the rocky planets, they find and for the sub-Neptunes with RV-measured masses, Due to the larger sample size and more robust fitting procedure, we adopt Equations 5 and 6 as our preferred M-R relationships, but for completeness we consider the Weiss & Marcy (2014) relationships as well. We find that the choice of M-R relationships has a minimal im- pact on our final results. Both sets of relationships are plotted in Figure 1. It is important to note that the above relationships for sub-Neptunes exclude masses measured by TTV, since TTV masses have been found to be systematically lower than RV masses. This could indicate a selection bias or systematic error in the method used, but since Prox- ima b was detected through RV measurements, we be- lieve it is proper to exclude the TTV masses either way. It is also clear that there is a significant spread in the masses of the observed planets. Wolfgang et al. (2016) suggest a spread of ±1.9 M⊕ for the sub-Neptune plan- ets, which we adopt for our simulations. For rocky plan- ets, the spread is noticeably smaller. There are too few planets to constrain this spread, but it should most likely increase with mass, so we arbitrarily define the spread to be 30% of the calculated mass. 3. METHOD 3.1. Simulated sample The fitted occurrence rates and their uncertainties (f ± df ) are given in even bins in log-space. We use them to generate a random sample of radii, where the number of radii in each bin (r0) is selected from a nor- mal distribution with mean value f (r0) and standard deviation df (r0). We find that the results converge for 1,000 samples of the occurrence rates, with each sample containing ∼ 106 radii. 4 To each radius, we assign a composition ('rocky' or 'sub-Neptune') based on the model of Rogers (2015) (Equation 2), with Rthresh = 1.5 R⊕ and ∆R = 0. We then assign a mass to each radius and composition from a Gaussian distribution with mean value M (R) - calcu- lated using our chosen M-R relationships (Equations 5 and 6) - and a standard deviation dM which represents the spread. We choose a spread proportional to the cal- culated mass for rocky planets (dM = 0.3× M (R)), but a constant spread for sub-Neptunes (dM = 1.9 M⊕). We also reject negative masses, which could in principle bias the assigned masses towards higher-than-average values - however, we find that only a negligible number of masses are rejected. Finally, we assign a line-of-sight inclination parameter sin(i) to each planet, drawn from the isotropic probabil- ity distribution discussed in Section 2.1. 3.2. Prior and posterior probability distributions Prior mass distribution Posterior mass distribution The prior mass and radius distributions, P (M ) and P (R), can be derived directly from the simulated sam- ple. Factoring in the projected minimum mass M0, the posterior distributions P (MM0) and P (RM0) can be calculated from Bayes' formula: P (M0X)P (X) P (M0) P (XM0) = (7) where X represents mass or radius. Since M0 is known, P (M0) is just a normalizing constant. Taking M0 = 1.27+0.20−0.17 M⊕ as the projected mass of Proxima b and the upper limit σM0 = 0.20 M⊕ as its standard devia- tion, we calculate for each simulated planet Pi(M0X) = exp(cid:0)−(M0 − Mi sini(i))2/2σ2 (cid:1) (8) Then P (M0M ) and P (M0R) are the average values of Pi(M0X) for each bin in mass or radius. The prior and posterior distributions are calculated for each sam- ple of 106 planets, and the final results are taken to be the mean result of 1,000 samples. M0 3.3. Posterior compositional probability The prior probability that a planet in a given mass bin is rocky is equal to the number of simulated rocky plan- ets in that bin divided by the total number of planets in the same bin. Since we want to know the likelihood that Proxima b belongs to the 'rocky' population, we multiply this prior composition probability distribution by the posterior mass distribution from the previous sec- tion and integrate over all masses. 4. RESULTS 4.1. Mass distributions The prior and posterior mass probability distributions for Proxima b are plotted in Figure 2. The shaded re- Figure 2. Prior (top) and posterior (bottom) mass distribu- tions for the simulated sample. The blue and red shaded re- gions represent contributions due to rocky and sub-Neptune planets, respectively. The dash-dotted line is the posterior distribution assuming a flat prior distribution. The binning is 0.01 M⊕. gions demonstrate the relative contributions of the pop- ulations at each mass. The prior distribution is valid for RV-detected planets around M dwarfs with intermedi- ate periods (2 < P < 50 days) and radii (0.7 < R < 4 R⊕), while the posterior distribution can be taken as the mass probability distribution for Proxima b. For reference, we include the posterior distribution given no prior constraints on the mass; that is, the dis- tribution resulting from an isotropic sin(i) distribution and the measured M0 with its uncertainty. We find that this nearly matches our result, since both P (sin(i)) and P (M ) are bottom-heavy. Figure 3 shows the cumulative probability that M < X for both of the considered M-R relationships (Section 2.4) as well as for the case of no prior mass distribution. Cumulative mass probability distribution Figure 3. The cumulative mass probability distribution for our simulated posterior mass distribution (solid) and assum- ing a flat prior P (M ) (dashdot). The dotted lines intersect 68% and 95% confidence upper limits on the mass. We find that there is little difference between the results for each M-R relationship. 4.2. Escape velocity In order to verify that sub-Neptune planets can main- tain H2 envelopes in the habitable zone, we compare the escape velocities of our simulated sub-Neptunes to the ∼ 8 km/s cutoff for H2 atmospheric escape (assuming an Earth-like stellar flux) defined by Zahnle & Catling (2013). In both the prior and posterior distributions of escape velocities, we find that fewer than 1% of the sub-Neptunes have escape velocities below this thresh- old, with most having ve (cid:38) 15 km/s. Therefore, we do not believe that Proxima b will be subject to significant atmospheric loss if it has a sub-Neptune composition. 4.3. Composition Table 1 lists the sets of parameters for which we run the simulation, including the mass spread dM for each composition and the central value (Rthresh) and width (∆R, if nonzero) of the transition region defined by Equation 2. The following results for each case are given: the probability Procky that Proxima b belongs to the 'rocky' category of planets, i.e. that its density is consistent with a fully iron and silicate composition, and the expectation values (cid:104)M(cid:105)rocky and (cid:104)R(cid:105)rocky of the mass and radius under the assumption that it belongs to this population. Case A is most consistent with the previous work we have cited, so we take it as our primary result. In this case, there is a ∼ 90% probability that Proxima b be- longs to the 'rocky' population, with an ∼ 10% likeli- hood that it belongs to the 'sub-Neptune' population. In the case that it is rocky, the expectation values (and 5 95% confidence intervals) for the mass and radius are (cid:104)M(cid:105)rocky = 1.63+1.66−0.72 M⊕ and (cid:104)R(cid:105)rocky = 1.07+0.38−0.31 R⊕. We investigate the effect of increasing (Case B) and decreasing (Case C) the mass spread for each composi- tion, which results in lower and higher values of Procky, respectively. This results from low-radius (R ∼ 1.5 R⊕), low-mass sub-Neptunes; when dM is large, they can lie significantly below the M-R relation with masses be- tween 1 and 2 M⊕, so that they are indistinguishable from the rocky planets in the mass domain. In Cases D and E, we determine the effect of raising or lowering the threshold radius Rthresh at which the rocky and sub-Neptune populations are split. A 0.2 R⊕ offset in either direction, which encompasses most of the values suggested in the literature, results in a ∼ 5% to 8% shift in Procky, where higher threshold radii allow for more rocky planets and therefore a higher probability of a rocky composition. Furthermore, allowing for a non- zero width ∆R to the cutoff region allows sub-Neptunes to exist with lower radii and masses, thereby decreasing Procky. In all cases, we find Procky to be between 80% and 95% using the Wolfgang et al. (2016) M-R relationship, and we find similar values using the Weiss & Marcy (2014) relationship (e.g. Procky = 90.7% for Case A), so this result does not vary substantially over the range of rea- sonable values for the input parameters. 5. CONCLUSIONS By considering occurrence rates from the Kepler mis- sion and empirically derived M-R relationships, we de- rive a posterior probability distribution for the actual mass of Proxima b. If the planet has a rocky composi- tion, i.e. if it obeys the low-radius M-R relationship of Wolfgang et al. (2016), then the expectation values of the mass and radius (with 95% confidence intervals) are (cid:104)M(cid:105)rocky = 1.63+1.66−0.72 M⊕ and (cid:104)R(cid:105)rocky = 1.07+0.38−0.31 R⊕. In all of our simulations, we find a probability of 80% to 95% that Proxima b belongs to the 'rocky' popula- tion of planets defined in Section 2.3. In our 'best guess' scenario (Case A), this probability is 90%. Critically, we note that we have assumed all planets with fα = 1 (ac- cording to the Rogers (2015) criterion) are rocky plan- ets, while in reality their density is only consistent with such a composition. With this in mind, it is safest to say that there is at least a 10% chance that Proxima b has a sub-Neptune composition. If it is a sub-Neptune, then its surface gravity is high enough that it could maintain a thick hydrogen atmosphere. For future theoretical work involving the habitabil- ity and detectability of Proxima b, we advise caution regarding assumptions made about its mass or compo- sition; if Proxima b does possess a thick H/He enve- lope, then it is likely not habitable in the traditional Table 1. Monte-Carlo Simulation Parameters and Results Parameters dM (rocky) dM (sub-Neptune) Rthresh 0.3 × M 0.6 × M 0.1 × M 0.3 × M 0.3 × M 0.3 × M 0.6 × M 1.9 M⊕ 3.8 M⊕ 0.7 M⊕ 1.9 M⊕ 1.9 M⊕ 1.9 M⊕ 3.8 M⊕ 1.5 R⊕ 1.5 R⊕ 1.5 R⊕ 1.7 R⊕ 1.3 R⊕ 1.5 R⊕ 1.5 R⊕ ∆R - - - - - 1.2 R⊕ 1.2 R⊕ Results (cid:104)M(cid:105)rocky (cid:104)R(cid:105)rocky Procky 89.9% 1.63+1.66−0.72 M⊕ 1.07+0.38−0.31 R⊕ 1.03+0.42−0.36 R⊕ 84.6% 1.65+1.95−0.73 M⊕ 1.06+0.36−0.24 R⊕ 93.6% 1.65+1.52−0.73 M⊕ 1.10+0.50−0.33 R⊕ 94.6% 1.71+2.13−0.79 M⊕ 81.6% 1.52+1.15−0.62 M⊕ 1.02+0.26−0.27 R⊕ 1.06+0.53−0.30 R⊕ 84.8% 1.64+1.99−0.73 M⊕ 81.1% 1.65+2.13−0.75 M⊕ 1.02+0.63−0.36 R⊕ 6 Case Case A Case B Case C Case D Case E Case F Case G Note-The resulting values of Procky, (cid:104)M(cid:105)rocky, and (cid:104)R(cid:105)rocky for different mass spreads dM and compositional parameters Rthresh and ∆R. The expectation values are reported with 95% confidence intervals. sense. Even if the mass could be further constrained, sub-Neptunes have been measured with masses as low as ∼ 1 M⊕, so the composition cannot be conclusively inferred from the mass alone. Nevertheless, the rocky composition originally asserted by Anglada-Escud´e et al. (2016) remains the most likely possibility. The results reported herein benefited from collab- orations and/or information exchange within NASA's Nexus for Exoplanet System Science (NExSS) research coordination network sponsored by NASA's Science Mis- sion Directorate. We thank Benjamin Rackham and Gijs Mulders for their constructive advice and insights, and the anonymous referee for their comments. REFERENCES Anglada-Escud´e, G., Amado, P. J., Barnes, J., et al. 2016, Lovis, C., Snellen, I., Mouillet, D., et al. 2016, ArXiv e-prints, Nature, 536, 437 Barnes, R., Deitrick, R., Luger, R., et al. 2016, ArXiv e-prints, arXiv:1608.06919 Belikov, R., Bendek, E., Thomas, S., Males, J., & Lozi, J. 2015, in Proc. SPIE, Vol. 9605, Techniques and Instrumentation for Detection of Exoplanets VII, 960517 Brugger, B., Mousis, O., Deleuil, M., & Lunine, J. I. 2016, ApJL, 831, L16 Coleman, G. A. L., Nelson, R. P., Paardekooper, S. J., et al. 2017, MNRAS, Goldblatt, C. 2016, ArXiv e-prints, arXiv:1608.07263 Ho, S., & Turner, E. L. 2011, ApJ, 739, 26 Kervella, P., Th´evenin, F., & Lovis, C. 2017, A&A, 598, L7 arXiv:1609.03082 Luger, R., Lustig-Yaeger, J., Fleming, D. P., et al. 2016, ArXiv e-prints, arXiv:1609.09075 Mulders, G. D., Pascucci, I., & Apai, D. 2015, ApJ, 814, 130 Owen, J. E., & Mohanty, S. 2016, MNRAS, 459, 4088 Ribas, I., Bolmont, E., Selsis, F., et al. 2016, A&A, 596, A111 Rogers, L. A. 2015, ApJ, 801, 41 Schwieterman, E. W., Meadows, V. S., Domagal-Goldman, S. D., et al. 2016, ApJL, 819, L13 Turbet, M., Leconte, J., Selsis, F., et al. 2016, A&A, 596, A112 Weiss, L. M., & Marcy, G. W. 2014, ApJL, 783, L6 Wolfgang, A., Rogers, L. A., & Ford, E. B. 2016, ApJ, 825, 19 Zahnle, K. J., & Catling, D. C. 2013, Lunar and Planetary Science Conference, 44, 2787
1609.05053	1	1609	2016-09-16T13:53:59	First limits on the occurrence rate of short-period planets orbiting brown dwarfs	[ "astro-ph.EP", "astro-ph.SR" ]	Planet formation theories predict a large but still undetected population of short-period terrestrial planets orbiting brown dwarfs. Should specimens of this population be discovered transiting relatively bright and nearby brown dwarfs, the Jupiter-size and the low luminosity of their hosts would make them exquisite targets for detailed atmospheric characterisation with JWST and future ground-based facilities. The eventual discovery and detailed study of a significant sample of transiting terrestrial planets orbiting nearby brown dwarfs could prove to be useful not only for comparative exoplanetology but also for astrobiology, by bringing us key information on the physical requirements and timescale for the emergence of life. In this context, we present a search for transit-signals in archival time-series photometry acquired by the Spitzer Space Telescope for a sample of 44 nearby brown dwarfs. While these 44 targets were not particularly selected for their brightness, the high precision of their Spitzer light curves allows us to reach sensitivities below Earth-sized planets for 75% of the sample and down to Europa-sized planets on the brighter targets. We could not identify any unambiguous planetary signal. Instead, we could compute the first limits on the presence of planets on close-in orbits. We find that within a 1.28 day orbit, the occurrence rate of planets with a radius between 0.75 and 3.25 R$_\oplus$ is {\eta} < 67 $\pm$ 1%. For planets with radii between 0.75 and 1.25 R$_\oplus$, we place a 95% confident upper limit of {\eta} < 87 $\pm$ 3%. If we assume an occurrence rate of {\eta} = 27% for these planets with radii between 0.75 and 1.25 R$_\oplus$, as the discoveries of the Kepler-42b and TRAPPIST-1b systems would suggest, we estimate that 175 brown dwarfs need to be monitored in order to guarantee (95%) at least one detection.	astro-ph.EP	astro-ph	Mon. Not. R. Astron. Soc. 000, 1–11 (2015) Printed 19 September 2016 (MN LATEX style file v2.2) First limits on the occurrence rate of short-period planets orbiting brown dwarfs. Matthias Y. He1,2(cid:63), Amaury H.M.J. Triaud3,2,1, Michaël Gillon4 1Department of Astronomy & Astrophysics, University of Toronto, Toronto, Ontario, M5S 3H4, Canada 2Centre for Planetary Sciences, University of Toronto at Scarborough, 1265 Military Trail, Toronto, Ontario, M1C 1A4, Canada 3Institute of Astronomy, University of Cambridge, Madingley Road, Cambridge, CB3 0H4, UK 4Institut d'Astrophysique et de Géophysique, Université de Liège, Allée du 6 Août 17, Sart Tilman, 4000 Liège 1, Belgium Accepted ?. Received ?; in original form ? ABSTRACT Planet formation theories predict a large but still undetected population of short-period ter- restrial planets orbiting brown dwarfs. Should specimens of this population be discovered transiting relatively bright and nearby brown dwarfs, the Jupiter-size and the low luminosity of their hosts would make them exquisite targets for detailed atmospheric characterisation with JWST and future ground-based facilities. The eventual discovery and detailed study of a significant sample of transiting terrestrial planets orbiting nearby brown dwarfs could prove to be useful not only for comparative exoplanetology but also for astrobiology, by bringing us key information on the physical requirements and timescale for the emergence of life. In this context, we present a search for transit-signals in archival time-series photometry acquired by the Spitzer Space Telescope for a sample of 44 nearby brown dwarfs. While these 44 targets were not particularly selected for their brightness, the high precision of their Spitzer light curves allows us to reach sensitivities below Earth-sized planets for 75% of the sample and down to Europa-sized planets on the brighter targets. We could not identify any unambiguous planetary signal. Instead, we could compute the first limits on the presence of planets on close-in orbits. We find that within a 1.28 day orbit, the occurrence rate of planets with a radius between 0.75 and 3.25 R⊕ is η < 67 ± 1%. For planets with radii between 0.75 and 1.25 R⊕, we place a 95% confident upper limit of η < 87±3%. If we assume an occurrence rate of η = 27% for these planets with radii between 0.75 and 1.25 R⊕, as the discoveries of the Kepler-42b and TRAPPIST-1b systems would suggest, we estimate that 175 brown dwarfs need to be monitored in order to guarantee (95%) at least one detection. Key words: binaries: eclipsing – brown dwarfs – planets and satellites: detection – tech- niques: photometric 1 FOREWORDS Radial velocity and transit surveys have revealed that close-in packed systems of low-mass planets are very frequent around solar- type stars (Mayor et al. 2011; Howard et al. 2010; Batalha et al. 2013) and red dwarfs (Dressing & Charbonneau 2013). While still undetected, such systems could also be frequent around brown dwarfs. Observations confirm that brown dwarfs possess the same protoplanetary disc fraction as T Tauri stars do (Scholz et al. 2008). In addition Ricci et al. (2012, 2013) found evidence of dust growth, which has also been theoretically explored (Meru, Galvagni & Ol- czak 2013). Planet formation is thus expected to occur within these discs (Payne & Lodato 2007). Despite these encouraging signs, lit- tle evidence about short-period planets orbiting brown dwarfs ex- (cid:63) Contact e-mail: [email protected] c(cid:13) 2015 RAS ist, which is despite some of their very advantageous properties for planet detection, but can be explained by their intrinsic faintness. To this day, only a few long-period planetary mass objects have been associated to brown dwarf hosts using direct imaging and microlensing techniques (e.g. Chauvin et al. 2004). The mass ratios and orbital distances however make all these systems resem- ble substellar binaries more than planetary systems, at the excep- tion of the ∼ 2 MJup object OGLE-2012-BLG-0358Lb discovered by Han et al. (2013). The Carnegie Astrometric Planet Search Pro- gram attempts to detect gas giant planets around late M, L, and T dwarfs, as the smaller masses of these stars yield larger astrometric signals for a given planetary companion (Boss et al. 2009). Another astrometric search for long-period planets around brown dwarfs is also currently under way (Sahlmann et al. 2014) and has demon- strated sensitivity reaching well into the planetary domain. Blake, Charbonneau & White (2010) used the radial velocity technique 2 M. He Figure 1. For a given equilibrium temperature (here 255K, like Earth), the number of orbits (i.e. transits or occultations) per year (black), the transit depth (dashed red), and the probability of transit (dotted blue) as a function of the primary's mass. The green area is approximatively where a 5σ on spectral signatures can be reached with the mission lifetime of JWST. Stellar parameters were obtained from a 1 Gyr isochrone (Baraffe et al. 2003). The slopes steepen with older stellar ages. and produced a null result. A first attempt to detect short-period terrestrial planets transiting nearby brown dwarfs was performed by Blake, Latham & Bloom (2007). They used the PAIRITEL infrared telescope to monitor a sample of 20 ultra-cool dwarfs, including some brown dwarfs. This survey did not detect any transiting ob- ject, and its precision was too low – ∼1 % – with a sample too small to constrain the occurrence of short-period terrestrial plan- ets around brown dwarfs. In 2013, the newly detected nearby bi- nary brown dwarf Luhman-16AB (Luhman 2013) was intensively monitored in photometry by the TRAPPIST telescope (Jehin et al. 2011; Gillon et al. 2011) to search for transiting planets down to the radius of Earth. This project, that failed to detect any transit but revealed the fast-evolving weather of Luhman-16B (Gillon et al. 2013b), was done in the context of a transit survey targeting the ∼50 brightest Southern ultracool dwarfs ongoing since 2010 on TRAP- PIST (Gillon et al. 2013a). This same survey identified recently a trio of Earth-sized planets transiting a nearby star with a mass only ∼10% more massive than the Hydrogen-burning limit (Gillon et al. 2016). This recent discovery combined with the theoretical predic- tion that similar systems should be frequent around brown dwarfs is a strong motivation to intensify the search for transiting planets around the nearest brown dwarfs. In this context, we present here the results of the first space- based search for terrestrial planets transiting brown dwarfs, based on archive data gathered for a sample of 44 brown dwarfs by the Spitzer Space Telescope. In the next section we outline the impor- tance of a search for planets transiting brown dwarfs. In Section 3 we describe the sample we use and in Sect. 4 perform early calcu- lations on what type of planets can be detected. We then present a search algorithm in Section 5, and test it using synthetically in- serted transits. We apply this algorithm to first seek planets within the sample (Sect. 7) and then use it to compute upper limits on the occurence rate (Sect. 8). We then conclude. 2 A CASE FOR FINDING PLANETS TRANSITING BROWN DWARFS We briefly summarise here a case that we presented in a white pa- per (Triaud et al. 2013a) and which we detailed in a number of observing proposals, attempting to monitor brown dwarfs in search for systems of transiting planets. Some of our arguments are sim- ilar to those made in favour of M dwarfs by Nutzman & Char- bonneau (2008), particularly late M dwarfs (Kaltenegger & Traub 2009; Belu et al. 2011; Rodler & López-Morales 2014). Brown dwarfs have characteristics that make them ideal tar- gets to search for Earth-like rocky worlds, but also optimal for their atmospheric characterisation. Two hundred brown dwarfs have near-infrared magnitudes K < 13 (dwarfarchives.org), which are optimal for JWST. As an example, a first attempt to perform trans- mission spectroscopy on the TRAPPIST-1b & 1c planets has been presented by (de Wit et al. 2016) using the Hubble space telescope. Any planets found transiting a brown dwarf will offer similarly good conditions for the JWST, if not more, on account of their smaller radii, which enhance transmission features. Emission and reflection spectroscopy, as well as phase curves, will complement transmission spectroscopy and provide additional information about the atmospheres of any discovered exoplanet (Seager & Deming 2010), as will high-resolution spectroscopy (Collier Cameron et al. 1999; Snellen et al. 2015). A planet the size and temperature of the Earth has the same blackbody emission be it orbiting a G dwarf or a brown dwarf. The latter however is 6 to 10 magnitudes fainter, a very favourable case akin to a natural coronograph. Emission and reflection spectroscopy probe the full face of the planet, with flux emerging through only one airmass, which makes the technique less sensitive to clouds than transmis- sion spectroscopy. In addition, occultations permit the elaboration of thermal maps (de Wit et al. 2012). Measuring the reflected and emission spectra can also be obtained when the system is not in a transiting configuration (Snellen et al. 2015). In some configura- tions a 255 K, Earth-sized planet can have a signal of order that achieved these days for hot Jupiters orbiting by G dwarfs (∼ 1e−5, e.g. Leigh et al. 2003; Brogi et al. 2014). Other advantages are represented graphically in Figure 1, where we show two metrics of detection (probability of transit and transit depth), and one metric of characterisation (number of orbits per year), for an Earth-sized planet, with an equilibrium temper- ature of 255 K (like Earth), as a function of stellar mass. By fo- cusing on brown dwarfs, we gain one order of magnitude on the probability of having a habitable zone planet in a transiting config- uration (Bolmont, Raymond & Leconte 2011). We also improve by c(cid:13) 2015 RAS, MNRAS 000, 1–11 0.11.00.20.5 0 1 2 3 4 5probability of transit (%)mass (Msun) . . . .0.11.00.20.5 0 50 100 150 200 250 300number of orbits per yearmass (Msun) . . . .0.11.00.20.50.00.20.40.60.81.0transi depth (%)mass (Msun) . . . .0.11.00.20.50.00.20.40.60.81.0transi depth (%)mass (Msun) . . . .0.11.00.20.5 0 1 2 3 4 5probability of transit (%)mass (Msun) . . . .transit depth (%)probability of transit (%)Table 1. Examples of typically produced transit depths and transit widths W, with their corresponding signal to noise ratio (in brackets). We assumed a 1R⊕ planet, an impact parameter of b = 0.5, and a point to point rms of σ = 0.49% (taken as the weighted average of both channels for all the objects). M(cid:63) [ MJup ] R(cid:63) [ RJup ] D [ % ] W (P = 0.23d) [min] (SNR) W (P = 0.55d) [min] (SNR) W (P = 1.28d) [min] (SNR) 30 30 30 60 60 60 80 80 80 0.8 0.9 1.0 0.8 0.9 1.0 0.8 0.9 1.0 1.30 1.03 0.83 1.30 1.03 0.83 1.30 1.03 0.83 17.5 (7.5) 19.4 (6.2) 21.3 (5.3) 13.9 (6.7) 15.4 (5.6) 16.9 (4.7) 12.6 (6.4) 14.0 (5.3) 15.3 (4.5) 23.3 (8.7) 25.9 (7.2) 28.4 (6.1) 18.5 (7.7) 20.5 (6.4) 22.6 (5.5) 16.8 (7.4) 18.7 (6.1) 20.5 (5.2) 30.9 (10.0) 34.3 (8.3) 37.7 (7.1) 24.6 (8.9) 27.2 (7.4) 29.9 (6.3) 22.3 (8.5) 24.7 (7.1) 27.2 (6.0) two orders of magnitude the depth of the transit, easing detection, and reaching a comfortable 1% transit depth, which is routinely detected by many ground-based instruments (Pollacco et al. 2006; Bakos et al. 2007; Gillon et al. 2013b, 2016). Finally, for a true Earth analog orbiting a solar analog, there is only one transit per year. If 15 transits need to be co-added to reach a significant detec- tion of atmospheric features, 15 years of observations are required. For brown dwarfs, we can collect in excess of one transit, and one occultation, every week. Belu et al. (2011) point out that a formal detection of atmo- spheric features can only be made during the limited mission life- time of JWST, if that planet orbits a star with a mass inferior to 0.2 M(cid:12). This is assuming that all transit events will be recorded. Due to scheduling concerns, it is more likely closer to 0.15 M(cid:12). Brown dwarfs form a significant fraction of the remaining systems. Any planet (rocky or not, habitable or not), found to be transiting a nearby brown dwarf could make an exquisite target for a detailed atmospheric characterisation. 3 OUR SAMPLE The only sizeable sample of brown dwarfs that has been monitored photometrically in a sufficiently precise and consistent manner, has been presented by Metchev et al. (2015). They acquired their data using the Spitzer Space Telescope (in its warm phase), as part of an Exploration Science program called Weather on Other Worlds (see also Heinze et al. 2013). Their aim was to study the mid-infrared variability of brown dwarfs across the L to T spectral class tran- sition. 44 brown dwarfs ranging from L3 to T8 were monitored using IRAC (Fazio et al. 2004), in both channel 1 and 2 (3.6 µm and 4.5 µm respectively, or [3.6] & [4.5]). The main results are presented in Metchev et al. (2015). Out of 44 brown dwarfs, 23 exhibit constant photometric timeseries while the other 21 display significant variability. Variability is produced by any inhomogene- ity on the surface of brown dwarfs that moves in and out of view due to rotation. Metchev et al. (2015) provide arguments imply- ing that all brown dwarfs may in fact be variable. Those that have their rotation axis aligned with the line of sight, present a similar hemisphere at any given time, and produce constant timeseries. The physical origin of this variability remains debated but is now fre- quently interpreted as being produced by clouds (e.g. Ackerman & Marley 2001; Burgasser et al. 2002; Artigau et al. 2009; Radigan et al. 2012; Gillon et al. 2013b; Radigan et al. 2014; Crossfield et al. 2014). c(cid:13) 2015 RAS, MNRAS 000, 1–11 Brown dwarf planets 3 For our analysis, we used the data already detrended from ro- tational variability that is presented in Metchev et al. (2015). In or- der to study variability and reach their conclusions, Metchev et al. (2015) fit Fourier components to their data. Most often, one term is clearly significant. When several alternate models are proposed, we adopt the model according to the following criteria: we resorted to the lowest number of Fourier terms unless each extra term im- proved the point to point rms by more than 10%, or significant vari- ability could still be seen after correction using the lowest number of terms. We now have a set of 44 photometric timeseries that has been corrected for both instrumental and astrophysical effects. We search for transits into each lightcurve and use the set to place the first constraints on the presence of planets orbiting brown dwarfs. The median observation time per brown dwarf is 20.9 hours, over both channels. Typically, observations start with the 3.6 µm channel, observing for 13.8h (exceptions are 2M0036, 2M1632, and DENIS1058, which only have about 7-8h) and then proceed to 4.5 µm for 6.8h (except 2M2224, 2M0825, and 2M1507, which have 8-9h of data). There is a small gap while the spacecraft moves from channel 1 to channel 2. The median gap between is 15 min- utes, although 2M0949 and 2M1511 have gaps that are over an hour long (1.9h and 1.4h respectively). All targets, in both channels, ex- hibit a similar cadence, with new frames acquired every 2.18 min ± 0.07 min (the error is taken as standard deviation of the gaps in time between consecutive data points, ignoring the larger gap between the two channels). Statistically, the two photometric chan- nels produce a similar precision: on average, the flux at [3.6] has a point to point rms of 0.0050 ± 0.0037 while [4.5] has an rms of 0.0046 ± 0.0024 (the errors are taken as the standard deviations of the rms over all the objects, per channel). We plot their distribution in Fig. 2. We note a difference between the rms of the objects with con- stant light curves and that of the variable objects. Constant time- series have a higher dispersion than variable targets. On average constant timeseries have an rms = 0.0060 ± 0.0032 while this value is 0.0035 ± 0.0023 for those light curves with variability removed. We do not elaborate much on this topic, but remark that this may suggest that the Fourier treatment overfits the data. Alternatively, it is possible that some constant sources are in fact variable, but that their variability is too weak to be significant, or that it is not coherent in time (e.g. Artigau et al. 2009; Gillon et al. 2013b). 4 EXPECTED SIGNAL The important variables used to analyse planetary transits are described by Mandel & Agol (2002); Seager & Mallén-Ornelas (2003) and Winn (2010). The mass of the brown dwarf, the orbital period of the planet, the relative size of the brown dwarf and the planet, as well as the orbital inclination on the sky, essentially pro- duce two observables, the depth D, and the width W. The depth is the ratio of areas squared, D = R2 (cid:63), for Rp, the planet's radius, and R(cid:63), a brown dwarf's. p/R2 We assume a fiducial brown dwarf for the remainder of the paper. Its mass is M(cid:63) = 60MJup and its size is R(cid:63) = 0.9RJup. Brown dwarfs > 1 Gyr all have approximately this size (e.g. Baraffe et al. 2003; Triaud et al. 2013b; Moutou et al. 2013; Díaz et al. 2013; Di- eterich et al. 2014), which eases greatly how we sample parameter space. Further justifications for these numbers will come through the paper. However what we assume for the brown dwarf does not affect our conclusions much. Furthermore, our retrieval only uses D and W as observables, which are independent of what we assume 4 M. He Figure 2. Histogram and cumulative distributions of the rms of the Spitzer photometry on our sample of 44 brown dwarfs, in both channels. Vertical lines translate the rms into detectable planetary radii, assuming transits with a signal to noise ratio of at least 3.5. Figure 3. A few examples of our ability to retrieve planetary transits that we added to the Spitzer lightcurves (locations marked with a small vertical line). Left, we have four brown dwarfs' lightcurves, as observed with Spitzer showing artificially implanted planetary transits. Right, we show the depth detected by our top-hat model, as a function of time for each of this timeseries. We also display the mean and three times the rms above the mean, with solid blue lines. Large red dots with accompanying vertical red lines indicate where transit-like signals are identified. From top to bottom: (a) Lightcurve of SDSS 2052 containing a correct retrieval of an inserted transit produced by a planet of 1.75 R⊕. (b) This is 2MASS 0516, with a 1.41 R⊕ planet transiting twice; both instances are correctly recovered, but a third, false positive signal too (indicated by the red stripe). (c) We have two simulated transits of a 1.0 R⊕ planet on the time series of SDSS 1520, both of which were easily retrieved. (d) Shows 2MASS 2148, where two transits were artificially inserted, to simulate a 0.59 R⊕ planet on a short orbit; we identify only one of the transits, failing to recover the second, a false negative. We think the second event is missed by the search algorithm due to it being barely at the very end of the time series. for the central object. In Table 1, we produce a series of typical depth and width values, for a few assumed brown dwarf parame- ters. Since the cadence of observation is the same for each series, we will also frequently refer to the width as a number of points in transit, or n. We now assess how good the Spitzer lightcurves are in terms of what type of planets they are be able to yield. For this purpose, we define a signal to noise ratio (SNR) such that: √ S NR = n D σ (1) where n is the number of data points that fall within transit, D is the transit depth, our signal, and σ is the point to point rms of the c(cid:13) 2015 RAS, MNRAS 000, 1–11 02468101214Frequency0.25 Rearth0.75 Rearth1.25 Rearth0.0000.0050.0100.0150.020Channel 1 rms0.00.20.40.60.81.0Cumulative fraction0.25 Rearth0.75 Rearth1.25 Rearth024681012140.25 Rearth0.75 Rearth1.25 Rearth0.0000.0050.0100.0150.020Channel 2 rms0.00.20.40.60.81.00.25 Rearth0.75 Rearth1.25 Rearth5101520Time (h)0.9641.015Normalized flux5101520Time (h)0.9671.032Normalized flux5101520Time (h)0.9821.010Normalized flux5101520Time (h)0.9951.004Normalized flux05101520Time (h)0.02880.00000.0288DepthW=15.3(min)05101520Time (h)0.02390.00000.0239DepthW=15.3(min)05101520Time (h)0.01320.00000.0132DepthW=10.9(min)05101520Time (h)0.00320.00000.0032DepthW=10.9(min)(a) 2M2254 (b) 2M0103 Figure 4. Examples of our recovery rate for two brown dwarfs. Detections and non-detections are plotted in black and red, respectively. Top, we plot how systems distribute in bins of signal to noise ratios. Bottom, we draw the respective fraction of recovered to non recovered for each of the bins. For clarity, only ten bins were used for each histogram, and include all the inserted transits that have an SNR less than the highest SNR of the un- detected transits plus one. The SNR corresponding to the 50% and 95% recoverability rates (represented by vertical black lines) were calculated by linear interpolation between the bins (grey regions). timeseries. A typical transit (Table 1) has a width W that lasts 22.2 minutes; we chose here n = 10 (meaning a width of 22 minutes). For a set SNR, we can calculate approximately which planetary radii Rp would be detected for a certain rms. For reasons elaborated in Section 5 and shown in Fig 4, we set SNR = 3.5. We draw the histogram and cumulative distributions of the rms of our sample in Fig. 2, and indicate which rms can allow the identification of planets with 0.25, 0.75 and 1.25 R⊕. For roughly 75% of the sam- ple, we are sensitive to planets with Rp down to 0.75 R⊕, which is remarkable. 5 RETRIEVAL OF ARTIFICIALLY INSERTED TRANSITS The Spitzer timeseries that we have at our disposal are relatively short (20.9 hours). This means that we need a detection algorithm able to retrieve single transit-like signals. Because our sample is also small, we can allow ourselves fairly loose detection criteria. The systems that are picked up can be analysed in detail and could lead, in a second stage, to follow-up observations whose aim would be the confirmation of a transit signal, the measure of the planet's period, as well as possibly, evidence for the presence of other plan- etary companions. We insert and retrieve planetary transits into our light series by using a simplified model. A transiting planet produces a box-like event, which can be approximated with a top-hat model, essentially consisting of a vertical drop in flux, of a depth D and for a duration (or width), W. These observable parameters can then be converted into physical quantities. For simplicity, we will neglect any limb darkening effect (which are irrelevant in the mid-infrared (Claret & c(cid:13) 2015 RAS, MNRAS 000, 1–11 Brown dwarf planets 5 Table 2. Signal to noise ratios for the time series of each object correspond- ing to a 50% and 95% chance of transit retrieval, as well as the maximum SNR out of all assumed transit widths for each object. Their respective planet radii were calculated from adopting an orbital period equal to the av- erage coverage of the time series (20.9h) and an impact parameter b = 0.5. The maximum SNR marked by * correspond to signals that are both de- tected by our 3 rms criterion and over the 50% SNR threshold. Object SNR Rp for 50% (R⊕) SNR Rp for 95% (R⊕) Max. SNR 2MASS 0036 2MASS 0050 2MASS 0103 SDSS 0107 SDSS 0151 2MASS 0328 2MASS 0421 2MASS 0516 2MASS 0820 2MASS 0825 SDSS 0858 2MASS 0949 SDSS 1043 DENIS 1058 2MASS 1059 2MASS 1122 2MASS 1126 SDSS 1150 2MASS 1209 SDSS 1254 ROSS 458C 2MASS 1324 ULAS 1416 SDSS 1416 2MASS 1507 SDSS 1511 SDSS 1516 SDSS 1520 SDSS 1545 2MASS 1615 2MASS 1632 2MASS 1721 2MASS 1726 2MASS 2224 2MASS 1753 2MASS 1821 SDSS 2043 SDSS 2052 HN Peg B 2MASS 2148 2MASS 2208 2MASS 2228 SDSS 2249 2MASS 2254 5.4 3.6 3.7 3.5 3.5 3.0 3.5 4.1 3.3 3.6 3.1 3.0 4.8 4.0 2.9 3.1 3.5 3.2 2.8 3.2 4.1 3.3 4.3 3.7 3.9 3.2 2.8 3.8 3.9 3.0 4.7 2.6 3.3 3.4 3.3 5.4 3.5 3.3 2.9 3.4 3.0 4.2 3.0 3.0 0.33 0.85 0.48 0.42 0.69 0.57 0.42 0.86 0.52 0.37 0.55 0.68 0.65 0.40 0.85 0.48 0.38 0.55 0.59 0.44 1.04 0.44 0.81 0.34 0.31 0.51 0.51 0.53 0.75 0.44 0.57 0.31 0.46 0.32 0.33 0.35 0.72 0.59 0.56 0.27 0.46 0.77 0.48 0.57 8.9 5.8 6.8 6.4 5.5 6.7 7.4 5.1 4.7 6.1 6.0 5.6 8.8 6.8 4.6 5.8 6.3 6.1 5.2 4.6 6.8 4.9 6.4 6.7 7.4 6.6 4.7 6.1 6.2 7.9 8.5 4.3 6.8 5.5 6.5 8.0 4.8 6.4 6.0 7.0 4.5 6.3 5.6 6.1 0.43 1.08 0.65 0.57 0.87 0.85 0.60 0.95 0.63 0.48 0.76 0.92 0.89 0.51 1.06 0.66 0.51 0.76 0.79 0.53 1.34 0.54 0.99 0.45 0.43 0.73 0.67 0.67 0.96 0.71 0.76 0.40 0.66 0.41 0.46 0.43 0.85 0.82 0.80 0.39 0.57 0.93 0.66 0.80 4.7 3.4 3.7 3.7 3.3 4.0* 3.1 3.3 3.1 3.0 2.5 3.1 5.0 2.9 3.0 3.2 3.1 3.1 3.3* 3.0 3.4 3.0 3.4 3.7 3.8 2.7 4.0* 4.1* 3.9 2.9 3.6 3.5 3.1 3.1 2.9 4.8 4.4 2.9 2.8 3.3 3.7* 4.8* 2.5 3.3 Bloemen 2011), and unknown for brown dwarfs). We also neglect the typically brief ingress and egress durations, as is often done for detection algorithms (Kovács, Zucker & Mazeh 2002; Collier Cameron et al. 2007). We search for transits in the following way. We start the top- hat model such that the first point is at the beginning of the time- series and assume a certain width. We compute the average flux within transit, and normalise the flux outside of the transit. The depth is the difference between both values. We then step the model by one point forward and repeat the procedure until we reach the 01234567061FrequencyDetectionsNon-detections01234567SNR bins01Probability012345678036FrequencyDetectionsNon-detections012345678SNR bins01Probability6 M. He (a) 2M0328 (b) 2M1209 (c) SDSS1520 (d) 2M1516 (e) 2M2208 (f) 2M2228 Figure 5. Six lightcurves showing candidate transit-like events. Below each of the timeseries we plot the step by step depth found by the top-hat in black, for two two assumed width: n = 5 (top) and n = 7 (bottom). Other widths are not shown here. The red horizontal line indicate the three rms threshold, above which we count a detection. The blue horizontal line correspond to depths that have a depth producing an empirically determined SNR with a 50% confidence for detection (i.e., derived from Figure 4). end of the timeseries. This results in a temporal series of depth measurements. We calculate the mean and the rms of this series. Our criterion for a possible transit detection is when the transit depth reaches a value larger than three times the rms, above the mean depth value. The entire procedure is repeated, assuming six different values for the width (5, 7, 10, 13, 16, and 20 points wide, corresponding to the range of actual transit durations in Table 1). Figure 3 shows examples of a few light curves that have in- serted transits and through which we ran the top-hat model, to iden- tify interesting locations. We provide a range of scenarios: true de- tection, false positive, and false negative. We also display how the depth varies as a function of temporal location, for the width yield- ing the clearest signal. We show that we can identify individual events, which in some cases would have allowed us to identify both transit signals caused by a planet orbiting in less than 20.9 hours (or equivalently, that of two planets, each producing one event). Although our procedure recovers D and W only, we insert transits using a range of physical parameters. This range will later be used to estimate our limits on the occurrence rate of planets or- biting brown dwarfs, described in Section 8. The orbital period P, its inclination from the plane of the sky ip, the transit phase φ0, and the planet radius Rp were randomly sampled. The orbital period was sampled linearly in log from about 0.23 to 3 days, where the lower limit was derived from assum- ing a minimum orbital separation of two times the Roche limit, aRoche = 1.26R(cid:63)(ρ(cid:63)/ρp)1/3 where we set the planet density ρp equal to the mean density of Earth by adopting ρ⊕ = 5.51 g cm−3. We drew the inclination angle sampling uniformly in cos ip with ip = 0◦ to 90◦. φ0 was sampled linearly between 0 to 1. For our purposes of investigating Earth-sized planets, we also sampled planetary radii uniformly, between 0.25 and 3.25 R⊕. Every time a set of parameters is drawn we verify the planet's impact parameter and select only those that are in a transiting configuration. We stop when we reach 50 000 transiting systems (which is on average 8.6% of the total number of systems). We then randomly assign these systems to one of the 44 Spitzer lightcurves, and insert the transits where they fall by subtracting the depth D for a number of points equivalent to the width W (for the edge points of the transits, we scale the depth linearly by the fraction of the cadence that the in-transit duration covers). Finally we run our de- tection algorithm. A large fraction of the 50 000 synthetic planets have a phase such that a transit happens during the Spitzer obser- vations. Most of them are then identified by our search algorithm, with the detection rate falling quickly for systems with large or- bital periods or very small planetary radii. We grouped the 50 000 planets into three categories: • the inserted transit falls outside the lightcurve; c(cid:13) 2015 RAS, MNRAS 000, 1–11 0.9791.021Flux0.00000.00000.00950.0095Depth05101520Time (h)0.00000.00000.00870.00870.9801.018Flux0.00000.00000.00880.0088Depth05101520Time (h)0.00000.00000.00870.00870.9881.010Flux0.00000.00000.00550.0055Depth05101520Time (h)0.0000.0000.0050.0050.9811.016Flux0.00000.00000.01040.0104Depth05101520Time (h)0.00000.00000.00710.00710.9881.013Flux0.0000.0000.0060.006Depth05101520Time (h)0.00000.00000.00580.00580.9731.033Flux0.00000.00000.01410.0141Depth05101520Time (h)0.00000.00000.01360.0136• the inserted transit • the inserted transit is properly recovered. lightcurve); is not recovered (but falls inside the 13.7% fall in the first category, 11.5% are not detected, and 74.8% are properly detected. However, our algorithm gave a 37% rate of falsely detected systems. A fraction of this value arises from six transit-like candidates in the original 44 Spitzer lightcurves (see Sec. 7). We plot the number of detections and non-detections into Fig- ure 4 as a function of the signal to noise of the signal that had been inserted, as defined in eq. 1. For each Spitzer timeseries, we find the signal to noise ratios for which we have a 50% and a 95% probabil- ity to recover a correct transit. Those values are recorded in Table 2 where we also include what planet radii they correspond to (assum- ing an impact parameter of b = 0.5 and an orbital period equal to the average coverage of the timeseries, P = 20.9h, which translates to a width of about W = 28 min, or n = 13). The 50% recovery rate corresponds to an average signal to noise of 3.5, supplying support for the value we provided in Section 4 and Fig. 2. We adopt this number as verification criterion for identifying transit-like signals from the data. 6 RETRIEVAL OF KEPLER-42B AND TRAPPIST-1B We take this paper as an occasion to compute an order of magnitude occurence rate for Earth-sized planets transiting objects at the bot- tom of the main sequence. We will later extrapolate that number to brown dwarfs and compute approximate yield estimates. We looked into the Kepler Input Catalog, and removed giant stars by applying colour-cuts following Dressing & Charbonneau (2013) (J-K < 1.0 and Kepmag > 14). We then restricted ourselves to objects with Kepmag < 17.5, with r(cid:48) − K > 4.34, a colour compatible with ultra- cool dwarfs (Baraffe et al. 1998, M/H = 0, age = 8 Gyr). We obtain 107 stars, two of which have confirmed planets: Kepler-42 (Muir- head et al. 2012) and Kepler-445 (Muirhead et al. 2015). These studies collected additional data that indicate that both systems or- bit stars that are not ultra-cool dwarfs, but mid-M dwarfs instead. It is likely that the other members in the sample suffer similarly. Nevertheless, we keep this sample as the coldest within the Kepler data and therefore the most representative to the systems we seek. Kepler-445's innermost planet would have less than 30% chance of producing one transit within the duration of our Spitzer timeseries, so we conservatively discard it for our estimate. Kepler- 42b has a 7% geometric probability of transit. TRAPPIST-1b has a 5% transit probability and the system was identified amongst 50 who have not all been fully monitored yet. The successful detection of these two transiting systems amongst a total of 157 surveyed combined with a mean chance of transit of 6% suggests that the occurrence rate of Earth-like planets is likely on the order of 22%, within a two-day orbital period. We performed a slightly more advanced analysis to confirm this number. We assumed a period distribution linear in log between the orbital periods of 0.55 and 6 days (so as to englobe most of the planets contained in the TRAPPIST-1, Kepler-42 and Kepler-445 systems, and ignoring the period box where we are the most sen- sitive). We modelled multiple sample of 157 stars, whose masses were randomly drawn from a uniform distribution in mass bounded by 80 and 150 M(cid:12). We then computed which occurence rate was necessary to obtain the identification of two systems, using Poisson statistics. This yielded η = 27% (including Kepler-445, this num- c(cid:13) 2015 RAS, MNRAS 000, 1–11 Brown dwarf planets 7 Figure 6. Close up view on the transit-shaped shaped signal for the six candidates that passed our selection criteria. ber rises to 45%). We therefore conclude that the occurence is in the region of 20-30%. We extend our simulations to investigate whether we could detect Kepler-42 (Muirhead et al. 2012) and TRAPPIST-1 (Gillon et al. 2016). Both systems involve three Earth-sized planets at vary- ing orbital periods: • Kepler-42: radii of 0.73, 0.78, 0.57 R⊕ with periods of 0.453, • TRAPPIST-1: radii of 1.11, 1.05, 1.16 R⊕ with periods of 1.51, 1.1214, 1.865 days respectively 2.42, 18.202 days respectively In order to determine the suitability of Spitzer's photometry to detect these systems, we only considered the innermost planet of these systems (the planets with periods of 0.453 days and 1.51 days for Kepler-42b and TRAPPIST-1b respectively, i.e. 0.73 R⊕ and 1.11 R⊕), and simulate transit retrieval per each of the 44 lightcurves. For both planets, we sampled all other physical pa- rameters (inclination ip and phase φ0) as described earlier. We still assume a M(cid:63) = 60MJup and R(cid:63) = 0.9RJup brown dwarf. We then ran the simulated data through our retrieval algorithm. Our results indicate that for Kepler-42b, 10/44 lightcurves yield > 95% detection rate with none of them giving > 99% de- tection. For TRAPPIST-1b, 33/44 lightcurves give > 95% detec- tion rate with 15/44 yielding > 99% retrieval, results which are markedly better due to the planet being larger. To summarise, 75% of our 44 Spitzer series provide a 95% confidence for detecting a 1.11 R⊕ planet with a period of 1.51 days, indicating a remarkable sensitivity to transiting Earth-sized planets. 1.0210.9791.0180.9801.0160.9811.0100.9881.0330.9731.0130.988Time [hours]Time [hours]FluxFluxFluxFluxFluxFlux2M15162M22082M22282M03282M1209DSS15208 M. He Table 3. Transit candidates retrieved using our 3 rms depth criterion that have SNR greater than the 50% threshold (i.e., corresponding to the objects in Figure 5). The transit epoch corresponds to the front-end position of our top-hat model that gives detection. For each object, the width W corresponding to the maximum SNR (as shown in the third column) out of the six widths we tried is presented, along with the respective depth D for that width. The radii Rp of the potential planets are calculated from the depths assuming R(cid:63) = 0.9 RJup. We also show the magnitudes of the targets in 2MASS J and K bands, and in the Wise 1 and 2 channels, which are similar to Spitzer's. Object Transit epoch [h] Max. SNR W [min] 2MASS 0328 2MASS 1209 2MASS 1516 SDSS 1520 2MASS 2208 2MASS 2228 1.6 4.3 5.5 4.0 3.4 12.0 4.0 3.2 4.0 4.1 3.7 4.3 21.8 15.3 10.9 43.6 15.3 15.3 D [%] 0.77 0.87 1.04 0.36 0.58 1.36 Rp [R⊕] 0.96 1.02 1.12 0.66 0.84 1.28 J K [mag] [mag] 16.69 15.91 16.85 15.54 15.80 15.66 14.87 15.06 15.08 14.00 14.15 15.30 W1 [mag] 14.13 14.66 14.14 13.44 13.38 15.23 W2 [mag] 13.60 13.48 13.40 12.92 12.91 13.33 7 SEARCH FOR TRANSIT-LIKE SIGNALS We search each of the Spitzer lightcurves for transit-like signals, us- ing the retrieval procedure described above. Despite Spitzer's pre- cise photometry, we detect no obvious candidate for transiting ex- oplanets, only recovering a number of doubtful events. In 20 cases we obtained identified locations where the top-hat indicates a lower flux for at least one of our six model widths. In 14 systems, the low- est depths recovered have a SNR high enough for a 50% likelihood that the event is real (as produced in Table 2), although only a num- ber of them are detected according to our criterion. The overlap includes 6 candidates which have a signal that is both over the 3 rms depth threshold and over the 50% SNR threshold. We present these candidates in Figure 5. None of these events repeat, implying periods larger than 21 hours. None of these reaches a 95% confi- dence detection and are therefore likely false positives. In Table 3 we present the details of these detected best candidates. The transit epoch is displayed as the time of the beginning of the transit model which gives the detection. Since we ran six different top-hat model widths through each light curve and thus in some of these cases the same transit-like signal is detected by different widths, in order to choose the one model (width W and corresponding depth D) we opted for the one yielding the highest SNR. 8 ESTIMATES OF MAXIMUM OCCURRENCE RATE We use the population of 50 000 synthetic transiting planets that we described in Sec. 5, to measure upper limits on the occurence rate of planets orbiting brown dwarfs. We divide the planet radius and orbital period domain into a grid of 6 rows (equally spaced in Rp from 0.25 to 3.25 R⊕) and 3 columns (equally spaced in log P, from P = 0.23 to 3 days), consisting of 18 boxes. Inside each box we compute: • the mean transit probability, • the fraction of transits missed (not in the light curve), • the fraction of properly recovered transits. • the fraction of false positives. The results are plotted in Figure 7, for a subset of the total popula- tion to avoid over-crowding (10% of the total). We then proceed to estimate the occurrence rate, which we denote as η. By setting η = 1 and multiplying it by the transit prob- ability p, and the fraction of properly recovered transits f , we can get a rapid idea of how constraining the observations are. We place those numbers in Fig. 8 (bottom right of each box). For instance, in the top-left box, 6.6 planets would be expected. Since we have no detection, we can only estimate upper limits in each box, and for a combination of boxes. Each of the values that we provide are 95% confident. η is calculated by considering the probability of detect- ing a transit given a random brown dwarf and demanding that the chance of having 44 observations with no detections in the box is 5%. Eq. 2 encapsulates this calculation by rearranging to solve for η: 0.05 = (1 − η f p)n (2) where f is the fraction of properly recovered transits, p is the prob- ability of transits, and n is the number of targets in our sample (44). We gather the results in Fig. 8. In our example of the top-left box, we find η < 43 ± 1%, which is the tightest limit using this portion of parameter space. The uncertainty on η is derived by dividing the 50 000 transits into ten equal sets, repeating the calculations, and computing the rms of the ten trials. Note that η > 100% means that the brown dwarfs would have more than one planet each on average in a particular box. Several trends are evident from the values presented in Figure 8. The sharp increase in η for the bottom row (planets from 0.25 to 0.75 R⊕) is mostly caused by the fact that such small planets are extremely hard to detect, even for the relatively precise pho- tometry obtained of [3.6] and [4.5]. The probability of transit also slightly reduces with planet radius; smaller planets are slightly less likely to transit for a given orbital inclination. This creates a drop in p by a few percentage points where otherwise detection frac- tions are nearly 100%. The very large values of η for the longer orbital periods reflects two important effects: the orbital period ex- ceeds the timeseries' coverage (thus there is an increased chance that the transit is not captured), and the geometric probability of transit plummets with orbital period. We will now combine different regions of parameter space and produce a number of scenarios. We also compute how many brown dwarfs need to be surveyed in order to have a 95% chance of ob- taining a detection. In this exercise we assume that new targets will follow the distributions described in Fig. 2, when in fact one could bias the sample to earlier spectral types and brighter targets. 8.1 Between 0.75 and 3.25 R⊕ We estimate the overall η for planets with sizes between 0.75 and 3.25 R⊕ over all three period boxes. Our 95% confident upper limit is η < 92 ± 2%. Restricting ourselves to the two columns at shorter periods, we obtain η < 67 ± 1%. If in reverse, there are no planets c(cid:13) 2015 RAS, MNRAS 000, 1–11 Brown dwarf planets 9 Figure 7. Left: Map of transiting planets versus planet radius and orbital period. Blue filled in dots represent transiting planets that are detected, hollow red circles represent transiting planets that are captured on the light curve but are not detected, and red x's represent transiting planets that are missed by the time series. Right: The corresponding transit fractions and detection statistics per box: the top right corner indicates the probability of transit, while the three percentages on the left side of each box represent, from top to bottom, the fraction of transits missed by the observation (red x's), the fraction of transits undetected using our methods (hollow red circles), and the fraction of detected transits (blue dots). The percentage in parenthesis in the bottom right corner of each box represent the rate of false positives. 20.9 hour monitoring, and 365 targets if we instead observe for 42 hours. If we assume no planets for the inner period box and a 10% rate for the two outer boxes, those numbers rise to 717 and 542 respectively. 8.2 Earth-sized planets We estimate η for planets with sizes between 0.75 and 1.25 R⊕, likely to be rocky (Rogers 2015), for which we also retain a good rate of recoverability. On the ensemble of three period boxes, we place a 95% confident upper limit of η < 120 ± 4%. Restricting ourself to the inner two boxes, we find η < 87 ± 3%. Assuming η = 10% spread over the three boxes, we expect detectable Earth-like transiting planets for 0.63% of brown dwarfs, over a 20.9h timeseries. This value increases to 0.72%, assuming we extend the photometric coverage to 42 hours. For a 20.9h and a 42h monitoring per brown dwarf, we would need to survey 476 and 417 systems respectively, in order to have a 95% confidence for a successful detection. If we adopt the results of the Kepler-42b and TRAPPIST-1b systems and set η = 27% for the three Earth-sized boxes as derived in a previous section, the number of systems we need to survey for a 95% chance of detecting one planet drops to just 175 (of which 44 have already been monitored), given a 20.9h monitoring. Doubling the observation time to 42h monitoring reduces this number to 153 brown dwarfs, with brightness, and variability similar to the WoW sample. 9 DISCUSSION We have searched for transiting planets signals within photometric timeseries obtained on 44 brown dwarfs by the Weather on other Worlds programme (Metchev et al. 2015), that used the Spitzer Figure 8. Info-graphic showing the fractions of brown dwarfs that would have a planet (in black, center of each box), in order to be consistent with having no detections for 44 observations, 95% of the time. The top right corner of each box gives the probability of transit (top-most) and detection chance of those transits as derived from sampling. The bottom right value gives the number of detections we would expect for η = 1, of our 44 brown dwarfs. in the inner period column, we get η < 159 ± 4% for just the two outer columns, and η < 102 ± 3% simply for the central column. Now, we assume there is a flat η = 10%. In order to make one detection (with 95% confidence), we need to observe 413 targets with brightness and variability similar to our sample of 44, for a c(cid:13) 2015 RAS, MNRAS 000, 1–11 0.51.01.52.02.53.0Planet radius (Earth radii)0.51.01.52.02.53.012.85%0.24%69.49%30.27%(55.2%)6.25%12.37%63.92%23.71%(63.49%)4.36%56.12%31.65%12.23%(80.46%)12.17%0.26%13.27%86.48%(33.53%)7.94%15.0%14.23%70.77%(38.26%)4.08%60.77%7.69%31.54%(56.84%)12.81%0.0%1.19%98.81%(34.18%)7.08%16.52%2.23%81.25%(35.69%)3.93%51.56%0.78%47.66%(51.2%)13.29%0.0%0.0%100.0%(33.75%)8.22%13.38%0.37%86.25%(30.75%)4.41%57.24%0.0%42.76%(50.79%)14.58%0.66%0.0%99.34%(33.28%)7.97%11.46%0.0%88.54%(31.5%)4.88%56.96%0.0%43.04%(51.77%)15.15%0.82%0.0%99.18%(30.65%)7.88%13.67%0.0%86.33%(28.48%)5.24%59.65%0.0%40.35%(55.48%)Orbital period (days, log scale)0.230.551.283.0Orbital period (days, log scale)0.230.551.283.00.51.01.52.02.53.0Planet radius (Earth radii)11.49%29.33%1.46200%±20%6.79%22.53%0.63460%±70%3.88%11.35%0.191540%±430%12.13%86.74%4.662%±2%7.12%70.03%2.2130%±10%4.03%35.24%0.6480%±40%13.34%98.49%5.7949%±2%7.5%84.61%2.76104%±4%4.0%43.67%0.81360%±50%13.16%99.65%5.7750%±2%7.67%86.37%2.8999%±5%4.25%43.7%0.83350%±30%14.19%99.77%6.2446%±1%8.26%87.86%3.1691%±5%4.47%43.31%0.85340%±50%15.0%99.61%6.5943%±1%8.44%86.36%3.2189%±5%4.86%43.33%0.89320%±30%Orbital period (days, log scale)0.230.551.283.010 M. He Figure 9. Spitzer's photometric precision as a function of the magnitude of the source in IRAC's channel 1 and 2 (3.6 and 3.5 µm). We show a histogram of the magnitudes in the top panel, and indicate to which planet size, photometric precision correspond to (for a SNR of 3.5). space telescope. After studying our ability to recover transits with synthetic signals, we identify six transit event candidates. They have depths at the threshold of our detection criteria and are there- fore likely to be false positives. We assumed the lightcurves con- tained no transits and generated a size/period map of upper limits on the occurence rates of planets orbiting brown dwarfs. Due to the limited number of targets, and short observation time, our results are not very constraining, but can inform us about how much more of an observational effort can be made to discover planets orbiting brown dwarfs, using the transit method. The dis- covery of an Earth-sized planet transiting a nearby and relatively bright brown dwarf would provide an exquisite target for the de- tailed characterisation of a terrestrial extrasolar world, one that will be similar to Earth in some aspects (size, temperature...), but differ- ent in others (irradiation, tidal locking...). With the current sample and assuming a 27% occurence rate (as implied by the discoveries of TRAPPIST-1 (Gillon et al. 2016) and Kepler-42 (Muirhead et al. 2012)), we estimate that 175 brown dwarfs need to be monitored for 21 hours, of which 44 have already been done. This sample of brown dwarfs, that composed the WoW survey, is centred on the L-T transition, the focus of most of brown dwarf variability. In ad- dition many of these targets were not selected for their brightness (see Fig. 9). A careful sample selection concentrating on brown dwarfs with magnitudes brighter than 12 (in both Spitzer channels) would likely lower this required number down. A doubling of the observing time only marginally helps transit recovery (also because we used a 'single transit' detection approach). Finding one Earth-sized planet transiting a nearby brown dwarf, would provide an exquisite target for the detailed character- isation of a terrestrial extrasolar world using transmission (like for TRAPPIST-1; Gillon et al. 2016; de Wit et al. 2016), and emission spectroscopy. In addition, the discovery of multiple planet orbiting brown dwarfs may help answer fundamental questions in Astrobi- ology, yielding precious information on the physical requirement and timescale for the emergence of life. Stars evolve, which changes the irradiation received by a planet, and therefore its ability to retain liquid water at the surface. This implies that many planets remain within the habitable zone only for a while. Some enter it at a late stage in their history. This argument has been used by Ramirez & Kaltenegger (2016). They argue that as solar-like stars leave the main sequence and evolve to the stage of giants, the habitable zone sweeps worlds that were pre- viously too cold. We would counter-argue that as a star becomes a giant, it becomes harder to distinguish the planet from the star due to a worsening flux and size ratio. However the main idea, that of evolution remains valid, and very useful. Brown dwarfs evolve all the time. They shrink and cool down (eg. Burrows et al. 1997; Baraffe et al. 2003) constantly. Planets that were previously not within the habitable zone will eventually enter it, sometimes Gyrs after their formation. Identifying such planets, and finding traces of life inside their atmospheres would inform us that Archean conditions are not necessary for the emergence of life. In addition, the time a given planet will remain habitable will vary as a function of the mass of the brown dwarf host. Once a sta- tistically significant sample of habitable worlds has been gathered orbiting a variety of brown dwarfs, it will become possible to mea- sure a timescale for the emergence of biology from the amount of time spent being habitable and the relative number of planets. ACKNOWLEDGEMENTS We are grateful, and thank our anonymous referee for taking the time to read and assess our paper. MYH would acknowledge and thank Yanqin Wu for supporting this research financially. AHMJT was at the time a fellow of the Centre for Planetary Sciences, hosted at the University of Toronto. MG is a Research Associate at the Belgian Scientific Research Foundation (F.R.S-FNRS). MYH and AHMJT acknowledge the many discussions and friendly collabo- rations that existed and allowed this project to fruition with Yan- qin Wu, Daniel Tamayo, Hanno Rein, Diana Valencia and Kristen Menou. In addition, our thoughts about planets transiting brown dwarfs were influenced via interactions with Frank Selsis, Sean Raymond, Émeline Bolmont, Jérémy Leconte, Brice-Olivier De- mory, Josh Winn, Stanimir Metchev and Adam Burgasser. We are grateful to Stanimir Metchev and to Aren Heinze for sending us the lightcurves resulting from their analysis of the WoW programme. This publication makes use of data products from the Wide- field Infrared Survey Explorer (Cutri & et al. 2013), which is a joint project of the University of California, Los Angeles, and the Jet Propulsion Laboratory/California Institute of Technology, funded by the National Aeronautics and Space Administration. We also used data products from the Two Micron All Sky Survey (Skrutskie et al. 2006), which is a joint project of the Uni- versity of Massachusetts and the Infrared Processing and Analysis Center/California Institute of Technology, funded by the National Aeronautics and Space Administration and the National Science Foundation. REFERENCES Ackerman A. S., Marley M. S., 2001, ApJ, 556, 872 c(cid:13) 2015 RAS, MNRAS 000, 1–11 0.00.20.4 10 11 12 13 14 150.0000.0050.0100.0150.020 . . . .[3.6 µm][4.5 µm]IRAC magnitudePhotometric rms0.25 Rearth0.75 Rearth1.25 RearthBrown dwarf planets 11 Muirhead P. S. et al., 2012, ApJ, 747, 144 -, 2015, ApJ, 801, 18 Nutzman P., Charbonneau D., 2008, PASP, 120, 317 Payne M. J., Lodato G., 2007, MNRAS, 381, 1597 Pollacco D. L. et al., 2006, PASP, 118, 1407 Radigan J., Jayawardhana R., Lafrenière D., Artigau É., Marley M., Saumon D., 2012, ApJ, 750, 105 Radigan J., Lafrenière D., Jayawardhana R., Artigau E., 2014, ApJ, 793, 75 Ramirez R. M., Kaltenegger L., 2016, ApJ, 823, 6 Ricci L., Isella A., Carpenter J. M., Testi L., 2013, ApJL, 764, L27 Ricci L., Trotta F., Testi L., Natta A., Isella A., Wilner D. J., 2012, A&A, 540, A6 Rodler F., López-Morales M., 2014, ApJ, 781, 54 Rogers L. A., 2015, ApJ, 801, 41 Sahlmann J., Lazorenko P. F., Ségransan D., Martín E. L., Mayor M., Queloz D., Udry S., 2014, A&A, 565, A20 Scholz A., Jayawardhana R., Wood K., Lafrenière D., Schreyer K., Doyon R., 2008, ApJL, 681, L29 Seager S., Deming D., 2010, ARA&A, 48, 631 Seager S., Mallén-Ornelas G., 2003, ApJ, 585, 1038 Skrutskie M. F. et al., 2006, AJ, 131, 1163 Snellen I. et al., 2015, A&A, 576, A59 Triaud A. H. M. J. et al., 2013a, ArXiv e-prints -, 2013b, A&A, 549, A18 Winn J. N., 2010, Exoplanet Transits and Occultations, Seager S., ed., pp. 55–77 This paper has been typeset from a TEX/ LATEX file prepared by the author. Artigau É., Bouchard S., Doyon R., Lafrenière D., 2009, ApJ, 701, 1534 Bakos G. Á. et al., 2007, ApJ, 656, 552 Baraffe I., Chabrier G., Allard F., Hauschildt P. H., 1998, A&A, 337, 403 Baraffe I., Chabrier G., Barman T. S., Allard F., Hauschildt P. H., 2003, A&A, 402, 701 Batalha N. M. et al., 2013, ApJS, 204, 24 Belu A. R., Selsis F., Morales J.-C., Ribas I., Cossou C., Rauer H., 2011, A&A, 525, A83 Blake C. H., Charbonneau D., White R. J., 2010, ApJ, 723, 684 Blake C. H., Latham D. W., Bloom J. S., 2007, in Astronomical Society of the Pacific Conference Series, Vol. 366, Transiting Extrapolar Planets Workshop, Afonso C., Weldrake D., Henning T., eds., p. 87 Bolmont E., Raymond S. N., Leconte J., 2011, A&A, 535, A94 Boss A. P. et al., 2009, PASP, 121, 1218 Brogi M., de Kok R. J., Birkby J. L., Schwarz H., Snellen I. A. G., 2014, A&A, 565, A124 Burgasser A. J., Marley M. S., Ackerman A. S., Saumon D., Lod- ders K., Dahn C. C., Harris H. C., Kirkpatrick J. D., 2002, ApJL, 571, L151 Burrows A. et al., 1997, ApJ, 491, 856 Chauvin G., Lagrange A.-M., Dumas C., Zuckerman B., Mouillet D., Song I., Beuzit J.-L., Lowrance P., 2004, A&A, 425, L29 Claret A., Bloemen S., 2011, A&A, 529, A75 Collier Cameron A., Horne K., Penny A., James D., 1999, Nature, 402, 751 Collier Cameron A. et al., 2007, MNRAS, 380, 1230 Crossfield I. J. M. et al., 2014, Nature, 505, 654 Cutri R. M., et al., 2013, VizieR Online Data Catalog, 2328, 0 de Wit J., Gillon M., Demory B.-O., Seager S., 2012, A&A, 548, A128 de Wit J. et al., 2016, Nature, 537, 69 Díaz R. F. et al., 2013, A&A, 551, L9 Dieterich S. B., Henry T. J., Jao W.-C., Winters J. G., Hosey A. D., Riedel A. R., Subasavage J. P., 2014, AJ, 147, 94 Dressing C. D., Charbonneau D., 2013, ApJ, 767, 95 Fazio G. G. et al., 2004, ApJS, 154, 10 Gillon M., Jehin E., Delrez L., Magain P., Opitom C., Sohy S., 2013a, in Protostars and Planets VI Posters Gillon M. et al., 2016, Nature, 533, 221 Gillon M., Jehin E., Magain P., Chantry V., Hutsemékers D., Man- froid J., Queloz D., Udry S., 2011, in European Physical Journal Web of Conferences, Vol. 11, European Physical Journal Web of Conferences, p. 6002 Gillon M., Triaud A. H. M. J., Jehin E., Delrez L., Opitom C., Magain P., Lendl M., Queloz D., 2013b, A&A, 555, L5 Han C. et al., 2013, ApJ, 778, 38 Heinze A. N. et al., 2013, ApJ, 767, 173 Howard A. W. et al., 2010, Science, 330, 653 Jehin E. et al., 2011, The Messenger, 145, 2 Kaltenegger L., Traub W. A., 2009, ApJ, 698, 519 Kovács G., Zucker S., Mazeh T., 2002, A&A, 391, 369 Leigh C., Collier Cameron A., Udry S., Donati J.-F., Horne K., James D., Penny A., 2003, MNRAS, 346, L16 Luhman K. L., 2013, ApJL, 767, L1 Mandel K., Agol E., 2002, ApJL, 580, L171 Mayor M. et al., 2011, eprint arXiv:1109.2497 Meru F., Galvagni M., Olczak C., 2013, ApJL, 774, L4 Metchev S. A. et al., 2015, ApJ, 799, 154 Moutou C. et al., 2013, A&A, 558, L6 c(cid:13) 2015 RAS, MNRAS 000, 1–11
1608.03484	1	1608	2016-08-11T14:40:28	Secular and tidal evolution of circumbinary systems	[ "astro-ph.EP", "astro-ph.IM", "math.DS" ]	We investigate the secular dynamics of three-body circumbinary systems under the effect of tides. We use the octupolar non-restricted approximation for the orbital interactions, general relativity corrections, the quadrupolar approximation for the spins, and the viscous linear model for tides. We derive the averaged equations of motion in a simplified vectorial formalism, which is suitable to model the long-term evolution of a wide variety of circumbinary systems in very eccentric and inclined orbits. In particular, this vectorial approach can be used to derive constraints for tidal migration, capture in Cassini states, and stellar spin-orbit misalignment. We show that circumbinary planets with initial arbitrary orbital inclination can become coplanar through a secular resonance between the precession of the orbit and the precession of the spin of one of the stars. We also show that circumbinary systems for which the pericenter of the inner orbit is initially in libration present chaotic motion for the spins and for the eccentricity of the outer orbit. Because our model is valid for the non-restricted problem, it can also be applied to any three-body hierarchical system such as star-planet-satellite systems and triple stellar systems.	astro-ph.EP	astro-ph	Noname manuscript No. (will be inserted by the editor) Secular and tidal evolution of circumbinary systems. Alexandre C. M. Correia · Gwenael Bou´e · Jacques Laskar 6 1 0 2 g u A 1 1 . ] P E h p - o r t s a [ 1 v 4 8 4 3 0 . 8 0 6 1 : v i X r a August 12, 2016 Abstract We investigate the secular dynamics of three-body circumbinary systems under the effect of tides. We use the octupolar non-restricted approximation for the orbital interactions, general relativity corrections, the quadrupolar approximation for the spins, and the viscous linear model for tides. We derive the averaged equations of motion in a simplified vectorial formalism, which is suitable to model the long-term evolution of a wide variety of circumbinary systems in very eccentric and inclined orbits. In particular, this vectorial approach can be used to derive constraints for tidal migration, capture in Cassini states, and stellar spin-orbit misalignment. We show that circumbinary planets with initial arbitrary orbital inclination can become coplanar through a secular resonance between the precession of the orbit and the precession of the spin of one of the stars. We also show that circumbinary systems for which the pericenter of the inner orbit is initially in libration present chaotic motion for the spins and for the eccentricity of the outer orbit. Because our model is valid for the non-restricted problem, it can also be applied to any three-body hierarchical system such as star-planet-satellite systems and triple stellar systems. Keywords Extended Body · Dissipative Forces · Planetary Systems · Rotation 1 Introduction Circumbinary bodies are objects that orbit around a more massive binary system. In the solar system, the small satellites of the Pluto-Charon system are the best example (e.g. Brozovi´c et al, 2015). Planets orbiting two stars, often called circumbinary planets, have also been reported, at present we know about 20 of them1. These kind of systems are particularly interesting from a dynamical point of view, as they can be stable for very eccentric and inclined orbits and thus present uncommon behaviors. Moreover, A.C.M. Correia CIDMA, Departamento de F´ısica, Universidade de Aveiro, Campus de Santiago, 3810-193 Aveiro, Portugal E-mail: [email protected] G. Bou´e · J. Laskar ASD, IMCCE-CNRS UMR8028, Obs. Paris, 77 Av. Denfert-Rochereau, 75014 Paris, France 1 http://exoplanet.eu/ 2 many circumbinary systems are observed within 1 AU, therefore they can undergo tidal dissipation, which slowly modifies the spins and the orbits, in particular for the inner binary pair (e.g. MacDonald, 1964). As a consequence, the final configuration of these systems can be totally different from the initial one (e.g. Correia et al, 2011). Half of the already known circumbinary planets were detected using the Kepler Spacecraft2 data, which is based on the transits observational technique (e.g. Doyle et al, 2011; Welsh et al, 2012; Orosz et al, 2012a,b; Kostov et al, 2015). These systems are thus almost coplanar, although this is not necessarily the standard configuration for circumbinary planets (e.g. Martin and Triaud, 2014). Indeed, misaligned circumbinary discs were already observed (e.g. Kennedy et al, 2012; Plavchan et al, 2013). Moreover, for systems detected with different observational techniques, the mutual inclination is usually not constrained, but it is compatible with no coplanar orbits (e.g. Ford et al, 2000a; Couetdic et al, 2010). The origin and evolution of circumbinary systems can be analyzed with direct numerical integrations of the full equations of motion (e.g. Verrier and Evans, 2009; Martin and Triaud, 2014), but a theoretical understanding of the dynamics often re- quires the development and application of theories with valid analytic approximations. Additionally, tidal effects usually act over very long time-scales and therefore approx- imate theories also allow to speed-up the numerical simulations and to explore the parameter space much more rapidly. Stability studies have shown that for semi-major axis ratios a1/a2 > 1/2 there is a large overlap of mean motion resonances and circumbinary orbits are most likely unstable (e.g. Doolin and Blundell, 2011; Bosanac et al, 2015; Correia et al, 2015). For smaller ratios we can use secular perturbation theories based on series development in the ratio of the semi-major axis a1/a2. The development to the second order, called the quadrupole approximation, was used by Lidov (1962) and Kozai (1962) for the restricted inner problem (the outer orbit is unperturbed). Farago and Laskar (2010) derived a simple model for the non-restricted quadrupolar problem of three masses and Correia et al (2011) added the effect from tides. However, the quadrupole approxima- tion is insufficient for studying nearly coplanar systems, because it fails to reproduce the eccentricity oscillations. For that purpose we need to extend the series development in a1/a2 to the third order, that is, to the octopole order (e.g. Marchal, 1990; Ford et al, 2000b; Laskar and Bou´e, 2010). In this paper we intend to get deeper into the study of circumbinary three-body systems, where all bodies undergo tidal interactions. We do not make any restrictions on the masses of these bodies, and use the octupolar approximation for the orbital gravitational interactions with general relativity corrections. It has been shown that the tidal evolution of the orbits cannot be dissociated from the spin evolution (e.g. Correia et al, 2012). Therefore, we also consider the full effect on the spins of all bodies (up to the quadrupolar approximation), including the rotational flattening of their figures. This allows us to correctly describe the precession of the spin axis and subsequent capture in Cassini states. We adopt a viscous linear model for tides (Singer, 1968; Mignard, 1979), as it provides simple expressions for the tidal torques for any eccentricity value. For gaseous planets and stars, this model has also shown to be a very good approximation (e.g. Ferraz-Mello, 2013; Correia et al, 2014). Since we are interested in the secular behavior, we average the motion equations over the mean 2 http://kepler.nasa.gov/ 3 anomalies of the orbits and express them using the vectorial methods (e.g. Bou´e and Laskar, 2006, 2009; Correia, 2009; Farago and Laskar, 2010; Bou´e and Fabrycky, 2014). In Section 2 we derive the averaged equations of motion that we use to evolve cir- cumbinary systems by tidal effect. In Section 3 we obtain the secular evolution of the spin and orbital quantities in terms of reference angles and elliptical elements, that are useful and more intuitive to understand the outcomes of the numerical simulations. In Section 4 we include the contribution of tidal effects and perform some numerical simu- lations to illustrate some unexpected behaviors for circumbinary systems. In Section 5 we explain how our model can be extended to any three-body hierarchical system. Finally, last section is devoted to the conclusions. 2 Model We consider here a system composed of an inner pair of bodies with stellar masses m0 and m1, together with an external planetary companion with mass m2: m0 ≥ m1 (cid:29) m2 . (1) All bodies are considered oblate ellipsoids with gravity field coefficients given by J2,i, rotating about the axis of maximal inertia along the directions si (gyroscopic approx- imation), with rotation rates Ωi, such that (e.g. Lambeck, 1988) J2,i = k2,i Ω2 i R3 i 3Gmi , (2) where G is the gravitational constant, Ri is the radius of each body, and k2,i is the second Love number for potential. The index i = 0, 1, 2 pertains to the body with mass mi. In order to obtain the equations of motion we use Jacobi canonical coordinates, which are the distance between the two innermost bodies, r1, and the distance from the external body to the inner's orbit center of mass, r2 (see Fig. 1). We additionally assume that r1 (cid:28) r2 and adopt the octupolar three-body problem approximation. In the following, for any vector u, u = u/(cid:107)u(cid:107) is the unit vector. 2.1 Orbital motion The potential energy U of a hierarchical three-body system of punctual masses is given in Jacobi coordinates by (e.g. Smart, 1953): − G U = −G m2m01 m0m1 (cid:48) + U ; (3) r1 r2 with (cid:18) r1 r2 (cid:48) U = −G m2β1 r2 (cid:19)2 P2(r2 · r1) − G m2β1 r2 m0 − m1 m0 + m1 (cid:19)3 P3(r2 · r1) , (cid:18) r1 r2 (4) where P2(x) = (3x2 − 1)/2 and P3(x) = (5x3 − 3x)/2 are the Legendre polynomial of degree two and three, respectively, and terms in (r1/r2)4 have been neglected. We also have m01 = (m0 + m1), β1 = m0m1/m01, β2 = m2m01/(m2 + m01), µ1 = Gm01, and 4 Fig. 1 Jacobi coordinates, where r1 is the position of m1 relative to m0 (inner orbit), and r2 the position of m2 relative to the center of mass of m0 and m1 (outer orbit). All bodies are considered oblate ellipsoids, where si is the spin axis. µ2 = G(m2 + m01). The evolution of the orbits can be tracked by the orbital angular momenta, Gi = Gi ki , with Gi = βi µiai(1 − e2 i ) where ki is the unit vector Gi, ai is the semi-major axis, and ei is the eccentricity. The mean motion is defined as ni = i . The Laplace-Runge-Lenz vector ei points along the major axis in the direction of periapsis with magnitude e1 and is expressed as µi/a3 The contributions to the orbits are easily computed from the potential as (cid:113) (cid:113) ei = ri × Gi βiµi − ri ri . Gi = ri × Fi , (cid:32) (cid:33) , Fi × Gi βi + ri × Gi (5) (6) (7) (8) and where Fi = −∇ri U(cid:48). ei = 1 βiµi Because we are only interested in the secular evolution of the system, we further av- erage the equations of motion over the mean anomalies of both orbits (see appendix C). The resulting equations are: Quadrupole: G1 = −γ2 (cid:104) (1 − e2 1) cos I k2 × k1 − 5(e1 · k2) k2 × e1 G2 = − G1 , (cid:105) , (9) (10) m1m0m2r1r2s01s2sr21r20(cid:104) cos I k2 × e1 − 2 k1 × e1 − 5(e1 · k2) k2 × k1 (cid:105) 5 , (11) 1) (1 − e2 1) cos I k1 × e2 − 5(e1 · k2) e1 × e2 1) cos2 I + 25(e1 · k2)2(cid:17) k2 × e2 1 − 5(1 − e2 (cid:105) , 1 − 6e2 e1 = − γ2(1 − e2 (cid:107)G1(cid:107) (cid:104) e2 = − γ2 (cid:107)G2(cid:107) 1 2 (cid:16) + with and Octupole: cos I = k1 · k2 , γ2 = 3Gm2β1a2 1 2(1 − e2 2)3/2 4a3 . (cid:104) G1 = γ3 (B e2 + C k2) × e1 + (D e2 + E k2) × k1 (cid:105) , G2 = − G1 , (1 − e2 1)(A e1 + B e2 + C k2) × k1 + (D e2 + E k2) × e1 (cid:104)(cid:0)F + C(e1 · k2) + E cos I(cid:1) e2 × k2 (cid:104) e1 = γ3(cid:107)G1(cid:107) e2 = with γ3(cid:107)G2(cid:107) +(1 − e2 (cid:105) , (cid:105) 2)(B e1 + D k1) × k2 + (C e1 + E k1) × e2 A = 16(e1 · e2) , B = −(cid:104) 1 1 − 5(1 − e2 1) cos2 I + 35(e1 · k2)2 − 8e2 , C = 10(1 − e2 1)(k1 · e2) cos I − 70(e1 · k2)(e1 · e2) , 1)(e1 · k2) cos I , D = 10(1 − e2 E = 10(1 − e2 1) F = 5 (cid:104)B(e1 · e2) + D(k1 · e2) (e1 · e2) cos I + (e1 · k2)(k1 · e2) (cid:105) (cid:105) (cid:104) , , and γ3 = 15 64 Gm2β1a3 1 2(1 − e2 a4 2)5/2 m0 − m1 m0 + m1 . (12) (13) (14) (15) (16) , (17) (cid:105) (18) (19) (20) 6 Fig. 2 Reference planes for the definition of the direction cosines and the precession angles. 2.2 General relativity correction We can add to Newton's equations the dominant contributions from general relativistic effects. These effects are mainly felt by eccentric orbits on close encounters between the central stars, and contribute to the gravitational force with a small correction (e.g. Kidder, 1995) (cid:19) (cid:21) (cid:20)(cid:18) Fgr = − β1µ1 c2r2 1 (1 + 3η)r2 1 − 2(2 + η) µ1 r1 − 3 2 η r2 1 r1 − 2(2 − η) r1 r1 , (21) where c is the speed of light, and η = m0m1/m2 tivistic secular contribution is on the precession of the periapsis, with (Eq. 8): 01. To this order, the dominant rela- e1 = 3µ1n1 c2a1(1 − e2 1) k1 × e1 . (22) 2.3 Spin motion All bodies are oblate ellipsoids, so we need to take into account the deformation of the gravity field given by their structure. The additional contribution to the potential energy (Eq. 4) is given in Jacobi coordinates by (see appendix A): (cid:88) i=0,1 UR = G m0m1 r1 + G m2m01 r2 J2,2 (cid:18) Ri (cid:18) R2 (cid:19)2 J2,i r1 r2 (cid:19)2 P2(r1 · si) P2(r2 · s2) , (23) orbit 1equatororbit 2Iϕφεθ7 where terms in (Ri/rj )3, (R2/r2)2(r1/r2)2, and m2/mi(r1/r2)3 have been neglected (i = 0, 1, 2, j = 1, 2). As in previous section, we can obtain the contributions to the orbital motion directly from equations (7) and (8) using UR instead of U(cid:48), The evolution of the spins can be tracked by the rotational angular momenta, Li (cid:39) CiΩi si, where Ci are the principal moments of inertia. In Jacobi coordinates, the orbital angular momentum is equal to G1 + G2 (Poincar´e, 1905). Since the total angular momentum is conserved, the contributions to the spin of the bodies can be computed from their orbital contributions: L0 + L1 = − G1 , L2 = − G2 . (24) Then, averaging again the equations of motion over the mean anomalies of both orbits (see appendix C) we get: L0 = −α0 cos θ0 k1 × s0 , L1 = −α1 cos θ1 k1 × s1 , L2 = −α2 cos ε2 k2 × s2 , G1 = −α0 cos θ0 s0 × k1 − α1 cos θ1 s1 × k1 , G2 = −α2 cos ε2 s2 × k2 , e1 = − (cid:88) i=0,1 (cid:104) (cid:104) αi (cid:107)G1(cid:107) cos θi si × e1 + 1 2 (1 − 5 cos2 θi) k1 × e1 e2 = − α2 (cid:107)G2(cid:107) cos ε2 s2 × e2 + 1 2 (1 − 5 cos2 ε2) k2 × e2 (cid:105) , (cid:105) , where and α0 = α1 = α2 = 1(1 − e2 3Gm0m1J2,0R2 0 2a3 1)3/2 3Gm0m1J2,1R2 1 2a3 1)3/2 3Gm2m01J2,2R2 2 2a3 2)3/2 1(1 − e2 2(1 − e2 , , , cos θi = si · k1 , cos εi = si · k2 , (25) (26) (27) (28) (29) (30) (31) (32) (33) (34) (35) are the direction cosines of the spins: θi is the obliquity to the orbital plane of the inner orbit, and εi is the obliquity to the orbital plane of the outer companion. Using the mutual inclination between orbital planes (Eq. 13), the three direction cosines can also be related as cos εi = cos I cos θi + sin I sin θi cos φi , (36) 8 or cos θi = cos I cos εi − sin I sin εi cos ϕi , (37) where φi is the precession angle between the projections of si and k2 in the plane normal to k1, and ϕi is the precession angle between the projections of si and k1 in the plane normal to k2 (see Fig. 2). 2.4 Tidal effects (cid:88) i=0,1 Neglecting the tidal effects raised by the planet in each star (i = 0, 1), that is, neglecting terms in m2/mi(r1/r2)3, the tidal potential energy of the system is given in Jacobi coordinates by (see appendix B): UT = − G r3 1 k2,i m2 j R5 i r(cid:48)3 1i (cid:48) P2(r1 · r 1i) − k2,2 Gm2 01R5 2 2r(cid:48)3 r3 2 (cid:48) P2(r2 · r 2) , (38) where j = 1 − i, and r(cid:48) i is the position of the interacting mass at a time delayed of ∆ti, which corresponds to the deformation time-lag of the body i. The dissipation of the mechanical energy of tides in the body's interior is responsible for this delay between the initial perturbation and the maximal deformation. As the rheology of stars and planets is badly known, the exact dependence of ∆ti on the tidal frequency is unknown. Many different authors have studied the problem and several models have been developed so far, from the simplest ones to the more complex (for a review see Efroimsky and Williams, 2009; Ferraz-Mello, 2013; Correia et al, 2014). The huge problem in validating one model better than others is the difficulty to compare the theoretical results with the observations, as the effect of tides are very small and can only be detected efficiently after long periods of time. Nevertheless, for gaseous planets and stars, the amount of tidal energy that is dissipated in a cycle is rather small (e.g. Lainey et al, 2009, 2012; Penev et al, 2012), which is equivalent to short time responses. In those cases, most viscoelastic tidal models can be made linear (weak friction), with a constant ∆ti value (see Correia et al, 2014; Makarov, 2015). Therefore, for simplicity, we adopt here a weak friction model with constant ∆ti (Singer, 1968; Alexander, 1973; Mignard, 1979), for which: (cid:48) 1i (cid:39) r1 + ∆ti (Ωisi × r1 − r1) , r (cid:48) 2 (cid:39) r2 + ∆t2 (Ω2s2 × r2 − r2) . r (39) As for the spin motion, we can obtain the equations of motion directly from equa- tions (7), (8) and (24) using UT instead of U(cid:48), and then averaging over the mean anomalies of both orbits (see appendix C): G1 = − (cid:88) (cid:20) (cid:113) i=0,1 Li , G2 = − L2 . Li = Ki n1 f4(e1) 1 − e2 1 Ωi 2n1 −f1(e1) Ωi n1 si + f2(e1)k1 + (si − cos θi k1) (e1 · si)(6 + e2 1) 4(1 − e2 1)9/2 (40) (41) (cid:21) e1 , Ωi n1 (cid:20) f4(e2) L2 = K2 n2 (cid:113) 1 − e2 2 Ω2 2n2 −f1(e2) Ω2 n2 s2 + f2(e2)k2 + (s2 − cos ε2 k2) (e2 · s2)(6 + e2 2) 4(1 − e2 2)9/2 9 (42) (cid:21) e2 , Ω2 n2 f4(e1) (e1 · si) k1 − Ωi 2n1 f4(e1) cos θi − 9f5(e1) Ωi n1 2 (cid:33)5 (cid:33)(cid:32) (cid:32) mj mi Ri a1 f4(e1) k1 × e1 (cid:18) 11 (cid:32) (cid:33)(cid:32) (cid:33)5 R2 a2 m01 m2 f4(e2) k2 × e2 (cid:18) 11 (cid:19) (cid:21) e1 ,(43) (cid:19) (cid:21) e2 , (44) e1 = (cid:88) − (cid:88) i=0,1 i=0,1 e2 = 15 2 15 2 k2,i n1 (cid:20) Ki β1a2 1 k2,2 n2 (cid:20) − K2 β2a2 2 where and f4(e2) (e2 · s2) k2 − Ω2 2n2 f4(e2) cos ε2 − 9f5(e2) Ω2 n2 2 Ki = ∆ti 3k2,iGm2 j R5 i a6 1 , K2 = ∆t2 3k2,2Gm2 01R5 2 a6 2 , (45) f1(e) = 1 + 3e2 + 3e4/8 (1 − e2)9/2 , f2(e) = 1 + 15e2/2 + 45e4/8 + 5e6/16 (1 − e2)6 , (46) (47) f3(e) = 1 + 31e2/2 + 255e4/8 + 185e6/16 + 25e8/64 (1 − e2)15/2 , (48) f4(e) = 1 + 3e2/2 + e4/8 (1 − e2)5 , f5(e) = 1 + 15e2/4 + 15e4/8 + 5e6/64 (1 − e2)13/2 . (49) (50) The first term in expressions (43) and (44) corresponds to the permanent tidal deformation, while the second term corresponds to the dissipative contribution. 10 Table 1 Initial parameters for the Kepler-16 system (Doyle et al, 2011; Winn et al, 2011) and for a hypothetical circumbinary system (that we call "standard"), very similar to the Kepler-34 system (Welsh et al, 2012). The main differences between Kepler-34 and the standard system are in the mass of star B, and in the semi-major axis of the inner orbit (m1 ≈ M(cid:12) and a1 ≈ 0.2 AU for Kepler-34), in order to enhance tidal effects. The unknown parameters are compatible with observations for the Sun and giant planets in the Solar System. Kepler-16 standard parameter m (M(cid:12)) Prot (day) θ (deg) φ (deg) R (×106m) C/(mR2) k2 ∆t (s) parameter a (AU) e (deg) I (deg) star A star B 0.20 20. 10. 0. 0.69 35.1 1.6 0. 157. 0.08 0.028 0.1 452. 0.08 0.028 0.1 orbit 1 0.22 0.16 263. planet 0.0003 0.5 1. 0. 53. 0.20 0.50 100. orbit 2 0.70 0.01 318. star A star B planet 0.001 1.0 10. 5. 0. 0.2 1.0 10. 0. 0.5 20. 0. 70. 0.25 0.50 100. orbit 2 1.5 0.2 180. 695. 0.08 0.028 0.1 orbit 1 150. 0.08 0.028 0.1 0.1 0.5 0. 1. 10. 3 Secular dynamics In this section we look at the orbital and spin dynamics without tidal dissipation. We first consider only the orbital dynamics without spin (section 3.1) and then add to the equations of motion for the spin of a single body (section 3.2). In some special configurations the secular equations of motion become integrable, which allow us to easily compute the possible trajectories for the system given its initial conditions. Moreover, the equilibrium configurations for the orbit and spin, that correspond to the final outcomes of tidal evolution, become easy to identify. For that purpose, we will use a reference hypothetical circumbinary system (Table 1), which is similar to Kepler-34 (Welsh et al, 2012). In section 4, this particular choice of parameters will allow us to illustrate some interesting dynamical effects that were not yet described in the literature. 3.1 Orbital dynamics The non-resonant dynamics of circumbinary planets has been studied in great detail by Migaszewski and Go´zdziewski (2011). Here we recall some of the main features of the problem, in particular those corresponding to an integrable problem, which will allow us to understand the more complex dynamics when the effect from the spins and tides are included. 3.1.1 Coplanar systems When sin I = 0 (coplanar systems), the average of the orbital energy over the mean anomalies, in the octupolar approximation (Eq. 4) is given by (e.g. Lee and Peale, 2003; Laskar and Bou´e, 2010; Correia et al, 2012) 11 3 2 (cid:104)U(cid:105) = − γ2 3 e2 1) + 4γ3(1 + (1 + (51) where ∆ = 1 − 2, and i is the longitude of the pericenter of each orbit. The parameters γ2 and γ3 are given by expressions (14) and (20), respectively, and depend only on e2. For very eccentric inner orbits, the above orbital energy can be corrected for general relativity effects (Eq. 21) through the additional contribution (e.g. Touma et al, 2009): e2 1)e1e2 cos ∆ , 3 4 (cid:104)Ugr(cid:105) = − 3β1µ2 1 1c2(1 − e2 a2 1)1/2 . (52) The secular system (51,52) does not depend on the mean longitudes, so the semi-major axes are constant. Moreover, it depends on a unique angle ∆. It is thus integrable. The eccentricities are related through the total orbital angular momentum (Eq. 5): (cid:113) (cid:113) µ1a1(1 − e2 µ2a2(1 − e2 β1 (53) where σ = ±1, depending whether the orbits rotate in the same way or with opposite direction. Here we will only consider the σ = +1 case of planets orbiting in the same direction. 2) = C = cte . 1) + σβ2 In Figure 3 we plot the level curves of the total energy in the plane (e2, ∆) for Kepler-16 and for the standard system with I = 0 (Table 1). There are essentially two possible kinematic regimes for ∆ in Figure 3: oscillation around 0◦ or circulation be- tween −180◦ and 180◦. The standard system (Table 1) is in circulation, but for Kepler- 16 we cannot be sure. The best fit data (Doyle et al, 2011) gives ∆ ≈ (−55±20)◦. For the mean value of this estimate, the system is oscillating around 0◦, but for the lower bound ∆ ≈ −75◦ it is very near circulation zone (the dotted line gives the transition of kinematic regime). Although the angle ∆ may present a different behavior, the dynamics in the plane (e2 sin ∆, e2 cos ∆) always correspond to circulation of all trajectories around a fix point (e2 ≈ 0.033, ∆ = 0◦). For both systems, we observe that the eccentricity of the outer orbit only undergoes small variations. In the secular system, as the semi-major axes are constant, the difference of the angular momentum (53) with respect to the circular angular momentum, i.e. the AMD (Laskar, 2000) is also constant, that is µ2a2(1 −(cid:113) √ 1 − e2 1) + β2 1 − e2 2) = D = cte . µ1a1(1 −(cid:113) Thus, with χ(e) = 1 − √ √ β1 1 − e2, (54) (55) χ(e1) = D √ µ1a1 − β2 β1 β1 µ2a2 µ1a1 χ(e2) √ √ √ µ2a2 (cid:28) √ Since most of the orbital angular momentum is contained in the inner orbit (β2 β1 µ1a1), the variations in χ(e1) (resp. e1) are smaller than those in χ(e2) (resp. e2), so the inner orbit eccentricity e1 can be considered almost as constant. For Kepler-16 and the standard systems we have m0 (cid:29) m1, but for other circumbinary systems, such as Kepler-34 or Kepler-35 (Welsh et al, 2012), we have m0 ∼ m1. In those cases γ3 ≈ 0, so the octupolar contribution vanishes (Eq. 51), and both eccentricities remain almost constant. 12 Fig. 3 Level curves for the outer orbit eccentricity of Kepler-16 and the standard system (Table 1). There are essentially two possible kinematic regimes: oscillation around ∆ = 0◦ and circulation. The present position of each system is given by a dot. 0 0.05 0.1 0.15 0.2 0.25 0.3e2standard-180-135-90-4504590135180Δϖ (deg) 0 0.05 0.1 0.15 0.2 0.25 0.3-180-135-90-4504590135180e2Δϖ (deg)Kepler-1613 Fig. 4 Secular trajectories for the standard system (Table 1) seen in the plane (I, ω1). We show the trajectories using the quadrupolar approximation (top), corresponding to the level curves of constant energy, and using the octupolar approximation (bottom), obtained with numerical simulations. There are two possible dynamical regimes: libration around ω1 = ±90◦ and circulation (see also Migaszewski and Go´zdziewski, 2011). 14 Fig. 5 Secular trajectories for the standard system (Table 1) seen in the planes (e1, ω1) and (e2, ω1). We show the trajectories using the octupolar approximation, obtained with numerical simulations (for more details see Migaszewski and Go´zdziewski, 2011). 15 3.1.2 Inclined systems For inclined systems (sin I (cid:54)= 0), the average of the orbital energy (Eq. 4) over the mean anomalies additionally depends on I, 1 and 2 (e.g. Harrington, 1968; Lidov and Ziglin, 1976; Laskar and Bou´e, 2010), and the problem is no longer integrable. However, while for the coplanar problem the main dynamical features result from the octupolar contribution (term in ∆ω, Eq. 51), for the inclined problem the major contribution comes from the quadrupolar term. Expressing this term in the Laplace invariable plane, for which 1 = ω1 and 2 = π + ω2, where ωi is the longitude of the pericenter of each orbit measured from the line of the nodes between the two orbits, we get (Correia et al, 2013): (cid:105) . (56) (cid:10)Uqd (cid:11) = − γ2 (cid:104) (1 + 3 3 2 1)(1 − 3 e2 2 sin2 I) + 15 4 e2 1 sin2 I cos 2ω1 Moreover, cos I (and thus sin I) can be expressed in terms of G1, G2 through the conservation of the total orbital angular momentum (Eq. 5) G2 1 + G2 2 + 2G1G2 cos I = G2 tot (57) and ω1. For (cid:107)G1(cid:107) (cid:28) (cid:107)G2(cid:107) previous equation can be simplified to get(cid:112)(1 − e2 Thus, if we restrict the dynamics to the quadrupolar approximation, here again, the Hamiltonian depends on a single angle ω1 and the problem is integrable (Harrington, 1968; Lidov and Ziglin, 1976). Moreover, the outer orbit eccentricity e2 and γ2 are constant (Eq.14), since there is no contribution from ω2 (Harrington, 1968; Lidov and Ziglin, 1976; Farago and Laskar, 2010). Therefore, the orbital energy only depends on e1 1) cos I ≈ cte, which is at the origin of the Lidov-Kozai mechanism (Lidov, 1962; Kozai, 1962; Lidov and Ziglin, 1976). However, for circumbinary planets we expect (cid:107)G1(cid:107) > (cid:107)G2(cid:107), so we cannot neglect the first term in expression (57). In Figure 4 (top) we plot the level curves of the total energy in the plane (I, ω1) for the standard system (Table 1), with different values for the inclination. We observe that there are two possible dynamical regimes: circulation and libration around ω1 = ±90◦ for high inclinations values. In the example shown we have 40◦ (cid:46) I (cid:46) 140◦, but these boundaries strongly depend on the eccentricity of the inner orbit. For small e1 values the libration zone is restricted to the vicinity of I ∼ 90◦, while for large e1 values the libration region can reach very low inclinations (see Farago and Laskar, 2010). Moreover, these boundaries are not completely symmetric, since the separatrix between the libration and circulation regimes is shifted towards I > 90◦. This shift cannot be observed in the classic restricted case (e.g. Verrier and Evans, 2009), but in the planetary case with general relativity there is an increase in the precession rate of ω1 that breaks the symmetry. In Figure 4 (bottom) we plot the exact same trajectories performing numerical simulations using the full model form section 2.1 (octupole approximation). We observe that the phase space is almost unchanged, there is only some additional chaotic diffusion for the trajectories near the separatrix. The same is valid when we plot these trajectories in the plane (e1, ω1) as shown in Figure 5 (top). Although the inclination may undergo significant variations, the eccentricity of the inner orbit only varies by a very small amount due to the conservation of the total angular momentum (Eq. 57). We conclude that the quadrupolar approximation captures the main dynamical features for the e1, I and ω1 parameters. However, these conclusions cannot be extended to e2. In the quadrupolar approximation the eccentricity of the outer orbit is constant, 16 but in the octupolar approximation it undergoes some variations due to the presence of term in ω2 (Eq. 51). For the coplanar case we have 0.2 ≤ e2 ≤ 0.25 (Fig. 3). In Figure 5 (bottom) we show the same trajectories from Figure 4 in the plane (e2, ω1). For trajectories in circulation far from the separatrix, i.e., for sin I ≈ 0, the outer orbit eccentricity is still bounded by 0.2 ≤ e2 ≤ 0.25, like in the coplanar case. As the orbits become closer to the separatrix, the diffusion increases until a maximum 0.15 ≤ e2 ≤ 0.3. For trajectories in libration far from the separatrix the eccentricity oscillations are bounded by 0.17 ≤ e2 ≤ 0.23, except for those close to the equilibrium points, for which the eccentricity undergoes large amplitude variations 0.05 ≤ e2 ≤ 0.27. 3.2 Spin dynamics 3.2.1 Planetary spin We now consider the spin of the planet into the analysis. Averaging the rotational energy (Eq. 23) over the mean anomalies gives (e.g. Goldreich, 1966) (cid:10)UR,2 (cid:11) = − α2 2 (cos ε2)2 . (58) When we add the rotational energy to the orbital energy we get some additional de- grees of freedom, so the problem is not integrable. However, in previous section we saw that the eccentricity of the inner orbit only undergoes small variations (Fig. 5). Therefore, assuming e1 as constant, we can additionally average the orbital energy (Eq. 56) over the argument of the pericenter, ω1. Restricting the orbital contribution to the quadrupolar term, and suppressing the constant terms in e1, gives for the total energy: U2 =(cid:10)UR,2 (cid:11) +(cid:10)(cid:10)Uqd (cid:11)(cid:11) = − α2 2 (cos ε2)2 − γ2 2 ω1 (1 + 3 2 e2 1) cos2 I . (59) This reduced expression for the energy only depends on the direction cosines cos ε2 and cos I, which give the relative directions of the angular momentum components, together with cos θ2 (Eq. 35). The three directions are related through the total angular momentum (with Li = (cid:107)Li(cid:107)) L2G1 cos θ2 + L2G2 cos ε2 + G1G2 cos I = K2 = cte . (60) This problem is then integrable (Bou´e and Laskar, 2006; Bou´e and Fabrycky, 2014; Correia, 2015), since the three degrees of freedom (given by the direction cosines) can be related through the total energy (Eq. 59) and the total angular momentum (Eq. 60). Moreover, we usually have L2 (cid:28) G2 (cid:28) G1, so previous expression can be simplified as cos I = cos I0 − L2 G2 cos2 I = cos2 I0 − 2 cos I0 L2 G2 cos θ2 + O( L2 G1 ) , cos θ2 + O( L2 G1 ) , with cos I0 = K2 G1G2 = cte . (61) (62) (63) 17 We conclude that the mutual inclination is almost constant (I ≈ I0) and, leaving out the constant terms, the total energy (Eq. 59) becomes: with U2 ≈ − α2 2 (cos ε2)2 − g2 cos θ2 , g2 = −γ2(1 + 3 2 e2 1) cos I0 L2 G2 = cte . (64) (65) g2/L2 is the precession rate of the line of nodes of the two orbital planes. The total energy expressed in this form is equivalent to the one obtained for previous works that study the spin evolution of a planet around a single star, whose orbit is perturbed by other planetary companions (e.g. Ward and Hamilton, 2004). Planetary spins are often described with respect to their orbital plane. In order to better understand the secular trajectories for the spin of the planet, we perform a change of variables from polar to rectangular coordinates (ε2, ϕ2) −→ (u2, v2) that gives the projection of the spin in the orbit (Correia, 2015): u2 = sin ε2 cos ϕ2 ; v2 = sin ε2 sin ϕ2 , (66) where ϕ2 is the precession angle measured along the outer orbit from the inner orbit to the equatorial plane of the planet (Fig. 2). Thus, and, with I ≈ I0, cos ε2 = (cid:113) 1 − u2 2 − v2 2 , (cid:113) (67) (68) cos θ2 = cos I cos ε2 − sin I sin ε2 cos ϕ2 ≈ cos I0 1 − u2 2 − v2 2 − u2 sin I0 . Replacing the above expressions for the direction cosines in expression (64) finally gives for the total energy: (cid:16) (cid:18) (cid:17) − g2 (cid:113) U2 ≈ α2 2 2 + v2 u2 2 cos I0 1 − u2 2 − v2 2 − u2 sin I0 . (69) (cid:19) In Figure 6 (left) we plot the level curves of the total energy in the plane (u2, v2) for Kepler-16 b (Doyle et al, 2011), adopting a rotation period of 0.5 day and k2,2 = 0.5 to compute the J2,2 value for the planet (Eq. 2), and C/(m2R2 2) = 0.2 to compute the rotational angular momentum (Table 1). Since α2 (cid:28) −g2 and I0 ≈ 1◦, the total 2 − v2 energy is dominated by the middle term U2 ≈ −g2 cos I0 2 (Eq. 69). We thus observe that the spin describes almost circular trajectories with constant obliquity. The equilibria for the spin can be obtained by finding the critical points of the total energy (Correia, 2015): (cid:112)1 − u2 ∂U2 ∂u2 = 0 ; ∂U2 ∂v2 = 0 . (70) These equations impose that v2 = 0 (which is equivalent to ϕ2 = 0), meaning that the three angular momentum vectors lie in the same plane. These equilibria are known in the literature as Cassini states (e.g. Colombo, 1966; Ward, 1975; Correia, 2015). The first equation combined with v2 = 0 gives: α2 u2 + g2 + sin I0 = 0 , (71) (cid:32) (cid:112)1 − u2 u2 cos I0 2 (cid:33) 18 Fig. 6 Secular trajectories for the spin of Kepler-16 b in the present system (left) and for a modified system with m1 = 0.8 MJup (right). We show the spin projected on the orbit normal (top), and its projection on the orbital plane (bottom). Cassini states are marked with a dot. where u2 = sin ε2 (Eq. 66). This expression is equivalent to the commonly used condi- tion to find the Cassini states equilibria (e.g Ward and Hamilton, 2004): α2 sin ε2 cos ε2 + g2 sin (ε2 + I0) = 0 . (72) When α2 (cid:28) −g2, there is one equilibrium point for ε2 ≈ −I0 (Fig. 6 left). This is often the case for circumbinary planets, so we do not expect any significant oscillations in their obliquities. In addition, as long as the system is nearly coplanar, the equilibrium obliquity remains very small. However, for α2 ≈ g2, the phase space becomes much more interesting. This situation occurs when the precession of the planet's spin and the precession of its orbit are near resonance (α2 cos ε2/L2 ≈ γ2 cos I0/G2), which can be obtained if Kepler-16's secondary is replaced by another Jupiter-mass planet. In Figure 6 (right) we plot the level curves of the total energy for a modified Kepler-16 system, with m1 = 0.8 MJup. In this case, high obliquity Cassini states are possible for the outer planet, similarly to what is observed for Saturn in the Solar System (Ward and Hamilton, 2004). 3.2.2 Stellar spins We now consider the effect of the spin of one of the stars in our analysis, for instance, the primary with mass m0. As in previous section, we restrict the orbital contribution to the quadrupolar term, so that the problem remains integrable. The total energy in this case is then given by: = − α0 2 cos2 θ0 − γ2 2 ω1 (1 + 3 2 e2 1) cos2 I . (73) U0 =(cid:10)UR,0 (cid:11) +(cid:10)(cid:10)Uqd (cid:11)(cid:11) -1-0.8-0.6-0.4-0.2 0 0.2 0.4 0.6 0.8 1-1-0.5 0 0.5 1v2-u2 0 20 40 60 80 ε2 (deg)-1-0.8-0.6-0.4-0.2 0 0.2 0.4 0.6 0.8 1-1-0.5 0 0.5 1v2-u2 0 20 40 60 80 ε2 (deg)19 This reduced expression for the energy now depends on the projection of the spin in the inner orbit cos θ0, but the problem remains integrable as the three direction cosines can be related through the total angular momentum, L0G1 cos θ0 + L0G2 cos ε0 + G1G2 cos I = K0 = cte . (74) This expression is similar to expression (60) from previous section, where the rotational angular momentum of the planet is replaced by that of the star. However, the similar- ities with the planetary case end here, because for the star the three angular momenta may present similar magnitudes, i.e., L0 ∼ G1 ∼ G2. In order to better understand the secular trajectories for the spin of the star, we can nevertheless perform a similar variable change (θ0, φ0) −→ (u0, v0) that gives the projection of the spin of the star in the inner orbit: u0 = sin θ0 cos φ0 ; v0 = sin θ0 sin φ0 , (75) where φ0 is the precession angle measured along the inner orbit from the outer orbit to the equatorial plane of the star (Fig. 2). Thus, cos θ0 = and (cid:113) 1 − u2 0 − v2 0 , (cid:113) (76) (77) (78) (79) (80) cos ε0 = cos I cos θ0 + sin I sin θ0 cos φ0 = cos I 1 − u2 0 − v2 0 + u0 sin I . Replacing previous expression for cos ε0 in expression (74) for the total angular mo- mentum provides us an expression for cos I that depends only on the new variables (u0, v0), which can be explicitly solved as (Correia, 2015) √ 1 − Z2 (G1 + L0 cos θ0)Z − L0u0 cos I = G , with and Z = Z(u0, cos θ0) = K0 − L0G1 cos θ0 G2G , (cid:112) G = G(u0, cos θ0) = (G1 + L0 cos θ0)2 + (L0u0)2 . Therefore, cos θ0 and cos I depend only on the new variables (u0, v0), as well as the total energy given by expression (73). As for the planetary spin, the equilibria for the stellar spin are obtained from the critical points of the total energy: = 0 ; = 0 . (81) (cid:12)(cid:12)(cid:12)(cid:12)u0 ∂U0 ∂u0 (cid:12)(cid:12)(cid:12)(cid:12)u0 ∂U0 ∂v0 (cid:12)(cid:12)(cid:12)(cid:12)u0 (cid:12)(cid:12)(cid:12)(cid:12)u0 Since U0(u0, v0) = U0(u0, cos θ0(u0, v0)), the second equation becomes ∂U0 ∂v0 = ∂U0 ∂(cos θ0) ∂(cos θ0) ∂v0 = − ∂U0 ∂(cos θ0) v0 cos θ0 = 0 . (82) We hence conclude that v0 = 0 is still a possible solution (which is equivalent to φ0 = 0), meaning that the three angular momentum vectors lie again in the same plane. The first equation combined with v0 = 0 thus gives a generalised version of Cassini states for the spin of the star (Correia, 2015): 20 0 , with cos θ0 =(cid:112)1 − u2 (cid:18) u0Z√ (cid:12)(cid:12)(cid:12)(cid:12)v0 f (u0) = ∂Z ∂u0 1 − Z2 L0G = G1 G2 + α0 u0 + g(u0)f (u0) = 0 , g(u0) = −γ2(1 + (cid:12)(cid:12)(cid:12)(cid:12)v0 (cid:19) ∂Z (cid:12)(cid:12)(cid:12)(cid:12)v0 ∂G ∂u0 ∂u0 , 3 2 e2 1) cos I − u0Z L0G , −(cid:112) cos θ0 and ∂G ∂u0 1 − Z2− cos I L0 ∂G ∂u0 = − G1L0u0 G cos θ0 (cid:12)(cid:12)(cid:12)(cid:12)v0 . G1 L0 + cos θ0 u0 cos θ0 − Z G (83) (84) , (85) (86) (cid:12)(cid:12)(cid:12)(cid:12)v0 In Figure 7 we plot the level curves of the total energy in the plane (u0, v0) for the standard system (Table 1), adopting different rotation periods that range from 4.2 to 4.8 day. Although these rotation periods may seem too short for Solar-type stars, they are reliable for young stars (e.g. Skumanich, 1972). Moreover, close-binary systems undergo strong tidal effects that modify the rotation period until it is close to the orbital period, which corresponds to nearly 4.5 day in the case of the standard system (see section 4). For a rotation period of 4.8 day (Fig. 7 c), the stellar spin precesses around a direc- tion close to the binary orbit's normal. The obliquity is nearly constant as well as the mutual inclination between the two orbits. However, for the transition rotation period of 4.5 day (Figs. 7 b), there is a resonance between the precession of the stellar spin and the precession of the orbits (α0 cos θ0/L0 ≈ γ2 cos I0/G2), which completely modifies the evolution of the spin. In this case, the obliquity is no longer constant, and the secular trajectories resemble those for the modified Kepler-16 system (Fig. 6, right). Moreover, unlike the Kepler-16's case, the mutual inclination between the orbits also undergoes significant variations. Thus, for a star increasing its rotation rate, the spin can be captured in resonance (Fig. 8). This event can produce a significant variations in the obliquity of the star and in the mutual inclination of the orbits. In Figure 8 we show the Cassini states equilibria for the standard system (Table 1) as a function of the rotation period of the primary star. This figure is equivalent to the classic Cassini states for the planet that are obtained when solving equation (72) as a function of the ratio α2/g2 (see, for example, Fig. 3 in Ward and Hamilton, 2004). We observe some differences in the number of Cassini states, but the key feature is the resonance near P0 ≈ 4.5 day, that allows two distinct evolutions for the stellar spin. For close-in stars, in the expression of the total energy (73), we also need to take into account the contribution of the secondary star:(cid:10)UR,1 (cid:11) = −α1(cos θ1)2/2. In this case the problem is no longer integrable. However, if the perturbation introduced by the secondary star dominates the perturbation from the planet (α1 (cid:29) γ2), we can neglect the term in γ2 and the problem becomes integrable again (see Bou´e and Laskar, 2009; Correia, 2016) by solving the following equations using the same steps explained in this section: UR =(cid:10)UR,0 (cid:11) +(cid:10)UR,1 (cid:11) = − α0 (cos θ0)2 − α1 2 (cos θ1)2 , 2 and L0G1 cos θ0 + L1G1 cos θ1 + L0L1 cos θ01 = K0 = cte , where θ01 is the angle between the spin axis of the two stars (Correia, 2016). (87) (88) 21 ) b ( , y a d 2 . 4 = t o r P ) a ( , s d o i r e p n o i t a t o r t n e r e ff d i r o f ) 1 e l b a T ( m e t s y s d r a d n a t s e h t n i ) 0 m ( r a t s n i a m e h t f o i n p s e h t r o f s e i r o t c e j a r t r a l u c e S 7 . g i F n o i t c e j o r p s t i d n a , ) e l d d m i ( l a m r o n t i b r o e h t n o d e t c e j o r p i n p s r a l l e t s e h t , ) p o t ( n o i t a n i l c n i l a u t u m e h t w o h s e W . y a d 8 . 4 = t o r P ) c ( d n a , y a d 5 . 4 = t o r P . t o d a h t i w d e k r a m e r a s e t a t s i n i s s a C . ) m o t t o b ( e n a l p l a t i b r o e h t n o Prot= 4.8 dayProt= 4.5 dayProt= 4.2 day(a)(b)(c)-0.3 0 0.3-0.3 0 0.3v0u0 0 10 20 θ0 (deg) 0 5 10 I (deg)-0.3 0 0.3-0.3 0 0.3v0u0 0 10 20 θ0 (deg) 0 5 10 I (deg)-0.3 0 0.3-0.3 0 0.3v0u0 0 10 20 θ0 (deg) 0 5 10 I (deg)22 Fig. 8 Cassini states equilibria for the standard system (Table 1) as a function of the rotation period of the primary star, P0. These equilibria are obtained by solving equation (83), and each color corresponds to a different dynamical state. 4 Tidal evolution In this section we add the contribution of tides to the secular dynamics of circumbinary systems. Tidal dissipation modifies the rotational and orbital angular momenta and the system evolves into some equilibrium configuration. For the unknown geophysical parameters, we adopt for the stars C/(mR2) = 0.08 and k2 = 0.028 (Eggleton and Kiseleva-Eggleton, 2001), and for the Jupiter-mass planets C/(mR2) = 0.25 and k2 = 0.50 (Yoder, 1995). For tidal dissipation, we adopt the constant time-lag linear model presented in section 2.4. For stars we use ∆t = 0.1 s (Penev et al, 2012), and for planets ∆t = 100 s (Lainey et al, 2009), which roughly corresponds to Q ∼ 107 and Q ∼ 104, respectively, with Q−1 ≡ n∆t. 4.1 Inner binary evolution We first look at the evolution of the inner binary without the presence of the companion planet, i.e., we restrict our analysis to the two body problem. This simplification has been widely studied (e.g., Kaula, 1964; Goldreich, 1966; Alexander, 1973; Efroimsky and Williams, 2009; Correia and Laskar, 2010; Migaszewski, 2012; Ferraz-Mello, 2013; Correia et al, 2014; Makarov, 2015), but most previous studies focus on the spin and orbital evolution of a single component disturbed by a point-mass companion. This as- sumption is usually a good approximation while studying the evolution of a star-planet system. However, it becomes less realistic for star-star systems, since both companions -0.5-0.4-0.3-0.2-0.1 0 0.1 0.2 0.3 0.4 0.5 3 3.3 3.6 3.9 4.2 4.5 4.8 5.1 5.4 5.7 6u0P0 (day)123423 Fig. 9 Tidal evolution of the inner binary in the standard system (Table 1), for a point mass primary star with no spin (k2,0 = 0), and in absence of the planetary companion (m2 = 0). We show the relative rotation rate of the secondary star (a) and its obliquity (b), the relative semi- major axis variation (c), and the relative eccentricity variation (d). The dotted line corresponds to the equilibrium rotation given by expression (89). have similar angular momenta values. As a consequence, in this case resonant inter- actions between the spin of the two stars can occur, which modify the intermediary evolution. In Figure 9 we plot the tidal evolution of the inner binary in absence of the planetary companion for the standard system (Table 1). The primary star is initially considered as a point mass with no spin, which is equivalent to assume k2,0 = 0. Therefore, all the modifications in the system result from tidal dissipation in the secondary star. Its initial rotation period of is chosen to be 1 day, corresponding to Ω1/n1 ≈ 10.5, while its initial obliquity is set at θ1 = 10◦. Tidal interactions with the primary decrease the rotation rate of the secondary (Fig. 9 a) until it reaches the "pseudo-synchronization" equilibrium rotation (e.g., Correia, 2009) Ω1 n1 = 2 cos θ1 1 + cos2 θ1 f2(e1) f1(e1) , while the obliquity tends to the equilibrium value cos θ1 = 2n1 Ω1 f2(e1) f1(e1) . (89) (90) 0 1 2 3 0 0.2 0.4 0.6 0.8 1(c)∆a1/a10 (x 10-4)time (Gyr) 0 1 2 0 0.2 0.4 0.6 0.8 1(d)∆e1/e10 (x 10-4)time (Gyr) 0 4 8 12 (a)Ω1/n1 0 5 10 15 (b)θ1 (deg)24 Fig. 10 Obliquity evolution of the inner binary in the standard system (Table 1), without dissipation in the primary star (∆t0 = 0), and in absence of the planetary companion (m2 = 0). We show the evolution for both stars using two different initial configurations of the primary star's spin: (a) initial obliquity θ0 = 5◦; (b) initial obliquity θ0 = 0.5◦. We plot the evolution through 1 Gyr (left), and near the resonance crossing around 0.4 Gyr (right). The dotted line corresponds to the obliquity of the secondary for a point mass primary (Fig. 9 b). For initial fast rotation rates (Ω1 (cid:29) n1), the equilibrium obliquity is close to 90◦, that is why we observe an initial increase in the obliquity of the secondary star (Fig. 9 b). However, as the rotation rate slows down and approaches the equilibrium value (89), the final obliquity tends to zero. Concerning the orbit, the semi-major axis and eccentricity also initially increase, because the angular momentum is transferred from the spin to the orbit. These vari- ations are almost imperceptible, because in the example shown the orbital angular momentum is much larger than the rotational angular momentum of the secondary. As the rotation rate comes close to the equilibrium value (89), the angular momentum transfer ceases, and the only consequence of tidal dissipation is to decrease the orbital energy, hence we observe a decrease in the semi-major axis (Fig. 9 c). Since the orbital angular momentum must be conserved, i.e., a1(1 − e2 1) = cte. (Eq. 5), the eccentricity also decreases (Fig. 9 d). The final evolution of the system is obtained when the orbit becomes circular, the obliquity is zero, and the rotation rate is synchronous with the mean motion (e.g., Hut, 1980; Correia, 2009), although in most stellar systems this process takes longer than the maximum life-span of the stars. 0 5 10 15 0 0.2 0.4 0.6 0.8 1(b)θ0 , θ1 (deg)time (Gyr) 0 5 10 15 0.38 0.39 0.4 0.41 0.42(b)time (Gyr) 0 5 10 15 (a)θ0 , θ1 (deg) 0 5 10 15 (a)25 In Figure 10 we plot the evolution of the binary system when the primary star is no longer a point mass object. The rotation of the primary is taken to be 10 day (Ω0 ≈ n1), while for the initial obliquity we adopt θ0 = 5◦ (Table 1), or 10 times smaller, θ0 = 0.5◦. In order to better compare with the previous simulation, we neglect tidal dissipation on the primary star (∆t0 = 0). The orbital evolution and the evolution of the rotation rate are almost identical to those shown for a point mass companion (Fig. 9), so they are not shown. However, the obliquity of the secondary can now experience quite different intermediary evolutions. Indeed, around 0.4 Gyr, there is a resonance between the precession rates of the spin of both stars α0 cos θ0/L0 ≈ α1 cos θ1/L1, which modifies their obliquities. In one case, there is no capture in resonance, and the obliquities only receive a kick (Fig. 10 a) (see also Appendix A in Laskar et al, 2004). In the other situation, capture occurs, and the obliquity of the secondary is brought near zero degrees in a time-scale much shorter than tidal effects alone (Fig. 10 b). The resonant equilibrium is only broken for low obliquity when the tidal torque becomes stronger than the precession torque. After the resonant encounter, the obliquity of the secondary decreases again towards zero degrees. The final evolution of the system is the same as the one described for a point-mass companion (Fig. 9), since it corresponds to the minimum of the total energy of the system (Hut, 1980). However, the time-scales involved can be significantly different. Moreover, when additional bodies are present, such as a circumbinary planet, they will interact with a different stellar configuration that may drastically change their future evolution (see next section). Therefore, when we inspect the tidal evolution of multi-body systems, we need to take their spin states into account. 4.2 Planetary evolution The tides raised by the planet on the stars is much weaker than mutual tides between stars, so we only take into account tidal effects raised by the stars on the planet (section 2.4). These tides can be described as the tidal effect raised by the center of mass of the inner binary in the circumbinary planet (Eq. 107). Therefore, the tidal evolution of the planet is similar to the evolution of a single star described in previous section (see also Figure 9). Moreover, the orbital angular momentum of the circumbinary planet is much larger than the rotational angular momentum, so tides on the planet are only expected to significantly modify its spin. In Figure 11 we show the spin evolution of the planet Kepler-16 b using the initial values from Table 1. Although the orbits of this system are well established (Doyle et al, 2011), we ignore the present values for the rotation period and obliquity. Assuming initial fast rotating planets like the gaseous planets of the solar system, the general trend for the spin evolution corresponds to a progressive increase in the rotation period and obliquity as the one shown in Figure 11. In the particular case of Kepler-16 b, we additionally observe a small kick on its obliquity around t ≈ 0.6 Gyr. This corresponds to a secular resonance between the precession of the spin and the frequency p−g1 +g2 ≈ −3.58(cid:48)/yr, where p is the precession frequency of the node, and g1, g2 are the precession frequencies for the pericenters of both orbits. 26 Fig. 11 Tidal evolution of the planet Kepler-16 b (Table 1). We show the relative rotation rate (a), and the obliquity (b). 4.3 Coupled evolution For binary systems with separations a1 > 0.1 AU and moderate eccentricities, the orbits are only marginally modified by tides during the age of the system (Fig. 9c,d). The spins of all bodies can be significantly modified, but they present a general trend of synchronising the rotation with the orbit and decreasing the obliquity (Fig. 9a,b). However, the spin can cross some secular resonances that may accelerate or delay its evolution (Figs. 10 and 11). In the examples previously shown, these resonances had no impact on the orbits, either because they corresponded to spin-spin interactions (Fig. 10), or because the orbital angular momentum is much larger than the rotational angular momentum (Fig. 11). 4.3.1 Secular spin-orbit resonances In Figure 12 we show the tidal evolution of the standard system (Table 1). Unlike the previous examples, we observe some resonant interactions that considerably modify the spins and the orbits. The main interactions correspond to 1) a resonance between the spin precession rate of the secondary and the precession of the node at t ≈ 85 Myr; 2) a resonance between the spin precession rate of the primary and the precession of the node for 150 (cid:46) t (cid:46) 250 Myr. A spin-spin resonance between the spin precession rates of both stars at t ≈ 125 Myr is also noticeable. The two spin-orbit resonances result in an obliquity increase of about 20 degrees. However, while for the secondary the increase occurs in a short time-scale, less than 1 Myr, for the primary it lasts for more than 100 Myr. The different behaviors cor- respond to two different kinds of resonance crossing. For the secondary, the rotation period is increasing from lower values, while for the primary it is the opposite (Fig. 12a). Therefore, the secondary cannot be trapped in resonance, while for the primary this is possible (Fig. 8). The two different evolutions can be well understood using the diagrams with all possible secular trajectories given in Figure 7. These diagrams cor- respond to the spin of the primary, but for the secondary we obtain a similar picture, 1 10 100 1000 0 5 10 15(a)Ω2/n2time (Gyr) 0 5 10 15 0 5 10 15(b)θ2 (deg)time (Gyr)27 Fig. 12 Tidal evolution of the standard system (Table 1). We show the rotation rate of the stars (a), their obliquity (b), the eccentricity of the orbits (c), and the mutual inclination (d). The dotted line corresponds to the equilibrium rotation given by expression (89). only the rotation period for which the resonance is crossed would change (≈ 1.5 day, instead of 4.5 day). In Figure 13 we show the projection of the spin of both stars during the resonant encounter. For the secondary, the spin is initially circulating around state 1. As the the rotation period increases, this state merges with the hyperbolic state 3 and the resonant equilibria disappear. The spin is then forced to circulate around state 2, with higher obliquity (Fig. 8). For the primary, the spin is initially circulating around state 2. As the rotation period decreases, it crosses the resonance around P0 ≈ 4.6 day. At this point, the spin can either start circulating around state 1 or follow librating around state 2, as it happens in our example. For rotation periods faster than the resonant period, the obliquity of state 2 increases (Fig. 8). As long as tidal dissipation is not too strong, the evolution is adiabatic and the spin remains trapped in resonance (state 2). Therefore, although tidal effects tend to damp the obliquity, we observe that the obliquity actually increases such that the resonant ratio can be maintained. More interestingly, we also observe that while the obliquity of the primary increases, the mutual inclination is damped. This unexpected behavior is also a result of a change in the topology of the system. Indeed, in Figure 7 we can see that while the obliquity 0.1 0.2 0.3 0.4 0.5 0 0.2 0.4 0.6 0.8 1(c)e2 , e1time (Gyr) 0 5 10 15 0 0.2 0.4 0.6 0.8 1(d)I (deg)time (Gyr) 0 2 4 6 8 10 (a)P1 , P0 (day) 0 5 10 15 20 25 30 (b)θ0 , θ1 (deg)28 Fig. 13 Tidal evolution of the standard system (Table 1). We show the obliquity (top) and the projection of the spin on the orbital plane (bottom) of the secondary (left) and of the primary (right) during the resonance crossing. Each dot corresponds to a different time. of state 2 increases for faster rotation periods, the corresponding mutual inclination decreases. The equilibrium inclination for a given Cassini state can be obtained directly from expression (78). In the case of the standard system we have L0 (cid:28) G2 (cid:28) G1, so we can simplify this expression as cos I ≈ Z ≈ K0 G1G2 − L0 G2 cos θ0 . (91) Assuming L0 approximately constant, we see that an increase in the obliquity cor- responds to a decrease of the mutual inclination and vice versa. However, L0 is not completely constant, since the rotation period is varying during the resonant crossing. Therefore, in Figure 14 we plot the exact solution of equation (78) as a function of the rotation period, where the obliquity curve corresponds to the obliquity of state 2 shown in Figure 8. We observe there is a very good agreement with the numerical sim- ulations shown in Figure 12. The small differences observed result from the fact that our analytical model is obtained with the quadrupolar approximation for the orbits and considering only the spin of the primary. As the rotation period decreases, the libration width of the resonant state 2 de- creases (Fig. 7a). When the obliquity is near to its maximum value (and the mutual inclination is near zero), the tidal torque becomes stronger than the resonant coupling. -0.5-0.25 0 0.25 0.5-0.5-0.25 0 0.25 0.5 70 75 80 85 90 95 100v1u1 0 5 10 15 20 25 30 70 75 80 85 90 95 100θ1 (deg)time (Myr)-0.5-0.25 0 0.25 0.5-0.5-0.2500.250.5 100 150 200 250 300 350 400v0u0 0 5 10 15 20 25 30 100 150 200 250 300 350 400θ0 (deg)time (Myr)29 Fig. 14 Cassini state 2 equilibria for the standard system (Table 1) as a function of the rotation period of the primary star, P0. These equilibria are obtained by solving equation (83). We show the obliquity of the primary (blue) and the mutual inclination (green). At this stage, the spin ends its libration around state 2, and starts to circulate around state 1. In absence of the secular resonant forcing, the obliquity is damped by tides until it reaches the equilibrium Cassini state 1 value. In Figure 15 we show the evolution of the obliquity and the mutual inclination for a modified standard system with an initial inclination of 60◦ (instead of 10◦). Although this new configuration presents larger oscillations of the mutual inclination (due to the secular quadrupolar interactions between the two orbits), its average value is also damped to zero while crossing the secular resonance with the spin of the primary. For larger initial mutual inclinations we would observe a similar behavior, at least as long as the pericenter of the inner orbit is in circulation (Fig. 4). The damping of the mutual inclination is a consequence of the angular momentum transfer between the spin of the primary and the orbit of the planet through the secular resonance. We thus conclude that this resonance is an efficient mechanism of transforming circumbinary systems with arbitrary initial mutual inclination in coplanar systems like the ones observed by the Kepler Spacecraft surveys (e.g. Doyle et al, 2011; Welsh et al, 2012; Orosz et al, 2012a,b; Kostov et al, 2015). 4.3.2 Chaotic evolution In section 3.1 we saw that there are two different dynamical regimes for the orbits: the pericenter of the inner orbit, ω1, can be in circulation or in libration around ±90◦ (Fig. 4). In previous section we have increased the initial mutual inclination of the 0 5 10 15 20 25 3 3.3 3.6 3.9 4.2 4.5 4.8 5.1 5.4 5.7 6θ0 , I (deg)P0 (day)30 Fig. 15 Tidal evolution of the standard system (Table 1) with initial mutual inclination I = 60◦. We show the obliquity of the primary (blue) and the mutual inclination (green). standard system up to 60◦, but we kept ω1 = 0◦ (Fig. 15). As a consequence, the orbits were always in circulation (Fig. 4). We now keep initial I = 60◦, but with initial ω1 = 90◦ for the standard system (Table 1), which places the system in libration. In Figure 16 we show two different evolutions with slightly different initial precession angle for the primary star, φ0 = 0.0◦ (a) and φ0 = 0.1◦ (b). The initial evolution of the system is similar to the circulation regime (Fig. 12): the average mutual inclination is constant, the spin of the secondary encounters a resonance at t ≈ 0.1 Gyr, and the primary at t ≈ 0.2 Gyr. At this stage, the obliquity of the primary increases, which is accompanied by an increase in the amplitude of the mutual inclination. The mechanism behind this exchange is the same described for the circulation regime (see Fig. 14). As the amplitude of the mutual inclination increase, the system approaches the separatrix between the libration and circulation regimes (Fig. 4). Around t ≈ 1 Gyr the stellar spins and the eccentricity of the planet's orbit become excited by several resonances and become chaotic. The chaotic regime can last for several Gyr, until the separatrix is crossed and the system enters in circulation. For long times spent in the libration region near the separatrix, the eccentricity of the outer orbit can rise to values close to the unity, so the planet can experiment close encounters with the remaining bodies and the system becomes unstable. In this case the secular model presented in this paper is no longer adapted to follow the orbits. However, as soon as the system enters in the circulation regime, the eccentricity of the planet's orbit stabilises, as well as the spins. The obliquity of all bodies is strongly chaotic in the libration region, but it becomes dominated by tides in the circulation region. 0 10 20 30 40 50 60 70 0 0.5 1 1.5 2θ0 , I (deg)time (Gyr)31 Fig. 16 Tidal evolution of the standard system (Table 1) with initial I = 60◦, ω1 = 90◦, and φ0 = 0.0◦ (a) or φ0 = 0.1◦ (b). We show the obliquity of the stars together with the mutual inclination (top), and the eccentricity of the orbits (bottom). The evolution in the chaotic zone is very sensitive to the initial conditions. In the example shown in Figure 16 we have just modified slightly the initial precession angle of the spin of the primary by ∆φ0 = 0.1◦. We have run many more simulations, with different phase angles, and also with faster and slower initial rotation rates for the stars. In half of the cases, the separatrix is crossed with inclination larger than 90◦, so the orbit of the planet becomes retrograde (Fig. 16 a), while in another half the system becomes prograde (Fig. 16 b). The initial resonant crossing at t < 0.2 Gyr can be avoided, but we observed that for systems initially in libration, the spins always become excited after some time. As a consequence, in all runs the amplitude of the mutual inclination increased and the systems ultimately quit the libration regime. Their final evolution is therefore always similar to the one shown in Figure 16. We hence conclude that tidally evolved circumbinary systems are most likely in circulation. In Figure 17 we show two evolutions for the standard system (Table 1) starting with initial I = 43.5◦ and ω1 = 90◦. These initial conditions still correspond to the circulation regime, but place the system very close to the separatrix. We adopt again slightly different initial precession angle for the primary star, φ0 = 0.0◦ (a) and φ0 = 0.1◦ (b). Moreover, we adopt an initial fast rotation period for the primary, P0 = 1.5 day. Thus, tidal effects will increase the semi-major axis and the eccentricity of 0 0.2 0.4 0.6 0.8 1 0 1 2 3 4 5(a)e2 , e1time (Gyr) 0 0.2 0.4 0.6 0.8 1 0 1 2 3 4 5(b)time (Gyr) 0 30 60 90 120 150 θ0 , θ1 , I (deg) 0 30 60 90 120 150 (b)(a)32 Fig. 17 Tidal evolution of the standard system (Table 1) with initial I = 43.5◦, ω1 = 90◦, P0 = 1.5 day, and φ0 = 0.0◦ (a) or φ0 = 0.1◦ (b). We show the obliquity of the stars together with the mutual inclination (top), and the eccentricity of the orbits (bottom). the inner orbit during the initial stages (Fig. 9), which forces the system to cross the separatrix and enter in the libration region. Once the system enters in libration, the mutual inclination undergoes variations ranging from 44◦ to 136◦ (Fig. 4), meaning that the orbit of the planet oscillates from prograde to retrograde. As in the previous example (Fig. 16), the spins of all bodies and the eccentricity of the outer orbit become chaotic. As the rotation period of the stars approach their equilibrium values, the semi-major axis and the eccentricity of the inner orbit decrease again and the system returns to circulation. However, in half of the cases, the separatrix can be crossed with high inclination and the orbit of the planet becomes retrograde (Fig. 17 a). This mechanism is then able to transform previous prograde orbits in retrograde ones, as it was already reported by Correia et al (2011). 5 Additional applications The secular model presented in this paper (section 2) is very general and therefore its validity is not restricted to circumbinary systems. Indeed, it can also be applied to any three-body hierarchical system for which terms in (r1/r2)4 can be neglected (octupolar 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1(a)e2 , e1 (deg)time (Gyr) 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1(b)time (Gyr) 0 30 60 90 120 150 θ0 , θ1 , I (deg) 0 30 60 90 120 150 (b)(a)33 Fig. 18 Tidal evolution of a fictitious system with a Sun-like star and two Jupiter-mass planetary companions using the same initial conditions as for Figure 2 in Naoz et al (2011). approximation), as well as the torque of the outermost body on the spin of the inner bodies. Nevertheless, this last effect can also be considered as in Correia et al (2011), by keeping in the rotational potential (23) the contribution from m2 (Eq. 100): UR,i = G m0m1 r1 J2,i P2(r1 · si) + m2 mj P2(r2 · si) . (92) (cid:19)2(cid:34) (cid:18) Ri r1 (cid:19)3 (cid:18) r1 r2 (cid:35) A straightforward application is, for instance, the formation of hot Jupiters from secular planet-planet interactions (Naoz et al, 2011; Beaug´e and Nesvorn´y, 2012). In Figure 18 we reproduce a simulation for a Sun-like star with two Jupiter-mass planetary companions using the same initial conditions as for Figure 2 by Naoz et al (2011). We observe there is a good agreement between the two models3. Another suitable applications are star-planet-satellite systems and triple stellar systems. 3 In order to reproduce the results in Naoz et al (2011) we cannot take into account the flattening of the star (Eq. 23). The evolution is also highly chaotic, so a slightly change in the reference angles lead to different final mutual inclination. 10-310-210-1100 0 20 40 60 80 100 1201 - e1time (Myr) 0 30 60 90 120 150 180 I (deg)34 6 Conclusion In this paper we provide a secular model to describe the evolution of circumbinary three-body systems, where all bodies undergo tidal interactions. We use the octupolar non-restricted approximation for the orbital interactions, including general relativity corrections, the quadrupolar approximation for the spins, and the viscous linear model for tides. Many of the already known circumbinary planets evolve in coplanar orbits, since they were detected through the transits method. However, this is not necessarily the standard configuration for circumbinary planets (e.g. Martin and Triaud, 2014). Our model is suitable to study the long-term evolution of a wide variety of circumbinary systems in very eccentric and inclined orbits. We have shown that tidal effects coupled with the secular evolution can totally modify the final configuration of the system. For instance, tides alone are unable to damp the mutual inclination of planetary systems during their life-times. A most strik- ing example is that circumbinary planets with initial arbitrary orbital inclination can become coplanar through a secular resonance between the precession of the spin of one star and the precession of the orbit. We also show that circumbinary systems for which the pericenter of the inner orbit is initially in libration present chaotic motion for the spins and for the eccentricity of the outer orbit. We have presented in this paper a few examples, which are representative of the diversity of behaviors among circubinary systems. Many other systems are awaiting to be studied. The fact that we use average equations for both tidal and gravitational effects, makes our method suitable to be applied in long-term studies. It allows to run many simulations for different initial conditions in order to explore the entire phase space and evolutionary scenarios. In particular, it can be very useful to derive constraints for the past and future tidal evolution of circumbinary systems. Our model can also be applied to any three-body hierarchical system such as star-planet-satellite systems and triple stellar systems. Acknowledgements We acknowledge support from PNP-CNRS, and from from CIDMA strategic project UID/MAT/04106/2013. A Oblate spheroid potential The gravitational potential of an oblate body of mass mi symmetric about its rotation axis s is given by (e.g. Goldstein, 1950): (cid:34) (cid:18) Ri (cid:19)2 r (cid:35) Vi(r) = −G mi r 1 − J2,i P2(r · si) , (93) where we neglected terms in (Ri/r)3. The gravity field coefficient J2,i is obtained from the principal moments of inertia Ai = Bi and Ci as J2,i = (Ci − Ai)/miR2 i . When the asymmetry in the body mass distribution results only from its rotation, J2,i is given by expression (2). The main term in the above expression is responsible for the orbital motion (Eq. 4), while the contribution in J2,i is responsible for a perturbation of this motion, since J2,i(Ri/r)2 (cid:28) 1. Thus, retaining only the terms in J2,i, the resulting perturbing potential energy of a system composed of three oblate bodies is given by: UR = UR,0 + UR,1 + UR,2 , (94) where we have for the planet (cid:88) UR,2 = miV2(r2i) = i=0,1 i=0,1 35 (cid:18) R2 (cid:19)2 r2i G mim2 r2i J2,2 P2(r2i · s2) , (95) (cid:88) and for each star (i = 0, 1) UR,i = G m0m1 r1 J2,i We also have (Fig. 1) (cid:18) Ri (cid:19)2 r1 P2(r1 · si) + G mim2 r2i J2,i (cid:18) Ri (cid:19)2 r2i P2(r2i · si) . (96) r2i = r2 + δir1 , where δ0 = m1/m01 and δ1 = −m0/m01. Since we assume that r1 (cid:28) r2, we can write (cid:2)r1 · r2 − 5(r1 · r2)(r2 · sj )2 + 2(r1 · sj )(r2 · sj )(cid:3) , P2(r2i · sj ) r3 2i ≈ P2(r2 · sj ) r3 2 + 3δi 2r3 2 r1 r2 (97) (98) where we neglected terms in (r1/r2)2, that is, we neglect terms in J2,i(Ri/ri)2(r1/r2)2 in the potential energy. Replacing in expressions (96) and (95) we get for the planet (cid:18) R2 (cid:19)2 r2 UR,2 = G m2m01 r2 J2,2 P2(r2 · s2) , (99) since m0δ0 + m1δ1 = 0, and for each star (i = 0, 1; j = 1 − i) (cid:19)2(cid:34) (cid:19)2 (cid:18) Ri (cid:18) Ri r1 r1 P2(r1 · si) + m2 mj P2(r1 · si) , (cid:18) r1 r2 (cid:19)3 (cid:35) P2(r2 · si) (100) UR,i = G m0m1 r1 J2,i ≈ G m0m1 r1 J2,i since terms in m2/mj (r1/r2)3 can also be neglected. B Tidal potential The tidal potential of a body of mass mi when deformed by another body of mass m(cid:48) at the position r(cid:48) is given by (e.g. Kaula, 1964): (101) where we neglected terms in (Ri/r)4(Ri/r(cid:48))4. The resulting perturbing potential energy of a system composed of three bodies is given by: Vi(r, r(cid:48), m(cid:48)) = −k2,i Gm(cid:48)R5 r3r(cid:48)3 i P2(r · r(cid:48)) , UT = UT,0 + UT,1 + UT,2 , where we have for the planet (cid:88) UT,2 = miV2(r2i, r(cid:48) 2j , mj ) = i,j=0,1 i,j=0,1 Gmimj R5 2 2ir(cid:48)3 r3 2j (cid:88) −k2,2 (102) (103) P2(r2i · r(cid:48) 2j ) , and for each star (i = 0, 1; j = 1 − i) UT,i = mj 1, mj ) + Vi(r1, r(cid:48) (cid:2)Vi(r1, r(cid:48) 2i, m2)(cid:3) + m2 (cid:2)Vi(r2i, r(cid:48) 1, mj ) + Vi(r2i, r(cid:48) 2i, m2)(cid:3) .(104) 36 Neglecting the tidal interactions with the external body m2, i.e., neglecting terms in m2/mj (r1/r2)3, the above potential can be simplified as UT,i ≈ miVi(r1, r(cid:48) 1, mj ) = −k2,i j R5 Gm2 i 1r(cid:48)3 r3 1 P2(r1 · r(cid:48) 1) . (105) Using expression (97) we can rewrite P2(r2i · r(cid:48) 2ir(cid:48)3 r3 2j ) + 2j 2 ≈ P2(r2 · r(cid:48) 2) 2r(cid:48)3 r3 r(cid:48) 3δj 2r(cid:48)3 r(cid:48) 2r3 + 2 2 2 1 r1 r2 3δi 2r(cid:48)3 2r3 2 − 5(r(cid:48) 1 · r(cid:48) (cid:2)r(cid:48) (cid:2)r1 · r2 − 5(r1 · r2)(r2 · r(cid:48) 2)2 + 2(r1 · r(cid:48) 2)(r2 · r(cid:48) 1 · r(cid:48) 2)(r2 · r(cid:48) 2)2 + 2(r2 · r(cid:48) 1)(r2 · r(cid:48) 2)(cid:3) , 2)(cid:3) (106) where we neglected terms in (r1/r2)2, that is, we neglect terms in (R2/r2)6(r1/r2)2 in the potential energy. Replacing in expression (103) we get for the planet UT,2 = −k2,2 Gm2 01R5 2 2r(cid:48)3 r3 2 P2(r2 · r(cid:48) 2) , (107) since m2 0δ0 + m0m1(δ0 + δ1) + m2 1δ1 = 0. C Averaged quantities For completeness, we gather here the average formulae that are used in the computation of secular equations. Let F (r, r) be a function of a position vector r and velocity r, its averaged expression over the mean anomaly (M ) is given by (cid:90) 2π 0 (cid:104)F(cid:105)M = 1 2π F (r, r) dM . (108) Depending on the case, this integral is computing using the eccentric anomaly (E), or the true anomaly (v) as an intermediate variable. The basic formulae are dM = r a dE = r2 1 − e2 dv , r = a(cos E − e) e + a 1 − e2(sin E) k × e , a2(cid:112) √ r = r cos v e + r sin v k × e , k × (r + e) , na(cid:112) r = 1 − e2 r = a(1 − e cos E) = a(1 − e2) 1 + e cos v , (109) where k is the unit vector of the orbital angular momentum, and e the Laplace-Runge-Lenz vector (Eq. 6). We have then (cid:42) (cid:43) 1 r3 = 1 a3(1 − e2)3/2 , and = 1 2a3(1 − e2)3/2 where ut denotes the transpose of any vector u. This leads to (cid:28) rrt (cid:29) r5 (cid:16) 1 − kkt(cid:17) , (110) (cid:42) (cid:43) r3 P2(r · u) 1 = − 1 2a3(1 − e2)3/2 P2(k · u) , (111) for any unit vector u. In the same way, (cid:10)r2(cid:11) = a2 give (cid:10)rrt(cid:11) = a2 1 − e2 (cid:18) 2 The other useful formulae are 37 a2eet , (112) (cid:16) 1 − kkt(cid:17) + 5 2 (cid:19) (1 − e2)P2(k · u) − 5e2P2(e · u) . (113) , e2 3 2 1 + and (cid:19) (cid:18) (cid:10)r2P2(r · u)(cid:11) = − a2 (cid:42) (cid:43) (cid:42) 2 (cid:42) (cid:43) rrt r8 (cid:112) 1 r8 1 − e2 2a6 (cid:43) 1 r6 = = (cid:42) f4(e) r r8 = 5 2 (cid:42) (cid:42) (cid:43) (cid:43) (cid:43) r r10 (r · r)r r10 f2(e) , + 6 + e2 4a6(1 − e2)9/2 eet , = 1 a6 f1(e) , a8 √ 1 1 − e2 1 − kkt(cid:17) (cid:16) a7(cid:112) 1 1 − e2 1 a9(1 − e2) f4(e)e , f5(e)e , n 1 − e2 f5(e) k × e , = 7 2 = 2a7(cid:112) (cid:90) 2π (cid:18) (cid:10)eet(cid:11) ω = 1 2π 0 eet dω = e2 2 (cid:0)1 − kkt(cid:1) , (cid:19) (cid:104)(e · u) e(cid:105)ω = e2 2 u − (k · u)k . (114) (115) (116) (117) (118) (119) (120) (121) where the fi(e) functions are given by expressions (46) to (50). Finally, for the average over the argument of the pericenter (ω), we can proceed in an identical manner: which gives References Alexander ME (1973) The Weak Friction Approximation and Tidal Evolution in Close Binary Systems. Astrophys. Space Sci. 23:459–510, DOI 10.1007/BF00645172 Beaug´e C, Nesvorn´y D (2012) Multiple-planet Scattering and the Origin of Hot Jupiters. Astrophys. J. 751:119, DOI 10.1088/0004-637X/751/2/119, 1110.4392 Bosanac N, Howell KC, Fischbach E (2015) Stability of orbits near large mass ratio bi- nary systems. Celestial Mechanics and Dynamical Astronomy 122:27–52, DOI 10.1007/ s10569-015-9607-6 Bou´e G, Fabrycky DC (2014) Compact Planetary Systems Perturbed by an Inclined Compan- ion. II. Stellar Spin-Orbit Evolution. Astrophys. J. 789:111, DOI 10.1088/0004-637X/789/ 2/111, 1405.7636 Bou´e G, Laskar J (2006) Precession of a planet with a satellite. Icarus 185:312–330, DOI 10.1016/j.icarus.2006.07.019 Bou´e G, Laskar J (2009) Spin axis evolution of two interacting bodies. Icarus 201:750–767, DOI 10.1016/j.icarus.2009.02.001 Brozovi´c M, Showalter MR, Jacobson RA, Buie MW (2015) The orbits and masses of satellites of Pluto. Icarus 246:317–329, DOI 10.1016/j.icarus.2014.03.015 Colombo G (1966) Cassini's second and third laws. Astron. J. 71:891–896 Correia ACM (2009) Secular Evolution of a Satellite by Tidal Effect: Application to Triton. Astrophys. J. 704:L1–L4, DOI 10.1088/0004-637X/704/1/L1, 0909.4210 38 Correia ACM (2015) Stellar and planetary Cassini states. Astron. Astrophys. 582:A69, DOI 10.1051/0004-6361/201525939 Correia ACM (2016) Cassini states for black hole binaries. Mon. Not. R. Astron. Soc. 457:L49– L53, DOI 10.1093/mnrasl/slv198, 1511.01890 Correia ACM, Laskar J (2010) Tidal Evolution of Exoplanets. In: Exoplanets, University of Arizona Press, pp 534–575 Correia ACM, Laskar J, Farago F, Bou´e G (2011) Tidal evolution of hierarchical and in- clined systems. Celestial Mechanics and Dynamical Astronomy 111:105–130, DOI 10.1007/ s10569-011-9368-9, 1107.0736 Correia ACM, Bou´e G, Laskar J (2012) Pumping the Eccentricity of Exoplanets by Tidal Effect. Astrophys. J. 744:L23, DOI 10.1088/2041-8205/744/2/L23, 1111.5486 Correia ACM, Bou´e G, Laskar J, Morais MHM (2013) Tidal damping of the mutual inclination in hierarchical systems. Astron. Astrophys. 553:A39, DOI 10.1051/0004-6361/201220482, 1303.0864 Correia ACM, Bou´e G, Laskar J, Rodr´ıguez A (2014) Deformation and tidal evolution of close- in planets and satellites using a Maxwell viscoelastic rheology. Astron. Astrophys. 571:A50, DOI 10.1051/0004-6361/201424211, 1411.1860 Correia ACM, Leleu A, Rambaux N, Robutel P (2015) Spin-orbit coupling and chaotic rotation for circumbinary bodies. Application to the small satellites of the Pluto-Charon system. Astron. Astrophys. 580:L14, DOI 10.1051/0004-6361/201526800, 1506.06733 Couetdic J, Laskar J, Correia ACM, Mayor M, Udry S (2010) Dynamical stability analysis of the HD 202206 system and constraints to the planetary orbits. Astron. Astrophys. 519:A10, DOI 10.1051/0004-6361/200913635, 0911.1963 Doolin S, Blundell KM (2011) The dynamics and stability of circumbinary orbits. Mon. Not. R. Astron. Soc. 418:2656–2668, DOI 10.1111/j.1365-2966.2011.19657.x, 1108.4144 Doyle LR, Carter JA, Fabrycky DC, Slawson RW, Howell SB, Winn JN, Orosz JA, Prsa A, Welsh WF, Quinn SN, Latham D, Torres G, Buchhave LA, Marcy GW, Fortney JJ, Shporer A, Ford EB, Lissauer JJ, Ragozzine D, Rucker M, Batalha N, Jenkins JM, Borucki WJ, Koch D, Middour CK, Hall JR, McCauliff S, Fanelli MN, Quintana EV, Holman MJ, Caldwell DA, Still M, Stefanik RP, Brown WR, Esquerdo GA, Tang S, Furesz G, Geary JC, Berlind P, Calkins ML, Short DR, Steffen JH, Sasselov D, Dunham EW, Cochran WD, Boss A, Haas MR, Buzasi D, Fischer D (2011) Kepler-16: A Transiting Circumbinary Planet. Science 333:1602–, DOI 10.1126/science.1210923, 1109.3432 Efroimsky M, Williams JG (2009) Tidal torques: a critical review of some techniques. Celestial Mechanics and Dynamical Astronomy 104:257–289, DOI 10.1007/s10569-009-9204-7, 0803. 3299 Eggleton PP, Kiseleva-Eggleton L (2001) Orbital Evolution in Binary and Triple Stars, with an Application to SS Lacertae. Astrophys. J. 562:1012–1030, DOI 10.1086/323843, arXiv: astro-ph/0104126 Farago F, Laskar J (2010) High-inclination orbits in the secular quadrupolar three-body problem. Mon. Not. R. Astron. Soc. 401:1189–1198, DOI 10.1111/j.1365-2966.2009.15711.x, 0909.2287 Ferraz-Mello S (2013) Tidal synchronization of close-in satellites and exoplanets. A rheo- physical approach. Celestial Mechanics and Dynamical Astronomy 116:109–140, DOI 10.1007/s10569-013-9482-y, 1204.3957 Ford EB, Joshi KJ, Rasio FA, Zbarsky B (2000a) Theoretical Implications of the PSR B1620-26 Triple System and Its Planet. Astrophys. J. 528:336–350, DOI 10.1086/308167, astro-ph/ 9905347 Ford EB, Kozinsky B, Rasio FA (2000b) Secular Evolution of Hierarchical Triple Star Systems. Astrophys. J. 535:385–401, DOI 10.1086/308815 Goldreich P (1966) History of the Lunar Orbit. Reviews of Geophysics and Space Physics 4:411–439, DOI 10.1029/RG004i004p00411 Goldstein H (1950) Classical mechanics. Addison-Wesley, Reading Harrington RS (1968) Dynamical evolution of triple stars. Astron. J. 73:190–194, DOI 10.1086/ 110614 Hut P (1980) Stability of tidal equilibrium. Astron. Astrophys. 92:167–170 Kaula WM (1964) Tidal dissipation by solid friction and the resulting orbital evolution. Rev. Geophys. 2:661–685 Kennedy GM, Wyatt MC, Sibthorpe B, Duchene G, Kalas P, Matthews BC, Greaves JS, Su KYL, Fitzgerald MP (2012) 99 Herculis: host to a circumbinary polar-ring debris disc. Mon. 39 Not. R. Astron. Soc. 421:2264–2276, DOI 10.1111/j.1365-2966.2012.20448.x, 1201.1911 Kidder LE (1995) Coalescing binary systems of compact objects to (post)5/2-Newtonian order. V. Spin effects. Phys. Rev. D 52:821–847, DOI 10.1103/PhysRevD.52.821, gr-qc/9506022 Kostov VB, Orosz JA, Welsh WF, Doyle LR, Fabrycky DC, Haghighipour N, Quarles B, Short DR, Cochran WD, Endl M, Ford EB, Gregorio J, Hinse TC, Isaacson H, Jenkins JM, Jensen ELN, Kull I, Latham DW, Lissauer JJ, Marcy GW, Mazeh T, Muller TWA, Pepper J, Quinn SN, Ragozzine D, Shporer A, Steffen JH, Torres G, Windmiller G, Borucki WJ (2015) KOI-2939b: the largest and longest-period Kepler transiting circumbinary planet. ArXiv e-prints 1512.00189 Kozai Y (1962) Secular perturbations of asteroids with high inclination and eccentricity. As- tron. J. 67:591–598, DOI 10.1086/108790 Lainey V, Arlot JE, Karatekin O, van Hoolst T (2009) Strong tidal dissipation in Io and Jupiter from astrometric observations. Nature 459:957–959, DOI 10.1038/nature08108 Lainey V, Karatekin O, Desmars J, Charnoz S, Arlot JE, Emelyanov N, Le Poncin-Lafitte C, Mathis S, Remus F, Tobie G, Zahn JP (2012) Strong Tidal Dissipation in Saturn and Constraints on Enceladus' Thermal State from Astrometry. Astrophys. J. 752:14, DOI 10.1088/0004-637X/752/1/14, 1204.0895 Lambeck K (1988) Geophysical geodesy : the slow deformations of the earth Lambeck. Oxford [England] : Clarendon Press ; New York : Oxford University Press, 1988. Laskar J (2000) On the Spacing of Planetary Systems. Physical Review Letters 84:3240–3243 Laskar J, Bou´e G (2010) Explicit expansion of the three-body disturbing function for arbi- trary eccentricities and inclinations. Astron. Astrophys. 522:A60, DOI 10.1051/0004-6361/ 201014496, 1008.2947 Laskar J, Robutel P, Joutel F, Gastineau M, Correia ACM, Levrard B (2004) A long-term numerical solution for the insolation quantities of the Earth. Astron. Astrophys. 428:261– 285, DOI 10.1051/0004-6361:20041335 Lee MH, Peale SJ (2003) Secular Evolution of Hierarchical Planetary Systems. Astrophys. J. 592:1201–1216, DOI 10.1086/375857, arXiv:astro-ph/0304454 Lidov ML (1962) The evolution of orbits of artificial satellites of planets under the action of gravitational perturbations of external bodies. Plan. Space Sci. 9:719–759, DOI 10.1016/ 0032-0633(62)90129-0 Lidov ML, Ziglin SL (1976) Non-restricted double-averaged three body problem in Hill's case. Celestial Mechanics 13:471–489, DOI 10.1007/BF01229100 MacDonald GJF (1964) Tidal friction. Revs Geophys 2:467–541 Makarov VV (2015) Equilibrium rotation of semiliquid exoplanets and satellites. ArXiv e-prints 1507.07383 Marchal C (1990) The Three-Body Problem. Elsevier, Amsterdam Martin DV, Triaud AHMJ (2014) Planets transiting non-eclipsing binaries. Astron. Astrophys. 570:A91, DOI 10.1051/0004-6361/201323112, 1404.5360 Migaszewski C (2012) The generalized non-conservative model of a 1-planet system re- visited. Celestial Mechanics and Dynamical Astronomy 113:169–203, DOI 10.1007/ s10569-012-9413-3, 1203.2358 Migaszewski C, Go´zdziewski K (2011) The non-resonant, relativistic dynamics of circumbinary planets. Mon. Not. R. Astron. Soc. 411:565–583, DOI 10.1111/j.1365-2966.2010.17702.x, 1006.5961 Mignard F (1979) The evolution of the lunar orbit revisited. I. Moon and Planets 20:301–315 Naoz S, Farr WM, Lithwick Y, Rasio FA, Teyssandier J (2011) Hot Jupiters from secular planet-planet interactions. Nature 473:187–189, DOI 10.1038/nature10076, 1011.2501 Orosz JA, Welsh WF, Carter JA, Brugamyer E, Buchhave LA, Cochran WD, Endl M, Ford EB, MacQueen P, Short DR, Torres G, Windmiller G, Agol E, Barclay T, Caldwell DA, Clarke BD, Doyle LR, Fabrycky DC, Geary JC, Haghighipour N, Holman MJ, Ibrahim KA, Jenkins JM, Kinemuchi K, Li J, Lissauer JJ, Prsa A, Ragozzine D, Shporer A, Still M, Wade RA (2012a) The Neptune-sized Circumbinary Planet Kepler-38b. Astrophys. J. 758:87, DOI 10.1088/0004-637X/758/2/87, 1208.3712 Orosz JA, Welsh WF, Carter JA, Fabrycky DC, Cochran WD, Endl M, Ford EB, Haghighipour N, MacQueen PJ, Mazeh T, Sanchis-Ojeda R, Short DR, Torres G, Agol E, Buchhave LA, Doyle LR, Isaacson H, Lissauer JJ, Marcy GW, Shporer A, Windmiller G, Barclay T, Boss AP, Clarke BD, Fortney J, Geary JC, Holman MJ, Huber D, Jenkins JM, Kinemuchi K, Kruse E, Ragozzine D, Sasselov D, Still M, Tenenbaum P, Uddin K, Winn JN, Koch DG, Borucki WJ (2012b) Kepler-47: A Transiting Circumbinary Multiplanet System. Science 40 337:1511–, DOI 10.1126/science.1228380, 1208.5489 Penev K, Jackson B, Spada F, Thom N (2012) Constraining Tidal Dissipation in Stars from the Destruction Rates of Exoplanets. Astrophys. J. 751:96, DOI 10.1088/0004-637X/751/2/96, 1205.1803 Plavchan P, Guth T, Laohakunakorn N, Parks JR (2013) The identification of 93 day periodic photometric variability for YSO YLW 16A. Astron. Astrophys. 554:A110, DOI 10.1051/ 0004-6361/201220747, 1304.2398 Poincar´e H (1905) Le¸cons de M´ecanique C´eleste, Tome I. Gauthier-Villars. Paris Singer SF (1968) The Origin of the Moon and Geophysical Consequences. Geophys. J. R. Astron. Soc. 15:205–226 Skumanich A (1972) Time Scales for CA II Emission Decay, Rotational Braking, and Lithium Depletion. Astrophys. J. 171:565, DOI 10.1086/151310 Smart WM (1953) Celestial Mechanics. London, New York, Longmans, Green Touma JR, Tremaine S, Kazandjian MV (2009) Gauss's method for secular dynamics, softened. Mon. Not. R. Astron. Soc. 394:1085–1108, DOI 10.1111/j.1365-2966.2009.14409.x, 0811. 2812 Verrier PE, Evans NW (2009) High-inclination planets and asteroids in multistellar systems. Mon. Not. R. Astron. Soc. 394:1721–1726, DOI 10.1111/j.1365-2966.2009.14446.x, 0812. 4528 Ward WR (1975) Tidal friction and generalized Cassini's laws in the solar system. Astron. J. 80:64–70 Ward WR, Hamilton DP (2004) Tilting Saturn. I. Analytic Model. Astron. J. 128:2501–2509, DOI 10.1086/424533 Welsh WF, Orosz JA, Carter JA, Fabrycky DC, Ford EB, Lissauer JJ, Prsa A, Quinn SN, Ragozzine D, Short DR, Torres G, Winn JN, Doyle LR, Barclay T, Batalha N, Bloemen S, Brugamyer E, Buchhave LA, Caldwell C, Caldwell DA, Christiansen JL, Ciardi DR, Cochran WD, Endl M, Fortney JJ, Gautier TN III, Gilliland RL, Haas MR, Hall JR, Holman MJ, Howard AW, Howell SB, Isaacson H, Jenkins JM, Klaus TC, Latham DW, Li J, Marcy GW, Mazeh T, Quintana EV, Robertson P, Shporer A, Steffen JH, Windmiller G, Koch DG, Borucki WJ (2012) Transiting circumbinary planets Kepler-34 b and Kepler-35 b. Nature 481:475–479, DOI 10.1038/nature10768, 1204.3955 Winn JN, Albrecht S, Johnson JA, Torres G, Cochran WD, Marcy GW, Howard AW, Isaacson H, Fischer D, Doyle L, Welsh W, Carter JA, Fabrycky DC, Ragozzine D, Quinn SN, Shporer A, Howell SB, Latham DW, Orosz J, Prsa A, Slawson RW, Borucki WJ, Koch D, Barclay T, Boss AP, Christensen-Dalsgaard J, Girouard FR, Jenkins J, Klaus TC, Meibom S, Morris RL, Sasselov D, Still M, Van Cleve J (2011) Spin-Orbit Alignment for the Circumbinary Planet Host Kepler-16 A. Astrophys. J. 741:L1, DOI 10.1088/2041-8205/741/1/L1, 1109. 3198 Yoder CF (1995) Astrometric and geodetic properties of Earth and the Solar System. In: Global Earth Physics: A Handbook of Physical Constants, American Geophysical Union, Washington D.C, pp 1–31
1508.00931	1	1508	2015-08-04T22:12:15	The Solar System as an Exoplanetary System	[ "astro-ph.EP" ]	With the availability of considerably more data, we revisit the question of how special our Solar System is, compared to observed exoplanetary systems. To this goal, we employ a mathematical transformation that allows for a meaningful, statistical comparison. We find that the masses and densities of the giant planets in our Solar System are very typical, as is the age of the Solar System. While the orbital location of Jupiter is somewhat of an outlier, this is most likely due to strong selection effects towards short-period planets. The eccentricities of the planets in our Solar System are relatively small compared to those in observed exosolar systems, but still consistent with the expectations for an 8-planet system (and could, in addition, reflect a selection bias towards high-eccentricity planets). The two characteristics of the Solar System that we find to be most special are the lack of super-Earths with orbital periods of days to months and the general lack of planets inside of the orbital radius of Mercury. Overall, we conclude that in terms of its broad characteristics our Solar System is not expected to be extremely rare, allowing for a level of optimism in the search for extrasolar life.	astro-ph.EP	astro-ph	Draft version August 6, 2015 Preprint typeset using LATEX style emulateapj v. 5/2/11 THE SOLAR SYSTEM AS AN EXOPLANETARY SYSTEM 1Department of Physics and Astronomy, University of Nevada, Las Vegas, 4505 South Maryland Parkway, Las Vegas, NV 89154, USA and 2Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218, USA Rebecca G. Martin1 and Mario Livio2 Draft version August 6, 2015 ABSTRACT With the availability of considerably more data, we revisit the question of how special our Solar System is, compared to observed exoplanetary systems. To this goal, we employ a mathematical transformation that allows for a meaningful, statistical comparison. We find that the masses and densities of the giant planets in our Solar System are very typical, as is the age of the Solar System. While the orbital location of Jupiter is somewhat of an outlier, this is most likely due to strong selection effects towards short-period planets. The eccentricities of the planets in our Solar System are relatively small compared to those in observed exosolar systems, but still consistent with the expectations for an 8-planet system (and could, in addition, reflect a selection bias towards high-eccentricity planets). The two characteristics of the Solar System that we find to be most special are the lack of super-Earths with orbital periods of days to months and the general lack of planets inside of the orbital radius of Mercury. Overall, we conclude that in terms of its broad characteristics our Solar System is not expected to be extremely rare, allowing for a level of optimism in the search for extrasolar life. Subject headings: planetary systems – planets and satellites: formation – protoplanetary disks 1. INTRODUCTION of thousands of The discovery extrasolar (see, 2014; Rowe et al. 2014; Han et al. plan- e.g., ets and planet candidates in recent years 2014; Wright et al. 2011; Batalha 2014, Lissauer, Dawson & Tremaine and references therein and see exoplanet.org for a complete list), coupled with the rapidly increasing interest in the potential existence of extrasolar life, raise again in a big way the question of whether or not our Solar System is special in any sense. More specifically, we are interested in understanding whether the planetary and orbital properties in our Solar System are typical or extremely unusual compared to those of extrasolar planets. The Solar System contains eight planets and two main belts (the asteroid belt and the Kuiper belt). While tens of debris disks and warm dust belts (similar perhaps to the Solar Sys- tem’s asteroid belt) have been observed and resolved, belts with dust masses as low as those in the Solar System would currently be undetectable in extrasolar systems (e.g. Wyatt 2008; Pani´c et al. 2013). Consequently, we can quantitatively assess in detail how special the Solar System is, only on the basis of its planetary components and properties such as its age and metallicity. However, there are now hundreds of un- resolved debris disk candidates (e.g. Chen et al. 2014). Of these, about two thirds of the systems are better modelled by a two component dust disk rather than a single dust disk. The two temperature components likely arise from two separate belts (Kennedy & Wyatt 2014). Thus, the two belt configura- tion of our own Solar System is plausibly fairly typical. Beer et al. (2004) made an initial attempt to explore to what extent Jupiter’s periastron could be considered atypi- cal compared to those of the giant planets known at the time. Their analysis, however, included only fewer than 100 exo- planets, most of which had been detected via radial veloc- ity measurements. Consequently, selection effects dominated their conclusions—a possibility fully acknowledged by the authors. In the present work we re-examine the question of how special the Solar System is. In Section 2 we use the much larger currently available database to consider the planetary orbital parameters. We identify the semi–major axis of the innermost planet as the most discrepant characteristic of the solar system and the low mean eccentricity as being somewhat special. In Section 3 we compare the masses and densities of the planets in our Solar System with those in exosolar sys- tems. While the lack of a super–Earth in the Solar System is somewhat unusual, we argue that none of the characteristics identified make the Solar System very special. We discuss po- tential implications of our results in Section 4. Some of the apparent differences between the Solar System and exoplane- tary systems continue to be driven by strong selection effects that affect the sample. We draw our conclusions in Section 5. 2. PLANETARY ORBITAL PROPERTIES We first consider the statistical distribution of orbital sep- aration and eccentricity of the observed planetary orbits. To allow for a more meaningful quantitative analysis, we perform a Box & Cox (1964) transformation on the data. This trans- formation makes the data closer to a normal distribution so that we can more accurately evaluate properties such as the mean and the standard deviation. The transformation takes a skewed data set to approximate normality. It is based on the geometric mean of the measurements and is independent of measurement units. It is possible to do a multivariate Box- Cox transformation (e.g. Velilla 1993). However, because of the selection effects associated with different parameters we choose to consider each parameter separately. We transform the data with the function if λ , 0 log a if λ = 0, yλ(a) = ( aλ−1 λ (1) where a is the parameter we are examining, such as the semi- major axis or eccentricity, and λ is a constant that depends upon the original distribution, that we discuss below. The maximum likelihood estimator of the mean of the transformed data is yλ = yλ,i n , (2) n Xi=1 2 where yλ,i = yλ(ai) and ai is the i-th measurement of a to- tal of n. Similarly, the maximum likelihood estimator of the variance of the transformed data is s2 λ = n Xi=1 (yλ,i − yλ)2 n . (3) We choose λ such that we maximise the log likelihood func- tion l(λ) = − n 2 log(2π) − n 2 − n 2 log s2 λ + (λ − 1) n Xi=1 log ai. (4) This new distribution, yλ(a), will be an exact normal distribu- tion if λ = 0 or 1/λ is an even integer. We can measure how well the transformed distribution compares to a normal distribution with two parameters. The skewness is , (5) S = n Xi=1 1 n yλ,i − yλ σ !3 where σ is the standard deviation of the distribution and thus we take σ = sλ. The skewness measures the asymmetry of the distribution, a positive number implying the right hand tail of the distribution is longer, and a negative number that the left hand tail is longer. Furthermore, we can compare the kurtosis, K = n Xi=1 1 n yλ,i − yλ σ !4 − 3. (6) This is a measure of the “peakedness” of the distribution and the heaviness of the tails. A normal distribution has a kurto- sis value of zero. A positive value means that the distribution is tightly peaked but the tails are broad, and vice versa for a negative value. In the following subsections we take the sam- ples of the eccentricity and semi-major axis of the observed exoplanets and compare them to the planets in our own Solar System. 2.1. Eccentricity There are a total of 539 exoplanets with a measured eccen- tricity. In the left hand panel of Fig. 1 we show the eccen- tricity distribution for this sample. We find the maximum of the log likelihood function to be when λ = 0.30. The right hand panel of Fig. 1 shows the histogram of the Box- Cox transformed data. For our data, we find the skewness to be S = −0.065. For a normal distribution we expect the magnitude of the skewness to be up to √6/n = 0.11. Thus the data are not heavily skewed. The kurtosis is K = −0.53 whereas for a normal distribution we would expect the magni- tude to have values up to √24/n = 0.21. Thus, the data have a slightly large kurtosis, or the distribution is not so tightly peaked as a normal one. For the transformed data, the mean is −1.42 and the standard deviation is 0.60. Jupiter lies at −0.97σ from the mean. Similarly the Earth lies at −1.60σ. Thus, the eccentricities of the planets in our Solar System (that range from Venus at e = 0.0068 to Mercury at e = 0.21) are all relatively small compared to those of exoplanets, but not altogether significantly different. Recently Limbach & Turner (2014) considered how the mean eccentricity of planets in a system is correlated with the number of planets in the system. They found a strong anti-correlation of eccentricity with multiplicity in systems observed by radial velocity. An extrapolation of their relation up to 8 planets fits well with the mean eccentricity observed in the Solar System. Furthermore, Van Eylen & Albrecht (2015) used the Kepler exoplanets with asteroseismically determined stellar mean densities to derive a rather low eccentricity dis- tribution of the multi–planet Kepler systems. Thus, while the eccentricities in our Solar System are low, those may be ex- pected in a system with so many planets. There is some bias in the exoplanet eccentricity data. For the RV planets, the best fit eccentricity is biased up- wards from the true value leading to a reduced number of systems with a low eccentricity (e.g. Shen & Turner 2008; Hogg, Myers & Bovy 2010; Zakamska, Pan & Ford 2011; Moorhead et al. 2011). However, the detection efficiency decreases only mildly with increasing eccentricity because despite being more difficult to detect, they have a larger RV amplitude for a fixed planet mass and semi-major axis (Shen & Turner 2008). While planets found with the tran- sit method require follow up observations (for example, with RV) to determine the eccentricity, the distribution of eccen- tricities is consistent with those of the RV planets (Kane et al. 2012). Without the bias, the eccentricities of the planets in our Solar System would appear to be less special. 2.2. Semi-Major Axis To date, there are a total of 1580 planets with a determined semi-major axis. This increases up to a total of 5289 if we in- clude planet candidates from Kepler. The false positive rate of the Kepler exoplanets is low, especially for multi planet sys- tems and the non–giant planets (e.g. D´esert et al. 2015) and thus we consider this much larger dataset also. We find the maximum of the log likelihood function to be when λ = −0.29 (λ = −0.20, including planet candidates). Fig. 2 shows the histogram of the Box-Cox transformed data. The left hand panel includes only confirmed exoplanets. For this data, we find the skewness to be S = 0.11. This is only slightly higher than the upper value expected for a normal distribution of √6/n = 0.062. We find the kurtosis to be K = −0.48 whereas for a normal distribution we would expect magnitudes smaller than √24/n = 0.12. Thus, the data have a large kurtosis. For the transformed data, the mean is −2.77 and the standard devi- ation is 2.16. Thus, Jupiter lies at 1.9σ. The right hand panel of Fig. 2 repeats this analysis but includes unconfirmed Kepler planets. The skewness for this is small at S = 0.022 (expected magnitude less than 0.034) and the kurtosis is much smaller also, K = −0.18 (expected magnitude less than 0.067). Jupiter lies at 2.4σ, suggesting on the face of it that Jupiter is rather special. However, as we explain below, this is most likely a result of selection effects. The majority of the planets in this distribution have been found by transit methods. The planet with the largest semi-major axis found by this method is only at 0.996 AU (the planet is KIC 11442793 h, Cabrera et al. 2014). Ke- pler, is thought to be complete only for planets at least as large as the Earth and for orbital periods up to a year (e.g Winn & Fabrycky 2014). Microlensing surveys prefer- entially find planets at radial distances of a few AU from their host star, that is often an M dwarf (e.g Gould et al. 2010; Cassan et al. 2012). This scale is dictated by the size of the Einstein ring radius around the lensing star. The ra- dial velocity method has found planets in the range 0.01 to 5.83 AU (e.g. Bonfils et al. 2013). Direct imaging can de- tect planets at much larger distances (e.g. Marois et al. 2008; Lafreni`ere, Jayawardhana & van Kerkwijk 2010), but so far JE 200 150 N d 100 50 0 0.0 N d 100 80 60 40 20 0 0.2 0.4 0.6 0.8 1.0 e 3 Earth Jupiter -3.0 -2.5 -2.0 -1.5 -1.0 -0.5 0.0 y_Λe Fig. 1.— Left: Eccentricity distribution for all observed exoplanets with a measured orbital eccentricity. Right: Box-Cox transformed distribution of exoplanet eccentricities. The total number of exoplanets is 539. only 8 planets have been detected by the method. In order to test the possibility that Jupiter’s outlier status is largely due to selection effects, we repeated the analysis but removed planets found by the transit method. There remain 473 planets in the sample. We find the maximum of the log likelihood function to be when λ = 0.11. Fig. 3 shows the histogram of the transformed data. For our data, we find the skewness to be S = 0.015. This is less than the value expected for a normal distribution of √6/n = 0.11. For our data we find a high kurtosis of K = 0.71 whereas for a normal distribution we would expect values with magnitude less than √24/n = 0.23. For the transformed data, the mean is −0.22 and the standard deviation is 1.41. Thus, Jupiter lies at 1.44σ and continues to be somewhat of an outlier, but the trend suggests that this is most likely still due to selection effects. The fact that direct imaging repeatedly reveals planets at separations much larger than Jupiter’s also may indicate that the current relative dearth of planets at large separations could be due to selection biases but more complete observations are required to test this possibility. We should also note that Beer et al. (2004) used only the planet with the largest velocity semi-amplitude in each ob- served system in their plots. However, they also performed the analysis with the most massive planet in each system and again with all the planets. They reported no difference in the significance of Jupiter as an outlier. In 2004, they found that Jupiter was at 2.3σ and half a sigma from its nearest neigh- bour. We find that Jupiter is not such an outlier as it was with the much smaller data set in 2004, and selection effects con- tinue to affect the distribution. This analysis should again be repeated once we have more reliable observations around the orbital radius of Jupiter. 2.3. Inner Solar System Currently, our inner Solar System appears to be rather spe- cial compared to observations of exoplanet systems. The in- ner edge of our Solar System is at the orbit of Mercury at 0.39 AU, while exoplanetary systems are observed to habor planets much closer to their star. We find that Mercury lies at 0.78 σ above the mean, while the Earth is at 1.28 σ above the mean of the distribution of confirmed exoplanet semi-major axes (as shown in the left hand panel of Fig. 2). When we in- clude all of the Kepler candidates (right hand panel of Fig. 2), these increase to 0.98σ for Mercury and 1.57σ for the Earth. Thus, all of the planets in our Solar System have orbital semi- major axes that are larger than the mean observed in exoplan- etary systems. However, it is possible that this could be the result of selection effects as it is easier to find planets in this region, if they are there. In terms of the radial location of observed exoplanets, the lack of close–in planets in our So- lar System is the parameter that makes our Solar System most special. We should note though that if we use only the planets found by methods other than the transit, then Mercury is at −0.48σ and the Earth is at 0.16σ. Consequently it is difficult to say how significant this discrepancy is. Batygin & Laughlin (2015) suggested that the migration of Jupiter and Saturn into the terrestrial planet forming region of our Solar System (down to a ≈ 1.5 AU) led to the depletion of mass in a < 0.39 AU. The planets are then thought to mi- grate outwards to their current location (see also Walsh et al. 2011). However, whatever the formation mechanism for the giant planets in our Solar System might have been, it is not thought to have been specific to our Solar System, and thus neither is a depleted inner Solar System. We discuss this point further in Section 4. 2.4. Migration Finally in this Section, we note that migration of plan- ets through the protoplanetary disk or planet-planet inter- actions or secular interactions of a binary star could af- fect these distributions, especially that of the semi-major axes. For example, it may be theoretically impossible for Jupiter mass planets to form at the radial location of hot– Jupiters (e.g. Bodenheimer, Hubickyj & Lissauer 2000). In- stead, they are supposed to form outside of the snow line and migrate inwards through the protoplanetary disk be- fore the latter disperses (e.g. Goldreich & Tremaine 1980). Migration could also occur by the Kozai–Lidov mecha- nism increasing the eccentricity of the planet followed by tidal circularization (Kozai 1962; Lidov 1962; Wu & Murray 2003; Takeda & Rasio 2005; Nagasawa, Ida & Bessho 2008; Perets & Fabrycky 2009). Evidence obtained by the Rossiter– McLaughlin effect suggests that some fraction of the hot Jupiters may have been produced through dynamical in- teractions (see e.g. Winn et al. 2005; Lubow & Ida 2010; Triaud et al. 2010; Albrecht et al. 2012). Hot–Jupiter planets dominated the initial planet discoveries because they are large and close to their host star. However, we now know that they are quite rare and orbit only about one percent of solar type stars (e.g. Wright et al. 2012). Opinions vary on whether in the Solar System Jupiter has significantly migrated. On one hand Morbidelli et al. (2010) 4 300 250 200 N d 150 100 50 0 Jupiter 1200 1000 800 N d 600 400 200 Jupiter -10 -5 y_ΛaAU 0 0 -10 -8 -6 -2 -4 y_ΛaAU 0 2 Fig. 2.— Box-Cox transformed distribution of exoplanet semi-major axis. Left: Only including confirmed exoplanets. The total number of exoplanets is 1580. Right: Including all planet candidates. The total number of exoplanets is 5289. Jupiter 3. PLANETARY PROPERTIES In this Section we consider how the masses and densities of the planets in our Solar System compare to those in ex- osolar systems. In this context, we discuss also the potential significance of the lack of a super-Earth in our Solar System. 3.1. Planet Masses 150 100 50 N d 0 -4 -2 0 2 y_ΛaAU 4 6 Fig. 3.— Box-Cox transformed distribution of exoplanet semi-major axis not including planets found by the transit method. The total number of exo- planets is 473. suggested that Jupiter did not migrate much from its forma- tion location, and on the other, Walsh et al. (2011) proposed that the low mass of Mars could be explained by gas-driven early migration of Jupiter. Batygin & Laughlin (2015) fur- ther suggested that this formation process could explain the lack of objects in our inner Solar System. Armitage et al. (2002) assumed that planets form constantly at a radius of 5 AU and found theoretically that about 10 to 15% of systems will have a Jupiter mass planet that does not migrate signifi- cantly. However, this conclusion will be affected by the pres- ence of a dead zone (a region of the disk with no turbulence, e.g. Gammie 1996; Martin et al. 2012a,b) that may slow or halt migration altogether. Furthermore, the inner and outer edges of a dead zone may act as planet traps that stop migra- tion (e.g. Hasegawa & Pudritz 2011, 2013). Thus, the distri- bution of planet semi-major axes definitely does not represent the initial distribution at the time of planet formation. Given that Jupiter mass planets are thought to form in the vicinity of Jupiter’s current radial location, theory suggests that the radial location of Jupiter is not particularly special. The current observational bias towards planets that are close to their host star means that it is easier to find planets that have migrated inwards, rather than those that have not, or even those that may have migrated outwards. We expect in the future that with more complete observations of Jupiter mass planets at Jupiter’s radial location we will be able to constrain the migration mechanisms and uncover how special Jupiter really is for its small (net) distance of migration. Fig. 4 shows the distribution of the approximate masses of the exoplanets that have been observed to date1. The masses of the planets within our Solar System are shown with arrows at the top (but are not included in the data). The masses of the gas giants fit well with those of exosolar planets, but the ter- restrial planets are all on the low side. This is most likely due (at least partially) to the difficulty in finding low-mass plan- ets. The masses of the exoplanets are strongly biased towards high–mass and short–period planets. Kepler has shown us that small planets are very common but the mass measurement of small mass planets is difficult and thus currently they appear to be rare. There are two peaks in the data, the first of which is at a mass between that of the Earth and that of Uranus, at around 0.01 MJ, where MJ is the mass of Jupiter. Plan- ets with a mass in the range of 1 M (where M ⊕ ⊕ is the mass of the Earth) are known as super-Earths (e.g. Valencia, Sasselov & O’Connell 2007). Our Solar System does not contain any super-Earths thus in that sense it is some- what unusual. We discuss this further in subsection 3.3. The second peak is around the mass of Jupiter. to 10 M ⊕ We fit the exoplanet mass data with a binormal probability density function (PDF) P(z) dz ∝ (z−µ1)2 σ2 1 + e 1 σ1 (z−µ2 )2 σ2 2 , w2 σ2 e (7) where z = log10(m). With a Kolmogorov-Smirnov (KS) test we find the best fitting parameters to be µ1 = −1.80, σ1 = 0.28, w2 = 0.69, µ2 = 0.12, σ2 = 0.56 and we show the PDF as the solid line in Fig. 4. With this distribution, we find that Jupiter is very typical at only −0.26σ from the higher-mass peak, while Saturn is at −1.3σ. The terrestrial planets are hard to compare because there are so few data points for those 1 The masses for planets that have been observed by the Doppler method are Mp sin i, where i is the orbital inclination. For directly imaged planets, the mass is predicted by theoretical models of the planets’ evolution. For the planets that have been observed by microlensing, it’s the ratio of the planet to star mass that is measured with accuracy. F D P 0.8 0.6 0.4 0.2 0.0 Me Ma V E U N S Jupiter -4 -3 -2 -1 logm 0 1 2 Fig. 4.— Exoplanet mass distribution. The arrows show the masses of the planets in our Solar System. The vertical lines shows range of planets that are considered to be super–Earths, the lower limit is the mass of the Earth and the upper limit is the lower limit to the mass of a giant planet at 10 M . The ⊕ total number of exoplanets is 1516. Super-Earths Giant Planets 2 ^ m c g y t i s n e D 100 10 1 0.1 0.001 0.01 0.1 MM_J 1 10 Fig. 5.— Densities of planets as a function of their mass. The blue circles show the exoplanets (total number of 287) and the red squares the planets in our Solar System. The vertical lines shows range of planets that are consid- ered to be super–Earths, the lower limit is the mass of the Earth and the upper limit is the lower limit to the mass of a giant planet at 10 M ⊕ . small masses. However, Neptune lies at 1.89σ and Uranus at 1.57σ from the lower-mass peak. Gas giants are thought to form outside of the snow line ra- dius in the protoplanetary disk where there is more solid mate- rial available to form massive planets (e.g. Pollack et al. 1996; Martin & Livio 2012, 2013a). The surface density of the pro- toplanetary disk decreases with increasing distance from the star and the timescale to form a planet increases with radius. Thus, lower mass gas giants could preferentially form farther away from the star. Low mass and large orbital radius planets are certainly harder to detect than those with higher mass and lower orbital radius. This can explain the lack of observed small mass planets, but also perhaps the dip in the observed distribution. For example, if Uranus and Neptune were around another star, at their large orbital radii, they would also be dif- ficult to detect. The only method that could currently detect them is direct imaging. However, the smallest mass planet that has been found with this method is Formaulhaut b that has an approximate mass of . 2 MJ (Currie et al. 2012). It therefore remains a possibility that the double peaked mass distribution is solely the result of selection effects. 3.2. Planet Densities Fig. 5 shows the approximate densities of the observed planets as a function of their mass. The exoplanets are shown 5 in blue and the planets in our Solar System in red. While the lower mass exoplanets have a large range in their density for a given mass (see also Wolfgang & Laughlin 2012; Marcy et al. 2014a; Howe, Burrows & Verne 2014; Knutson et al. 2014), the giant planets show a clear correlation of increasing density with mass. Thus, the super-Earths may have a wide range of compositions (Valencia, Sasselov & O’Connell 2007). Despite the large spread in the data for the low mass planets, there appears to be a trend of decreasing den- sity with increasing mass. This could be attributed qualita- tively to the peak in the theoretical radius against mass of a planet (see e.g. Zapolsky & Salpeter 1969; Seager et al. 2007; Fortney, Marley & Barnes 2007; Mordasini et al. 2012). to 2 R ⊕ ⊕ < R < 4 R ⊕ More recently, it has been suggested that the super-Earths (with radii in the range 1 R ) show two trends separated by a critical planet radius. The smallest planets increase in density with radius while those that are larger decrease suggesting that the larger planets have a large amount of volatiles on a rocky core. There is some un- certainty over the value of the critical radius, as estimates range from about 1.5 R Petigura, Marcy & Howard (2013); Lopez & Fortney (2014); Weiss & Marcy (2014); Marcy et al. (2014b), if it exists at all (Morton & Swift 2014). The data suggest that the densities of the giant planets within our Solar System are very typical of those of observed exoplanets. The masses of the terrestrial planets in our Solar System are on the edge of our current sensitivity and thus it is hard to draw any conclusions about their densities. How- ever, recently, Dressing et al. (2015) found that the Earth (and Venus) can be modelled with the same ratio of iron to mag- nesium silicate as the low mass exoplanets observed and thus the Earth may not be special in this respect. ⊕ 3.3. Lack of a super-Earth ⊕ It is interesting to examine whether our Solar System’s lack of a super-Earth is truly unusual. There have been several at- tempts to calculate an occurrence rate for super-Earths taking into account the selection biases. The results for RV obser- vations predict an occurrence rate in the range 10 − 20 % in the period range Pb < 50 d (Howard et al. 2010; Mayor et al. 2011). The transiting planet observations imply a range of occurrence rates that is at most as high as 50 % in the pe- riod range Pb < 85 d (Fressin et al. 2013). More recently Burke et al. (2015) examined the Kepler sample for planets with radii in the range 0.75 < R < 2.5 R with orbital periods in the range 50 to 300 d and found an occurrence rate of 77%. Although these results include Earth-size planets and some of the periods are longer than that of Mercury, the occurrence rate increases with short orbital periods, making the existence of close-in planets more likely. The high occurrence rate of these types of planets offers perhaps the strongest argument against the Solar System being very common, but even that does not necessarily make it extremely rare. Typically, sys- tems that have an observed super-Earth, have more than one and this is theoretically expected if the planets form by merg- ers of inwardly migrating cores (e.g. Terquem & Papaloizou 2007; Cossou et al. 2014). It is possible that the presence of a super-Earth can af- fect terrestrial planet formation. Many of the super-Earths observed are at small radial locations, where theoretically they could not have formed (e.g. Raymond & Cossou 2014; Schlichting 2014). Izidoro, Morbidelli & Raymond (2014) found that if a super-Earth migrates sufficiently slowly through the habitable zone (defined as the radial range of dis- 6 y t i c i r t n e c c E e g a r e v A 0.8 0.6 0.4 0.2 0.0 0.1 10 Minimum Semi-Major AxisAU 1 100 Fig. 6.— The average eccentricity in a planetary system vs the smallest planet semi-major axis. The size of the point is proportional to the number of planets in the system. The small blue points have 1 planet, the red points have 2 planets, the purple have 3, the green have 4, the orange have 5 and the Solar System with 8 is shown in the large blue point. The Solar System has an average eccentricity of 0.06 and Mercury, being the inner most planet, at 0.39 AU. tances from the star at which a rocky planet can maintain liq- uid water on its surface) then any terrestrial planet that later forms there would be volatile-rich and not very Earth-like. In conclusion, the masses and densities of the planets of our Solar System appear to be very typical of those of ex- oplanets. However, the lack of a planet with a mass in the , a super-Earth, makes the Solar System ap- range 1 − 10 M ⊕ pear somewhat special. 4. DISCUSSION Generally, the physical properties of the planets in our So- lar System are quite typical when compared to those of the observed exoplanets, although the lack of a super–Earth is unusual. The orbital properties, however, may be some- what special and perhaps more conducive to life. Low ec- centricity planets have a more stable temperature throughout the orbit and therefore may be more likely to host life (e.g. Williams & Pollard 2002; Gaidos & Williams 2004). Further- more, planetary systems with a low average eccentricity are more likely to have long term dynamical stability. For exam- ple, the terrestrial planets in our Solar System are expected to be dynamically stable at least until the Sun becomes a red giant and engulfs the inner planets (e.g. Laughlin 2009). There are a few other factors that could, in principle at least, make our Solar System special with respect to the emergence of life. First we can consider the age. The current age of our Sun is about half the age of the disk of our Galaxy, and also half of the Sun’s total lifetime. Thus, we expect that roughly half of the stars in our Galaxy’s disk are older and half are younger than our Sun. This implies that the age of our Solar System is definitely not special. Furthermore, Behroozi & Peeples (2015) considered the planet formation history of the Milky way and determined that our Solar Sys- tem formed close to the median epoch for giant planet for- mation. They also found that about 80% of the currently- existing Earth-like planets were already formed at the time of the Earth’s formation. We also note that the fact that the Solar System contains a single star does not make it particu- larly special, since the binary fraction in the Kepler sample, for example, is about 50% (Horch et al. 2014). The presence of terrestrial planets in the habitable zone ⊕ around their host star appears to be quite common. For ex- ample, Dressing & Charbonneau (2015) examined the Kepler data for M dwarfs and found that for orbital periods shorter than 50 d, the occurrence rate of Earth-size planets in the hab- itable zone is around 18 − 27%. This conservative estimate could be as high as ∼ 50% depending on how the habitable zone is defined. This is consistent with radial velocity sur- veys that find 0.41 potentially habitable planets per M dwarf (Bonfils et al. 2013). Thus, an Earth-size planet in a habitable zone is not uncommon. A habitable planet may require a large moon which in turn may require an asteroid collision (Canup & Asphaug 2001). Thus, systems which contain an asteroid belt may be more conducive to initiating life. However, habitability may be sen- sitive to the size of the asteroid belt (Martin & Livio 2013b). As we have noted in the Introduction, asteroid belts could be a common feature of planetary systems (Chen et al. 2014). The metallicity of a protoplanetary disk (and hence the host star) determines the structure of a planetary system that forms (e.g. Buchhave et al. 2014; Wang & Fischer 2015). The higher the metallicity of a star, the more giant planets that are observed (e.g. Fischer & Valenti 2005; Sousa et al. 2011; Gonzalez 2014; Reffert et al. 2015). However, the correla- tion for lower mass planets is unclear (Buchhave et al. 2012; Mayor, Lovis & Santos 2014). Planets with radii less than four times that of the Earth are observed around stars with a wide range of metallicities. However, the average metallicity of stars hosting small planets (Rp < 1.7R ) in Buchhave et al. (2014) is very close to solar. Although such planets can form at a wide range of metallicities, the fact that the average metal- licity of the small planets is solar may not be a coincidence. Thus, while the metallicity of our Solar System may not be especially promotive to the formation of a habitable planet, it’s unclear whether the Solar System is special or not in this respect. The variability of our Sun has been compared to the ac- tivity of stars in the Kepler sample with conflicting conclu- sions. Basri et al. (2010) and Basri, Walkowicz & Reiners (2013) found that the Sun is rather typical with only a quarter to a third of stars in the Kepler sample being more active than the Sun. On the other hand, McQuillan, Aigrain & Roberts (2012) found the Sun to be relatively quiet with 60% of stars being more active. The difference in the conclusions stems from choices in defining the activity level of our Sun and the inclusion of stars with fainter magnitudes in McQuillan, Aigrain & Roberts (2012). However, the studies agree that the active fraction of stars becomes larger for cooler stars. M dwarfs have a fraction of 90% that are more active than the Sun. Thus, compared to other Sun-like stars, our Sun could be typical, but compared to cooler stars, our Sun is cer- tainly quiet. In general, there are three aspects in which the Solar Sys- tem differs most from other observed multi-planet systems. First, the low mean eccentricity of the planets in the Solar System maybe somewhat special, although this may be ac- counted for by selection effects. Secondly, there is in the total lack of planets inside Mercury’s orbit. Massive planets mi- grating through the habitable zone can change the course of planet formation in that region. Overall, however, processes that could act to clear the inner part of the Solar Systems (such as giant planet migration), are believed to be operating within a non-negligible fraction of the exoplanet systems (e.g. Batygin & Laughlin 2015). Third, the lack of super-Earths in our Solar System is somewhat special and could have allowed the Earth to become habitable. A close-in super-Earth could also affect the dynamical stability of a terrestrial planet in the habitable zone and this should be investigated in future work. We consider the first two of these special parameters in more detail in Fig. 6. For planets with measured eccentricity, we plot the mean eccentricity of the planetary system and the semi–major axis of the innermost planet observed within the system. In this parameter space, the Solar System appears to be somewhat special, but far from being rare. Although, most of the systems with three or more planets, do have a planet with an orbital semi-major axis smaller than that of Mercury, this could be at least partly due to selection effects. The ob- served semi–major axis of the innermost planet may be close to complete but the number of planets in each system is cer- tainly not. If the innermost planet is very close in, then it is easier to detect the planets outside of its orbit. If on the other hand the innermost planet is farther out (e.g. some of the small blue and red points in Fig. 6) then additional planets will be difficult to find (see also discussion in Section 2.2). There are many factors that may be required in order to form a habitable planet. When we multiply the probabilities for each together, we may end up with a small probability for such an event. However, since we currently do not know which factors are truly important for life to emerge, such an exercise does not make much sense. If we consider too many details, clearly the Solar System is special because all systems are different. However, at the moment we have not identified any parameter that makes the Solar System so significantly different that it would make it rare. 5. CONCLUSIONS We find that the properties of the planets in our Solar Sys- tem are not so significantly special compared to those in ex- 7 osolar systems to make the Solar System extremely rare. The masses and densities are typical, although the lack of a super– Earth sized planet appears to be somewhat unusual. The or- bital locations of our planets seem to be somewhat special but this is most likely due to selection effects and the difficulty in finding planets with a small mass or large orbital period. The mean semi-major axis of observed exoplanets is smaller than the distance of Mercury to the Sun. The relative deple- tion in mass of the Solar System’s terrestrial region may be important. The eccentricities are relatively low compared to observed exoplanets, although the observations are biased to- wards finding high eccentricity planets. The low eccentricity, however, may be expected for multi-planet systems. Thus, the two characteristics of the Solar System that we find to be most special are the lack of super-Earths with orbital periods of days to months and the general lack of planets inside of the orbital radius of Mercury. From the perspective of habitability the Solar System does not appear to be particularly special. If exosolar life happens to be rare it would probably not be because of simple ba- sic physical parameters, but because of more subtle processes that are related to the emergence and evolution of life. Since at the moment we don’t know what those might be, we can al- low ourselves to be optimistic about the prospects of detecting exosolar life. We should make every possible effort to detect and characterise the atmospheres of a few dozen Earth-size planets in the habitable zone, in the coming two decades. ACKNOWLEDGEMENTS We thank the anonymous referee for useful comments. This research has made use of the Exoplanet Orbit Database and the Exoplanet Data Explorer at exoplanets.org. REFERENCES 248 Albrecht S. et al., 2012, ApJ, 757, 18 Armitage P. J., Livio M., Lubow S. H., Pringle J. E., 2002, MNRAS, 334, Basri G. et al., 2010, ApJl, 713, L155 Basri G., Walkowicz L. M., Reiners A., 2013, ApJ, 769, 37 Batalha N. M., 2014, PNAS, 111, 12647 Batygin K., Laughlin G., 2015, Proceedings of the National Academy of Science, 112, 4214 Beer M. E., King A. R., Livio M., Pringle J. E., 2004, MNRAS, 354, 763 Behroozi P. S., Peeples M. S., 2015, MNRAS, submitted Bodenheimer P., Hubickyj O., Lissauer J. J., 2000, Icarus, 143, 2 Bonfils X. et al., 2013, A&A, 549, A109 Box G. E. P., Cox D. R., 1964, J. Roy. Stat. Soc., Ser. B, 26, 211 Buchhave L. A. et al., 2014, Nature, 509, 593 Buchhave L. A. et al., 2012, Nature, 486, 375 Burke C. J. et al., 2015, ArXiv e-prints Cabrera J. et al., 2014, ApJ, 781, 18 Canup R. M., Asphaug E., 2001, Nat, 412, 708 Cassan A., Kubas D., Beaulieu J.-P., et al., 2012, Nat, 481, 167 Chen C. H., Mittal T., Kuchner M., Forrest W. J., Lisse C. M., Manoj P., Sargent B. A., Watson D. M., 2014, ApJS, 211, 25 Cossou C., Raymond S. N., Hersant F., Pierens A., 2014, A&A, 569, A56 Currie T. et al., 2012, ApJL, 760, L32 D´esert J.-M. et al., 2015, ApJ, 804, 59 Dressing C. D., Charbonneau D., 2015, ArXiv e-prints Dressing C. D. et al., 2015, ApJ, 800, 135 Fischer D. A., Valenti J., 2005, ApJ, 622, 1102 Fortney J. J., Marley M. S., Barnes J. W., 2007, ApJ, 659, 1661 Fressin F. et al., 2013, ApJ, 766, 81 Gaidos E., Williams D. M., 2004, New Astronomy, 10, 67 Gammie C. F., 1996, ApJ, 457, 355 Goldreich P., Tremaine S., 1980, ApJ, 241, 425 Gonzalez G., 2014, MNRAS, 443, 393 Gould A., Dong S., Gaudi, et al., 2010, ApJ, 720, 1073 Han E., Wang S. X., Wright J. T., Feng Y. K., Zhao M., Fakhouri O., Brown J. I., Hancock C., 2014, PASP, 126, 827 Hasegawa Y., Pudritz R. E., 2011, MNRAS, 417, 1236 Hasegawa Y., Pudritz R. E., 2013, ApJ, 778, 78 Hogg D. W., Myers A. D., Bovy J., 2010, ApJ, 725, 2166 Horch E. P., Howell S. B., Everett M. E., Ciardi D. R., 2014, ArXiv e-prints Howard A. W. et al., 2010, Science, 330, 653 Howe A. R., Burrows A., Verne W., 2014, ApJ, 787, 173 Izidoro A., Morbidelli A., Raymond S. N., 2014, ApJ, 794, 11 Kane S. R., Ciardi D. R., Gelino D. M., von Braun K., 2012, MNRAS, 425, 757 Kennedy G. M., Wyatt M. C., 2014, MNRAS, 444, 3164 Knutson H. A. et al., 2014, ApJ, 794, 155 Kozai Y., 1962, AJ, 67, 591 Lafreni`ere D., Jayawardhana R., van Kerkwijk M. H., 2010, ApJ, 719, 497 Laughlin G., 2009, Nature, 459, 781 Lidov M. L., 1962, Planet. Space Sci., 9, 719 Limbach M. A., Turner E. L., 2014, ArXiv e-prints Lissauer J. J., Dawson R. I., Tremaine S., 2014, Nature, 513, 336 Lopez E. D., Fortney J. J., 2014, ApJ, 792, 1 Lubow S. H., Ida S., 2010, Planet Migration, Seager S., ed., pp. 347–371 Marcy G. W. et al., 2014a, ApJS, 210, 20 Marcy G. W., Weiss L. M., Petigura E. A., Isaacson H., Howard A. W., Buchhave L. A., 2014b, Proceedings of the National Academy of Science, 111, 12655 Marois C., Macintosh B., Barman T., Zuckerman B., Song I., Patience J., Lafreni`ere D., Doyon R., 2008, Science, 322, 1348 Martin R. G., Livio M., 2012, MNRAS, 425, L6 Martin R. G., Livio M., 2013a, MNRAS, 434, 633 Martin R. G., Livio M., 2013b, MNRAS, 428, L11 Martin R. G., Lubow S. H., Livio M., Pringle J. E., 2012a, MNRAS, 420, Martin R. G., Lubow S. H., Livio M., Pringle J. E., 2012b, MNRAS, 423, 3139 2718 Mayor M., Lovis C., Santos N. C., 2014, Nature, 513, 328 Mayor M. et al., 2011, ArXiv e-prints McQuillan A., Aigrain S., Roberts S., 2012, A&A, 539, A137 Moorhead A. V. et al., 2011, ApJS, 197, 1 Morbidelli A., Brasser R., Gomes R., Levison H. F., Tsiganis K., 2010, AJ, Mordasini C., Alibert Y., Georgy C., Dittkrist K.-M., Klahr H., Henning T., 140, 1391 2012, A&A, 547, A112 8 Morton T. D., Swift J., 2014, ApJ, 791, 10 Nagasawa M., Ida S., Bessho T., 2008, ApJ, 678, 498 Pani´c O. et al., 2013, MNRAS, 435, 1037 Perets H. B., Fabrycky D. C., 2009, ApJ, 697, 1048 Petigura E. A., Marcy G. W., Howard A. W., 2013, ApJ, 770, 69 Pollack J. B., Hubickyj O., Bodenheimer P., Lissauer J. J., Podolak M., Greenzweig Y., 1996, Icarus, 124, 62 Raymond S. N., Cossou C., 2014, MNRAS, 440, L11 Reffert S., Bergmann C., Quirrenbach A., Trifonov T., Kunstler A., 2015, A&A, 574, A116 Rowe J. F., Bryson S. T., Marcy G. W., et al., 2014, ApJ, 784, 45 Schlichting H. E., 2014, ApJL, 795, L15 Seager S., Kuchner M., Hier-Majumder C. A., Militzer B., 2007, ApJ, 669, Shen Y., Turner E. L., 2008, ApJ, 685, 553 Sousa S. G., Santos N. C., Israelian G., Mayor M., Udry S., 2011, A&A, 1279 533, A141 Takeda G., Rasio F. A., 2005, ApJ, 627, 1001 Terquem C., Papaloizou J. C. B., 2007, ApJ, 654, 1110 Triaud A. H. M. J. et al., 2010, A&A, 524, A25 Valencia D., Sasselov D. D., O’Connell R. J., 2007, ApJ, 665, 1413 Van Eylen V., Albrecht S., 2015, ArXiv e-prints Velilla S., 1993, Statiscics and Probability Letters, 17, 259 Walsh K. J., Morbidelli A., Raymond S. N., O’Brien D. P., Mandell A. M., 2011, Nature, 475, 206 Wang J., Fischer D. A., 2015, AJ, 149, 14 Weiss L. M., Marcy G. W., 2014, ApJl, 783, L6 Williams D. M., Pollard D., 2002, International Journal of Astrobiology, 1, 61 Winn J. N., Fabrycky D. C., 2014, ArXiv e-prints Winn J. N. et al., 2005, ApJ, 631, 1215 Wolfgang A., Laughlin G., 2012, ApJ, 750, 148 Wright J. T. et al., 2011, PASP, 123, 412 Wright J. T., Marcy G. W., Howard A. W., Johnson J. A., Morton T. D., Fischer D. A., 2012, ApJ, 753, 160 Wu Y., Murray N., 2003, ApJ, 589, 605 Wyatt M. C., 2008, ARA&A, 46, 339 Zakamska N. L., Pan M., Ford E. B., 2011, MNRAS, 410, 1895 Zapolsky H. S., Salpeter E. E., 1969, ApJ, 158, 809
1503.02476	1	1503	2015-03-09T13:48:17	Methane on Uranus: The case for a compact CH4 cloud layer at low latitudes and a severe CH4 depletion at high-latitudes based on re-analysis of Voyager occultation measurements and STIS spectroscopy	[ "astro-ph.EP" ]	Lindal et al. (1987, J. Geophys. Res. 92, 14987-15001) presented a range of temperature and CH4 profiles for Uranus that were consistent with 1986 Voyager radio occultation measurements. A localized refractivity slope variation near 1.2 bars was interpreted to be the result of a condensed CH4 cloud layer. However, models fit to near-IR spectra found particle concentrations in the 1.5-3 bar range (Sromovsky et al. 2006, Icarus 182, 577-593, Sromovsky and Fry 2008, Icarus 193, 211-229, Irwin et al. 2010, Icarus 208, 913-926), and a recent analysis of STIS spectra argued that aerosol particles formed diffusely distributed hazes, with no compact condensation layer (Karkoschka and Tomasko 2009, Icarus 202, 287-309). Trying to reconcile these results, we reanalyzed the occultation observations with a He volume mixing ratio reduced from 0.15 to 0.116, which is near the edge of the 0.033 range given by Conrath et al. (1987, J. Geophys. Res., 15003-10). This allowed us to obtain saturated CH4 mixing ratios within the putative cloud layer and to reach above-cloud and deep CH4 mixing ratios compatible with STIS spectral constraints. Using a 5-layer vertical aerosol model with two compact cloud layers in the 1-3 bar region, we find that the best fit pressure for the upper layer is virtually identical to the pressure range inferred from the occultation analysis for a methane mixing ratio near 4% at 5 deg S, arguing that Uranus does indeed have a compact methane cloud layer. While our cloud model can fit the latitudinal variations in spectra between 30 deg S and 20 deg N using the same temperature and CH4 profiles, closer to the pole, the model requires the introduction of an increasingly strong upper tropospheric depletion of CH4 at increased latitudes, in rough agreement with the trend identified by Karkoschka and Tomasko (2009, Icarus 202, 287-309).	astro-ph.EP	astro-ph	Journal reference: Icarus 215 (2011) 292-312. Preprint typeset using LATEX style emulateapj v. 5/2/11 METHANE ON URANUS: THE CASE FOR A COMPACT CH4 CLOUD LAYER AT LOW LATITUDES AND A SEVERE CH4 DEPLETION AT HIGH-LATITUDES BASED ON RE-ANALYSIS OF VOYAGER OCCULTATION MEASUREMENTS AND STIS SPECTROSCOPY† L. A. Sromovsky1, P. M. Fry1, J. H. Kim1 Journal reference: Icarus 215 (2011) 292-312. ABSTRACT Lindal et al. (1987, J. Geophys. Res. 92, 14987-15001) presented a range of temperature and methane profiles for Uranus that were consistent with 1986 Voyager radio occultation measurements of refractivity versus altitude. A localized refractivity slope variation near 1.2 bars was interpreted to be the result of a condensed methane cloud layer. However, models fit to near-IR spectra found particle concentrations much deeper in the atmosphere, in the 1.5-3 bar range (Sromovsky et al. 2006, Icarus 182, 577-593, Sromovsky and Fry 2008, Icarus 193, 211-229, Irwin et al. 2010, Icarus 208, 913-926), and a recent analysis of STIS spectra argued for a model in which aerosol particles formed diffusely distributed hazes, with no compact condensation layer (Karkoschka and Tomasko 2009, Icarus 202, 287-309). To try to reconcile these results, we reanalyzed the occultation observations with the He volume mixing ratio reduced from 0.15 to 0.116, which is near the edge of the 0.033 uncertainty range given by Conrath et al. (1987, J. Geophys. Res., 15003-10). This allowed us to obtain saturated mixing ratios within the putative cloud layer and to reach above-cloud and deep methane mixing ratios compatible with STIS spectral constraints. Using a 5-layer vertical aerosol model with two compact cloud layers in the 1-3 bar region, we find that the best fit pressure for the upper layer is virtually identical to the pressure range inferred from the occultation analysis for a methane mixing ratio near 4% at 5◦ S. This strongly argues that Uranus does indeed have a compact methane cloud layer. In addition, our cloud model can fit the latitudinal variations in spectra between 30◦ S and 20◦ N, using the same profiles of temperature and methane mixing ratio. But closer to the pole, the model fails to provide accurate fits without introducing an increasingly strong upper tropospheric depletion of methane at increased latitudes, in rough agreement with the trend identified by Karkoschka and Tomasko (2009, Icarus 202, 287-309). Subject headings: Uranus, Uranus Atmosphere; Atmospheres, composition, Atmospheres, structure 1. INTRODUCTION The existence of a thin methane ice cloud in the at- mosphere of Uranus was inferred by Lindal et al. (1987), henceforth referred to as L87, from their analysis of Voy- ager 2 radio occultation measurements. They also de- rived a suite of temperature and methane profiles that were all consistent with their measurements, including Model F, which had the greatest deep methane mixing ratio (4%), and their preferred Model D, which had a deep mixing ratio of 2.3% and an above-cloud relative humidity of 30%. (Here we use the term relative humid- ity of methane to refer to the ratio of its partial pressure to its saturation pressure at the same temperature, and the mixing ratio referred to is the volume mixing ratio or VMR.) A methane cloud layer near the 1.2-bar level in- ferred by L87 has been used successfully in the analysis of observations in the visible spectral range. For example, Rages et al. (1991) incorporated a methane cloud layer into their models of Voyager 2 imaging observations, find- ing modest optical depths of 0.66±0.18 at 22◦ S (assumed to be independent of wavelength), and about three times that level at 65 ◦ S, while assuming a deep methane 1 University of Wisconsin - Madison, Madison WI 53706 † Based in part on observations with the NASA/ESA Hub- ble Space Telescope obtained at the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Incorporated under NASA Contract NAS5-26555. mixing ratio of 4% (consistent with L87 Model F). On the other hand, an analysis of hydrogen S(0) and S(1) quadrupole line features by Baines et al. (1995), which also incorporated a methane cloud near 1.2 bars, found its opacity (weighted to high latitudes) to be about 0.4 at 0.6 µm, while inferring a deep methane mixing ratio of 1.6%. Neither of these authors tried to constrain the pressure of the cloud, however, so that consistency with occultation results was not fully established. challenge to the a serious Recently, exis- tence of the methane cloud layer was made by Karkoschka and Tomasko (2009), henceforth referred to as KT2009, based on their analysis of spatially resolved 0.3-1 µm spectra obtained from 2002 observations by the Hubble Space Telescope Imaging Spectrograph (STIS). They concluded that the most significant cloud opacity concentration was in a layer from 1.2-2 bars, with particles uniformly mixed with the gas in this layer, which had wavelength-independent optical depths between 1.2 and 2.2. They argued for no localized CH4 condensation layer at all, but instead for the existence of a global thick and diffuse tropospheric haze similar to that observed on Titan. This seemed to confirm the analysis of near-IR spectral observations, which had already questioned the existence of a methane ice cloud near 1.2 bars. From an analysis of the Fink and Larson (1979) near- IR spectrum, which made use of improved methane ab- 2 Sromovsky et al. sorption coefficients (Irwin et al. 2006), Sromovsky et al. (2006) obtained a cloud layer near 2 bars for the L87 Model D profiles and at 1.5-1.7 bars for their Model F profiles. A subsequent analysis of 2004 near-IR imag- ing observations by Sromovsky and Fry (2007) concluded that the methane relative humidity should be near 60%- 100% above the nominal cloud region (Models D and F had 30% and 53% respectively), and at low latitudes found no need for a cloud layer in the 1.2-2 bar region. Further refinements of methane absorption models for the near-IR (Karkoschka and Tomasko 2010) did not en- tirely fix these discrepancies. Irwin et al. (2010) used these improved methane coefficients and the L87 Model D T(P) profile and above-cloud methane mixing ratio, but used 1.6% instead of 2.26% below the putative cloud layer. With these assumptions they obtained a low- latitude cloud density peak near 2.5 bars. Even using the L87 Model F T(P) profile and a 4% deep methane VMR, Irwin et al. (2010) still obtained a cloud peak that was too deep to reach the methane condensation level, though the cloud pressure was then elevated to about the 1.7-bar level, which puts the peak in the middle of the main aerosol layer of KT2009. More methane seems to be required to bring the cloud pressure inferred from spectral observations to the same level as inferred from the refractivity profile. However, according to L87, their Model F is an upper limit on methane amounts, and they argued that Model D is re- ally preferable. That solution has a deep methane mix- ing ratio of 2.3% by volume, a cloud layer between 1.15 and 1.27 bars, and an above-cloud methane humidity of 30%. Other profiles that satisfy the occultation mea- surements have deep methane mixing ratios varying from zero to 4%, and above-cloud humidities varying from zero to 53%, while the in-cloud humidities vary from zero to 78%. Model D was preferred for three reasons: (1) it provides the best agreement with IRIS observations sampling the above-cloud region, (2) it has the high- est in-cloud humidity levels, (3) it yields a deep mix- ing ratio (2.3%) that is in close agreement with that of Orton et al. (1986). The first reason is weak because the IRIS observations in question (Conrath et al. 1987) are at a large zenith angle and rather uncertain. The second is weakened by the fact that only the nominal he- lium volume mixing ratio was considered, and that can strongly affect the methane humidity, as we will show here. The third reason is questionable because the Orton et al. temperature profile derived assuming 2% methane does not agree with the occultation profile using vir- tually the same mixing ratio. Much stronger external constraints are available from spectral observations in the CCD (∼0.3 - 1 µm) wavelength region, as shown by KT2009 and by the analysis presented here. In prior use of these spectral constraints however, there has been an unjustified deviation from occultation constraints in both thermal and methane profiles, in which a thermal profile derived for one methane profile is used for a very dif- ferent mixing ratio (Irwin et al. 2010; Baines et al. 1995; Sromovsky and Fry 2007) and above-cloud methane pro- files have been used that exceed all of the occultation so- lutions, e.g. KT2009 and Baines et al. (1995). KT2009 also inferred that the methane mixing ratio varies with latitude, which raises additional questions about the ef- fects of corresponding density variations with latitude. To summarize, where spectral observations have been used to test the location of cloud layers on Uranus, the inferred locations are considerably deeper than implied by the occultation observations. And the spectral con- straints on the methane mixing ratio range from a low of 1.6+0.7 −0.5% by Baines et al. (1995) to 4% by Rages et al. (1991), and include latitude dependent values between these values inferred by KT2009. Further, with the ex- ception of Rages et al., prior modelers have not followed the occultation constraints on the the vertical distribu- tion of methane. This motivates our efforts to redo the occultation analysis with consideration of a wider range of solutions, and to examine more carefully the plausi- bility of a compact methane cloud layer on Uranus. In the following, we pursue the point of view that the occultation measurements of sudden slope changes in refractivity do indicate a region of sudden changes in methane mixing ratio, which are indicative of the con- densation level of methane, and possibly of a thin region containing cloud particles. We describe a reanalysis of the occultation measurements that can actually achieve saturated vapor pressures in the same region as the sud- den changes in refractivity slope. We also find solutions with high methane amounts at and above the cloud level that are consistent with the adopted profile of KT2009. After finding a range of solutions with the desired char- acteristics, we then constrain these solutions by calcu- lating spectra for compact cloud layer models and com- paring the pressures inferred from matching spectra to the pressures inferred from the occultation analysis. We conclude that a methane cloud layer at the occultation pressure is consistent with the spectral observations, but that most of the cloud opacity is concentrated in a deeper layer that was not detected by the occultation measure- ments. We also confirm the conclusions of KT2009 that the methane is strongly depleted at high latitudes, but to shallow depths. 2. APPROACH TO OCCULTATION ANALYSIS We first describe the basic methods of analysis, how we reconstruct the refractivity profile, then use that profile to validate our methane retrieval methods by comparison with results of L87. 2.1. Physical Basis and Methods of Occultation Analysis The occultation measurements, after accounting for observing geometry, can be reduced to refractivity as a function of altitude, where refractivity is defined to be N = (index of refraction -1)×106. If the molecular com- position is known, then refractivity can be converted to number density and mass density. The pressure can then be determined by integrating the product of mass den- sity and gravity, assuming hydrostatic equilibrium. From pressure and density, the equation of state of the gas can then be used to infer temperature. The main complica- tion is that the variable distribution of methane is not known and cannot be directly inferred from the observa- tions. This allows a range of T(P) solutions that depend on what is assumed about the methane distribution. The procedure followed by L87 was to first select a molecular weight that yielded a thermal profile near the tropopause that was consistent with IRIS thermal in- frared observations. The only significant constituents at Methane on Uranus: Compact cloud layer and high-latitude depletion 3 that level of the atmosphere are hydrogen and helium, so that the molecular weight was determined by their ratio, which was taken to be the Conrath et al. (1987) value of He/H2=15/85. L87 did not consider other He/H2 ratios, even though the quoted uncertainty is large enough to permit substantially different profiles, as will be shown in Sec. 3. The next step in the procedure was to select altitudes bounding the cloud layer. These altitudes were not stated by L87, but are approximately between 5 and 7 km below the 1 bar level determined for the D model. These are the rough locations of the rapid changes in re- fractivity slope. While the altitude relative to the center of the planet is fixed for all the profile solutions, the pres- sure varies somewhat from one solution to the next, so that the altitude above the 1 bar level also varies slightly. To constrain the methane profile L87 generally as- sumed a constant relative methane humidity above the cloud layer and used the tropopause mixing ratio to set the constant stratospheric mixing ratio. Within the cloud region, L87 state that the temperature lapse rate was set equal to the wet adiabatic lapse rate, and ad- justed the number density and temperature to match the refractivity profile. However, the model D T (P ) profile of L87 does not match the wet adiabat within the assumed cloud layer. Instead, the profile matches a weighted av- erage of the form (dT /dz)cld = (dT /dz)dry × (1 − RH) + (dT /dz)wet × RH (1) where RH is the relative humidity. This weighted av- erage is consistent with the suggestion that the occul- tation sampled a broad horizontal region in which the average humidity was less than expected for a uniform cloud layer. That is also offered as an explanation for the sub-saturated humidity levels that were obtained in the cloud layer. The relatively smooth I/F profiles observed on Uranus (Sromovsky and Fry 2007; Sromovsky et al. 2009) are rarely disturbed by discernible discrete fea- tures, especially at low latitudes, and thus this expla- nation is not a compelling one. L87 also introduced a condensed fraction of the methane within the putative cloud layer, but did not publish the inferred values, nor how these values were constrained. If the temperature profile is constrained to follow the wet adiabat (or the weighted average of wet and dry adiabats given above) then the only remaining variable that needs to be adjusted is the fraction of to- tal methane. There is no need to partition a fraction of the total into condensed form unless leaving the total in gaseous form would lead to supersaturation. In the latter case, it is reasonable to treat the excess vapor as condensed material. However, it is not reasonable to al- low more than a tiny fraction of the methane to be in condensed form. The condensed fractions reaching 10% or so that are noted by L87 are grossly inconsistent with near-IR and CCD spectral constraints because those con- straints permit only a low opacity cloud layer. Even a few percent of methane in condensed form would lead to extremely large optical depths at the cloud layer, which would provide very obvious spectral signa- tures that are not seen. Models of near-IR and visible spectra require only small optical depths, typically unity or less at visible wavelengths. When treated as Mie parti- cles the inferred particle size of a compact cloud near 1.2 bars is of the order of r = 1 µm. The total mass per unit area for a given optical depth τ is given by m = 4 3 ρrτ /Q, where ρ is the density of solid methane and Q is the ex- tinction efficiency. Assuming r = 1 µm, τ = 1, Q = 1, and ρ = 0.5 g/cm3 (Costantino and Daniels 1975), the mass density is 6.7×10−5 g/cm2. This is a factor of 300,000 times smaller than the typical 20 g/cm2 of total gaseous methane within the cloud layer. Spectral con- straints thus require that the condensed fraction must be so small as to play an insignificant role in the refrac- tivity profile. While a T (P ) solution consistent with the refractiv- ity measurements is possible with no methane at all in the atmosphere (Model A of L87), this was rejected be- cause methane clearly plays a major role in shaping the spectrum of Uranus. As the mixing ratio above the cloud is increased, the inferred temperature must in- crease to maintain the observed refractivity, and so does the methane mixing ratio inferred for the deep atmo- sphere. The maximum cloud-top humidity inferred by L87 is 53%, which leads to a deep mixing ratio of 4%, although this solution did not yield the highest humid- ity level within the cloud layer. Trying to increase the above-cloud humidity any further results in rapidly in- creasing temperatures into the cloud layer and unaccept- able superadiabatic lapse rates. None of these profiles yield anything close to saturation at the altitudes where cloud condensation is suspected. 2.2. Reconstructing the refractivity profile. With the aim of conducting a reanalysis of the occul- tation profile with different assumptions, we first needed to create a detailed refractivity profile. We began with the tabulation of P, T, molecular weight, number density, and mixing ratio published by L87. We used refractivity values per molecule of KHe= 0.5062, KH2 =0.1302, and KCH4=1.629 (all in units of 10−17 cm−3), which are the same as those used by L87 and referenced therein. We then used the relation K(z) = KH2fH2 (z) + KHefHe(z) + KCH4 fCH4(z) (2) to compute the mean molecular refractivity, where sub- scripted f values denote numeric fractions for each molecule. We also used the same 15/85 ratio of He to H2 as L87. The total refractivity for the mixture is then given by N (z) = n(z)K(z), (3) where n(z) is the total number density at altitude z. Our computed refractivity profile versus altitude is shown in Fig. 1. To validate that the refractivity profile we computed was consistent with the temperature profile, we inverted the profile as follows. We computed the pressure vs al- titude under the assumption of hydrostatic equilibrium using P (z) = P0 + Z z z0 M (z)(n(z)/NA)g(z)dz, (4) where M(z) is the molecular weight in grams per mole, NA is Avogadro's number, and g(z) is the gravitational 4 Sromovsky et al. Fig. 1. -- Refractivity profile computed from the L87 model D profile. The inset provides a detailed view of the sudden slope change, and there horizontal lines indicate cloud boundaries for our D1 (dotted) and F1 (dashed) structure solutions (discussed in Sec. 3). acceleration as a function of altitude, which varies from 8.6843 m/s2 at 1 bar (0 km) to 8.5157 m/s2 at 240 km, assuming a retrograde wind of 100 m/s (Sromovsky et al. 2009), which reduces g by only 0.23% and could be ig- nored. We started the downward integration at 240 km above the 1-bar level, and took the starting pressure to be 2.5×10−4 bar to match the stratospheric profile of L87. (The specific value of P0 has an insignificant effect on the structure for pressures greater than a few hundred millibars.) Assuming an ideal gas equation of state we then computed T (P ) from pressure and number density. Our T (P ) profile thus constructed is compared with the L87 profile in Fig. 2, where we see that differences be- low the tropopause are generally smaller than 0.1◦ (the RMS deviation is 0.07 K). This provides a reasonable validation of our reconstructed refractivity profile, which we will reanalyze with different assumptions in a subse- quent section, after first validating our methane retrieval procedure. 2.3. Validation of methane profile retrieval We next tried to reproduce the methane profile re- trievals obtained by L87. We started with the refractivity profile, the assumed He/H2 ratio of 15/85, selected alti- tudes for the cloudy layer, and selected a constant rela- tive humidity above the cloud layer up to the tropopause, and above the tropopause assumed a constant methane mixing ratio equal to the tropopause value, although later we used the 10−5 upper limit of Orton et al. (1987). Within the cloud layer we forced the lapse rate to agree with Eq. 2. We then started at the top of the atmo- sphere and integrated downward the number density and pressure and iteratively solving for fCH4 and n(z) under Fig. 2. -- A: T (P ) profile derived from refractivity profile of Fig. 1 (solid) using L87 tabulated results, and the L87 model D profile (symbols). B: the difference profile. the constraints of the specified methane humidity above the cloud, the constraints of the temperature lapse rate within the cloud, and the fixed mixing ratio below the cloud, which was taken to be the mixing ratio at the bottom of the cloud layer. Our attempt to reproduce the Model D solution is dis- played in Fig. 3. In most respects our results are barely distinguishable from those of L87. We obtain a deep CH4 VMR of 2.22% compared to their value of 2.26%, and our maximum CH4 RH at the cloud base is 79.2% compared to their 78%. These might be brought into better agree- ment with slightly different choices for the cloud bound- aries. We did not assign any of the methane fraction to condensed material; we don't know whether L87 did or not for this model. We take these comparisons to be ad- equate validation of our inversion technique. Note that the T (P ) and CH4 profiles of KT2009, which are also plotted in Fig. 3, have significantly higher temperatures and much more methane above the cloud top. If we increase the above cloud relative humidity as much as plausible, which we estimate to be about 57% (instead of the 53% of L87), we get results very close to the L87 Model F profile. We obtained a deep CH4 VMR of 4.13% (instead of 4%) and a peak in-cloud humidity of 70% instead of 72%. The temperature profile is also close to the Model F profile of L87. Even at this upper limit, however, the above cloud methane is well below what was adopted by KT2009. Pushing the methane values above this level results in physically unacceptable results: the in-cloud humidity becomes lower than the humidity above the cloud and the sub-cloud lapse rate becomes more and more unstable. We would argue that even this upper limit leads to physically implausible pro- files because of the relatively low humidity in the cloud layer. Yet good fits to the spectral observations seem to need even more methane at these levels. In the following Methane on Uranus: Compact cloud layer and high-latitude depletion 5 Fig. 3. -- A: Inverted T (P ) profile for 30% RH above cloud (our Model D) compared to the L87 Model D (dotted) and the profile adopted by KT2009 (+). B: relative humidity profile of the inverted T (P ) profile. C: Methane mixing ratio profile we inverted (solid) compared to the L87 Model D (dotted) and that adopted by KT2009 (+), with the saturation mixing ratio shown as the dot-dash curve, using our T (P ) profile. D: detailed views of the methane mixing ratio profiles. The extra dashed curve is the saturation methane mixing ratio computed from the KT2009 T (P ) profile. Note the very close agreement between our inversion and that of L87, in several cases too close to distinguish their difference. The horizontal dotted lines are at pressures of 1.179 bars and 1.278 bars, which correspond to altitudes of 5.25 km and 7.75 km below the 1-bar level. section we show how these levels can be reached. 3. REVISED ANALYSIS OF THE REFRACTIVITY PROFILE. Now that we have validated our analysis techniques, we apply these techniques to obtain new solutions for temperature and methane profiles. We first revised the methane mixing ratio at the tropopause and throughout the stratosphere to equal the Orton et al. (1987) upper limit of 1×10−5, and used a variable relative humidity between the tropopause and the cloud top, using the for- mulation RH(z) = RHtrop + (RHctop − RHtrop)(1 − (z − zctop)/(ztrop − zctop))x, (5) where a tropopause humidity RHtrop of about 12% is needed to match the desired minimum methane VMR, and the tropopause height zctop is taken to be 45 km. An exponent x=1 provides linear interpolation, and a decrease of humidity above the cloud top that is similar to that adopted by KT2009. These changes made no sig- nificant difference in the plausible upper limit of methane mixing ratios in the vicinity of the cloud layer and in the deep atmosphere. We also tried adding neon to the at- mosphere, using the mixing ratio of 0.0004 suggested by Conrath et al. (1987). This has a very small but unde- sirable effect that reduces the upper limit on methane. What is really needed is a lower background molecular weight, which then requires increased methane to match the refractivity profile. The only plausible way to obtain a lower molecular weight is to decrease the mixing ratio of He, which is what we proceeded to do in the following fashion. Within the cloud layer we used the same formu- lation as given by Eq. 2, but decreased the He VMR (at all altitudes) as needed to obtain methane condensation within most of the putative cloud layer. Below the cloud layer we used the same methane mixing ratio as found at the bottom of the cloud layer. Decreasing the He mixing ratio has two beneficial ef- fects: it allows methane saturation mixing ratios to be attained within the layer where condensation is expected to occur, and it allows a higher maximum methane mix- ing ratio that is more likely to be consistent with near-IR and visible spectra of Uranus. Our first example solution (Model D1, Fig. 4) uses a He VMR of 0.126, and yields a saturated methane mixing ratio through the bottom half of the nominal cloud region, and a deep mixing ratio of 2.22% for an above-cloud humidity of 38%. This is sim- ilar to Model D of L87, but is more physically plausible. A second example (Model F1), which attains the same deep mixing ratio of Model F (4%) is shown in Fig. 5. This solution is notable in having much more methane above the cloud layer than Model F, and provides a close 6 Sromovsky et al. Fig. 4. -- As in Fig. 3 except that we plot our Model D1 profile, which uses a volume mixing ratio of 0.126 for He, an above cloud humidity of 38%, linearly interpolated to a tropopause humidity of 12%. The horizontal dotted lines are at pressures of 1.136 bars and 1.251 bars, which correspond to altitudes of 5.25 and 8.25 km below the 1-bar level. match to the CH4 VMR profile adopted by KT2009, although our corresponding T (P ) profile is somewhat cooler than their adopted profile, as needed to match the refractivity profile. The He VMR for this solution is 0.1155, which is only 1.05σ below the nominal value of 0.15±0.033, given by Conrath et al. (1987). Our maxi- mum methane solution (our Model G in Table 1) is ob- tained for a He VMR of 0.1063, which is only 1.3σ below the nominal value. This solution has an above-cloud hu- midity of 100%. This provides a deep CH4 VMR of 4.88% and even more methane above the cloud top than that adopted by KT2009. A summary of the above model profiles and interme- diate model EF is provided in Fig. 6 and detailed pa- rameter information in Table 1. The small difference in tropopause temperatures (at p = 0.1 bar) is due to the different helium mixing ratios used in each model. The maximum temperatures reached near p= 2.3 bars for our models D1-G are comparable to those obtained by L87 for their models C-F. We now have a range of solutions that are consis- tent with occultation results, as well as being consis- tent with the expectation that methane humidity levels should reach saturation levels in the putative condensa- tion region, which is presumably the physical reason for the sudden change in refractivity slope. The remaining question is whether it is possible to fit the spectra of Uranus with a cloud particle layer in the same pressure regime where occultation analysis implies a cloud layer. We first describe our chosen spectral constraints and ra- diation transfer and fitting methods, and then proceed to describe the results of applying the spectral constraints. 4. USE OF URANUS SPECTRAL CONSTRAINTS 4.1. Spectral observations We chose for our spectral comparisons the spatially re- solved STIS observations made on 19 August 2002, as corrected and calibrated by KT2009. These spectra are undoubtedly the most accurately calibrated and char- acterized spectra available for the purpose. The data provide two spatial dimensions, with a pixel size of 0.05 arc seconds on a side, providing 37 samples from center to limb, and one spectral dimension, which is sampled at 0.4-nm intervals, providing 1-nm resolution from 300 nm to 1000 nm. A sample image from this cube is pro- vided in Fig. 7A. These data provide a good view of the low latitude region sampled by the Voyager radio occul- tation experiment, although some 16 years later, which is a delay of 20% of a Uranus year. However, given the long radiative time constants in the Uranus atmosphere (Conrath et al. 1990), it is plausible to expect only small changes in cloud structure in low latitude regions. Stabil- ity at low latitudes also seems to be compatible with the analysis of HST images from 1994 to 2000 Karkoschka (2001), which indicate little change in albedo structure at these latitudes. In addition to spectral constraints, the KT2009 data provide important center-to-limb (CTL) information Methane on Uranus: Compact cloud layer and high-latitude depletion 7 Fig. 5. -- As in Fig. 4 except that we show our Model F1 profile, which uses a volume mixing ratio of 0.1155 for He, and an above cloud humidity of 83%. The horizontal dotted lines are at pressures of 1.179 bars and 1.278 bars, which correspond to altitudes of 5.25 and 7.75 km below the 1-bar level. Fig. 6. -- A summary plot comparing the profiles of temperature (left) and methane mixing ratios (right) for models D1, E1, EF, F1, and G. 8 Sromovsky et al. (Fig. 7B). To reduce the effects of noise, we fit the center- to-limb scans to a smooth function of µ (the cosine of the zenith angle), as illustrated in Fig. 7B for a planetocen- tric latitude of 5◦ S. We then sampled that function at µ = 0.3, 0.4, 0.6, and 0.8. The resulting substantial noise reduction is readily apparent in the figure, where uncer- tainties are shown by roughly parallel dot-dash lines at 1σ limits. We were able to ignore the difference between solar and observer zenith angles because the STIS obser- vations were taken at a very small phase angle (0.04◦). We also were able to ignore possible longitudinal varia- tions in cloud structure because the discrete cloud fea- tures of Uranus are generally so small and of such low contrast that they do not obscure the CTL information of the zonal bands. We also know from prior experi- ence with differences between Uranus images taken at substantially different central meridian longitudes that the only substantial longitudinal I/F variations are due to discrete features. At the few latitudes where discrete features were found in the STIS data the fits easily in- terpolated across them. The selected 5◦ S STIS spectra interpolated to five cosine values are displayed in Fig. 7. Note the strong limb darkening at short wavelengths and the strong limb brightening at centers of the methane ab- sorption bands at longer wavelengths. The spectrum for µ = 0.2, which is noisier and not well corrected because it is too close to the limb, was not used in our analysis. To further reduce noise in the observations, and to fa- cilitate model comparisons, we also smoothed the ob- served spectra to a uniform wavenumber resolution us- ing a 36 cm−1 boxcar. A uniform wavenumber resolution was chosen to fit the requirements of our Raman scatter- ing code (Sromovsky 2005a), and the specific resolution is chosen to be both commensurate with the wavenum- ber shifts of the three most important Raman transitions and to approximately match the resolution provided by the STIS observations at 550 nm. The spectral smooth- ing and sampling of the fits to the CTL variations made noise in the observations a fairly small contributor to the total uncertainty as demonstrated in Sec. 4.4. The sensitivity of these spectra to different atmo- spheric levels on Uranus is indicated in Fig. 8. This indi- cates the penetration depth of light into an aerosol-free atmosphere of Uranus by the plot of pressures at which one-way unit vertical optical depth is reached. Pene- tration depths for individual contributions by Rayleigh scattering, methane absorption and hydrogen collision- induced absorption (H2 CIA) are also shown. A key wavelength region is near 0.825 µm, where H2 CIA opac- ity exceeds the opacity of methane. Poor fits in this region are an indication of incompatible vertical distri- butions of methane and hydrogen absorptions, and may be a result of assuming an incorrect methane mixing ra- tio. 4.2. Radiation transfer calculations We used the radiation transfer code described by Sromovsky (2005a), which include Raman scattering and polarization effects on outgoing intensity. To save com- putational time we employed the accurate polarization correction described by Sromovsky (2005b). After trial calculations to determine the effect of different quadra- ture schemes on the computed spectra, we decided to use 14 zenith angle quadrature points per hemisphere and a 14-order azimuthal expansion for our compact model and 10 quadrature points for fitting models with the KT2009 structure. To characterize methane absorption we used the corrected coefficients of KT2009. To model collision- induced absorption (CIA) of H2-H2 and He-H2 interac- tions we interpolated tables of absorption coefficients as a function of pressure and temperature that were com- puted with a program provided by Alexandra Borysow (Borysow et al. 2000), and available at the Atmospheres Node of NASA'S Planetary Data System. We assumed equilibrium hydrogen for most calculations but did look into the effects of non-equilibrium distributions, which are discussed in Sec. 8.2 4.3. Cloud models We used two distinct models of cloud structure for comparison purposes. The first is nearly identical to the model of KT2009, which we will refer to as the KT2009 model, and the second is a modified version, which we will refer to as the compact cloud layer model. A plot of the vertical distribution of these layers can be found in Sec. 5.3 (Fig. 14). 4.3.1. The KT2009 model This model has four layers of aerosols, the uppermost being a Mie-scattering stratospheric haze layer charac- terized by an optical depth at 0.9 µm, a gamma size dis- tribution (Hansen 1971), with a mean radius of a =0.1 µm and a normalized variance of b =0.3. These particles are assumed to have a real index of 1.4, and an imaginary index following the KT2009 relation ni(λ) = 0.055 exp[(350 − λ)/100], (6) for λ in nm. This haze was distributed vertically above the 100 mb level with a constant optical depth per bar. The remaining layers in the KT2009 model are characterized by a wavelength-independent optical depth per bar and a wavelength-dependent single-scattering albedo, given by t(λ) = 1 − 1/[2 + exp[(λ − 290)/37]], (7) again for λ in nm. Their adopted double Henyey- Greenstein phase function for the tropospheric layers used g1 =0.7, g2=-0.3, and a wavelength-dependent frac- tion for the first term, given by f1(λ) = 0.94 − 0.47 sin4[(1000 − λ)/445], (8) which produces a backscatter that decreases with wave- length, as shown in Fig. 9. The three tropospheric lay- ers are uniformly mixed with gas molecules, with differ- ent optical depths per bar in three distinct layers: 0.1- 1.2 bars (upper troposphere), 1.2-2 bars (middle tropo- sphere), and P>2 bars (lower troposphere). These op- tical depths are the adjustable parameters we use to fit this model to the observations. 4.3.2. The compact cloud layer model This model is a modification of the KT2009 model. The main change we made is to replace their middle tropospheric layer with two compact layers: an upper tropospheric compact cloud layer and a middle tropo- spheric compact cloud layer. This allows the possibility Methane on Uranus: Compact cloud layer and high-latitude depletion 9 Occultation analysis parameters and characteristic results for our models. TABLE 1 Model name: D Tropopause RH Cloud top RH Maximum RH Neon VMR 30% 30% 79% 0.0 Cloud Top,km -5.25 Cloud Bot.,km -7.75 Cloud Top, bar 1.179 Cloud Bot., bar 1.278 He VMR ∆VMR/σVMR 0.15 0.00 D1 20% 38% E1 20% 70% EF 20% 78% F1 20% 83% G 20% 100% 100% 100% 100% 100% 100% 0.0004 0.0004 0.0004 0.0004 0.0004 -5.25 -8.25 1.136 1.251 0.126 -0.73 -5.50 -8.00 1.142 1.241 -5.70 -7.80 1.142 1.226 -5.90 -7.75 1.146 1.221 -6.50 -7.75 1.153 1.205 0.122 0.1179 0.1155 0.1063 -0.85 -0.97 -1.05 -1.32 He/H2 0.1765 0.1442 0.1390 0.1337 0.1306 0.1189 Deep CH4 VMR 2.22% 2.22% 3.24% 3.64% 4.00% 4.88% CH4 km-am to Cld Bot. 0.314 0.340 0.576 0.610 0.658 0.719 Fig. 7. -- A: Sample image from the unsmoothed KT2009 STIS data cube at 620 nm, with grayed pixels indicating our center-to-limb sampling at 5◦ S, with grid lines at 30◦ intervals. B: Sample center-to-limb scans at four wavelengths (after spectral smoothing with a 36 cm−1 boxcar), with fits shown by dot-dash lines bounding 1σ uncertainty limits. C: Sample interpolated STIS I/F spectra at 5◦ S planetocentric latitude after CTL fitting and spectral smoothing. 10 Sromovsky et al. from that value in preliminary fits. For the middle tro- pospheric cloud (MTC) in our model we used particles with the same scattering properties as given by KT2009 for their tropospheric particles. Both of these compact layers have the bottom pressure as a free (adjustable) parameter and a top pressure that is a fixed fraction of 0.93 of the bottom pressure. This degree of confinement is approximately the same as obtained for the cloud layer inferred from the occultation analysis. We assumed sim- ilar confinement for the deeper layer in our model, but it could easily be more vertically diffuse than we assumed, as long as the effective pressure is similar. The last change we made was to replace their lower tropospheric layer by a compact cloud layer at 5 bars (the LTC), with adjustable optical depth and with the KT2009 tropospheric scattering properties. We found that this layer was needed to provide accurate fits near 0.56 and 0.59 µm, but its pressure is not well constrained by the observations (the effect of varying the pressure can be compensated by varying the optical depth, to produce essentially the same fit quality). Whether this deep cloud is vertically diffuse or compact also cannot be well constrained, and thus our assumption of a compact cloud for this layer is a matter of convenience and is not compelled by observations. The wavelength dependence of the backscatter phase function, extinction efficiency, and single scattering albedo of the stratospheric haze are given in Fig. 9 for both the Mie particles we used for the putative methane layer (the UTC) and for the KT2009 tropospheric particles. The latter have wave- length independent optical depth, and thus the way its backscatter efficiency varies with wavelength is entirely determined by the phase function (defined by Eq. 8). The best-fit Mie particle size results in a smaller vari- ation in backscatter efficiency, although both decrease with wavelength. In summary, we consider two models. The diffuse one has the KT2009 structure, which provides a fitting stan- dard of comparison. The compact model, the main fea- ture of which is the splitting of the middle tropospheric layer of KT2009 into two layers, allows us to see if a compact layer of methane particles can provide good fits to the observed spectra, and to see which occultation- derived profile of temperature and methane mixing ra- tio provides (1) the best fit to the spectra and (2) the best agreement between the fit pressure for the middle tropospheric layer and the pressure inferred from the oc- cultation analysis. Hopefully, the best spectral fit would occur for the same profile that provided the best pressure match. As shown in the following, that is roughly what happened. 4.4. Error model To measure fit quality we use χ2, which requires an estimate of the expected difference between a model and the observations due to the uncertainties in both. We roughly characterized the uncertainty in measurements by the expected uncertainty in the samples of the smooth fits, as described previously. This uncertainty is shown by the purple curve in the upper panel of Fig. 10. There is also an overall calibration uncertainty, estimated by KT2009 to be 5%. However, calibration errors for spec- tra are similar to nearly wavelength-independent scale factor errors, which tend to cause changes in inferred op- Fig. 8. -- Pressure at which the vertical optical depth to space reached unity, shown for individual components including Raman scattering (dotted), Rayleigh scattering (dashed), methane absorp- tion (triple-dot-dash), and H2 CIA (dot-dash). The pressure for all effects combined is shown by the dark solid curve for unit optical depth and by the light solid curve for τ = 10. Fig. 9. -- Scattering properties of the stratospheric haze model (solid), the middle tropospheric Mie layer (dot-dash) and KT2009 tropospheric particles (plus signs). A: extinction efficiency (Q, black), single-scattering albedo (, red) and backscatter phase function (P (π), green). B: backscatter efficiency. of a better match to the observations if the aerosols be- tween 1 and 2 bars do not match the KT2009 assumption of being uniformly mixed with the gas. The upper tro- pospheric cloud (UTC) in our model is composed of Mie particles, which we characterized by a gamma size distri- bution with a mean particle radius of 1.2 µm and a fixed normalized variance of 0.1, a fixed refractive index of 1.4, and an imaginary index of zero. The particle radius was fixed at the given value because it did not vary much Methane on Uranus: Compact cloud layer and high-latitude depletion i.e. I(λ, µ) = I0(µ) exp(−ku/µ) 11 (9) where the optical depth is ku, k is the absorption coeffi- cient, and u is the absorber path amount. This directly leads to a fractional error in I/F given by δI/I = 1 − [I(λ, µ)/I0(µ)]α+ǫ0/k (10) which differs from the model of Sromovsky and Fry (2010) by using the maximum I0(µ) instead of unity, and by including the ǫ0/k term in the exponent. That term becomes dominant for small k, and expresses the fact that small absorptions that cause large I/F depressions lead to large uncertainties because the offset uncertainty begins to be comparable to the total absorption coeffi- cient. However, in this crude form the term becomes unstable when the absorption is very small and I/I0 is near 1, but where (because of our crude model) it is not close enough to 1 to make the ǫ0/k term unimpor- tant. To counteract this instability we replace ǫ0/k with ǫ0/(k + kmin), where we chose kmin=0.01 to avoid exces- sively large errors for λ < 0.5 µm. The final form of this model is shown as the CH4 uncertainty in Fig. 10. This was treated as normally distributed with the standard deviation as shown, which is another crude assumption. However, it is likely that the overall dominance of this error source at wavelengths exceeding about 0.6 µm is correct. Our final combined error estimate was the square root of the sum of the squares of the three sources. Although these error sources are not accurately known or well char- acterized by our noise model, they can be roughly vali- dated by comparing, for the best fit over all models and profiles, the minimum χ2 value with the expected value for a perfect fit. Our very best fit over this spectral range (at this latitude) yielded approximately 339 for χ2 instead of the value of 269 predicted by the error model. This would be the result of under-estimating the com- bined errors by 12%, or by a defect in the physical model we are using to fit to the observations (at other latitudes we obtained χ2 values as low as 269). 4.5. Fitting procedures To avoid the need for fitting an imaginary index in the methane layer (the UTC layer) we fit only the wavelength range from 0.55 µm to 1.0 µm. To provide a reasonable computational speed while still sampling a wide range of penetration depths for each spectral band, we chose a wavenumber step of 118.86 cm−1 for sampling the ob- served and calculated spectrum. This yielded 69 spectral samples, each at four different zenith angle cosines (0.3, 0.4, 0.6, and 0.8), for a total of 276 points of comparison. Our compact layer model has seven adjustable parame- ters (pressures and optical depths of the UTC and MTC layers, the optical depth per bar of the stratospheric haze, the optical depth per bar of the upper tropospheric haze (KT2009 referred to this as the upper tropospheric layer), and the optical depth of the LTC layer, leaving 269 degrees of freedom. We fixed the LTC base pressure to 5 bars and the UTC mean particle radius at 1.2 µm be- cause these values were consistently obtained for a vari- ety of fits, and reducing the number of adjustable param- eters improved fit algorithm performance. Furthermore, Fig. 10. -- A: Estimated contributions by measurement uncer- tainty (purple), methane coefficient uncertainty (blue), and overall modeling errors (red), to the combined relative I/F uncertainty for µ=0.6 (black). B: Combined Fractional error vs. wavelength at zenith angle cosines of 0.3 (purple), 0.6 (black), and 0.8 (green). tical depths of aerosols, with very little effect on their in- ferred vertical distributions, and thus we did not include this as an error source. KT2009 estimated relative uncer- tainties of their corrected spectra to be within 1%, which were estimated by comparison of the corrected spectrally weighted STIS observations to band-pass filtered images. To this relative calibration uncertainty we added an over- all modeling uncertainty of 1% for a combined relative fractional error of 1.4%, shown by the red curve in Fig. 10. Another important source of uncertainty is due to un- certainty in the methane absorption coefficients, which is not easy to characterize. We assumed that the un- certainty in k had the form ǫ(k) = αk + ǫ0, where we adopted values of α = 0.02 and ǫ0 = 5×10−4 km-am. The 2% scale-factor component (α) is roughly in agree- ment with the verification provided by the Descent Im- ager/Spectral Radiometer (DISR) measurements within Titan's atmosphere (Tomasko et al. 2008) over the 0.5- 0.75 µm wavelength range, though larger errors are in- dicated at longer wavelengths. The offset error is even less certain. KT2009 made changes as large as 0.01 km- am from previous coefficients (Karkoschka 1998), and the new values, which we use here, are likely to be much smaller, but by exactly what factor is unclear. We of the used a refinement of Sromovsky and Fry (2010) in approximating the ef- fect of methane absorption uncertainty on the spectrum. We used a similar crude approximation, that the reflected intensity at any wavelength could be expressed as a maximum times an exponential in optical depth, approach 12 Sromovsky et al. the LTC pressure is not well constrained by the observa- tions because its change can be compensated by changing the LTC optical depth. To fit the KT2009 model over the same range, we followed their approach by adjusting only the four dτ /dP (optical depth per bar) values. We use a modified Levenberg-Marquardt non-linear fitting algorithm (Sromovsky and Fry 2010) to adjust the fitted parameters to minimize χ2 and to estimate uncertain- ties in the fitted parameters. The uncertainty in χ2 is expected to be ∼25, and thus fit differences within this range are not of significantly different quality. 5. APPLICATION OF SPECTRAL CONSTRAINTS TO THE OCCULTATION SOLUTIONS 5.1. Fit results for 5◦ S The fit results for the compact model for profiles D1, E1, EF, F1, and G are given in Table 2 and key results are plotted as a function of methane volume mixing ra- tio in Fig. 11. The parameters of the aerosol model are very well constrained by the observations, with pressures constrained to a fraction of a percent and optical depths usually to within 5%. In Fig. 11 we show several results that can be used to constrain the methane mixing ratio at 5◦ S: (A) the pressure of the upper tropospheric cloud (UTC) in comparison with the occultation cloud pres- sure, (B) the overall quality of the spectral fit, (C) the fit error at 0.825 µm, and (D), the He/H2 ratio. The results for cloud pressure (Fig. 11 A) show that the upper compact cloud (modeled as methane) is in best agreement with the occultation pressure range at a methane mixing ratio of 4.0%, but the match is still fairly close down to a mixing ratio near 3.5%. An even stronger constraint on the CH4 mixing ratio is the fitting error in the region near 0.825 µm (Fig. 11B), where hy- drogen CIA exceeds methane absorption. If there is too much methane assumed in our model, then model cloud particles will need to move upward to compensate. That will place them relatively further above the absorption of hydrogen, and where the effects of hydrogen can be seen (as at 0.825 µm) the model will appear too bright rela- tive to the observations. Where the assumed CH4 mixing ratio is too low (as for Model D1 at 5◦ S) the model I/F will appear be too low at 0.825 µm. This constraint clearly favors a mixing ratio of 4.5% with an uncertainty of roughly 0.7%. The overlap of the first two uncertainty ranges is 3.8-4.5%. The third constraint is overall fit quality (Fig. 11 C). The best spectral fit (judged from the minimum χ2 in the middle panel) is for a methane mixing ratio near 3.6%, but the fit is nearly as good over a wide range from 3-4.9%. This is a relatively weak constraint that is easily compatible with the previous stronger constraints, which favor a mixing ratio of 4%. This makes our Model F1 profile the preferred profile, even though it slightly exceeds the Conrath et al. (1987) uncertainty limits of the He VMR, as shown in Fig. 11D. We consider E1, EF, and F1 profiles plausible candidates for further analysis. Our preferred model (F1) leads to a compact methane condensation cloud very close to pres- sure level expected from occultation results and it is also a model that provides an excellent fit to the STIS spectral observations, especially in the key region where hydrogen CIA is significant. The best compact cloud model fit to STIS spectra at Fig. 11. -- Fit results for compact cloud layer models as a function of the deep volume mixing ratio of methane (points are plotted for models D1, E1, EF, F1, and G from left to right). A: pressures of the upper (presumably CH4) and middle tropospheric cloud models compared to the cloud layer inferred from our occultation analysis (solid lines). B: Fit error at 0.825 µm where hydrogen CIA is prominent. C: Minimum χ2 values (solid line) for fits to spectra at 4 zenith angles and 69 wavelengths from 0.55 to 1.0 µm, with dotted lines indicating the 1σ uncertainty range expected for χ2. D: He volume mixing ratio (line) compared to the Conrath et al. (1987) value (symbol with error bars). The grayed areas indicate rough regions of uncertainty/acceptability. 5◦ S using the F1 profile (shown in Fig. 12) yielded an uncorrected χ2 of 342, which is significantly better than the best χ2 values of 409, 426, and 458 obtained by our fit of the KT2009 model to E1, EF, and F1 profiles re- spectively. Thus, we don't have to give up anything in fit quality to obtain the compatibility between aerosol models and occultation models. In fact, the fit quality improves, as one would expect with two more adjustable parameters available. Our presumed methane cloud has an effective particle radius near 1.2 µm and a rather small optical depth, only about 0.32 at λ = 0.5 µm, which about half the value re- ported by Rages et al. (1991) for their low latitude cloud, but close to the 0.4 value of Baines et al. (1995) for a disk average that weighted high latitudes most. The middle tropospheric cloud is considerably thicker, with an opti- Methane on Uranus: Compact cloud layer and high-latitude depletion 13 Best-fit parameters for compact cloud layer models at 5◦ S. TABLE 2 OCCULTATION PROFILE: D1 E1 EF F1 G KEY PROFILE PARAMETERS: Cloud Top, bar Cloud Bottom, bar 1.136 1.251 Deep Methane VMR 2.22% AEROSOL PARAMETERS: (dτ /dP )Strat.H, bar−1 (dτ /dP )UTH , bar−1 PUTC, bar PMTC, bar PLTC, bar 0.329 0.01 1.45 2.54 5. 1.142 1.241 3.24% 0.208 0.032 1.27 1.80 5. 1.142 1.226 3.64% 0.178 0.035 1.24 1.72 5. 1.146 1.221 1.153 1.205 4.00% 4.88% 0.158 0.038 1.23 1.68 5. 0.136 0.031 1.19 1.60 5. τ UTC τ MTC τ LTC rStrat.H, µm rUTC, µm FIT QUALITY: 2(total) 2(0.825 µm) Fit error at 0.825 µm, µ=0.8 χ χ 0.444 0.328 0.324 0.322 0.324 1.33 0.01 0.1 1.2 435.6 47.7 -4.54 1.16 2.30 0.1 1.2 340.5 11.0 -2.14 1.23 2.92 0.1 1.2 338.5 5.1 -1.34 1.28 3.45 0.1 1.2 342.3 2.2 -0.71 1.42 4.75 0.1 1.2 351.9 0.6 +0.54 Note: The uncertainty in χ2 is ∼25 and thus fits differing by less than this are not of significantly different quality. The stratospheric haze base extends upward from 100 mb, the upper tropospheric haze (UTH) extends from 900 mb to 100 mb, The upper tropospheric cloud (UTC) and middle tropospheric cloud (MTC) extend from the bottom pressures, tabulated here, to 0.93 times those pressures. The lower tropospheric cloud is bounded by the tabulated base pressure and 0.98 times that pressure. For these fits, the particle radii of the Mie layers were held fixed, as was the pressure of the lower tropospheric cloud. The fit error at 0.825 µm is the ratio of (model-measured) to estimated uncertainty. cal depth of 1.23 for the optimum methane profile. The much deeper lower tropospheric cloud is generally even thicker, except for the model D1 profile for which this cloud is not even needed. 5.2. Layer contributions The relative roles played by our five model layers in creating the observed spectral characteristics are illus- trated in Fig. 13B-F, which displays the effect of remov- ing each layer from the model spectrum. The distinctly different contributions of the various layers is what allows the model parameters to be so well constrained by the observations. We see in B that the stratospheric haze (layer 1) serves to reduce the I/F for λ < 0.6 µm and provide a slight (5-10%) boost to the I/F in the center of the deep absorption bands at longer wavelengths. As shown in C, the upper tropospheric haze layer (layer 2) makes a similar contribution but without as much short- wave absorption. The methane cloud (layer 3, panel D) contributes almost nothing at wavelengths less than 0.55 µm but provides a significant contribution to minima in the intermediate absorption bands and especially to the shoulders of the strong absorption bands. The effect of the middle tropospheric cloud (layer 4, panel E) is seen mainly in the strong contributions to the I/F peaks, and also in contributing a small absorption peaking near 0.4 µm. The lower tropospheric cloud (layer 5, panel F) con- Fig. 12. -- A: STIS 5◦ S spectra (solid lines) at four zenith an- gles, sampled at a wavenumber spacing of 118.86 cm−1, compared to model calculations (points) for the model that best fits the spec- trum when the F1 profile of temperature and methane mixing ratio is assumed. B: Ratio of Model to observed I/F values. C: Differ- ence of model and observed values divided by the expected uncer- tainty at each wavelength sample. Note that the larger fractional errors that occur for λ >0.78 µm are within the expected range of uncertainties. 14 Sromovsky et al. tributes mainly to the longer wave window regions, and is unique in providing key contributions at 0.55 and 0.59 µm. Note that, although the model fit was only con- strained by measurements from 0.55 to 1.0 µm, it pro- vides a relatively good fit to the shorter wavelengths as well (panel A). The discrepancies between model fit and measurements between 0.3 and 0.4 µm are only about 10% and could be largely removed by slight increases in the lower tropospheric cloud absorption and slight de- creases in the stratospheric haze absorption in this re- gion. It might also be possible to introduce some of the extra absorption needed at small zenith angles into the methane layer itself, as suggested by Rages et al. (1991). 5.3. Comparison of vertical structures The vertical structure of our compact cloud model and the diffuse vertical structure of the KT2009 model are compared directly in Fig. 14, which displays opti- cal depth per unit pressure versus pressure (A) and cu- mulative optical depth versus pressure (B). There is a very large difference in the vertical distribution of cloud particles, with our compact model providing high opac- ity concentrations in narrow pressure ranges. But when compared on the basis of cumulative opacity we see that the KT2009 model looks like a vertically averaged ver- sion of our results. The compact layer with the strongest effect on the CCD spectrum is our layer 4 (MTC), as shown in Fig. 13. The dashed line in Fig. 14A is for op- tical depth per bar at 1.6 µm from Fig. 5 of Irwin et al. (2010) with pressures scaled by the factor 1.8/2.6 to ac- count for the decrease indicated in their Fig. 8, when they used a vertical methane profile similar to the L87 Model F. Their relative vertical distribution of opacity is crudely consistent with that of the other models, when vertical smoothing is taken into account. However, the quantitative values of their opacities per unit pressure are hard to compare with our (or KT2009) results be- cause of very different assumptions made about particle scattering properties, including the extreme assumption that particle scattering properties were independent of latitude and altitude. While the top compact layer in our model is consis- tent with condensation of methane, the composition of the deeper layers is highly uncertain. Possible parent gases are H2S and NH3, but condensation layers at 1.7-3 bars would imply strong depletions from their expected deep mixing ratios, which is plausibly an effect of the formation of a much deeper cloud of NH4SH particles, as discussed by de Pater et al. (1989) and Fegley et al. (1991). 6. LATITUDE DEPENDENCE OF METHANE ON URANUS We first show that the occultation solution that is most consistent with low latitude spectra is not consistent with high-latitude spectra and then discuss how that profile can be modified in physically reasonable ways and which modifications provide the best fits at high latitudes. 6.1. Fit results assuming latitude-independent methane EF and F1 model temperature and methane profiles yielded high quality spectral fits not only at the occulta- tion latitude, but also for latitudes from 30◦ S to 20◦ N, as indicated in Fig. 15A for the F1 fits. Over this latitude range the fit quality was generally very high, reaching χ2 values close to those expected from the error model, and about half were notably better than we obtained at 5◦ S. However, between 30◦ S and 45◦ S fit quality began to deteriorate dramatically, with χ2 growing to ∼1100 by 60◦ S, which is four times the expected value. At the same time, at 0.825 µm, where hydrogen CIA is an im- portant contributor, the χ2 contribution (Fig. 15B) and the signed fit error (Fig. 15C) remained relatively flat over the 30◦ S to 20◦ N region, but grew dramatically from 30◦ S to 60◦ S, with the signed error reaching 7 times the expected uncertainty and the χ2 contribution reaching more than 40 times the expected value. Thus, there is little question that the methane mixing ratio at high latitudes is quite different from what it is at middle latitudes, a result already established by KT2009. To get a more realistic picture of the high latitudes, we need to estimate what the actual methane mixing ratio profile is at these latitudes. In the next section we discuss plausi- ble ways to modify our baseline (F1) profile so that we can investigate this issue. 6.2. Construction of depleted methane profiles The occultation places no direct constraints on the temperature or methane profiles at other latitudes. How- ever, there are some reasonable physical constraints to consider. First, the lack of significant latitudinal gradi- ents in the temperatures inferred from Voyager 2 IRIS observations for P>∼150 mb (Hanel et al. 1986) sug- gests that the occultation thermal profile is a good ap- proximation at all latitudes on Uranus. A second con- straint has to do with an extension of the relationship between the vertical wind shear and horizontal temper- ature gradients. The variation of mixing ratio with lati- tude, assuming the same pressure-temperature structure at all latitudes (as assumed by KT2009 as well), leads to a variation in density on constant pressure surfaces, which is similar in effect to horizontal temperature gradi- ents. Both kinds of gradients lead to vertical wind shear (Sun et al. 1991), a consequence of geostrophic and hy- drostatic balance. If such latitudinal variations occurred through great atmospheric depths, this would lead to great differences in cloud level winds and conflict with the observed wind structure of Uranus. Thus we expect that where methane is depleted, it is not depleted over great depths. And, as indicated by KT2009, the spec- tral observations do not require that methane depletions extend to great depths. In fact, we will later show that the spectral constraints favor relatively shallow methane depletions. For trial purposes we created several types of modi- fications of the F1 profile. We first defined an upper tropospheric mixing ratio and a set of transition pres- sures, P1 and P2. for P > P2 we kept the mixing ratio at the F1 deep value of 4%. For P < P1 we set the mixing ratio to the upper level value, except where that value exceeded the F1 value above the methane condensation level, at which point we reverted to the F1 model pro- file. Between the two transition pressures we interpolated between the upper troposphere and deep values. Sam- ple depleted profiles of this type are shown in Fig. 16A. These profiles are consistent with methane condensation at somewhat lower pressures than for the F1 profile. A depleted profile of this type might be created by descent Methane on Uranus: Compact cloud layer and high-latitude depletion 15 Fig. 13. -- A: Comparison of measured STIS spectrum (red) at 5◦ S with a compact model fit (black), with spectra shown at a solar zenith angle cosine of 0.6 and ratios shown for all four cosine values); B-F: Comparisons of the model spectrum with spectra for same model with layers 1-5 removed respectively. Layer 3 (D) is the methane cloud layer. Although these spectral sensitivities were calculated for the EF profile, they also closely represent results for E1 and F1 profiles. of the low-mixing-ratio gas from slightly above the 1.2 bar level downward to higher pressures, but to limited depths. We also considered "proportionally descended gas" profiles in which the model F1 mixing ratio profile α(P ) was dropped down to increased pressure levels P ′(α) us- ing the equation P ′ = P × [1 + (α/αd)vx(Pd/Pcb − 1)] (11) for Ptr < P < Pd, where Pd is the pressure depth at which the revised mix- ing ratio α′ = α(P ′) equals the uniform deep mixing ra- tio αd , Pcb is the cloud bottom pressure (which is where the unperturbed profile departs from the uniform deep value), Ptr is the tropopause pressure (100 mb), and the exponent vx controls the shape of the profile between 100 mb and Pd. A sampling of profiles of this type are shown in Fig. 16B. The profiles with vx = 1 are similar in form to those adopted by Karkoschka and Tomasko (2011). A profile of this type is consistent with the descending gas beginning at lower pressures than for our initial set of models. In both cases we would expect methane conden- sation to be inhibited by the downward mixing of up- per tropospheric gas of low methane mixing ratios. How these various types of profiles affected fits to the 45◦ S and 60◦ S spectra is discussed in the next section. 6.3. Constraining methane profiles at 60◦ S and 45◦ S We used our 5-layer model, allowed adjustment of all 9 parameters, and tried CH4 mixing ratios as low as 0.3% in the depleted region, and transition depths ranging from 1.5-2 bars to as deep as 9-10 bars. The bottom of the transition region is where the F1 deep mixing ratio of 4% is reached. The results were evaluated by com- paring their overall fit quality (how low their χ2 values were) and how big their fitting errors were at the key wavelength of 0.825 µm, where H2 CIA is an important contributor. The results are summarized in Fig. 17 for depletion depths of 1.5-2 bars (only for 45◦ S), 2-3 bars, 16 Sromovsky et al. Fig. 14. -- Optical depth per bar (A) and cumulative optical depth (B) for a vertically diffuse KT2009 model (dotted) and out compact layer model (solid) fit to 5◦ S STIS spectra using the F1 vertical profile of temperature and methane mixing ratio. For the KT2009 model we fit the four optical depths per bar, and used the KT2009 values for pressures and particle scattering properties. For our model we fit two optical depths per bar for the upper two layers and the pressures and optical depths of the three compact layers, except for the fixed 5 bar pressure we used for the bottom layer. The dashed line in A is optical depth per bar at 1.6 µm from Fig. 5 of Irwin et al. (2010) with pressures scaled by the factor 1.8/2.6 to account for the decrease indicated in their Fig. 8, when they used a vertical methane profile similar to the L87 Model F. 4-6 bars, and 9-10 bars. At neither latitude is it possible to obtain a good fit with deep depletions that extend to 9-10 bars. Although a low hydrogen fitting error can be obtained from deep depletions, the overall fit quality is much better for shallow and very strong depletions. At both latitudes lower mixing ratios lead to better fits, as long as the depletion depth is reduced as well. The mini- mum hydrogen error moves to lower methane mixing ra- tios as the depletion depth is decreased and the depletion factor is increased. It is apparent that the best combi- nation of overall and CIA fits requires very low methane mixing ratios, of the order of 1% or even less, and shal- low depletion depths, down to only 1.5-2 bars for 45◦ S, and perhaps 2-3 bars at 60◦ S, which are illustrated in Fig. 16A. These perturbations of the base profile are of such a small depth that the vertical wind shear caused by the resulting horizontal density gradients would not likely produce significant perturbations of the zonal wind field, though exact calculations would be needed to verify that. The depletion of methane at high latitude helps to explain the difference between prior estimates of 4% by Rages et al. (1991) and 1.1-2.3% by Baines et al. (1995), the latter weighting high latitudes more than the former. A limited fitting exploration using more physically ap- pealing methane profiles of the type shown in Fig. 16B did not yield significantly better fits than those shown in Fig. 16A. We did find constraints on the parameters of such profiles, however, with vx = 1 yielding better fits than either vx = 0.5 or vx = 2, and Ppd = 2 − 3 bars preferred at 45◦ S and Ppd = 4 − 6 bars preferred at 60◦ S. These results present a picture in which dry upper tropospheric gas descends at high latitudes after being dried out by condensation of methane in upwelling gas Fig. 15. -- Spectral fit quality measured by χ2 over the entire range of the fit (A) and χ2 for the 4 view angles at the 0.825 µm H2 CIA wavelength (B), and fit error at 825 nm and µ=0.8 (C), all as a function of planetocentric latitude, using the F1 temperature and methane profiles. Filled circles in A denote results of fitting the KT2009 structure model. The horizontal dotted lines in A and C indicate expected (central) and 1σ uncertainty bounds. Methane on Uranus: Compact cloud layer and high-latitude depletion 17 condensation level, is now found much deeper in the at- mosphere and thus must be composed of some other sub- stance, perhaps part of a somewhat extended H2S cloud layer. 7. LATITUDE DEPENDENCE OF CLOUD STRUCTURE Our cloud pressure and optical depth parameters as a function of latitude are given in in Table 3 and key re- sults are plotted in Fig. 19. The results for latitudes from 30◦ S to 20◦ N were all obtained using our F1 profile of methane and temperature. For 60◦ S we used a methane profile depleted to 0.9% with a transition to 4% between 3 and 4 bars. At 45◦ S, we used a depletion to 0.6% with a transition to 4% between 1.5 and 2 bars. All three of these methane profiles are shown in Fig. 16A. Looking at the low-latitude region (30◦ S to 20◦ N), we see that the upper compact layer (the putative methane layer, also known as the UTC or layer 3) changes pressure only slightly, remaining consistent with methane condensation at the position of that layer, while its optical depth in- creases slightly towards the equator before beginning a significant decline into the northern hemisphere. Some- what more modulation is seen in the pressure of the dominant middle tropospheric cloud (MTC, layer 4) and there is a substantial declining trend in optical depth to- wards the north, though the rate of decline decreases near the equator. Though these declines may seem modest in the logarithmic plot, they are 7-10 times the typical 5% fitting uncertainty. The upper tropospheric haze reaches a peak opacity near the equator, as noted by KT2009. The prominent bright band near 45◦ S that is observed in Uranus images at wavelengths of intermediate absorp- tion (0.1-0.4 /km-am, as shown in Fig. 23 of KT2009) is also prominent in near-IR H- and J-band images, as in Fig. 7 of Sromovsky and Fry (2007). According to Fig. 19, this band is in a region where cloud pressures have increased in both of the upper compact layers while the upper tropospheric methane mixing ratio has decreased. It is clear that the upper compact layer at 45◦ S and 60◦ S cannot be the putative methane ice layer, because the pressure levels are much too high to permit conden- sation. It is conceivable that in this region the upper tropospheric cloud is actually what we called the mid- dle tropospheric cloud at low latitudes. And the much deeper second compact layer may be a new cloud layer formed by the upwelling produced by the lower circu- lation branch in Fig. 18. This would suggest that the boundary between the upper and lower branches may be in the 1.5-1.7 bar region. The bright band contrast relative to 60◦ S is a result of the deeper cloud being much deeper at 60◦ S and hav- ing a much lower optical depth. The contrast relative to lower latitudes results from a combination of reduced methane opacity as well as the extra cloud opacity in the 2-bar region, which apparently more than compensates for the absence of the 1.2-bar cloud layer and the signif- icantly reduced opacity at 1.5 bars. At 60◦ S the middle tropospheric cloud reaches about the same pressure as Baines et al. (1995) inferred for the semi-infinite cloud in their model, which is heavily weighted towards high latitudes. 8. DISCUSSION Fig. 16. -- Sample profiles of depleted methane used to fit high- latitude STIS spectra. The undepleted base profile labeled by its valid latitude range of 30◦ S to 20◦ N is from Model F1. The profiles in A provided the best fits for the indicated latitudes. None of the B profiles yielded any better fits (see text). at low latitudes, with upwelling and descending regions connected by meridional cells as illustrated in Fig. 18. The general nature of this circulation is similar to that described by KT2009 except that they don't provide for the existence of a methane condensation cloud, which seems to us a necessary mechanism to dry out the as- cending gas. The high latitude descending gas appears to be very dry and to descend only a small distance below the original methane condensation level. However, this descent should inhibit the formation of methane clouds at these latitudes, and if any do form they would have to form at much lower pressures. When we use five lay- ers, our layer 3, which used to be found at the methane 18 Sromovsky et al. Fig. 17. -- Overall fit quality (A) and fit quality at 0.825 µm (B) versus depleted mixing ratio for planetocentric latitudes 60◦ S (Left) and 45◦ S (Right). The transition pressures between upper depleted regions and the deep region at 4% are noted in the legend. Fit parameters at 10 latitudes, using Model F1 profiles for all but 45◦ S and 60◦ S. TABLE 3 Centric PUTC PMTC (dτ /dP ) (dτ /dP ) Latitude bars bars Strat.Haze UT. Haze τ UTC τ MTC τ LTC 20◦ N 1.185 10◦ N 1.175 4◦ N 1.180 1.565 1.595 1.558 0◦ 5◦ S 10◦ S 20◦ S 30◦ S 45◦ S 60◦ S 1.192 1.585 1.225 1.676 1.224 1.649 1.222 1.622 1.192 1.500 1.533 2.074 1.589 3.131 0.230 0.334 0.148 0.185 0.158 0.280 0.262 0.075 0.234 0.206 0.0000 0.0019 0.0363 0.0531 0.0375 0.0098 0.0043 0.0246 0.0181 0.0493 0.163 0.213 0.269 0.291 0.322 0.297 0.248 0.265 0.393 0.445 0.86 1.17 1.15 1.22 1.28 1.40 1.48 1.53 2.24 0.67 2.74 4.16 3.83 3.89 3.45 4.68 5.48 5.85 21.4 23.8 χ2 372 269 394 351 342 281 282 311 435 444 Note: Fixed parameters (except at 45◦ S and 60◦ S) included the lower cloud base pressure (5 bars), the particle radius of the upper tropospheric cloud (1.2 µm ). At 45◦ S and 60◦ S, we used rU T C = 1.1 and 1.34 µm , and PLT C = 8 and 11 bars, respectively. 8.1. Occultation failure to detect the middle tropospheric cloud Because the occultation probes to higher pressures than the pressures we infer for the dominant middle tro- pospheric cloud, one might wonder why the occultation did not detect this layer if it is indeed compact and cre- ated by condensation? That turns out not to be a con- tradiction. If the mixing ratio of the condensible is low enough, then the change in molecular weight might actu- ally be very small, and the resulting refractivity change might be too small to be detected. For example, for H2S to form a cloud base near 1.7 bars its volume mixing ratio would need to be ∼10−7 (∼0.003 times the solar mixing ratio). The maximum change in molecular weight in- duced by condensation of that much H2S would then be about 3×10−6, which would be completely undetectable in the occultation measurements, even if it occurred in a thin layer. But could such a cloud actually produce enough condensible material to produce the needed re- flectivity of the cloud? The answer is yes. If half the Methane on Uranus: Compact cloud layer and high-latitude depletion 19 S. Pole 60o S 45o S 30o S Equator 100 mb Methane condensation cloud layer descent of dry gas rising methane-moist gas 1.2 bars 3 bars ? mixing of moist and dry gas deep reservoir of methane Fig. 18. -- Schematic of methane depletion through condensation at low latitudes as a drying process, poleward transport, then de- scent of dry gas, finally mixing with moist gas on the return flow. How far beyond 60◦ S the depletion extends towards the pole, or whether northern high latitudes are depleted is unknown. Also unknown is how deep the return flow extends and whether or not there is a deep meridional flow in the opposite sense. Fig. 19. -- Cloud model parameters as a function of latitude, from Table 3. The Upper Tropospheric cloud from 30◦ S to 20◦ N is consistent with a methane cloud. There appears to be no methane cloud at 45◦ S and 60◦ S. The largest effect on the STIS spectra is produced by the Middle Tropospheric Cloud. The filled circle in A marks the top of the optically thick cloud of Baines et al. (1995), plotted at the sub-earth latitude of the disk-integrated observations they analyzed. H2S gas in a 2-km thick layer were to condense into 1- µm droplets, we estimate that the resulting cloud would have about 1 optical depth. Convective processing of additional gas could increase its opacity to much higher values. NH3 is a much less suitable candidate because its corresponding mixing ratio would be three orders of magnitude smaller as would the mass available for cloud particle formation. 8.2. Effects of para-hydrogen variations All our calculations described so far have assumed equi- librium hydrogen, for which the para and ortho fractions are in local thermodynamic equilibrium. Since these two forms of hydrogen have different spectral features, this assumption deserves some consideration. The equilib- rium para faction for our F1 profile is displayed in Fig. 20, along with several non-equilibrium curves, which we discuss in the following. In the deep atmosphere high temperatures produce a para fraction of 0.25 (referred to as normal hydrogen), but above the methane condensa- tion level the para fraction increases dramatically. By vertically mixing deep atmospheric gas up to the cloud level and higher, a dramatic change in the para fraction can be produced, if done on a time scale short compared to the relaxation time. Baines et al. (1995) considered mixtures of equilibrium and normal hydrogen, as might be produced by vertical mixing of the high temperature para distribution with the locally equilibrated distribu- tion. Their results were consistent with mixtures that were no less than 85% equilibrium (and thus 15% nor- mal). KT2009 found that the difference between 100% and 85% equilibrium had a smaller effect than their un- certainties. They also stated that 50% equilibrium hy- drogen would shift their results toward lower aerosol opacities and lower mixing ratios, but not change their results about latitudinal variations unless the the equi- librium fraction itself varied with latitude, which they argued against on the basis of smooth latitudinal varia- tions in the observed hydrogen absorption regions. Mixtures of equilibrium and normal hydrogen cannot produce the very different para fraction profiles that can be produced by the descending or rising branches of our putative methane circulation. For convenience, we pa- rameterize the rising or descending gas by Pfac, which is the factor by which pressure at the midpoint of the equilibrium profile is displaced. In Fig. 20A, we show para fractions that result when gas is vertically trans- ported upward (Pfac=0.5) or downward (Pfac=1.5 and Pfac=3.0), assuming that the transport time is much less than the relaxation time to reach local equilibrium. Note that mixtures of equilibrium and normal hydrogen always decrease the para fraction, while vertical gas transport can dramatically increase or decrease the local para frac- tion. How changes in the para profile affect the Uranus spectrum can be seen in Fig. 21 for two profiles of the type shown in Fig. 20A: a descending gas case (Pfac=2.5) and an ascending gas case (Pfac=0.5). The effect is lo- cal, mainly between 0.77 and 0.86 µm, and is roughly the same percentage over a wide range of view angles, reaching a maximum difference of about 15%. For the case shown in Fig. 20B, where there is descending gas above 1.5 bars, and ascending gas below (deeper than) 1.5 bars, which is more consistent with high-latitude motions shown in Fig. 18, the spectral difference from 20 Sromovsky et al. Fig. 20. -- A: Profiles of para hydrogen assuming local thermodynamic equilibrium (solid), mixtures of equilibrium and normal hydrogen (dotted), and vertically displaced hydrogen before new equilibrium conditions are established (dot-dash). B: Example of a mixed rising- descending profile (dot-dash) compared to the equilibrium profile. (1981) and Conrath and Gierasch (1984). However, the in situ measurements of the Galileo Probe con- tradicted this value, instead obtaining 0.157±0.004 (von Zahn et al. 1998) and 0.156±0.006 (Niemann et al. 1998). Conrath and Gautier (2000) reviewed possible ex- planations for this discrepancy and ruled out IRIS cal- ibration errors, but were unable to identify any plau- sible error in the occultation measurements or analy- sis, although a detailed error analysis of the occulta- tion measurements was not conducted. This discrep- ancy also cast suspicions on the very low He/H2 ratio obtained by IRIS-RSS analysis for Saturn, and led to an upward revision of the Saturn ratio from 0.034±0.024 to 0.11-0.16, based on a rather uncertain IRIS-only analysis (Conrath and Gautier 2000). This also raises questions about the occultation results for Uranus, suggesting per- haps that the IRIS-RSS determined ratio is actually too low for Uranus, rather than our suggestion that the ra- tio is too high. But an even higher ratio, with the same refractivity profile would lead to very serious amplifica- tion of inconsistencies with near-IR and CCD spectra of Uranus that have already been noted. Furthermore, because the specific error that caused prior discrepan- cies is not known, it is not possible to predict with any confidence how it might affect the Uranus analysis, or even what direction it might take. The change in re- fractivity per molecule that results from changing the He VMR from 0.15 to 0.11, as computed from Eq. 2 is from 4.496×10−18 /cm3 to 4.647×10−18 /cm3, which is a change of 3.3%. We do not know if an error of this mag- nitude is even remotely conceivable for the occultation measurements. 8.4. The reasons for higher cloud pressures obtained from Near-IR observations The optical depth of the methane cloud is relatively small and the influence of the middle tropospheric cloud is much more significant because of its greater opacity. Fig. 21. -- A: Model spectra computed for ascending (red) and descending (black) gas profiles at view angle cosine =0.6. B: Ratio of the two spectra at four different view angle cosines. the equilibrium case is only a few percent, and not de- tectable within current uncertainties. The fact that we did not find evidence in the spectra for significantly al- tered para profiles is consistent with either such a com- bined upwelling/down-welling profile or no significant de- parture from equilibrium hydrogen. However, this issue may be worth further investigation, especially if methane absorption coefficients become better known. Part of the problem making use of this constraint is that we are working with a part of the spectrum where methane ab- sorption is weak and thus currently is significantly un- certain. 8.3. Plausibility of a reduced He VMR on Uranus The Voyager Infrared Radiometer Interferometer (IRIS) and Radio Science Sub- and Spectrometer system (RSS) together yielded a determination of Jupiter's He/H2 ratio equal to 0.114±0.025, which is a weighted mean of results by Gautier et al. Methane on Uranus: Compact cloud layer and high-latitude depletion 21 diffuse layers of the KT2009 model are split into three compact layers, when constrained by STIS spectra at 5◦ S, yield best-fit pressures for the top compact layer in excellent agreement with the loca- tion of the occultation cloud layer for profile mod- els with deep CH4 mixing ratios between 3.2 and 4.5%, with the best compromise fit being obtained at 4% (for our Model F1), although this fit is not significantly better than for the other models in this range. 3. As judged by fitting errors at 0.825 µm, where H2 CIA exceeds methane absorption, the best compat- ibility between methane and H2 CIA is obtained for a methane mixing ratio of 4.5% with an uncertainty range of about 0.7%. 4. When we consider constraints of upper cloud pres- sure, fit error at 0.825 µm , overall fit quality, and the uncertainty limits of the Conrath et al. (1987) helium abundance, our best compromise es- timate for the deep methane mixing ratio at 5◦ S is 4.0±0.5%, and the corresponding preferred vertical temperature and methane profiles are embodied in our F1 Model. 5. Our five-layer cloud model, using the EF and/or F1 profiles, can also provide excellent fits between lat- itudes of 30◦ S and 20◦ N, with relatively latitude- independent cloud pressure boundaries, but gener- ally decreasing optical depths from south to north. The main exception is the upper tropospheric haze, which shows a strong peak near the equator, as noted by KT2009. The lower tropospheric cloud is found to have the greatest opacity, but the 1-2 op- tical depth (at 0.5 µm) middle tropospheric cloud (in 1.5-1.7 bar range) has the most significant effect on the observed CCD spectrum. The pressure of the 0.15-0.3 optical depth methane cloud layer did not vary much with latitude, staying within 2% of its 1.2 bar mean. 6. Trying to fit our model at high southern latitudes made it clear that the methane mixing ratio profile could not remain the same as we used in the low lat- itude observations. Not only did overall fit quality deteriorate, but the fit quality at 0.825 µm wave- length of H2 CIA got suddenly very bad, and the direction and size of the error at that wavelength indicated a significant lowering of the methane mix- ing ratio, as first pointed out by KT2009. 7. We created depleted methane profiles in which de- pletions reached a limited depth at which point the mixing ratio transitioned to the same deep value of 4% used at low latitudes. We did not expect a depletion at great depths because the horizontal density gradients that would be generated would induce problematic vertical wind shears. Overall and 0.825 µm fit quality was found to be mini- mized with mixing ratios less than 1% and deple- tion depths of 3-4 bars at 60◦ S and only 1.5-2 bars at 45 ◦ S. 8. Creating the observed methane depletion is plausi- bly the result of upwelling at low latitudes, where Fig. 22. -- Comparison of methane profiles from our EF and F1 models with those used by KT2009, L87, and Irwin et al. (2011), the latter following the L87 profile above the point of intersection of their assumed deep 1.6% VMR with that profile. That and the limited vertical resolution inherent in all the spectral observations, will lead inversion methods such as the NEMESIS routine used in by Irwin and col- leagues (Irwin et al. 2010, 2011) to find peak opacity con- tributions at pressures deeper than that of the methane cloud. In addition, these and most prior modelers of near-IR spectra assumed profiles of methane with much smaller column amounts of methane, often by a factor of two or more (Fig. 22), which leads to much higher pres- sures for the level of significant aerosol scattering. The dependence of derived pressure on assumed methane was clearly demonstrated by Sromovsky and Fry (2007) and by Irwin et al. (2010). As noted previously, when Irwin et al. switched from Model D to the Model F profile, the pressure of their peak opacity concentration moved from 2.5 to 1.7 bars. Since our Model F1 has more col- umn abundance of methane down to the 1.2 bar level than even the L87 Model F, we expect similar pressures would be obtained using our recommended profile. 9. SUMMARY AND CONCLUSIONS After reanalysis of the L87 radio occultation profiles of refractivity vs altitude, and fitting compact cloud layer models to STIS spectra as corrected and calibrated by KT2009, we reached the following main conclusions: 1. By decreasing the stratospheric He mixing ratio from its nominal value of 0.15 by 1-1.3 times its uncertainty it is possible to achieve methane satu- ration within the layers suspected to have conden- sation and to achieve increased methane humidi- ties above the condensation level, even exceeding the high values adopted by KT2009 at low lati- tudes, which are inconsistent with the original L87 profiles. The maximum deep mixing ratio that we could obtain within reasonable physical constraints was 4.88% (for our model G). 2. A five-layer cloud model in which the bottom two 22 Sromovsky et al. the Uranus atmosphere is dried out by condensa- tion of cloud particles, and subsequent poleward transport to and descent of the dried-out atmo- sphere at high latitudes down to levels of a few bars, with greater descent nearer to the pole. This would inhibit methane condensation at high lati- tudes and perhaps help to explain the failure to detect any discrete cloud features from 45◦ S to the south pole. 9. The failure of modelers of near-IR spectra to de- tect a methane cloud at the proper location (where the occultation found a sudden change is refractiv- ity) is due to two effects: one is that most mod- elers used mixing ratio profiles providing too lit- tle methane absorption and the other is that the methane cloud layer has a significantly lower opac- ity than the main cloud layer that peaks in the 1.4-1.7 bar region, making it hard to resolve its separate contribution. 10. The failure of the occultation measurements to de- tect the deeper and more significant middle tro- pospheric cloud layer that so prominently affects both near-IR and CCD spectra is not at all sur- prising if it is created by condensibles at very low mixing ratios. In that case the condensation would not produce a detectable change in refractivity. 11. Although it is not possible for the STIS spectra alone to clearly distinguish between vertically ex- tended models, such as that of KT2009, and com- pact layer models of the type we presented here, there is a strong distinction in physical processes involved. Our model is consistent with cloud for- mation by condensation, while the KT2009 model presents a picture in which a vertically extended haze of stratospheric origin provides all the particu- late opacity, much like the vertically extended haze on Titan. We prefer the condensation model be- cause it provides consistency with the occultation observations and provides a mechanism for deplet- ing methane from the upper troposphere at high latitudes. Many of these results are dependent on models of methane absorption in spectral regions where absorp- tion is relatively weak and somewhat uncertain. Some of the absorption coefficients were adjusted by KT2009 to obtain better overall fits to the observations, which might tend to favor the particular models they were us- ing in their analysis. As we obtain more accurate infor- mation about methane absorption that is independent of the spectral observations we are trying to interpret, these conclusions may need to be modified. The view of Uranus' north polar regions at the be- ginning of 2011 is now adequate to address an impor- tant issue with new observations. The issue is whether the depletion of methane observed at southern high lati- tudes is also present at northern high latitudes, and if not currently depleted, whether it will become depleted as Uranus seasons proceed. There is some evidence suggest- ing high northern latitudes might not be depleted: dis- crete cloud features have been observed in 2007 Keck im- ages between 45◦ N and 74◦ N (Sromovsky et al. 2009), while no discrete cloud features have been seen at cor- responding southern latitudes. The lack of southern dis- crete features is correlated with the inferred descent of dry gas at those latitudes, suggesting that the descending gas and resulting methane depletion may not be present at high northern latitudes. The question of whether methane is currently depleted at high northern latitudes on Uranus can be answered unequivocally by new STIS spectral measurements. ACKNOWLEDGMENTS This research was supported by NASA Outer Plan- ets Research Grant NNG05GG93G and Planetary Atmo- spheres Grant NNX09AB67G. We thank E. Karkoschka and M. Tomasko for making their calibrated STIS data cubes available to the community. We thank two anony- mous reviewers for constructive comments. REFERENCES Baines, K. H., Mickelson, M. E., Larson, L. E., Ferguson, D. W., 1995. The abundances of methane and ortho/para hydrogen on Uranus and Neptune: Implications of New Laboratory 4-0 H2 quadrupole line parameters. Icarus 114, 328 -- 340. Borysow, A., Borysow, J., Fu, Y., 2000. Semi-empirical model of collision-induced absorption spectra of H2-H2 complexes in the second overtone band of hydrogen at temperatures from 50 to 500 K. Icarus 145, 601 -- 608. Conrath, B., Hanel, R., Gautier, D., Marten, A., Lindal, G., 1987. The helium abundance of Uranus from Voyager measurements. J. Geophys. Res. 92 (11), 15003 -- 15010. Conrath, B. J., Gautier, D., 2000. Saturn Helium Abundance: A Reanalysis of Voyager Measurements. Icarus 144, 124 -- 134. Conrath, B. J., Gierasch, P. J., 1984. Global variation of the para hydrogen fraction in Jupiter's atmosphere and implications for dynamics on the outer planets. Icarus 57, 184 -- 204. Conrath, B. J., Gierasch, P. J., Leroy, S. S., 1990. Temperature and circulation in the stratosphere of the outer planets. Icarus 83, 255 -- 281. Costantino, M. S., Daniels, W. B., 1975. Dielectric constant of compressed solid methane at low temperature. J. Chem. Phys. 62, 764 -- 770. de Pater, I., Romani, P. N., Atreya, S. K., 1989. Uranus deep atmosphere revealed. Icarus 82, 288 -- 313. Fegley, B. J., Gautier, D., Owen, T., Prinn, R. G., 1991. Spectroscopy and chemistry of the atmosphere of Uranus. In: Bergstralh, J. T.. and Miner, E. D. and Matthews, M. S. (Ed.), Uranus. Univ. of Arizona Press, pp. 147 -- 203. Fink, U., Larson, H. P., 1979. The infrared spectra of Uranus, Neptune, and Titan from 0.8 to 2.5 microns. Astrophys. J. 233, 1021 -- 1040. Gautier, D., Conrath, B., Flasar, M., Hanel, R., Kunde, V., Chedin, A., Scott, N., 1981. The helium abundance of Jupiter from Voyager. J. Geophys. Res. 86, 8713 -- 8720. Hanel, R., Conrath, B., Flasar, F. M., Kunde, V., Maquire, W., Pearl, J., Pirraglia, J., Samuelson, R., Horn, L., Schulte, P., 1986. Infrared observations of the Uranian system. Science 233, 70 -- 74. Hansen, J. E., 1971. Circular polarization of sunlight reflected by clouds. Journal of Atmospheric Sciences 28, 1515 -- 1516. Irwin, P. G. J., Sromovsky, L. A., Strong, E. K., Sihra, K., Bowles, N., Calcutt, S. B., Remedios, J. J., 2006. Improved near-infrared methane band models and k-distribution parameters from 2000 to 9500 cm-1 and implications for interpretation of outer planet spectra. Icarus 181, 309 -- 319. Methane on Uranus: Compact cloud layer and high-latitude depletion 23 Irwin, P. G. J., Teanby, N. A., Davis, G. R., 2010. Revised vertical cloud structure of Uranus from UKIRT/UIST observations and changes seen during Uranus' Northern Spring Equinox from 2006 to 2008: Application of new methane absorption data and comparison with Neptune. Icarus 208, 913 -- 926. Irwin, P. G. J., Teanby, N. A., Davis, G. R., Fletcher, L. N., Orton, G. S., Tice, D., Kyffin, A., 2011. Uranus' cloud structure and seasonal variability from Gemini-North and UKIRT observations. Icarus 212, 339 -- 350. Karkoschka, E., 1998. Methane, Ammonia, and Temperature Measurements of the Jovian Planets and Titan from CCD-Spectrophotometry. Icarus 133, 134 -- 146. Karkoschka, E., 2001. Uranus' Apparent Seasonal Variability in 25 HST Filters. Icarus 151, 84 -- 92. Karkoschka, E., Tomasko, M., 2009. The haze and methane distributions on Uranus from HST-STIS spectroscopy. Icarus 202, 287 -- 309. Karkoschka, E., Tomasko, M. G., 2010. Methane absorption coefficients for the jovian planets from laboratory, Huygens, and HST data. Icarus 205, 674 -- 694. Karkoschka, E., Tomasko, M. G., 2011. The haze and methane distributions on Neptune from HST-STIS spectroscopy. Icarus 211, 780 -- 797. Lindal, G. F., Lyons, J. R., Sweetnam, D. N., Eshleman, V. R., Hinson, D. P., 1987. The atmosphere of Uranus - Results of radio occultation measurements with Voyager 2. J. Geophys. Res. 92 (11), 14987 -- 15001. Niemann, H. B., Atreya, S. K., Carignan, G. R., Donahue, T. M., Haberman, J. A., Harpold, D. N., Hartle, R. E., Hunten, D. M., Kasprzak, W. T., Mahaffy, P. R., Owen, T. C., Way, S. H., 1998. The composition of the Jovian atmosphere as determined by the Galileo probe mass spectrometer. J. Geophys. Res. 103, 22831 -- 22846. Orton, G. S., Baines, K. H., Bergstralh, J. T., Brown, R. H., Caldwell, J., Tokunaga, A. T., 1987. Infrared radiometry of Uranus and Neptune at 21 and 32 microns. Icarus 69, 230 -- 238. Orton, G. S., Griffin, M. J., Ade, P. A. R., Nolt, I. G., Radostitz, J. V., 1986. Submillimeter and millimeter observations of Uranus and Neptune. Icarus 67, 289 -- 304. Rages, K., Pollack, J. B., Tomasko, M. G., Doose, L. R., 1991. Properties of scatterers in the troposphere and lower stratosphere of Uranus based on Voyager imaging data. Icarus 89, 359 -- 376. Sromovsky, L. A., 2005a. Accurate and approximate calculations of Raman scattering in the atmosphere of Neptune. Icarus 173, 254 -- 283. Sromovsky, L. A., 2005b. Effects of Rayleigh-scattering polarization on reflected intensity: a fast and accurate approximation method for atmospheres with aerosols. Icarus 173, 284 -- 294. Sromovsky, L. A., Fry, P. M., 2007. Spatially resolved cloud structure on Uranus: Implications of near-IR adaptive optics imaging. Icarus 192, 527 -- 557. Sromovsky, L. A., Fry, P. M., 2010. The source of 3-µm absorption in Jupiter's clouds: Reanalysis of ISO observations using new NH3 absorption models. Icarus 210, 211 -- 229. Sromovsky, L. A., Fry, P. M., Hammel, H. B., Ahue, W. M., de Pater, I., Rages, K. A., Showalter, M. R., van Dam, M. A., 2009. Uranus at equinox: Cloud morphology and dynamics. Icarus 203, 265 -- 286. Sromovsky, L. A., Irwin, P. G. J., Fry, P. M., 2006. Near-IR methane absorption in outer planet atmospheres: Improved models of temperature dependence and implications for Uranus cloud structure. Icarus 182, 577 -- 593. Sun, Z., Schubert, G., Stoker, C. R., 1991. Thermal and humidity winds in outer planet atmospheres. Icarus 91, 154 -- 160. Tomasko, M. G., B´ezard, B., Doose, L., Engel, S., Karkoschka, E., 2008. Measurements of methane absorption by the descent imager/spectral radiometer (DISR) during its descent through Titan's atmosphere. Planetary and Space Science 56, 624 -- 647. von Zahn, U., Hunten, D. M., Lehmacher, G., 1998. Helium in Jupiter's atmosphere: Results from the Galileo probe helium interferometer experiment. J. Geophys. Res. 103, 22815 -- 22830.
1710.01019	1	1710	2017-10-03T08:00:15	Centrifugal Experiments with Simulated Regolith:Effects of Gravity, Size Distribution, and Particle Shape on Porosity	[ "astro-ph.EP" ]	Porosity is a key characteristic of the regolith on the surface of small bodies. The porosity of the regolith on the surface of asteroids is changed by applied pressure, and the relationship between pressure and porosity depends on the particle properties of the regolith. We performed compression measurements on samples of different materials, particle size distributions, and shapes to examine the relationship between particle size and pressure required for compression. We used a centrifuge and a compression testing machine for the experiments. The applied pressure for the centrifuge and the compression testing machine experiments ranged from 10^2 to 5x10^3 Pa and from 10^4 to 5x10^6 Pa, respectively. The initial porosity before compression was generally higher for samples with smaller particles and narrower particle size distributions. A sample compressed more easily when it consisted of smaller particles, probably due to smaller frictional forces between particles. We estimated the porosity of granular asteroids based on the results of our experiments. The estimated porosity at the center of a homogeneous asteroid with a radius of 50 km is 0.43. This porosity is typical of the porosity of asteroids of similar size.	astro-ph.EP	astro-ph	Centrifugal Experiments with Simulated Regolith: Effects of Gravity, Size Distribution, and Particle Shape on Porosity By Tomomi OMURA1), Masato KIUCHI1), Carsten GÜTTLER2), Akiko M. NAKAMURA1) 1)Kobe University, Kobe, Japan 2)Max-Planck-Institute for Solar System Research, Göttingen, Germany (Received March 30th, 2016) Porosity is a key characteristic of the regolith on the surface of small bodies. The porosity of the regolith on the surface of asteroids is changed by applied pressure, and the relationship between pressure and porosity depends on the particle properties of the regolith. We performed compression measurements on samples of different materials, particle size distributions, and shapes to examine the relationship between particle size and pressure required for compression. We used a centrifuge and a compression testing machine for the experiments. The applied pressure for the centrifuge and the compression testing machine experiments ranged from 102 to 5×103 Pa and from 104 to 5×106 Pa, respectively. The initial porosity before compression was generally higher for samples with smaller particles and narrower particle size distributions. A sample compressed more easily when it consisted of smaller particles, probably due to smaller frictional forces between particles. We estimated the porosity of granular asteroids based on the results of our experiments. The estimated porosity at the center of a homogeneous asteroid with a radius of 50 km is 0.43. This porosity is typical of the porosity of asteroids of similar size. Key Words: Asteroids, Granular material, Porosity, Regolith Nomenclature p ρp r g A s p0 m n Ω  a0 G  : porosity : particle density : particle radius : gravity : Hamaker constant : cleanliness ratio : constant : constant : constant : diameter of O-2 ion : reduced shear modulus : radius of contact area : gravitational constant : bulk density of asteroid 1. Introduction Both recent and expected near-future spacecraft exploration of asteroids has included landing on and sampling from the surface. The surfaces of asteroids are known to be covered by regolith, and the porosity of this regolith its physical properties, such as strength of the surface against meteoroid impacts or spacecraft landings.1) The porosity of a granular bed consisting of particles of fine dust has been shown to decrease with the pressure applied to the important characteristic affecting is an in laboratory it experiments.2,3) A in protoplanetary disks. Therefore, similar sample phenomenon would occur in regolith on the surface of an asteroid. Models of the relationship between pressure and porosity of very fluffy aggregates of dust in protoplanetary disks have been developed.4) However, these models may not be directly applicable to regolith layers because the porosity of the aggregate in the models is larger than 0.9, too large to be applied to regolith. The porosities of regolith layers are generally lower than those of dust aggregates is expected that the physical process of compaction is different. Porosity is related to the coordination number of the constituent particles in particle layers, which increases as porosity decreases; thus, the force required to deform particle layers changes with coordination number.5) When particle layers have high porosity and low coordination numbers, particles are rearranged by rolling because the energy required to roll particles is smaller than the energy required to slide them. However, when porosity is low and the coordination number exceeds 6, particles are rearranged by sliding because they are locked in triaxial directions. The porosity of a particle layer at this transition has been shown to be ~0.7.5) The bulk porosity of regolith on small bodies has been estimated to be about 0.4-0.9.6,7) Therefore, the sliding force between particles can be expected to be relevant in the rearrangement of regolith particles and the compression of the regolith layer. In this study, we investigate the relationship between pressure and the porosity of simulated regolith using different materials and different material size distributions, and then estimate the porosity of granular asteroids based on our experimental results. 2. Experiments 2.1. Samples The powder samples used in this study are listed in Table 1. We used silica sand 1-3 of irregular shape and with different size distributions, fly ash of spherical shape, fused alumina 1-3 of irregular shape and with different size distributions, and basalt fragments smaller than 210 μm prepared by an impact experiment (hereafter called "basalt"). The size distributions of the samples, shown in Fig. 1, were determined by laser diffractometry. We sieved the samples into a cylindrical container with a diameter of 5.8 cm and depth of 3.3 cm (for basalt, diameter of 2.7 cm and depth of 1.4 cm) and then leveled off the part of the bed exceeding the height of the container with a spatula. All experimental manipulations were conducted under atmospheric pressure. The porosity Table 1. Sample description. Sample Density Median diameter Shape (g/cm3) (µm) Silica sand 1 Silica sand 2 Silica sand 3 Fly ash Fused alumina 1 Fused alumina 2 Fused alumina 3 Basalt 2.5 2.5 2.5 2.0 4.0 4.0 4.0 2.7 13 19 73 4.8 4.5 23 77 53 irregular irregular irregular spherical irregular irregular irregular irregular of the samples was determined by the mass weighed and volume measured after compression. The volumes of the samples were determined from their heights. 2.2. Experimental methods Compression of the samples was performed using a centrifuge and a compression testing machine at Kobe University. Figures 1 and 2 show schematic views of the experimental setups. Compression by centrifuge simulates self-gravitational compression. However, we could not achieve high pressure with the centrifuge used in this study. Therefore, we used a compression testing machine for the high-pressure range. 2.2.1. Centrifuge The centrifuge consisted of two ends. An experimental arrangement was mounted on one end and a personal computer (PC) for the recording of data was mounted on the opposite end. A counter balance was mounted on the PC side. Images of the surface of each sample were recorded during each experiment by a camera installed on the experimental apparatus. The images were saved to the hard disc of the PC and checked after the power supply of the centrifuge had been turned off. The simulated gravity for the samples resulted from the centrifugal force and the gravity of Earth. We could change the simulated gravity within 1-17 times Earth gravity. The rotation velocity was changed manually changing the voltage applied to the motor using a dial. The rotation velocity was determined precisely by a sensor consisting of a bar code mounted near the axis of rotation and a bar code sensor. The sensor detected the intensity of light reflected from the bar code, and returned a corresponding voltage. We could measure the time necessary for the sensor to pass over the bar code. In this study, the centrifugal acceleration at the middle depth of the sample container was calculated and then converted to overburden pressure. Fig. 1. Samples particle size distribution. Fig. 2. Centrifuge. 2.2.2. Compression testing machine The sample container and a compression piston were placed between the pressure plates of a compression testing machine. We used cylindrical containers with inner 0204060801000.11101001000Silica sand 1Silica sand 2Silica sand 3Fly ashFused alumina 1Fused alumina 2Fused alumina 3BasaltCumulative volume distribution, %Diameter, μm diameter of 58 mm and height of 33 mm. The diameter of the piston was slightly smaller than the inner diameter of the container. Before beginning the compression, the bottom of the piston was lowered into the top of the container. That is, the piston was lowered into the container to ensure alignment between the piston and the wall of the container. For silica sand 3 and fused alumina 2 and 3, the starting point of the piston was 1 mm below the top of the container; for the other samples, it was 2 mm lower. The loading velocity was 0.6 mm/min. During the measurements, downward displacement and compressive force were recorded by the PC at 1 s intervals. The pressure acting on the sample was calculated by dividing the compressive force by the cross-sectional area of the piston. The volume of the sample was estimated from the height of the sample, which was calculated by the displacement of the piston. We did not conduct this measurement for basalt powders because of the small volume of the sample. Fig. 3. Setups of the compression testing machine. 3. Results and Discussion 3.1. Porosity before compression Figure 4 shows the relationship between the median diameter of sample constituents and the porosity before compression. The error bars represent the standard deviation of the porosity measurements. Samples with the same composition tended to have higher porosity when the grain size was smaller. The relationship between particle size and porosity is expressed by the following formula:7) . (1) The results of Eq. (1) are also shown in Fig. 4. The blue solid, red solid, green dashed, and brown dashed curves are derived from Eq. (1) based on the density for each sample and the Hamaker constants for SiO2 (silica sand Fig. 4. Porosity before compression. this sand had and fly ash), Al2O3, and basalt, respectively. The porosities of the silica sand particles were roughly consistent with those expected from Eq. (1), especially the porosities of silica sands 1 and 3. The porosity of silica sand 2 was slightly lower than that expected from Eq. (1), presumably because the widest size distribution among the three silica sands. The porosities of the fused alumina were in agreement with or slightly higher than values estimated from Eq. (1). The porosities of the fly ash and the basalt fragments were slightly lower than expected based on Eq. (1), presumably because void space was filled by small particles. 3.2 Porosity after compression Figure 5 shows the results of the centrifuge and compression testing machine experiments. The results of the centrifuge experiments are shown by symbols, and the results of the compression testing machine experiments are shown by curves. The errors in Fig. 5, defined by errors in sample volume, are smaller than the symbols or the thickness of the curves. Although the pressure applied by the centrifuge was volume pressure, which is somewhat Fig.5. Results of compression experiments. npgrAsmppp2220064exp10.20.30.40.50.60.70.8102103104105106107Silica sand 1Silica sand 2Silica sand 3Fly ashFused alumina 1Fused alumina 2Fused alumina 3PorosityPressure, Pa0.50.60.70.80.9101020304050607080Silica sandFly ashFused aluminaBasaltEq. 1 for silica sandEq. 1 for fly ashEq. 1 for fused aluminaEq. 1 for basaltPorosityDiameter, μm different from piston compression, the results of the centrifuge and compressive testing machine experiments were roughly consistent, i.e., they showed on similar curves. This suggests that the overburden pressure, or gravitational pressure, applied by the centrifuge had a similar effect on the samples to that applied by the piston of the compression testing machine. The porosities of different samples differed even when the pressure applied to the samples was the same. We fit logarithmic functions to the data of Fig. 5. Note that the slope of the curve in Fig. 5 becomes steeper for those samples that were more easily compressed. Figure 6 shows the relationship between the slope of the curves in Fig. 5 and the frictional force between the particles. Errors defined by fitting deviation are smaller than the symbols in Fig. 6. Frictional force, the force required to slide particles over one another, is given by the following formula: 8) . (2) We assumed spherical particles and calculated the contact radius using the median diameter of the sample particles. The figure shows a systematic trend between the frictional force and the slope of the curves. Sample compression become more difficult as frictional force increased. Fly ash seemed to be compressed more easily than the other samples due to the smaller frictional force calculated from Eq. (2), and possibly also due to its spherical shape. Fig. 6. Relationship between the slope of the curve in Fig. 5 and the frictional force. 3.3. Compression by self-gravity The interiors of granular asteroids are subject to pressure from the soil due to self-gravity. We calculated the pressure at the center of an asteroid of radius R and determined the porosity using our experimental results. We assumed that the asteroid consists of regolith-sized particles. The pressure at the center of a homogeneous asteroid is given by the following formula: . (3) The porosity of a granular asteroid with a radius of 50 km and consisting of particles similar to the silica sand 1 particles would be 0.43. This porosity is typical of the observed porosity of similar-sized asteroids. However, based on thermal observations, the grain size of the regolith on the surface of Lutetia (diameter of 95.8 km) is estimated to be 210 µm,6) which is about 20 times larger than the median diameter of silica sand 1. We note that our experiment was performed the compressibility may be different in vacuum. We also note that this application is valid for only small bodies. Large bodies would have a density gradient due to self-gravity throughout the interior of the body. in air and that 4. Summary We performed compression experiments with granular samples of different materials, grain shapes, and size distributions. The porosity of the samples before compression tended to be higher when the median grain diameter was smaller. Our results were basically consistent with an empirical formula presented in the previous study. However, the particle samples with a narrow size distribution had a slightly higher porosity than the value expected from the empirical formula. The overburden pressure applied by centrifugal acceleration (one of our methods) or gravitational pressure had similar effects to the pressure exerted by compression with a piston (our second method). Different samples had different porosities, even if the same pressure was applied. Furthermore, the force required for compression of the sample increased as the particle diameter increased, possibly due to increased friction force between particles. The surfaces of large asteroids are covered with smaller particles, which may therefore be compressed more easily. We estimated the porosity of granular asteroids based on our experimental results, and the estimated porosity was consistent with the observed porosity of similar-sized asteroids. However, we note that our experiment was performed in air. Compression of regolith particles may be more difficult in vacuum. References 1) Schräpler, R., Blum, J., Borstel, I., and Güttler, C.: The Stratification of Regolith on Celestial Objects, Icarus 257 (2015), pp. 33-46. 2) Güttler, C., Krause, M., Geretshauser, R. J., Speith, R., Blum, J: The Physics of Protoplanetesimal Dust Agglomerates. IV. Toward a Dynamical Collision Model, Astrophys. J. 701 (2009), pp. 130-141. 3) Machii, N., Nakamura, A. M., Güttler, C., Beger, D., and Blum, J.: Collision of a Chondrule with Matrix: Relation between Static Strength of Matrix and Impact Pressure, Icarus 226 (2013), pp. 111-118. 220aFfric2232)0(GRP-0.08-0.07-0.06-0.05-0.04-0.03-0.02-0.0110-610-510-410-3Silica sand 1-3Fused alumina 1-3Fly ashThe slope of the curves in Fig. 5Ffric 4) Kataoka, A., Tanaka, H., Okuzumi, S., and Wada, K.: Static Compression of Porous Dust Aggregates, Astronomy & Astrophysics, 554 (2013), A4. 5) Wada, K., Tanaka, H., Suyama, T., Kimura, H. and Yamamoto, T.: The Rebound Condition of Dust Aggregates Revealed by Numerical Simulation of Their Collisions, Astrophys. J. 737 (2011) pp. 36-47. 6) Gundlach, B. and Blum, J.: A New Method to Determine the Grain size of Planetary Regolith, Icarus 223(2013), pp. 479-492. 7) Kiuchi, M. and Nakamura, A. M.: Relationship between Regolith Particle Size and Porosity on Small Bodies, Icarus 239 (2014), pp. 291-293. 8) Dominik and Tielens: The Physics of Dust Coagulation and The Structure of Dust Aggregates in Space, Astrophys. J. 480 (1997), pp. 647-673.
1811.09314	1	1811	2018-11-22T21:54:42	Photoevaporative Flows From Exoplanet Atmospheres: A 3-D Radiative Hydrodynamic Parameter Study	[ "astro-ph.EP" ]	The photoionization-driven evaporation of planetary atmospheres has emerged as a potentially fundamental process for planets on short period orbits. While 1-D studies have proven the effectiveness of stellar fluxes at altering the atmospheric mass and composition for sub-Jupiter mass planets, there remains much that is uncertain with regard to the larger-scale, multidimensional nature of such "planetary wind" flows. In this paper we use a new radiation-hydrodynamic platform to simulate atmospheric evaporative flows. Using the AstroBEAR AMR multiphysics code in a co-rotating frame centered on the planet, we model the transfer of ionizing photons into the atmosphere, the subsequent launch of the wind and the wind's large scale evolution subject to tidal and non-inertial forces. We run simulations for planets of 0.263 and 0.07 Jupiter masses and stellar fluxes of $2 \times 10^{13}$ and $2 \times 10^{14}$ photons/cm^2/s. Our results reveal new, potentially observable planetary wind flow patterns, including the development, in some cases, of an extended neutral tail lagging behind the planet in its orbit.	astro-ph.EP	astro-ph	MNRAS 000, 1 -- 16 (0000) Preprint 26 November 2018 Compiled using MNRAS LATEX style file v3.0 Photoevaporative Flows From Exoplanet Atmospheres: A 3-D Radiative Hydrodynamic Parameter Study Alex Debrecht1(cid:63), Jonathan Carroll-Nellenback1, Adam Frank1, John McCann2, Ruth Murray-Clay3, Eric G. Blackman1 1Department of Physics and Astronomy, University of Rochester, Rochester NY 14627 2Department of Physics, University of California, Santa Barbara, Santa Barbara, CA, 93106 3Physics and Astronomy Department, University of California, Santa Cruz, Santa Cruz, CA 95064 ABSTRACT The photoionization-driven evaporation of planetary atmospheres has emerged as a potentially fundamental process for planets on short period orbits. While 1-D studies have proven the effectiveness of stellar fluxes at altering the atmospheric mass and composition for sub-Jupiter mass planets, there remains much that is uncertain with regard to the larger-scale, multidimensional nature of such "planetary wind" flows. In this paper we use a new radiation-hydrodynamic platform to simulate atmospheric evaporative flows. Using the AstroBEAR AMR multiphysics code in a co-rotating frame centered on the planet, we model the transfer of ionizing photons into the atmosphere, the subsequent launch of the wind and the wind's large scale evolution subject to tidal and non-inertial forces. We run simulations for planets of 0.263 and 0.07 Jupiter masses and stellar fluxes of 2 × 1013 and 2 × 1014 photons/cm2/s. Our results reveal new, potentially observable planetary wind flow patterns, including the development, in some cases, of an extended neutral tail lagging behind the planet in its orbit. Key words: hydrodynamics -- planet-star interactions -- planets and satellites: at- mospheres 1 INTRODUCTION The characterization of planetary atmospheres comprises a set of challenges at the forefront of exoplanet science. While a great deal can be learned from studying planetary at- mospheres themselves, interactions between the atmosphere and the near-space environment of exoplanets is also a rich subject, promising a strong interplay between observational diagnostics and theoretical interpretation. In particular, at- mospheric blow-off, also known as hydrodynamic escape or evaporation, may hold the key to understanding important facets of planetary evolution, from end-state masses to final atmospheric compositions (and therefore habitability). Importantly, for some exoplanets (e.g. HD 209458b, Charbonneau et al. (2002); Vidal-Madjar et al. (2003, 2004, 2013); Ballester et al. (2007); Ben-Jaffel (2007, 2008); Ben- Jaffel & Sona Hosseini (2010)), the observational signatures of atmospheric blow-off have become well-determined. For the hot Jupiters that have been observed to have signatures of this blow-off, from 5% to 50% absorption has been seen (cid:63) [email protected] c(cid:13) 0000 The Authors both post- and pre-transit, with some systems showing sym- metric absorption (e.g. HD 209458b) and some highly asym- metric (e.g. GJ 436b, Kulow et al. (2014); Ehrenreich et al. (2015)). Interestingly, this absorption extends well into the Doppler-shifted wings of the Lyman-α line, with absorption out to Doppler shifts of ±150 km/s. Hydrodynamic planetary winds occur when irradiation from the central star, especially in the extreme ultraviolet (EUV), heats the upper layers of the atmosphere to produce an extended envelope of gas which transitions into a wind (Garc´ıa Munoz 2007; Murray-Clay et al. 2009). A charac- teristic measure of the strength of the wind is the ratio of gravitational potential to thermal energy at the top of the atmosphere. This is usually called the hydrodynamic escape parameter λ = GMpµ , where MP and Rp are the mass and RpkTp radius of the planet, and Tp and µ are the temperature and mean mass per particle in the atmosphere. For λ >> 10, the atmosphere is too tightly bound for a hydrodynamic wind to form. Note that weaker outflows may be produced via nonthermal processes, e.g. Hunten (1982). For λ ∼ 10, a Parker-type thermally driven hydrodynamic wind is ex- A. Debrecht et al. 2 pected (note that λ ≈ 15 for the sun with its T ∼ 106 K corona). Physical processes that can affect the wind structure and outflow rates, such as planetary magnetic fields (Owen & Adams 2014), time-dependent EUV flux (Lecavelier des Etangs et al. 2012), atmospheric circulation (Teyssandier et al. 2015), and the interaction between stellar and plane- tary winds (Stone & Proga 2009; Carroll-Nellenback et al. 2017; McCann et al. in press) have been incorporated into existing simulations. Fully 3-D simulations are expensive, but a growing number of groups are carrying them out (Schneiter et al. 2007; Cohen et al. 2011; Bisikalo et al. 2013; Tripathi et al. 2015; Matsakos et al. 2015; Carroll- Nellenback et al. 2017; Schneiter et al. 2016). The work of Matsakos et al. (2015) was noteworthy for including a vari- ety of processes in a 3-D MHD study that tracked the full orbital dynamics of the wind using a simplified atmospheric model. Resolving spatial and temporal structures (asymme- try, shocks, cometary tails via radiation pressure or stel- lar wind interactions) in exoplanet wind simulations is of particular importance for interpreting observations. Carroll- Nellenback et al. (2017) simulated the full 3-D global dy- namics of a planetary wind interacting with a stellar wind. Using the AMR code AstroBEAR Carroll-Nellenback et al. (2017) found a "2-armed" up-orbit and down-orbit pattern similar to that of Matsakos et al. (2015) and traced the origin of the flow to a combination of tidal and Coriolis forces. Carroll-Nellenback et al. (2017) also mapped out the observational consequences of these global flows, showing that the torus of planetary wind material was potentially observable far from the time of a transit. This conclusion was particularly relevant for WASP-12b, whose central star shows none of the expected MgII h&k lines (Fossati et al. 2010). Their absence has been posited to be due to the pres- ence of a relatively dense torus (Haswell et al. 2012). In Debrecht et al. (2018), a global planetary evaporation simu- lation of WASP-12b was carried out using AstroBEAR and a torus of the needed density was shown to be relatively easy to form. While these global simulation approaches show promise, for both Carroll-Nellenback et al. (2017) and Mat- sakos et al. (2015) the planetary winds were not calculated self-consistently but were imposed by boundary conditions at the planet's surface. Fully 3-D simulations that include the radiative transfer of stellar photons into the atmosphere, with its subsequent launch of the wind, are still relatively rare. In Tripathi et al. (2015) a self-consistent wind was driven by following the stellar flux as it was absorbed by the planet. These static mesh refinement simulations created winds whose properties were in good agreement with 1-D analytic models (Murray- Clay et al. 2009), as well as showing features such as back- flow on to the night side of the planet which had also been seen in models by Frank et al. (2016). Not included in Tri- pathi et al. (2015) were the effects of orbital motion. As shown in Matsakos et al. (2015) and Carroll-Nellenback et al. (2017), the Coriolis effect strongly distorts the streamlines of the wind after it launches, forming the up- and down-orbit arms. Our next step in understanding planetary photoevapo- rative winds is to model their launch via self-consistent 3-D radiative hydrodynamic calculations, while also increasing the global scale of simulations to capture the orbital motion of the planetary wind material. In this paper we present simulations of winds launched from the atmosphere of giant planets in both long and short period orbits, with variable planetary mass and subjected to variable incident EUV flux. We follow the radiative hydrodynamic behavior out to many planetary radii and run the simulations long enough to char- acterize the nature of the steady state flows. We note that a similar computational approach was taken in McCann et al. (in press). While that paper focused on the interaction of planetary with stellar winds, many of our conclusions about planetary outflows are supported by both papers. The plan of the paper is as follows: in section 2 we describe the numer- ical methods and simulations used; in section 3 we present the results of these simulations; in section 4 we analyze the results for important consequences; and in section 5 we dis- cuss the relationship of these simulations to previous studies and propose future avenues of inquiry. 2 METHODS AND MODEL Our simulations were conducted using AstroBEAR1 (Cun- ningham et al. 2009; Carroll-Nellenback et al. 2013), a mas- sively parallelized adaptive mesh refinement (AMR) code that includes a variety of multiphysics solvers, such as self-gravity, heat conduction, magnetic resistivity, radiative transport, and ionization dynamics. The equations solved for these simulations are those of fluid dynamics in a rotat- ing reference frame, with gravitational effects of both the planet and star included: + ∇ · ρv = 0, ∂ρ ∂t + ∇ · (ρv ⊗ v) = −∇p − ρ∇φ + fR, ∂ρv ∂t + ∇ · ((E + p)v) = G − L, ∂E ∂t (1) (2) (3) where ρ is the mass density, v is the fluid velocity, p is the thermal pressure, φ is the gravitational potential, fR combines the the Coriolis and centrifugal forces, so that fR = ρ (−2Ω × v − Ω × (Ω × r)) (where Ω is the orbital velocity), E = p/(γ − 1) + ρv2/2 is the combined internal and kinetic energies, and G and L are the heating and cool- ing rates. The simulation also tracked the advection, photoioniza- tion, and recombination of neutral and ionized hydrogen. We use the photon-conserving update scheme from Krumholz et al. (2007) to solve the following equations: ∂nH ∂t + ∇ · (nHv) = R − I, ∂nHII ∂t + ∇ · (nHII v) = I − R, (4) (5) where nH is the number density of neutral hydrogen, nHII is the number density of ionized hydrogen, and R and I are the recombination and ionization rates. 1 https://astrobear.pas.rochester.edu/ MNRAS 000, 1 -- 16 (0000) 2.1 Radiation transfer In these simulations we model the incident stellar radiation as a planar radiation front. This is an acceptable approxi- mation, as the spherical dilution of the radiation field is only 3% over the simulation domain. In addition, for computa- tional simplicity we use photons of a single frequency of 16 eV, representative of the integrated EUV flux of the quiet sun (Woods et al. 1998). Following Tripathi et al. (2015), and as in McCann et al. (in press), we therefore apply a constant, uniform, monochromatic flux F0 to the leftmost (stellar) side of the grid. The flux at a distance x from the edge of the simulation is then F (x) = F0e −τ (x), (6) where τ (x) is the optical depth, given by (cid:90) x 0 τ (x) = nHσphdl. (7) Here σph = 6.3 × 10−18cm2 is the cross-section for pho- toionization at the ionization threshold energy, 13.6eV. The photoionization rate can then be calculated: I(x) = σphnHF (x). (8) As discussed in Krumholz et al. (2007), this method con- serves photon number. The photon flux from each previous cell is propagated, the number of absorbed photons is calcu- lated from the optical depth of the cell, and that number is subtracted from the flux, which is then propagated forward again, so no photons are lost between cells. We assume case-B recombination (ignoring free to ground state transitions), giving a recombination rate R of R = αB(T )nenHII , (9) where the case-B recombination coefficient is αB(T ) = 2.59 × 10−13(T /104K)−0.7 cm3s−1 and ne is the number density of electrons. We also assume electrons are advected . It should be with the ionized hydrogen, so that ne = nHII noted that case-B recombination is a poor approximation throughout most of the simulation, as the winds are opti- cally thin. However, case B is appropriate at the base of the wind, where we are most concerned with determining accu- rate conditions in order to obtain the correct wind solution. In addition, the difference in ionization state resulting from assuming case B has few practical consequences, as we don't perform synthetic observations in this study. Each photon deposits energy above the ionization threshold as heat, so the photoionization heating rate is given by G = eγI, (10) with eγ = 2.4 eV the average EUV photon energy in excess of the ionization energy. We also include the cooling from both recombination and Lyman-α emission due to collisional excitations: Lrec = 6.11 × 10 −10cm3s −1kBT nenHII , (11) (cid:19)−0.89 (cid:18) T K LLyα = 7.5 × 10 −19erg cm3s −1e −118348 K/T nenHII . (12) In order to test our radiation transfer implementation, MNRAS 000, 1 -- 16 (0000) Photoevaporation of Exoplanet Atmospheres 3 we initialized a one-dimensional simulation at the theoretical equilibrium ionization state for a given flux and hydrogen mass density. The simulation was allowed to evolve for many crossing times, and after a small initial relaxation period was found to be extremely stable. 2.2 Description of simulation to [57, 10, 10]Rp, with the planet The Cartesian simulation domain ranges The input parameters of the simulation were chosen to test the response of the planetary wind to two key parameters: planetary mass and stellar flux. Our planet, which is highly- inflated for computational reasons, has a radius of 2.146RJ and a mass of either 0.07MJ or 0.263MJ (see Table 1). The planet orbits a star with a mass of 1.35M(cid:12) at a separation of a = 0.047AU. Table 1 lists the parameters used for each simulation. from [37,−10,−10]Rp lo- cated in the center at [47, 0, 0]. We apply outflow-only extrapolating boundary conditions at all boundaries, with the initial ambient conditions applied if the extrapolated conditions would result in inflow. The simulation has a base resolution of 803 and 3 levels of additional refinement, giving an effective resolution of 6403. The maximum resolution is forced in a sphere out to the termination radius of the hydrostatic atmosphere (defined further in Section 2.3). We allow the mesh to evolve outside of the planet based on the density gradient. The planetary radius is therefore resolved by 32 cells. The stellar location [0, 0, 0] is not included in these simulations; only the gravitational effects of the star are simulated. Runs 1 and 2 were each performed in both the co- rotating and non-rotating frames of reference. The planet temperatures were adjusted based on the planet mass in or- der to maintain reasonable scale heights near the surface of the planet. It was shown in Murray-Clay et al. (2009) (Ap- pendix A) that the planet's surface temperature had an in- significant effect on the wind structure for any temperature below the wind temperature at the base, generally around 104K. The surface densities of the planet were adjusted so that the surface with optical depth one in the unperturbed planet was near the nominal planet surface, Rp. This was initially calculated for the low-flux case, then scaled by(cid:112)F0,2/F0,1 for the high-flux case, based on the assumption that we are in the recombination-limited regime of Murray-Clay et al. (2009). Although this is shown to be false in section 4.1, it provides an acceptable estimate for the required change in density. Our simulations were run for 5.24 days (1.49 orbits), after which the outflows have all reached a steady state, defined as a stable ionization front and wind morphology. 2.3 Planet atmosphere model In these simulations, we have modeled the planet as a sphere of hydrogen held in hydrostatic equilibrium so that it satis- fies dP dr = − GMpρ r2 , (13) 4 A. Debrecht et al. Table 1. Run parameters Planet Radius Stellar Mass Orbital Separation Orbital Period Orbital Velocity Polytropic Index Planet Mass Planet Temperature Stellar Ionizing Flux Planet Surface Density Rp M(cid:63) a P Ω γ Mp Tp F0 ρp Run 1 Run 2 Run 3 Run 4 2.146RJ 1.35M(cid:12) 0.047 AU 3.525 days 1.78 rad/day 5 3 0.263MJ 3 × 103 K 0.07MJ 1 × 103 K 2 × 1013 phot cm−2s−1 1.324 × 10−16g cm−3 2 × 1014 phot cm−2s−1 5.05 × 10−16g cm−3 0.263MJ 3 × 103 K 2 × 1014 phot cm−2s−1 5.05 × 10−16g cm−3 0.07MJ 1 × 103 K 2 × 1013 phot cm−2s−1 1.324 × 10−16 g cm−3 giving a density profile of (cid:20) R0Rp R0 − Rp (cid:19)(cid:21) 1 γ−1 (cid:18) 1 r − 1 R0 . (14) ρatm(r) = (15) (cid:113) kB Tp In the equations above R0 = Rp 1 − (γc2 s,pRp)/([γ − 1]GMp) is the radius at which the atmospheric profile goes to zero (physically, the atmosphere ends), with cs,p = the speed of sound at Rp. Since we have taken an adiabatic index of γ = 5 3 , the hydrostatic pressure profile is given by mH Patm(r) = Cργ, (16) with C a constant set by the pressure at the planet surface: P (Rp) = ρpkBTp mH , (17) for the ideal gas case. This gives C = ρ1−γ p c2 s,p. In order to prevent the singularity at r = 0, we cut off the planet profile at r = Rmask = 0.2Rp. Interior to r = Rib = 0.35Rp, 5 grid cells outside Rmask, we reset the planet profile at every time step in order to replenish the supply of planet material blown out by the wind. Finally, at the outer boundary r = Rob = 1.35Rp, we cut the planet profile off on the outside edge and set up a static ambient with pressure matched to the final value of the planet profile, completely specifying the initial conditions of the simulation: ρ(r) = ρatm(r), ρatm(Rob) · 10−4, with corresponding pressures of ρatm(Rmask), Patm(Rmask), Patm(r), Patm(Rob), P (r) = r < Rmask Rmask (cid:54) r (cid:54) Rob r > Rob , (18) r < Rmask Rmask (cid:54) r (cid:54) Rob r > Rob . (19) 3 RESULTS In what follows we present the results of the simulations in terms of their steady state flow characteristics. We focus on hydrodynamic flow patterns and ionization conditions within the flows. 3.1 Long-period planets: Low Mass, Low Flux case All of our simulations are centered on the planet and carried out in the planet's orbiting frame of reference. The first set of two simulations we discuss are for cases where no non- inertial or tidal forces are applied. Although the primary purpose of this paper is to study the behavior of short pe- riod giant planets (for which these forces will be significant), the non-rotating cases allow us to make contact with previ- ous work (Tripathi et al. 2015), as well as similar cases in McCann et al. (in press), and establish a baseline for the sub- sequent short period models. In addition, these simulations may be relevant to giant planets orbiting more luminous O-, B-, and potentially A-type stars at orbital radii comparable with that of Jupiter in our solar system (Sternberg et al. 2003). We begin with the long period version of Run 1, the low flux, low mass planet, shown in figure 1. The leftmost panel of this figure shows the density (hue, logarithmically scaled), in g cm−3, and velocity field (quivers) for this sim- ulation. In addition, in this panel we show the τ = 1 surface (black contour), Mach surface (magenta contour), and nom- inal planetary radius Rp (green contour). Because tidal and non-inertial forces are not present, the wind is symmetric for rotations about the x-axis; therefore, we show only a slice of the x-z plane cutting through the planet and looking up-orbit. The τ = 1 surface can be considered the surface of the planet from which the wind is launched. At this point, most of the ionizing photons have been absorbed. Below the τ = 1 surface the planetary hydrogen density and the correspond- ing optical depth quickly increase such that by R = 0.5Rp, 99.9% of the incident radiation has been absorbed. Murray- Clay et al. (2009) found that the wind solution is insensitive to conditions below the τ = 1 surface. Although the details of the flow below τ = 1 are not expected to accurately model conditions of a real giant planet, this region still plays a role in the simulation by providing a flow of neutral material to larger radii that is subsequently ionized and continually supplies the wind. The τ = 1 contour also denotes the extent of shadow- ing by the planet. In the non-rotating cases, this shadow extends uniformly in the −x direction. This results in ma- terial launched from the planet's night side remaining neu- tral. Much of this flow also remains subsonic. By following MNRAS 000, 1 -- 16 (0000) movies2 of the simulations we can see that the subsonic ma- terial in the planetary shadow eventually passes off the grid, although the Mach surface in figure 1 has yet to reach that point. On the day side of the planet the wind follows nearly radial streamlines as expected. The Mach surface is nearly spherical on the day side and is close to the wind launch- ing radius. Moving from the day to the night side of the planet, the sonic surface becomes aspherical due to pres- sure effects. Because the wind is stronger on the day side, pressure gradients drive lateral flows from one hemisphere to the other, as seen in figure 2 of Tripathi et al. (2015) as well as figure 3 of Carroll-Nellenback et al. (2017). These lateral flows influence the evolution of the Mach surface. Ini- tially the Mach surface is coincident with the τ = 1 surface in the planet's shadow. In time, however, the day-side wind sweeping around the planet collides with the night-side wind material, compressing it to a cylinder of radius less than Rp. This effect can also be seen in the center panel of fig- ure 1, which shows the neutral hydrogen fraction for the same slice through the planet. Here we see that the subsonic tail is almost entirely neutral due to the planet's ionization shadow. A linear scaling was chosen to highlight the com- pression of the tail material. Also seen in the center panel is the fact that the rest of the planetary wind (outside the shadow) carries a small fraction of neutrals with it. The neu- tral fraction nH/nH is of order 10−3 in the bulk of the wind. This is more obvious in the logarithmically scaled plots used for figures 5 through 12. The final panel in figure 1 shows the temperature, in K, on a logarithmic scale. Because the wind is mostly trans- parent to ionizing radiation outside of the planet's atmo- sphere, the temperature of the wind is determined hydro- dynamically, beginning at 3500 K at the base of the wind and cooling primarily by expansion (though radiative and recombination cooling are still present). The temperature map also highlights the interior por- tions of the planet which remain at the initial conditions. The temperature transition from the upper atmosphere to the base of the wind is sharply delineated, as would be ex- pected from the previous discussion of the optical depth. Al- though the initial planet profile is maintained at a greater radius on the night side, there are some atmospheric dynam- ics present there as well. We now consider the flow pattern in the model via figure 2, which was created by convolving random noise integrated along the streamlines of the velocity field with a color plot of the density. In this figure, hue shows density (as in the first panel of figure 1) and the texture of the plot represents flow streamlines. The nearly spherical symmetry of the flow on the day side of the planet is conspicuous. The slower, confined subsonic flow in the tail is also clearly delineated. The backflow towards the night side seen in previous studies is highlighted by this plot as well, particularly on the night side close to the planet. 2 To tions, (https://www.youtube.com/user/URAstroBEAR) of AstroBEAR movies view see the the YouTube simula- page MNRAS 000, 1 -- 16 (0000) Photoevaporation of Exoplanet Atmospheres 5 3.2 Long-period planets: High Mass, Low Flux Case Before we consider the results of the high mass case we note that the momentum flux of the wind is its ram pressure, which is determined in our case by the escape velocity vesc and the density at the base of the wind, with (cid:115) vesc = 2GMp Rp . (20) In a more general sense, the outflow velocity is not necessar- ily set by only by the escape velocity at the base of the wind, but also by the distance it takes for the flow to accelerate to the sound speed. However, in our cases, the sonic radius and the base of the wind are located within a small fraction of a planetary radius of each other, which allows us to neglect this factor when determining the relative ram pressures of our outflows. The wind material must be given an energy sufficient to achieve this velocity in order to be liberated from the planet's gravity. Since the (nominal) planet radius Rp is the same for all of our simulations, only the planet mass affects the velocity with which the winds are launched. The escape velocity for Runs 1 and 3 is 10.9 km s−1, while the escape velocity for the higher-mass Runs 2 and 4 is 21.1 km s−1. This difference in the wind speed will bear directly on the models, since it affects the ram pressure of the wind. The other factor in the ram pressure is the density of the wind. Although the density is diluted as the wind ex- pands (mostly) spherically away from the planet, the rela- tive density at particular radii is set by the density at the base of the wind. The parameters at the base of the wind are given in table 2. The low-flux runs have densities within 20% of each other, and the high-flux runs have densities within 10% of each other; however, the low-flux runs differ from the high-flux runs by a factor of 4.5. Note that this is greater than the increase in density between the two sets of runs (see table 1), because the base of the wind in the high flux cases is deeper inside the planet and launching from a higher ionization fraction. While this could explain a dichotomy in the behavior of the low-flux and high-flux runs, only the in- creased planetary mass could produce a significant difference between the two low-flux runs. We now consider the details of the long period Run 2 simulation. Figure 3 is the same as figure 1 for the higher- mass planet. Although the large-scale behavior is very sim- ilar, still showing a nearly spherically symmetric outflow, there are differences (particularly in the lateral flow around the planet) that will become more significant in the rotat- ing cases. First note that although the ionization front is in a similar location, the Mach surface is located significantly outside Rp. In addition, while the sonic surface is still com- pressed by the lateral flows far from the planet, it has a width comparable to Rp near the planet. The neutral density (center panel) highlights another significant change due to the stronger wind blowing around the planet. Here we see stronger compression of the neutral tail than found previously, which is almost entirely due to the lateral flows from the day side of the planet. Finally, due to the higher planet surface temperature the ionization front cannot be seen as clearly in the rightmost panel. Figure 4 shows the flow streamlines for the higher-mass 6 A. Debrecht et al. Table 2. Atmospheric properties at base of wind Simulation Low Mass, Low Flux High Mass, Low Flux Low Mass, High Flux High Mass, Low Flux T K 3300 5302 5312 6856 ρ g cm−3 1.31 × 10−16 1.62 × 10−16 5.80 × 10−16 5.30 × 10−16 R Rp 0.80 0.86 0.61 0.80 X 0.24 0.24 0.54 0.55 planet. This figure serves to highlight the significant flow around the planet from the day side to the night side, which alters the characteristics of the neutral tail as discussed with figure 3. In general, the two long period models establish the code's ability to accurately capture the details of the launch- ing of the wind via ionizing stellar flux as well as the result- ing larger-scale flow out to many planetary radii. 3.3 Short-period planets: Low Mass Low Flux Case We now consider the models which comprise the main focus of this work. In Carroll-Nellenback et al. (2017) the redi- rection of the planetary wind into a torus-like configuration comprised of up-orbit and down-orbit arms was investigated. In that study it was found that the formation of the torus could be understood, to first order, via Coriolis forces. This was expressed via a "Coriolis length" for wind flows, given by (cid:114) RΩ a = aq λRp (q + 1) , (21) where a is the orbital radius and q = Mp/M∗. For small q and a it can be seen that the ratio RΩ a < 1 and the wind streamlines are turned 90◦ by the Coriolis force on length scales smaller than the orbit. In the Carroll-Nellenback et al. (2017) study the planetary outflow was driven from preset temperature and density conditions at a preset radius that was intended to represent the outer edge of the planet (its exobase). The day side of the planet was set to a temper- ature T (θ) = Tp max [0.01, cos(θ)] where θ is the angle of incidence of the light from the star (and the angular dis- tance from the sub-stellar point). The night side was kept at .01Tp = 100K (Stone & Proga 2009). We now explore the flow patterns produced for short period planets (strong Coriolis and tidal forces) when the wind is generated self- consistently via radiation deposition in the atmosphere. We first consider the low mass, low flux, short orbit case. Figure 5 again shows the density (left), neutral frac- tion (center, now logarithmically scaled), and temperature (right) with contours of Mach number (magenta), optical depth (black), and planet radius (green). The simulation is again carried out in the co-rotating frame, but because we now have a rapidly orbiting planet, we show a 2-D slice looking down on the orbital plane (top row) and a side view (bottom row). As in Carroll-Nellenback et al. (2017) and Matsakos et al. (2015), we again see the formation of up- and down- orbit wind trajectories (top row). The wind flow no longer fills the whole computational domain. Rather, it is confined to a torus with a quasi-cylindrical cross section (bottom row). Note that the wind is bounded in all directions with shocks at the interface with the original ambient medium, as well as additional shocks in the interior of the flow. While the structure of the wind has changed significantly from the non-rotating cases, one important similarity can be seen on the day side of the planet in the density panel (top left). Here, the ionization front and Mach surface are nearly iden- tical to those found in the non-rotating case, suggesting, as one would expect, that these are set only by the incident flux and the planet parameters. In addition to the changes in global flow structure, there is no longer a subsonic tail present in this model. A small portion of the material flowing from the night side that has been turned down-orbit does remain subsonic near the planet, which creates shocks with gas from the up-orbit arm that has been turned completely around the planet and now flows in the −y direction. This interior shock creates a high- density region which can also be seen in the τ = 1 surface. Rather than forming directly behind the planet, the shadow is now extended in the −y direction by the tail on the down- orbit side of the planet. The neutral fraction nH/nH (top center panel) shows that the tail remains neutral with an extent defined by the ionization timescale and wind velocities via advection. In this model, the neutral tail extends a significant distance from the planet. There is a similarly-shaped feature mirrored across the y axis, where ionized material originating from the down-orbit side of the planet turns in the up-orbit direction. Finally, note that, as in the long period simulations, the bulk of the wind far from the ionization shadow maintains a small neutral fraction of order 10−2. The temperature (top right panel) highlights the inter- nal shock structures present in the wind. There are three primary internal shocks of interest. The first appears in the left-most boundary of the flow. This shock is created by ma- terial leaving the planet at latitudes closer to the sub-stellar point colliding with material leaving closer to the −y termi- nator and nightside of the planet (see also 6). The second shock occurs near the right edge of the flow and forms from material leaving latitudes near the sub-stellar point collid- ing with material leaving the closer to the +y terminator and night side. Finally, we see material leaving the planet near the +y terminator colliding with material leaving the night side, which forms the shock structure at the supersonic neutral tail. The density in the x−z plane (bottom left panel) shows that the rotational forces creating the up-orbit and down- orbit arms of the wind also confine it to a torus surrounding the planet. Oblique internal shocks can be seen in the cor- responding temperature plot bounding the rim of the torus. Note that the τ = 1 surface is once again defined by the ex- tent of the planet in this plane. The edges of the torus in the x − z plane also show some corrugations, which may be due to instabilities at the interface with the ambient medium. The neutral fraction in the x − z plane (bottom center panel) shows that the neutral tail extends through the ion- ization shadow of the planet. The fact that the tail does not reach the torus edge in this cut is consistent with its being driven down-orbit via the shocks visible in the cut through the orbital plane. Figure 6 shows the flow texture of the low flux, low mass planet run from the top view (left) and side view (right). The MNRAS 000, 1 -- 16 (0000) Photoevaporation of Exoplanet Atmospheres 7 Figure 1. Density (left), neutral fraction (center), and temperature (right) for Run 1 in the non-rotating case, standing in the orbital plane and looking in the up-orbit direction. The quivers describe the velocity field, and the contours are of the Mach surface (magenta), the τ = 1 surface (black), and Rp (green). The extended neutral tail is due to the shadow of the planet. 3.4 Short-period planets: High Mass, Low Flux Case In figure 7 we see the effect of raising the planet's mass while keeping the ionizing flux constant. As seen in the density (top left), significant portions of the up-orbit and down-orbit arms of the flow are now subsonic, in contrast with Run 1. Additionally, the Coriolis force confines the wind much more strongly in this case, with the boundaries of the outflow no longer extending to the edge of the grid in the x direction. As in the non-rotating case, the Mach surface is located farther from the base of the wind here. Due to the lack of a strong shock enhancing the neutral density on the night side of the planet, the τ = 1 contour no longer protrudes significantly beyond the shadow of the planet. The neutral fraction (top center panel) highlights one consequence of the planetary wind's stronger confinement to the up/down orbit torus. Here, the neutral tail no longer shows an extended arc behind and down-orbit from the planet. As in the non-rotating, high mass planet case (fig. 3), this model has a stronger wind in terms of its ram pressure Pram = ρwv2 w which, in this case, sweeps around the planet and truncates the neutral tail. This is also shown promi- nently in the temperature map (top right panel), where the temperature jump seen in the low planet mass case is absent. In addition, the oblique shocks are notably weaker (and in the case of the down-orbit shocks, completely absent) here. The slices in the x − z plane (bottom set of panels) again show the strong confinement of the wind. From this perspective, we can see the torus of wind material has a smaller cross section and none of the wind leaves the grid in this plane. The density map in this plane (bottom left panel) again shows the Mach surface being held farther from the planet than in Run 1. The bottom center panel also shows a faint increase in the neutral fraction in the shadow of the planet due to wind material recombining as it passes through the shadow. Figure 8 shows the flow texture map for the higher- mass planet. Comparing figure 6, we see that the wind has a smaller turning radius. The cause of the down-orbit subsonic flow is also revealed to be material leaving in the up-orbit direction and being turned sharply down-orbit to impact material launching from the night side. The side view (right Figure 2. Flow-texture plot of Run 1 in the non-rotating case, showing density (hue) and local streamline orientation (texture). The outflow from the planet is nearly spherically symmetric due to the lack of rotational forces, with a small flow around the planet from the day side to the night side. redirection of the flow at the shock structures is most appar- ent in the top view, with the origins of the shocked material simple to trace as described previously. Note the similarity of the streamlines in this plane to the semi-analytic mod- els of figure 6 in Carroll-Nellenback et al. (2017). In those calculations Carroll-Nellenback et al. (2017) were able to predict the wind streamlines from ballistic trajectories of wind parcels launched from the planet's surface and subject to Coriolis and tidal forces (in particular see the Ξp = 0.2 and τ = 1 models as defined in that paper). The side view shown in 6 shows that the wind flow near the planet has a symmetric structure and that stream- lines encounter the internal shocks close to the wind-ambient medium interface. The stretching of the wind due to stellar tidal forces is also apparent in the right panel. MNRAS 000, 1 -- 16 (0000) Low Mass, Low Flux, Non-Rotating-10-50510x (Rp)-10-50510z (Rp)1e-201e-191e-181e-171e-161e-15 (g/cm3)-10-50510x (Rp)-10-50510z (Rp)00.20.40.60.81Neutral Fraction-10-50510x (Rp)-10-50510z (Rp)1e+024e+021.6e+036.3e+032.5e+041e+05T (K)Low Mass, Low Flux, Non-Rotating-10-50510x (Rp)-10-50510z (Rp)1e-201e-191e-181e-171e-161e-15 (g/cm3)8 A. Debrecht et al. Figure 3. Density (left), neutral fraction (center), and temperature (right) for Run 2 in the non-rotating case, standing in the orbital plane and looking in the up-orbit direction. The quivers describe the velocity field, and the contours are of the Mach surface (magenta), the τ = 1 surface (black), and Rp (green). The extended neutral tail is again due to the shadow of the planet, but in this case is diluted and compressed by the stronger planetary wind. 1. We also again see a high-density neutral tail present on the night side. The shock formed from day side material sweeping around to the night side of the planet again creates a protrusion of the τ = 1 surface down-orbit, beyond the shadow of the planet. The total extent of the wind flow (up- orbit and down-orbit arms) is similar to Run 1 (and again larger than in Run 2). The strong oblique internal shocks (most apparent in the temperature panel) are also present as in Run 1. One of the key differences between this high flux case and the previous low flux version can be seen in the neutral density (top center panel). In the high flux case we see a thicker neutral tail due to the the increased planetary mass flux which occurs thanks to the higher radiative flux in Run 3. From Table 3 we see that MRun3 ∼ 10 MRun1. In addition, the increase in neutral fraction across the shadow of the planet can be attributed to the lower equilibrium ionization state in the bulk of the wind for the high-flux runs, which leads to greater recombination in the tail (due to increased electron density) and increased contrast between the highly- ionized wind material and the recombined material in the planet's shadow. The temperature map (top right panel) also highlights changes in the structure of the tail. Run 3 shows two shocks in the tail, one when the material turned from the up-orbit outflow shocks with the tail material, and another when the tail material shocks with the remaining night-side and down- orbit outflow (these flows are more easily seen in Fig 10). In addition, the temperature map highlights again the discon- tinuity between the base of the wind and the undisturbed inner structure of the atmosphere. Finally, we note the pres- ence of what appear to be instabilities at the boundary of the planetary wind and ambient medium at the sub-stellar regions of the flow, in contrast with the smooth flow in the low-flux runs. The density panel in the x− z plane (bottom left) again shows the significant oblique shock structure in the confined planetary wind torus, with the Mach surface located nearer the planet at the poles. The neutral density (bottom center) highlights both the greater extent of the neutral tail driven from the night side and the more extended recombination shadow of the planet. MNRAS 000, 1 -- 16 (0000) Figure 4. Flow-texture plot of Run 2 in the non-rotating case, showing density (hue) and local streamline orientation (texture). The outflow is again nearly spherical, but the flow around the planet can be seen much more strongly than in Run 1. panel) also provides evidence for the larger effect of the ro- tational forces, with only a very small portion of the outflow appearing azimuthally symmetric when compared to figure 4. 3.5 Short-period planets: Low Mass High Flux Case We now consider simulations for planets exposed to an in- cident flux that is an order of magnitude higher than the previous models. Figure 9 includes the same 6 panels shown previously, but this time for the low planet mass, high stellar flux case (Run 3). The overall flow pattern is similar to that seen in the low planet mass, low stellar flux case (Run 1) with a few key differences. In the density panel (top left) we see that the Mach surface is very close to the base of the wind, as in Run High Mass, Low Flux, Non-Rotating-10-50510x (Rp)-10-50510z (Rp)1e-201e-191e-181e-171e-161e-15 (g/cm3)-10-50510x (Rp)-10-50510z (Rp)00.20.40.60.81Neutral Fraction-10-50510x (Rp)-10-50510z (Rp)1e+024e+021.6e+036.3e+032.5e+041e+05T (K)High Mass, Low Flux, Non-Rotating-10-50510x (Rp)-10-50510z (Rp)1e-201e-191e-181e-171e-161e-15 (g/cm3)Figure 10 shows the flow texture map for Run 3. Once again, the streamlines show the reorientation of the gas parcels launched from the planet's exobase due to Corio- lis and tidal forces. Subtle differences in the flows between Run 3 and Run 1 can been seen in terms the orientation of the night-side shocks and neutral flow. 3.6 Short-period planets: High Mass, High Flux Case The results for Run 4 are shown in figure 11. In the density panels, we note the strong similarities between the Mach surfaces for Run 4 and Run 2, with the two subsonic exten- sions in the up-orbit and down-orbit arms of the wind. Once again there is no high-density region extending beyond the night side to cause the ionization front to protrude beyond the planet's shadow. The neutral fraction panels display the truncation of the neutral tail, as seen in Run 2, and also show the recombination occurring in the shadow of the planet as in Run 3. In the neutral density in the x − y plane, we see again the strong shock in the up-orbit arm that is absent in the down-orbit arm. In addition, we see the same shadow behind the planet due to recombination as in Run 3. The density in the x − z plane displays a new flow fea- ture, wherein the subsonic area behind the planet is split into two lobes toward the north and south poles of the planet. In addition, we see as in figure 9 the turbulence of the star- ward edge of the toroidally confined wind. The new lobe structure can also be seen in the bottom center panel, where the lobes are traced out by highly neutral material. Finally, the bottom right panel shows that the oblique shocks along the orbital direction have moved in significantly toward the planet on the night side. Figure 12 shows the flow texture map for the Run 4. The flow is generally similar to what is seen in the low flux, high mass case. Differences occur mainly in the details of where stagnation regions occur behind the planet and where shocks lead to redirection of the planetary wind. 4 ANALYSIS 4.1 Wind Mass Loss Rates and Regimes Photoevaporative winds can be classified based on the way incoming stellar flux is distributed in the flow (Murray-Clay et al. 2009; Owen & Alvarez 2016). For "energy-limited" mass loss, most of the energy deposited by photoionization as heat is ∝ FU V R2 p and goes into P dV work. Losses due to radiation and internal energy changes are small. The P dV work, measured per unit mass, can be expressed as which yields an energy-limited mass loss rate of Mel ∼ 6×109(cid:16) (cid:17)(cid:18) Rp P δV ρR2 pH ∼ GMp Rp (cid:19)3(cid:18) 0.7MJ (cid:19)(cid:18) , (22) (cid:19) −1. g s FU V 450 erg cm−2 0.3 1010cm Mp (23) Note that the mass loss rate is linear with FU V . In the more detailed numerical models of Murray-Clay et al. (2009) the actual dependence was slightly weaker: Mel ∼ F 0.9 U V . Alternatively, for "radiation/recombination limited" MNRAS 000, 1 -- 16 (0000) Photoevaporation of Exoplanet Atmospheres 9 Table 3. Mass loss rates Simulation Low Mass, Low Flux, non-rotating High Mass, Low Flux, non-rotating Low Mass, Low Flux High Mass, Low Flux Low Mass, High Flux High Mass, High Flux M g s−1 3.42 × 1010 3.11 × 1010 3.87 × 1010 3.35 × 1010 1.70 × 1011 2.04 × 1011 (cid:18) (cid:19)1/2 mass loss, the photoevaporative flows are such that the input UV power is mostly lost to radiative cooling. The balance of radiative heating and cooling then keeps the gas temper- ature at T ∼ 104 K. In this case it can be shown that the mass loss rate goes as Mrl ∼ 4 × 1012 FU V 5 × 105 erg cm−2 −1. g s (24) We can examine this question directly with figure 13, which shows the relative size of heating and cooling terms in one of our runs. The remainder are nearly identical. From this plot it is clear that all of the cases examined in this study are in the energy-limited regime, with nearly all of the energy deposited by EUV radiation being used to launch the wind. Mass loss rates for each run were calculated by integrat- ing the flux through a spherical surface at 3Rp. Despite the significant differences in flow structure, the mass loss rates of each low flux run were similar. We find mass loss rates of between 3 × 1010 and 4 × 1010 for the low flux cases and approximately 2 × 1011 for the high flux cases, comparable to the mass loss rates found in Tripathi et al. (2015) for the same radiation flux. Note that we see the wind mass loss increasing with flux at a lower rate than predicted for an energy-limited flow M ∼ F .9 U V , which for a given fac- tor of 10 increase should yield a mass loss increase factor of 7.9. In the models we find MRun3/ MRun1 ∼ 4.39 and MRun4/ MRun2 ∼ 6.09. However, because the mass loss rate models were derived with a planet of constant radius, the fact that our winds were launched from slightly different radii in each run explains the difference in mass loss rates τ =1F .9 here. If we include the radius dependence, U V (Murray-Clay et al. 2009), we find theoretical mass loss ra- MRun3/ MRun1 ∼ 3.53 and MRun4/ MRun2 ∼ 6.39, tios of similar to our measured values. M ∼ R3 4.2 Rotational effects In order to discuss the effects of rotation on the wind, we define a few quantities (following Carroll-Nellenback et al. (2017)). φc is the critical angle (measured clockwise from the substellar ray) for determining whether material leaving the planet joins the up-orbit or down-orbit arm, given by (cid:18)−2rΩ (cid:19) v φc = arctan . (25) For our analysis, we take the radius and sound speed at the sonic radius along the substellar ray, which leads to a slight overestimate of the value of φc. In addition, the orbital angle of the subsonic eddy which occurs in the up-orbit arm of the 10 A. Debrecht et al. Figure 5. Density (left), neutral fraction (center), and temperature (right) for Run 1 in the short period case, with a view down onto the orbital plane (top row) and standing in the orbital plane looking in the up-orbit direction (bottom row). The quivers describe the velocity field, and the contours are of the Mach surface (magenta), the τ = 1 surface (black), and Rp (green). There is an extended thin supersonic tail of neutral material leaving the night side of the planet, and complex oblique shock structures along the tube of the wind, which is confined in the radial and perpendicular directions. Figure 6. Flow-texture plot of Run 1 in the short period case, showing density (hue) and local streamline orientation (texture), looking down onto the orbital plane (left) and standing in the orbital plane and looking up-orbit (right). The redirection of the flow at the shocks is apparent, as well as the azimuthal symmetry of the streamlines out to ∼ 4Rp. wind is given by ΘD = 12v Ωa , (26) where the speed is again taken at the sonic radius along the substellar ray. These quantities are summarized for the suite of simulations in Table 4. It is noteworthy that our simulations find the same ro- Table 4. Rotational quantities Simulation φc ΘD −0.25π Low Mass, Low Flux High Mass, Low Flux −0.24π Low Mass, High Flux −0.17π High Mass, High Flux −0.19π 0.139π 0.173π 0.156π 0.198π MNRAS 000, 1 -- 16 (0000) Low Mass, Low Flux-10-50510x (Rp)-10-50510y (Rp)1e-201e-191e-181e-171e-161e-15 (g/cm3)-10-50510x (Rp)-10-50510y (Rp)1e-061.6e-050.000250.0040.0631Neutral Fraction-10-50510x (Rp)-10-50510y (Rp)1e+024e+021.6e+036.3e+032.5e+041e+05T (K)-10-50510x (Rp)-10-50510z (Rp)1e-201e-191e-181e-171e-161e-15 (g/cm3)-10-50510x (Rp)-10-50510z (Rp)1e-061.6e-050.000250.0040.0631Neutral Fraction-10-50510x (Rp)-10-50510z (Rp)1e+024e+021.6e+036.3e+032.5e+041e+05T (K)Low Mass, Low Flux-10-50510x (Rp)-10-50510y (Rp)1e-201e-191e-181e-171e-161e-15 (g/cm3)-10-50510x (Rp)-10-50510z (Rp)1e-201e-191e-181e-171e-161e-15 (g/cm3)Photoevaporation of Exoplanet Atmospheres 11 Figure 7. Density (left), neutral fraction (center), and temperature (right) for Run 2 in the short period case, with a view down onto the orbital plane (top row) and standing in the orbital plane looking in the up-orbit direction (bottom row). The quivers describe the velocity field, and the contours are of the Mach surface (magenta), the τ = 1 surface (black), and Rp (green). Here the neutral tail is disrupted by material turned from the up-orbit arm by the Coriolis force, and the wind is more strongly confined along the radial and perpendicular directions. Figure 8. Flow-texture plot of Run 2 in the short period case, showing density (hue) and local streamline orientation (texture), looking down onto the orbital plane (left) and standing in the orbital plane and looking up-orbit (right). The wind has a larger velocity here, and is therefore turned much more strongly than in Run 1. The azimuthal symmetry in the side view is limited due to this turning. tational effects as found in the Rotating case of McCann et al. (in press), conducted using the hydrodynamics code Athena. In particular, note the similarities between figures 2(c) and (f) and our figure 7. 4.3 Planetary ionization shadow The change in the neutral fraction behind the planet seen in Runs 3 and 4 can be understood in terms of the shadowing of the higher flux by the planet. The shadow is enhanced by the recombination of material temporarily protected by the planet from ionizing radiation. We can compare the re- combination timescale τrecom and the shadow crossing time MNRAS 000, 1 -- 16 (0000) High Mass, Low Flux-10-50510x (Rp)-10-50510y (Rp)1e-201e-191e-181e-171e-161e-15 (g/cm3)-10-50510x (Rp)-10-50510y (Rp)1e-061.6e-050.000250.0040.0631Neutral Fraction-10-50510x (Rp)-10-50510y (Rp)1e+024e+021.6e+036.3e+032.5e+041e+05T (K)-10-50510x (Rp)-10-50510z (Rp)1e-201e-191e-181e-171e-161e-15 (g/cm3)-10-50510x (Rp)-10-50510z (Rp)1e-061.6e-050.000250.0040.0631Neutral Fraction-10-50510x (Rp)-10-50510z (Rp)1e+024e+021.6e+036.3e+032.5e+041e+05T (K)High Mass, Low Flux-10-50510x (Rp)-10-50510y (Rp)1e-201e-191e-181e-171e-161e-15 (g/cm3)-10-50510x (Rp)-10-50510z (Rp)1e-201e-191e-181e-171e-161e-15 (g/cm3)12 A. Debrecht et al. Figure 9. Density (left), neutral fraction (center), and temperature (right) for Run 3 in the short period case, with a view down onto the orbital plane (top row) and standing in the orbital plane looking in the up-orbit direction (bottom row). The quivers describe the velocity field, and the contours are of the Mach surface (magenta), the τ = 1 surface (black), and Rp (green). Note again the presence of a dense tail of neutrals streaming from the night side of the planet, with greater cross-section than was found in Run 1. Figure 10. Flow-texture plot of Run 3 in the short period case, showing density (hue) and local streamline orientation (texture), looking down onto the orbital plane (left) and standing in the orbital plane and looking up-orbit (right). τshadow for Run 4 to get an estimate of the expected neutral fraction on the far side of the tail: τrecom = 1 αBne = 4.2 × 105 s, τshadow = = 1.34 × 104 s. 2Rp vy (27) (28) This gives us an estimate of the total change in neutral frac- tion over the length of the shadow of τshadow/τrecom = 0.032. As figure 14 demonstrates, the neutral fraction climbs to 0.035 at the edge of the shadow before dropping sharply as expected. It is also possible to determine whether there is a "sweet spot" in parameter space such that the recombined material from the planet's shadow remains neutral for a sufficiently long time to be detectable after it passes out of the shadow (in the absence of a denser neutral tail, such as in Runs 2 MNRAS 000, 1 -- 16 (0000) Low Mass, High Flux-10-50510x (Rp)-10-50510y (Rp)1e-201e-191e-181e-171e-161e-15 (g/cm3)-10-50510x (Rp)-10-50510y (Rp)1e-061.6e-050.000250.0040.0631Neutral Fraction-10-50510x (Rp)-10-50510y (Rp)1e+024e+021.6e+036.3e+032.5e+041e+05T (K)-10-50510x (Rp)-10-50510z (Rp)1e-201e-191e-181e-171e-161e-15 (g/cm3)-10-50510x (Rp)-10-50510z (Rp)1e-061.6e-050.000250.0040.0631Neutral Fraction-10-50510x (Rp)-10-50510z (Rp)1e+024e+021.6e+036.3e+032.5e+041e+05T (K)Low Mass, High Flux-10-50510x (Rp)-10-50510y (Rp)1e-201e-191e-181e-171e-161e-15 (g/cm3)-10-50510x (Rp)-10-50510z (Rp)1e-201e-191e-181e-171e-161e-15 (g/cm3)Photoevaporation of Exoplanet Atmospheres 13 Figure 11. Density (left), neutral fraction (center), and temperature (right) for Run 4 in the short period case, with a view down onto the orbital plane (top row) and standing in the orbital plane looking in the up-orbit direction (bottom row). The quivers describe the velocity field, and the contours are of the Mach surface (magenta), the τ = 1 surface (black), and Rp (green). Figure 12. Flow-texture plot of Run 4 in the short period case, showing density (hue) and local streamline orientation (texture), looking down onto the orbital plane (left) and standing in the orbital plane and looking up-orbit (right). and 4). Using the ionization timescale, τion = 1 σphF0 = 800 s, (29) we can calculate the distance we would expect it to take to return to the equilibrium wind ionization: wind is still recombining as it exits the shadow, so that the ionization fraction does not increase as quickly as it would without the recombination. Therefore, it may be possible to detect the material that has recombined in the planet's shadow even for flux as high as we use here. x = τionvy = 1.8 × 109 cm = 0.12Rp. (30) 4.4 Neutral tail In fact, it takes ∼ 3 4 Rp for the material to ionize back to the equilibrium level of the wind in Run 4. The extended distance here can be partially explained by the fact that the In section 3, we identified the importance of neutral mate- rial streaming laterally from the night side of the planet, as well as its absence in the simulations with high planet MNRAS 000, 1 -- 16 (0000) High Mass, High Flux-10-50510x (Rp)-10-50510y (Rp)1e-201e-191e-181e-171e-161e-15 (g/cm3)-10-50510x (Rp)-10-50510y (Rp)1e-061.6e-050.000250.0040.0631Neutral Fraction-10-50510x (Rp)-10-50510y (Rp)1e+024e+021.6e+036.3e+032.5e+041e+05T (K)-10-50510x (Rp)-10-50510z (Rp)1e-201e-191e-181e-171e-161e-15 (g/cm3)-10-50510x (Rp)-10-50510z (Rp)1e-061.6e-050.000250.0040.0631Neutral Fraction-10-50510x (Rp)-10-50510z (Rp)1e+024e+021.6e+036.3e+032.5e+041e+05T (K)High Mass, High Flux-10-50510x (Rp)-10-50510y (Rp)1e-201e-191e-181e-171e-161e-15 (g/cm3)-10-50510x (Rp)-10-50510z (Rp)1e-201e-191e-181e-171e-161e-15 (g/cm3)14 A. Debrecht et al. Figure 13. Plot of the heating and cooling terms along the star- planet axis for Run 4. Because the cooling is primarily by expan- sion (PdV work), we see that we are in the energy-limited regime. Also note the compressive heating on the night side of the planet. Figure 14. Neutral fraction (left axis) and recombination and shadow crossing timescales (right axis) for Run 4, taken at x = 4Rp. As we enter the planetary wind material, the recombination timescale drops sharply and the neutral fraction increases to its steady-state bulk wind value of approximately 0.002. Upon enter- ing the planet's shadow around 1Rp, the neutral fraction begins to rise as the ionizing radiation is blocked. Exiting the planet's shadow around −1Rp, the wind material ionizes precipitously as it is again exposed to the ionizing radiation. mass (Runs 2 and 4). While we don't believe any such phe- nomenon has yet been detected, it could provide useful di- agnostic information about the planet. The vector field of velocities in figure 15 provide a guide to the depth at which the neutral tail material originates. In addition to its pres- ence as an indicator of low planet mass, the fact that the material in the tail is drawn from within the planet could Figure 15. A close-up of the steady state of the planet in Run 1, with the red contour showing Rib, the radius held fixed during the simulation, the blue contour showing the radius of the ioniza- tion front at the substellar point Rτ =1 = 0.8Rp, and the green contour showing the nominal planet radius Rp. The vectors show the velocity field, and the hue is the neutral fraction. Note that much of the material in the neutral tail appears to be originating within the blue contour. allow its use as a probe into the composition of the planet at deeper radii than we are currently able to detect. 5 DISCUSSION AND CONCLUSIONS In the preceding analysis, we have studied the effects of planet mass and stellar ionizing flux on photoevapora- tive winds launched from gaseous giant planets using self- consistent radiative transfer. We found that low mass plan- ets produce a potentially observable extended neutral tail of material being advected from within the night side of the planet that is not present in the simulations of high mass planets. We also see that low mass planets create a torus of wind material of larger cross section than that produced by high mass planets. In addition, the high flux simulations result in a more pronounced recombination region in the ionization shadow of the planet than seen in the low flux cases. Finally, we observed mass loss rates a factor of ap- proximately 5x greater in the high flux cases than those seen in the low flux simulations. Here we compare to two previous works which investi- gated the physics most relevant to the current study, Carroll- Nellenback et al. (2017) and Tripathi et al. (2015). We be- gin with Carroll-Nellenback et al. (2017). In that study the planetary wind was launched from a fixed pressure and den- sity boundary condition rather than by deposition of energy via ionizing radiation. We have extended that study by in- cluding a wind launched by self-consistent radiative heating in both rotating and non-rotating frames. We first compare panel ANISO of figure 3 of Carroll-Nellenback et al. (2017) (which shows the effect of orbital motion on wind stream- lines) to the flow-texture plots of our simulations. The com- parison shows that the streamlines close to the planet match well, particularly in the low mass planet case. We also wish to investigate whether orbital motion (i.e., MNRAS 000, 1 -- 16 (0000) -3-2.5-2-1.5-1x (Rp)-10-8-10-9010-910-8Heating/Cooling Rates (erg/s)Energy Deposition: High Mass, High FluxPdV workLy- coolingRecombination coolingIonization heatingTotal-10-50510y (Rp)00.0050.010.0150.020.0250.030.035Neutral Fraction103104105106107Timescale (s)Recombination Across Planetary ShadowNeutral FractionShadow Crossing TimeRecombination Timescalethe Coriolis and centrifugal forces) has any effect on the mass loss rate from the planet. In Carroll-Nellenback et al. (2017), it was found that uneven planetary heating (intrin- sic in tidally-locked planets) reduced the mass loss rate by almost a factor of 2, but that placing an unevenly-heated planet into a rotating frame of reference had no effect on the mass loss rate. Although we did not run a simulation with uniform planetary heating, we also find that rotation has little effect on the mass loss rate through the wind, as the mass loss rates for Runs 1 and 2 in the rotating and non-rotating cases differ by only 10% (see table 3). We now compare the results of our study that of Tri- pathi et al. (2015), who performed simulations of the radia- tive heating of the atmosphere of similar planets in the pres- ence of tidal forces but without non-inertial forces, which we have now included. In order to compare similar simula- tions, we consider the long-period planet with high planet mass, with no rotational or tidal effects (figure 1. Note that our masses differ by almost a factor of two and our surface densities by an order of magnitude. Tripathi et al. (2015) found that the sonic surface was significantly farther from the planet, at a distance of approximately 0.6Rp compared to a distance of 0.1Rp for our simulation (compare Tripathi et al. (2015) figure 3 and the left panel of figure 3). This dif- ference is likely explained by the lower planet mass, which we have seen in comparing our own models leads to a de- crease in the sonic radius. We can also compare the overall flow patterns in the wind, shown in the bottom left panel of figure 2 in Tripathi et al. (2015) and figure 4. Although the flow patterns are broadly similar, with axisymmetric outflow and lateral flows toward the night side, the lateral flows are flowing back slightly more strongly onto the planet in the previous study. We find a mass loss rate approximately 1.4x larger for the low flux case than Tripathi et al. (2015). Although we did not simulate the high flux cases without rotational effects, as argued above, the addition of rotation has little effect on the mass loss rate and we can conclude that the high flux non- rotating case would display a mass loss rate of approximately 2 × 1011 g/s, nearly identical to that found in the high-flux case of Tripathi et al. (2015). Finally, we can also locate our simulated planetary winds in the regimes identified by both Murray-Clay et al. (2009) and Matsakos et al. (2015). As discussed in section 4, all of our simulations are located in the energy-limited regime rather than the radiation/recombination (or cool- ing) limited regime, with the majority of the input EUV energy being used for the expansion of the wind rather than radiative cooling. In addition, we have no stellar wind or magnetic fields. Our low ambient pressure and relatively low planet mass (even in the high mass cases) allow us to place all of our simulations in the "Type III" regime of Matsakos et al. (2015), with the wind flow dominating planetary tidal forces. To produce their Type I or Type IV interactions, a strong magnetic field and/or a weak planetary wind (rela- tive to the ambient/stellar wind pressure) are required. This confines the wind, preventing it from forming the up-orbit arm. A high-mass planet might produce a Type I interaction, with the wind confined to the vicinity of the planet, while a low-mass planet would produce a Type IV interaction, with the wind overflowing the Roche lobe and being captured by the star. To produce a Type II interaction, a higher-mass MNRAS 000, 1 -- 16 (0000) Photoevaporation of Exoplanet Atmospheres 15 planet or stronger stellar wind pressure would be required, so that the wind material is again confined to the vicinity of the planet rather than being captured by the star. The fact that our interactions are Type III means we would expect in a simulation of the full system a long up-orbit arm extend- ing a significant distance around the star with accretion at the far end of this arm, as well as an extended tail. 5.1 Phenomena not yet considered There are a number of physical processes which we have not considered in these simulations, but which may have impor- tant consequences for the structure of the wind and for the observational signatures of evaporating planets. The inter- action of the planetary wind with a strong stellar wind has been shown to be able to confine the up-orbit arm of the wind (Matsakos et al. 2015; Schneiter et al. 2016; McCann et al. in press). Radiation pressure has also been shown to be a potentially important mechanism for the acceleration of neutral hydrogen to speeds required to reproduce the ob- served Lyman-α absorption (Schneiter et al. 2016; Bourrier et al. 2015). Another mechanism proposed to produce such a fast neutral population is charge exchange between the ionized stellar wind and the planetary wind (Bourrier et al. 2016; Tremblin & Chiang 2013; Christie et al. 2016). Mag- netic fields are also expected to affect both the launching of the wind and its interaction with the stellar wind (Owen & Adams 2014; Matsakos et al. 2015; Villarreal D'Angelo et al. 2018). Finally, we note that it has recently been pro- posed that the metastable helium line at 10830 Awill provide an excellent source of observational diagnostics for evapo- rating exoplanets, since it is relatively abundant and unaf- fected by ISM absorption (Oklopci´c & Hirata 2018). This re- quires modeling the photoionization dynamics of additional species. Only McCann et al. (in press) of the works above, however, has included a self-consistent treatment of photo- ionization in driving the wind from the planet's atmosphere as was done in this study. Thus many of the listed processes can be added productively to the kind of simulation platform developed here. Although it is a more difficult computational problem, it will also be important to include the effects of spherical dilution of the radiation field in order to examine the full orbital behavior of the wind material. As an example of a phenomenon where the full orbital treatment is required, ac- cretion onto the star (as seen in Matsakos et al. (2015)) could provide another important observational diagnostic for the behavior of evaporating hot Jupiters. In addition, the full orbital treatment is likely to be important for cases such as WASP-12b, where we have previously shown (Debrecht et al. 2018) that it is possible for a torus to form completely surrounding the star. Such a torus is likely to be significantly affected by radiation pressure. 6 ACKNOWLEDGMENTS We thank Luca Fossati for many helpful conversations. We also thank the Other Worlds Laboratory (OWL) at Univer- sity of California, Santa Cruz for facilitating this collabo- ration by way of the OWL Exoplanets Summer Program, 16 A. Debrecht et al. funded by the Heising-Simons Foundation. This work used the computational and visualization resources in the Center for Integrated Research Computing (CIRC) at the Univer- sity of Rochester and the computational resources of the Texas Advanced Computing Center (TACC) at The Uni- versity of Texas at Austin, provided through allocation TG- AST120060 from the Extreme Science and Engineering Dis- covery Environment (XSEDE) (Towns et al. 2014), which is supported by National Science Foundation grant number ACI-1548562. Financial support for this project was pro- vided by the Department of Energy grant DE-SC0001063, the National Science Foundation grants AST-1515648 and AST-1411536, and the Space Telescope Science Institute grant HST-AR-12832.01-A. REFERENCES Ballester G. E., Sing D. K., Herbert F., 2007, Nature, 445, 511 Ben-Jaffel L., 2007, ApJ, 671, L61 Ben-Jaffel L., 2008, ApJ, 688, 1352 Ben-Jaffel L., Sona Hosseini S., 2010, ApJ, 709, 1284 Bisikalo D., Kaygorodov P., Ionov D., Shematovich V., Lammer H., Fossati L., 2013, ApJ, 764, 19 Bourrier V., Ehrenreich D., Lecavelier des Etangs A., 2015, A&A, 582, A65 Bourrier V., Lecavelier des Etangs A., Ehrenreich D., Tanaka Y. A., Vidotto A. A., 2016, A&A, 591, A121 Carroll-Nellenback J. J., Shroyer B., Frank A., Ding C., 2013, Journal of Computational Physics, 236, 461 Carroll-Nellenback J., Frank A., Liu B., Quillen A. C., Blackman E. G., Dobbs-Dixon I., 2017, MNRAS, 466, 2458 Charbonneau D., Brown T. M., Noyes R. W., Gilliland R. L., 2002, ApJ, 568, 377 Christie D., Arras P., Li Z., 2016, ApJ, 820, 3 Cohen O., Kashyap V. L., Drake J. J., Sokolov I. V., Garraffo C., Gombosi T. I., 2011, ApJ, 733, 67 Cunningham A. J., Frank A., Varni`ere P., Mitran S., Jones T. W., 2009, ApJS, 182, 519 Debrecht A., Carroll-Nellenback J., Frank A., Fossati L., Black- man E. G., Dobbs-Dixon I., 2018, MNRAS, 478, 2592 Ehrenreich D., et al., 2015, Nature, 522, 459 Fossati L., et al., 2010, ApJ, 714, L222 Frank A., Lui B., Carroll-Nellenback J., Quillen A. C., Black- man E. G., Kasting J., Dobbs-Dixon I., 2016, in Kastner J. H., Stelzer B., Metchev S. A., eds, IAU Symposium Vol. 314, Young Stars & Planets Near the Sun. pp 237 -- 240 (arXiv:1508.00880), doi:10.1017/S1743921315006675 Garc´ıa Munoz A., 2007, Planet. Space Sci., 55, 1426 Haswell C. A., et al., 2012, ApJ, 760, 79 Hunten D. M., 1982, Planet. Space Sci., 30, 773 Krumholz M. R., Stone J. M., Gardiner T. A., 2007, ApJ, 671, 518 Kulow J. R., France K., Linsky J., Parke Loyd R. O., 2014, ApJ, 786, 132 Lecavelier des Etangs A., et al., 2012, A&A, 543, L4 Matsakos T., Uribe A., Konigl A., 2015, A&A, 578, A6 McCann J., Murray-Clay R. A., Kratter K., Krumholz M. R., in press Murray-Clay R. A., Chiang E. I., Murray N., 2009, ApJ, 693, 23 Oklopci´c A., Hirata C., 2018, ApJ, 855, L11 Owen J. E., Adams F. C., 2014, MNRAS, 444, 3761 Owen J. E., Alvarez M. A., 2016, ApJ, 816, 34 Schneiter E. M., Vel´azquez P. F., Esquivel A., Raga A. C., Blanco- Cano X., 2007, ApJ, 671, L57 Schneiter E. M., Esquivel A., D'Angelo C. S. V., Vel´azquez P. F., Raga A. C., Costa A., 2016, MNRAS, 457, 1666 Sternberg A., Hoffmann T. L., Pauldrach A. W. A., 2003, ApJ, 599, 1333 Stone J. M., Proga D., 2009, ApJ, 694, 205 Teyssandier J., Owen J. E., Adams F. C., Quillen A. C., 2015, MNRAS, 452, 1743 Towns J., Cockerill T., Dahan M., Foster I., 2014, Computing in Science and Engineering, 16, 62 Tremblin P., Chiang E., 2013, MNRAS, 428, 2565 Tripathi A., Kratter A., Murray-Clay R., Krumholz M., 2015, ApJ, 808, 173 Vidal-Madjar A., Lecavelier des Etangs A., D´esert J.-M., Ballester G. E., Ferlet R., H´ebrard G., Mayor M., 2003, Nature, 422, 143 Vidal-Madjar A., et al., 2004, ApJ, 604, L69 Vidal-Madjar A., et al., 2013, A&A, 560, A54 Villarreal D'Angelo C., Esquivel A., Schneiter M., Sgr´o M. A., 2018, MNRAS, 479, 3115 Woods T. N., Rottman G. J., Bailey S. M., Solomon S. C., 1998, Sol. Phys., 177, 133 This paper has been typeset from a TEX/LATEX file prepared by the author. MNRAS 000, 1 -- 16 (0000)
1804.06859	2	1804	2018-11-25T20:50:26	Shepherding in a Self-Gravitating Disk of Trans-Neptunian Objects	[ "astro-ph.EP" ]	A relatively massive and moderately eccentric disk of trans-Neptunian objects (TNOs) can effectively counteract apse precession induced by the outer planets, and in the process shepherd highly eccentric members of its population into nearly-stationary configurations which are anti-aligned with the disk itself. We were sufficiently intrigued by this remarkable feature to embark on an extensive exploration of the full spatial dynamics sustained by the combined action of giant planets and a massive trans-Neptunian debris disk. In the process, we identified ranges of disk mass, eccentricity and precession rate which allow apse-clustered populations that faithfully reproduce key orbital properties of the much discussed TNO population. The shepherding disk hypothesis is to be sure complementary to any potential ninth member of the Solar System pantheon, and could obviate the need for it altogether. We discuss its essential ingredients in the context of Solar System formation and evolution, and argue for their naturalness in view of the growing body of observational and theoretical knowledge about self-gravitating disks around massive bodies, extra-solar debris disks included.	astro-ph.EP	astro-ph	Draft version November 27, 2018 Typeset using LATEX twocolumn style in AASTeX62 8 1 0 2 v o N 5 2 . ] P E h p - o r t s a [ 2 v 9 5 8 6 0 . 4 0 8 1 : v i X r a SHEPHERDING IN A SELF-GRAVITATING DISK OF TRANS-NEPTUNIAN OBJECTS Antranik A. Sefilian1 and Jihad R. Touma2 1Department of Applied Mathematics and Theoretical Physics, University of Cambridge, Centre for Mathematical Sciences, Wilberforce Road, Cambridge CB3 0WA, UK 2Department of Physics, American University of Beirut, PO BOX 11-0236, Riad El-Solh, Beirut 11097 2020, Lebanon (Received -- ; Revised -- ; Accepted -- ) Submitted to AJ ABSTRACT A relatively massive and moderately eccentric disk of trans-Neptunian objects (TNOs) can effectively counteract apse precession induced by the outer planets, and in the process shepherd highly eccentric members of its population into nearly-stationary configurations which are anti-aligned with the disk itself. We were sufficiently intrigued by this remarkable feature to embark on an extensive exploration of the full spatial dynamics sustained by the combined action of giant planets and a massive trans- Neptunian debris disk. In the process, we identified ranges of disk mass, eccentricity and precession rate which allow apse-clustered populations that faithfully reproduce key orbital properties of the much discussed TNO population. The shepherding disk hypothesis is to be sure complementary to any potential ninth member of the Solar System pantheon, and could obviate the need for it altogether. We discuss its essential ingredients in the context of Solar System formation and evolution, and argue for their naturalness in view of the growing body of observational and theoretical knowledge about self-gravitating disks around massive bodies, extra-solar debris disks included. Keywords: celestial mechanics -- Kuiper belt: general -- planets and satellites: dynamical evolution and stability 1. INTRODUCTION Trans-Neptunian (phase)-space appears to be popu- lated with bodies that show signs of orbital sculpting, then shepherding. With the discovery of 2012 VP113, a Sedna-like object, Trujillo & Sheppard (2014) first ar- gued for a ninth planet of 5 M⊕ on a circular orbit at 200 AU as a potential shepherd of several TNOs with ec- centric and inclined orbits showing peculiar clustering in the argument of periapse. Later on, Batygin & Brown (2016) noted remarkable spatial nodal alignment of the same objects. They reexamined the proposition of an additional planet, and argued instead for a super-Earth (dubbed "Planet Nine") on a larger eccentric and in- clined orbit, while appealing to an alternative resonant [email protected] [email protected] process for the aligning trap (Batygin & Brown 2016). Further indirect evidence for such a planet was sought around apparent deviations in the orbit of the Cassini spacecraft (Fienga et al. 2016), and in the potential to explain the Sun's obliquity (Bailey et al. 2016; Lai 2016). To date, twenty-three trans-Neptunian objects (TNOs) have been identified on eccentric and inclined orbits, with semi-major axes ap % 150 AU and perihelion dis- tance qp > 30 AU. Out of these, thirteen roam with ap % 250AU and have had their notorious kinematic properties classified in the course of Planet-Nine related studies which propose to explain them. They are in- terpreted as either: spatially clustered and anti-aligned with Planet Nine (ten objects); spatially clustered and aligned with Planet Nine (two objects); neither here nor there, though strongly perturbed by Planet Nine (one object). These classes are of course expected to grow in size and definition by proponents of a ninth planet 2 Sefilian & Touma which is requested to structure, along with the rest of the Solar system, the phase space in which TNOs are presumed to evolve. Alternatively, Shankman et al. (2017a) argued that the spatial clustering which Planet Nine is supposed to explain is fraught with observational bias. Running their own orbital simulations, they disputed the claim that a planet alone could maintain clustering for the required duration. They further noted that to observe this group of TNOs within existing campaigns, implies a parent population of 6− 24 M⊕. Such a massive reser- voir of trans-Neptunian icy bodies is nearly two orders of magnitude larger than currently favored estimates (Gladman et al. 2011). Shankman et al. (2017a) took this requirement as further evidence against significant clustering, and gave no further consideration to the dy- namical signature of a massive trans-Neptunian popula- tion. Here, we go precisely after the dynamical impact of an extended and relatively massive disk of trans-Neptunian objects, and demonstrate that it alone can provide a fair amount of shepherding, perhaps obviating the need for an extra planetary member in the Solar System pan- theon, surely complementing it. We describe results in a progression of complexity around a fiducial razor thin disk. We then comment briefly on parametric variations on such a disk, discuss its properties and their origin, together with the poten- tial interplay between the dynamical features it stim- ulates, and those associated with a hypothetical few Earth mass planet in post-Neptunian realm. 2. COPLANAR DYNAMICS We study the secular, orbit-averaged coplanar dynam- ics of trans-Neptunian test particles characterized by their semi-major axis ap, eccentricity ep and apsidal an- gle p, in the combined gravitational potential of: a- the outer planets and b- a hypothetical extended disk, lying in the plane of the giant planets, and built out of confocal eccentric apse-aligned orbits. The outer planets are included via the quadrupo- lar potential of a sequence of fixed concentric circular rings. The coplanar disk is parametrized by its non- axisymmetric surface density Σ (Eq. A1), eccentricity profile ed, global apsidal angle d (fixed at π, except otherwise stated), and inner and outer boundaries ain and aout respectively. We work with disks sity/eccentricity profiles that have power-law den- ad (cid:19)p Σd(ad) = Σ0(cid:18) aout (1) and ad (cid:19)q ed(ad) = e0(cid:18) aout (2) for ain ≤ ad ≤ aout. Here, Σ0 and e0 are the pericentric surface density and eccentricity at the outer edge of the disk respectively. Surface density profiles with p < 2 (p > 2) are associated with disks which have more mass concentrated in the outer (inner) parts of the disk than in the inner (outer) regions. Total disk mass Md can be Σd(ad)addad yielding estimated with Md ≃ 2πR aout out(cid:20)1 −(cid:18) ain Md = Σ0a2 ain 2π 2 − p aout(cid:19)2−p(cid:21) ≈ 2π 2 − p Σ0a2 out (3) where the approximation is valid as long as the disk edges are well-separated and more mass is found in the outer parts. Disk models which were thoroughly ex- plored in this work are listed in Table 1. Table 1. Power-law disk models. Disk Model p q DM1 DM2 DM3 DM4 0.5 0.5 0.5 2.5 -1 -1 -1 -1 ain (AU) 40 40 40 40 π aout d (AU) (rd) 750 750 750 750 π π π e0 Md (M⊕) 0.165 0.165 0.165 0.165 10 2.5 20 10 Note -- Disk Model 1 (DM1) is the fiducial disk configuration adopted in this work. Our basic shepherding mechanism is best articulated in planar dynamics which will ultimately provide the skeleton around which fully inclined behavior is struc- tured (see Section 3). At the outset it is important to remind the reader that hot nearly-Keplerian disks induce negative apse preces- sion in their constitutive particles, in contrast to the fa- miliar prograde apse precession expected from cold disks of isolated planets. This fact was recently noted to argue for the role of massive gaseous disks in mitigating the destructive role of perturbations induced on planetes- imal disks by wide binary companions (Rafikov 2013; Silsbee & Rafikov 2015; Rafikov & Silsbee 2015; Sefilian 2017). We exploit that feature here by appealing to the neg- ative precession induced by an extended and massive trans-Neptunian debris disk to mitigate against and, if possible, freeze the prograde differential precession in- duced by the outer planets on a distinguished population of TNOs which is yet to be identified. Shepherding of trans-Neptunian objects With this in mind, we recover the secular orbit- averaged disturbing potential, Rd, of power law disks up to fourth order in the orbital eccentricity of a copla- nar test particle (see Appendix A): Rd = K(cid:20)ψ1ep cos ∆ + ψ2e2 p(cid:21), p cos ∆ + ψ5e4 + ψ4e3 p + ψ3e2 p cos(2∆) (4) -0.5 1 0.5 0 3 ep = 0.1 ep = 0.2 ep = 0.3 ep = 0.4 ep = 0.5 ep = 0.6 ep = 0.7 ep = 0.8 ep = 0.9 where outa1−p p > 0, K ≡ πGΣ0ap ∆ ≡ p − d. (5a) (5b) The dimensionless coefficients ψi are given by equations A9 -- A14. Orbit averaged quadrupolar action of the outer plan- ets is captured via with Rp = + 1 3 Γ(1 − e2 p)− 3 2 , Γ = 3 4 GM⊙ ap 4 Xi=1 mia2 i M⊙a2 p , (6) (7) and [(mi, ai), i = 1..4] the masses and current semi ma- jor axes of the four giant planets. Combining both contributions, Hamilton's equations p and for the signed "angular" momentum lp = ±q1 − e2 the conjugate longitude of the apse p are given by: Lp lp =−K(cid:20) sin ∆(cid:2)ψ1q1 − l2 p) sin(2∆)(cid:21), + 2ψ3(1 − l2 p + ψ4(1 − l2 p) 3 2(cid:3) (8) and Lp p = + 2ψ2 + 2ψ3 cos(2∆) Γ l4 p + Klp(cid:20) ψ1 cos ∆ q1 − l2 + 3ψ4 cos ∆q1 − l2 p p + 4ψ5(1 − l2 p)(cid:21), (9) with Lp = pGM⊙ap, the constant angular momentum conjugate to the mean anomaly which has been averaged out of the game. Disturbing functions [Eqs.(4, 6)], and equations of motion [Eqs.(8, 9)] govern the dynamics of both prograde and retrograde orbits which are coplanar with disk and planets. Below, and in keeping with ob- served aligned TNOs, we concentrate primarily on the prograde phase space. ep -1 150 200 250 a p 300 (AU) 350 400 450 500 Figure 1. Rate of apse precession p of TNOs which are anti-aligned with the disk (∆ = π), over a range of semi- major axis ap for different values of eccentricity ep. Preces- sion is here driven by the combined action of the giant planets and the fiducial disk model DM1 (see Table 1). Zero apse precession obtains for all the considered values of ep, and semi-major axes ap between 150 and 500 AU. Given that the torque vanishes for ∆ = π (see Eq. 8), we have here a family of stationary orbits, which are anti-aligned with the disk, and whose eccentricity grows with ap. In Figure 1, we display the apsidal precession rate in- duced by the outer planets and the fiducial power law disk model (Table 1, model DM1) on orbits which are anti-aligned with the disk's spatial orientation (i.e or- bits with ∆ = π), over a range of semi-major axis ap, and for different values of TNO eccentricity ep. Evi- dently, there is an eccentric anti-aligned orbit with zero net apse precession at all semi-major axes in the con- sidered range. Keeping in mind that the torque (Eq. 8) is null for ∆ = π, we have here evidence for a one parameter family of anti-aligned stationary orbits which will provide the skeletal structure around which the ob- served TNOs, and the rest of our paper will be fleshed out! Before we examine the full dynamical behavior of this family, we thought it reasonable to probe the robustness of this remarkable broad-ranged cancellation of apse- precession to variations in disk properties (mass den- sity profile, disk eccentricity, disk radial extent). We thus computed the disk mass Md which is required to apse-freeze an anti-aligned orbit (∆ = π) of given ec- centricity ep and semi-major axis ap when embedded in a disk of given mass distribution (dictated by p), inner and outer edge, and e0. The outcome of this exercise for a test particle with ap = 257 AU and ep = 0.82 is shown in Figure 2, and permits the following conclusions: a- the required disk mass can be as low as ∼ 1M⊕ and as high as ∼ 30M⊕; b- lower Md is required at higher disk eccentricity, an effect which is surely due to en- 4 101 100 0 0.2 0.4 0.6 e 0 0.8 M1 M2 M3 M4 M5 M6 M7 M8 M9 Figure 2. The required disk mass Md as a function of disk eccentricity (q = 0, ed(ap) = e0) to obtain stationary anti-aligned (∆ = π) TNO orbits at ap = 257 AU with ep = 0.82. The calculation is performed for disk parameters (p, ain and aout) given in Table 2 . When combined with re- sults shown in Figure 1, this figure speaks for the robustness of our proposed mechanism. Table 2. Power-law disks used to generate Figure 2. Model ain(AU) aout(AU) 200 1200 200 1000 200 800 M1 M2 M3 M4 M5 M6 M7 M8 M9 p 0.1 0.5 0.9 0.1 0.5 0.9 0.1 0.5 0.9 Note -- We have adopted a constant disc eccentricity by setting q = 0, with 0 ≤ ed(ap) = e0 - 0.90. hancement of disk induced retrograde precession with increasing e0; c- the critical Md increases with increas- ing aout: this behavior is evident in axisymmetric disks where the disk induced precession is well approximated by the following expression1 p(cid:12)(cid:12)disk ≃ −4.2× 10−10yr−1 Md 1M⊕ 103AU aout (cid:18) ap 500AU(cid:19)−0.5 (10) (cid:12) p 1 The approximate expression of (cid:12)disk is obtained for cir- cular disks with p = 1 under the reasonable assumption of ain ≪ ap ≪ aout. This assumption allows us to drop contribu- tions to Rd (Eq. 4) from the disk edges rendering the coefficients ψi mild functions of only p and q (see Eq. A15-A19). For instance, in a circular disk ψ2 = −0.5 for p = 1; see Eq. A16. Sefilian & Touma for circular TNO orbits. What is evident for axisymmet- ric disks is clearly maintained in eccentric ones. Further- more, we checked that our conclusions for a single anti- aligned equilibrium orbit (with ep = 0.82 and ap = 257 AU) holds for all. Exhaustive exploration of the dynamics sustained by our orbit averaged Hamiltonian shows that the fiducial disk model (Table 1, DM1) harbours three distinct fam- ilies of orbits: • A family of stable, highly eccentric, and anti- aligned orbits (∆ = π): this family shows equi- librium ep(ap)-behavior which is remarkably con- sistent with the trend followed by clustered TNOs. It is the family of most interest to us in relation to the shepherding phenomenon. • A family of stable aligned (∆ = 0) and low ec- centricity orbits: interestingly enough, this family follows in its trend the eccentricity distribution of the disk that hosts it. • A family of highly eccentric and aligned orbits (∆ = 0): this family parallels the behavior of the stable high ep anti-aligned family but is doomed to instability. Taking it for granted that the stable anti-aligned family correlates with the observed family of clustered TNOs, we conclude that DM1 naturally excludes stable high ec- centricity clustering in the opposite apse orientation, an orientation where significant high eccentricity clustering is apparently not observed2. All three families are shown in Figure 3 together with the eccentricity distribution of the underlying disk. These families can be further situated within the global phase space structure which is captured in Fig. 4 at three distinct semi-major axes. In addition to equilibria and their bifurcations, the phase diagrams re- veal aligned and anti-aligned islands (AI and A-AI re- spectively) of bounded motion around the parent sta- ble equilibrium orbits. These islands host orbits which will show signs of clustering (aligned and anti-aligned respectively) when considered collectively, and in time. The A-AI shelters high eccentricity orbits which strad- dle, as they oscillate in their eccentricity and longitude of apse, the parent equilibrium family which so closely follows the ep − ap trend of the observed TNOs. We 2 We know of two highly eccentric TNOs [2013 FT28 (Sheppard & Trujillo 2016) and 2015 KG163 (Shankman et al. 2017b)] having, ap > 250 AU and qp > 30 AU, and which are currently anti-aligned with the much discussed clustered bunch. Their dynamical behavior is reviewed in Section 5. Shepherding of trans-Neptunian objects 5 , stable = = 0, unstable = 0, stable ed(ap) Observed TNOs 1 0.8 0.6 0.4 0.2 p e ed (ap) 0 50 100 150 200 250 300 350 400 450 500 (AU) a p Figure 3. The stationary TNO families (ep, ∆), both sta- ble and unstable, which are sustained by DM1 (Table 1) act- ing together with the giant planets. The stable anti-aligned (∆ = π) family follows quite closely the observed ep − ap trend of the seven clustered TNOs which are considered in this study (Table 3). A stable family of non-precessing orbits which are aligned with the disk (∆ = 0) has an eccentricity profile which is almost identical to the imposed disk eccen- tricity profile ed(ap). will have more to say about this population when we discuss it in full 3D glory below. The AI on the other hand is populated by orbits which share, on average, the orientation and orbital eccentricity of the host disk, thus providing a rich supply of orbits with which to con- struct a self-consistent deformation of DM1 (an exercise on which we comment in Section 5). It further includes orbits which have large amplitude eccentricity variations that bring them close to the unstable high eccentricity aligned orbits. Such orbits tend to linger around that unstable aligned configuration, projecting a temporary sense of eccentric alignment with the disk which is then lost to evolution on timescales which are long enough3. The AI and A-AI are both surrounded by high ec- centricity orbits which circulate in the longitude of the apse, while maintaining large and near constant eccen- tricity. More on these populations when we discuss cu- rious members of the TNO population in Section 5. In sum, DM1 shepherds eccentric anti-aligned orbits (∆ = π) whose properties favor them as coplanar analogs of the family identified by Trujillo & Sheppard (2014), while at the same time supporting aligned and non-precessing orbits of moderate eccentricity which promise to reproduce the disk that supports them, in 3 Similar such orbits which tend to linger around eccentric aligned orientation can be found in the A-AI, when one moves suf- ficiently far from the equilibrium. These orbits are fairly eccentric, with some suffering encounters with Neptune on their journey. a self-consistent treatment of the dynamics4. It would thus seem that a massive eccentric trans-Neptunian de- bris disk together with the action of the outer planets provides significant and profoundly suggestive clustering of embedded test particles. Whether such a disk obvi- ates the need for a Planet Nine-like perturber altogether will be discussed further below after we explore out-of- plane dynamics, close to where the observed TNOs tend to roam. However, what is already clear at this coplanar stage is that the action of such a potential disk (which is evidently felt by highly eccentric orbits for disks with mass as low as ∼ 1 M⊕) cannot be ignored, and will have to be considered together with any putative extra planet. 3. LIFE OUTSIDE THE PLANE: GOING 3D Freezing coplanar orbits is interesting enough. How- ever, the observed TNO bunch is held together in in- clined orbits. Can we say anything about inclina- tions? No hurdle in principle to generalizing the pro- posed mechanism to inclined orbits. An attempt to work it out with our orbit-averaged treatment of a razor thin disk potential faces an insurmountable singularity (Heppenheimer 1980). Disk height comes in for rescue but then expressions become arbitrary without a specific prescription for vertical disk structure (Hahn 2003). A fix is to use a local approximation in the averages which regularizes expressions (Ward 1981) and allows us to ex- plore eccentricity-inclination dynamics for both axisym- metric and eccentric disks. But we went further. Starting with the mass and eccentricity distribu- tions in DM1, we computed the full 3D potential and recovered associated spherical harmonics numerically (Kazandjian, Sefilian & Touma, in preparation). Then we orbit-averaged spherical harmonics (again numer- ically), and obtained closed form expressions for any given semi-major axis, and to the desired (arbitrary) order in eccentricity and inclination. A brief expla- nation of the steps involved is provided in Appendix B. With the orbit-averaged mean field of the razor thin eccentric disk in hand (Eq. B23), we added the secular contribution of the outer planets (Eq. B25) to study the 4 The coplanar dynamics we just mapped out is naturally in dialog with Beust (2016) who considers secular dynamics which is controlled by the outer planets together with a coplanar, eccen- tric, Planet Nine like object. He identifies eccentric non-aligned secular equilibria which are analogous to the ones we recover here. Perhaps, the analogy can be pushed further to argue for the com- bined action of a pre-existing disk and a scattered planet. We discuss this possibility, but do not explore its detailed workings in the present article. 6 Sefilian & Touma 0.4 0.2 0 -0.2 AI 0.4 0.2 0 -0.2 -0.4 A-AI -0.4 -0.2 0 0.2 0.4 -0.4 A-AI -0.4 -0.2 0 0.2 0.4 0.8 0.4 0 -0.4 -0.8 AI A-AI -0.8 -0.4 0 0.4 0.8 (a) ap = 198 AU (b) ap = 207 AU (c) ap = 453 AU Figure 4. Phase portraits corresponding to the total Hamiltonian, Hp + Hd = −(Rp + Rd), in ep(sin ∆, cos ∆) space, at three different TNO semi-major axes ap for the fiducial disk model DM1. The semi-major axes were chosen to illustrate the existence and bifurcation of the families identified in Figure 3. The stable (unstable) secular equilibria are highlighted in red (blue). Panel a shows the phase portrait at ap = 198 AU with a single stable anti-aligned equilibrium and its associated Anti-Aligned Island (A-AI) situated at ∆ = π. In Panels b and c, we show the phase portrait at ap = 207 AU and ap = 453 AU respectively, with two new aligned equilibria (∆ = 0), one unstable and one stable, the latter coming with an Aligned Island (AI) of librating orbits. The ep − ap trends of Fig. 3 are evident with the progression in semi-major axis, through panels a, b and c respectively. coupled eccentricity-inclination dynamics of a particle in a perfectly smooth fashion. With the help of these expansions, we could study off-plane dynamics of TNOs which are clustered in the plane, determine stability to small inclinations, as well as the long term evolution of populations of initially clustered and inclined objects. A brief report on global dynamics follows: quickly silenced by the realization that both inner quadrupole and eccentric disk induce retrograde nodal precession. While varying with location in and/or inclination to the disk, the reinforced ret- rograde precession excludes the possibility of apse aligned orbits that further share the same spatial orientation. • As evident in Fig. 15, planar phase space structure (including families of equilibria, their stability, and their behavior as a function of semi-major axis) is recovered quite accurately within this generalized formalism; • An involved linear stability analysis confirmed that families of stable planar eccentric equilibria (both aligned and anti-aligned with the disk's ap- sidal line) are further stable to small perturbations in inclination: this is quite encouraging because it suggests that the flock of stationary orbits that were identified in the plane is maintained when subject to small out-of-plane perturbations. • Small-amplitude variations in the inclinations, ec- centricities, and longitude of apse around sta- ble coplanar equilibria were numerically shown to maintain near alignment in the longitude of apse, all the while the argument of apse and longitude of node circulate. • Moving to large amplitude variations in inclina- tion: Any temptation to inquire about fixed anti- aligned, eccentric and sufficiently inclined orbits is We could of course proceed to provide a complete classi- fication of orbital dynamics in the combined field of disk and planets. This is a two degree of freedom problem which is amenable to description in terms of Poincar´e sections at any given semi-major axis. We think that such an exercise is best relegated to a separate purely dynamical treatment. Instead, we opt to follow popula- tions of judiciously chosen particles over the underlying complex phase space, with a view to characterizing the extent to which our setup can reproduce observed met- rics. 3.1. Populations over Phase Space The reference disk, together with the outer planets, sustains two families of stable coplanar equilibria, one aligned (∆ = 0) and of low eccentricity, the other anti- aligned (∆ = π) and of large eccentricity. Anti-aligned equilibria follow the observed trend of eccentricity with semi-major axis (see Fig. 3). It is natural to ask what remains of this trend when vertical heating is included, and when eccentricity-inclination dynamics kicks in. We spoke of linear stability of planar equilibria to slight in- clination change. We further reported results of numer- Shepherding of trans-Neptunian objects and specific angular momentum h = qGM⊙ap(1 − e2 p)  sin ip sin Ωp − sin ip cos Ωp cos ip   7 (12) Table 3. Heliocentric semi major axis (ap) , perihelion dis- tance (qp), inclination (ip), argument of perihelion (ωp), lon- gitude of ascending node (Ωp), and longitude of perihelion (p) of the clustered TNOs with ap > 250AU and qp > 30AU considered in this study. Data obtained from the Minor Planet Center. TNO ap qp (AU) (AU) 2012 VP113 2014 SR349 2004 VN112 2013 RF98 2010 GB174 2007 TG422 Sedna 260.8 289.0 317.7 350.0 369.7 483.5 499.4 80.3 47.6 47.3 36.1 48.8 35.6 76.0 ip (◦) 24.1 18.0 25.6 29.6 21.5 18.6 11.9 ωp (◦) 292.8 341.4 327.1 311.8 347.8 285.7 311.5 Ωp (◦) 90.8 34.8 66.0 67.6 130.6 112.9 144.5 p (◦) 23.6 376.2 33.1 379.4 118.4 38.6 96.0 ical simulations showing that the long term evolution of perturbed planar orbits, shows stability in inclinations for small enough inclination, while maintaining confine- ment in the longitude of the apse. That would suggest that populations of particles initiated around the islands of planar stability would maintain the planar alignment, though not immediately clear for how long in the pres- ence of nonlinearities. With this in mind, we explored the dynamics of popu- lations of particles in the combined orbit-averaged grav- itational field of DM1 and the outer planets over the age of the Solar System. Particles were initiated around the AI and A-AI of stable planar equilibria (see Fig. 4) at the semi-major axis locations of seven of the clustered objects listed in Table 3 5. Islands of stability were sam- pled uniformly in eccentricity, with inclinations assigned uniformly in a 10◦ range. The argument of pericenter ωp and the longitude of ascending node Ωp were picked to guarantee uniform ∆ sampling in the range 180◦± 20◦ for the anti-aligned family, and ±20◦ for the aligned fam- ily. This way, we end up with 300 particles at each of the seven observed semi-major axis, and follow their orbits over the age of the Solar system. We characterize an orbit's orientation and eccentricity with the Lenz ep = ep cos wp cos Ωp − cos ip sin wp sin Ωp cos wp sin Ωp + cos ip sin wp cos Ωp  sin ip sin wp   = epe (11) vectors. The Lenz vector lives in the plane of a parti- cle's orbit and points to its periapse; the angular mo- mentum vector is perpendicular to the orbital plane, and its dynamics encodes nutation and precession of the orbit. Orbits can have aligned Lenz vectors while being spread out in node and inclination. Orbits can be spread out in Lenz vector while sharing the same or- bital plane. In other words, behavior of both vectors is required for a complete characterization of the degree of spatial alignement of a population of orbits, with the following metrics being particularly useful in that regard (Millholland et al. 2017): • The departure of Lenz vector orientation from the mean as captured by S(t) = 7 Xi=1 em(t).ei(t), (13) where [ei(t), i = 1, . . . , 7] are unit Lenz vectors and em(t) is their mean unit vector. This definition of S(t) allows us to quantify the degree to which the Lenz vectors are clustered about their mean. Specifically, if the Lenz vectors of the seven ob- jects coincide with their mean at a given time, then S(t) = 7. • A measure of the anti-alignment with respect to the trans-Neptunian disk as given by A(t) = em(t).ed, (14) where the disk orientation is fixed such that ed = (cos d, sin d, 0)(cid:12)(cid:12)d=π unless otherwise stated. A measure of A(t) = −1 (+1) corresponds to configurations where the mean Lenz unit vector of the seven bodies is perfectly anti-aligned (aligned) with that of the disk at a given time. • A measure of clustering in p, wp and Ωp sepa- rately as provided by the mean over unit vectors; rp , rwp and rΩp , that circulate with these an- gles respectively. When the associated vectors are homogeneously distributed on a circle they have zero mean, while perfect alignment results in their mean having unit length. 5 Performing the study at all positions in the disk, and not only at the seven considered semi-major axes (Table 3), does not affect the conclusions drawn from the population studies around AI and A-AI. The objects of interest to us listed in Table 3 yield a current value of S ≈ 4.57 and A ≈ −0.77, as- suming a disk whose apsidal angle is 180◦ away from 8 0.18 0.16 0.14 p e 0.12 0.1 0.08 °) ( 2 0 -2 250 Eccentricity AI Inclination Sefilian & Touma 0.4 0.3 0.2 0.1 °) ( 0 , p i - p i 300 350 400 450 500 ap (AU) Figure 5. The average behavior of eccentricity, inclination and longitude of periapse, and excursions thereabout, for a flock of objects initiated in the neighborhood of the AI. The top panel shows particles at all semi-major axes executing slight excursions in eccentricity and inclinations: values of ep remain close to the planar equilibria while inclinations oscillate around the initial conditions ip,0. The bottom panel shows how all these particles maintain alignment with the disk with ∆ ≈ 0◦, with negligible spread, over the relevant range in ap. the mean of the clustered inclined bunch. Moreover, the measures of r for the group of clustered objects are: rωp = 0.93, rΩp = 0.81 and rp = 0.80, in- dicating confinement in both ωp and Ωp as noted by Batygin & Brown (2016). Based on extensive orbit integrations of our samples, we learned the following: • Particles initiated around the AI (Fig. 4 [b, c]) stay tightly bunched while showing small ampli- tude variations in their inclination, eccentricity and longitude of apse (see Fig. 5). Indeed, we find a time-averaged value of rp = 0.989 ± 0.011 indicating strong p-confinement which is main- tained by the opposite circulation of ωp and Ωp as reflected in rωp = 0.041 ± 0.009 and rΩp = In other words, the orbit struc- 0.040 ± 0.003. ture that is expected to self-consistently reproduce the planar disk is stable enough to inclined mo- tion to hold the promise of sustaining a thick ver- In Fig. 6, we show ensemble sion of that disk. averaged behavior of A(t) and S(t) which sup- ports the conclusion above, with the time-averaged S ∼ 6.85 and A ∼ 0.98 indicating Lenz vector confinement, together with disk alignment in the neighborhood of the AI. • Populations of particles initiated around the A- AI (Fig. 4b) show greater complexity as a func- tion of semi-major axis and inclination. Particles Figure 6. Time evolution of S and A for objects initially near the AI. The calculation is based on an ensemble of 10 sets of particles, each set consisting of 7 particles randomly picked within the AI at each of the considered semi-major axes. The thick lines represent the ensemble averaged val- ues and the shaded regions enclose the spread around the average. with semi-major axes 250 (cid:22) ap (cid:22) 350 AU librate around the planar island with slight excursion in inclination and eccentricity for all initial inclina- tions (see Fig. 7): linear stability translates into longterm stability in this case, with low inclina- tion orbits maintaining Lenz vector alignment all the while displaying a spread in node and peri- apse. Indeed, our simulations show that around 93% of all the considered orbits with ap (cid:22) 350AU are strongly confined in p with rp > 0.9 while the node and periapse circulate (rωp (cid:22) 0.05 and rΩp (cid:22) 0.03). On the other hand, particles with ap % 350 AU show long term behavior which depends on the initial inclination: i. Particles launched with ip - 4 − 5◦ are stable to off-plane motion, similar to the objects at ap (cid:22) 350AU ; ii. In contrast, particles launched with ip > 5◦ show large amplitude variations in ep and ip, with in- clinations growing somewhat erratically to values higher than 20◦, all the while ep evolves to smaller values. Such particles still show significant clus- tering in p with rp ∼ 0.6, though weaker than what is observed with stably inclined populations. As evident in Fig. 7 and Fig. 8, particles initiated with low inclinations (ip < 5◦) around the A-AI remain clus- tered and anti-aligned with the disk, with small ampli- tude variations in their eccentricities and inclinations. Their stability guarantees the survival of a puffed up version of the backbone of coplanar anti-aligned equilib- ria (∆ = π) whose ep − ap behavior is consistent with the observed TNOs. Shepherding of trans-Neptunian objects Eccentricity A-AI Inclination 0.9 0.8 0.7 0.6 0.5 0.4 p e ) ( 185 180 175 2 1.5 1 0.5 0 ) ( 0 p , i - p i 250 300 350 400 450 500 ap (AU) Figure 7. Average values of eccentricity, inclination and longitude of periapse, and spread thereabout, for objects ini- tiated near the A-AI with initial inclinations ip,0 < 5◦. It is evident that slightly inclined particles initiated around the A-AI exhibit an average behavior of ep − ap consistent with the planar equilibrium profile. At the same time, as shown in the bottom panel, the orbits remain clustered in and anti-aligned with the disk, with ∆ ∼ 180◦ at all ap. Clones of observed TNOs 1 0.8 p e 0.6 0.4 ) ( 220 180 140 9 ) ( p i 50 40 30 20 10 Simulated ep Observed ep Simulated ip Observed ip 250 300 350 400 ap (AU) 450 500 Figure 9. The average eccentricity, inclination, and longi- tude of apse along with their time-averaged spread for all the "successful clones" in our simulations. The top panel reveals reasonable agreement between observed and average values of both ep and ip. The bottom panel indicates that clustering and anti-alignment with the proposed disk (DM1) is maintained at all considered values of ap. Figure 8. Time evolution of the ensemble averaged S and A (thick-lines) and the spread thereabout (shaded regions) for objects sampled near the A-AI, with ip(t = 0) < 5◦. Clus- tering of Lenz vectors is maintained at all times (S(t) ∼ 6.5) together with disk anti-alignement (A(t) ∼ −0.96). 3.2. Clones of Observed TNOs We probe the orbital evolution of "clones" of ob- served TNOs (Table 3) over the age of the Solar Sys- tem. At each of the considered semi-major axes, we build samples of 300 particles with orbital elements ran- domly picked in the neighborhood of the observed ones (δe = 1%eobs and δi = δω = δΩ = 5◦). The disk is again coplanar with the outer planets and anti-aligned with the mean apse direction of the clustered TNOs. Figure 10. Evolution of both S and A as a function of time for clones of the observed TNOs. The calculation is done with ten ensembles of seven particles each, randomly picked from the population of SCs at each of the seven con- sidered values of ap. Thick lines represent the mean, and the shaded regions the spread about that mean. S(t) and A(t) oscillate around their initial values indicating p- confinement which, on average, is 180◦ away from the fixed disk apsidal line. We find that more than 60% of the clones maintain a perihelion distance which is larger than the orbital radius of Neptune at all times. We dub these objects "successful clones" (SCs for short) and analyze their or- bital evolution to conclude that: • SCs follow quite closely the eccentricity and incli- nation of their progenitors (see Fig. 9); 10 Sefilian & Touma • SCs, on average, maintain anti-alignment with the disk apsidal line, while showing slight oscillations in the longitude of apse around the mean; see Fig. 9. Considering all successful simulations we find rp ≈ 0.785 which compares well with that of the observed bunch; rp = 0.80; • SCs show no confinement in Ωp and ωp, with rΩp = 0.020 ± 0.006 and rωp = 0.033 ± 0.019. Computing A(t) and S(t) as before, except with ensembles of SCs, we recover A ≈ −0.78 ± 0.03 and S ≈ 4.63 ± 0.34 (refer to Fig. 10 for the full be- haviour of both metrics) . The latter is in agreement with S(t = 0) ≈ 4.57 for the observed TNOs while the former is consistent with the expected value of A(t = 0) ≈ −0.77 assuming that the mean apsidal angle of the observed TNOs is 180◦ away from that of the hypothesized disk. In short, the simulations we carried out show that the envisioned disk of trans-Neptunian icy bodies (DM1 to be specific) can provide a fair amount of -confinement for particles whose orbits are seeded in the neighbour- hood of the observed clustered TNOs. 4. VARIATIONS ON A THEME Given uncertainties, about disk mass, eccentricity, self-consistent precession,...etc, we thought it reasonable to explore a range of disk properties around the fiducial ones adopted in what preceded. Below is a brief account of what we learned, supplemented by the appropriate figure, when called for. These variations will be as- sessed with observed properties and disk self-consistency in mind. They will serve to inform our discussion of how a disk of the desired properties forms in the first place. Disk Mass. More massive disks, all else being kept the same, maintain planar equilibria of higher eccentric- ity, which as is clear by now, will carry over to properties of spatially aligned populations. Figure 11 illustrates the effect in disks that are less and more massive than the adopted reference disk (DM2 and DM3 in Table 1). This behavior is not too difficult to recover from model equations 8 and 9, which reduce to ep ≃ (cid:20)1 − c × M − 2 5 d × a 2 5 (p−4) p (cid:21)0.5 (15) under the assumption of axisymmetry (ed = 0) where the constant c = f (p) ×(cid:0)a2−p out P4 i=1 mia2 i(cid:1) 2 5 > 0.6 6 Important to note that Eq. 15 captures the trend of growing equilibrium ep with ap although the relation was derived for ed = 0. Also note that increasing aout (at constant Md with p < 2) tends to lower equilibrium eccentricities. 1 0.8 0.6 0.4 0.2 p e 0 50 1 0.8 0.6 0.4 0.2 0 50 p e , stable = = 0, unstable = 0, stable ) p (a e d Observed TNOs 100 150 200 250 300 350 400 450 500 (AU) a p (a) 100 150 200 250 300 350 400 450 500 (AU) a p (b) Figure 11. Coplanar families of stable and unstable equi- libria sustained by disk models DM2 and DM3 (see Table 1). DM2 and DM3 are identical to the fiducial DM1 except that DM2 is less massive with 2.5M⊕ (panel a) and DM3 more massive with 20M⊕ (panel b). Evidently, increasing Md drives up the equilibrium ep while maintaining the sta- bility of the three families, in agreement with our expectation (see Eq. 15). Furthermore, and as evident in panel b, mas- sive disks provide a supply of aligned orbits with ep ∼ ed over a broader range of ap . Thus, a better fit with the observed eccentricities (ignoring eccentricity-inclination dynamics) can be achieved with a disk mass which is higher than the one adopted in our analysis (see Fig. 11b). Furthermore, the bifurcation of equilibria into aligned and anti-aligned families is seeded earlier in a massive disk, such as DM3, implying that such disks will have a supply of aligned orbits (∆ = 0) over a broader range of semi-major axes, with which to build themselves up! If explanation be required, the reader should keep in mind that a more massive disk allows for stronger precession, hence the ability of low eccentricity orbits to withstand the dif- ferential precession induced by the planets, at smaller semi-major axis than otherwise possible with a lighter disk - see Figure 11. So far we have taken for granted disk models with mass increasing with ap. We now ask how would things differ in a disk with mass dropping outwards. As evident Shepherding of trans-Neptunian objects 11 , stable = = 0, unstable = 0, stable ed(ap) Observed TNOs 1 0.8 0.6 0.4 0.2 p e 1 0.8 0.6 0.4 0.2 p e ed (ap) (t) = ; stable Observed TNOs 0 50 100 150 200 250 300 350 400 450 500 0 100 150 200 250 300 350 400 450 500 ap (AU) ap (AU) Figure 12. The equilibrium TNO families sustained by a disk model analogous to DM1 but with more mass concen- trated in the inner parts, p = 2.5 (DM4 in Table 1). It is evident that although the assumed disk model gives rise to a coplanar stable family of anti-aligned high ep orbits, such a disk cannot harbour TNO orbits aligned with the disk such that ep ∼ ed(ap). Such is the case in all disks with mass dropping outwards (p > 2). in Figure 12, in a disk which is otherwise analogous to the fiducial model but with p = 2.5 (DM4 in Table 1), it appears impossible to support orbits which are aligned with the disk, and having ep ∼ ed(ap). Such (apse- aligned) disks appear unable to sustain themselves, and will not be discussed any further. Disk Eccentricity. Here one can change both the eccentricity profile (via q) as well as the eccentricity at the outer edge of the disk (via e0), for a given pro- file. Reporting on our thorough exploration of the rich set of bifurcations that obtain as a function of disk ec- centricity and their implication for the structure of the disk itself will take us too far afield. Suffice it to say that increasing the outer-edge eccentricity in negative-q disks (i.e. ded/dad > 0) or adopting eccentricity profiles which drop outwards, keeping all else invariant, intro- duces greater complexity in the structure of both anti- aligned and aligned planar equilibria, but unfortunately at the cost of loosing those disk-aligned orbits which we believe will be essential in any self-consistent reconstruc- tion of an eccentric disk. We find that for disks which are so structured that the bulk of their mass is in the outer parts, a disk eccentricity of e0 ≈ 0.20− 0.25 (with negative q) is the maximum that can be tolerated be- fore the eccentricity behavior of the disk-aligned family no longer follows that of the underlying disk. Disk Precession. A self-gravitating eccentric disk, which is further torqued by the outer planets, is likely to precess as a whole. The actual rate of precession of a saturated, nonlinear, eccentric mode is difficult to ascertain. The timescale associated with self-sustained precession is on the order of the secular timescale in the (t) = 0; stable 0.2 p e 0.1 ed (ap) 0 100 150 200 250 300 350 400 450 500 ap (AU) Figure 13. The effect of rigid disk precession, both retro- grade (in blue) and prograde (in red), on the two coplanar stable equilibrium TNO families sustained by non-precessing DM1 (in black). Precession rates of increasing order, 0.02, 0.2, and 2 × 10−8yr−1, are depicted by dotted, half-dashed and full lines respectively. The results show that prograde disk precession increases (lowers) the eccentricity of the high(low)-ep family; one has the opposite effect with d < 0, and this with increasing severity for larger d. For instance, equilibrium orbits aligned with the disk (∆(t) = 0) are not d = −2 × 10−8yr−1. Note that the stability of sustained if each equilibrium family is maintained in precessing disks. disk, ≃ (M⊙/Md) × TKepler, and comes out to ≃ 1010 years for a circular TNO orbit in a 1 M⊕ axisymmetric trans-Neptunian disk7. This is then superposed with differential precession induced by the outer planets, with a timescale of 1010 years. Actually, we can write the contribution of the giant planets to the total TNO precession rate as p(cid:12)(cid:12)planets ≈ +1.93 × 10−10yr−1(cid:18) 500AU ap (cid:19)3.5 for circular TNO orbits. (16) We explore the structure of equilibria in uniformly precessing disks at three progressively faster pattern 7 This characteristic timescale can be reduced by at least an order of magnitude upon accounting for the disc and TNO eccen- tricities (see Eq. 9) for a given disc mass. Furthermore, more massive discs drive faster precession since p ∝ Md (see Eq. 10). 12 Sefilian & Touma speeds (both prograde and retrograde). The results are shown in Figure 13. For the disk mass being consid- ered, it is evident that with prograde precession, agree- ment with the observed eccentricity profile improves at the risk of loosing dynamical support from the eccentric aligned and stable family of orbits which acquire lower values of ep with increasing d. On the other hand, retrograde precession worsens agreement with the ob- served family, while shifting the eccentricity profile of the aligned family to eccentricities that are too large to sustain the precessing disk. That the desirable features of our fiducial disk are dis- turbed by imposed precession is of course not surprising. Its properties were optimized under the assumption of zero precession. But now that we have a sense of the ef- fect of uniform precession, we can consider scenarios in which we optimize over disk properties and precession simultaneously. In particular, one can foresee a lower mass disk undergoing prograde precession while at the same time matching observed high eccentricity orbits, and sustaining a family of stable aligned orbits over a broader range of semi-major axes. 5. DISCUSSION Our proposition, with its pros and cons, is perhaps not as singular in the context of planetary system formation as the Planet Nine hypothesis. Still the ingredients that go into it, the origin of the disk, its mass, its eccentricity, as well as the self-consistent maintenance of the disk itself, require a closer look which we attempt below. Disk mass: There is to be sure much uncertainty concerning the mass that lies beyond Neptune, let alone question of eccentricity, and self-organization of that mass. We require an eccentric, lopsided, equilib- rium disk (precessing or not) of 10 Earth masses or so. Standard pictures allow for at most a few tenth of Earth masses to be scattered in that region in a primordial disk of planetesimals (Gladman et al. 2011; Silsbee & Tremaine 2018). Arguments that put a few tenth of Earth masses in that region are either based on extrapolations of ob- served size distributions, or on numerical simulations of a scattered disk that would invariably allow for such low amounts in that region. As noted in the text, such low masses can contribute to -confinement but they make for eccentricities in disagreement with what is ob- served (at least in the coplanar, non-precessing disk case - see Fig. 11a). But the question is how serious are these constraints? Well the distributions themselves are poorly constrained in their tail, and the dynamical ar- guments constrained by primordial assumptions which may or may not be legitimate. There are of course al- ternatives considered in the literature. Hills (1981) en- visions an intermediate zone between the Kuiper belt and the Oort cloud which is expected to harbor up to a few tens of Earth masses. Then there are suggestions that massive planetesimal disks may be a natural out- come of planet formation processes (Kenyon & Bromley 2004; Eriksson et al. 2018; Carrera et al. 2017). Most important for us however is the exercise of Hogg et al. (1991) who consider the question of hidden mass, and its gravitational signature and conclude that by looking at planetary motion and more importantly at cometary orbits, it is not unreasonable to expect up to a hundred or so Earth masses in the region in question. The exercise has not been revisited since (Tremaine, Pri- vate communication), though related questions were re- cently examined in relation to the Planet Nine hypoth- esis (Fienga et al. 2016; Bailey et al. 2016; Lai 2016). With the above in mind, it would seem that a mas- sive trans-Neptunian debris disk is not securely ruled out, hence our suggestion: rather than lump the per- turbing mass in a 10 Earth mass planet, and then find a way to push it out on an inclined and eccentric orbit (Kenyon & Bromley 2016; Eriksson et al. 2018; Parker et al. 2017), allow for the less daring hypothe- sis of a distribution of coplanar trans-Neptunian objects with a total of 10M⊕ and see what it does for you. Disk eccentricity: Of course, the existence of eccentric particle distributions with inner quadrupo- lar forcing is not foreign to the Solar System with Uranus's ǫ ring providing an early example of the type (Goldreich & Tremaine 1979). In that context, it was argued that self-gravity provides resistance to differen- tial precession induced by the planet's quadrupole to maintain an eccentric equilibrium configuration for the ring. Similar arguments in favor of self-gravitating eccentric distributions are brought to bear on the lopsided double nuclei of galaxies (Tremaine 1995; Peiris & Tremaine 2003). Such a mechanism may also structure eccentric circumbinary and/or circumpri- mary protoplanetary disks (Paardekooper et al. 2008; Kley et al. 2008; Meschiari 2012). Here, we gave evi- dence for a family of aligned and moderately eccentric orbits which promises to self-consistently build the disk that maintains it! So, little that is unusual about an eccentric disk though the details of its origin remain to be explored. Origin: Observations of ring systems, extra-solar de- bris disks, stellar disks, as well as theoretical mod- els and associated simulations suggest that eccentric disks are ubiquitous, and are rather easy to stimulate and apparently easy to sustain (with and without in- ner quadrupolar forcing). There are as many propo- Shepherding of trans-Neptunian objects 13 sitions for the origin of self-gravitating eccentric disks as there are dynamicists working in various contexts and at various scales: perturbation by passing objects (Jacobs & Sellwood 2001); dynamical instabilities that afflict low angular momentum (perhaps counter-rotating distributions) (Touma 2002; Tremaine 2005; Kaur et al. 2018); and forcing by eccentric inner or outer binary companion (Kley et al. 2008; Paardekooper et al. 2008; Marzari et al. 2009; Meschiari 2012; Pelupessy & Zwart 2013). In numerical experiments with an eccentric, self- gravitating, narrow ring-like disk, Madigan & McCourt (2016) noted an inclination instability which was accom- panied with a pattern of alignment in argument of pe- riapse. The authors took that as an indication that the process may underly the inclination-eccentricity behav- ior of the observed clustered TNOs. Intriguing though the proposition maybe, it suffers from various limita- tions: a- simulations do not allow for inner quadrupolar forcing by the planets, or earlier their being embedded in massive gaseous disk; b- simulations are not pursued long enough to follow the unfolding of the instability, and its eventual relaxation; c- hard to imagine how to form a disk of the required mass in the envisioned hot kinematic state. It is likely that inner quadrupolar forcing, if strong enough, can quench the inclination-eccentricity instabil- ity altogether, a suggestion which is motivated by re- lated effect in Kozai-Lidov type instability. As to the observed pattern of alignment in argument of periapse, it is surely a transient of an in-plane in- stability, which is expected for high eccentricity disks of the type considered (Kaur et al. 2018) and which ultimately relaxes into a lopsided uniformly precess- ing state of lower mean eccentricity. Gauss wire nu- merical simulations (Kazandjian, Sefilian & Touma, in preparation) confirm our expectations, with the disk of Madigan & McCourt (2016) relaxing into a thick lop- sided uniformly-precessing configuration in the presence of outer planet quadrupolar forcing, and remaining ax- isymmetric when we allow for outer planets which are ten times more massive. So while we believe the clustering mechanism of Madigan & McCourt (2016) is simply a transient which dissolves in time, it seems to do so in just the right sort of self-gravitating, eccentric, thick and uniformly pre- cessing disk that in combination with the outer planets is expected to sustain anti-aligned orbits with behavior comparable to what is observed! The difficulty of course is that the initial conditions for the required instability (bias towards low angular momentum, highly eccentric orbits) seem far from what is expected of distributions of planetesimals at formation. Promising in its ultimate state, but somewhat unlikely in its origin! Comment on disk self-consistency: Our proposi- tion is predicated on the properties of an idealized disk and its gravitational impact on test-particles that are embedded within it. We showed how a power-law disk can support stable equilibrium families of eccentric or- bits which align with the lopsidedness of the disk, as they reproduce its eccentricity profile. We further showed how, in such a disk, particles which librate in the AI are stable to off-plane perturbations, maintaining disk- alignment. This is all encouraging, in the sense that it suggests that a fully self-consistent thick, lopsided and precessing disk can be constructed. We would very much like to carry over our dynamical analysis to self-consistent equilibrium disks. For now, we note that in such disks a dispersion of apse directions will surely replace the apse-aligned eccentricity profiles of the present work. Apse dispersion, in the same razor thin disks, will mainly contribute to enhance the potential contribution of the axisymmetric mode over the lopsided one. Such relative adjustments are expected to leave the present qualitative picture pretty much unchanged, all the while inducing variations in the eccentricities of the various equilibrium families. The extreme of course is a disk that is hot enough to have a uniform distribution in the apses, an axisymmetric disk which, depending on its radial density profile, will sustain a degenerate family of equilibria. Slight non-axisymmetry will then break the degeneracy, and nucleate families of aligned and anti- aligned equilibria akin to the ones shown in Fig. 3 and Fig. 11. Comment on odd TNOS: Currently, three of the eccentric, inclined TNOs with ap > 250 AU and qp > 30 AU fall outside of the p-confinement which we have sought to account for in terms of an anti-aligned massive eccentric trans-Neptunian disk. Here we briefly review how these objects were analyzed with Planet Nine in the picture, as we further situate them within the phase space structured by DM1 and giant planets: • 2013 FT28 (Sheppard & Trujillo 2016) and 2015 KG163 (Shankman et al. 2017b): These two objects have apse-orientations which are nearly anti-aligned with the clustered bunch of ten. For Planet Nine activists, the detection of these two objects was reassuring, for it was understood early on that an eccentric and inclined super-Earth will shelter sta- ble eccentric objects with ∆ = 0, i.e. which are planet-aligned in apse (Batygin & Brown 2016; Beust 2016). Sheppard & Trujillo (2016) take 2013 FT28 as symptomatic of a larger cluster (dubbed the "secondary cluster") of TNOs which 14 Sefilian & Touma are stabilized in aligned orientations by/with Planet Nine. Batygin & Morbidelli (2017) pointed out that the alignment of FT28 and KG163 might be transient with a relatively short lifetime (100- 500 Myrs). We have here argued that DM1 shel- ters a family of stable aligned equilibrium orbits of moderate eccentricity which share the disk's eccentricity profile. While discussing planar phase space dynamics (Section 2), we showed how this family of orbits seeds aligned islands of stability (the so called AIs) in which particles undergo peri- odic oscillations in eccentricity and around the parent orbit. We further pointed out that mem- bers of the AI (and of the A-AI) find themselves on orbits which bring them close to the unstable aligned orbit, where they will tend to linger in transient disk-aligned states (see Fig. 4). While it is tempting to suggest that 2013 FT28 and 2015 KG163 are in similar such transient states8, lingering around the family of unstable aligned orbits, only further analysis with variants of DM1 (broad enough to include 2015 KG163) can decide that. • 2015 GT50 (Shankman et al. 2017b): a nonaligned TNO, almost orthogonal to the preferred apse ori- entation and which, for Shankman et al. (2017b), is yet another indication that confinement is due to observational bias (more on bias below). For Batygin & Morbidelli (2017), 2015 GT50 is one in a class of eccentric objects which are predomi- nantly controlled by Planet Nine on orbits with circulating longitude of the apse. We again refer to the discussion of planar phase space dynamics in Sec. 2 to remind the reader that DM1, together with the outer planets, orchestrates a copious pop- ulation of highly eccentric apse-circulating orbits, at semi-major axes which keep them safely out Neptune's way. In particular, clones of 2015 GT50 within our model demonstrate circulatory behav- ior in their p while undergoing small-amplitude oscillations in their orbital inclination and main- taining perihelion distance marginally larger than Neptune's orbital radius. 8 Simulations of coplanar particles with the semi-major axis of 2013 FT28 (ap = 310.1 AU) show lingering around the unstable aligned equilibrium (∆ = 0) for more than 5 Gyr when those particles are initiated in the AI or the A-AI (of Fig. 4) on orbits with large enough amplitudes to bring them close to that unstable equilibrium. Naturally, such orbits maintain eccentricities akin to that of the unstable equilibrium (∼ 0.71; Fig. 3) which is not so different from that of the inclined TNO 2013 FT28 (≈ 0.86). Comment on observational bias: Shankman et al. (2017b) scrutenized observational bias in the OSSOS sample and concluded that there is no evidence of clustering in ωp, Ωp and p distributions. A similar conclusion was drawn by Lawler et al. (2017). Indeed, Shankman et al. (2017b) report that although the OS- SOS survey is biased towards detecting TNOs with p near the region of observed clustering, it was able to detect TNOs across all values of , 2015 GT50 included (Shankman et al. 2017b). On the other hand, Brown (2017) concluded that al- though the observed sample is not free of biases, the statistical significance of the signal remains solid. In- deed, Brown (2017) estimates a rather low probability (∼ 1.2%) for the observed sample (with ap > 230 AU) to be drawn from a uniform population. Controversy over observational bias may or may not remove the need of a shepherding mechanism responsi- ble for the spatial alignment noted by Batygin & Brown (2016). Additional TNO discoveries will surely help clar- ify the matter further. Here, we would like to build on the ep−ap relationship we noted and explored in our sec- ular models (for instance, see Eq. 15 and Fig. 3) to pro- pose the following: if it proves that further (seemingly) clustered TNOs maintain the ep − ap trend currently correlated with dynamical models, then we take that as a strong observational signature favoring secularly in- duced clustering in the presence of a massive disk, an outer planet or both. Alternatively, if the ep − ap distri- bution of such objects reveals significant scatter, above and beyond that implied by the eccentricity-inclination dynamics of our models, then we would take that to weaken the case for dynamical clustering, and weigh more in favor of bias in the observed clustering. 6. CONCLUSION We probed dynamical behavior stimulated by a rel- atively massive disk of icy bodies in trans-Neptunian space to flesh out a hunch concerning the interplay be- tween the retrograde apse precession induced by such a disk and prograde precession forced by the outer planets: what if the clustered TNO population inhabits regions of phase space where the two effects cancel? Analysis of coplanar dynamics yielded a family of ec- centric, clustered, and apse frozen orbits, which showed remarkable agreement with the observed eccentricity- semi major axis distribution. It further yielded a family of low eccentricity orbits, aligned with the disk, which if properly populated is expected to reproduce the disk that helps sustain them: a self-consistency argument which we require for our disk's mass and eccentricity distributions. Shepherding of trans-Neptunian objects 15 We then allowed for out-of-plane motion and learned that an eccentric disk promotes linear stability to verti- cal motion, and gave evidence for the persistence of the planar backbone of stable apse-aligned orbits in inclined dynamics. We further analyzed the orbital evolution of the observed population of spatially aligned, eccentric and inclined bodies, without addressing its origin, and concluded that the envisioned self-gravitating disk main- tains what we like to think of as robust observables (i.e. eccentricity, inclination, longitude of apse) over the age of the Solar System. We carried out orbital simulations over the age of the Solar System while assuming a fixed planetary configu- ration, and a stationary disk. We ignored a dissipating gaseous disk, planetary migration and/or the scatter- ing of objects into the region where our disk resides. We were naturally concerned with the range of behav- ior sustained in the the "present" phase space of our hypothesized system. However, a massive gaseous disk could initially quench an eccentricity-inclination insta- bility in a kinematically hot debris disk, then, with its dissipation, the instability kicks in and allows an ini- tially axisymmetric disk to settle into a thick lopsided configuration which could harbor the apse-aligned or- bits that we observe (Kazandjian, Sefilian & Touma, in preparation). Furthermore, migration of planets, and secular resonances sweeping along with them might play a role in stirring an extended disk into an eccen- tric configuration(e.g Hahn & Malhotra 2005, and refer- ences therein)9. Our endeavor takes observational "evidence" for granted. Shankman et al. (2017a) cast doubt on the significance of the signal, further arguing that the mere observation of clustered TNOs, when taken at face value, requires a massive (∼ 6 − 24M⊕) extended reservoir of TNOs. We are of course happy to hear about indi- cations for a massive population of TNOs, while our colleagues see it as problematic given the currently favored estimates for mass in this region of the Solar System. These estimates put the total mass at ∼ 0.1M⊕ (Gladman et al. 2011). They are largely based on em- pirically constrained size distributions with significant uncertainty in their tails. We question those estimates as we point to recent global simulations of protoplan- etary disks suggesting the production of rather mas- 9 To the perspicacious reader who wonders about the evolution of equilibrium families with migrating outer planets, we note that such migration will primarily modify the strength of planetary quadrupolar forcing, thus shifting the location of zero net apse precession in the disk, and moderately perturbing the eccentric- ities of aligned and anti-aligned equilibria ( ∼ 10 − 20% in the course of migration). sive (% 60 M⊕) planetesimal disks beyond 100 AU (Carrera et al. 2017). The last thorough attempt at dynamical modelling of baryonic dark matter in the outer Solar System was undertaken in the early nineties (Hogg et al. 1991). By considering variations in cometary orbital elements, Hogg et al. (1991) argued for a few (perhaps hundreds) of Earth masses on scales of 100 AU. We like to think that this early exercise [which incidentally was un- dertaken to carefully examine the evidence (or lack thereof) for a tenth planet] is being revisited piece- meal with Planet Nine in mind (Batygin & Brown 2016; Fienga et al. 2016; Bailey et al. 2016)! We propose that similar such indirect measures be undertaken with an extended moderately eccentric few Earth mass disk perhaps replacing, perhaps combined with, a trans- Neptunian planet. Of course we can draw comfort in our hypothesis from observations of extra-solar debris disks particularly mas- sive ones around planet-hosting stars10, as we hope for the mechanism of -confinement identified in this study to shed light on the dynamics and structure of these disks. Ultimately though, we do not have secure and di- rect observational evidence for our proposed disk, pretty much like we do not have full proof arguments against Planet Nine. Still, we hope to have given sufficiently many dynamical indicators for the game changing role of such a disk in shepherding eccentric TNOs over a broad range of semi-major axes. Of course, a massive eccentric disk could operate simultaneously with a post- Neptunian planet to assure full secular spatial confine- ment if and when called for. And the converse is also true in the sense that, with the proper mass distribu- tion and orbital architecture, a disk-planet combination may prove capable of stabilizing orbits in configurations which are difficult to maintain with Planet Nine acting alone. TNO 2013 SY99 (Bannister et al. 2017) provides a case in point. A newly discovered object, its orbit is highly eccentric, apparently clustered with the wild bunch, but unlike its companions nearly in the ecliptic with ∼ 4◦ orbital inclination. This object is so tenuously held on its orbit that it is exposed to the randomizing influence of Neptune promoting diffusion at the inner edge of the Oort cloud. Interestingly enough, a Planet Nine like influence is not of much stabilizing help here. In fact, when allowance is made for an inclined eccen- 10 For instance, the analog of the Kuiper belt around τ Ceti has been estimated to contain 1.2ME in r < 10km objects (Greaves et al. 2004) with four candidate planets orbiting the star in the inner region (Feng et al. 2017). 16 Sefilian & Touma tric ninth planet, a la Batygin & Brown (2016), all hell breaks loose in the orbital evolution of this curious TNO (Bannister et al. 2017). Well, forgetting about Planet Nine for the moment, and entertaining, as we like to do, the possibility of a massive trans-Neptunian disk, we naturally find stable anti-aligned coplanar equilibria at a few hundreds AUs, and with nearly the observed eccentricity! In fact, following the orbital evolution of 2013 SY99 under the action of our hypothesized disk, we learned that its current orbit can be sustained over the age of the Solar System, executing only mild oscillations in inclination and perihelion distance, while maintaining near-alignment in as the pericenter and node circu- late. These two limits, along with the eccentricity-semi major axis distribution which we highlighted, speak in favor of the combined action of a self-gravitating trans- Neptunian disk, together with a trans-Neptunian terres- trial core (e.g. something akin to what was recently sug- gested by Volk & Malhotra (2017), or perhaps the result of a scattering event a la Silsbee & Tremaine (2018)). We end with the hope for this combined action to be the subject of parametric studies akin to the ones un- dertaken with Planet Nine acting alone. This work grew out of a graduate seminar on the Oort Cloud conducted in Fall 2016 at the American University of Beirut. Discussions with participants M. Khaldieh and R. Badr are gratefully acknowledged. The authors would like to thank S. Tremaine, S. Sridhar, L. Klushin and M. Bannister for helpful discussions, and an anonymous referee for constructive comments. We thank M. Kazandjian for making his modal analy- sis toolbox available to us. J.T. acknowledges insightful comments by R. Touma concerning the potential inter- play between Planet Nine and a massive eccentric disk. A.S. acknowledges a scholarship by the Gates Cam- bridge Trust (OPP1144). Open Access for this article was funded by Bill & Melinda Gates Foundation. A. THE DISTURBING POTENTIAL DUE TO THE DISK APPENDIX We present explicit expressions for the expansion of the orbit-averaged disturbing function (per unit mass) Rd due to a disk composed of coplanar, apse-aligned, and confocal ellipses given in equation (4). The expansion is carried out to fourth order in test-particle eccentricity, generalizing the work of Silsbee & Rafikov (2015), and is valid for arbitrary semi-major axes by the aid of the classical (unsoftened) Laplace coefficients. The mathematical details and techniques involved; from expanding the disturbing function in eccentricities and retaining the secular part, are discussed in Sefilian (2017). We consider a finite razor-thin disk, orbiting a central massive object, composed of confocal eccentric apse-aligned streamlines. The non-axisymmetric surface density of such a disk can be written as (Statler 1999), Σ(ad, φd) = Σd(ad) 1 − ed(ad)2 − ade 1 − ed(ad)2 − ade′ ′ d[1 + ed(ad)] d[ed(ad) + cos Ed] (A1) where Ed (φd) is the eccentric (true) anomaly of the disk element and e In what follows, we have assumed power-law relations for Σd(ad) and ed(ad) as given in equations (1) and (2), although it is possible to use the same methods involved to recover expansions for arbitrary profiles as well as for cases with d = d(ad) (Statler 2001; Ogilvie 2001) . Our aim is to compute the secular potential Φd experienced by a test-particle embedded in the disk, ed(ad). ′ d ≡ d dad Φd = −Rd = −G(cid:28)Z Σ(rd, φd)rddrddφd ∆ (cid:29) (A2) where the integration is performed over the area of the eccentric disk, < .. > represents time-averaging over the test- particle orbit, G is the gravitational constant, and ∆2 = r2 d − 2rprd cos θ with rp and rd being the instantaneous position vectors of the test-particle and disk element respectively such that (rp, rd) = θ. Following the classical formulation developed by Heppenheimer (1980) who first computed Φd for axisymmetric disks without softening the kernel ∆ (see Ward 1981), we make use of the techniques laid down in Silsbee & Rafikov (2015) and extend their derivation to fourth-order in eccentricities rendering our forumalae applicable in general astrophysical setups which need not be cases where the assumption of e << 1 holds. After a laborious task of algebraic manipulations (expanding, averaging...) and ignoring constant terms which have no dynamical effect at the secular level, we find that Φd takes the following form p + r2 Φd = −Rd = −K(cid:20)ψ1ep cos(p − d) +(cid:0)ψ2 + ψ3 cos(2p − 2d)(cid:1)e2 p + ψ4e3 p(cid:21) p cos(p − d) + ψ5e4 (A3) Shepherding of trans-Neptunian objects 17 where K = πGΣ0ap power-law indices p and q. To compactify the expressions of ψi, we use the definition of Laplace coefficients p > 0 and the dimensionless coefficients ψi depend on α1 ≡ ain/ap, α2 ≡ ap/aout, and the outa1−p bm s (α) = 2 π and define the following auxiliary functions π Z 0 cos(mθ) (1 + α2 − 2α cos θ)s/2 dθ I(x, y, z) = to write 1 Z x αybz 1(α)dα and Dm i [f ] = dm dαm i f (αi) for i = 1,2 and m (cid:23) 0 J1 = I(α1, 1 − p, 0) + I(α2, p − 2, 0) J2 = I(α1, 1 − p − q, 0) + I(α2, p + q − 2, 0) J3 = I(α1, 1 − p − q, 1) + I(α2, p + q − 2, 1) These definitions allow us to cast the coefficients ψi in the following form: (A4) (A5) (A6) (A7) (A8) ψ1 = ψ2 = ed(ap) 2 (cid:26) − (p + q)(p + q − 3)J3 + αp+q−1 1(cid:3) + α2−p−q 2b1 (cid:2)(2 − p − q)D0 1 + α2D1 2b1 α2−p−2q (cid:18)D0 1(cid:19)(cid:21) 1(cid:19) − 1b1 1 − α1D1 1b1 2b1 p − 2 2[αp−2b0 α2D0 (cid:18)D0 2b1 α1D0 1[α1−p b0 1] + α2 2 1 + D1 2 2 1 1 2 + ed(ap)2(p + 2q)(cid:20)αp+2q−1 (1 − p)(2 − p) 1 − p ed(ap) J1 + 2 4 1[α1−p (cid:20)q(1 − p − q)(2 − p − q)J2 + 2q(1 − p − q)α1D0 b0 1] − D1 1 1 + 4 α2 1 4 2 + 2q(p + q − 2)α2D0 2[αp+q−2b0 2 (cid:18)D1 (cid:20)αp+2q + ed(ap)2 p + 3q 2b0 (cid:18)D0 8 (cid:20)3αp+2q−1 2b2 4 ψ3 = ed(ap)2 p + q 1 + 2 2D1 1] − qα2 D2 α2 2 1 + α2D1 1D1 1[α1−p−qb0 1] − qα2 2 (cid:18)D1 α2 1] + 2αp+q 2b0 2[αp+q−2b0 D2 2 (cid:18)D1 1(cid:19) − α3−p−2q 1b0 2b0 1(cid:19) − α2−p−2q 2b2 2b2 1(cid:19)(cid:21) 1b0 (cid:18)D0 1b2 1 − α1D1 1b2 α2 2 6 α1 2 1 + 1 + 1 + D2 D2 1 1 1b1 (cid:2)(p + q − 1)D0 1 + α1D1 1b1 1(cid:3)(cid:27) (A9) 2[αp−2b0 1] D1 1] − α2 2 4 1[α1−p−qb0 1] 1(cid:19) − 2α3−p−q 2b0 1 (cid:18)D1 1b0 1 + α1 2 D2 1(cid:19)(cid:21) 1b0 1 + D2 α2 1 2 1(cid:19)(cid:21) 1b2 2[αp+q−2b1 1] ψ4 = ed(ap) 16 (cid:20)(p + q)2(1 − p − q)(p + q − 3)J3 − (p + q)(cid:2)(p + q − 2)(3p + 3q − 5) − 2(cid:3)α2D0 2D1 + (p + q)(3p + 3q − 7)α2 + (p + q)(3p + 3q − 1)α2 − 2α2−p−q (cid:18)D0 1b1 1 − α1D1 1b1 1D1 1 5α2 1 2 D2 1b1 1 − α3 1 2 D3 1 − 1D2 1(cid:19)(cid:21) 1b1 2[αp+q−2b1 1[α1−p−qb1 1] − (p + q)α3 1] + (p + q)α3 2D2 2[αp+q−2b1 1[α1−p−qb1 1] + (p + q)2(3p + 3q − 5)α1D0 1[α1−p−qb1 1] 1(cid:19) 1] + αp+q+1 2b1 (cid:18)4D2 2b1 1 + α2D3 2 and ψ5 = + 64 p(1 − p2)(2 − p) α2 16 J1 − (p3 − 3p2 + 2p)D0 α1 16 p(p2 − 1)D0 3α2 2 32 1] − 1] − 1[α1−pb0 3α2 1 32 (p − 1)(p − 2)D1 2[αp−2b0 2[αp−2b0 1] − α3 1 1] − 16 (2 − p)D2 α3 2 16 2[αp−2b0 1] − α4 2 64 D3 α4 1 64 1] − 2[αp−2b0 1] (p2 + p)D1 1[α1−pb0 (p + 1)D2 1[α1−pb0 D3 1[α1−pb0 1] We note that the disk potential Φd presented here reduces to that of Silsbee & Rafikov (2015) upon limiting it to second order in eccentricities and naturally, to that of Heppenheimer (1980) for axisymmetric disks. We end with a brief remark about the behavior of the coefficients ψi. The expressions of ψi depend on the disk boundaries through α1 and α2 such that the magnitudes of ψi diverge when the test-particle is situated nearby the disk (A10) (A11) (A12) (A13) (A14) 18 Sefilian & Touma edges. However, in the opposite limit, when the test-particle semi-major axis is well separated from the disk boundaries ain and aout, we can ignore the edge effects provided that the disk spans several order of magnitude in radius. In such a case (α1, α2 → 0) we can get relatively simple closed form expressions for ψi valid for ain << ap << aout using the series expansions of complete elliptic integrals (Gradshteyn & Ryzhik 1994) such that; ∞ Xn=2 (2n − 1)(2n + 1 − p − q)(2n + p + q − 2)(cid:21) 2nAn(4n − 1) (4n + 1)An (2n + 2 − p)(2n + p − 1) ∞ Xn=1 (4n + 1)An (2n + 2 − p − q)(2n + p + q − 1)(cid:21) ψ1 = ed(ap)(cid:20) 3 2 − (p + q)(p + q − 3) ψ2 =− 1 2 + (1 − p)(2 − p) 2 ∞ Xn=1 + qed(ap) 2 (cid:20)1 + (1 − p − q)(2 − p − q) ψ3 = 0 ψ4 = ψ5 = where √An = (2n)! 22n(n!)2 . ψ1 8 (p + q)(p + q − 1) p(p + 1) ψ2(cid:12)(cid:12)q=0 16 (A15) (A16) (A17) (A18) (A19) B. NUMERICAL 3D POTENTIAL OF A DISK Here, we present a brief recipe to recover numerically the full 3D gravitational potential generated by a disk formed of coplanar, apse-aligned, eccentric rings. The numerical tool that we developed is indeed general and can be employed to recover the potential due to any configuration of rings (warped, inclined, spiral-armed disks, etc...). The ramifications of this tool will be explored in the future (Kazandjian, Sefilian & Touma, in preparation). For the purposes of this study, we distributed N (=1024) coplanar, apse-aligned rings over the range of the disk with specified mass and eccentricity distributions (Eq. 1 and 2). We then computed the full 3D potential experienced by test-particles by recovering the associated spherical harmonics Y m l (θ, φ) numerically. Identifying the most dominant modes al,m(r), we then orbit averaged numerically the arising terms and obtained closed form expressions for the modes in question for any given semi-major axis. This numerical procedure allows us to express the secular potential of any disk as ¯Φd = Xl,m (cid:10)al,m(r)Y m l (θ, φ)(cid:11) (B20) where < .. > stands for time-averaging, and the angles θ and φ can be expressed in terms of the usual orbital elements using x = r sin θ cos φ = r[cos(Ω − d) cos(w + f ) − sin(Ω − d) sin(w + f ) cos(i)] y = r sin θ sin φ = r[sin(Ω − d) cos(w + f ) + cos(Ω − d) sin(w + f ) cos(i)] z = r cos θ = r sin(w + f ) sin(i) (B21a) (B21b) (B21c) where f is the true anomaly. Equipped with the relevant dominant modes al,m(r), we find that the potential due to the reference disk (DM1; Table 1) takes the following form ¯Φd = −r 1 4π < a0,0(r) > +r 5 4π < a2,0P 0 2 (cos θ) > −r 3 8π < a1,1(r)P 1 1 (cos θ) cos φ > +r 7 48π < a3,1(r)P 1 3 (cos θ) cos φ > (B22) where we have assumed that we have averaged over the orbit of the test particles ( < .. > ) and P m l (x) are the Legendre polynomials. After a long, tedious, but straightforward algebra, we express ¯Φd in terms of the orbital elements and Shepherding of trans-Neptunian objects 19 10 8 6 10 8 6 0 -5 -10 f0,0,0 106 f2,0,0 107 f3,1,1 108 0 0.2 0.4 e p = 260.8AU a p = 369.7AU a p = 499.4AU a p f1,1,1 107 f2,0,2 107 f3,1,3 108 0 -1 -2 -3 8 4 0 0 -5 0.6 0.8 1 0 0.2 0.4 0.6 0.8 e p -10 1 Figure 14. The behavior of numerically-averaged functions fl,m,n =< al,m(r) cos(nf ) > appearing in the disk potential (Eq. B23) as a function of eccentricity ep for orbits with different semi-major axes ap as shown. The calculation is performed for the fiducial disk configuration DM1 (see Table 1). The chosen values of ap correspond to semi-major axes at which TNOs are observed (see Table 3). Note that the transition in behavior of fl,m,n with varying semi-major axis is smooth. = , stable; A = , stable; N =0, unstable; A =0, unstable; N =0, stable; A =0, stable; N 1 0.8 0.6 0.4 0.2 p e 0 50 100 150 200 250 300 350 400 450 500 550 (AU) a p Figure 15. Comparison between the coplanar equilibrium families obtained via orbit-averaged disk potential (A; Eq. 4) and orbit-averaged harmonics (N; Eq. B23 with ip = 0) of the potential generated by DM1 (Table 1). It is evident that both treatments of the disk potential, when combined with that of the giant planets, yield very similar equilibrium structure, including stability and behavior as a function of semi-major axis. write ¯Φd = F (ep) + G(ep)(cid:2) cos(Ωp − d) cos(wp) − cos(ip) sin(Ωp − d) sin(wp)(cid:3) 4π 4r 5 8 r 7 15 sin2(ip)(cid:20)f2,0,0 − cos(2wp)f2,0,2 +r 35 cos(ip) sin2(ip) sin(Ωp − d) sin(wp)(cid:2)3f3,1,1 − f3,1,1 − 2f3,1,3 cos(2wp)(cid:3) cos(Ωp − d)(cid:2) cos(3wp)f3,1,3 − cos(wp)f3,1,1(cid:3)(cid:21) 48 3 + + 48π (B23) 20 Sefilian & Touma where fl,m,n(ep; ap) =< al,m(r) cos(nf ) > are computed numerically at a given semi-major axis to the desired (arbi- trary) order in eccentricity, and we have defined F = −r 1 G = r 3 8π 4π f0,0,0 − 3 f1,1,1 + 4π 1 2r 5 2r 7 48π f2,0,0 f3,1,1 (B24a) (B24b) It is noteworthy that all coplanar, apse-aligned, power-law disks share this general form of the Hamiltonian with the only difference being in the functions al,m(r) which depend on the characteristics of the disk: mass distribution, eccentricity profile and the boundaries. For reference, in Fig. 14 we show the time-averaged behavior of the functions fl,m,n for DM1 as a function of test-particle eccentricity at three values of semi-major axes. With the orbit-averaged mean field of the razor thin eccentric disk in hand, we add the secular contribution of the outer planets and write the total Hamiltonian H as H = ¯Φd − Γ 6l3 p [3 cos(ip)2 − 1] (B25) where Γ is given by equation 7. This allows us to express the equations of motion governing the dynamics of a TNO under the effect of the considered disk and the giant planets as: T1 + 3 + l3 p p (cid:18) Γ Lp Ωp =−Cil−1 4r 5 Lp lp = G(ChSw + CiShCw) − 5 π 3 4 S2 i(cid:18)r 5 π − S2wf2,0,2 − Lp αp = G(ShCw + CiChSw) + 5 5 π π 16r 21 32r 21 8r 21 32r 21 π 5 ChT2(cid:19) − ShSwl−1 p (cid:20) 15 32r 21 π C2 i T3 +(cid:18)G − 5 32r 21 π CiS2 i Sh(cid:0)4SwS2wf3,1,3 + CwT3(cid:1) ChSwf3,1,1(cid:19) 24r 21 π 5 ChS3wf3,1,3 + T3(cid:19)(cid:21) (B26) (B27) (B28) π p CiShSw + S2 i (ShT2 − CiChSwT3) 32r 21 5 (ChCw − CiShSw) + S2 p Ci(3C2 l−1 8r 5 i(cid:18) 3 T π π ′ i − 3) + Gl−1 (cid:20)F + G ′ ′ i − 1)ShSwT3 + l−1 2(cid:19) + 32r 21 4r 5 i(cid:20) 3 p C2 32r 21 ChT 1 + π π π 5 5 ′ T1 + 5 π 16r 21 ChT2(cid:21) 3(cid:21) i T ′ ShSwCiS2 (B29) Lp ωp = Γ 6l4 p (15C2 − lp q1 − l2 p where we have written Lp = pGM⊙ap T1 = f2,0,0 − C2wf2,0,2 T2 = C3wf3,1,3 − Cwf3,1,1 T3 = 3f3,1,1 − f3,1,3(1 + 2C2w) (B30a) (B30b) (B30c) (B30d) and S and C are shorthand for sine and cosine of the angles given as subscript: i is the inclination determined by p as before, h = Ωp − d and w is argument of pericenter. Note that the primed . This set of equations represents the basis of our population study α ≡ Lz/Lp = lp cos(ip), lp = q1 − e2 ′ terms in Eq. B29 are defined such that f performed in the body of this work (Section 3). = ∂f (ep) ∂ep Finally, we comment on the accuracy of our numerical disk potential. Insisting on coplanar test-particle orbits by setting ip = 0 in Eq. B25, we solved for the equilibria and analyzed their stability. The results are shown in Fig. 15. It is evident that the coplanar equilibria recovered by the numerical formulation of the disk potential (Eq. B23) and that of the analytical (Eq. 4) agree very well both in orbital structure and stability. Shepherding of trans-Neptunian objects 21 REFERENCES Bailey, E., Batygin, K., & Brown, M. E. 2016, AJ, 152, 126 Kley, W., Papaloizou, J. C., & Ogilvie, G. I. 2008, A&A, Bannister, M. T., Shankman, C., Volk, K., et al. 2017, AJ, 487, 671 153, 262 Batygin, K., & Brown, M. E. 2016, AJ, 151, 22 Batygin, K., & Morbidelli, A. 2017, AJ, 154, 229 Beust, H. 2016, A&A, 590, L2 Brown, M E. 2017, AJ, 154, 65 Carrera, D., Gorti, U., Johansen, A., & Davies, M. B. 2017, ApJ, 839, 16 Eriksson, L., Mustill, A. J., & Johansen A. 2018, MNRAS, 475, 4609 Lai, D. 2016, AJ, 152, 215 Lawler, S. M., Shankman, C., Kaib, N., et al. 2017, AJ, 153, 33 Madigan, A. M., & McCourt, M. 2016, MNRAS, 457, L89 Marzari, F., Scholl, H., Thebault, P., & Baruteau, C. 2009, A&A, 508, 1493 Meschiari, S. 2012, ApJL, 761, L7 Millholland, S., & Laughlin, G. 2017, AJ, 153, 91 Ogilvie, G. 2001, MNRAS, 325, 231 Paardekooper, S.-J., Thebault, P., & Mellema, G. 2008, Feng, F., Tuomi, M., Jones, H. R., et al. 2017, AJ, 154, 135 MNRAS, 386, 973 Fienga, A., Laskar, J., Manche, H., & Gastineau, M. 2016, A&A, 587, L8 Gladman, B., Kavelaars, J. J., Petit, J. M., et al. 2011, AJ, 122, 1051 Goldreich, P., & Tremaine, S. 1979, AJ, 84, 1638 Gradshteyn, I. S., & Ryzhik, I. M. 1994, Table of Integrals, Parker, R. J., Lichtenberg, T., & Quanz, S. P. 2017, MNRAS, 472, L75 Peiris, H. V., & Tremaine, S. 2003, ApJ, 599, 237 Pelupessy, F. I., & Zwart, S. P. 2013, MNRAS, 429, 895 Rafikov, R. R. 2013, ApJL, 765, L8 Rafikov, R. R., & Silsbee, K. 2015, ApJ, 798, 69 Sefilian, A. A. 2017, MSc Thesis, American University of Series and Products (New York: Academic) Beirut Greaves, J. S., Wyatt, M. C., Holland, W. S., & Dent, Shankman, C., Kavelaars, J. J., Lawler, S. M., et al. 2017a, W. R. F. 2004, MNRAS, 351, L54 Hahn, J. M. 2003, ApJ, 595, 531 AJ, 153, 63 Shankman, C., Kavelaars, J. J., Bannister, M. T., et al. Hahn, J. M., & Malhotra, R. 2005, AJ, 130, 2392 2017b, AJ, 154, 50 Heppenheimer, T. A. 1980, Icarus, 41, 76 Hills, J. G. 1981, AJ, 86, 1730 Hogg, D. W., Quinlan, G. D., & Tremaine, S. 1991, AJ, 101, 2274 Iorio, L. 2014, MNRAS, 444, L78 Jacobs V., & Sellwood J. A. 2001, ApJL, 555, L25 Kaur, K., Kazandjian. M. V., Sridhar, S., & Touma, J. R. 2018, MNRAS, 476, 4104 Kenyon, S. J., & Bromley, B. C. 2004, AJ, 127, 513 Kenyon, S. J., & Bromley, B. C. 2016, ApJ, 825, 33 Sheppard, S. S., & Trujillo, C. 2016, AJ, 152, 221 Silsbee, K., & Rafikov, R. R. 2015, ApJ, 798, 71 Silsbee, K., & Tremaine, S. 2018, AJ, 155, 75 Statler, T. S. 1999, ApJL, 524, L87 Statler, T. S. 2001, AJ, 122, 2257 Touma, J. 2002, MNRAS, 333, 583 Tremaine, S. 1995, AJ, 110, 628 Tremaine, S. 2005, ApJ, 625, 143 Trujillo, C. A., & Sheppard, S. S. 2014, Nature, 507, 471 Volk, K., & Malhotra, R. 2017, AJ, 154, 62 Ward, W. R. 1981, Icarus, 47, 234
1510.00858	1	1510	2015-10-03T19:37:09	The Structure and Evolution of Protoplanetary Disks: an infrared and submillimeter view	[ "astro-ph.EP", "astro-ph.SR" ]	Circumstellar disks are the sites of planet formation, and the very high incidence of extrasolar planets implies that most of them actually form planetary systems. Studying the structure and evolution of protoplanetary disks can thus place important constraints on the conditions, timescales, and mechanisms associated with the planet formation process. In this review, we discuss observational results from infrared and submillimeter wavelength studies. We review disk lifetimes, transition objects, disk demographics, and highlight a few remarkable results from ALMA Early Science observations. We finish with a brief discussion of ALMA's potential to transform the field in near future.	astro-ph.EP	astro-ph	Young Stars & Planets Near the Sun Proceedings IAU Symposium No. 314, 2015 J. H. Kastner, B. Stelzer, & S. A. Metchev, eds. c(cid:13) 2015 International Astronomical Union DOI: 00.0000/X000000000000000X The Structure and Evolution of Protoplanetary Disks: an infrared and submillimeter view Lucas A. Cieza1 1N´ucleo de Astronom´ıa, Universidad Diego Portales, Chile email: [email protected] Abstract. Circumstellar disks are the sites of planet formation, and the very high incidence of extrasolar planets implies that most of them actually form planetary systems. Studying the structure and evolution of protoplanetary disks can thus place important constraints on the conditions, timescales, and mechanisms associated with the planet formation process. In this review, we discuss observational results from infrared and submillimeter wavelength studies. We review disk lifetimes, transition objects, disk demographics, and highlight a few remarkable results from ALMA Early Science observations. We finish with a brief discussion of ALMA's potential to transform the field in near future. Keywords. protoplanetary disks, infrared: planetary systems, submillimeter: planetary systems 1. Introduction Protoplanetary disks are complex systems that evolve through various physical mech- anisms, including accretion onto the star (Hartmann et al. 1998), grain growth and dust settling (Dominik, C., & Dullemond, 2008), dynamical interactions (Artymowicz & Lubow, 1994), photoevaporation (Alexander et al. 2006), and planet formation itself (Lissauer 1993; Boss et al. 2000). However, the relative importance and timescales of these processes are still not fully understood. Describing each of these processes is be- yond the scope of this review. Instead, we focus on IR and submillimeter† observational results, while providing a global view of disk evolution and their connection to planet for- mation theory. In §2 we discuss results from infrared observations, while in §3 we present submillimeter results previous to the commissioning of the Atacama Large Millimeter Array (ALMA). In §4 we discuss some of the highlights of ALMA Early Science obser- vations. In §5 we speculate on connections between disk demographics and extrasolar planet statistics. In §6 we finish with a discussion on the prospects for studies with the full ALMA array and present our conclusions. Complementary discussions on planet for- mation, chemistry, disk evolution theory, and observational studies in other wavelength regimes can be found in other contributions to this conference proceedings. 2. Infrared constraints 2.1. Disk lifetime The lifetime of protoplanetary disks is perhaps the strongest astrophysical constraint on planet formation theory, as it provides valuable information on the time available to complete the process. For rocky planets, their formation could continue beyond the † Submillimeter in this context refers to the wavelength regime extending from ∼0.3 mm to a few mm. 1 5 1 0 2 t c O 3 . ] P E h p - o r t s a [ 1 v 8 5 8 0 0 . 0 1 5 1 : v i X r a 2 Lucas A. Cieza dispersal of the gas. In the case of giant planets, the dissipation of the primordial (gas rich) disk sets a hard limit for their formation timescale. Since very small amounts of hot dust (approximately equivalent to an asteroid mass at 1000 K) are needed to produce an optically thick near-IR (NIR) excess above the stellar photosphere, NIR observations are particularly useful to trace the presence of an inner accretion disk (r . 0.1 AU). In fact, there is a very close correspondence between NIR excesses and accretion signatures (e.g. Hartigan et al. 1995), the only exception being transition disks with completely depleted inner holes. From NIR observations of clusters at different ages, we know that ∼80% of stars with ages ∼1 Myr have an inner accretion disk, and that the inner-disk fraction drops close to ∼0 % by 10 Myr (Mamajek, 2009). From these studies we can derive a mean inner- disk lifetime of 2-3 Myr with a large dispersion: some pre-main-sequence (PMS) stars accrete for less < 1 Myr, other accrete for up to 10 Myr. Mid-infrared (MIR) observations with Spitzer later showed that most targets lacking NIR excesses also lack MIR excesses (Cieza et al. 2007; Wahhaj et al. 2010), again, with the exception of transition disks with clean inner holes. This indicates that, in general, once accretion stops and the inner disk dissipates, the entire disk is dispersed completely rather quickly, in agreement with the predictions of photoevaporation models (e.g., Alexander et al. 2014). In any given cluster, disk fractions tend to be lower around higher-mass stars than lower-mass objects (Carpenter et al. 2006), indicating that disk dissipation proceeds faster around higher mass stars. This could be due to the combination of higher accre- tion and/or higher photoevaporation rates. Disk lifetime is also a function of multiplicity. In young clusters, disk frequency is close to 100% for single stars and much lower for medium-separation (5-50 AU) binaries (Kraus et al. 2012; Cieza et al. 2009). This is easy to understand, as mid-separation binaries truncate each other's outer disks, drastically decreasing the amount of material available for accretion. Spectroscopic binaries (sepa- ration . 1 AU) and wide binaries (separation > 100 AU) seem to have little effect on disk lifetimes. 2.2. Disk structure and evolution IR observations are also useful to derive disk structures. Most protoplanetary disks (∼80%) have "full disks" extending inward to the dust sublimation radius and are opti- cally thick at all IR wavelengths. As a result, they all have very similar spectral energy distributions (SEDs). The other 20%, the transition disks, have a wide range of structures and SEDs. Protoplanetary disks can be thought of as a series of optically thick annuli at different temperatures. Each annulus dominates the emission at a different wavelength; the closer to the star and the hotter the annulus, the shorter the corresponding wave- length. Therefore, the presence of an inner hole in the disk translates to a reduced level of NIR excess. If the hole is completely empty, no detectable excess will be seen above the stellar photosphere. If some detectable dust remains inside the cavity, an optically thin excess will still be present, but the NIR SED will stay below the typical value of an opti- cally thick inner-disk. Similarly, if the disk contains a wide enough gap that is optically thin in the IR, its SED will show a dip at the wavelength corresponding to the temper- ature of the "missing" dust annulus. Furthermore, most young protoplanetary disks are flared and thus intercept and reprocess significantly more starlight than a flat disk. As disks evolve and dust grows and settles to the mid-plane, the outer disk becomes flatter with time, which translates to reduced levels of mid- and far-IR emission with respect to a fully flared disk. Finally, as the primordial disk dissipates, it becomes optically thin at all IR wavelengths, resulting in reduced levels of IR excess at all wavelengths. In practice, most young stellar clusters and star-forming regions contain two categories The structure and evolution of protoplanetary disks 3 of disks: normal "full" disks that are optically thick and transition disks. The latter group includes objects with inner holes and gaps (sometimes called classical transition disks, cold disks, or pre-transition disks) and objects that are physically very flat or optically thin in the IR (some times called anemic, weak-excess or homologously depleted disks)†. The relative number of these types of objects seems to be a function of age, with more transition disks in older star-forming regions and clusters (Currie et al. 2009). 3. Pre-ALMA submillimeter results 3.1. Disk mass While most protoplanetary disks remain optically thick in the IR, they become optically thin at longer wavelengths. Submillimeter wavelength observations are hence sensitive to fundamental disk properties such as total gas and dust masses and grain size distribu- tions, and are highly complementary to IR data. The main constituent of protoplanetary disks, H2, is a very poor emitter under the relevant temperatures and densities. Cir- cumstellar dust is much easier to observe from continuum observations. Estimating the mass of a disk, and hence its ability to form different types of planets, typically entails making very strong assumptions about the dust opacity and the gas to dust mass ratio (Williams & Cieza, 2011). Submillimeter fluxes are routinely used to estimate the masses of protoplanetary disks using: Mdust = Fν d2 κνBν (Tdust) (3.1) where d is the distance to the target, T is the dust temperature and κν is the dust opacity. Following Beckwith et al. (1990), and making standard (although uncertain) assumptions about the disk temperature (Tdust = 20 K), the dust opacity (κν = 10[ν/1200 GHz] cm2g−1), and the gas to dust mass ratio (100), Equation 3.1 becomes: Mdisk(gas + dust) = 1.7 × 10−4(cid:18) Fν(1.3mm) mJy (cid:19) ×(cid:18) d 140pc(cid:19)2 M⊙ (3.2) The gas to dust mass ratio of 100 that is usually adopted is appropriate for the in- terstellar medium, but highly questionable for protoplanetary disks that are known to undergo significant grain growth and photoevaporation. In fact, understanding the evo- lution of the gas (and the gas to dust mass ratio) in protoplanetary disks remains one of the main challenges in the fields of disk evolution and planet formation. The second most common molecule after H2 is CO, and its isotopologues, 13CO and C18O, provide a viable approach to estimate the gas content in disks (Williams & Best, 2014). From continuum submillimeter surveys of nearby star-forming regions such as Taurus and Ophiuchus, the following benchmark properties have been derived: Mdisk ∝ M⋆ and Mdisk ∼0.5% M⋆ (Andrews & Williams 2005, 2007; Andrews et al. 2013). However, the dispersion is large (± 0.7 dex) and the submillimeter luminosities of protoplanetary disks in the same regions, and with almost identical IR SEDs, can differ by up to two orders of magnitude (e.g., Cieza et al. 2008; 2010), implying drastically different disk masses and/or grain-size distributions. Understanding disk evolution requires both IR † For a discussion on disk nomenclature, see the Diskionary by Evans et al. (2009). 4 Lucas A. Cieza and submillimeter observations. Unfortunately, millimeter surveys of disks in molecular clouds and young stellar clusters clearly lag far behind their IR counterparts. While deep and complete IR censuses of disks exist for tens of regions spanning a wide range of ages and IR disk fractions, Taurus is the only region in which the entire disk population has been observed at submillimeter wavelengths with enough depth to detect the majority of the IR-detected disks (Andrews et al. 2013). 3.2. Resolved observations Using interferometers and imaging synthesis techniques, nearby protoplanetary disks can be spatially resolved, providing direct information on the surface density profile of the disk, which is critical for planet formation theory. Simultaneous modeling of the SED and submillimeter images is a powerful technique to constrain the basic properties of disks. Thus far, resolved submillimeter studies have been limited to continuum observations of bright sources (Fmm >50 mJy) and/or gas observations of optically thick lines. Therefore, they are very biased towards massive disks around relatively massive stars and provide little direct information on the gas content of disks. In particular, most imaging studies have focused on transition objects (see Williams & Cieza, 2011 for a review). These early resolved observations indicated that the surface density profiles in many of the bright protoplanetary disks are consistent with the amount of material needed for the formation of the planets in the Solar System (Andrews et al. 2010). With the commissioning of ALMA, we should be able to 1) extend the imaging surveys to less massive, more typical, protoplanetary disks and 2) image the disks in optically thin gas tracers such as 13CO and C18O in order to trace the surface density profile of the gas more directly. 4. ALMA Early Science results Since the start of Early Science observations in 2011, ALMA is shedding new light on every stage of disk evolution, from the deeply embedded Class I stage† to the dissipation of the primordial disk and the transition to the debris disk phase. In the following, we highlight some of remarkable ALMA Early Science results. 4.1. Rings in Class I circumstellar disks Long-baseline ALMA observations at 0.025" to 0.075" (3.5 -- 10 AU) resolution of the Class I circumstellar disk around HL Tau revealed a set of impressive concentric gaps. These gaps show evidence for grain growth, are slightly eccentric, and many of them appear to be in resonance with each other (Partnership et al. 2015). All these features are highly suggestive of advanced planet formation through core accretion, although other explanations have merit, including the growth of pebbles at the snow lines of different species (Zhang et al. 2015). If the gaps in the HL Tau disk are in fact due to orbiting planets massive enough to carve them, it would imply that the core accretion mechanism is significantly more efficient that previously thought and that the planet formation process can be fairly advanced by the Class I stage and an age of ∼1 Myr. 4.2. Structures in transition objects High-resolution Early Science observations of transition disks have also produced some surprises such as the discovery of huge asymmetries in the outer disks of Oph IRS 48 (van der Marel et al. 2013), SR 21 (Perez et al. 2014 ), and HD 142527 (Casassus et al. 2013). These asymmetries have been interpreted as large-scale vortices or dust traps, which † The Class I stage corresponds to a disk that is still embedded in its natal envelope. The structure and evolution of protoplanetary disks 5 could play an important role in the formation of planetesimals and planets at large radii. HD 142527 also shows a set of spiral arms (Christiaens et al. 2014) of unknown origin and gaseous accretion flows connecting the inner and the outer disk components of the system (Casassus et al. 2013). Such accretion flows have been predicted by hydrodynam- ical simulations of multiple planets embedded in a primordial disk (Dodson-Robinson & Salyk, 2011). All these results demonstrate that transition disks have complex structures and that the models derived from SED fitting and low-resolution images can only be considered zero-order approximations to their true structures. 4.3. The primordial to debris disk transition As discussed in §2.2, star-forming regions include very diverse types of young stellar objects: "full" disks, disks with inner-holes and gaps, optically thin disks, and diskless stars. The evolutionary status of the optically thin disks has remained a matter of intense debate over the last few years and these objects have often been considered the final stage of the primordial disk phase: evolved, primordial (gas rich) disks where most of the primordial dust has been depleted. In an ALMA survey of 24 non-accreting PMS stars (weak-line T Tauri stars; WTTSs) with known IR excesses, Hardy et al. (2015) found that these systems have dust masses . 0.3 M⊕ and no evidence of gas, and therefore are more akin to young debris disks than to evolved primordial disks. This would indicate that they are already in a debris stage, as opposed to transitioning into that phase. Non-accreting PMS with IR excesses represent ∼20% of the WTTS population (Cieza et al. 2007), and their incidence has been used to estimate the dissipation timescale of the primordial disks once accretion stops and Classical T Tauri stars (CTTSs) evolve into WTTSs (Wahhaj et al. 2010). If WTTS disks are produced by second-generation dust, it would imply that the final inside out dispersal of the primordial disk happens even faster than previously thought (∼1% vs ∼10% of the disk lifetime). These ALMA results also raise questions on how to initiate the debris disk phenomenon within a few Myr of the formation of the star and why this is seen in ∼20% of the PMS stars that have already lost their primordial disks. 5. Disk evolution and planet formation It is usually assumed that "full disks" are the starting point of disk evolution and that the transition disks with holes and gaps are more "evolved" systems. However, demographic studies (Cieza et al. 2010; Romero et al. 2012) in nearby star-forming regions suggest that there are at least two different evolutionary paths that disks can follow. Some systems develop gaps and inner-holes while their disks are relatively massive and still accreting, while other objects lose most of their disk mass while keeping a perfectly "normal" IR SED for millions of years until they reach a point in which the entire disk dissipates from the inside out in a small fraction of the disk lifetime (most likely through photoevaporation). The former group includes most of the famous transition disks found in the literature. In general, these are early-type stars (early-K to A-type) have massive disks with enough material to form the Solar System and large inner-holes, again, consistent with the size of the Solar System. Despite their high profile in the literature, such systems are not the typical young stellar object in a nearby molecular cloud or even the typical transition disk. Instead, they represent a minority of the order of 20% of the transition disk population, and 5% of the general disk population (Cieza et al. 2012). It is tempting to speculate that such systems are the minority of disks that form planetary systems with one or more giant planets. On the other hand, the "typical" disk in a young (2-3 Myr) stellar cluster seems to be a low-mass accreting CTTS with a normal 6 Lucas A. Cieza SED, but not enough mass left in the disk to form a giant planet (Cieza et al. 2015). Such systems can be considered "evolved" in the sense that they have probably already lost most of their initial disk mass and have also undergone significant grain growth and dust settling (Lada et al. 2006). Given the fact that rocky planets are ubiquitous in the Galaxy (Batalha et al. 2013), it is also tempting to speculate that these typical disks are the ones that form the more typical planetary systems, i.e., those hosting rocky planets but no giant planets. 6. Prospects for full-ALMA observations and conclusion With unprecedented sensitivity and resolution, ALMA is clearly poised to revolutionize the fields of disk evolution and planet formation in the near future. It will have an impact in all stages of disk evolution. We can expect to see many detailed studies of disks in early stages, when they still remain embedded in their natal envelope. High resolution images at long wavelengths and the use of optically thin lines allow us to zoom in and peek inside the envelope emission and reveal the disk structure. Some of the questions that the full ALMA array will address include: are rings like those of HL Tau common in Class I objects? What is their origin? Are very young protoplanetary disks gravitationally unstable? Can we image collapsing clumps in the process of forming brown dwarfs or giant planets? We can also expect to see many more detailed studies of transition systems which will investigate the origin of their intriguing features (e.g., large cavities, spiral arms, asymmetries, clumps) and their connection to planet formation. However, as discused in §5, such systems are not typical and might be related to the formation of some particular types of planetary systems. If we want to learn how more common planetary systems form, we will also need to study more common protoplanetary disks. In this regard, ALMA's sensitivity will make it possible to perform complete surveys and resolve entire disk populations in nearby molecular clouds using both continuum and gas tracers. This will help us to investigate the full distribution of disk properties (gas and dust masses, sizes, surface density profiles, etc.) as a function of stellar parameters such as mass, age, and multiplicity. In particular, the observations of optically thin line tracers such as 13CO and C18O will allow us to place much needed constraints on the evolution of gas, which is far less established than that of the dust. Thanks to the exponential growth of extrasolar planet studies and the construction of ALMA, we are approaching an era in which we can start making direct connections between disk properties and the statistics of planetary systems we see in the Galaxy. This will help us to understand the kind of disks that form different types of planetary systems and to develop a coherent picture of the demographics of protoplanetary disks and the planets they form. References Alexander, R. D., Clarke, C. J., & Pringle, J. E. 2006, MNRAS, 369, 229 Alexander, R., Pascucci, I., Andrews, S., Armitage, P., & Cieza, L. 2014, Protostars and Planets VI, 475 Andrews, S. M., & Williams, J. P. 2005, ApJ, 631, 1134 Andrews, S. M., & Williams, J. P. 2007, ApJ, 671, 1800 Andrews, S. M., Wilner, D. J., Hughes, A. M., Qi, C., & Dullemond, C. P. 2010, ApJ, 723, 1241 Andrews, S. M., Rosenfeld, K. A., Kraus, A. L., & Wilner, D. J. 2013, ApJ, 771, 129 Artymowicz, P., & Lubow, S. H. 1994, ApJ, 421, 651 Batalha, N. M., Rowe, J. F., Bryson, S. T., et al. 2013, ApJS, 204, 24 Beckwith, S. V. W., Sargent, A. I., Chini, R. S., & Guesten, R. 1990, AJ, 99, 924 Boss, A. P. 2000, ApJL, 536, L101 The structure and evolution of protoplanetary disks 7 Carpenter, J. M., Mamajek, E. E., Hillenbrand, L. A., & Meyer, M. R. 2006, ApJL, 651, L49 Casassus, S., van der Plas, G., M, S. P., et al. 2013, Nature, 493, 191 Cieza, L., Padgett, D. L., Stapelfeldt, K. R., et al. 2007, ApJ, 667, 308 Cieza, L. A., Padgett, D. L., Allen, L. E., et al. 2009, ApJL, 696, L84 Cieza, L. A., Schreiber, M. R., Romero, G. A., et al. 2010, ApJ, 712, 925 Cieza, L. A., Schreiber, M. R., Romero, G. A., et al. 2012, ApJ, 750, 157 Cieza, L., Williams, J., Kourkchi, E., et al. 2015, arXiv:1504.06040 Christiaens, V., Casassus, S., Perez, S., van der Plas, G., & M´enard, F. 2014, ApJL, 785, L12 Currie, T., Lada, C. J., Plavchan, P., et al. 2009, ApJ, 698, 1 Dominik, C., & Dullemond, C. P. 2008, A&A, 491, 663 Dodson-Robinson, S. E., & Salyk, C. 2011, ApJ, 738, 131 Evans, N., Calvet, N., Cieza, L., et al. 2009, arXiv:0901.1691 Hardy, A., Caceres, C., Schreiber, M. R., et al. 2015, arXiv:1504.05562 Hartigan, P., Edwards, S., & Ghandour, L. 1995, ApJ, 452, 736 Hartmann, L., Calvet, N., Gullbring, E., & D'Alessio, P. 1998, ApJ, 495, 385 Kraus, A. L., Ireland, M. J., Hillenbrand, L. A., & Martinache, F. 2012, ApJ, 745, 19 Lada, C. J., Muench, A. A., Luhman, K. L., et al. 2006, AJ, 131, 1574 Lissauer, J. J. 1993, ARA&A, 31, 129 Mamajek, E. E. 2009, American Institute of Physics Conference Series, 1158, 3 Partnership, A., Brogan, C. L., P´erez, L. M., et al. 2015, ApJL, 808, L3 P´erez, L. M., Isella, A., Carpenter, J. M., & Chandler, C. J. 2014, ApJL, 783, L13 Romero, G. A., Schreiber, M. R., Cieza, L. A., et al. 2012, ApJ, 749, 79 van der Marel, N., van Dishoeck, E. F., Bruderer, S., et al. 2013, Science, 340, 1199 Wahhaj, Z., Cieza, L., Koerner, D. W., et al. 2010, ApJ, 724, 835 Williams, J. P., & Best, W. M. J. 2014, ApJ, 788, 59 Williams, J. P., & Cieza, L. A. 2011, ARA&A, 49, 67 Zhang, K., Blake, G. A., & Bergin, E. A. 2015, ApJL, 806, L7
1004.4180	1	1004	2010-04-23T16:59:28	A Peculiar Family of Jupiter Trojans: the Eurybates	[ "astro-ph.EP" ]	The Eurybates family is a compact core inside the Menelaus clan, located in the L4 swarm of Jupiter Trojans. Fornasier et al. (2007) found that this family exhibits a peculiar abundance of spectrally flat objects, similar to Chiron-like Centaurs and C-type main belt asteroids. On the basis of the visible spectra available in literature, Eurybates family's members seemed to be good candidates for having on their surfaces water/water ice or aqueous altered materials. To improve our knowledge of the surface composition of this peculiar family, we carried out an observational campaign at the Telescopio Nazionale Galileo (TNG), obtaining near-infrared spectra of 7 members. Our data show a surprisingly absence of any spectral feature referable to the presence of water, ices or aqueous altered materials on the surface of the observed objects. Models of the surface composition are attempted, evidencing that amorphous carbon seems to dominate the surface composition of the observed bodies and some amount of silicates (olivine) could be present.	astro-ph.EP	astro-ph	A Peculiar Family of Jupiter Trojans: the Eurybates∗ F. De Luise1,2, E. Dotto1, S. Fornasier3,4, M.A. Barucci3, N. Pinilla-Alonso5,6, D. Perna1,3,7, and F. Marzari8 June 4, 2018 1 INAF-Osservatorio Astronomico di Roma, Via Frascati 33, 00040 Monteporzio Catone (Roma), Italy; 2 INAF-Osservatorio Astronomico di Collurania-Teramo, Via Mentore Maggini, s.n.c., 64100 Teramo (TE), Italy; 3 LESIA, Paris Observatory, 5 Place Jules Janssen, 92195 Meudon Cedex, France; 4 University of Paris VII "Denis Diderot", 4 rue Elsa Morante, 75013 Paris, France; 5 Fundaci´on Galileo Galilei & Telescopio Nazionale Galileo, PO Box 565, 38700 S/C de La Palma, Tenerife, Spain; 6 Oak Ridge Associated Universities (ORAU) - NASA Ames Research Center, MS 245-3, Moffett Field, CA 94035-1000; 7 University of Rome "Tor Vergata", Via della Ricerca Scientifica 1, 00133 Roma, Italy; ∗Based on observations made with the Italian Telescopio Nazionale Galileo (TNG) operated on the island of La Palma by the Fundacion Galileo Galilei of the INAF (Istituto Nazionale di Astrofisica) at the Spanish Observatorio del Roque de los Muchachos of the Instituto de Astrofisica de Canarias, programmes: TAC41(AOT14) and TAC69(AOT15). 1 8 Dipartimento di Fisica - University of Padova, Via Marzolo 8, 35131 Padova, Italy. Manuscript pages: 23 Figures: 2; Tables: 2 Running head: Jupiter Trojans: near-infrared investigation of Eury- bates Family Send correspondence to: De Luise Fiore INAF-Osservatorio Astronomico di Collurania-Teramo Via Mentore Maggini, s.n.c. I-64100 Teramo Italy e-mail: [email protected] phone: +39-0861-439712 fax: +39-0861-439740; 2 Abstract The Eurybates family is a compact core inside the Menelaus clan, located in the L4 swarm of Jupiter Trojans. Fornasier et al. (2007) found that this family exhibits a peculiar abundance of spectrally flat objects, similar to Chiron-like Centaurs and C-type main belt aster- oids. On the basis of the visible spectra available in literature, Eury- bates family's members seemed to be good candidates for having on their surfaces water/water ice or aqueous altered materials. To improve our knowledge of the surface composition of this peculiar family, we carried out an observational campaign at the Telescopio Nazionale Galileo (TNG), obtaining near-infrared spectra of 7 mem- bers. Our data show a surprisingly absence of any spectral feature referable to the presence of water, ices or aqueous altered materials on the surface of the observed objects. Models of the surface com- position are attempted, evidencing that amorphous carbon seems to dominate the surface composition of the observed bodies and some amount of silicates (olivine) could be present. Keywords: Jupiter Trojans – Dynamical families – Spectroscopy – Near- infrared 3 1 Introduction Jupiter Trojans (JTs) are small bodies of the Solar System located in the Jupiter's Lagrangian points L4 and L5. Their origin is not yet well understood and it is still matter of debate. Several mechanisms were proposed to model their origin (Marzari & Scholl, 1998a,b, 2000, 2007; Marzari et al., 2002; Morbidelli et al., 2005), and it is widely accepted that they formed in the outer Solar System, in regions rich in frozen volatiles. The JT population is supposed to have undergone a significant collisional evolution, and to be at least as collisionally evolved as main belt asteroids. The discovery of dynamical families in both L4 and L5 clouds supports this hypothesis (e.g. Milani, 1993; Milani & Knezevi´c, 1994; Beaug´e & Roig, 2001; Dell'Oro et al., 1998). Physical properties of JTs are poorly known. The presently available spectroscopic data set is largely unsatisfactory, covering only about 10% of the entire JT population. To improve our knowledge of the nature of these bodies, in the last years several surveys have been carried out, both in vis- ible and infrared wavelengths (Dotto et al., 2008, and reference therein). Although JTs formed at large heliocentric distances, the data so far acquired have shown a lack of any evidence of ices on their surfaces. Emery & Brown (2003, 2004) analysed the content of water ice and hydrated materials on the surface of 17 JTs, obtaining upper limits of a few% for water ice and of 30% for hydrated materials. More recently, Yang & Jewitt (2007) suggested that water ice can occupy no more than 10% of the total surface of (4709) 4 Ennomos. JTs belonging to dynamical families do not exibit any spectral feature related to the presence of ices on their surface (Dotto et al., 2006). The data so far available in the literature put in evidence a great homo- geneity in the whole population: all of the known JTs are low albedo bodies belonging to the primitive C, P or D classes. The same uniformity is also found in JTs belonging to dynamical families (Fornasier et al., 2004, 2007; Dotto et al., 2006). However, some differences between the L4 and L5 swarms are evident: as discussed by Fornasier et al. (2007), the majority of L5 JTs are D-types, while an higher presence of C and P types is observed among the L4 objects. A peculiar case is given by the Eurybates dynamical family, in the L4 swarm. This family is a strong cluster inside the Menelaus clan (Roig et al., 2008), which survives also at a very low relative velocity cut-off, as defined by Beaug´e and Roig (2001). The family population, up to date, is composed by 28 members at a cut-off of 70 m/s, 22 of them surviving also at 40 m/s. On the basis of the dynamical properties, it is still not possible to understand if the Eurybates members constitute a distinct family that lies in the same space of proper elements of Menelaus or, as suggested by Roig et al. (2008), they formed by a secondary break-up of a former Menelaus member. Although the Eurybates family is clustered, in the space of proper ele- ments, in a small portion of the region occupied by the Menelaus clan, its members show spectral properties quite different from those of Menelaus: as shown by Roig et al. (2008) the Menelaus clan evidences a larger diversity of taxonomic classes including C, P, and D-type objects in agreement with 5 the whole JT population, while the Eurybates members are characterized by almost flat visible spectra (see e.g. Fig. 12 in Roig et al., 2008), with spectral slopes strongly clustered around 2 %/103 A, and spectral behaviors similar to those of C–type main belt asteroids and/or Chiron-like Centaurs (Fornasier et al., 2007). The Eurybates family assumes a great importance in the study of JTs because such a peculiar clustering of spectrally flat objects strongly affects the color-size-orbital parameter distributions of the whole JT population investigated up to now. Fornasier et al. (2007) noted how this family fills the distribution of spectrally neutral JTs at low inclination and appears to be the major responsible of a color-inclination trend (bluer bodies concentrated at lower inclination) of the whole JT population. In the same paper, the Eurybates family appears also to be the major cause of the abundance of C- and P-types among the L4 objects, which would imply a more heterogeneous composition of this swarm than the L5 one. Moreover, the Eurybates family strongly contributes to the population of L4 small JTs (with a D<40 km) having low spectral slopes. The observations made by Fornasier et al. (2007) showed the presence of a drop-off of reflectance shortward of 0.52 µm in the visible spectra of four Eurybates members (18060, 24380, 24420 and 39285). This behavior is de- tected on the spectra of many main belt C-type asteroids (Vilas, 1994; 1995), and it is often associated to other spectral features due to aqueous alteration products. Since no other absorption features were found on the visible spec- tra of Eurybates members, we still do not have a final proof that aqueous 6 alteration processes occurred on the surface of these bodies. Nevertheless, the presence of the ultraviolet drop-off could suggest that subsurface water or water ice could have been present on Eurybates members at a certain mo- ment of their life, in order to cause aqueous alteration on their surfaces. They are therefore good candidates to preserve still detectable spectral signatures of water/water ice or aqueous altered materials. 2 Observations and Data Analysis To constrain the surface composition of Eurybates family's members, we per- formed an observational spectroscopic campaign in the near infrared (NIR) wavelength range. The observations were carried out at the 3.6 m Telescopio Nazionale Galileo (TNG) at Roque de Los Muchachos in La Palma (Canary Islands, Spain) in 2006 and 2007 (AOT14-TAC41 and AOT15-TAC69, respectively). The targets were selected using the list defined by Beaug´e and Roig (2001) and the P.E.Tr.A. project1. We observed 7 objects already investigated in visible range by Fornasier et al. (2007). With the exception of 163135, all of our targets survive at a velocity cut-off of 40 m/s. The observational circumstances are summarized in Tab. 1. We used the Near Infrared Camera Spectrometer (NICS), a multimode instrument based on a HgCdTe Hawaii 1024x1024 array, with a field of view of 4.2 x 4.2 arcmin, coupled with the AMICI prism. Our observations were carried out in low resolution spectroscopic mode, covering the 0.9–2.4 µm 1http://www.daf.on.br/froig/petra/ 7 spectral range, using a 5 arcsec wide slit, oriented in the object moving di- rection. The total exposure time was divided into several sub-spectra of 120 s each, to reduce the noise contribution typical of sky at NIR wavelengths. The observations were done by nodding the object along the slit by 30 arcsec be- tween two positions A and B. Flat-fields were also acquired at the beginning of each night. Data were reduced using the standard procedure (e.g. Dotto et al., 2006) with MIDAS and the IDL software packages. The two averaged A and B images were subtracted from each other. The A − B and B − A images were flat-fielded, corrected for spatial and spectral distortion and finally combined with a 30-arcsec offset. The spectra were hence extracted from the resulting combined images. Wavelength calibration was obtained using a look-up table, available on the TNG website, which is based on the theoretical dispersion predicted by ray-tracing and adjusted to best fit the observed spectra of calibration sources. The telluric absorption correction and the removal of the solar contribution were obtained by dividing the spectrum of each object by the spectrum of the solar analog star closest in time and airmass to the target (see Tab. 1). The resulting spectra were smoothed with a median filtering technique, to reach a spectral resolution of about 20. The edges of each spectral region were cut to avoid low S/N regions at wavelength lower than about 0.90 µm and greater than about 2.2 µm. Only for 163135 we cut the spectrum at 1.65 µm, as, due to sky variability, it was not possible to properly remove the sky contribution. The obtained NIR spectra are shown in Fig. 1. 8 Our NIR spectra were finally combined with the visible spectra published by Fornasier et al. (2007), overlapping the common region between 0.9 and 0.95 µm. The resulting V+NIR spectra, normalized at 0.55 µm, are shown in Fig. 2. We computed the spectral slopes of all the observed objects between 1.0 and 1.6 µm (see Tab. 2). The obtained values span a small range of values, from 0.11 to 3.60 %/103A, with a mean value of 1.43 ± 0.41 %/103A. The obtained spectral behaviors allow us to confirm the taxonomic clas- sification published by Fornasier et al. (2007) (see Tab. 2). Our investigation in NIR wavelengths has surprisingly shown featureless spectra: we did not detect any spectral feature between 0.9 and 2.2 µm referable to the presence, on the surface of the observed bodies, of water, ices or hydrated minerals. To model the surface composition of the observed Eurybates members we used the radiative transfer model, based on the Hapke theory, already applied to JTs by Dotto et al. (2006). We took into consideration the following materials: amorphous carbon (by Zubko et al., 1996), organic solids (e.g. kerogens by Clark et al., 1993 and Khare et al., 1991; Triton tholins by McDonald et al., 1994; titan tholins (by Khare et al., 1984) all the minerals present in the RELAB database2, bitumen (by Moroz et al., 1998), and ices (H2O, CH4, CH3OH, NH3, and ice tholin by McDonald et al., 1996 and Khare et al., 1993). To model the surface composition of each observed object, we considered several geographical mixtures of all these compounds. For each mixture, the modeling procedure produced a synthetic spectrum, to be compared with the observed one, and calculated the geometric albedo 2http://www.planetary.brown.edu/relabdocs/relab disclaimer.htm 9 value at 0.55 µm. A χ2-test was applied to compare the different models tentatively considered for each target, and to select the model which better reproduces the observed spectrum. In this analysis we did not take into account the critical regions of the spectrum, around 1.4 and 1.9 µm, where telluric bands occur. We considered as best model the geographycal mixture best fitting the asteroid spectrum, and having an albedo value compatible with the typical value of C- or P-type dark asteroids and the mean value for JTs (0.041±0.002, as computed by Fern´andez et al., 2003). In Fig. 2 the synthetic spectra of final models (continuous lines) are superimposed on the observed spectra. Table 2 reports, for each object, the model of surface composition, as well as the computed albedo. The obtained spectra suggest the predominance of amorphous carbon on the surface of the observed members of the Eurybates family. In particular, the spectra of 13862, 18060 and 163135 are similar to the one of pure amor- phous carbon. The spectral behaviors of (3548) Eurybates, 24380 and 24420 suggest the presence on their surface of a few amount of olivine. The slightly spectral reddening of (9818) Eurymachos has been modelled using a small percentage of a reddening agent (e.g. Triton tholin). 3 Discussion and conclusions All the spectra of Eurybates members presented in this work appear flat and featureless and confirm the taxonomic classification published by Fornasier et al. (2007). Our surface modeling has shown that the amorphous carbon seems to dominate the surface composition of the observed bodies and some 10 amount of silicates (olivine) could be present. The proposed models are not unique, since they depend on many parameters (e.g. physical properties of the surface, optical constants and particle size), but a complete lack of di- agnostic features typical of water, ices and hydrated minerals is evident in our spectra. This result does not allow us to definitively exclude that some percentage of water ice is still present on the observed bodies, hidden by dark materials. Brunetto & Roush (2008) showed that few tens of microns of an organich-rich layer (e.g. irradiated methan ice), produced by space weath- ering, are enough to mask spectroscopically, in the near-infrared wavelength range, the presence of water and other volatiles below the surface. The spectral evidences presented in this paper leave open several possi- bilities about the origin of the Eurybates family. A first scenario implies the formation of the Eurybates family by the disruption of an exogenous body, i.e. coming from other Solar System re- gions, probably captured by Jupiter gravitational field and trapped in L4 Lagrangian point. In this case, the origin of the parent body is a crucial point that must yet be assessed, as well as the nature and the efficiency of the capture mechanism. Of course, it is plausible that this captured parent body is not the only one in the Trojan clouds, but the population of these objects must be still assessed. Other scenarios take into account the action of space weathering pro- cesses, still efficient at 5.2 AU from the Sun where JTs are presently orbiting (see e.g. Strazzulla et al. 2005; Melita et al. 2009). We know, from labo- ratory experiments, that the effect of ageing mechanisms strongly depends 11 on the composition and nature of the surfaces exposed to space weathering. Since we still do not know the origin and primordial composition of JTs, and therefore we do not know how space weathering processes acted on JT surfaces, several scenarios have to be considered: a) if JTs had icy surfaces, the space weathering processes would have pro- duced an irradiation mantle spectrally red and with low albedo (Moore et al., 1983; Thompson et al., 1987; Strazzulla, 1998; Hudson & Moore, 1999; Brunetto et al., 2006; Brunetto & Roush, 2008); b) a similar result would have been produced on silicatic composition, where space weathering produces a gradual spectral reddening, as already observed in several dynamical families in the asteroid main belt (e.g. the Eunomia family by Lazzaro et al., 1999; the Flora clan by Florczak et al., 1998; and the Eos family by Doressoundiram et al., 1998) and shown by laboratory experiments (e.g. Strazzulla et al. 2005; Lazzarin et al. 2006); c) an opposite result would have been produced on a surface covered of natural complex hydrocarbons, where ion irradiation would have produced gradually neutralized spectra (Moroz et al., 2004). The first two cases bring to the possibility that the Eurybates is a young fam- ily, produced either by a the fragmentation of an object coming from outside the present Trojan population and trapped around L4, or by a secondary collision involving one of Menelaus' family members. Under the scenario (c), Eurybates should be an old family, with an initial hydrocarbon composition, on which space weathering flattened the members' spectra, wiping out the primordial differences. 12 Whether JTs experienced a phase of cometary activity, water ice on their surface could have been devolatized, or they could have formed a thin dust mantle as shown by Rickman et al. (1990) and Tancredi et al. (2006). This mechanism is quite probable for large JTs not belonging to dynamical families, and it is still plausible for members of dynamical families, since we cannot exclude that they suffered some episodic cometary activity during their life after the fragmentation of the parent body. In a recent work Melita et al. (2009) estimated the timescales of the sublimation of amorphous water ice, the collisional resurfacing, and the flattening of the spectral slopes by solar irradiation in the region of JTs. According to these authors, a dust layer, probably with a red spectroscopic slope, does exist on the surface of JTs as a result of the rapid sublimation of water ice after resurfacing impact events. If this dust layer remains unaltered for more than 103 years, its spectroscopic slope is flattened by the action of solar protons. According to the estimated timescales, impacts are so frequent that the irradiation mantle is usually disrupted for JTs, but in the case of the Eurybates family the flat spectra would suggest that we are seeing aged surfaces. The knowledge of the age of the Eurybates family is hence fundamental for understanding the parent body origin and nature. More observations are also absolutely needed to see whether this peculiar family is unique or if more Eurybates-like families are present in both L4 and L5 swarms. At the same time more observations of Menelaus objects are needed to com- pare Eurybates and Menelaus members with the whole population of JTs, to have more hints on the relation between Eurybates family and Menelaus 13 clan, and to cast some light on their (common or not) origin. Acknowledgments FDL want to thank the TNG support astronomers and telescope operators, in particular W. Boschin and G. Tessicini, for their support and for the kind help to optimize the observations. The authors thank G.A. Baratta for the helpful discussions. The authors thank as well both the anonymous reviewers for providing comments and suggestions helpful to improve the manuscript. 14 References - Beaug´e, C. & Roig, F., 2001. A Semianalytical Model for the Motion of the Trojan Asteroids: Proper Elements and Families, Icarus 153, 391-415 - Brunetto & Roush, 2008. Impact of irradiated methane ice crusts on compositional interpretations of TNOs (Research Note). Astron. Astrophys 481, 879-882 - Brunetto, R., Barucci, M. A., Dotto, E., Strazzulla, G., 2006. Ion Irra- diation of Frozen Methanol, Methane, and Benzene: Linking to the Colors of Centaurs and Trans-Neptunian Objects, Astroph. J. 644, 646-650 - Clark, R.N., Swayze, G.A., Gallagher, A.J., King, T.V.V., Calvin, W.M., 1993. U.S. Geological Survey Open File Report 93-592, http://speclab.cr.usgs.gov. - Dell'Oro, A., Marzari, P., Paolicchi, F., Dotto, E., Vanzani, V., 1998. Trojan collision probability: A statistical approach. Astron. Astrophys 339, 272-277. - Dotto, E., Fornasier, S., Barucci, M.A., Licandro, J., Boehnhardt, H., Hainaut, O., Marzari, F., de Bergh, C., De Luise, F., 2006. The surface com- position of Jupiter Trojans: Visible and Near-Infrared Survey of dynamical families. Icarus 183, 420-434. - Dotto, E., Emery, J.P., Barucci, M.A., Morbidelli, A., Cruikshank, D.P., 2008. De Troianis: The Trojans in the Planetary System. In: The Solar System Beyond Neptune (M. A. Barucci, H. Boehnhardt, D. P. Cruikshank, and A. Morbidelli Eds.), Univ. of Arizona Press, Tucson, 592, 383-395 - Doressoundiram, A., Barucci, M.A., Fulchignoni, M., Florczak, M., 1998. EOS Family: A Spectroscopic Study. Icarus 131, 15-31 15 - Emery, J.P. & Brown, R.H., 2003. Constraints on the surface com- position of Trojan asteroids from near-infrared (0.8-4.0 mum) spectroscopy. Icarus 164, 104-121. - Emery, J.P. & Brown, R.H., 2004. The surface composition of Trojan asteroids: Constraints set by scattering theory. Icarus 170, 131-152. - Fern´andez, Y.R., Sheppard, S.S., Jewitt, D.C., 2003. The Albedo Dis- tribution of Jovian Trojan Asteroids. Astron. J. 126, 1563-1574 - Florczak, M., Barucci, M.A., Doressoundiram, A., Lazzaro, D., Angeli, C.A., Dotto, E., 1998. A Visible Spectroscopic Survey of the Flora Clan Icarus 133, 233-246 - Fornasier, S., Dotto, E., Marzari, F., Barucci, M.A., Boehnhardt, H., Hainaut, O., de Bergh, C., 2004. Visible and near-infrared spectroscopic survey of Jupiter Trojan asteroids: investigation of dynamical families, 172, 221-232 - Fornasier, S., Dotto, E., Hainaut, O., Marzari, F., Boehnhardt, H., De Luise, F., Barucci, M. A., 2007. Visible spectroscopic and photometric survey of Jupiter Trojans: Final results on dynamical families. Icarus 190, 622-642 - Khare, B.N., Sagan, C., Arakawa, E.T., F. Suits, Callcott, T.A., Williams, M.W., 1984. Optical constants of organic tholins produced in a simulated titanian atmosphere - From soft X-ray to microwave frequencies. Icarus 60, 127-137. - Khare, B.N., Thompson, W.R., Sagan, C., Arakawa, E.T., Meisse, C., Gilmour, I., 1991. Optical constants of kerogen from 0.15 to 40 micron: 16 Comparison with meteoritic organics. In: Origin and Evolution of Inter- planetary Dust, IAU Colloq. 126 (Levasseur-Regours, A.C., Hasegawa, H. Eds.), Kluwer Academic, Dordrecht. ASSL 173, 99. - Khare, B.N., Thompson, W.R., Cheng, L., Chyba, C., Sagan, C., Arakawa, E.T., Meisse, C., Tuminello, P.S., 1993. Production and optical constraints of ice tholin from charged particle irradiation of (1:6) C2H6/H2O at 77 K. Icarus 103, 290-300. - Hudson, R.L. & Moore, M.H., 1999. Laboratory studies of the forma- tion of methanol and other organic molecules by water + carbon monoxide radiolysis: Relevance to comets, icy satellites, and interstellar ices. Icarus 140, 451-461. - Lazzarin, M., Marchi, S., Moroz, L. V., Brunetto, R., Magrin, S., Paolic- chi, P., Strazzulla, G., 2006. Space Weathering in the Main Asteroid Belt: The Big Picture. ApJ 647, 179-182 - Lazzaro, D., Moth´e-Diniz, T., Carvano, J.M., Angeli, C.A., Betzler, A.S., Florczak, M., Cellino, A., Di Martino, M., Doressoundiram, A., Barucci, M.A., Dotto, E., Bendjoya, P., 1999. The Eunomia Family: A Visible Spec- troscopic Survey. Icarus 142, 445-453 - Marzari, F. & Scholl, H., 1998a. Capture of Trojans by a growing proto-Jupiter. Icarus 131, 41-51. - Marzari, F. & Scholl, H., 1998b. The growth of Jupiter and Saturn and the capture of Trojans. Astron. Astrophys. 339, 278-285. - Marzari, F. & Scholl, H., 2000 The Role of Secular Resonances in the History of Trojans. Icarus 146, 232-239. 17 - Marzari, F. & Scholl, H., 2007. Dynamics of Jupiter Trojans during the 2:1 mean motion resonance crossing of Jupiter and Saturn. Mon. Not. R. Astron. Soc. 380, 479-488 - Marzari, F., Scholl, H., Murray, C., Lagerkvist, C., 2002. Origin and evolution of Trojan asteroids. In: Asteroids III (Bottke Jr., W.F., Cellino, A., Paolicchi, P., Binzel, R.P. Eds.). Univ. of Arizona Press, Tucson, 725- 738. - McDonald, G.D., Thompson,W.R., Heinrich, M., Khare, B.N., Sagan, C., 1994. Chemical investigation of Titan and Triton tholins. Icarus 108, 137-145. - McDonald, G.D., Whited, L.J., Deruiter, C., Khare, B.N., Patnaik, A., Sagan, C., 1996. Production and chemical analysis of cometary ice tholins. Icarus 122, 107-117. - Melita, M. D., Strazzulla, G., Bar-Nun, A., 2009, Collisions, Cosmic Radiation and the Colors of the Trojan Asteroids. Icarus 203, 134-139. - Milani, A., 1993. The Trojan asteroid belt: Proper elements, stability, chaos and families. Celestial Mechanics and Dynamical Astronomy 57, 59-94. - Milani, A., and Knezevi´c, Z., 1994. Asteroid proper elements and the dynamical structure of the asteroid main belt. 107, 219-254. - Moore, M.H., Donn, B., Khanna, R., A'Hearn, M.F., 1983. Studies of protonirradiated cometary-type ice mixtures. Icarus 54, 388-405. - Morbidelli, A., Levison, H.F., Tsiganis, K., Gomes, R., 2005. Chaotic capture of Jupiter's Trojan asteroids in the early Solar System. Nature 435, 462-465. 18 - Moroz, L.V., Arnold, G., Korochantsev, A.V., Wasch, R., 1998. Natural solid bitumens as possible analogs for cometary and asteroid organics. Icarus 134, 253-268 - Moroz, L., Baratta, G., Strazzulla, G., Starukhina, L., Dotto, E., Barucci, M.A., Arnold, G., Distefano, E., 2004. Optical alteration of com- plex organics induced by ion irradiation. 1. Laboratory experiments suggest unusual space weathering trend. Icarus 170, 214-228. - Rickman, H., Fern´andez, J. A., Gustafson, B. A. S., 1990. Formation of stable dust mantles on short-period comet nuclei Astron. Astrophys. 237, 524-535. - Roig, F., Ribeiro, A.O., Gil-Hutton, R., 2008. Taxonomy of asteroid families among the Jupiter Trojans: comparison between spectroscopic data and the Sloan Digital Sky Survey colors. Astron. Astrophys. 483, 911-931. - Strazzulla, G., 1998. Chemistry of ice induced by bombardment with energetic charged particles. In: Solar System Ices International Symposium (Schmitt, B., de Bergh, C., Festou, M. Eds.), Kluwer Academic Dordrecht, ASSL 227, 281. - Strazzulla, G., Dotto, E., Binzel, R., Brunetto, R., Barucci, M. A., Blanco, A., Orofino, V., 2005. Spectral alteration of the Meteorite Epinal (H5) induced by heavy ion irradiation: a simulation of space weathering effects on near-Earth asteroids. Icarus 174, 31-35. - Tancredi, G., Fern´andez, J. A., Rickman, H., Licandro, J., 2006. Nuclear magnitudes and the size distribution of Jupiter family comets. Icarus 182, 527-549. 19 - Thompson,W.R., Murray, B.G.J.P.T., Khare, B.N., Sagan, C., 1987. Coloration and darkening of methane clathrate and other ices by charged particle irradiation - Applications to the outer Solar System. J. Geophys. Res. 92, 14933-14947. - Vilas, F., 1994. A quick look method of detecting water of hydration in small Solar System bodies. Lunar Planet. Sci. 25, 1439-1440. - Vilas, F., 1995. Is the U-B color sufficient for identifying water of hydration on solar system bodies? Icarus 115, 217-218. - Yang, B. & Jewitt, D., 2007. Spectroscopic Search for Water Ice on Jovian Trojan Asteroids. Astron. J. 134, 223-228. - Zubko, V.G., Mennella, V., Colangeli, L., Bussoletti, E., 1996. Optical constants of cosmic carbon analogue grains. I. Simulation of clustering by a modified continuous distribution of ellipsoids. Mon. Not. R. Astroph. Soc. 282, 1321-1329. 20 TABLES and FIGURES Table 1: Observational circumstances. For each object, we report observ- ing date, exposure time, airmass and solar analog used (with correspondent airmass). Object Solar Analog Date Texp Airmass 3548 Eurybates 13862 18060 9818 Eurymachos 24380 24420 163135 18 Aug 2006 15 Jul 2007 16 Jul 2007 17 Jul 2007 17 Jul 2007 04 Aug 2007 04 Aug 2007 (s× nacq) 120 x 8 120 x 16 120 x 16 120 x 24 120 x 16 120 x 16 120 x 32 (airmass) 1.8 1.6 1.3 1.7 1.4 1.5 Hip102491 (1.9) HD210078 (1.4) HD210078 (1.3) HD210078 (1.4) HD210078 (1.3) HD210078 (1.3) 1.5-1.8 HD210078 (1.3) Table 2: For each object, the spectral slope SN IR (computed between 1.0 and 1.6 µm), taxonomic classification given by Fornasier et al. (2007), the model of the surface composition and computed albedo value at 0.55 µm are reported. The used acronyms are: AC = amorphous carbon, Tr. th. = Triton tholin, Ol = olivine. Object 3548 Eurybates 13862 18060 9818 Eurymachos 24380 24420 163135 SN IR (%/103A) 1.82 ± 0.41 0.11 ± 0.31 1.11 ± 0.27 3.60 ± 0.38 0.98 ± 0.55 2.12 ± 0.50 0.25 ± 0.55 Tax. class C C P P C C P Model Albedo 99% AC – 1% Ol 100% AC 100% AC 98% AC – 2% Tr. th. 99% AC – 1% Ol 99% AC – 1% Ol 100% AC 0.03 0.03 0.03 0.03 0.03 0.03 0.03 21 Figure 1: Near-infrared spectra of Jupiter Trojans belonging to Eurybates family. All the spectra are normalized at 1.25 µm and shifted by 1.0 in reflectance for clarity. Their mean S/N ratio value, measured at about 1.25 µm, is around 15. 22 Figure 2: Near-infrared spectra of Eurybates family members obtained by our observations, together with the visible part already published by Fornasier et al. (2007). All the spectra are normalized at 0.55 µm and shifted by 1 in reflectance for clarity. The superimposed continuous lines represent the synthetic spectra obtained modeling the surface composition. 23
1308.5221	1	1308	2013-08-23T19:59:13	Exploring Io's atmospheric composition with APEX: first measurement of 34SO2 and tentative detection of KCl	[ "astro-ph.EP" ]	The composition of Io's tenuous atmosphere is poorly constrained. Only the major species SO2 and a handful of minor species have been positively identified, but a variety of other molecular species should be present, based on thermochemical equilibrium models of volcanic gas chemistry and the composition of Io's environment. This paper focuses on the spectral search for expected yet undetected molecular species (KCl, SiO, S2O) and isotopes (34SO2). We analyze a disk-averaged spectrum of a potentially line-rich spectral window around 345 GHz, obtained in 2010 at the APEX-12m antenna (Atacama Pathfinder EXperiment). Using different models assuming either extended atmospheric distributions or a purely volcanically-sustained atmosphere, we tentatively measure the KCl relative abundance with respect to SO2 and derive a range of 4x10^{-4}-8x10^{-3}. We do not detect SiO or S2O and present new upper limits on their abundances. We also present the first measurement of the 34S/32S isotopic ratio in gas phase on Io, which appears to be twice as high as the Earth and ISM reference values. Strong lines of SO2 and SO are also analyzed to check for longitudinal variations of column density and relative abundance. Our models show that, based on their predicted relative abundance with respect to SO2 in volcanic plumes, both the tentative KCl detection and SiO upper limit are compatible with a purely volcanic origin for these species.	astro-ph.EP	astro-ph	Draft version August 14, 2018 Preprint typeset using LATEX style emulateapj v. 12/16/11 EXPLORING IO'S ATMOSPHERIC COMPOSITION WITH APEX: FIRST MEASUREMENT OF 34SO2 AND TENTATIVE DETECTION OF KCL A. Moullet1, E. Lellouch2, R. Moreno2, M. Gurwell3, J. H Black4, and B. Butler5 1National Radio Astronomy Observatory, Charlottesville VA-22902, U.S.A 2LESIA-Observatoire de Paris, 5 place J. Janssen, 92195 Meudon CEDEX, France 3Harvard-Smithsonian Center for Astrophysics, Cambridge MA-02138, U.S.A 4Department of Earth and Space Sciences, Chalmers University of Technology, Onsala Space Observatory, 43992 Onsala, Sweden and 5National Radio Astronomy Observatory, Socorro NM-87801, U.S.A Draft version August 14, 2018 ABSTRACT The composition of Io's tenuous atmosphere is poorly constrained. Only the major species SO2 and a hand- ful of minor species have been positively identified, but a variety of other molecular species should be present, based on thermochemical equilibrium models of volcanic gas chemistry and the composition of Io's environ- ment. This paper focuses on the spectral search for expected yet undetected molecular species (KCl, SiO, S2O) and isotopes (34SO2). We analyze a disk-averaged spectrum of a potentially line-rich spectral window around 345 GHz, obtained in 2010 at the APEX-12m antenna (Atacama Pathfinder EXperiment). Using different models assuming either extended atmospheric distributions or a purely volcanically-sustained atmosphere, we tentatively measure the KCl relative abundance with respect to SO2 and derive a range of 4×10−4-8×10−3. We do not detect SiO or S2O and present new upper limits on their abundances. We also present the first measure- ment of the 34S/32S isotopic ratio in gas phase on Io, which appears to be twice as high as the Earth and ISM reference values. Strong lines of SO2 and SO are also analyzed to check for longitudinal variations of column density and relative abundance. Our models show that, based on their predicted relative abundance with respect to SO2 in volcanic plumes, both the tentative KCl detection and SiO upper limit are compatible with a purely volcanic origin for these species. 1. INTRODUCTION The galilean moon Io holds a very tenuous atmosphere (∼1-10 nbar pressure at the ground) that is unusual in many respects. This is the only known SO2-dominated atmosphere (∼90% of the total pressure), and is ultimately dependent upon the volcanic activity of the moon, the most intense in the solar system. The atmosphere acts as a reservoir feeding a plasma torus in orbit around Jupiter that sweeps the upper gas layers at a rate as high as 1 ton/second (Schneider & Bagenal 2007), thus requiring a continuous and efficient atmospheric replenishment mechanism. It is yet unclear whether the bulk of the day-side atmosphere is replenished directly by the outgassing of volcanic plumes or by sublimation of SO2 volcanic condensates (see review in Lellouch 2005). The latest results obtained on SO2 spatial distribution (Feaga et al. 2009; Moullet et al. 2010) and column density variation with heliocentric distance (Tsang et al. 2012) tend to support sublimation as the major immediate source. Io's atmosphere displays marked longitudinal column density variations on the day-side, by a factor up to 10 along the equator (Spencer et al. 2005). In addition, it is expected that the pressure collapses almost entirely on the night-side and at high latitudes, where the colder surface temperature cannot maintain a significant SO2 column density through sublimation, except possibly around volcanic centers. Primarily due to instrumental sensitivity limitations the composition of Io's atmosphere has been poorly constrained obser- vationally. Beyond SO2 only three molecular species have been identified in the lower atmosphere: - SO is the main expected photochemistry product of SO2 and was first detected by Lellouch et al. (1996) with an abundance of 3-10% with respect to SO2 (i.e., mixing ratio). Measurements of significant temporal variations in SO mixing ratio, as well as mapping of SO lines emission (de Pater et al. 2007; Moullet et al. 2010) that shows a spatial distribution more localized and linked to volcanic centers than the bulk SO2 atmosphere, support the hypothesis that atmospheric SO is at least partially sustained by plume outgassing. - NaCl has been detected by sub-millimeter spectroscopy (Lellouch et al. 2003) with a mixing ratio of 0.3-1.3%, and could be the main carrier for the Na and Cl ions present in the plasma torus. This molecule has a very short lifetime in the atmosphere due to photolysis and rapid condensation on the surface (Moses et al. 2002). The presence of NaCl in volcanic plumes had been predicted by thermochemical equilibrium models of volcanic gas chemistry (Fegley & Zolotov 2000b). It is expected to be injected in the atmosphere directly through volcanic outgassing or sputtering of Na-bearing volcanic frosts by high-energy particles from Io's torus (Johnson & Burnett 1990), but not sustained via sublimation. - S2 was detected by HST over the Pele plume (Spencer et al. 2000) with a temporally variable (Jessup et al. 2004) yet significant mixing ratio (8-33%), hinting at a primarily volcanic origin. As S2 is expected to quickly condense on the surface, it should remain localized mostly around volcanic centers. On the ground, it could polymerize into S3 and S4, thought to be responsible for the observed ring-shaped red-colored deposits. 2 In parallel to these observations, numerous theoretical models predicting the composition of Io's atmosphere have been de- veloped. Summers & Strobel (1996) showed that SO2 photochemistry should produce highly volatile, poorly condensible molecular species in the lower atmosphere such as O2, SO, S2, as well as atomic species S and O, that, through atmospheric transport, can build-up in the night-side where O2 and SO may be the main constituents. Surface sputtering could also trigger the production of non- or partially-condensible atomic and molecular gases (S, K, Na, NaS, Na2O,... Wiens et al. 1997; Chrisey et al. 1988; Haff et al. 1981). An extensive literature focuses on detailed modeling of the composition of the gas mixture expelled by volcanic plumes, for a variety of temperature and pressure conditions in volcanic conduits, as well as different bulk magmatic atomic ratios, resulting in a wide range of expected species and abundances (i.e., Zolotov & Fegley 1998; Fegley & Zolotov 2000b; Moses et al. 2002; Schaefer & Fegley 2004). Assuming bulk magma composition based on Io's plasma torus and chondritic abundances, sulfur- (S2, S3, S2O), carbon- (CO, CO2, OCS), silicate- (SiO, Si), potassium- (KCl, K) and sodium-bearing molecules and atoms (NaCl, Na), among others, may be present in volcanic plumes at mixing ratios higher than 1×10−4. On the other hand, hydrogen-bearing molecules that are often found in terrestrial plumes (such as H2O and H2S) may not be present at significant levels as Io appears to be hydrogen depleted (Fegley & Zolotov 2000a). Many of the volcanic species expected on Io have a short lifetime in the atmosphere due to destruction by photochemical reactions and/or high condensibility on the cold ground (∼90-130 K), thus forming condensate deposits around the plumes. These deposits can be partially re-injected back in the atmosphere by sputtering or, for the more volatile species (such as SO2), through sublimation. By relating measurements of the mixing ratios of different species to the above models, one can in principle better charac- terize volcanic regimes, the interactions of the atmosphere with its environment (surface and plasma torus), as well as check the validity of photochemical models. A more detailed knowledge of the chemical composition of Io's atmosphere, including the detection of expected yet undetected species, is essential for the understanding of the moon as a whole. The study of Io's spectrum in the sub(millimeter) wavelength range, which contains many rotational transitions of species that are or could be present in the atmosphere, is the most promising technique to pursue for the exploration of Io's chemistry, and is now more feasible with the recent developments of very sensitive instruments. In this paper, we report on a spectroscopic exploration project carried out with the 12-m APEX antenna (Atacama Pathinder EXperiment) coupled with the APEX-2 heterodyne receiver. Spectra were obtained in the summer and fall of 2010, and targeted rotational transitions of confirmed molecules (SO2, SO), as well as expected yet undetected species (KCl, SiO, S2O) and isotopes (34SO2). We describe the data and the models used to derive column densities and mixing ratios. We present the first measure- ment of the 34S/32S isotopic ratio in gaseous SO2, a tentative detection of KCl, upper limits on SiO and S2O, and inferences on the role of volcanism for atmospheric replenishment. 2. OBSERVATIONS The data analyzed in this paper were obtained on the APEX antenna (Atacama Pathfinder EXperiment), a 12 m diameter dish located on Llano de Chajnantor, at an altitude of 5100 m, in Chile (Gusten et al. 2006). The observations were performed in August and October 2010, aggregating a total of ∼3.3 hours of on-source integration time (corresponding to ∼16.5 total hours of observation, taking into account time on the off position and overheads), with 1.4 hours on the leading hemisphere of Io (Eastern elongation, longitudes 0-180◦W), and 1.9 hours on the trailing hemisphere (Western elongation, longitudes 180-360◦W). The relevant observational parameters for each observation period are gathered in Table 1. The single-sideband Swedish Heterodyne Facility Instrument (SHeFI, Lapkin et al. 2008) was tuned to cover the 344.100- 345.100 GHz and 346.430-347.430 GHz spectral windows. This setup was designed so as to target at once multiple rotational transitions of different species: two SO2 transitions (346.523 and 346.652 GHz), two SO transitions (344.310 and 346.528 GHz), one KCl transition (344.820 GHz), one SiO transition (347.330 GHz), five 34SO2 transitions (344.245, 344.581, 344.808, 344.987 and 344.998 GHz) and five S2O transitions (344.851, 346.543, 346.804, 346.862 and 347.123 GHz). The Fast Fourier Transform Spectrometer backend (FFTS, Klein et al. 2006) provided a spectral resolution of 122 kHz (equivalent to 106 m/s at 345 GHz), which is sufficient to resolve the Doppler-broadened lines (∼600-800 m/s width). At the observed frequency, the antenna beam is ∼17" and does not resolve Io's disk (∼1.2"), so that all the presented measurements are disk-averaged spectra. The observations were performed in on-off mode, hence in the raw spectra most of the sky thermal contribution is already removed, and the instrumental bandpass response corrected for. We noticed that the continuum level could vary significantly from scan to scan, and even reach large negative values, however line emission above the continuum is not affected by the continuum variations. These variations are probably related to the contribution of a strong continuum source (Jupiter) in the sidelobes. In the scans that are the most affected, the spectra can also display a large variability of the bandpass response on scales of a few hundred MHz (ripples). These issues prevent us from correctly estimating Io's continuum emission level, but still allow us to measure the emission contrast of each line. To extract the spectra, using the GILDAS-CLASS package1, we carefully removed the continuum emission in each scan after fitting the continuum baseline by a polynomial (of order lower or equal to 2). A few scans where the continuum level could not be well estimated were flagged and removed. To account for Io's angular size variation from August to October 2010, we rescaled the October 2010 data by normalizing each scan's antenna temperature to the equivalent antenna temperature for the geocentric distance of August 2010. We then combined together the obtained scans by averaging them on a common rest frequency scale, taking into account the line-of-sight projected velocity of the source at the time of each scan. The continuum-subtracted spectrum obtained is measured in the TA ⋆ scale, corresponding to the antenna temperature corrected 1 http://www.iram.fr/IRAMFR/GILDAS/ Date Angular size Central longitudes On-source time Rms on spectrum August 20, 2010 August 21, 2010 October 13, 2010 October 15, 2010 Combined spectrum (") 1.226 1.227 1.247 1.242 1.226 (◦W) 90-150 293-315 279-312 300-329 (h) 1.4 0.5 0.7 0.7 3.3 (mK/channel) 15 20 23 23 9.5 3 Io's observational parameters at the time of observations. To obtain the combined spectrum, the antenna temperature scale of the individual scans is rescaled to the same apparent angular size, before averaging all the scans together TABLE 1 . The rms is determined by the uncertainty on the spectrum continuum baseline around 344.8 GHz, using the original 122 kHz spectral resolution. for atmospheric absorption. To convert this scale to a disk-averaged brightness temperature scale, we multiplied the TA ⋆ values by the dilution factor (taking into account Io's ephemeris size and the beam size at each frequency), and by the antenna efficiency equal to Be f f /Fe f f , where the telescope forward efficiency Fe f f is estimated at 0.95 and the main-beam efficiency Be f f at 0.73 (Gusten et al. 2006). We estimate an error of 5% on the determination of the brightness temperature scale. After combining all the obtained scans, the rms on the spectrum obtained (combined spectrum) is 9.5 mK at 344.8 GHz on the antenna temperature scale (for each spectral channel), corresponding to 3.8 K in equivalent disk-averaged brightness temperature on Io. When binned to obtain a spectral resolution of 488 kHz (424 m/s at 345 GHz), closer to the expected line-widths, the rms decreases to 4.8 mK (1.9 K in equivalent disk-averaged brightness temperature). 3. ATMOSPHERIC MODELING 3.1. Radiative transfer To create synthetic atmospheric disk-averaged lines against which to compare the data, we use a numerical radiative transfer model described in detail in Moullet et al. (2008). For a given transition, considering the local gas temperature, density and velocity in each cell of a 3-D spatial grid encompassing the whole atmosphere, the code computes each local rotational line opacity profile, using spectroscopic parameters retrieved from the Splatalogue catalogue2 and the CDMS database3. Pressure broadening of the line profile is neglected at the low pressures considered here, so that the line profiles are assumed to be only thermally and dynamically Doppler-broadened. The brightness temperature corresponding to each line-of-sight direction is then computed through the corresponding atmospheric column, under the assumption that the emission is in local thermodynamical equilibrium. A continuum thermal surface emission at the brightness temperature of 95 K is assumed, corresponding to the average brightness temperature measured at ∼345 GHz (Moullet et al. 2010). The synthetic disk-averaged line is finally derived by summing all local brightness temperatures in the grid and rescaling the result to Io's disk size. 3.2. Extended atmospheric distributions (hydrostatic case) We first consider the case of an atmospheric global structure mimicking a horizontally extended atmosphere in hydrostatic equilibrium - akin to the case of a sublimation-sustained bulk atmosphere. The primary purpose of this modeling work is to determine the disk-averaged SO2 column density and the abundances of the observed minor species with respect to SO2 (i.e., mixing ratios). We use a computing grid with 30 km resolution in the plane of sky, and 0.25 km resolution in altitude. The airmass at each cell is computed assuming plane parallel geometry. Our atmospheric distribution model assumes that SO2 and all other species are co-located vertically and horizontally. Emission on the limbs, corresponding to the terminator region, is ignored since sublimation is unlikely to support significant amounts of SO2 on the night-side. Otherwise, we assume either a horizontally constant column density d (homogeneous distribution model) or a spatially variable column density, taking as a reference the atmospheric distribution model proposed by Feaga et al. (2009) based on HST disk-resolved SO2 observations. The latter model proposes an atmosphere restricted to longitudes lower than 70◦, with column densities increasing towards the equatorial region. It also displays significant longitudinal variations (from 1×1016mol.cm−2 to 5×1016mol.cm−2 at the equator), with lower column densities on the sub-jovian hemisphere than on the anti-jovian hemisphere. We utilize the Feaga model distribution as a reference for relative column density variations across the disk, but allow the overall distribution to be scaled by a single parametric factor χ. For both homogeneous and Feaga distribution models, the vertical density profile within each atmospheric column is computed assuming hydrostatic equilibrium, using the SO2 scale height for all species. We assume that the atmospheric temperature T is uniform over the entire atmospheric column, and horizontally constant across the disk. Doppler-shifts produced by atmospheric winds (and Io's solid rotation of 75 m/s at the equator) are also included. While most dynamic models of Io's atmosphere predict a day-to-night global circulation driven by pressure gradients (Ingersoll 1989; Austin & Goldstein 2000), the only detection of planetary winds on Io showed a structure similar to a prograde zonal wind, that can also conveniently explain the observed width of rotational lines (Moullet et al. 2008). We hence introduce in our model a prograde zonal wind, characterized by the value of its equatorial velocity V, and for which, similarly to a solid rotation, velocity varies as a function of latitude as cos(lat). The projection of the zonal wind velocity on the line of sight is computed for each atmospheric column. We use the minimum χ 2 method on a grid of parameters to estimate the best set of atmospheric parameters (d or χ, T, V) that can 2 www.splatalogue.net 3 http://www.astro.uni-koeln.de/cdms/catalog 4 reproduce each observed line. The line contrast is mainly dependent on T and d, and, if the lines are not saturated, the Doppler line-width is mostly controlled by T and V. To break the possible degeneracy between retrieved parameters, when possible, we fit together different lines corresponding to distinct rotational transitions of the same molecule. Indeed, for a given molecule, the relative contrast between two lines of different intrinsic strength is mostly related to column density d, thus allowing for a quasi-independent determination of d. Here, by fitting alltogether the absolute and relative line contrasts and widths on the two strong SO2 lines at 346.523 and 346.652 GHz, we obtain an optimal solution for dS O2 (or scaling parameter χ in the case of non-uniform distribution), allowing for a quasi-independent determination of T and V. Assuming that all species share the same T and V parameters and are co-located with SO2, we can then independently adjust for each minor species disk-averaged column density (and hence mixing ratio q). 3.3. Volcanically-sustained case To investigate the possible role of volcanism for minor species (SO, KCl, SiO and S2O), we use a different atmospheric model than the one described above, based on realistic models of gaseous plumes in a rarefied environment developed by Zhang et al. (2003) to simulate a plume-sustained atmosphere (see details in Moullet et al. 2008). These models do not include a background sublimation-sustained SO2 component, so they only represent the direct volcanic contribution to the atmosphere. The plume models fix the temperature and wind fields within a volcanic plume, as well as the spatial distribution of SO2, corre- sponding to a canopy-shaped ballistic structure. The SO2 total content in the plume, that ultimately relates to the SO2 production rate from the vent, is also fixed so as to match SO2 column densities measured over plumes by McGrath et al. (2000) and Jessup et al. (2004). Two types of plumes are considered corresponding to those identified on Io (McEwen & Soderblom 1983): Pele-type plumes, which are very large (∼600 km radius) and rather rare - and the more common Prometheus-type (∼200 km radius). Hence, the SO2 emission from a given plume type (Prometheus or Pele) is entirely specified from the plume model (with no free parameters). Within each plume, the only adjustable parameter is the minor species mixing ratio q in the plume. By using a single q parameter for the whole plume, we hence assume that SO2 and all other species share the same spatial distribution in the plume. The total volcanic emission from an observed hemisphere can be modeled using a variable number of simultaneously active plumes of each category, placed at different locations on the disk. Our radiative transfer model allows simulation of any specified distribution of such plumes. In particular, we consider as the reference plume distribution the case where all plumes observed and localized by Galileo are active (Geissler & Goldstein 2007). This reference distribution includes 16 plumes, concentrated in particular on the anti-jovian side, two of which being Pele-type plumes while the others are Prometheus-type plumes. The corresponding volcanically-sustained atmospheric distribution hence exhibits large spatial variations of temperature and column density across the disk. Volcanic models are used here to fit the SO, KCl and SiO observations, using different plausible plumes distributions, rang- ing from the most favorable case (reference plume distribution), to a minimal case where only one small plume near the sub-earth point is active. For each plume distribution case, we determine the best fitting q parameter in plumes that would be needed to reproduce the observed line contrast (or the upper limit on line contrast). The range of retrieved q parameters is then compared, for each species, to the range of mixing ratios predicted by thermochemical equilibrium models of volcanic gas chemistry. The goal eventually is to determine what fraction of the minor species' atmospheric content can plausibly be directly sustained by active volcanic plumes. 4. SO2 AND SO: LONGITUDINAL DISTRIBUTION 4.1. SO2 As measured on the combined spectrum (average of all obtained scans), the two SO2 lines at 346.652 and 346.523 GHz offer a peak signal-to-noise ratio (SNR) of 14 and 10 per spectral channel respectively. The best fitting parameters obtained with the extended atmosphere models are gathered in Table 2 (line 'combined'). For the homogeneous distribution model, we derive a column density d=5.5(±0.7)×1015mol.cm−2, a temperature T=220±20 K, and a wind velocity V=230±30 m/s. The addition of a prograde wind appears to be necessary to reach a satisfactory fit, and the V values found are compatible with the direct wind measurement of Moullet et al. (2008). The T values retrieved are typical of findings from (sub)millimeter and UV spectral data (Jessup et al. 2004), but higher than the upper limit derived from IR data (∼160 K, Tsang et al. 2012). The data are also well reproduced for a column density distribution very similar to the Feaga model (χ=0.9). We notice that the disk-averaged column density in our rescaled Feaga model is somewhat higher than the value for an homogeneous distribution (8.3 versus 5.5×1015mol.cm−2), since the former contains higher-density regions in which lines are saturated. The best model for an homogeneous atmospheric distribution is plotted against the observed lines in Figure 1. For these strong SO2 lines, the quality of the data is sufficient to split the dataset into five (5) portions covering ∼30◦ of ro- tational phase each (see Table 2). The same modeling approach was performed on each portion in order to estimate atmospheric longitudinal variations. On these shorter datasets, uncertainties on the best fitted parameters reach ∼1.2×1015mol.cm −2 on d, 40 K on T, and 50 m/s on V. Comparing the results obtained on different dates, it is not straightforward in this dataset to distinguish between longitudinal and temporal variations, as the data were obtained on two observation runs three months apart. The only direct temporal comparison can be done between the August 21 and October 13 datasets, that span approximately the same longitudes, and give compatible modeling results. SO2 disk-averaged column densities on both hemispheres appear to be of the same order, although they tend to be smaller on the Homogeneous model Date Central longitudes Combined August 20, 2010 August 20, 2010 August 20, 2010 August 21, 2010 October 13, 2010 October 15, 2010 (◦W) 90-114 115-150 90-150 293-315 279-312 300-329 October 13-15, 2010 279-312/ 300-329 Column density (1016mol.cm−2) 0.55 0.65 0.65 0.70 0.50 0.65 0.35 0.45 Temperature Wind Speed Scaling factor (K) 220 280 240 230 190 150 260 190 (km/s) 230 180 240 250 110 110 90 170 0.9 0.9 0.7 0.8 1.1 0.9 0.9 0.9 Feaga model Temperature Wind Speed (K) 260 350 320 330 220 190 290 300 (m/s) 150 60 140 130 40 40 20 20 5 Best-fit SO2 column density (or scaling factor χ), temperature and zonal wind velocity assuming homogeneous and Feaga distribution models, for different TABLE 2 portions of the dataset. ) K ( t s a r t n o c m u u n i t n o c - o t - e n L i 70 60 50 40 30 20 10 0 -10 -2 SO2 @ 346.523 GHz SO2 @ 346.652 GHz Homogeneous model -1.5 -1 -0.5 0 0.5 1 1.5 2 Velocity (km/s) Fig. 1. -- SO2 lines measured on the combined spectrum (average of all scans), plotted against the best synthetic line models for an homogeneous atmospheric distribution. trailing hemisphere than on the leading hemisphere. With scaling factors χ between 0.7-1.1, our rescaled Feaga models roughly follow the longitudinal column densities variations trend proposed by the original Feaga distribution model. We find the only significant departure from the original Feaga model between central longitudes 90◦W and 150◦W, where our best models propose a quasi-constant disk-averaged column density while the original Feaga model displays a ∼20% increase. The prograde wind velocities appear to be systematically higher on the leading hemisphere than on the trailing hemisphere, where, at least for the case of rescaled Feaga models, the addition of a zonal wind is actually not required to provide a good fit. We note that a prograde zonal wind has been directly observed only on the leading hemisphere (Moullet et al. 2008). Temperatures tend to be lower on the trailing hemisphere than on the leading hemisphere, with the exception of the high tempera- tures obtained for the October 15th dataset. This particular dataset spans a large portion of the sub-jovian hemisphere where lower SO2 column densities (by a factor 2-4) are measured, a predicted consequence of daily eclipses (Feaga et al. 2009; Walker et al. 2012) which may make it more sensitive to plasma collisional heating (Wong & Smyth 2000). 4.2. SO To derive the SO mixing ratio, we average on a common velocity scale the 344.310 and 346.528 GHz SO lines measured on the combined spectrum, assigning an equal weight to each transition. On the resulting averaged line, a SNR of 9 per channel is reached on the line contrast (Figure 2). We consider the atmospheric models discussed in Section 3.2 and assume that SO and SO2 are co-located. For a rigorous comparison, the radiative transfer code is run for both SO transitions, and the two obtained synthetic SO lines are averaged before being compared to the average observed SO line. We find the best line fit for a SO mixing ratio of q=0.07±0.007 (homogeneous distribution) or q=0.06±0.007 (rescaled Feaga distribution), corresponding to a disk-averaged SO column density of respectively 3.8 and 4.5×1014mol.cm−2. Given the data quality, it is possible to split the data in 3 different portions - August 20, August 21 and October (13 and 15) - to look for mixing ratio variations. The values found (respectively q=0.06, 0.05 and 0.08 for the homogeneous case), given the error bars (∼0.012), do not support significant mixing ratio variations, and are compatible with a scenario where SO is continuously present and detectable in the atmosphere. These values fall in the mixing ratio range derived from previous (sub)mm observations (q=0.03-0.1 Lellouch et al. 1996; Moullet et al. 2010). The assumption on co-location between SO and SO2 in our extended atmosphere models could corre- 6 ) K ( t s a r t n o c m u u n i t n o c - o t - e n L i 30 25 20 15 10 5 0 -5 -2 SO combined line Homogeneous model (q=0.07) Volcanic model (q=0.1) -1.5 -1 -0.5 0 0.5 1 1.5 2 Velocity (km/s) Fig. 2. -- Average of the two SO lines measured on the combined spectrum , plotted against the best synthetic line model for an homogeneous atmospheric distribution and for a volcanic model assuming the reference plume distribution and qS O=0.1. The volcanic model line appears to be slanted towards redshifted velocities, as it is dominated by the emission of plumes' infalling gas. spond to the case of SO being entirely sustained by SO2 photolysis, for which SO mixing ratios lower than 0.15 are indeed predicted (Summers & Strobel 1996). However several arguments in favor of an additional volcanic source for SO have been proposed, including a more localized and plume-linked distribution for SO compared to SO2 (Moullet et al. 2010) and the detec- tion of temporal mixing ratio variations (de Pater et al. 2007). While our disk-averaged spectrum does not carry information on SO spatial distribution, we can investigate the possibility of a purely volcanically sustained SO atmosphere with our volcanic models. As explained in section 3.3, we assume plausible plume distributions, and derive the SO mixing ratio in plumes needed to reproduce the observed SO line. Assuming the most favorable case where all known plumes are active (reference plume distribution), the line contrast observed is reached for qS O=0.7, which is a much larger value than the mixing ratios predicted from volcanic thermochemical models (<10%, e.g., Zolotov & Fegley 1998). Consequently, even higher SO mixing ratios would be required if fewer plumes were active. These results demonstrate that the SO content measured in our observations cannot be entirely sustained by plume activity, at least in the context of the latest thermochemical models. To determine an upper limit to the volcanic contribution to SO total content, we simulate the emission produced in the most favorable yet realistic volcanic case, by fixing the SO mixing ratio in the plumes at q=10% (corresponding to the upper range of predicted mixing ratios) and using the reference plume distribution. We then find that, within our assump- tions on realistic SO abundance and plumes distribution, volcanic SO emission could only explain up to 30% of the total emission observed. Hence our data are in agreement with the interpretation of previous (sub)mm observations, in which direct volcanic input of SO can only sustain a fraction of the SO content in the atmosphere, while SO2 photolysis is the main source for SO. 5. NEW DETECTIONS AND UPPER LIMITS 5.1. First detection of gaseous 34SO2 While the most common isotope of sulfur, 32S, has been repeatedly detected both in solid and gaseous compounds, the sulfur isotopic composition has been only poorly constrained. A detection of the second-most common sulfur isotope (34S) was only reported once, in a 4 µm 34SO2 frost band (Howell et al. 1989). However, the 34S/32S ratio could not be derived, as the observed bands of 32SO2 were too saturated to directly compare bands depths. To attempt the first determination of the 34S/32S ratio on Io, we averaged on a common velocity scale 5 different rotational lines of 34SO2 that were each marginally detected on the combined spectrum, assigning an equal weight to each transition. The resulting average line is detected with a SNR of ∼4 on the line contrast (Figure 3). In a similar way than for SO, the data is compared to the average of the same 5 34SO2 synthetic lines produced by the radiative transfer code. The best 34SO2/32SO2 ratio derived is of 0.095±0.015 assuming an homogeneous distribution, and 0.08±0.015 assuming the rescaled Feaga distribution 5.2. Tentative detection of KCl The 344.820 GHz KCl line is tentatively detected (Figure 4). At the original spectral resolution (top panel), a SNR per channel of just ∼2 is reached; however, when the data are spectrally binned down to a resolution of 488 kHz, corresponding to ∼half the linewidth (middle panel), a SNR of 3 per channel is reached. Assuming that SO2 and KCl are co-located (extended atmosphere model), the KCl line is well modeled using a KCl mixing ratio of (5±2)×10−4 assuming an homogeneous distribution and (4±2)×10−4 for a rescaled Feaga distribution, both corresponding to a KCl disk-averaged column density of ∼(3±1)×1012mol.cm−2. To explore the case of a purely volcanic origin for KCl, we derived the range of KCl mixing ratios in plumes that is compatible with the data, as explained in section 3.3. In the most favorable situation where all known plumes are erupting (reference plume 34SO2 lines combined Homogeneous model (q=0.095) 7 10 8 6 4 2 0 -2 ) K ( t s a r t n o c m u u n i t n o c - o t - e n L i -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 Velocity (km/s) Fig. 3. -- Average of the five 34SO2 lines measured on the combined spectrum, plotted against the best synthetic line models for an homogeneous atmospheric distribution. distribution), a KCl mixing ratio as low as 2×10−3 is sufficient. With such a distribution, the major part of the line emission is produced by the two large Pele-type Tvashtar and Pele plumes. Due to their geographic location (high northern latitudes on the leading hemisphere for Tvashtar, and as close as 15◦ from the western terminator on the trailing hemisphere for Pele), these plumes appear near the limbs, which is the geometry that most enhances the line emission. With a single erupting Pele plume near the sub-earth point (or 5 to 10 active Prometheus plumes distributed across each hemisphere), a KCl mixing ratio of 8×10−3 in plumes would be needed. With fewer active plumes, the observed contrast cannot be reached, as the KCl line saturates for mixing ratios higher than 8×10−3. The extended atmospheric model proposed above also reproduces the contrast of the 253.271 GHz KCl line tentatively detected at the IRAM-30m antenna on January 2002 (see Figure 4, bottom panel Lellouch et al. 2003). Those observations were interpreted, based on the comparison with NaCl lines, as a maximum KCl/NaCl ratio of 1 for a NaCl mixing ratio of 2.5-5×10−4 (corre- sponding to a maximum disk-averaged KCl column density of 2.75-5.5×1012mol.cm−2 assuming that all species are colocated). While the column density values appear to be consistent with the result obtained with our extended atmosphere model, a direct comparison of KCl mixing ratios cannot be drawn as the SO2 atmospheric model used in Lellouch et al. (2003) is quite different from our model; in particular it proposes a spatially concentrated and denser atmosphere to account for the large linewidths, as opposed to a global atmospheric distribution in our models. 5.3. Upper limits on SiO and S2O The search for the strong 347.330 GHz SiO line was unsuccessful. With a rms of 2 K on the spectrally binned data, the 3-σ upper limit on the SiO mixing ratio is of ∼1.3×10−3 for an homogeneous distribution and ∼0.8×10−3 for a rescaled Feaga distribution, corresponding to an upper limit on the disk-averaged column density of ∼7×1012mol.cm−2. We used volcanic models to determine a range of SiO mixing ratios in plumes that would just reach the measured upper limit on the line contrast, assuming that SiO is purely maintained in the atmosphere by active plumes. In the most favorable case (reference plume distribution), the upper limit would be reached for an SiO mixing ratio in plumes of 6×10−3. With a single Pele plume near the sub-earth point, a mixing ratio of 2×10−2 would be required, and with a single Prometheus plume, a mixing ratio of 0.7. Finally, even when averaging the five spectral regions where S2O transitions are expected within our observed spectral windows, S2O could not be detected, only allowing us to a put a very loose mixing ratio upper limit of q∼0.6. 6. DISCUSSION 6.1. Interpretation of the 34S/32S isotopic ratio result The main observational result in this paper is the first reported detection of a 34S-bearing molecule in gas phase. The range of values that we derive for the 34S/32S abundance ratio (within error bars, 0.065-0.120) is surprisingly higher than what was expected, as compared with the solar system reference 34S/32S value of 0.044 (Lodders 2003). While Howell et al. (1989) could not estimate the 34S/32S abundance ratio, they detected a 33SO2 band that allowed them to measure the 33S/34S abundance ratio in the solid phase to be 0.13±0.07. This result is consistent with the solar system reference value (0.18, Lodders 2003), suggesting that the global sulfur isotopic composition on Io may be similar to the solar system reference. Our measurement of the 34S/32S abundance ratio in the gas phase appears to be roughly twice as high as the solar system ref- erence, as well as solar and ISM values (Asplund et al. 2009; Lucas & Liszt 1998). A 34S/32S abundance ratio as high as our 8 10 10 ) ) [email protected] GHz [email protected] GHz K K ( ( t t s s a a r r t t n n o o c c m m u u u u n n i i t t n n o o c c - - o o t t - - e e n n L L i i ) K ( t s a r t n o c m u u n i t n o c - o t - e n L i ) K ( t s a r t n o c m u u n i t n o c - o t - e n L i -4 -4 -3 -3 -2 -2 -1 -1 0 0 1 1 2 2 3 3 4 4 Velocity (km/s) Velocity (km/s) [email protected] GHz Homogeneous model (q=5e-4) Volcanic model (q=2e-3) -4 -3 -2 -1 0 1 2 3 4 Velocity (km/s) [email protected] GHz Homogeneous model (q=5e-4) 5 5 0 0 -5 -5 6 4 2 0 -2 10 5 0 -5 -10 -4 -3 -2 -1 0 1 2 3 4 Velocity (km/s) Fig. 4. -- Top: KCl line tentatively detected at APEX with the original spectral resolu- tion. Middle: KCl line tentatively detected at APEX (with a spectral resolution degraded to 488 kHz), plotted against the best models for an homogeneous atmospheric distribution (q=5×10−4) and a purely volcanic model (q=2×10−3in plumes). Bottom: KCl line ten- tatively detected at IRAM-30m (Lellouch et al. 2003), plotted against the homogeneous atmospheric distribution model. 9 measurement has actually only been reported in redshifted absorption at z ∼ 0.89 toward a quasar (Muller et al. 2006). There are different ways to understand our result, depending on how and where fractionation occurs. First, the atmospheric 34S enrichment observed may directly reflect the sulfur isotopic composition in ionian magmas. This could be the same as the ionian bulk sulfur isotopic distribution, or result from fractionation occurring within the lava. 34S isotopic anomalies have been reported in terrestrial volcanic rocks, and are thought to be linked to variations in magma temperature and composition (Marini et al. 1994; Sakai et al. 1982). Anomalies have also been measured in chondrites, achondrite and iron meteorites (Gao & Thiemens 1993a,b; Rai et al. 2005; Gao & Thiemens 1991). However, the anomalies reported in terrestrial and meteoritical materials are at most a few per hundred. Another possibility is that mass-independent fractionation processes at work in the atmosphere significantly change the gas-phase isotopic composition with respect to the solid-phase isotopic distribution. SO2 photolysis, chemical reactions and atmospheric transport have for example all been invoked to explain the enrichment in 34S observed in some terrestrial volcanic gases (Farquhar et al. 2001; Baroni et al. 2007), although the isotopic anomalies reported and modeled in those cases are of the order of a few per thousand only. Finally, since the immediate source of SO2 is dominated by sublimation equilibrium, it could be possible that sublima- tion/condensation cycles produce mass-dependent fractionation between 32SO2 and 34SO2, which are expected to have slight differences in equilibrium vapor pressure. These three possibilities may all be important to varying degree, but remain specula- tive and require quantitative evaluation to verify if they could account for the 34S enrichment that we measure. 6.2. Potassium chloride The accumulation of tentative detections of KCl (sub)millimeter lines is an encouraging sign for definitive KCl detection using more sensitive instruments. The presence of K-bearing molecules in Io's atmosphere is strongly anticipated, as K atoms are sus- pected to be present in the upper atmosphere based on the spectral analysis of UV-airglows observed by Cassini (Geissler et al. 2004), and have been detected in Io's corona (Brown 2001). In addition, K atoms and K-bearing dust have been detected respec- tively in Io's neutral clouds (Trafton 1975) and in Jupiter's rings (Postberg et al. 2006), that are both believed to be ultimately fed by Io's atmosphere. Using terrestrial volcanism analogies, Fegley & Zolotov (2000a), Moses et al. (2002), Schaefer & Fegley (2004) and Schaefer & Fegley (2005) suggest that potassium is originally present in Io's erupting magma, from which it is easily volatilized due to the high temperatures of volcanic regions and low atmospheric pressures. KCl is the expected main K-carrier in these models and could be injected to the atmosphere directly from volcanic vents. Assuming Cl/S and K/S abundance ratios of 0.045 and 0.005, respectively, Schaefer & Fegley (2005) predict a KCl mixing ratio in the plumes of ∼8×10−3 for a wide range of conduit pressures. They also predict that KCl should have a short lifetime in the atmosphere (< 2 hours) due to condensation and photolysis. In regions where the bulk atmosphere is thinner, for example because of low sublimation-sustained SO2 (nightside), gaseous K-bearing molecules could also be produced by sputtering of volcanic-ash coated surfaces by high-energy particles, in particular on the trailing hemisphere that is hit by plasma torus particles (Haff et al. 1981). In this context, most of the KCl content on the dayside should correspond to a short-lived volcanically sustained atmosphere, indicative of on-going volcanic activity. Our volcanic modeling results on KCl are consistent with this hypothesis. We have shown that assuming the KCl mixing ratio proposed by Schaefer & Fegley (2005) (∼8×10−3), a plausible distribution of active plumes could explain the tentatively detected KCl line. Our data are compatible with volcanism being the only KCl source on the dayside atmosphere. Considering the plume distribution models proposed in Section 5.2 to explain the KCl tentative detection, we derive the corre- sponding total volcanically produced KCl. Given the SO2 emission fluxes fixed by our individual plume models (1.1×104 kg/s for a Prometheus plume, 2.2×104 kg/s for a Pele plume), we estimate a production rate ranging between 7.8×109 and 4×1010 KCl molecules/s.cm−2 (accounting for the emission from both hemispheres). We can compare this result to atmo- spheric escape rates measured at higher altitudes. Mendillo et al. (2004) measured the Na escape rate during volcanically active periods to be between 1.8-5.6×109 Na atoms/s.cm−2. Assuming the Na/K ratio measured in the corona by Brown (2001) (10±3), the K atoms escape rate inferred is of one to two orders of magnitude lower than our proposed volcanic KCl production rate. Our models are compatible with volcanically-produced KCl in Io's bulk atmosphere being the sole source of K-atoms for the corona. This result is consistent with the general behavior that atmospheric escape represents only a small fraction of the volcanic output, and in particular with the 50-100 times larger NaCl volcanic input compared to the upward flux of atomic Na and Cl at the top of the atmosphere (Moses et al. 2002; Lellouch et al. 2003). In addition to these results on the possible contribution of volcanism to atmospheric KCl content, our data allows us to constrain the Na/K atmospheric ratio. As explained in Schaefer & Fegley (2004), obtaining a stringent determination of the Na/K ratio in the bulk atmosphere can be a powerful tool to distinguish between different types of lavas (silicate, basalt, K-rich) and constrain vaporization temperature. In the lower atmosphere, where most of gaseous K and Na is believed to be carried in KCl and NaCl respectively (Fegley & Zolotov 2000a; Moses et al. 2002; Schaefer & Fegley 2005), this can be done by measuring the NaCl/KCl ratio. Our observed spectral windows did not include NaCl transitions that would allow a direct comparison. Using the 338.021 GHz NaCl line detected at the SMA (Moullet et al. 2010), we derive disk-averaged NaCl column densities between 7-9×10−12mol.cm−2, leading to a NaCl/KCl ratio of 2.7+1.8 −1 . This result is consistent and close to the Na/K value proposed for the upper atmosphere by Geissler et al. (2004) (∼3.3). Both these measurements are much lower, by several orders of magnitude, than most Na/K ratios modeled for gases vaporized over lavas that could be Ionian analogues (Schaefer & Fegley 2004), and significantly lower than cosmic and chondritic values (16 10 and 15 respectively, Asplund et al. 2009; Lodders 2003). Schaefer & Fegley (2004) address this issue in detail by proposing that, since Na is more volatile than K, fractional vaporization effects could over time lead to a depletion of the Na content in lavas and therefore in the atmosphere. Another proposition is the presence of originally K-rich lavas (ultra-potassic lavas with Na/K down to ∼2.8). We also note that our Na/K estimate is lower than the value measured in the corona based on Na and K atoms (10±3, Brown 2001). This apparent Na/K ratio increase from the bulk atmosphere to Io's environment needs to be confirmed by higher- precision measurements, as it would be indicative of processes acting differently on K-bearing and Na-bearing molecules and atoms. Brown (2001) actually argue that their corona measurement should be considered as an upper limit on the Na/K ratio near the surface, as K atoms are more quickly photo-ionized than Na atoms. 6.3. Silicon oxide The present understanding of Io's volcanism is that both silicate-based and sulfur-based magmas are driving volcanic activity (Williams & Howell 2007). The very high temperatures measured in some volcanic regions (Lopes-Gautier et al. 1999) are much higher than the sulfur boiling point but consistent with silicate flows akin to those commonly found on Earth (basalts), or to ultramafic lavas (McEwen et al. 1998). The rigidity of the surface topographic features, including several km high mountains (Clow & Carr 1980), also points to a silicate-based crust that would explain Io's bulk density. While sulfur-bearing volatiles can be dissolved in silicate melts (Carr et al. 1979), the overall predominance of sulfur on Io's surface, atmosphere and environment, suggests the possibility of co-existent sulfur-based magmas. Most measured hot-spot temperatures are indeed low enough to be consistent with molten-sulfur magma (McEwen & Soderblom 1983), and the morphology of some bright volcanic regions is indicative of sulfur flows (McEwen et al. 2000). No Si-carrying molecules have been positively identified on Io, and such a detection would provide an additional strong proof of the existence of silicate volcanism on Io. The science case behind the interpretation of SiO content is similar to that of KCl. Vaporization models show that SiO should be the main Si-bearing volatile over silicate lavas at very high temperature (Schaefer & Fegley 2004), and that it is expected to condense very quickly, so that its presence should be restricted to active volcanic plumes. However, unlike KCl, the amount of vaporized SiO depends strongly on lava temperature, with column densities above lavas expected to span four orders of magnitude (for a temperature range between 1700-2400 K). Based on the examples given in Schaefer & Fegley (2004), the SiO column density over a plume can vary between 2×1015mol.cm−2 and 8×1018mol.cm−2. Considering that the SO2 column density over Pele and Prometheus plumes is of the order of 2×1018mol.cm−2 based on plume-resolved measurements (Jessup et al. 2004; Spencer et al. 2000), this means a wide possible range for SiO mixing ratio in plumes from 1×10−3 to 4. Our modeling shows that it is possible to reproduce our measured upper limit on the SiO line for a variety of plausible plume distributions, assuming mixing ratios in the range of 6×10−3 - 0.7, that are within the range of predicted SiO mixing ratios. Hence our upper limit measurement is entirely consistent with a purely volcanic source for SiO. Using the same results presented above on NaCl content, and given that SiO is expected to be the main Si-carrier, we derive an upper limit on the Si/Na ratio of ∼1. This value is much less stringent than the value obtained by Na et al. (1998) on the Si/Na ratio in the corona (0.014). Our upper limit also appears to be consistent with the Si/Na ratio measured in the vaporized phase above most terrestrial basalts (of the order of 1×10−3-1×10−6 Schaefer & Fegley 2004), but is too imprecise to constrain lava composition. 7. CONCLUSIONS This work demonstrates how submillimeter observations can be used to derive unique constraints on the composition and sources of Io's lower atmosphere, which are essential to understand Io's volcanism and atmospheric processes, but also demon- strates the sensitivity limits reached by APEX observations that prevent us from performing spectroscopic searches deep enough to detect further minor species. In particular, obtaining better upper limits on volcanic tracers such as SiO would strongly constrain the characterization of ionian volcanic regimes. In addition, the interpretation of disk-averaged spectra of a strongly spatially variable atmosphere is inherently uncertain, and in particular for optically thick lines, the necessary assumptions on spatial distribution can lead to significantly different results in column density. With an apparent size of 0.9-1.2", the use of inter- ferometric facilities is necessary to resolve Io's disk, and observations with the SMA and IRAM-PdBI offer a spatial resolution down to 0.4" allowing for a first-order assessment of atmospheric coverage. The Atacama Large Millimeter Array (ALMA), a 64-element array that is approaching completion near the APEX site, will pro- vide the necessary boosts in sensitivity and spatial resolution. Due to its very large collecting area and state-of the art receivers, detection limits on minor species could be lowered by a factor of ∼50-70, allowing for searches of more species and isotopes. Extended array configurations could be used to reach spatial resolution down to 0.15" (corresponding to ∼400-500 km, i.e., the size of a Prometheus-type plume) with a sufficient SNR at ∼350 GHz, allowing direct determination of the local SO2 column density as well as helping to establish the link between atmospheric species and their respective sources. ALMA observations are then one of the most promising prospects in the next decade for the advancement of the understanding of Io at large. This paper is based on observations obtained at the Atacama Pathfinder EXperiment (APEX) telescope. APEX is a col- laboration between the Max Planck Institute for Radio Astronomy, the European Southern Observatory, and the Onsala Space Observatory. We are grateful to the APEX staff for scheduling this challenging observing program, and in particular to M. Dumke for developing dedicated observation software. We thank L. Feaga and C. Moore who kindly shared their atmospheric and vol- canic models, which were used for the analysis presented in the paper. A.M. is a Jansky Fellow at the National Radio Astronomy Observatory, a facility of the National Science Foundation operated under cooperative agreement by Associated Universities, Inc. REFERENCES 11 Chrisey, D. B., Johnson, R. E., Boring, J. W., & Phipps, J. A. 1988, Icarus, Icarus, 146, 476 Farquhar, J., Savarino, J., Airieau, S., & Thiemens, M. H. 2001, Moullet, A., Lellouch, E., Moreno, R., Gurwell, M. A., & Moore, C. 2008, Asplund, M., Grevesse, N., Sauval, A. J., & Scott, P. 2009, ARA&A, 47, 481 Austin, J. V. & Goldstein, D. B. 2000, Icarus, 148, 370 Baroni, M., Thiemens, M. H., Delmas, R. J., & Savarino, J. 2007, Science, Brown, M. E. 2001, Icarus, 151, 190 Carr, M. H., Masursky, H., Strom, R. G., & Terrile, R. J. 1979, Nature, 280, 315, 84 729 75, 233 Clow, G. D. & Carr, M. H. 1980, Icarus, 44, 268 de Pater, I., Laver, C., Marchis, F., Roe, H. G., & Macintosh, B. A. 2007, Icarus, 191, 172 J. Geophys. Res., 106, 32829 Feaga, L. M., McGrath, M., & Feldman, P. D. 2009, Icarus, 201, 570 Fegley, B. & Zolotov, M. Y. 2000a, Icarus, 148, 193 Fegley, Jr., B. & Zolotov, M. Y. 2000b, Meteoritics and Planetary Science Supplement, 35, 52 Gao, X. & Thiemens, M. H. 1991, Geochim. Cosmochim. Acta, 55, 2671 -- . 1993a, Geochim. Cosmochim. Acta, 57, 3159 -- . 1993b, Geochim. Cosmochim. Acta, 57, 3171 Geissler, P., McEwen, A., Porco, C., et al. 2004, Icarus, 172, 127 Geissler, P. E. & Goldstein, D. B. 2007, Plumes and their deposits, ed. R. M. C. Lopes & J. R. Spencer (Springer Praxis Books / Geophysical Sciences), 163 Gusten, R., Nyman, L. Å., Schilke, P., et al. 2006, A&A, 454, L13 Haff, P. K., Watson, C. C., & Yung, Y. L. 1981, J. Geophys. Res., 86, 6933 Howell, R. R., Nash, D. B., Geballe, T. R., & Cruikshank, D. P. 1989, Icarus, 78, 27 Ingersoll, A. P. 1989, Icarus, 81, 298 Jessup, K. L., Spencer, J. R., Ballester, G. E., et al. 2004, Icarus, 169, 197 Johnson, M. L. & Burnett, D. S. 1990, Geophys. Res. Lett., 17, 981 Klein, B., Philipp, S. D., Kramer, I., et al. 2006, A&A, 454, L29 Lapkin, I., Nystrom, O., Desmaris, V., et al. 2008, in Ninteenth International Symposium on Space Terahertz Technology, ed. W. Wild, 351 Lellouch, E. 2005, Space Sci. Rev., 116, 211 Lellouch, E., Paubert, G., Moses, J. I., Schneider, N. M., & Strobel, D. F. 2003, Nature, 421, 45 Lellouch, E., Strobel, D. F., Belton, M. J. S., et al. 1996, ApJ, 459, L107 Lodders, K. 2003, ApJ, 591, 1220 Lopes-Gautier, R., McEwen, A. S., Smythe, W. B., et al. 1999, Icarus, 140, 243 Lucas, R. & Liszt, H. 1998, A&A, 337, 246 Marini, L., Paiotti, A., Principe, C., Ferrara, G., & Cioni, R. 1994, Bulletin McEwen, A. S., Belton, M. J. S., Breneman, H. H., et al. 2000, Science, 288, of Volcanology, 56, 487 1193 McEwen, A. S., Keszthelyi, L., Spencer, J. R., et al. 1998, Science, 281, 87 McEwen, A. S. & Soderblom, L. A. 1983, Icarus, 55, 191 McGrath, M. A., Belton, M. J. S., Spencer, J. R., & Sartoretti, P. 2000, Mendillo, M., Wilson, J., Spencer, J., & Stansberry, J. 2004, Icarus, 170, 430 Moses, J. I., Zolotov, M. Y., & Fegley, B. 2002, Icarus, 156, 107 Moullet, A., Gurwell, M. A., Lellouch, E., & Moreno, R. 2010, Icarus, 208, 353 A&A, 482, 279 458, 417 Muller, S., Gu´elin, M., Dumke, M., Lucas, R., & Combes, F. 2006, A&A, Na, C. Y., Trafton, L. M., Barker, E. S., & Stern, S. A. 1998, Icarus, 131, 449 Postberg, F., Kempf, S., Srama, R., et al. 2006, Icarus, 183, 122 Rai, V. K., Jackson, T. L., & Thiemens, M. H. 2005, Science, 309, 1062 Sakai, H., Casadevall, T. J. C., & Moore, J. G. 1982, Geochim. Cosmochim. Acta, 46, 729 Schaefer, L. & Fegley, B. 2004, Icarus, 169, 216 -- . 2005, Icarus, 173, 454 Schneider, N. M. & Bagenal, F. 2007, Io's neutral clouds, plasma torus, and magnetospheric interaction, ed. R. M. C. Lopes & J. R. Spencer (Springer Praxis Books / Geophysical Sciences), 265 Spencer, J. R., Jessup, K. L., McGrath, M. A., Ballester, G. E., & Yelle, R. 2000, Science, 288, 1208 Spencer, J. R., Lellouch, E., Richter, M. J., et al. 2005, Icarus, 176, 283 Summers, M. E. & Strobel, D. F. 1996, Icarus, 120, 290 Trafton, L. 1975, Nature, 258, 690 Tsang, C. C. C., Spencer, J. R., Lellouch, E., et al. 2012, Icarus, 217, 277 Walker, A. C., Moore, C. H., Goldstein, D. B., Varghese, P. L., & Trafton, L. M. 2012, Icarus, 220, 225 Wiens, R. C., Burnett, D. S., Calaway, W. F., et al. 1997, Icarus, 128, 386 Williams, D. A. & Howell, R. R. 2007, Active Volcanism: Effusive Eruptions, ed. R. M. C. Lopes & J. R. Spencer (Springer Praxis Books / Geophysical Sciences), 265 Wong, M. C. & Smyth, W. H. 2000, Icarus, 146, 60 Zhang, J., Goldstein, D. B., Varghese, P. L., et al. 2003, Icarus, 163, 182 Zolotov, M. Y. & Fegley, B. 1998, Icarus, 132, 431
1501.03231	1	1501	2015-01-14T01:42:19	Long-lived Chaotic Orbital Evolution of Exoplanets in Mean Motion Resonances with Mutual Inclinations	[ "astro-ph.EP" ]	We present N-body simulations of resonant planets with inclined orbits that show chaotically evolving eccentricities and inclinations that can persist for at least 10 Gyr. A wide range of behavior is possible, from fast, low amplitude variations to systems in which eccentricities reach 0.9999 and inclinations 179.9 degrees. While the orbital elements evolve chaotically, at least one resonant argument always librates. We show that the HD 73526, HD 45364 and HD 60532 systems may be in chaotically-evolving resonances. Chaotic evolution is apparent in the 2:1, 3:1 and 3:2 resonances, and for planetary masses from lunar- to Jupiter-mass. In some cases, orbital disruption occurs after several Gyr, implying the mechanism is not rigorously stable, just long-lived relative to the main sequence lifetimes of solar-type stars. Planet-planet scattering appears to yield planets in inclined resonances that evolve chaotically in about 0.5% of cases. These results suggest that 1) approximate methods for identifying unstable orbital architectures may have limited applicability, 2) the observed close-in exoplanets may be produced during the high eccentricity phases induced by inclined resonances, 3) those exoplanets' orbital planes may be misaligned with the host star's spin axis, 4) systems with resonances may be systematically younger than those without, 5) the distribution of period ratios of adjacent planets detected via transit may be skewed due to inclined resonances, and 6) potentially habitable planets in resonances may have dramatically different climatic evolution than the Earth. The GAIA spacecraft is capable of discovering giant planets in these types of orbits.	astro-ph.EP	astro-ph	Draft version August 2, 2018 Preprint typeset using LATEX style emulateapj v. 04/17/13 5 1 0 2 n a J 4 1 . ] P E h p - o r t s a [ 1 v 1 3 2 3 0 . 1 0 5 1 : v i X r a LONG-LIVED CHAOTIC ORBITAL EVOLUTION OF EXOPLANETS IN MEAN MOTION RESONANCES WITH MUTUAL INCLINATIONS Rory Barnes1,2,3, Russell Deitrick1,2, Richard Greenberg4, Thomas R. Quinn1,2, Sean N. Raymond2,5 Draft version August 2, 2018 ABSTRACT We present N-body simulations of resonant planets with inclined orbits that show chaotically evolv- ing eccentricities and inclinations that can persist for at least 10 Gyr. A wide range of behavior is possible, from fast, low amplitude variations to systems in which eccentricities reach 0.9999 and incli- nations 179.9◦. While the orbital elements evolve chaotically, at least one resonant argument always librates. We show that the HD 73526, HD 45364 and HD 60532 systems may be in chaotically-evolving resonances. Chaotic evolution is apparent in the 2:1, 3:1 and 3:2 resonances, and for planetary masses from lunar- to Jupiter-mass. In some cases, orbital disruption occurs after several Gyr, implying the mechanism is not rigorously stable, just long-lived relative to the main sequence lifetimes of solar-type stars. Planet-planet scattering appears to yield planets in inclined resonances that evolve chaotically in about 0.5% of cases. These results suggest that 1) approximate methods for identifying unstable orbital architectures may have limited applicability, 2) the observed close-in exoplanets may be pro- duced during the high eccentricity phases induced by inclined resonances, 3) those exoplanets' orbital planes may be misaligned with the host star's spin axis, 4) systems with resonances may be systemat- ically younger than those without, 5) the distribution of period ratios of adjacent planets detected via transit may be skewed due to inclined resonances, and 6) potentially habitable planets in resonances may have dramatically different climatic evolution than the Earth. The GAIA spacecraft is capable of discovering giant planets in these types of orbits. 1. INTRODUCTION Exoplanetary systems with multiple planets show a wide variety of orbital behavior, such as large amplitude oscillations of eccentricity (e.g. Laughlin & Adams 1999; Barnes & Greenberg 2006), mean motion resonances (MMRs) (e.g. Lee & Peale 2002; Raymond et al. 2008), and, in one case, oscillations of inclination (McArthur et al. 2010; Barnes et al. 2011). Here we consider exo- planets in MMRs with mutual inclinations and find that these systems can evolve with large, chaotic fluctuations of eccentricity and inclination for at least 10 Gyr, but the MMR is maintained throughout. In general, orbital interactions are broken into two main categories: secular and resonant. The former treats the orbital evolution as though the planets' masses have been spread out in the orbit, i.e. the mass distribution averaged over timescales much longer than the orbital periods. If the orbits are non-circular or non-coplanar, the gravitational forces change the orbital elements peri- odically. MMRs occur when two or more orbital frequencies are close to integer multiples of each other. The planets pe- riodically reach the same relative positions, introducing repetitive perturbations that can dominate over secular effects. The combination drives the long-term evolution of the system. For example, Rivera & Lissauer (2001) 1 Astronomy Department, University of Washington, Box 951580, Seattle, WA 98195 2 NASA Astrobiology Institute -- Virtual Planetary Labora- tory Lead Team, USA 3 E-mail: [email protected] 4 Lunar and Planetary Laboratory, 1629 E. University Blvd., Tucson, AZ 86716 5 CNRS, Laboratoire d'Astrophysique de Bordeaux, UMR 5804, F-33270, Floirac, France considered the long-term behavior of Gl 876 (Marcy et al. 2001) and their integrations show hints of chaotic evolu- tion on 100 Myr timescales (see their Fig. 4b). In this study we expand the research on exoplanets in MMRs to include significant mutual inclinations. While most bodies in our Solar System are on low-e, low-i orbits, the exoplanets do not share these traits. Ec- centricities span values from 0 to > 0.9 (e.g. Butler et al. 2006), and at least the two outer planets of υ And have a large mutual inclination of 30◦ (McArthur et al. 2010). Numerous exoplanetary pairs in MMR are known, in- cluding the systems of Gl 876, HD 82943, and HD 73526 in 2:1 (Marcy et al. 2001; Mayor et al. 2004; Tinney et al. 2006; Vogt et al. 2005), HD 45364 in 3:2 (Correia et al. 2009) and HD 60532 in 3:1 (Desort et al. 2008). The or- bital plane of Gl 876 b has been measured astrometrically with HST (Benedict et al. 2002), and it was reported that this observation was compatible with a mutual inclina- tion between orbital planes (Rivera & Lissauer 2003), but no formal publication demonstrated that suggestion. The HD 128311 planetary system lies close to the 2:1 res- onance, and recent astrometric measurements have mea- sured the orbital plane of HD 128311 c, but not the other planet (McArthur et al. 2014). Without knowledge of both orbital planes, the mutual inclinations are unknown and use of the minimum masses and coplanar orbits may not reveal the true orbital behavior, so this system could in fact lie in the 2:1 resonance. Several studies have also explored MMRs with mu- inclinations in the context of planet formation. tual Thommes & Lissauer (2003) found that convergent mi- gration in a planetary disk can excite large inclinations in exoplanetary systems. Lee & Thommes (2009) per- formed a similar experiment and found that migrat- ing planets could become temporarily captured in an 2 inclination-type MMR, in which conjunction librates about the midpoint between the longitudes of ascending node (inclination resonances are described in more detail in § 2). Libert & Tsiganis (2009) explored the forma- tion of higher-order inclination resonances and also found that temporary capture in an inclination resonance can occur. They also artificially turned off migration shortly after capture and found the resulting system was stable. Teyssandier & Terquem (2014) considered this process and identified several constraints on the planetary mass ratio and eccentricities that permit entrance into an incli- nation resonance. All these studies find that MMRs with inclination can form in the protoplanetary disk, but only for a few specific scenarios, and even then the inclination resonance is likely to be fleeting. None of these studies considered the long-term evolution of the resonant pairs. Significant mutual inclinations may also be formed via gravitational scattering events that typically result in the ejection of one planet (Marzari & Weidenschilling 2002; Chatterjee et al. 2008; Raymond et al. 2010; Barnes et al. 2011). Raymond et al. (2010) also showed that scattering could produce systems in MMRs about 5% of the time, but they did not consider the inclinations of the resul- tant systems. All these studies reveal that formation of MMRs with mutual inclination is possible, but probably rare. While HST has successfully characterized several nearby exoplanetary systems astrometrically, the GAIA space telescope could astrometrically detect hundreds of Jupiter-sized planets (see e.g. Lattanzi et al. 2000; Sozzetti et al. 2001; Casertano et al. 2008; Sozzetti et al. 2014), revealing how common are systems in an MMR with significant mutual inclinations. Against this back- drop, we explore the dynamics of planets with orbital period commensurabilities, significant eccentricities and mutual inclinations. This paper is organized as follows. In § 2 we describe the physics of resonance and our numerical methods. In § 3, we present results of hypothetical systems, including planets in the habitable zone (Kasting et al. 1993; Kop- parapu et al. 2013), as well as giant and dwarf planets in the 2:1, 3:1 and 3:2 commensurabilities. In § 4 we show that planet-planet scattering can produce systems in the 2:1 MMR and with significant mutual inclinations. In § 5 we analyze several known exoplanetary systems and find several that could be evolving chaotically. In § 6 we discuss the results and then conclude in § 7. 2. METHODS 2.1. Resonant Dynamics MMRs are well-studied, and can be explained intu- itively for low, but non-zero, values of e and inclination i (Peale 1976; Greenberg 1977; Murray & Dermott 1999). Stable resonances can be divided into two types: eccen- tricity and inclination. The difference lies in the loca- tions of the stable longitudes of conjunction, sometimes called the libration center. In an eccentricity (e-type) resonance, the stable longitudes are located at the lon- gitude of periastron of the inner planet (1), and the apoastron of the outer planet (α2 ≡ 2 + π). For in- clination (i-type) resonances, the stable longitudes lie halfway between the longitudes of ascending node, Ω, of each planet (Ω1,2 ± π/2) when the reference plane is the fundamental plane. When a system is formed, if con- junction initially occurs near one of these stable points, and if there is a commensurability of mean motions such that the conjunction longitude varies slowly, then con- junction will tend to librate. Furthermore, because the orbits are farthest apart at these libration centers, res- onances can reduce the likelihood of close encounters, further maintaining long-term orbital stability. Both e-type and i-type resonances are observed in our Solar System, but e-type is far more prevalent, including the Galilean satellite system and the Neptune-Pluto pair. An i-type resonance is observed in the Saturnian satellite pair of Mimas and Tethys (Allan 1969; Greenberg 1973). Neptune and Pluto are particularly relevant to this study, as the pair is in the 3:2 e-resonance (Cohen & Hubbard 1965), and later studies found that the i-resonance ar- guments also librate on short timescales, but the libra- tion drifts slowly such that the resonant argument actu- ally circulates with a period of ∼ 25 Myr (Williams & Benson 1971; Applegate et al. 1986). Higher order secu- lar resonances are also operating on Neptune and Pluto (Milani et al. 1989; Kinoshita & Nakai 1996), and likely contribute to their configuration being formally chaotic (Sussman & Wisdom 1988). Nonetheless, the pair is sta- ble for at least 5.5 Gyr (Kinoshita & Nakai 1996). Thus, Pluto and Neptune's orbital evolution is impacted by the i-resonance but it is likely a small effect. The first step in identifying a mean motion resonance is to examine the ratio of orbital periods P or, equivalently, the mean motions n ≡ 2π/P . If the ratio is close to a ratio of two integers, i.e. n1/n2 ≈ j1/j2 where j1 and j2 are integers, then the system may be in resonance. An MMR requires a periodic force to applied near the same longitude relative to an apse or a node, which pre- cess with time. To account for this evolution, celestial mechanicians have introduced the resonant argument, a combination of angles that account for both the mean motions and the orbital orientations. For e-type resonances, the resonant arguments are θ1,2 = j1λ2 − j2λ1 − j31,2, (1) where the j terms are integers that sum to 0, and the sub- scripts 1 and 2 to correspond to the inner and outer planet, respectively. Should any θ librate with time, or even circulate slowly, then resonant dynamics are impor- tant. Conjunction (j1λ2 − j2λ1) lies close to a particular longitude relative to the apsides. Usually θ will librate about either 0 or π, but other stable values are possible and are referred to as asymmetric resonances. For i-type resonances, the dominant resonant argu- ment is φ = j1λ2 − j2λ1 − j3Ω1 − j4Ω2, (2) if i = 0 corresponds to the invariable, or fundamen- tal, plane, i.e. the plane perpendicular to the total angular momentum of the system. With this choice, Ω1 = Ω2 ± π. Inclination resonances are fundamen- tally different from e-type in that first order resonances are technically not possible, i.e. j1 − j2 > 1. The sta- ble conjunction longitudes correspond to the two mid- node longitudes. Note that for circular orbits, there is no substantive difference between the two stable longi- tudes: The geometry at ± 90◦ from the mutual node are identical. See Greenberg (1977) for more discussion on the physics of the i-resonance. It is often convenient to think of resonances in terms of a pseudo-potential. The conjunction longitude, λc is accelerated toward the stable points, and oscillates sinu- soidally about it. The acceleration of λc in a 2:1 (4:2) MMR can be written as λc = c1e1 sin(λc−1)+c2e2 sin(2−λc)+c3i2 1 sin 2(λc−Ω1), (3) where c1, c2 and c3 are constants that depend on the masses and semi-major axes (Greenberg 1977), and λc = 2λ2 − λ1. Eq. (3) is analogous to that of a compound pendulum, as long as the e's and i's are approximately constant. The pseudo-potential contains minima at 1, α2, and Ω1 ± π/2, but each has different depths which vary as the orbits evolve. Moreover, as e and/or i become large, this simple picture breaks down and more minima will likely appear. Thus, we should anticipate the motion to become complicated, an expectation that is borne out in §§ 3 -- 5. 2.2. Numerical Methods The classical analytic theory summarized above be- comes less accurate as the values of eccentricity and in- clination increase. To lowest order, the evolution of e and i are decoupled, but as either or both become large, pathways for the exchange of angular momentum open up, and the motion can become very complicated (e.g. Barnes et al. 2011). Thus, we rely on N -body numer- ical methods to analyze resonances with mutual incli- nations, while appealing to published analytic theory to help interpret the outcomes. In particular we use the well-tested and reliable Mercury (Chambers 1999) and HNBody (Rauch & Hamilton 2002) codes. We do not include general relativistic corrections, which are mostly negligible for the planetary systems we consider here. We used both mixed variable symplectic and Bulirsch-Stoer methods to integrate our systems. While the former can integrate our hypothetical systems to 10 Gyr within about 10 days on a modern workstation, the Bulirsch- Stoer method, which is more accurate, requires about 80 -- 90 days with HNBody; hence we used it sparingly. Regard- less of our choice of software and/or integration scheme, identical initial conditions produced qualitatively similar results. All integrations presented here conserved energy to better than 1 part in 105, usually better by many or- ders of magnitude. Unless stated otherwise, all systems maintained an MMR for 10 Gyr and met our energy con- servation requirements. The reference plane for i and Ω for all cases is the invariable plane so that the inclina- tions and longitudes of ascending node are more physi- cally meaningful6. Here we ignore planetary and stellar spins: Planets and stars are point masses. We consider several broad categories of exoplanetary systems. First we model systems in the 2:1 resonance. In Set #1, one planet is always 1 Earth-mass (1 M⊕) with a semi-major axis a of 1 AU, and the primary is a solar-mass star, i.e. it is in the habitable zone (Kasting et al. 1993; Kopparapu et al. 2013). The other planet may have a mass between 3 and 40 M⊕. In Set #2, we 6 We have made our source code to rotate any astrocen- tic orbital elements into the invariable plane, as well as gen- erate input files for Mercury and HNBody, publicly available at https://github.com/RoryBarnes/InvPlane. 3 explore the case of two giant planets orbiting a 0.75 M(cid:12) star in 2 and 4 year orbits. Set #3 considers two dwarf planets orbiting a solar-mass star. In Sets #4 and #5, we simulated ∼Earth-mass planets in the 3:1 and 3:2 MMR, respectively. For all these simulations, the period ratios are at exact resonance, but the other orbital elements are chosen randomly and uniformly over a wide range of val- ues. For each set, we integrated 100 systems for 10 Myr as an initial survey, and for systems that appeared inter- esting, i.e. showed chaotic evolution, we continued them to 10 Gyr using 0.01 year timesteps. Table 1 shows the ranges of initial conditions we used for these sets of hy- pothetical systems. In § 5 we examine known systems in or near the 2:1 (HD 73526 and HD 128311), 3:2 (HD 60532) and 3:1 (HD45364) MMRs. Each of these systems was discov- ered via radial velocity data, and hence suffer from the mass-inclination degeneracy. However, HD 128311 c has also been detected astrometrically with HST (McArthur et al. 2014), hence its mass and full orbit are known. The best fits and uncertainties in parentheses for these known systems are listed in Table 2. The position of each planet in its orbit, the "phase," is crucial information and differ- ent authors present the information in different formats and so we present the parameter used in the most recent paper in Table 2. Tp is the time of periastron passage, λ is the mean longitude and µ is the mean anomaly. For each of the known systems, we vary each orbital element uniformly within its published uncertainties and simulate the orbital evolution with HNBody. Our goal is not to calculate a probability that each system is in an i-type resonance, but rather to determine if it is possible at all. We simulate 100 versions of the these systems and allow i and Ω to take any value, and m is adjusted accordingly. If chaotic motion in the MMR is apparent, we integrate it for a further 10 Gyr. 3. HYPOTHETICAL SYSTEMS In this section we present the results of the simulations described in the previous section. We separate the results by resonance: 2:1, then 3:1, and finally 3:2. In all cases we find evidence of chaotic orbital evolution, often with very large amplitudes of eccentricity and inclination. 3.1. The 2:1 Resonance In this section we examine example cases in Sets #1 -- 3, i.e. in the 2:1 MMR. The e-resonance arguments are θ1,2 = 2λ2 − λ1 − 1,2, and the i-resonance argument is φ = 4λ2 − 2λ1 − Ω1 − Ω2. (4) (5) 3.1.1. Earth with an Exterior Companion (Set #1) Set #1 consists of an Earth-mass planet with a 1 year period and a larger exterior planet with an orbital period of 2 years. In Fig. 1 we show the evolution of the resonant arguments, e and i for the first example, System A in Table 3, for the first 105 years. The top two panels show that the resonance arguments switch between libration and circulation. Initially θ1 (black dots) librates about 0, the classic libration cen- ter for e-type resonances, while θ2 (red dots) librates on short timescales, but circulates on long timescales. Also 4 Set MMR M∗ (M(cid:12)) 1 2 3 4 5 2:1 2:1 2:1 3:1 3:2 1 0.75 1 1 1 Initial Conditions for Hypothetical Systems TABLE 1 m1 1 M⊕ 1 M⊕ 1 M⊕ 0.314 MJup 0.0775 MM oon m2 a1 (AU) a2 (AU) e 3.3 -- 36.6 M⊕ 0.1 -- 1.15 MJup 2.84 MM oon 3.3 -- 36.6 M⊕ 3.3 -- 36.6 M⊕ 1 1.44219 1 1 1 1.5874 2.2893 1.5874 2.0801 1.3104 0 -- 0.5 0 -- 0.5 0 -- 0.5 0 -- 0.5 0 -- 0.5 i (◦) Ω (◦) 0 -- 360 0 -- 30 0 -- 360 0 -- 30 0 -- 30 0 -- 360 0 -- 360 0 -- 30 0 -- 30 0 -- 360 ω (◦) 0 -- 360 0 -- 360 0 -- 360 0 -- 360 0 -- 360 λ (◦) 0 -- 360 0 -- 360 0 -- 360 0 -- 360 0 -- 360 Best Fits and Uncertainties for Selected Known Exoplanet Systems TABLE 2 System M∗ (M(cid:12)) Planet m (MJup) P (d) e ω (◦) Phase HD 128311 0.828 HD 73526 HD 60532 HD 45364 1.08 1.44 0.82 b c b c b c b c 1.769 (0.023) 3.125 (0.069) 453.019 (0.404) 921.538 (1.15) 0.303 (0.011) 0.159 (0.006) 57.864 (3.258) 15.445 (6.87) 2400453.019 (4.472)a 2400921.538 (18.01)a 2.8 (0.2) 2.5 (0.3) 1.03 (0.05) 2.46 (0.09) 0.1872 (0) 0.6579 (0) 188.3 (0.9) 377.8 (2.4) 201.3 (0.6) 604 (9) 0.19 (0.05) 0.14 (0.09) 0.28 (0.03) 0.02 (0.02) 226.93 (0.37) 342.85 (0.28) 0.1684 (0.019) 0.0974 (0.012) 203 (9) 13 (76) 351.9 (4.9) 151 (92) 162.58 (6.34) 7.41 (3.4) 86 (13)b 82 (27)b 2453987 (2)a 2453723 (158)a 105.76 (1.41)c 269.52 (0.58)c a Tperi (JD) b µ (◦) c λ (◦) Initial Conditions for Selected Set #1 Systems TABLE 3 System Body m (M⊕) a (AU) e A B C Db E F G 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 10.07 1 26.91 1 4.39 1 35.62 1 14.82 1 10 1 10 1 1.5874 1 1.5874 1 1.5874 1 1.5874 1 1.5874 1 1.5874 1 1.5874 0.0866 0.1883 0.0251 0.0531 0.2373 0.4225 0.00296 0.266 0.2952 0.0961 0.1 0.15 0.15 0.2 i (◦) 10.322 0.82 1.61 0.0475 2.918 0.565 19.83 0.449 38.51 1.832 12.04 0.9551 23.526 1.832 Ω (◦) 223.75 43.747 277.19 97.19 217.13 37.13 278.01 98.01 247.92 67.92 270.26 90.26 232.96 52.96 ω (◦) 340.2 74.21 353.67 219.07 330.25 246.36 332.27 69.73 40.44 253.92 140.77 23.77 31 247.9 µ (◦) 268.33 296.5 261.33 359.38 102.45 112.88 268.52 343.09 225.93 318.12 0 10 0 10 b Stable for only 73 Myr. note that the e- and i-resonance arguments appear to be coupled. During this initial phase, φ librates with large amplitude. After about 14,000 years, the behav- ior changes dramatically and all arguments librate, but with sudden jumps between libration centers. Then at 25,000 years, the motion appears to return to the ini- tial state. This switching between modes of oscillation demonstrates the system is chaotic, and is often an indi- cator of impending instability (e.g. Laskar 1990). Hence we would naively expect a similar outcome for this sys- tem. The eccentricity of the inner planet shows un- usual behavior that also changes with the resonance ar- guments. The evolution is approximately periodic, but does not appear regular. The inclinations do not appear to be strongly impacted by the changes in the resonance argument behavior. In Fig. 2, we extend the evolution of e and i by a factor of 105, to 10 Gyr. Here we see that the hints of chaotic behavior in Fig. 1 remain present, but are not indica- tive of the true scale of the chaotic motion for this sys- tem. Remarkably, the system survives for 10 Gyr despite the chaos. The eccentricity of the inner planet aperiodi- cally reaches values larger than 0.65, while its inclination reaches 40◦. The outer planet is more massive, so its evo- lution does not have as large an amplitude as the inner, but it, too, evolves chaotically. While the variations in e and i are chaotic, there are clearly bounds to their permitted values. The libration and slow circulation of the resonant ar- guments suggest that both e and i-type resonances are important in this system. We further explore the evo- lution of the system in Fig. 3, which only considers the first 30,000 years of orbital evolution. In the left panel we compare the conjunction longitude (black dots) with the 5 Fig. 1. -- The first 105 years if evolution of System A. Black corresponds to the inner Earth-mass planet initially at 1 AU, red to a larger planet in the outer 2:1 resonance. Variations of the inclination resonance argument (top), eccentricity arguments (top middle) with black dots corresponding to the 1 argument and red to 2, eccentricities (bottom middle) and inclinations (bottom). 090180270360φ (o) 090180270θ1, θ2 (o) 0.000.050.100.150.200.25Eccentricity02•1044•1046•1048•1041•105Time (yr)024681012Inclination (o)6 Fig. 2. -- The evolution of e and i for System A over 10 Gyr. Black points correspond to the inner planet and red to the outer. 0.00.20.40.60.8Eccentricity0246810Time (Gyr)010203040Inclination (o)various stable longitude predicted from classic resonant theory, 1 (purple curve), α2 ≡ 2 + π (orange curve), 1 ≡ Ω1 + π/2 (blue curve), and Ω− 1 ≡ Ω1 − π/2 Ω+ (green curve). Conjunction librates, but, surprisingly, not always about the expected stable longitudes. For most of the time, λc ∼ 1, but it also librates about other locations that are not associated with any classic libration center. Note that conjunction avoids α2. In the right panel, we plot the two mean motions and see that they librate about the resonant frequencies, but with varying amplitudes. Careful inspection of the two panels shows that the different mean motion amplitudes correspond to the different librations seen in the left panel. There appear to be 5 different resonant oscilla- tions over this cycle, which approximately repeats for at least the first 1 Myr. The long-term behavior in Fig. 2 shows mode-switching can occur on longer timescales as well. For example, from 6.5 -- 6.8 Gyr, e and i are confined to narrow regions, but then return to the large amplitude oscillations. Fig. 4 shows the evolution over 105 years starting at 6.75 Gyr within this alternative mode. In comparison to Fig. 1, the i-resonance argument is circulating, as is θ2. How- ever, θ1 is librating about 0 indicating that the system is exclusively in an e-type resonance. After 10 Gyr the evolution is qualitatively similar to the initial evolution. The e- and i-resonant arguments are switching between libration and circulation, a quasi- stable behavior. We are unaware of any study that has found qualitatively similar behavior in a planetary or satellite system. While long-lived chaos is evident in our Solar System (e.g. Sussman & Wisdom 1988), the am- plitudes of the variations are much smaller. Moreover, the switching between different quasi-periodic states is also usually an indicator of instability. System A is not "nearly integrable," as is often argued for stable plane- tary systems. Nonetheless, these planets remains bound and in resonance for the main sequence lifetime of a solar- type star. Although the orbital behavior of System A is surpris- ing, this case is not anecdotal. Table 3 contains 6 exam- ples from Set #1 that show chaotic evolution, yet remain in resonance for 10 Gyr. In Figs. 5 -- 7, we show 3 more cases as representative examples of the 2:1 MMR. These simulations demonstrate that chaotic inclined MMRs can produce behavior that spans a range from high frequency, low amplitude oscillations to low frequency and high am- plitude oscillations to cases in which all available phase space is sampled. System B is an example of high frequency, low ampli- tude evolution. Fig. 5 shows the first 25,000 years of its evolution. In this case the resonant arguments oscillate with a period of a few hundred years and switch states at ∼18,000 years. The eccentricities and inclinations of the Earth-like planet oscillate with an amplitude of 0.1 and 0.25◦, respectively, on this timescale. Note that System B begins with a mutual inclination of just 1.65◦, reveal- ing that relatively small mutual inclinations can lead to chaotic orbital evolution. Over 10 Gyr, e1 remains below 0.14 and e2 0.06, while i1 remains below 18◦ and i2 below 1◦. System C is similar to System A in its evolution, as shown in Fig. 7. Over 10 Gyr e1 varies from 0 to 0.9 and 7 i1 varies from 0 to 60◦. From ∼ 3 Gyr to ∼ 6 Gyr (not shown) the system enters a different mode in which the eccentricities and inclinations are confined to narrower regions. System D only survives in the resonance for 73 Myr, but we it include here to illustrate the extreme eccen- tricity evolution that is possible in inclined MMRs. At 87,170 years, e1 reaches a value of 0.99998, implying a periastron distance that would place it inside the core of its solar-mass primary. Clearly, the point-mass approxi- mation for the orbital dynamics has broken down for this system. In particular, we expect that at when e1 ∼ 1 that strong tidal effects should dramatically alter the or- bits. We return to this point in § 6. Rather than show similar plots for Systems E -- G, we only comment on their behavior. System E shows a cir- culating φ for the first 105 years, while θ1 librates with large amplitude about 0, and θ2 drifts. The inner planet's e and i vary from 0 to 0.75 and 55◦, respectively. After ∼ 6 Gyr, the system slowly moves into a different state with i1 varying between 20◦ and 50◦, and θ1 aperiodi- cally circulating. System F has resonant arguments that switch between libration and circulation as in Fig. 1 and eccentricities that remain below 0.3 and inclinations be- low 20◦. System G is similar to System A. These examples are only illustrative and do not rep- resent the full range of possibilities. The 7 cases listed in Table 3 met our stability criteria (see § 2), and we suspect many more cases also would, but computational constraints prevented a more thorough analysis. 3.1.2. Two Giant Planets Orbiting a 0.75 M(cid:12) Star (Set #2) Next we consider Set #2: Two Saturn- to Jupiter-mass planets orbiting a 0.75 M(cid:12) star in 2 and 4 year orbital periods. Such planets induce a larger astrometric signal in the host star, and are large enough that radial velocity observations could break the 180◦ ambiguity in Ω implicit in astrometric measurements of exoplanets. In Table 4 we present 7 systems which survived for ∼ 10 Gyr and conserved energy adequately. In Fig. 8 we show the evolution of System N on two timescales. The left panels show the variation in the reso- nant arguments and e and i for 105 years in the same for- mat as Fig. 1. As before the resonant arguments switch between libration and circulation leading to chaotic evo- lution of e and i. In the right panels we show the evolu- tion of e and i for 10 Gyr. e1 aperiodically reaches values of ∼ 0.85, and i1 reaches 50◦. Note as well at 2.4 and 6.0 Gyr the systems enters qualitatively different states for about 100 Myr. In Fig. 9 the orbit of the host star about the system's barycenter is shown over 7 years if viewed face-on, i.e. the invariable plane is parallel to the sky plane. In this ex- ample we assume the system is located at 25 pc, and the combined system induces a ∼ 100 µas astrometric signal, which is 4 -- 5 times larger than the expected GAIA uncer- tainties for stars with G-band magnitudes <∼13, repre- sented by the line labeled "GAIA Uncertainty" (Sozzetti et al. 2014). This system would be relatively easy to characterize with radial velocity data as well, and hence could be fully characterized in the next 10 years. Of course, the details of actually modeling resonant systems are non-trivial (e.g. Marcy et al. 2001), but the discov- 8 Fig. 3. -- Initial evolution of System A. Left: Evolution of the conjunction longitude λc (black dots) in relation to the 4 classic stable longitudes, 1 (purple curve), α2 ≡ 2 + π (dashed orange), Ω+ 1 ≡ Ω1 − π/2 (green). Right: Evolution of the mean motions of the inner planet (top) and outer planet (bottom). 1 ≡ Ω1 + π/2 (blue), and Ω− Initial Conditions for Selected Set #2 Systems TABLE 4 System Body m (MJup) a (AU) e H Ie J K L M N 1 2 1 2 1 2 1 2 1 2 1 2 1 2 0.314 0.573 0.314 0.303 0.314 0.234 0.314 0.328 0.314 0.736 0.314 0.399 0.314 0.902 1.4422 2.2893 1.4422 2.2893 1.4422 2.2893 1.4422 2.2893 1.4422 2.2893 1.4422 2.2893 1.4422 2.2893 0.0276 0.2021 0.1155 0.4692 0.2374 0.3988 0.0364 0.223 0.0263 0.3696 0.4596 0.3402 0.0561 0.4466 i (◦) 2.504 1.118 3.302 3.076 7.089 8.019 11.267 8.761 19.459 6.989 3.282 1.936 21.623 6.53 Ω (◦) 247.16 67.12 39.32 219.29 24.56 204.54 126.11 306.18 176.25 356.23 106.68 286.57 247.55 67.64 ω (◦) 94.57 126.15 273.88 247.62 256 185.86 284.5 273.49 202.97 153.08 44.59 131.7 325.64 60.26 µ (◦) 7.74 313.87 195.14 48.35 77.1 127.68 41.42 160.47 229.91 261.74 350 73.93 274.59 326.55 e Stable for only 9.761 Gyr. ery of such a chaotic system would mark an important milestone in exoplanet science. In Fig. 10 we show the evolution of System I in the same format as Fig. 8. This system appears qualitatively similar to System N, but with lower amplitudes in e and i. However, this system destabilizes at 9.761 Gyr, as seen on the right side of the right panels. We found several other systems in which an ejection occurred after 1 Gyr. The chaotic resonant behavior shown in this study is not necessarily stable on arbitrarily long timescales. Like our own Solar System, these hypothetical systems are just relatively long-lived (see, e.g. Lecar et al. 2001). 3.1.3. Dwarf Planets in the 2:1 MMR (Set #3) In this section we consider Set #3, dwarf planets in the 2:1 MMR. While these planets are not detectable in the near future, they display remarkable dynamics and illustrate the extreme chaos that the 2:1 MMR with in- clinations can produce. In Table 5 we present 4 cases that are stable for 10 Gyr, and in Fig. 11 we show the 01•1042•1043•104Time (Years)090180270360Angle (o)Ω1+ Ω1- α2ϖ1λc01•1042•1043•104Time (Years)090180270360 0.9960.9981.0001.0021.004n1 (/yr)01•1042•1043•104Time (Years)0.49960.49980.50000.50020.5004n2 (/yr)9 Fig. 4. -- Same as Fig. 1, but at 6.75 Gyr. 090180270360φ (o) 090180270θ1, θ2 (o) 0.000.050.100.150.200.250.30Eccentricity02•1044•1046•1048•1041•105Time - 6.75 Gyr05101520Inclination (o)10 Fig. 5. -- The first 25 kyr of evolution of System B in the same format as Fig. 1. 090180270360φ (o) 090180270θ1, θ2 (o) 0.000.020.040.060.080.10Eccentricity05.0•1031.0•1041.5•1042.0•1042.5•104Time (yr)0.00.51.01.5Inclination (o)11 Fig. 6. -- The first 1 Myr of evolution of System C in the same format as Fig. 1. 090180270360φ (o) 090180270θ1, θ2 (o) 0.00.20.40.6Eccentricity02•1054•1056•1058•1051•106Time (yr)0123Inclination (o)12 Fig. 7. -- The first 105 years of evolution of System D in the same format as Fig. 1. This system destabilized after 73 Myr. 090180270360φ (o) 090180270θ1, θ2 (o) 0.00.20.40.60.8Eccentricity02•1044•1046•1048•1041•105Time (yr)050100150Inclination (o)13 Fig. 8. -- Orbital evolution of System N. The left panels are in the same format as Fig. 1; the right is in the same format as Fig. 2. evolution of System P over 10 Gyr. Note that the e- resonance arguments sometimes librate for long periods of time, such as near 2.5 Gyr. The i argument does not appear to librate in this visualization, but higher reso- lution plots over shorter periods show similar behavior as above. Systems O and Q are similar to P, but the inclinations and eccentricities remain lower. These systems show a diversity of behavior from small- scale chaos (System Q), to dramatic chaos in which e1 and i1 sample all available phase space (System P). The resonant arguments, particularly the eccentricity argu- ments, switch between circulation and libration. e1 in Systems O (not shown), P and R (not shown) aperiodi- cally reach values in excess of 0.99, and hence they should tidally circularize prior to 10 Gyr. The resonant arguments in these systems behave dif- ferently than for the larger planets. On short timescales (not shown) the resonant arguments switch modes as in previous cases, but these systems can remain in one mode for long timescales, particularly the e-resonance. In Fig. 11 note that there are intervals when the two e-arguments librate for more than 100 Myr. and the i-resonance argument is φ = 3λ2 − λ1 − Ω1 − Ω2. (7) Table 6 lists three configurations that are stable for 10 Gyr and show chaotic evolution. Systems T and U are shown in Figs. 12 and 13, respectively. The former has (relatively) modest inclination and eccentricity vari- ations, while the latter samples all the phase space avail- able, with 0 <∼ i <∼ π and 0 <∼ e <∼ 1. In both systems the resonance arguments switch between libration and circu- lation, as seen in the previous 2:1 MMR cases. System S (not shown) is similar to System U, but the inclinations only reach ∼ 65◦ and e only 0.9. 3.3. The 3:2 Resonance Next we consider the 3:2 MMR with an interior 1 M⊕ planet at 1 AU from a solar-mass star and a larger exter- nal companion at 1.3104 AU. The e-resonance arguments are (8) (9) θ1,2 = 3λ2 − 2λ1 − 1,2, and the i-resonance argument is φ = 6λ2 − 4λ1 − Ω1 − Ω2. 3.2. The 3:1 Resonance In this section we consider an Earth-like planet at 1 AU from a 1 M(cid:12) star with an exterior companion with an orbital period of 3 years, i.e. in the 3:1 MMR. In this case the e-resonance arguments are θ1,2 = 3λ2 − λ1 − 21,2, (6) Table 7 lists two systems that are stable for 10 Gyr and showed chaotic evolution. In Fig. 14 we plot the evolu- tion of the resonant arguments, e and i on two timescales for System W. The evolution is qualitatively similar to that in the other resonances. System V is quite similar, with e reaching 0.9 and i reaching 60◦ aperiodically over 10 Gyr. 090180270360φ (o) 0100200300θ1, θ2 (o) 0.00.20.40.60.8Eccentricity02•1044•1046•1048•1041•105Time (yr)01020304050Inclination (o) 0.00.20.40.60.8Eccentricity0246810Time (Gyr)01020304050Inclination (o)14 Fig. 9. -- Orbit of the host star in Fig. 8 (System N) about the system barycenter for 7 years if the system lies at 25 pc and the fundamental plane is parallel to the sky plane. The line labeled "GAIA Uncertainty" is 25 µas long, and represents GAIA's typical per-measurement uncertainty for bright stars. K Dwarf at 25 pc-150-100-50050100150Dec (µ")-150-100-50050100150RA (µ")GAIAUncertainty15 Fig. 10. -- Orbital evolution of System I in the same format as Fig. 8. The system destabilizes after 9.761 Gyr. Initial Conditions for Selected Set #3 Systems TABLE 5 System Body m (MM oon) a (AU) e O P Q R 1 2 1 2 1 2 1 2 0.0775 1.171 0.0775 2.817 0.0775 0.568 0.0775 2.513 1 1.5874 1 1.5874 1 1.5784 1 1.5784 0.3032 0.2768 0.04305 0.3192 0.1036 0.1544 0.4475 0.2093 i (◦) 19.835 1.013 24.34 0.544 5.298 0.577 14.98 0.332 Ω (◦) 283.31 103.31 57.06 237.06 169.03 349.03 289.05 109.05 ω (◦) 290.53 50.74 90.49 231.06 21.27 171.19 94.15 352.99 µ (◦) 231.72 54.65 57.09 90.68 348.68 158.69 272.54 154.19 Initial Conditions for Selected Set #4 Systems TABLE 6 System Body m (M⊕) a (AU) e S T U 1 2 1 2 1 2 1 6.5 1 27.05 1 22.2 1 2.0801 1 2.0801 1 2.0801 0.1798 0.3808 0.02514 0.05311 0.149 0.2755 i (◦) 2.06 0.234 1.04 0.0362 43.64 1.27 Ω (◦) 178.04 358.4 246.96 66.96 260.51 80.51 ω (◦) 261.79 113.3 151.63 17.04 18.15 51.82 µ (◦) 66.85 286.96 261.33 359.38 220.4 6.72 090180270360φ (o) 0100200300θ1, θ2 (o) 0.00.10.20.30.40.5Eccentricity02•1044•1046•1048•1041•105Time (yr)051015Inclination 0.00.10.20.30.40.50.6Eccentricity0246810Time (Gyr)05101520Inclination (o)16 Fig. 11. -- Evolution of System P in the same format as Fig. 1. 090180270360φ (o) 090180270θ1, θ2 (o) 0.00.20.40.60.8Eccentricity0246810Time (Gyr)04590135Inclination (o)17 Fig. 12. -- Orbital evolution of System T in the same format as Fig. 8. Initial Conditions for Selected Set #5 Systems TABLE 7 System Body m (M⊕) a (AU) e V W 1 2 1 2 1 5.96 1 21.83 1 1.3104 1 1.3104 0.1404 0.3498 0.0144 0.1671 i (◦) 27.34 4.08 36.05 1.37 Ω (◦) 232.78 52.78 316.73 136.73 ω (◦) 39.79 314.14 13.72 307.8 µ (◦) 5.92 108.38 293.02 308.84 0100200300φ (o) 0100200300θ1, θ2 (o) 0.000.010.020.030.040.05Eccentricity02•1044•1046•1048•1041•105Time (yr)0.00.51.01.5Inclination (o) 0.00.10.20.3Eccentricity0246810Time (Gyr)051015Inclination (o)18 Fig. 13. -- Orbital evolution of System U in the same format as Fig. 8. 4. FORMATION BY SCATTERING The large amplitude chaotic orbital evolution shown above is remarkable, but can such a system form? As described in § 1, several studies have examined the for- mation of inclination resonances in the context of con- vergent migration during the protoplanetary disk phase. Those studies found that i-resonances can form under the proper circumstances, but they did not consider the post- formation evolution. In this section, we show that grav- itational scattering among planets can produce systems in the 2:1 MMR with mutual inclinations that evolve chaotically for 10 Gyr. Our sample comes from the data set used in Raymond et al. (2008) that found MMRs resulting from scattering events. The reader is referred to that paper for details, but briefly, systems consisting of initially 3 planets were allowed to interact gravitationally and in many cases 1 -- 2 planets were ejected from the system. Raymond et al. (2008) examined the e-resonance arguments of systems in which 2 planets remained, and reported that 5% of systems had at least one e-resonance argument librating. Systems like those shown in § 3 were deemed unstable and thrown out. In light of the results of § 3 we have re-examined sys- tems near an MMR to search for chaotic, but long-lived evolution. Specifically we focus on the "Mixed2" distri- bution of Raymond et al. (2008) in which the planets' masses follow a power-law distribution with exponent −1.1. This distribution has recently been shown to repro- duce many observed dynamical properties of exoplanets (Timpe et al. 2013). This suite of simulations consisted of 1000 systems, and we examined 49 that had period ratios with 10% of 2, of which 24 were identified as being in an e-resonance by Raymond et al. (2008). Of these, 27 had mutual inclinations less than 5◦ and the largest was 27◦. We find three that appear qualitatively similar to those in the previous section. They are listed in Ta- ble 8, and System X is shown in Fig. 15. System Y (not shown) evolved such that the eccentricities and inclina- tions remained below 0.12 and 7◦, respectively. System Z evolved such that they remained below 0.15 and 3◦. We also searched for systems evolving chaotically in the 3:1 or 3:2 MMR, but did not find any. The amplitudes of the variations of the orbital elements is lower than some cases shown in § 3, but similar to others, e.g. System B. Given the small number of systems that produced chaotic resonant behavior, it remains to be seen if evolution that reaches e ∼ 1 and i ∼ π can be naturally produced. Nonetheless we conclude that planet-planet scattering can produce systems that evolve chaotically for 10 Gyr in an MMR. 5. KNOWN SYSTEMS The previous two sections established the typical char- acteristics of planets in inclined MMRs and a viable for- mation mechanism. The next question that naturally arises is if any known systems might be evolving in a long-lived and chaotic configuration. In this section we examine four candidates in 3 different MMRs and with observed properties listed in Table 2. These planets were all discovered by the radial velocity technique and hence 090180270360φ (o) 090180270θ1, θ2 (o) 0.00.20.40.60.8Eccentricity02•1054•1056•1058•1051•106Time (yr)04590135Inclination (o) 0.00.20.40.60.81.0Eccentricity0246810Time (Gyr)04590135Inclination (o)19 TABLE 8 Initial Conditions for Planets in a Chaotic 2:1 MMR Formed by Scattering System Body m (MJup) a (AU) e X Y Z 1 2 1 2 1 2 0.118 0.213 0.65 0.0737 0.168 0.934 4.83696 7.71971 6.23933 9.97199 5.45138 8.78422 0.1498 0.1548 0.026033 0.04723 0.13299 0.01305 i (◦) 18.845 8.15 0.802 5.60 1.453 0.204 Ω (◦) 291.14 111.12 168.16 348.18 133.88 313.83 ω (◦) 83.60 195.95 30.05 298.78 186.20 137.53 µ (◦) 330.83 118.57 284.95 49.35 113.99 146.93 their orbital planes were undetected. HD 128311 c has been detected astrometrically (McArthur et al. 2014), but its companion planet has not, so the mutual inclina- tion remains unknown. 5.1. 2:1 Systems: HD 128311 and HD 73426 We performed 100 integrations of HD 128311 and HD 73526 for 10 Myr each. We found no stable con- figurations of the former if the orbits were allowed to be non-planar. This should not be taken to mean the sys- tem must be coplanar as we may not have considered enough cases. Rein & Papaloizou (2009) found the sys- tem could form in a coplanar configuration through con- vergent migration, and McArthur et al. (2014) found the best-fit coplanar solution to the system is dynamically stable and not in resonance. At this point, we conclude that this system is likely to be in a coplanar configu- ration, which precludes the chaotic evolution we report here. HD 73526, on the other hand, could be evolving chaot- ically. In Table 9 we list three versions that are stable for 10 Gyr and show chaotic evolution. Note that we always use a stellar mass of 1.08 M(cid:12). In Fig. 16 we show the evolution of System AB. Unlike the previous systems, the resonant arguments do not appear to switch between li- bration and circulation, and the i-arguments do not show any libration at all. The θ1 argument librates about 0 for the full 10 Gyr year duration. Nonetheless, the ec- centricities and inclinations appear to be coupled and to switch between modes, as was seen in § 3. Systems AA and AC show similar behavior with similar amplitudes. Of the previous systems, HD 73526 is most similar to System B (c.f. Fig. 5). 5.2. The 3:1 System HD 60532 In this section we consider the HD 60532 system which is in the 3:1 MMR. In Table 10 we list 3 cases which show chaos for 10 Gyr, and present the evolution of case BC in Fig. 17. As with HD 75326, the i-resonance arguments circulate, but the e-arguments switch between libration and circulation. This system appears qualitatively sim- ilar to those in § 3.2. Over 10 Gyr, the system evolves chaotically, as shown in the right panels. The BA and BB systems are qualitatively similar, but the mode-switching is not as dramatic. 5.3. The 3:2 System HD 45364 We found no configurations of HD 45364 that were sta- ble for 10 Gyr, but one trial did survive for 4.557 Gyr while conserving energy to 1 part in 106. Its initial con- ditions are in Table 11, and its evolution is shown in Fig. 18. 6. DISCUSSION The previous three section have demonstrated that 1) planets in an MMR and with mutual inclinations can experience chaotic evolution of orbital elements for Gyr while at least 1 resonant argument librates throughout, 2) planet-planet scattering, in which one planet is re- moved from a planetary system by gravitational inter- actions, can leave behind two planets in the 2:1 MMR and with significant mutual inclinations, and 3) several systems known to be in an MMR could have mutual incli- nations that induce chaotic evolution, but maintain the resonance. In this section, we discuss the theoretical and observational implications of these results, as well as de- scribe the limits of our analysis, which naturally leads to directions for future research on this topic. Fig. 3 shows that conjunction can librate about mul- tiple centers, some of which, such as 1, are predicted by classic celestial mechanics. These libration centers are derived from low-order expansions of the "disturbing function" (for a review see Murray & Dermott (1999)). If one term dominates, there are specific stable longitudes for conjunction, e.g. 1 if the term 2λ2 − λ1 − 1 domi- nates. However, with large and varying values of e and i multiple terms are important. Stable longitudes can mi- grate or become unstable, allowing transitions of libra- tion to different kinematic modes. This movement could be due to the deepening of nearby minima as e and/or i changes. As the orbits evolve, conjunction could gain access to another minimum, leading to a change in the libration center. The behavior has multiple drivers that each behave like a pendulum. In effect, the system is analogous to a com- pound pendulum, e.g. Eq. (3). As the system evolves, the planets are able to move into different modes of os- cillation. For some modes, the stable longitude is repre- sented by classic longitudes like 1, but others may only be derivable using higher order theory. Our hypothesis should be testable from derivation of high-order models of resonant behavior from the disturbing function. Such an analysis was beyond the scope of this work, which is just a demonstration of the amplitude and duration of the chaos, but is clearly desirable. This result has important implications for theoretical work on orbital stability. Common approaches for iden- tifying unstable orbits rely on short integrations (∼ 1000 orbits) and a subsequent analysis of the orbital evolution to count the number of frequencies in the orbital oscil- lations: a larger number could indicate the system is chaotic and unstable. Examples include the Mean Expo- nential Growth of Nearby Orbits (MEGNO; (e.g. Cin- cotta & Sim´o 2000; Go´zdziewski et al. 2001)) and fre- quency maps (e.g. Laskar 1990; Laskar & Correia 2009). 20 Initial Conditions for Selected HD 73526 Cases TABLE 9 System Body m (MJup) a (AU) e AA AB AC b c b c b c 2.89 2.576 2.881 2.405 2.87 2.691 0.66857 1.0615 0.64347 1.0311 0.6572 1.0274 0.1728 0.1126 0.1824 0.05134 0.1552 0.08992 f Measured relative to invariable plane i (◦)f Ω (◦)f 1.37 182.66 1.21 1.27 1.18 3.37 2.85 262.07 81.74 342.84 162.79 2.44 ω (◦) 193.96 147.97 57.63 8.87 349.04 324.05 µ (◦) 96.27 107.05 95.97 92.51 75.29 77.07 Initial Conditions for Selected HD 60532 Cases TABLE 10 System Body m (MJup) a (AU) e BA BB BC c b c b c b 0.997 2.42 1.053 2.464 1.062 2.545 0.759 1.5943 0.7582 1.5661 0.7587 1.572 0.3037 0.1578 0.2978 0.0354 0.2816 0.0207 f Measured relative to invariable plane i (◦)f Ω (◦)f 2.56 267.21 0.692 1.403 0.399 5.806 1.62 253.26 73.39 293.03 113.05 87.1 ω (◦) 185.73 252.05 212.28 250.17 79.61 341.14 µ (◦) 162.6 322.57 277.17 322.34 209.64 321.88 Initial Conditions for the HD 45364 Case TABLE 11 Body m (MJup) a (AU) e b c 0.1874 0.6581 0.6822 0.8969 0.1782 0.09935 i (◦)f Ω (◦)f 359.31 2.07 0.509 179.38 ω (◦) 110.52 320.73 µ (◦) 199.03 265.9 f Measured relative to invariable plane Our results suggest that those methods are susceptible to labeling long-lived systems as short-lived. The inves- tigation of viable orbital architectures of planets in an inclined MMR should seek configurations that can per- sist for the age of the host star as described here, which appear unstable on the short term, but are in fact long- lived. For example, HD 202206 is likely in a 5:1 MMR, but Correia et al. (2005) used a frequency analysis to con- strain the orbital parameters rather than N-body mod- els. Future work should explore the the veracity of these approximate methods for the case of inclined MMRs. At epochs of very high e, we expect some planets to tidally circularize. For example, Systems D and U could not persist as shown because tides would circularize the planet. Two scenarios are plausible, depending on the dissipation rate in the planet. If the dissipation is rela- tively weak, the resonance may be maintained such that the resonant pair migrates inward together. Such sys- tems should be represented in radial velocity and transit surveys, which are biased toward the detection of plan- ets on relatively short orbital periods. The three cases presented in § 5 are possible examples. In principle, the inward migration could be arrested if the tidal dissipa- tion forces the pair into a configuration that diminishes the maximum eccentricity. Thus, planets found far from the host star today could nonetheless have formed at larger distances and moved in. As the dissipation can be episodic, it could take a long time for the pair to migrate inward significantly, further increasing the likelihood to detect the planets where tidal forces are expected to be insignificant. This possibility has been discussed for non- resonant systems (Li et al. 2014; Dawson & Chiang 2014), and we find it can also occur for MMRs. Future work should explore the role of tidal dissipation in these sys- tems and determine if there is any predicted or observed signature of tidal evolution resulting from an inclined MMR. On the other hand, if the tidal dissipation in the planet is very large, it may break the resonance, resulting in rapid tidal circularization and migration, leaving one close-in planet and one more distant a planet. Although many close-in planets are singletons (Steffen et al. 2012), some are known to host more distant companions. These companions have been suggested as perturbers that could drive eccentricities to high values through Kozai-type oscillations (Takeda & Rasio 2005), but the possibility of resonant excitation of eccentricity has not previously been proposed. Note that these periods of high-e can occur when i has a wide range of values. Thus, the planet's final or- bital plane could be independent of its initial orbital plane and be misaligned with its host star's spin axis. Such spin-orbit misalignment is observed (Triaud et al. 2010; Hirano et al. 2014), and numerous mechanisms have been proposed, such as planet-planet scattering fol- lowed by tidal circularization (Chatterjee et al. 2008), tidal circularization during phases of high eccentricity due to interactions with very distant perturbers (Fab- 21 Fig. 14. -- Orbital evolution of System W in the same format as Fig. 8. 0100200300φ (o) 0100200300θ1, θ2 (o) 0.00.20.40.60.8Eccentricity02•1044•1046•1048•1041•105Time (yr)01020304050Inclination (o) 0.00.20.40.60.8Eccentricity0246810Time (Gyr)02550Inclination (o)22 Fig. 15. -- Orbital evolution of System X in the same format as Fig. 8. 23 Fig. 16. -- Orbital evolution of System AB (HD 73526) in the same format as Fig. 8. 090180270360φ (o) 090180270θ1, θ2 (o) 0.000.050.100.150.20Eccentricity02•1044•1046•1048•1041•105Time (yr)0123Inclination (o) 0.000.050.100.150.20Eccentricity0246810Time (Gyr)01234Inclination (o)24 rycky & Tremaine 2007; Storch et al. 2014), and interac- tions with the gas disk and a distant stellar binary com- panion (Batygin 2012). The chaotic evolution of e and i in an inclined MMR is another process to add to this list. However, given the low occurrence rate of chaotic incli- nation resonances formed by scattering, it is unlikely to be a dominant mechanism. But note that this inference is based on only one set of scattering simulations with specific initial conditions -- others may be more efficient and producing these types of systems. Inclined MMRs could impact the distribution of period ratios of planets found via transit by making one planet's orbital plane significantly different from the other. After applying a geometric transit correction, Steffen & Hwang (2014) find a relative excess of planets in the 3:2 reso- nance and relative deficits in 2:1 and 3:1 in Kepler data. However, for the systems in which only 2 planets are de- tected, the excess of 3:2 pairs disappears, the 2:1 is still depleted, and the 3:1 appears to be unaffected. If addi- tional planets destabilize inclined MMRs, then the trend in the 3:2 MMR may be indicative of inclined MMRs, as we would not expect systems that evolve with large amplitude to have companions nearby. However, caution is necessary when interpreting these data, as knowledge of the underlying population and the role of tidal evo- lution is critical. Migration during the protoplanetary disk phase often leads to capture into resonance (Snell- grove et al. 2001; Lee & Peale 2002), producing a pri- mordial excess of planets in MMRs. On the other hand, tidal evolution of planets in MMRs tends to pull them to period ratios that are not exactly commensurate (Lith- wick & Wu 2012; Batygin & Morbidelli 2013; Delisle & Laskar 2014). Moreover, the formation of inclined reso- nant orbits of close-in planets is unknown, so they could be intrinsically rare. Thus, it is not obvious that inclined MMRs are sculpting the close-in exoplanet population, but our results suggest that they could. Future work should identify boundaries to the long- lived but chaotic resonances discovered here and employ a quantitative description of the chaos. All of our sim- ulations of hypothetical systems began with planetary orbital periods at exact commensurability, but this state is not typically observed for exoplanets, see Table 2 and § 5. A large suite of N-body simulations that integrate systems to 10 Gyr is required to map out the bound- aries of the chaotic and resonant behavior. Throughout this study we have referred to systems as chaotic as they clearly are. However, as a system moves toward regular motion, the motion might no longer be obviously chaotic. The use of Lyapunov exponents or other metrics would be necessary to elucidate the true boundaries of the long- lived chaotic motion. Identifying inclined MMRs in transit and/or RV data is possible but difficult (see e.g. Dawson et al. 2014), but astrometric measurements are probably the best route to find their existence. HST has successfully detected some planets for which RV detections suggest an MMR is present (Benedict et al. 2002; McArthur et al. 2014) and two planets not in an MMR (McArthur et al. 2010), but has not detected two planets in an MMR. The outer planet of HD 128311 has been detected, and the in- ner is potentially detectable with HST (McArthur et al. 2014), but its precarious architecture relative to dynam- ical instability suggests it is in a coplanar configuration. The most likely instrument to discover planets in an in- clined MMR is GAIA (Casertano et al. 2008), as shown in Fig. 9. The known systems we examined in § 5 are all good candidates for GAIA astrometry, and so in the next 5 years we should find out if any of them could be experiencing chaotic evolution. We note that our simulations of known systems failed to reveal any in which the i arguments librate. Instead the e-arguments switch between circulation and libra- tion, which likely drives the chaos. Note that System B evolves in a similar manner, and many other cases do as well. Of our 3 systems formed by scattering, 2 were similar to the known systems (Fig. 15). A further ex- ploration of known systems, either by including more or broadening the parameter space survey, could reveal if this trend is real. If known systems are only consistent with i-argument circulation, it could provide important clues to the formation and frequency of exoplanets in inclined MMRs. Throughout this study we have described systems that survive for 10 Gyr as "stable." As most of our host stars are solar analogs, this usage is reasonable as the systems survive for the main sequence lifetimes of the host stars. However, in some cases we find destabilization can occur on > 1 Gyr timescales. Hence, these systems are not "stable" in the sense that they could survive indefinitely. Late-term destabilization of systems in inclined MMRs may explain the observation that the host stars of planets in the 2:1 MMR tend to be younger than other host stars (Koriski & Zucker 2011). If this observation is validated, it would support the hypothesis that some MMRs evolve chaotically and disrupt on long timescales. In the previous sections we did not consider the role of additional planets, which could significantly shrink the longevity of a chaotically evolving system. It is clear that for system in which e1 ∼ 1 that interior planets are for- bidden. However, exterior planets could exist provided they are distant and/or small. On the other hand, some systems show low amplitude chaotic variations and may therefore be robust to perturbations from a third planet. We do not map out the role of a third companion in the stability of our systems, but future work should explore how they modify the results presented here. Most of our simulations begin with an Earth-mass planet at 1 AU from a solar-mass star, and would be con- sidered potentially habitable (Kasting et al. 1993; Kop- parapu et al. 2013). If habitable, these worlds would be markedly different from the Earth, with unpredictable climates on geologic timescales. For planets in which the eccentricity grows large, the planet could occasion- ally enter a runaway greenhouse (incident radiation flux scales as (1−e2)−1/2) rendering the planet uninhabitable. Large inclination fluctuations can also lead to large vari- ations in obliquity, which is a major driver of climate evolution (e.g. Williams & Kasting 1997; Spiegel et al. 2009). While extremely fast and large variations of obliq- uity could be detrimental to habitability, at the outer edge of the habitable zone, these variations can suppress ice sheet growth and, in principle, increase a planet's habitability (Armstrong et al. 2014). For planets that occasionally reach very high eccentricity, tidal dissipa- tion could ultimately pull the planet out of the habitable zone (Barnes et al. 2008). Should any potentially habit- 25 Fig. 17. -- Orbital evolution of System BC (HD 60352) in the same format as Fig. 2. 090180270360φ (o) 090180270θ1, θ2 (o) 0.00.10.20.3Eccentricity02•1044•1046•1048•1041•105Time (yr)0123456789Inclination (o) 0.00.10.20.30.4Eccentricity0246810Time (Gyr)0510Inclination (o)26 Fig. 18. -- Orbital evolution of the HD 45364 case in the same format as Fig. 8. due to eccentricity excitation by inclined MMRs followed by rapid tidal circularization. • Some short-period planets with orbital planes mis- aligned with the stellar spin axis may be produced by systems initially in inclined MMRs. • MMR pairs may be episodically migrating inward due to weak dissipation occurring during epochs of very large eccentricity. • Systems in an MMR may be systematically younger than other multiplanet systems due to the destabilization of older MMR systems. • The distribution of period ratios of adjacent plan- ets detected via transit may be skewed by inclined MMRs. • Potentially habitable planets may be severely im- pacted by the orbital architecture of the system. Although no systems are currently known to demon- strate the behavior we have outlined here, the GAIA space telescope has the power to detect hundreds of giant exoplanets in inclined MMRs. This work was supported by NASA's Virtual Plan- etary Laboratory under Cooperative Agreement No. NNA13AA93A and NSF grant AST-1108882. able planets be found in an MMR, it will be imperative to understand the orbital evolution, and its connection to climate, prior to investing spectroscopic observations of its atmosphere to search for biosignatures (see e.g. Dem- ing et al. 2009; Kaltenegger & Traub 2009; Misra et al. 2014). 7. CONCLUSIONS We have simulated the orbital evolution of exoplanets in mean motion resonances and inclinations and found the orbits can evolve chaotically for at least 10 Gyr. We hypothesize that these systems behave like compound pendula, which are naturally chaotic systems that can switch between modes of oscillation, as seen in our sim- ulations (see Fig. 3). We find this chaotic motion over a range of mass ratios and for the 2:1, 3:2 and 3:1 res- onance. We also tested different N-body codes using different integration schemes, and conclude the results are robust. Inclined MMRs can be produced by planet- planet scattering and the resultant systems are qualita- tively similar to our simulations of known systems in an MMR. These results have numerous implications for both the- ory and observations: • Approximate methods to estimate stability with short integrations may be unreliable near MMRs. • Close-in planets may arrive at their current orbits Allan, R. R. 1969, AJ, 74, 497 Applegate, J. H., Douglas, M. R., Gursel, Y., Sussman, G. J., & Wisdom, J. 1986, AJ, 92, 176 REFERENCES 0100200300φ (o) 0100200300θ1, θ2 (o) 0.000.050.100.150.200.25Eccentricity02•1044•1046•1048•1041•105Time (yr)0.00.51.01.52.02.5Inclination (o) 0.000.050.100.150.200.250.30Eccentricity012345Time (Gyr)012345Inclination (o)Armstrong, J. C., Barnes, R., Domagal-Goldman, S., et al. 2014, Astrobiology, 14, 277 Barnes, R., & Greenberg, R. 2006, ApJ, 652, L53 Barnes, R., Greenberg, R., Quinn, T. R., McArthur, B. E., & Benedict, G. F. 2011, ApJ, 726, 71 Barnes, R., Raymond, S. N., Jackson, B., & Greenberg, R. 2008, Astrobiology, 8, 557 Batygin, K. 2012, Nature, 491, 418 Batygin, K., & Morbidelli, A. 2013, AJ, 145, 1 Benedict, G. F., McArthur, B. E., Forveille, T., et al. 2002, ApJ, 581, L115 Butler, R. P., Wright, J. T., Marcy, G. W., et al. 2006, ApJ, 646, 505 Casertano, S., Lattanzi, M. G., Sozzetti, A., et al. 2008, A&A, 482, 699 Chambers, J. E. 1999, MNRAS, 304, 793 Chatterjee, S., Ford, E. B., Matsumura, S., & Rasio, F. A. 2008, ApJ, 686, 580 Cincotta, P. M., & Sim´o, C. 2000, A&AS, 147, 205 Cohen, C. J., & Hubbard, E. C. 1965, AJ, 70, 10 Correia, A. C. M., Udry, S., Mayor, M., et al. 2005, A&A, 440, 751 -- . 2009, A&A, 496, 521 Dawson, R. I., & Chiang, E. 2014, Science, 346, 212 Dawson, R. I., Johnson, J. A., Fabrycky, D. C., et al. 2014, ApJ, 791, 89 Delisle, J.-B., & Laskar, J. 2014, ArXiv e-prints, arXiv:1406.0694 Deming, D., Seager, S., Winn, J., et al. 2009, PASP, 121, 952 Desort, M., Lagrange, A.-M., Galland, F., et al. 2008, A&A, 491, 883 Fabrycky, D., & Tremaine, S. 2007, ApJ, 669, 1298 Go´zdziewski, K., Bois, E., Maciejewski, A. J., & Kiseleva-Eggleton, L. 2001, A&A, 378, 569 Greenberg, R. 1973, MNRAS, 165, 305 -- . 1977, Vistas in Astronomy, 21, 209 Hirano, T., Sanchis-Ojeda, R., Takeda, Y., et al. 2014, ApJ, 783, 9 Kaltenegger, L., & Traub, W. A. 2009, ApJ, 698, 519 Kasting, J. F., Whitmire, D. P., & Reynolds, R. T. 1993, Icarus, 101, 108 Kinoshita, H., & Nakai, H. 1996, Earth Moon and Planets, 72, 165 Kopparapu, R. K., Ramirez, R., Kasting, J. F., et al. 2013, ApJ, 765, 131 Koriski, S., & Zucker, S. 2011, ApJ, 741, L23 Laskar, J. 1990, Icarus, 88, 266 Laskar, J., & Correia, A. C. M. 2009, A&A, 496, L5 Lattanzi, M. G., Spagna, A., Sozzetti, A., & Casertano, S. 2000, MNRAS, 317, 211 Laughlin, G., & Adams, F. C. 1999, ApJ, 526, 881 Lecar, M., Franklin, F. A., Holman, M. J., & Murray, N. J. 2001, ARA&A, 39, 581 Lee, M. H., & Peale, S. J. 2002, ApJ, 567, 596 Lee, M. H., & Thommes, E. W. 2009, ApJ, 702, 1662 27 Li, G., Naoz, S., Kocsis, B., & Loeb, A. 2014, ApJ, 785, 116 Libert, A.-S., & Tsiganis, K. 2009, MNRAS, 400, 1373 Lithwick, Y., & Wu, Y. 2012, ApJ, 756, L11 Marcy, G. W., Butler, R. P., Fischer, D., et al. 2001, ApJ, 556, 296 Marzari, F., & Weidenschilling, S. J. 2002, Icarus, 156, 570 Mayor, M., Udry, S., Naef, D., et al. 2004, A&A, 415, 391 McArthur, B. E., Benedict, G. F., Barnes, R., et al. 2010, ApJ, 715, 1203 -- . 2014, ApJ, in press Milani, A., Nobili, A. M., & Carpino, M. 1989, Icarus, 82, 200 Misra, A., Meadows, V., Claire, M., & Crisp, D. 2014, Astrobiology, 14, 67 Murray, C. D., & Dermott, S. F. 1999, Solar system dynamics Peale, S. J. 1976, ARA&A, 14, 215 Rauch, K. P., & Hamilton, D. P. 2002, in Bulletin of the American Astronomical Society, Vol. 34, AAS/Division of Dynamical Astronomy Meeting #33, 938 Raymond, S. N., Armitage, P. J., & Gorelick, N. 2010, ApJ, 711, 772 Raymond, S. N., Barnes, R., Armitage, P. J., & Gorelick, N. 2008, ApJ, 687, L107 Rein, H., & Papaloizou, J. C. B. 2009, A&A, 497, 595 Rivera, E. J., & Lissauer, J. J. 2001, ApJ, 558, 392 Rivera, E. J., & Lissauer, J. J. 2003, in Bulletin of the American Astronomical Society, Vol. 35, AAS/Division of Dynamical Astronomy Meeting #34, 1042 Snellgrove, M. D., Papaloizou, J. C. B., & Nelson, R. P. 2001, A&A, 374, 1092 Sozzetti, A., Casertano, S., Lattanzi, M. G., & Spagna, A. 2001, A&A, 373, L21 Sozzetti, A., Giacobbe, P., Lattanzi, M. G., et al. 2014, MNRAS, 437, 497 Spiegel, D. S., Menou, K., & Scharf, C. A. 2009, ApJ, 691, 596 Steffen, J. H., & Hwang, J. A. 2014, ArXiv e-prints, arXiv:1409.3320 Steffen, J. H., Ragozzine, D., Fabrycky, D. C., et al. 2012, Proceedings of the National Academy of Science, 109, 7982 Storch, N. I., Anderson, K. R., & Lai, D. 2014, ArXiv e-prints, arXiv:1409.3247 Sussman, G. J., & Wisdom, J. 1988, Science, 241, 433 Takeda, G., & Rasio, F. A. 2005, ApJ, 627, 1001 Teyssandier, J., & Terquem, C. 2014, MNRAS, 443, 568 Thommes, E. W., & Lissauer, J. J. 2003, ApJ, 597, 566 Timpe, M., Barnes, R., Kopparapu, R., et al. 2013, AJ, 146, 63 Tinney, C. G., Butler, R. P., Marcy, G. W., et al. 2006, ApJ, 647, 594 Triaud, A. H. M. J., Collier Cameron, A., Queloz, D., et al. 2010, A&A, 524, A25 Vogt, S. S., Butler, R. P., Marcy, G. W., et al. 2005, ApJ, 632, 638 Williams, D. M., & Kasting, J. F. 1997, Icarus, 129, 254 Williams, J. G., & Benson, G. S. 1971, AJ, 76, 167
1611.02798	1	1611	2016-11-09T02:00:54	Results from a triple chord stellar occultation and far-infrared photometry of the trans-Neptunian object (229762) 2007 UK126	[ "astro-ph.EP" ]	A stellar occultation by a trans-Neptunian object (TNO) provides an opportunity to probe its size and shape. Very few occultations by TNOs have been sampled simultaneously from multiple locations, while a robust estimation of shadow size has been possible for only two objects. We present the first observation of an occultation by the TNO 2007 UK126 on 15 November 2014, measured by three observers, one nearly on and two almost symmetrical to the shadow's centerline. This is the first multi-chord dataset obtained for a so-called detached object, a TNO subgroup with perihelion distances so large that the giant planets have likely not perturbed their orbits. We revisit Herschel/PACS far-infrared data, applying a new reduction method to improve the accuracy of the measured fluxes. Combining both datasets allows us to comprehensively characterize 2007 UK126. We use error-in-variable regression to solve the non-linear problem of propagating timing errors into uncertainties of the ellipse parameters. Based on the shadow's size and a previously reported rotation period, we expect a shape of a Maclaurin spheroid and derive a geometrically plausible size range. To refine our size estimate of 2007 UK126, we model its thermal emission using a thermophysical model code. We conduct a parametric study to predict far-infrared fluxes and compare them to the Herschel/PACS measurements. The favorable geometry of our occultation chords, combined with minimal dead-time imaging, and precise GPS time measurements, allow for an accurate estimation of the shadow size (best-fitting ellipse with axes 645.80 $\pm$ 5.68 km $\times$ 597.81 $\pm$ 12.74 km) and the visual geometric albedo (15.0 $\pm$ 1.6 %). By combining our analyses of the occultation and the far-infrared data, we can constrain the effective diameter of 2007 UK126 to 599 - 629 km. We conclude that subsolar surface temperatures are $\approx$ 50 - 55 K.	astro-ph.EP	astro-ph	Astronomy & Astrophysics manuscript no. Schindler Friday 5th October, 2018 ©ESO 2018 Results from a triple chord stellar occultation and far-infrared photometry* of the trans-Neptunian object (229762) 2007 UK126 K. Schindler1, 2, J. Wolf1, 2, J. Bardecker3,†, A. Olsen3, T. Müller4, C. Kiss5, J. L. Ortiz6, F. Braga-Ribas7, 8, J. I. B. Camargo7, 9, D. Herald3, and A. Krabbe1 6 1 0 2 v o N 9 . ] P E h p - o r t s a [ 1 v 8 9 7 2 0 . 1 1 6 1 : v i X r a 1 Deutsches SOFIA Institut, Universität Stuttgart, Pfaffenwaldring 29, 70569 Stuttgart, Germany e-mail: [email protected] 2 SOFIA Science Center, NASA Ames Research Center, Mail Stop N211-1, Moffett Field, CA 94035, USA 3 International Occultation Timing Association (IOTA) 4 Max Planck Institute for Extraterrestrial Physics, Giessenbachstrasse 1, 85748 Garching, Germany 5 Konkoly Observatory, Research Centre for Astronomy and Earth Sciences, Hungarian Academy of Sciences, Konkoly Thege 15-17, 1121 Budapest, Hungary 6 Instituto de Astrofisica de Andalucia-CSIC, Glorieta de la Astronomia, 3, 18080 Granada, Spain 7 Observatório Nacional/MCTI, Rua Gal. José Cristino 77, Rio de Janeiro, RJ 20921-400, Brazil 8 Federal University of Technology - Paraná (UTFPR / DAFIS), Rua Sete de Setembro, 3165, CEP 80230-901, Curitiba, PR, Brazil 9 Laboratório Interinstitucional de e-Astronomia - LIneA, Rua Gal. José Cristino 77, Rio de Janeiro, RJ 20921-400, Brazil Received 01 April 2016; revised 10 August 2016; accepted 10 October 2016 ABSTRACT Context. A stellar occultation by a trans-Neptunian object (TNO) provides an opportunity to probe the size and shape of these distant solar system bodies. In the past seven years, several occultations by TNOs have been observed, but mostly from a single location. Only very few TNOs have been sampled simultaneously from multiple locations. Sufficient data that enable a robust estimation of shadow size through an ellipse fit could only be obtained for two objects. Aims. We present the first observation of an occultation by the TNO 2007 UK126 on 15 November 2014, measured by three observers, one nearly on and two almost symmetrical to the shadow's centerline. This is the first multi-chord dataset obtained for a so-called detached object, a TNO subgroup with perihelion distances so large that the giant planets have likely not perturbed their orbits. We also revisit Herschel/PACS far-infrared data, applying a new reduction method to improve the accuracy of the measured fluxes. Combining both datasets allows us to comprehensively characterize 2007 UK126. Methods. We use error-in-variable regression to solve the non-linear problem of propagating timing errors into uncertainties of the ellipse parameters. Based on the shadow's size and a previously reported rotation period, we expect a shape of a Maclaurin spheroid and derive a geometrically plausible size range. To refine our size estimate of 2007 UK126, we model its thermal emission using a thermophysical model code. We conduct a parametric study to predict far-infrared fluxes and compare them to the Herschel/PACS measurements. Results. The favorable geometry of our occultation chords, combined with minimal dead-time imaging, and precise GPS time mea- surements, allow for an accurate estimation of the shadow size (best-fitting ellipse with axes 645.80 ± 5.68 km × 597.81 ± 12.74 km) and the visual geometric albedo (pV = 15.0 ± 1.6%). By combining our analyses of the occultation and the far-infrared data, we can constrain the effective diameter of 2007 UK126 to deff = 599−629 km. We conclude that subsolar surface temperatures are in the order of ≈ 50 − 55 K. Key words. Occultations -- Kuiper belt objects: individual: (229762) 2007 UK126 -- Radiation mechanisms: thermal -- Methods: data analysis 1. Introduction Stellar occultations are the best opportunity to directly and ac- curately determine the size and shape of a solar system body re- motely from Earth. They allow improvements of orbital elements and have the potential to discover previously unknown satel- * Herschel is an ESA space observatory with science instruments pro- vided by European-led Principal Investigator consortia and with impor- tant participation from NASA. † Also affiliated with the Western Nevada Astronomical Society, 2699 Van Patten Avenue, Carson City, NV 89703, USA and the Re- search and Education Collaborative Occultation Network (RECON). lites (e.g. Timerson et al. 2013; Descamps et al. 2011), or even ring systems, as recently revealed for Centaur (10199) Chariklo (Braga-Ribas et al. 2014a). The shape of the light curve can re- veal the presence of an atmosphere, and enable the study of its properties (e.g. Person et al. 2013). While observations of occul- tations only require basic photometric tools, the technical chal- lenge is to acquire images at high cadence with little-to-no dead time between acquisitions, and with very precise timing infor- mation to measure disappearance and reappearance times of the occulted star with little uncertainty. Knowledge of the target's apparent orbital velocity and length of the occultation allow for a direct calculation of its size at the location of the sampled plane Article number, page 1 of 16 A&A proofs: manuscript no. Schindler or chord. Multiple observers distributed across the shadow path can sample the occulting object at different locations, which pro- vides information on its shape. Trans-Neptunian objects (TNOs) are considered the most pristine objects in our solar system. Being left-overs from the very early stage of the accretion phase, estimates of their sizes, densities and albedos, as well as constraints on their composition can provide critical information on the evolution of the solar sys- tem. Given their large geocentric distance and very small angular diameter, the prediction of a TNO's shadow path on the Earth's surface during an occultation is difficult. Owing to the very long orbital periods of TNOs, only a very small fraction of their orbits has been observed since their discovery, leaving relatively large uncertainties in their orbital elements. Another issue for shadow path predictions are astrometric uncertainties of currently avail- able star catalogs, which are inherited by orbit and ephemeris calculations for TNOs and which blur the true position of the occulted star. Besides random astrometric uncertainties that typ- ically increase towards fainter stars, catalogs usually have zonal errors. Another dominating error that can shift a shadow path significantly is potential stellar duplicity, which usually cannot be excluded in advance. We also have no information on albedo variations on the TNO's surface, which could cause a periodic shift of the center of light. All these error sources make predic- tions and successful observation campaigns for TNO occulta- tions a difficult task (Bosh et al. 2016). To improve shadow path predictions, extensive astrometric observations with high preci- sion of both the target star and the occulting body are usually conducted for many weeks in advance of an event. The first data release of the GAIA star catalog (Lindegren et al. 2016; Gaia Collaboration et al. 2016) and all subsequent data releases1 up to the final GAIA catalog in 2022 are expected to improve the ac- curacy of occultation predictions significantly, since GAIA will offer the most accurate astrometry ever obtained. GAIA will also significantly contribute towards stellar duplicity measurements. A number of occultation events covering at least 11 differ- ent TNOs (see e.g. Ortiz et al. 2014; Braga-Ribas et al. 2014b) have been successfully observed to date, not counting the well- studied Pluto system. However, most occultations by TNOs could only be observed at a single location, while the acqui- sition of multiple chords during an event has been extremely rare. Table 1 gives an overview of all successful multi-chord observations of occultations by TNOs (except Pluto) that have been published in peer-reviewed literature so far, covering only four objects to date. (50000) Quaoar (Braga-Ribas et al. 2013) is currently the best sampled TNO, having five unique chords (i.e. chords that are geographically sufficiently distant to each other to sample the occulting object at distinguishable loca- tions) that were recorded during a single event (04 May 2011). In this dataset, the time resolution and accuracy of the two chords above the centerline have been low, resulting in large uncertainties of ingress and egress times and consequently in some ambiguity of the ellipse fit. (136472) Makemake (Ortiz et al. 2012) has been generally sampled with good time reso- lution, but three of the four unique chords were very close to each other and almost located at the centerline of the ellipse fit, while the fourth chord was located below the centerline, and no chord was located above, likewise resulting in ambi- guity of the shadow geometry. For the remaining two objects, (136199) Eris and (55636) 2002 TX300, only two unique chords are available, which results in an under-determined ellipse fit. To our knowledge, multi-chord events were also recorded for 1http://www.cosmos.esa.int/web/gaia/release Article number, page 2 of 16 the following objects, but have not been published in peer- reviewed literature so far: (208996) 2003 AZ84, (20000) Varuna and (84922) 2003 VS2. In this paper, we report results from the very first multi-chord observation of a stellar occultation by a TNO that is classified as a detached object according to the widely used definitions by Gladman et al. (2008). This term denotes a sub class of TNOs with perihelion distances so large that Neptune and the other gi- ant planets have likely not perturbed their orbits in the past. In addition to being the first comprehensive study of an object of this dynamical class, the dataset presented in this paper is also extremely rare compared to previously achieved observations of occultations by TNOs: To our knowledge, it is only the second after (50000) Quaoar that provides a sufficient number of chords (at least three chords are required for an ellipse fit) that are well spaced (in the present case sampling the target simultaneously above, very close, and below the centerline by extremely fa- vorably distributed, quasi symmetrically located observers). The size derived from the occultation measurements is used as an important new constraint for thermophysical modeling based on Herschel far-infrared (FIR) flux measurements. This combined analysis allows us to comprehensively characterize 2007 UK126. 2. Known properties of 2007 UK126 from literature 2007 UK126 was discovered at Palomar Observatory on 19 Oc- tober 2007. According to the database of the Minor Planet Cen- ter (MPC)2, it has been identified subsequently on older images taken at Siding Spring Observatory and Palomar dating back to August 1982, which allowed improvements of orbit calculations (i = 23.34°, e = 0.49, a = 74.01 AU). Having an orbital pe- riod of 636.73 yr, the object is approaching perihelion (at a he- liocentric distance of rH = 37.522 AU), which it will pass on 18 March 2046, based on current ephemeris data provided by the JPL Small-Body Database3. Around perihelion, the target will have its largest apparent magnitude of mV ≈ 19.13 mag (in V-band) during its entire orbit according to MPC estimates. Photometric and spectroscopic data of 2007 UK126 were acquired with several instruments at ESO's VLT in a coordi- nated campaign on 21 and 22 September 2008. Perna et al. (2010) report visible and infrared photometry (VRIJH) taken with VLT/FORS2 and VLT/ISAAC. They estimated the abso- lute magnitude (defined as the apparent magnitude at a distance of 1 AU both to the Sun and to the observer at zero phase an- gle) of 2007 UK126 in V-band as HV = 3.69 ± 0.04 mag. Their R-band measurement allows us to derive HR = 3.07 ± 0.04 mag. While these photometric measurement uncertainties ∆Hphot re- sult from noise in the data, they do not consider possible peri- odic magnitude variations ∆m owing to the rotation of the body. To take these into account as well, Santos-Sanz et al. (2012) pro- posed an additional error term of ∆Hrot = 0.88 · ∆m 2 . This is jus- tified by assuming a sinusoidal light curve, where roughly 88% of its amplitude contains 68.3% of the function values. As the light curve amplitude was not measured back in 2012, a peak- to-peak amplitude of ∆m = 0.2 mag was assumed, arguing that roughly 70% of a sample of TNO light curves studied by Duf- fard et al. (2009) showed less magnitude variations. Consider- ing both independent error sources, the total measurement un- + (∆Hrot)2. This leads to certainty becomes ∆H = (cid:113)(cid:16) (cid:17)2 ∆Hphot 2http://www.minorplanetcenter.net 3http://ssd.jpl.nasa.gov/sbdb.cgi http://ssd.jpl. nasa.gov/horizons.cgi Schindler et al.: Results from an occultation and FIR photometry of (229762) 2007 UK126 Table 1. Multi-chord observations of stellar occultations by trans-Neptunian objects (TNOs) published in peer-reviewed literature to date. Object (55636) 2002 TX300 (136199) Eris (136472) Makemake (50000) Quaoar Dynamical class HC SDO HC HC (229762) 2007 UK126 DO Date (UTC) 09 October 2009 06 November 2010 23 April 2011 04 May 2011 17 February 2012 15 November 2014 Location Hawaii Chile Chile, Brazil Chile, Brazil, Uruguay France, Switzerland USA No. of recorded chords (No. of unique chords) Key reference 2 3 (2) 7 (4) 6 (5) 4 (2) 3 1 2 3 4 this work Notes. Numbers in brackets indicate the number of chords that were geographically sufficiently distant to each other to sample the occulting object at distinguishable (i.e. unique) locations. Abbreviations: SDO - Scattered disk object; HC - Hot classical Kuiper belt object; DO - Detached Object. References. (1) Elliot et al. (2010); (2) Sicardy et al. (2011); (3) Ortiz et al. (2012); (4) Braga-Ribas et al. (2013) refined absolute magnitude estimates of HV = 3.69 ± 0.10 mag and HR = 3.07 ± 0.10 mag. Interestingly, Perna et al. (2010) note that 2007 UK126 was the only object in their study that they were unable to catego- rize into any of the four TNO taxonomic classes. By compar- ing color indices (B-V, V-R, V-I, V-J, V-H, V-K) to those of the Sun, a TNO can be classified as BB (neutral, or "blue" in color), BR (intermediate "blue-red"), IR (moderately "red") and RR (very "red"); details are given in Barucci et al. (2005) and Fulchignoni et al. (2008). While visible photometry provides classifications of 2007 UK126 both as IR or RR, infrared pho- tometry allows classifications both as BB or BR. This could in- dicate that the current taxonomic scheme needs to be refined in the future. Fornasier et al. (2009) report visible spectroscopy of 2007 UK126 obtained with VLT/FORS2 (0.44 -- 0.93 µm). The spectral continuum between 500 -- 800 nm has a slope of (19.6 ± 0.7)% per 100 nm, which is considered moderately red and well within the slope range of various samples of the TNO population (see e.g. Lacerda et al. 2014). Barucci et al. (2011) report spectroscopy performed with VLT/ISAAC (1.1 -- 1.4 µm) and VLT/SINFONI (1.49 -- 2.4 µm). They calculated the rela- tive flux difference D between 1.71 -- 1.79 µm and 2.0 -- 2.1 µm, which could hint at a possible water ice feature around 2.0 µm. Owing to the limited signal-to-noise ratio (S/N) of the data, a presence of this type of feature could not be confirmed with suf- ficient statistical significance (D = (11 ± 7)%). Using Hapke's radiative transfer model, a synthetic spectrum was calculated by considering various compounds. The best Hapke model fit is based on a geometric albedo (in V-band) of pV = 0.20 (a value that we disprove in our subsequent analysis) and results in an estimated surface composition of 12% amorphous wa- ter ice, 20% kaolinite, 17% Titan tholins, 32% Triton tholins, 4% kerogen and 15% carbon. However, the authors conclude that "much of the information obtained from spectral modeling is nonunique, especially if the albedo is not available, the SNR is not very high and / or there are no specific features of particular components", so these findings need to be viewed with caution. For a full discussion of this spectral analysis, see Barucci et al. (2011). All spectral data available from literature is summarized in Figure 1. We conclude that the spectral data available at this point cannot reliably constrain surface composition. Estimating a size from the magnitude of an object is im- possible without the knowledge of its geometric albedo, as a small object with a highly reflective surface cannot be distin- guished from a large object with a dark surface. The thermal emission of the target can be used as a second constraint to over- come this ambiguity. Santos-Sanz et al. (2012) report results de- Fig. 1. Visible and near-infrared reflectance spectra and photometry, ob- tained using multiple instruments at ESO's VLT on 21 and 22 Septem- ber 2008. To concatenate spectral data from different instruments, pho- tometry (purple) was converted to relative reflectance (black) and used as an indicator for alignment (see discussion in Barucci et al. 2011). Zones marked in gray denote telluric bands of the Earth's atmosphere. See text for details. rived from far-infrared photometry in three bands (60 -- 85 µm, 85 -- 120 µm, 130 -- 210 µm) using Herschel/PACS. By applying a hybrid standard thermal model, they conclude that 2007 UK126 has a diameter in the range of d = 599 ± 77 km, a geometric albedo (in V-band) of pV = 16.7+5.8−3.8%, and a beaming factor of η = 1.20 ± 0.35, a factor that is empirically found to scale modeled surface temperature to observations (see discussion in Müller et al. 2009). Santos-Sanz et al. (2012) also conclude that a beaming factor close to η = 1 indicates "very low thermal inertia and/or very large surface roughness", which could im- ply a highly porous surface with very low thermal conductivity as a possible scenario. The derived size implies that the body is a dwarf planet candidate that could have formed into a reg- ular, round shape due to its own gravity. This depends on the density and material strength of its compounds and its rotation period. For incompressible fluids, figures of hydrostatic equi- librium have been derived, of which two are of particular im- portance: A Maclaurin spheroid (an oblate spheroid with axes a = b > c) and a Jacobi ellipsoid (all three axes a > b > c hav- Article number, page 3 of 16 Wavelength [µm]0.40.81.21.622.4Relative Reflectance0.20.40.60.811.21.4VLT/FORS2 (Fornasier et al. 2009)VLT/ISAAC (Barucci et al. 2011)VLT/SINFONI (Barucci et al. 2011)VRIJH Reflectance (Barucci et al., 2011)VRIJH Photometry (Perna et al., 2010)Apparent Magnitude1818.51919.52020.521A&A proofs: manuscript no. Schindler ing different length, Chandrasekhar 1967). However, the discus- sion of these shapes only represents a theoretical limiting case, since bodies made of solid matter have mechanical strength. As pointed out by Sheppard & Jewitt (2002), a fractured interior as a result of impacts could lead to fluid-like behavior, but conditions are never determined solely by hydrodynamics. Thirouin et al. (2014) report photometric light curves taken at the Telescopio Nazionale Galileo (TNG) over a time span of ≈ 10 h in total, distributed over three nights in October 2011. The obtained light curve of 2007 UK126 is very flat and varies by just ∆m = 0.03 ± 0.01 mag peak-to-peak. The light curve amplitude depends on the shape of the body, the viewing ge- ometry and albedo variation across the surface, all three be- ing unknown factors. Both a single-peaked (Maclaurin spheroid with albedo inhomogeneity) and double-peaked (Jacobi ellip- soid) light curve are plausible. Thirouin et al. (2014) have used a criterion of ∆m = 0.15 mag peak-to-peak amplitude to distin- guish between albedo and shape-related light curve variations; consequently, this object was analyzed assuming a single-peak solution. Still, they were unable to estimate a secure sidereal ro- tation period P, since data indicate multiple possible solutions of P = {11.05 h, 14.30 h, 20.25 h}, giving the P = 11.05 h so- lution only a minimal higher likelihood compared to the other two. They conclude that their data only allow the rotation pe- riod to be constrained with a lower limit of P > 8 h at this time. As a rough first indicator, Thirouin et al. (2014) also derived a lower density limit of ρ > 0.32 g cm−3 based on the assump- tion of a homogeneous Jacobi ellipsoid in hydrostatic equilib- rium that rotates in P = 8 h. They note that this assumption on shape is in conflict with the preference for an oblate spheroid, and emphasize that this estimate is very vague and can be un- realistic. Thirouin et al. (2014) also state an absolute magnitude of H = 3.4 mag without specifying a filter band (presumably R) or uncertainty estimate, and indicate that no absolute photome- try has been derived; hence we discarded their value. As we now have additional information on the peak-to-peak amplitude of the light curve, we can reapply the approach from Santos-Sanz et al. (2012), as explained earlier, to derive an improved uncertainty estimate of their absolute magnitude measurements. We obtain ∆H = (cid:113)(cid:0)0.04 mag(cid:1)2 + (cid:16) 0.88 · 0.04 mag (cid:17)2 = 0.044 mag. 2 Hubble Space Telescope (HST)/WFPC2-PC observations on 13 November 20084 have revealed the existence of a satellite (Noll et al. 2009; Santos-Sanz et al. 2012), but its orbit could not be determined so far. The satellite has a magnitude difference of ∆m = 3.79 ± 0.24 mag, measured in the HST/WFPC2-PC F606W band (599.7 nm±75 nm, see Mikulski Archive for Space Telescopes for more details5). 3. Observations 3.1. Visual observations of the stellar occultation 2007 UK126 occulted USNO CCD Astrograph Catalog 4 star UCAC4 448-006503 in the constellation Eridanus (RAJ2000 = 04h 29m 30.6s, DecJ2000 = −00° 28(cid:48) 20.9(cid:48)(cid:48), apparent magni- tude mV = 15.86 mag (V-band), mB = 17.00 mag (B-band), mJ = 14.34 mag (J-band)6) on 15 November 2014 UTC. At the 4http://www2.lowell.edu/users/grundy/tnbs/229762_ 2007_UK126.html 5https://archive.stsci.edu/ 6http://vizier.u-strasbg.fr/viz-bin/VizieR-5?-ref= VIZ5599d6063ad7&-out.add=.&-source=I/322A/out&UCAC4= ==448-006503 Article number, page 4 of 16 time of the occultation, the target had a geocentric distance of rG = 42.572 AU and a heliocentric distance of rH = 43.47 AU. It was close to opposition (01 December 2014, phase angle χ = 0.56) and had an approximate apparent visual magnitude of mV ≈ 19.84 mag. Details on the prediction of this event can be found in Camargo et al. (2014). The event was announced in October 2014 via the RIO TNO Events data feed7 in OccultWatcher8, a software program writ- ten by H. Pavlov (Pavlov 2014) that allows coordination among professional and amateur observers, planning of campaigns, and deployment of mobile setups across the predicted path to acquire as many chords as possible. Before the event, 25 stations regis- tered to attempt an observation of the event, while 16 stations reported back afterwards. In addition, this event was announced in a private email by J. L. Ortiz. Unfortunately, weather conditions clouded out most ob- servers in California and the south western U.S.. Three observers reported a successful observation of the event on OccultWatcher. Table 2 provides an overview of their locations and setups. It turned out that all successful observers were coincidentally po- sitioned with almost equal spacing to each other in relation to the shadow path, sampling the shadow almost symmetrically to the centerline. Schindler & Wolf observed at about 89 -- 91% humidity and low transparency sky conditions in California; for- tunately, the event happened during a larger gap in a high cir- rus cloud cover seen around the time of the event. The 60 cm Astronomical Telescope of the University of Stuttgart (ATUS)9, located at Sierra Remote Observatories about an hour's drive north-east of Fresno, was controlled remotely from Germany via internet connection. Bardecker, located in Nevada, also re- ported that high cirrus clouds moved in about two minutes after the event. Olsen reported clear and stable weather conditions in Illinois. Schindler & Wolf were observing with an Andor iXon DU- 888 camera with a back-illuminated EMCCD sensor in 2×2 binning mode. Being cooled to −80°C via an internal thermo- electrical cooler, exposures are virtually free of dark current. The frame-transfer design of the EMCCD allows a fast shift of the accumulated charge from the light-sensitive image area to a masked storage area that is read out subsequently, while the next exposure is already under way in the image area. This allows vir- tually gap-free imaging, with a dead time between frames of only 3.45 ms (the time to shift an image from the light-sensitive to the storage area). The video signal is digitized in the camera and transmitted to a proprietary PCIe frame grabber and controller card. Image acquisition was controlled with Andor SOLIS run- ning on Windows 7 64 bit. Running the camera in frame-transfer mode, subsequent series of 50 frames with 2 s integration time were taken and written directly to the PC's hard drive into three- dimensional FITS files without buffering image data in the com- puter's main memory (spooling). Between two image series, a gap of about 300 ms cannot be avoided owing to the non-real- time behavior of the operating system. A number of 50 images per series was chosen as a compromise between achieving a high probability to capture the disappearance and reappearance event during an image series while still achieving a timing accu- racy well below 1 ms. Precise time stamps were logged with a 7Maintained by D. Gault, Australia, publishing predictions from the RIO TNO Group (with members from Observatório Nacional/MCTI and Observatório do Valongo/UFRJ, Rio de Janeiro, Brazil; and Ob- servatoire de Paris-Meudon/LESIA, Meudon, France). 8www.occultwatcher.com 9https://www.dsi.uni-stuttgart.de/forschung/atus. html Schindler et al.: Results from an occultation and FIR photometry of (229762) 2007 UK126 Table 2. Locations, setups and image acquisition parameters of successful observers, and disappearance / reappearance times derived from obtained light curves. Observer Closest city Latitude N (deg mm ss.ss) Longitude W (deg mm ss.ss) Altitude (m) Telescope type Aperture, focal ratio Camera Sensor type Sensor cooling (temperaturea ) Frame grabber card File format GPS time logger Integration time (s) Camera-internal frame accumulation Cycle time per frame (s) Disappearance (D) time from square-well fit (UTC) Reappearance (R) time from square-well fit (UTC) Duration of event (s) Chord length (km) SNRc , star & target combined Schindler & Wolf Alder Springs, CA 37 04 13.50 119 24 45.00 1405 Ritchey-Chrétien 600 mm, f/8 Andor iXon DU-888E-C00-BV Frame transfer EMCCD (e2v CCD201-20, back-illuminated, grade 1) Thermo-electrical (−80°C) Andor CCI-24 PCIe FITS, uncompressed Spectrum Instruments TM-4, triggered by electronic shutter of camera 2.000 none Olsen Urbana, IL 40 05 12.40 88 11 46.30 224 Newtonian 500 mm, f/4 Watec 120N+ Interline transfer CCD (Sony ICX418ALL) None (≈ −9°C) Pinnacle DVC-100 Video, uncompressed Bardecker Gardnerville, NV 38 53 23.53 119 40 20.32 1534 Schmidt-Cassegrain 304 mm, f/3.3 MallinCam B/W Special Interline transfer CCD (Sony ICX428ALL-A with micro lenses) None (≈ +4°C) Hauppauge USB-Live2 Video, Lagarithb codec Kiwi Video Time Inserter (VTI), overlaying GPS time stamp on camera frames IOTA Video Time Inserter (VTI), overlaying GPS time stamp on camera frames 4.271 128 × 1/29.97 2.135 64 × 1/29.97 2.00345 10:19:24.356 ± 0.159 0.03337 10:18:05.447 ± 1.032 0.03337 10:19:29.151 ± 0.452 10:19:50.249 ± 0.159 10:18:26.802 ± 1.032 10:19:48.370 ± 0.452 25.893 ± 0.225 641.0 ± 3.9 ≈ 12.6 21.355 ± 1.459 525.9 ± 25.4 ≈ 4.1 19.219 ± 0.639 475.7 ± 11.2 ≈ 4.7 Notes. None of the observers have used any broadband filter. (a) Given temperatures are measured at the image sensor (Schindler & Wolf) or close to the telescope setup at the time of recording (Olsen, Bardecker). (b) Lossless video compression algorithm. (c) Signal-to-noise (S/N) ratio per data point. Spectrum-Instruments TM-4 GPS receiver that was triggered di- rectly via a TTL signal from the shutter port of the camera every time a series of 50 images started. Time stamps of the 2nd -- 50th image in a series have been extrapolated based on the time stamp of the first image and the well-known cycle time. The total un- certainty of the time stamp for each exposure is estimated to be < 0.25 ms and consists of the following components: (a) the uncertainty of the TM-4 GPS time measurement of ± 10 ns (Spectrum Instruments, Inc. 2014), (b) the synchronization of the TTL signal generated at the shut- ter port of the camera with the beginning of an exposure bet- ter than 1 µs (Andor Technology, personal communication), (c) cycle time that is known with an accuracy of ± 5 µs, which leads to an accumulated timing uncertainty of ± 245 µs after extrapolating the time stamp for the 50th frame of a series. Consequently, the cycle time uncertainty dominates the total uncertainty of the time measurement. The event was captured in frames 30 -- 43 of an image series. A time-lapse animation of the frames acquired during the event is provided in Figure 2 and online. The setup was inspired by an earlier system described by Souza et al. (2006) and is based on very good experiences with this type of camera during characterization measurements of the SOFIA telescope (Pfüller et al. 2012) and the subsequent upgrade of SOFIA's focal plane imager (Wolf et al. 2014). Dif- ferential aperture photometry was performed relative to five field stars in AstroImageJ10 (Collins & Kielkopf 2013; Collins 2015; Collins et al. 2016), an image processing program written in Java on the basis of ImageJ (Schneider et al. 2012). The aperture ra- dius was initially set to 3 pixels and subsequently varied based on the estimated full width at half maximum (FWHM) of the star image (radial profile mode). The background annulus size was set to an inner radius of 8 and an outer radius of 16 pixels; no background stars were visible within the annuli. Olsen and Bardecker used monochrome video cameras that employ uncooled interline CCDs. Ambient temperature during data recording was −9°C (Olsen) and +4°C (Bardecker); sensor temperatures could not be measured and were somewhat higher than ambient owing to heat dissipated by the camera's sensor and electronics. An interline CCD has a masked storage col- 10http://www.astro.louisville.edu/software/ astroimagej/ Article number, page 5 of 16 A&A proofs: manuscript no. Schindler the measured signal. Therefore, data have to be binned accord- ingly over n frames, corresponding to each image accumulation interval. During data reduction, internal camera delays specific to each camera model (e.g. the accumulated image is delayed by a number of fields at the camera output) and VTI (e.g. the time stamp is delayed by a field) need to be considered to precisely establish the start of each individual exposure. The captured digital video has been analyzed with Tangra 3 (Pavlov 2014) and Limovie (K. Miyashita12) to extract the un- corrected time stamp for each frame from the video and to conduct differential aperture photometry of the target and field stars. While a classical annulus was used in Olsen's dataset for background estimation, two annulus sectors had to be used in Bardecker's dataset as a satellite trail was recorded in di- rect proximity of the target star in the frame integrated from 10:19:48.370 to 10:19:50.505 UTC. This accumulated frame recorded the reappearance of the target star at Bardecker's loca- tion; the profile of the target star could be clearly distinguished from the background. The two annular sectors were carefully po- sitioned to avoid any pixels being contaminated by the satellite trail for local background estimation. A slight contamination of the central aperture could not be avoided, which resulted in a slightly overestimated flux of the target star, as seen in Figure 3 in the first data point after the estimated end of the occultation at Bardecker's site. Usage of a classical annulus for background es- timation would have caused an overestimated background level, similar in counts to the measurement in the central aperture on the reappearing star, thus hiding the reappearance of the star in the derived light curve, and overestimating the length of the event at Bardecker's location by 2.135 s. Time stamp corrections were applied as described in George (2014) based on measurements by G. Dangl13, with the help of the software package R-OTE14 (George & Anderson 2013): For Bardecker's dataset, a correction of −2.1355 s accounting for camera delay and −0.01668 s accounting for VTI delay was made; for Olsen's dataset, applied corrections were −4.2876 s and −0.01668 s, respectively. The difference in S/N between the three datasets is caused by differences in telescope aperture size, noise characteristics of the respective camera (e.g. dark noise, amplifier noise, read noise of A/D converters), and sensor design affecting quantum efficiency (back-illuminated frame-transfer CCD with fill fac- tor = 1 vs. front-illuminated interline CCD with fill factor < 1, with or without micro lenses; differences in AR coating). Given the small telescopes used in this campaign, the low S/N was a necessary compromise to sample the faint star at an acceptable cadence, while integration times have been chosen wisely to go to the limit of the respective equipment. 3.2. Far-infrared observations with Herschel/PACS In addition to the data obtained from the occultation, we have revisited far-infrared photometric data that were previously ac- quired with the PACS photometer (Poglitsch et al. 2010) on- board Herschel (Pilbratt et al. 2010). Table 3 provides a sum- mary of all observations of 2007 UK126 that were made on 08 and 09 August 2010. This dataset has been reduced by images optimized by source constructing double-differential 12http://www005.upp.so-net.ne.jp/k_miyash/occ02/ limovie_en.html 13http://www.dangl.at/ausruest/vid_tim/vid_tim1.htm 14http://www.asteroidoccultation.com/observations/ NA/ Fig. 2. Time-lapse animation of the stellar occultation as recorded by Schindler & Wolf (click here to view the animation in a web browser). The time stamp indicates the start of each individual exposure; frames have been cropped to 210 x 210 pixels. umn next to each light sensitive imaging column. All accumu- lated charges are simultaneously shifted sidewards into masked columns that are read out subsequently while the next frame is already acquired in the light sensitive area. Thanks to this de- sign, these image sensors are also virtually free of any dead time and widely used in video cameras. The disadvantages compared to frame-transfer CCDs (that mostly find application in scientific cameras) are their reduced light-sensitive area and less optimized noise characteristics. The disadvantage of a reduced fill factor is partly mitigated by the use of micro lenses that are placed on each pixel of the CCD, which is the case in the MallinCam B/W Special used by Bardecker. This results in a higher sensitivity compared to the Watec 120N+ used by Olsen that has a sensor without micro lenses. Both types of cameras are widely used by amateurs who routinely conduct observations of occultations and report their results to the International Occultation Timing As- sociation (IOTA). Based on the monochrome EIA video signal standard (Engineering Industries Association (EIA) 1957), these cameras are running with a fixed frame rate of 29.97 frames/s (59.94 fields/s) and have an analog video output. One frame has 525 lines and consists of two fields that are interlaced (first field even-numbered lines, second field odd-numbered lines). To achieve longer integration times, a number of n frames can be accumulated internally in the camera. Since the camera and its output are running at a fixed frame rate, the camera provides an accumulated frame n-times at its output, while it is accumulating the next frame (see Figure 3). A video time inserter (VTI) was used to overlay a GPS-provided time stamp on the lower part of each field coming from the camera's analog video signal out- put. The time-stamped analog video signal was then digitized directly to a Windows PC via a USB video capture device by using the software VirtualDub11. Olsen captured the video into uncompressed video files of 4 min length, while Bardecker used the lossless Lagarith video codec and recorded in segments of 5 min. The event was captured by both observers well within a video segment. Data reduction differs from classical CCD imaging in two ways: The camera returns the same accumulated frame n times, superimposed with noise from analog amplifiers and the analog- digital conversion of the video capture device. This means that the accumulated frame is sampled n times, so the average value of all samples in the respective bin is a representative value of 11http://www.virtualdub.org/ Article number, page 6 of 16 Schindler et al.: Results from an occultation and FIR photometry of (229762) 2007 UK126 Table 3. Summary of observations of 2007 UK126 with Herschel/PACS. OBSID 1342202277 1342202278 1342202279 1342202280 1342202324 1342202325 1342202326 1342202327 Start time (UTC) 2010 Aug 08 17:52:17 2010 Aug 08 18:02:48 2010 Aug 08 18:13:19 2010 Aug 08 18:23:50 2010 Aug 09 12:47:55 2010 Aug 09 12:58:26 2010 Aug 09 13:08:57 2010 Aug 09 13:19:28 Bands B/R B/R G/R G/R B/R B/R G/R G/R β (deg) 110 70 110 70 70 110 110 70 Notes. OBSID denotes a Herschel internal ID for each observation. All observations were done in standard mini scan-map mode with a scan- leg length of 3(cid:48), a scan-leg separation of 4(cid:48)(cid:48), and a total of 10 scan legs, resulting in a duration of 603 s per acquisition. For more details on the mini scan-map mode, its data reduction and calibration, see Balog et al. (2014). The scan-maps were taken with the specified orientation angle β relative to the instrument. During each observation, data were taken simultaneously in two bands: B - blue band (60 - 85 µm), G - green band (85 - 120 µm), R - red band (130 - 210 µm). All observations were conducted with a repetition factor of 2, i.e. all maps have been observed twice. matching, which leads to a higher S/N compared to supersky- subtracted images used in a previous reduction by Santos-Sanz et al. (2012). Also, the PACS pipeline flux calibration is based on standard stars and asteroids that are significantly brighter than the target. The absolute photometric calibration of our reduc- tion is instead based on a number of faint standard star mea- surements. The applied data reduction technique is described in detail in Kiss et al. (2014). The fluxes obtained in each band are summarized in Table 4. We have also checked data from the Wide-field Infrared Sur- vey Explorer (WISE, Wright et al. 2010) satellite for possible detections of 2007 UK126. Although two possible detections in images taken on 10 January 2010 are listed in the WISE All- Sky Known Solar System Object Possible Association List15, we concluded that those are false positives for several reasons: The associated sources are off by several arcsec from their predicted positions, and our thermophysical modeling (see Section 5) pre- dicts fluxes in the WISE bands that are well below the respective detection limits. Both Herschel/PACS and WISE measurements were taken close in time, i.e. at very similar observing geometry, so it appears safe to say that 2007 UK126 was below the detection limits of WISE. Also, the Minor Planet Center does not list any entries for 2007 UK126 reported from WISE. 4. Results from the stellar occultation 4.1. Occultation light curves and square-well fits The integration time and achieved S/N of each camera dictate the fundamental limit of how accurate the moment of disappearance and reappearance of the occulted star can be determined on each chord. The expected light curve resembles a square-well if the following conditions are fulfilled: 1. The body has no atmosphere, or one that is so thin that it cannot be detected, given the sampling rate and S/N. 15http://irsa.ipac.caltech.edu/cgi-bin/Gator/ nph-scan?submit=Select&projshort=WISE 2. Effects of Fresnel diffraction and the finite stellar diameter are negligible. Considering the peak sensitivity wavelength of the Si-based CCD cameras (λ ≈ 650 nm) and the geocentric distance of 2007 UK126, the Fresnel scale is F ≈ 1.44 km (Roques et al. 1987). The estimated angular stellar diameter is about 0(cid:48)(cid:48).0097 (van Belle 1999), or 0.3 km projected at the geocentric distance of 2007 UK126. Given the very high speed of the shadow on the Earth's surface of about v = 24.1 km s−1 and the three orders of magnitude larger size of 2007 UK126, it becomes clear that both effects can be neglected for the dataset at hand. The light curves that have been obtained by the three ob- servers are illustrated in Figure 3, together with derived square- well fits. All relative fluxes have been normalized based on the average combined relative flux of star and TNO in all available exposures before and after the occultation ("baseline"). None of the light curves indicate a gradual decline and emergence of the star, so a square-well fit is a legitimate approximation. Given the achieved sampling rates, the presence of a thin at- mosphere cannot be excluded; this question remains to be stud- ied during future occultations that allow for faster sampling (i.e. brighter star or availability of larger telescopes). Given the quasi- zero dead time of all cameras used in this study, disappear- ance and reappearance of the occulted star must have occurred (with near-100% certainty) during integration of an exposure (a single frame for Schindler/Wolf, or n accumulated frames for Bardecker and Olsen). The measured relative flux in the respec- tive exposure is consequently smaller than the combined relative flux of the star and the TNO, but larger than the TNO's relative flux alone. The S/N of the dataset of Schindler & Wolf is sufficient to clearly isolate the frames that recorded disappearance and reappearance, and to interpolate their time stamps to subframe accuracy. As the response of the CCD is linear, the ratio be- tween the measured relative flux in the disappearance or reap- pearance frame and the baseline is directly proportional to the offset of disappearance or reappearance in time relative to the start of the respective exposure16. An upper and lower limit of this ratio has been derived using the normalized relative flux er- ror σSignal = (S/N)−1, which propagates directly into the un- certainty of the estimated disappearance and reappearance times σD/R = σSignal tint, as given in Table 2. Unfortunately, the S/Ns of the datasets of Olsen and Bardecker do not allow for the identification of the accumu- lated frame that recorded disappearance and reappearance. This becomes clear from the light curves: None of the accumulated frames has a relative flux that separates it from the upper or lower baseline with statistical significance. Thus, we decided to apply square-well fits that are coincident with the sampling frequency. Again, the normalized relative flux error has been used as an un- certainty estimate to calculate timing uncertainties (see Table 2). We note that the R-OTE software package uses the Akaike in- formation criterion (AIC) as a logic to decide objectively if a square-well fit shall be applied to exposure timing (3 parame- ter model) or sub-exposure timing (4 parameter model). While the goodness of fit improves with an additional parameter, the fit is not necessarily a better representation of the data at hand. By weighing the number of parameters of a model against its 16This simplification is only accurate for a body without any atmo- sphere. Assuming a gradual dimming would have happened on shorter time scales than the integration time, the gradual transition could have been integrated by a single image. The length of the chord would then be slightly overestimated. Article number, page 7 of 16 Table 4. Herschel/PACS fluxes of 2007 UK126, derived from a combination of all data available in a given band. A&A proofs: manuscript no. Schindler Band λref (µm) Mid-time B G R 70 100 160 2010 Aug 09 03:30 2010 Aug 09 03:51 2010 Aug 09 03:41 Herschel-centric rH (AU) 44.94437 44.94436 44.94437 ∆ (AU) 45.16375 45.16351 45.16363 α (deg) 1.27 1.27 1.27 In-band flux (mJy) FD (mJy) 1σ (mJy) 11.43 ± 1.20 13.20 ± 1.47 8.08 ± 2.03 11.55 13.33 8 1.34 1.63 2.05 Notes. For all details on the data reduction, see Kiss et al. (2014). The photometric uncertainty is derived from aperture photometry on 200 artificial sources, resembling the PACS PSF in the respective band, that have been planted individually on the double-differential image. The distribution of measured artificial source fluxes resembles a Gaussian, so its standard deviation defines the photometric uncertainty of the respective image. λref - reference wavelength for respective band, rH - heliocentric distance at mid-time, ∆ - Herschel-centric distance at mid-time, α - phase angle at mid-time, FD - color-corrected monochromatic flux density at reference wavelength, 1σ - estimated uncertainty of FD (includes the 5% absolute flux calibration error of PACS). Fig. 3. Obtained light curves for the occultation on 15 November 2014. The occultation was recorded at different times due to different longitudes and latitudes of the observatories; the shadow was moving from east to west (c.f. Figure 5). Light curves were normalized based on the combined relative flux of the star UCAC4 448-006503 and TNO 2007 UK126, averaged over all available images before and after the event. The occulted star's apparent V-band magnitude is mV = 15.8 mag, which was four magnitudes brighter than 2007 UK126 (mV ≈ 19.8 mag). Raw data from Bardecker and Olsen is plotted in light red and light green; their video cameras were running at a fixed frame rate, accumulating n frames internally and returning the result n times at the analog video output (see Section 3.1 for details). Consequently, the relative flux was averaged over each bin of 64 (Bardecker, dark red) and 128 samples (Olsen, dark green; see discussion in text). Data points by Schindler & Wolf (blue) were derived from single images, with relative flux errors (1σ) plotted in light blue. The black lines represent square-well fits that were used to derive the disappearance and reappearance times for each chord (see text for details). goodness of fit, AIC provides a formal way to decide if sub- exposure timing can be applied. For the two datasets of Olsen and Bardecker, the AIC ruled in favor of the simpler model. 4.2. Size of the shadow from a best-fitting ellipse The reconstruction of shadow size and shape is done in the geo- centric fundamental (Besselian) plane, which is perpendicular to the line connecting the apparent position of the occulted star UCAC4 448-006503 (at the epoch of the observation, trans- formed from the J2000 UCAC4 position) and the position of 2007 UK126 as given by the ephemeris. In the fundamental plane, the shadow of 2007 UK126 is not distorted by Earth's curvature, i.e. shape and size of the shadow can be directly derived. Each Article number, page 8 of 16 observer's position on the Earth's surface (latitude, longitude) is projected twice onto the fundamental plane: At the disappear- ance (D) and reappearance (R) time of the star as measured by the respective observer. This results in six event coordinates on the fundamental plane that represent points on the limb profile of 2007 UK126. A reference frame is then defined on the fun- damental plane, which is fixed with respect to the shadow of 2007 UK126, or in other words, moving in the fundamental plane with the shadow. After transforming the D and R coordinates from the fundamental plane into the moving reference frame, the shadow is reconstructed. Details of this reduction approach can be found in, for example, Millis & Dunham (1989), Millis & Elliot (1979) and Wasserman et al. (1979). Time Stamp (UTC)10:17:3010:18:0010:18:3010:19:0010:19:3010:20:0010:20:30Relative Flux, Normalized, Shifted0123456Bardecker (Binned Raw Readout)Schindler / Wolf (with 1< error bars)Olsen (Binned Raw Readout)Schindler et al.: Results from an occultation and FIR photometry of (229762) 2007 UK126 Based on the assumption that 2007 UK126 is of sufficient size that its shape can be approximated either by a Maclaurin or a Jacobi ellipsoid, its projected shadow on the Earth's surface is expected to be an ellipse. For a subsequent analysis of size and shape, we need to find the ellipse that best fits to the measured disappearance (D) and reappearance (R) times (or respectively, the D and R coordinates in the reference frame moving with the shadow). An ellipse can be described by five geometrical param- eters: Its center coordinates xEl, yEl, the major and minor axes aEl and bEl, and the orientation angle of the major axis θEl. Fortu- nately, the availability of three chords covering both sides of the shadow path allows for an ellipse fit without taking additional assumptions. We used Occult17, a software program written and continu- ously developed by D. Herald and used by occultation observers worldwide for predictions of events and subsequent data anal- ysis. Using Occult, we translated measured D and R times and their derived uncertainties (see previous section) into D and R coordinates with propagated uncertainties in the moving refer- ence frame, and calculated an initial ellipse fit (for details, see Herald 2016). To verify these results independently, we recalcu- lated the ellipse fit with our own code written in Matlab and re- alized that the algorithm implemented in Occult has a significant shortcoming. Occult uses a direct least-square ellipse fit algo- rithm, where uncertainties of D and R timings have no effect on the resulting ellipse and are not propagated into the uncertainties associated to the ellipse parameters. Timing error estimates are collected merely for archiving purposes at this moment. The un- certainties of the geometric ellipse parameters provided by Oc- cult are solely derived from the residuals between the ellipse and each D or R coordinate, measured along the line between ellipse center and D or R. Although Occult allows the user to specify weights to differentiate between different levels of data quality, these weights only influence the estimated uncertainties of the ellipse parameters, not the ellipse fit itself. It became clear that the desired propagation of timing errors into the ellipse fit leads to a non-linear errors-in-variable prob- lem. We found a perfectly suited algorithm recently described by Szpak et al. (2015) that offers a solution for exactly this type of problem. The D and R uncertainties can be described as inde- pendent Gaussian noise with zero mean that is inhomogeneous, i.e. that depends on the respective setup, integration time and observing conditions. Therefore, we can use the propagated D and R uncertainties in the moving reference frame to define a covariance matrix for each D and R coordinate. The algorithm uses these covariance matrices as weights in a cost function that needs to be minimized to determine the best-fitting ellipse. After the algorithm has calculated the algebraic parameters of the el- lipse and their corresponding covariances, a conversion and co- variance propagation to the ellipse's geometrically meaningful parameters is conducted. In this way, we get a reliable and ro- bust estimate of the uncertainties of all parameters of interest, while the ellipse fit itself considers the noise in the data points. Finally, a confidence region in the plane of the ellipse is calcu- lated, illustrating the zone in which the true ellipse is located at a given probability. We chose a planar 68.3% confidence region to be consistent with other literature in the field instead of the 95% region suggested by Szpak et al. (2015), e.g. for industrial machine vision applications. The resulting geometrical parameters of the best fitting el- lipse and their associated uncertainties are listed in Table 5. The ellipse fit and its 68.3% confidence region is illustrated in 17http://www.lunar-occultations.com/iota/occult4.htm Table 5. Geometric parameters of the best fitting ellipse and associated uncertainties. Parameter aEl (km) bEl (km) xEl (km) yEl (km) θEl (deg) Value 645.80 ± 5.68 597.81 ± 12.74 6.57 ± 2.06 −10.82 ± 3.13 21.25 ± 5.65 Fig. 4. The derived ellipse fit and its 68.3% confidence region (gray). Uncertainties of each individual ellipse parameter are summarized in Table 5. The angle θ gives the rotation of the major axes, measured from east to north. The plotted error bars of the D and R locations have been derived from the timing deviations given in Table 2. The major and minor axes are indicated by the dotted lines. The distances provided in the legend are topocentric with respect to the centerline derived for the best fit (dash-dotted line). The chord by Schindler & Wolf sampled 2007 UK126 almost at the centerline, while the chords by Bardecker and Olsen sampled the upper and lower part quasi symmetrically. The geocentric distance of 2007 UK126 was rG = 42.572 AU at the time of the observation. Figure 4. The positive y axis is directed to the north, while the positive x axis indicates east. The origin (x,y) = (0,0) represents the average of the minimum and maximum event coordinate in north-south and east-west direction. Figure 5 illustrates the reconstructed shadow path on the sur- face of the Earth, where the shadow's upper and lower bound- ary and centerline are based on the size and orientation of the ellipse as illustrated in Figure 4. An animation that illustrates the shadow traveling across the continental US and the measure- ments conducted in parallel is provided as supplementary ma- Article number, page 9 of 16 x [km]-400-2000200400y [km]-400-2000200400600800aElbElBest fitting ellipseDerived center line68.3% confidence region of fitBardecker (+201.7 km)Schindler / Wolf (+32.0 km)Olsen (-174.6 km)NEA&A proofs: manuscript no. Schindler Substituting the magnitudes in equation 1 with the R-band estimates HR,TNO = 3.07 ± 0.044 mag (see Section 2) and HR,Sun = −27.12 mag (Bessel R18), we obtain a geometric albedo in the R-band of pR = 19.5 ± 2.0%. Fig. 5. Reconstructed shadow path and geographic locations of all ob- servers across the continental US. The three lines indicate the shadow path's northern boundary, centerline, and southern boundary, assuming an elliptical shape for 2007 UK126 based on the ellipse fit presented in Figure 4. The arrows indicate the direction of travel of the shadow on the Earth's surface from east to west. An animation illustrating the event and the acquired light curves is provided as supplementary material on the A&A website. terial on the A&A website. We conclude that the relative mea- surement uncertainty of the ellipse axes is in the order of 0.9% and 2.1%, respectively. The axial ratio of the ellipse translates to aEl/bEl = 1.080 ± 0.025, i.e. a circular fit and hence a pole-on viewing geometry can already be ruled out geometrically. 4.3. Albedo El b−1 El , Thanks to the occultation measurement, we can improve the ge- ometric albedo estimate. Following the formula pV = 4 · 100.4(HV,Sun−HV,TNO) · (1 AU)2 a−1 (1) where HV,Sun = −26.78 mag is the absolute magnitude of the Sun (Bessel V18), HV,TNO = 3.69 ± 0.044 mag is the ab- solute magnitude of 2007 UK126 in V-Band (see Section 2, al- though this value might include a small flux contribution from a potential satellite) and 1 AU = 1.49598 · 108 km. Consider- ing propagation of all measurement errors (see Appendix B), we derive pV = 15.0 ± 1.6%. This calculation assumes that, between the epochs of the photometric measurements (Septem- ber 2008) and the occultation (November 2014), the projected area did not change, i.e. the pole orientation of the body as seen from Earth remains virtually the same. This assumption is rea- sonable: As pointed out by Sheppard & Jewitt (2002), with ref- erence to Harris (1994), an object of the size of 2007 UK126 has a damping timescale for its non-principal axis rotation ("wob- ble") that is significantly smaller than the age of the solar sys- tem. Considering the orbital period of 636.73 yr, a body in pure principal axis rotation would appear in almost the same projec- tion on such short timescales. The new albedo estimate is within the previously derived range by Santos-Sanz et al. (2012) of pV = 16.7+5.8−3.8%. 2007 UK126 is one of eight detached objects that have been observed with Herschel in the far-infrared and subsequently an- alyzed via thermophysical modeling. Based on this sample, Lac- erda et al. (2014) derived a median albedo for detached objects of pV,DO = 16.7%. Considering our new estimate, the median albedo for this class of objects shifts to pV,DO = 15.0%, but the sample size is too small for reliable statistical analysis. 4.4. Shape and size The very small light curve variations (∆mV = 0.03 ± 0.01 mag peak-to-peak, see Section 2) are another strong indicator of a Maclaurin spheroid with albedo variations that is not seen pole- on. From spacecraft fly-bys, we know about objects in the so- lar system that were shaped into oblate spheroids by their own gravity, but which are considerably smaller than 2007 UK126: The icy Saturnian satellites Mimas (ρ = 1.149 ± 0.007 g cm−3, deq = 396.4 ± 0.8 km, Roatsch et al. 2009), Enceladus (ρ = 1.609 ± 0.005 g cm−3, deq = 504.2 km, Roatsch et al. 2009) and Miranda (ρ = 1.20 ± 0.14 g cm−3, deq = 472 ± 3 km, Jacob- son et al. 1992). We feel it is a reasonable assumption to con- strain our following analysis to a Maclaurin spheroid that rotates around its principal axis c, and to discard the Jacobi solution. Viewing a Maclaurin spheroid pole-on (θ = 0°) would lead to a circular shadow during an occultation, while viewing it equator-on (θ = 90°) would lead directly to the elliptical shadow that was measured. While a circular fit (and hence a pole-on view) has been ruled out geometrically, the ellipse derived in the previous section represents a distorted projection of the true shape of 2007 UK126 in any other pole orientation than equator- on. For a Maclaurin spheroid, we can calculate the flattening ra- tio a/c from the major axis aEl and minor axis bEl of the shadow ellipse and the angle θ between the spheroid's rotation axis and the line of sight: (cid:118)(cid:117)(cid:117)(cid:116) sin2 θ a c = . (2) − cos2 θ b2 El a2 El We do not know θ at the time of the occultation. This im- plies that we can only specify a plausible range of flattening ra- tios based on the uncertainties of the shadow ellipse parameters and an arbitrary tilt angle. The upper limit of the flattening ratio is given by the bifurcation point between a Maclaurin spheroid and a Jacobi ellipsoid at a/c = 1.71609, corresponding to an ellipse eccentricity of e = 0.81267 (Chandrasekhar 1967). As illustrated in Figure 6 and summarized in Table 6, the flattening ratio of 2007 UK126, and therefore its size, are relatively poorly constrained. We also do not have a density estimate for 2007 UK126. Mod- els of volatile retention (see e.g. Schaller & Brown 2007) could be used to derive a lower bulk density limit when surface ices or a thin atmosphere are present, but the featureless near-infrared spectrum (see Section 2) does not indicate the presence of any prominent volatiles (within the S/N limits of the dataset). As pointed out by Brown (2013a), the absence of volatiles and a measurable atmosphere cannot be used as an argument to derive an upper density limit, as Jeans escape is unfortunately not the only mechanism that can cause a body to lose its volatiles, even though it is the slowest. One way to estimate a lower bulk density limit would be a precise determination of the rotation period. At bifurcation, we can calculate the density of the spheroid as shown by Chan- drasekhar (1967) from 18Willmer (2006), http://mips.as.arizona.edu/~cnaw/sun. html using Bohlin & Gilliland (2004) and Fukugita et al. (1995) ρ = ω2 0.37423 π G , Article number, page 10 of 16 (3) 2131 = Bardecker2 = Schindler3 = OlsenSchindler et al.: Results from an occultation and FIR photometry of (229762) 2007 UK126 Table 6. Possible size range of the axes of a Maclaurin spheroid. Parameter a/c < 1.7161 a = b (km) c (km) dSphere,eff (km) 640 − 651 373 − 611 535 − 638 a/c ≤ 1.4870 ρ ≥ 0.73 g cm−3 P ≥ 8 h 640 − 651 432 − 611 563 − 638 Notes. The equivalent diameter of a sphere dSphere has been calculated to compare with a previous size estimate. for which densities have been estimated thanks to their satellites are Ceto (dsystem,eff = 281 ± 11 km, ρ = 0.64+0.16−0.13 g cm−3, pV = 5.6±0.6, Santos-Sanz et al. 2012) and Eris (deff = 2326±12 km, ρ = 2.52 ± 0.05 g cm−3, pV = 96+9−4%, Sicardy et al. 2011). Both objects could not be more contrary in character; their properties are at complete opposite ends of the parameter space, illustrating the difficulties at hand. Fig. 6. Flattening ratio a/c of the Maclaurin spheroid as a function of angle θ between the spheroid's rotation axis and the line of sight. With- out a density estimate, true flattening ratios up to the bifurcation point are theoretically plausible. We argue that a more realistic lower density limit is ρ = 0.73 g cm−3, which would exclude extreme axis ratios and therefore constrain the volume of the body considerably. where ω denotes the angular velocity and G = 6.67384 · 10−11 m3 kg−1 s−2 the gravitational constant. We only know that 2007 UK126 takes more than P > 8 h for a full rotation (see Section 2), which means its bulk density at bifurcation would be ρ < 0.61 g cm−3. Rotation period and density are indirectly pro- portional to the square, i.e. if the body takes twice as long for a full rotation, its lower density limit would be one quarter of this estimate. We can therefore not constrain a lower limit on bulk density for 2007 UK126 at this time. Estimated bulk densities and sizes of TNOs have been col- lected, e.g. in Ortiz et al. (2012) (see supplementary informa- tion), Brown (2013b) and Johnston (2014). We compiled a list of all TNOs whose size has been constrained by thermophysi- cal modeling and/or occultations, and bulk density has been es- timated owing to the presence of a satellite with a known orbit (see Appendix A). We did not include TNOs for which bulk den- sity estimates have been derived solely from light curves (as an assumption on body figure has to be made in these cases), or limits on bulk density have been stated mistakenly based on the absence of volatiles or a measurable atmosphere. Our list covers 19 objects with a diameter range of 157 − 2374 km. We em- phasize that this compilation can only be used for qualitative statements; it is not meaningful to derive a correlation function between size and density -- the sample is too small, could be ob- servationally biased, and uncertainties are, in general, very large. Also, the large diversity among the TNO population is not under- stood, and bodies could have undergone entirely different evo- lutions. To our knowledge, no density estimate has been derived for any detached object to date. Only few objects have been stud- ied through far-infrared observations, and 2007 UK126 is the first detached object that has been studied in detail during an occul- tation. In addition, the limits of orbital elements of the detached object population are unclear, and transitional objects between the scattered disk and the inner Oort cloud could belong to this population as well. The only two scattered disk objects (SDOs) From our list, we find that, except for two targets (both about half the size of 2007 UK126), no TNO has an estimated density ρ < 0.6 g cm−3. This corresponds to our estimated lower density limit at bifurcation based on P = 8 h. A subset of 13 objects in our list have a size deff < 800 km; the average density of this subset is ¯ρ = 0.87 g cm−3, while the median density is ρmedian = 0.73 g cm−3. Eleven of these 13 objects are smaller than deff < 400 km, so this selection is strongly biased towards objects that are significantly smaller than 2007 UK126. We feel that it is unlikely that 2007 UK126 has a bulk density below ρ = 0.73 g cm−3 given its size, since this would require a significant porosity comparable to a comet. The Rosetta mission revealed a density of ρ = 0.533 ± 0.006 g cm−3 and a porosity of 72 − 74% for the nucleus of 67P/Churyumov-Gerasimenko (Pätzold et al. 2016). These properties appear feasible for highly fractured rubble piles, but are hard to imagine for a dwarf planet candidate that is larger than the three Saturnian satellites men- tioned earlier. Arguing from a different perspective, finding an oblate spheroid with a flattening ratio close to bifurcation is very unlikely, since 2007 UK126 is certainly not composed of a strengthless fluid. Solving the following equation provided by Chandrasekhar (1967) numerically, 1 − e2(cid:17)−1 (cid:16) , (4) arcsin e − 6 e2 π G √ 1 − e2 e3 2 (cid:16) 3 − 2e2(cid:17) ρ = ω2 leads to an ellipse eccentricity of e = 0.7401, or a/c = 1.4870, assuming ρ = 0.73 g cm−3 and P = 8 h. A longer rotation period would lower the flattening ratio. We feel that a flattening ratio of a/c = 1.4870 represents an acceptable, albeit qualitative, upper limit. Table 6 summarizes the possible size range of all three spheroid axes for both the generic case (unknown density) and considering the added qualitative constraint on density. It can be seen that the occultation data is able to improve the previous size estimate of dSphere,eff = 599± 77 km for an equivalent sphere that was derived by Santos-Sanz et al. (2012) solely through ther- mophysical modeling based on Herschel/PACS data. In the next section, we refine our size estimate from the occultation further through thermophysical modeling, using the occultation data as constraints. Article number, page 11 of 16 Tilt angle 3 of principal axis c [deg]0153045607590Spheroid flattening ratio a/c11.251.51.752Best fitting ellipse axes ratio (aEl/bEl = 1.080)Confidence interval for best fitting ellipse axes ratioBifurcation point (a/c = 1.716)Limit for ; 6 0.73 g cm-3, P > 8 h (a/c 5 1.487)Possible solutionsPossible solutions with constraint ; 6 0.73 g cm-3, P > 8 hA&A proofs: manuscript no. Schindler 5. Results from thermophysical modeling To improve our geometrically obtained size estimate of 2007 UK126, we model the body's thermal emission using a ther- mophysical model (TPM) code. We conduct a parametric study to predict far-infrared fluxes for a wide range of plausible ge- ometries and physical properties, and compare them to our re- reduced Herschel/PACS measurements presented in Section 3.2. A description of the thermophysical model, its parameters and further details can be found in Müller & Lagerros (1998, 2002). The model considers the heliocentric distance of 2007 UK126 (rH = 44.944 AU) at the epoch of Herschel's observations. 2007 UK126 is considered a binary system. Since we do not know the position of the satellite relative to the primary during Herschel's observations, we need to consider that the satellite might have contributed some flux to the Herschel/PACS mea- surements. We therefore studied the following cases to constrain which combination of parameters are compatible to the occulta- tion and to the Herschel/PACS data: I The satellite did not contribute flux to the observed FIR II The satellite contributed a maximum plausible flux to the fluxes; observed FIR fluxes; Γ = {0.3, 0.7, 1, 3, 5, 10} J m−2 s−0.5 K−1; each considering (a) A thermal inertia of (b) A surface roughness with a slope of s = {0.1, 0.5, 0.9} rms; (c) A rotation period of P = {8, 11.05, 14.30, 20.25} h for three representative shadow ellipses: The best fitting ellipse, and the largest and smallest ellipse according to the estimated uncertainties of the major and minor axes. For each ellipse ge- ometry, we considered the possible tilt angle range of the body's principal axis c starting at the lowest limit constrained by bifur- cation, and then discretized in multiples of 5° up to θ = 45°, and in multiples of 10° between θ = 50° and an equator-on viewing geometry (θ = 90°). The qualitatively derived limit of a/c = 1.4870 discussed in the previous section was discarded to enable an unbiased analysis. To study the described param- eter space, 4320 flux predictions had to be made, not counting additional predictions to test, for example, an even more ex- treme thermal inertia. Figure 7 illustrates how all flux predic- tions made for P = 8 h scatter and compare with our improved Herschel/PACS flux measurements, which are plotted with their respective uncertainties. 5.1. No flux contribution from a satellite We consider a thermophysical model as plausible if it predicts far-infrared fluxes that fit to the measured Herschel/PACS fluxes with a reduced χ2 ≤ 1.7. Table 7 summarizes the parameter space of models that fulfilled this criteria, assuming P = 8 h. We consider the very high roughness case (s = 0.9 rms) as rather unrealistic, as such high roughness values are typi- cally not used in thermophysical simulations. For near-Earth ob- jects, typical roughness values in simulations are s = 0.2 rms, while values above s > 0.5 rms do not find application. Also, the thermal inertia that is required in this case to produce plau- sible models is rather high. Recent work by Lellouch et al. (2013) finds evidence that thermal inertia decreases with in- creasing heliocentric distance, implying a trend towards more and more porous surfaces. According to their study, typical val- ues at heliocentric distances of rH > 41 AU are expected to Article number, page 12 of 16 Fig. 7. Far-infrared fluxes of 2007 UK126 with estimated 1σ uncertain- ties as measured by Herschel/PACS in three bands (see Table 4), in comparison to far-infrared fluxes calculated from the modeled thermal emission of all cases covered by our parametric study, discarding a flux contribution from the satellite and assuming P = 8 h. For clarity, flux predictions have been plotted with a slight offset in wavelength based on the applied surface roughness (s = {0.1, 0.5, 0.9} rms). In general, predicted fluxes increase with decreasing thermal inertia. The dashed line represents the best model obtained for an intermediate roughness (s = 0.5 rms) with Γ = 3 J m−2 s−0.5K−1, θ = 70° and a/c = 1.11, resulting in Deff = 618 km. See text for details. be around Γ ≈ 2 J m−2 s−0.5 K−1, while values do not exceed Γ ≈ 6 J m−2 s−0.5 K−1 that far out in the solar system. This was also the reason why we limited our parameter study to thermal inertia values of Γ ≤ 10 J m−2 s−0.5 K−1. For a rotation period of P = 8 h, all plausible solutions cover a range of the principal axis tilt of θ = 45 − 90°. The estimated effective diameter (of a sphere with equal volume) is deff = 599 − 629 km. When assuming a rotation period of P > 8 h, it is more difficult to find models that are in agree- ment with the Herschel/PACS measurements. At P = 20.25 h, we obtain deff = 605 − 625 km at principal axis tilt angles of θ = 60 − 90°; solutions can then only be found for low and intermediate surface roughness levels. To explain the size mea- surement from the occultation under the assumption of a surface with high roughness and a long rotation period, a thermal inertia would be required that is too high to be physically plausible. 5.2. Some flux contribution from a satellite Given that the broad HST F606W band pass is reasonably close to the standard V band pass from the perspective of solar sys- tem studies, we translate the satellite's magnitude difference of ∆m = 3.79 ± 0.24 mag into an absolute magnitude of HSat = 7.48 ± 0.29 mag. This implies that the satellite emits about 3% of the combined flux. Assuming an identical albedo and using the equation from Harris (1998), − 1 V , 2 5 1329 p dSat = 10− HV,Sat (5) we obtain a diameter estimate of dSat = 112.2 ± 75.5 km for the satellite. For this scenario, we estimated with our TPM that the satellite would contribute a flux of 0.3 / 0.4 / 0.3 mJy in the Herschel/PACS bands (70 / 100 / 160 µm). 100Wavelength [µm]0510152025Best solutionPsid = 8 hrms=0.10.50.920050high / intermediate / lowthermal inertiaFlux [mJy]Schindler et al.: Results from an occultation and FIR photometry of (229762) 2007 UK126 Table 7. Results from a parametric study with a TPM code, assuming a rotation period of P = 8 h and no flux contribution from the satellite. Parameter very low roughness intermediate roughness very high roughness θ (degrees) Γ (J m−2 s−0.5 K−1) a = b (km) c (km) a/c dSphere,eff (km) reduced χ2 (0.1 rms) 45 − 90 0.7 − 10 640 − 646 524 − 598 1.08 − 1.22 599 − 629 1.40 − 1.63 (0.5 rms) 60 − 70 3 − 10 640 − 646 565 − 591 1.09 − 1.13 614 − 627 1.42 − 1.70 (0.9 rms) 60 − 90 3 − 10 640 565 − 585 1.09 − 1.13 614 − 621 1.46 − 1.66 Using the same albedo for the satellite and for the primary is an assumption that is often made in literature, but we do not know how realistic it is. For example, the New Horizons mission proved that Charon is significantly darker than Pluto (Buratti et al. 2016). It therefore appears more meaningful to study a worst case scenario for the satellite: We assume that its albedo is lower than for the primary (pV,Sat = 5%) and that its brightness is at the upper uncertainty limit of the available mea- surements (HV,Sat = 7.19 mag), implying an equivalent diam- eter of dSat = 217 km. Using our TPM, we estimate that this type of satellite produces 1.6 / 1.8 / 1.3 mJy (corresponding to ≈ 14 − 16% flux) in the Herschel/PACS bands if we use typ- ical simulation parameters: An equator-on viewing geometry, a P = 8 h rotation period (assuming a tidally locked motion), a low thermal inertia of Γ = 1 J m−2 s−0.5 K−1, and an intermedi- ate level of surface roughness of s = 0.5 rms. Acting as a com- parison, the NEATM model (Harris 1998) produces consistent flux estimates of 1.6 / 1.9 / 1.5 mJy for such a satellite using a beaming parameter of η = 1.2. Subtracting the satellite's worst- case flux estimates from the measured Herschel/PACS fluxes provided earlier in Table 4 gives us a minimum flux estimate of 9.9 / 11.5 / 6.7 mJy for 2007 UK126. Thermophysical modeling considering a satellite flux contri- bution did not lead to meaningful results at any studied rotation period: Although predicted fluxes typically match measurements in the 70 µm and 100 µm bands well, they poorly fit the measure- ments in the 160 µm band. This is indicated by a much degraded reduced χ2 ≥ 3.32. Table 8 lists the range of possible properties of 2007 UK126 according to models with a goodness of fit in the range of 3.32 ≤ χ2 ≤ 4.0. Again, a very high roughness and a very high thermal inertia are rather unrealistic. Given the very poor fit of our modeled fluxes to Her- schel/PACS measurements, it is very likely that the flux contri- bution from the satellite is much less than the derived worst-case contribution, indicating that the discussion in Section 5.1 is a realistic approximation. 5.3. Implications The TPM simulations enable us to constrain the parameter space marked in Figure 6 significantly. Figure 8 illustrates the sur- face temperature distribution on 2007 UK126, as predicted by the TPM for one of the best model fits: A viewing geometry of θ = 70° of an oblate sphere with an effective diameter of deff = 618 km, a flattening ratio of a/c = 1.11, a thermal in- ertia of Γ = 3 J m−2 s−0.5 K−1, an intermediate level of surface roughness of s = 0.5 rms, and a rotation period of P = 8 h. This fit results in a reduced χ2 = 1.43 and predicted fluxes of 10.98 / 13.28 / 10.29 mJy in Herschel's 70 / 100 / 160 µm bands. Fig. 8. Surface temperature distribution on 2007 UK126, as predicted by the TPM for the best model with an intermediate roughness (see Figure 7 and text for details). The TPM takes the epoch of the Herschel observation into account. In our range of plausible model fits, we find maximum sur- face temperatures of about ≈ 50 − 55 K in the subsolar point. Since 2007 UK126 is still on its way to perihelion in 2046, it spends a large fraction of its orbit at comparable or even higher surface temperatures. From two perspectives, it is unlikely that 2007 UK126 has retained volatile ices (CH4, N2, and CO): Be- cause of its estimated temperature levels and its small size. At the given surface temperatures, all volatiles would be lost solely based on Jeans escape, as indicated by a greatly sim- plified model by Schaller & Brown (2007). However, as dis- cussed by Stern & Trafton (2008), an atmosphere on a body as small as 2007 UK126 would be governed by hydrodynamic es- cape (owing to the body's low gravity, even considering mod- erate to high densities), or by a combination of Jeans and hy- drodynamic escape. It is unlikely that Jeans escape alone, the slowest loss mechanism, is depleting the atmosphere. Given the eccentricity of the orbit of 2007 UK126, the proportion of each escape mechanism could vary over time, and seasonal freeze out and sublimation of volatile ices could play a role. In addi- tion to classic escape mechanisms, impacts that almost certainly occurred throughout the lifetime of the solar system would ac- celerate the escape of an atmosphere. The lack of volatiles is supported by available near-infrared spectra (see Section 2) that could not identify any ice features (within the S/N limits of the dataset). However, small amounts of involatile amorphous wa- ter ice, below the detection limit of the available spectra, cannot be excluded since they could have been preserved at the calcu- lated surface temperatures. Since no volatiles, and hence no at- mosphere are expected on 2007 UK126, the approach of fitting square-well profiles to the occultation light curve data as dis- cussed in Section 4.1 is reconfirmed. Article number, page 13 of 16 Table 8. Results from a parametric study with a TPM code, considering a maximum plausible flux contribution from the satellite and assuming a rotation period of P = 8 h for the primary body. A&A proofs: manuscript no. Schindler Parameter very low roughness intermediate roughness very high roughness θ (degrees) Γ (J m−2 s−0.5 K−1) a/c dSphere,eff (km) reduced χ2 (0.1 rms) 50 − 70 3 − 10 1.09 − 1.18 606 − 627 3.43 − 3.87 (0.5 rms) 70 − 90 5 − 10 1.07 − 1.11 618 − 637 3.32 − 4.00 10 (0.9 rms) 80 − 90 1.09 − 1.10 3.47 − 3.56 621 6. Conclusions For the first time, it was possible to measure the size of a de- tached object of the TNO population directly during a stellar oc- cultation at three locations. Our findings from the occultation measurements are: at any given rotation period that would be compatible with the occultation shadow size and Herschel/PACS measure- ments, so it is very likely that any satellite flux contributions would be much less. ≈ 50 − 55 K in the subsolar point. 4. Our models result in maximum surface temperatures of 1. The shadow ellipse that fits best to the times of disappearance and reappearance on each chord has a major axis of aEl = 645.80±5.68 km and a minor axis of bEl = 597.81±12.74 km (aEl/bEl = 1.080 ± 0.025). 2. The estimated projected area, combined with previous pho- tometric measurements, suggests an albedo estimate of pV = 15.0± 1.6%. This estimate is consistent with and more accu- rate than an earlier estimate that was based on a size derived solely from a thermophysical model. 3. Owing to the very low light curve amplitude and rotation pe- riod reported in literature and the size estimated from the oc- cultation, we assume by analogy to other solar system bodies that 2007 UK126 resembles an oblate spheroid. 4. Given the shadow ellipse geometry and the non-zero light curve amplitude, 2007 UK126 was not seen pole-on. 5. We cannot constrain a bulk density at this time. Purely by geometric means, we can constrain the effective diameter of 2007 UK126 to deff = 535 − 638 km. We argue qualitatively that the spheroid's flattening ratio is likely a/c ≤ 1.4870, which would constrain the effective diameter range slightly more to deff = 563 − 638 km. 6. No secondary occultation has been detected, so the satellite could not be reobserved (although chances would have been extremely small). This implies that it was either behind or in front of the primary, not in the primary's shadow path, or so small that its shadow moved in between the observers' locations. With sufficient distance to the primary at the time of the occultation, the shadow of the secondary may not have even hit the Earth's surface. Modeling the thermal emission of 2007 UK126, using geo- metrical constraints obtained from the occultation, and compar- ing modeled fluxes with improved Herschel/PACS far-infrared flux measurements, we conclude that: 1. The effective diameter of 2007 UK126 is likely of the order of deff = 599 − 629 km. The flattening ratio is of the order of a/c = 1.08 − 1.22. 2. The body was likely seen near an equator-on viewing geom- etry (θ = 45 − 90°). 3. Assuming an albedo of 5% for the satellite and a worst-case photometric error, its effective diameter would be 217 km, from which we derive the worst-case satellite flux contribu- tion in the Herschel/PACS bands. Considering a flux contri- bution from the satellite, it was not possible to derive models Article number, page 14 of 16 Combining our analyses of the occultation and the thermal emission, we conclude: 1. While a true flattening ratio close to bifurcation cannot for- mally be ruled out from the occultation data, given the un- known density, high flattening ratios can be ruled out thanks to our thermal simulations. This constrains the size of the body considerably. 2. Owing to its estimated surface temperature levels and its small size, it is unlikely that 2007 UK126 has retained volatile ices (CH4, N2, and CO), which is also supported by its fea- tureless near-infrared spectra reported in literature. This re- confirms fitting square-well profiles to the occultation light curve, as no atmosphere is expected. The estimated size supports the status of 2007 UK126 as a dwarf planet candidate. While faster sampling of an occultation would help to decrease the uncertainties of the shadow ellipse fit, it would not help to significantly constrain the tilt angle θ of the rotation axis, which is the dominating source of uncertainty for a geometrically derived size estimate. Without a density es- timate, only the observation of multiple occultations, extensive light curve measurements, and work on a shape model could eventually provide better geometrical constraints on the body's true size. Thermophysical modeling allows to overcome this is- sue. Future occultations of 2007 UK126 or acquisition of addi- tional high-resolution image data will hopefully provide oppor- tunities to reobserve its satellite. Knowledge of the satellite's size and orbit is crucial to constrain the system mass and bulk density, which would immediately constrain plausible shapes. While size and albedo estimates solely derived from far-infrared measure- ments still have considerable uncertainties, the combined anal- ysis of occultation observations and far-infrared measurements is a powerful tool to thoroughly characterize bodies in our solar system. Acknowledgements. For their efforts on long-term occultation predictions and updates, we thank B. Sicardy (LESIA, Observatoire de Paris), J. Desmars (IM- CEE, Paris), N. Morales, P. Santos-Sanz, R. Duffard (Instituto de Astrofisica de Andalucia-CSIC) and the following members from the Rio Group: R. Vieira- Martins, M. Assafin, A. Dias-Oliveira, B. Morgado, G. Benedetti-Rossi and A. Ramos Gomes Jr.. We would like to thank M. Hümmer for his work on creat- ing a video sequence via ray tracing illustrating the occultation based on the re- constructed geometry of 2007 UK126 (see supplementary material to this article). K.S. has been supported financially by a one-year fellowship from the German Academic Exchange Service (DAAD) during his stay at NASA Ames Research Schindler et al.: Results from an occultation and FIR photometry of (229762) 2007 UK126 Center, CA as part of his PhD research. K.S. would also like to thank the Uni- versities Space Research Association (USRA) for their support throughout his thesis. T.M., C.K. and J.L.O. have received funding from the European Union's Horizon 2020 Research and Innovation Programme, under Grant Agreement No. 687378. C.K. has been supported by the PECS grant # 4000109997/13/NL/KML of the Hungarian Space Office and the European Space Agency, the K-104607 grant of the Hungarian Research Fund (OTKA), and the GINOP-2.3.2-15-2016- 00003 grant of the National Research, Development and Innovation Office of Hungary (NKFIH). J.L.O. acknowledges support from Proyecto de Excelencia J.A. 2012-FQM1776. The Astronomical Telescope of the University Stuttgart (ATUS) has been funded in parts by the German Aerospace Center (DLR). K.S. and J.W. would like to thank all involved parties at Sierra Remote Obser- vatories (SRO) for providing and maintaining the infrastructure that allows us to remotely operate ATUS. L. Van Vleet was manually monitoring and controlling the observatory roof that night which allowed us to conduct this observation de- spite of critical humidity levels. K.S. thanks T. George and R.L. Anderson, the developers of R-OTE, for vivid discussions on statistical methods and camera- and VTI-specific time offset corrections. This research has made use of NASA's Astrophysics Data System. References Balog, Z., Müller, T., Nielbock, M., et al. 2014, Exp. Astron., 37, 129 Barucci, M. A., Alvarez-Candal, A., Merlin, F., et al. 2011, Icarus, 214, 297 Barucci, M. A., Belskaya, I. N., Fulchignoni, M., & Birlan, M. 2005, AJ, 130, 1291 Bohlin, R. C. & Gilliland, R. L. 2004, AJ, 127, 3508 Bosh, A. S., Person, M. J., Zuluaga, C. A., et al. 2016, Icarus (submitted) Braga-Ribas, F., Sicardy, B., Ortiz, J. L., et al. 2013, ApJ, 773, 26 Braga-Ribas, F., Sicardy, B., Ortiz, J. L., et al. 2014a, Nature, 508, 72 Braga-Ribas, F., Vieira-Martins, R., Assafin, M., et al. 2014b, in Revista Mexi- cana de Astronomia y Astrofisica Conference Series, Vol. 44, 3 -- 3 Brown, M. E. 2013a, ApJL, 767, L7 Brown, M. E. 2013b, ApJL, 778, L34 Buratti, B. J., Hofgartner, J. D., Hicks, M. D., et al. 2016, ArXiv e-prints [arXiv:1604.06129] Camargo, J. I. B., Vieira-Martins, R., Assafin, M., et al. 2014, A&A, 561, A37 Chandrasekhar, S. 1967, Comm. Pure Appl. Math., 20, 251 Collins, K. & Kielkopf, J. 2013 [ascl:1309.001] Collins, K. A. 2015, PhD thesis, University of Louisville Collins, K. A., Kielkopf, J. F., & Stassun, K. G. 2016, ArXiv e-prints Descamps, P., Marchis, F., Berthier, J., et al. 2011, Icarus, 211, 1022 Duffard, R., Ortiz, J. L., Thirouin, A., Santos-Sanz, P., & Morales, N. 2009, [arXiv:1601.02622] A&A, 505, 1283 Elliot, J. L., Person, M. J., Zuluaga, C. A., et al. 2010, Nature, 465, 897 Engineering Industries Association (EIA). 1957, Standard RS 170, Electrical Performance Standards - Monochrome Television Studio Facilities Fornasier, S., Barucci, M. A., de Bergh, C., et al. 2009, A&A, 508, 457 Fornasier, S., Lellouch, E., Müller, T., et al. 2013, A&A, 555, A15 Fukugita, M., Shimasaku, K., & Ichikawa, T. 1995, PASP, 107, 945 Fulchignoni, M., Belskaya, I., Barucci, M. A., de Sanctis, M. C., & Doressoundi- ram, A. 2008, in The Solar System Beyond Neptune, ed. M. A. Barucci, H. Boehnhardt, D. P. Cruikshank, A. Morbidelli, & R. Dotson (The University of Arizona Press), 181 -- 192 Gaia Collaboration, Brown, A. G. A., Vallenari, A., et al. 2016, ArXiv e-prints [arXiv:1609.04172] George, T. 2014, Analysis of Camera Delay Corrections for use in R-OTE and Occular Occultation Timing Extraction Programs, White Paper George, T. & Anderson, R. L. 2013, in 2013 IOTA Annual Meeting, Toronto, Canada 2068 Gladman, B., Marsden, B. G., & Vanlaerhoven, C. 2008, in The Solar System Be- yond Neptune, ed. M. A. Barucci, H. Boehnhardt, D. P. Cruikshank, A. Mor- bidelli, & R. Dotson (The University of Arizona Press), 43 -- 57 Harris, A. W. 1994, Icarus, 107, 209 Harris, A. W. 1998, Icarus, 131, 291 Herald, D. 2016, Occult v4 Help Jacobson, R. A., Campbell, J. K., Taylor, A. H., & Synnott, S. P. 1992, AJ, 103, Johnston, W. R. 2014, TNO and Centaur Diameters, Albedo, and Densities V2.0, EAR-A-COMPIL-5-TNOCENALB-V2.0, NASA Planetary Data System Kiss, C., Müller, T., Vilenius, E., et al. 2014, Exp. Astron., 37, 161 Lacerda, P., Fornasier, S., Lellouch, E., et al. 2014, ApJL, 793, L2 Lellouch, E., Santos-Sanz, P., Lacerda, P., et al. 2013, A&A, 557, A60 Lindegren, L., Lammers, U., Bastian, U., et al. 2016, ArXiv e-prints [arXiv:1609.04303] Millis, R. L. & Dunham, D. W. 1989, in Asteroid II, ed. R. P. Binzel, T. Gehrels, & M. S. Matthews (University of Arizona Press), 148 -- 170 Millis, R. L. & Elliot, J. L. 1979, in Asteroids, ed. T. Gehrels (University of Arizona Press), 98 -- 118 Mommert, M., Harris, A. W., Kiss, C., et al. 2012, A&A, 541, A93 Müller, T. G. & Lagerros, J. S. V. 1998, A&A, 338, 340 Müller, T. G. & Lagerros, J. S. V. 2002, A&A, 381, 324 Müller, T. G., Lellouch, E., Böhnhardt, H., et al. 2009, EM&P, 105, 209 Noll, K. S., Grundy, W. M., Benecchi, S. D., Levison, H. F., & Barker, E. A. 2009, in AAS/Division for Planetary Sciences Meeting Abstracts, Vol. 41, 47.07 Ortiz, J. L., Sicardy, B., Braga-Ribas, F., et al. 2012, Nature, 491, 566 Ortiz, J. L., Sicardy, B., Braga-Ribas, F., et al. 2014, in European Planetary Sci- ence Congress 2014 Abstracts, Vol. 9, EPSC2014 -- 525 Pätzold, M., Andert, T., Hahn, M., et al. 2016, Nature, 530, 63 Pavlov, H. 2014, in Eighth Trans-Tasman Symposium on Occultations (TTSO8), Melbourne, Australia Perna, D., Barucci, M. A., Fornasier, S., et al. 2010, A&A, 510, A53 Person, M. J., Dunham, E. W., Bosh, A. S., et al. 2013, AJ, 146, 83 Pfüller, E., Wolf, J., Hall, H., & Röser, H.-P. 2012, Proc. SPIE, 8444, 844413 Pilbratt, G. L., Riedinger, J. R., Passvogel, T., et al. 2010, A&A, 518, L1 Poglitsch, A., Waelkens, C., Geis, N., et al. 2010, A&A, 518, L2 Roatsch, T., Jaumann, R., Stephan, K., & Thomas, P. 2009, in Saturn from Cassini-Huygens, ed. M. Dougherty, L. Esposito, & S. Krimigis (Springer Netherlands), 763 -- 781 Roques, F., Moncuquet, M., & Sicardy, B. 1987, AJ, 93, 1549 Santos-Sanz, P., Lellouch, E., Fornasier, S., et al. 2012, A&A, 541, A92 Schaller, E. L. & Brown, M. E. 2007, ApJL, 659, L61 Schneider, C. A., Rasband, W. S., & Eliceiri, K. W. 2012, Nat. Methods, 671 Sheppard, S. S. & Jewitt, D. C. 2002, AJ, 124, 1757 Sicardy, B., Ortiz, J. L., Assafin, M., et al. 2011, Nature, 478, 493 Souza, S. P., Babcock, B. A., Pasachoff, J. M., et al. 2006, PASP, 118, 1550 Spectrum Instruments, Inc. 2014, Intelligent Reference/TM-4™ GPS Time & Frequency System - User Manual Spencer, J. R., Stansberry, J. A., Grundy, W. M., & Noll, K. S. 2006, in AAS/Division for Planetary Sciences Meeting Abstracts, Vol. 38, 546 Stansberry, J., Grundy, W., Brown, M., et al. 2008, in The Solar System Beyond Neptune, ed. M. A. Barucci, H. Boehnhardt, D. P. Cruikshank, A. Morbidelli, & R. Dotson (The University of Arizona Press), 161 -- 179 Stansberry, J., Grundy, W., Mueller, M., et al. 2012, Icarus, 219, 676 Stern, S. A., Bagenal, F., Ennico, K., et al. 2015, Science, 350 Stern, S. A. & Trafton, L. M. 2008, in The Solar System Beyond Neptune, ed. M. A. Barucci, H. Boehnhardt, D. P. Cruikshank, A. Morbidelli, & R. Dotson (The University of Arizona Press), 365 -- 380 Szpak, Z. L., Chojnacki, W., & van den Hengel, A. 2015, J. Math. Imaging Vi- sion, 52, 173 Thirouin, A., Noll, K. S., Ortiz, J. L., & Morales, N. 2014, A&A, 569, A3 Timerson, B., Brooks, J., Conard, S., et al. 2013, Planet. Space Sci., 87, 78 van Belle, G. T. 1999, PASP, 111, pp. 1515 Vilenius, E., Kiss, C., Mommert, M., et al. 2012, A&A, 541, A94 Vilenius, E., Kiss, C., Müller, T., et al. 2014, A&A, 564, A35 Wasserman, L. H., Millis, R. L., Franz, O. G., et al. 1979, AJ, 84, 259 Wolf, J., Wiedemann, M., Pfüller, E., et al. 2014, Proc. SPIE, 9145, 91450W Wright, E. L., Eisenhardt, P. R. M., Mainzer, A. K., et al. 2010, AJ, 140, 1868 Appendix A: Density estimates Table A.1 summarizes all TNOs for which a density estimate could be derived to this date. The size and mass of Pluto and its satellite Charon were measured during the New Horizons flyby. All remaining objects are binary systems, which allowed an esti- mate of the respective system mass based on the satellite's orbit. Only the size of Eris and Quaoar have been probed through ob- servations of stellar occultations so far. All other size estimates were derived from thermophysical modeling of Spitzer and/or Herschel data. Article number, page 15 of 16 Table A.1. Summary of TNOs with estimated bulk densities. A&A proofs: manuscript no. Schindler Effective Diameter (km) Bulk Density (g cm−3) Target (134340) Pluto (136199) Eris (134340) I Charon (50000) Quaoar (90482) Orcusa (120347) Salaciaa (174567) Vardaa (55637) 2002 UX25 a (47171) 1999 TC36 (79360) Sila-Nunama a (148780) Altjiraa 2001 QC298 a (26308) 1998 SM165 a (65489) Cetoa (275809) 2001 QY297 a 2001 XR254 a (88611) Teharonhiawakoa (66652) Borasisia (42355) Typhona 2374 ± 8 2326 ± 12 1212 ± 6 1110 ± 5 958.4 ± 22.9 901 ± 45 792+91−84 692 ± 23 393.1+25.2−26.8 343 ± 42 331+51−187 303+27−30 279.8+29.7−28.6 281 ± 11 229+22−108 221+41−71 220+41−44 163+32−66 157 ± 34 1.860 ± 0.013 2.52 ± 0.05 1.702 ± 0.021 1.99 ± 0.46 1.53+0.15−0.13 1.29+0.29−0.23 1.27+0.41−0.44 0.82 ± 0.11 0.64+0.15−0.11 0.73 ± 0.28 0.30+0.50−0.14 1.14+0.34−0.30 0.51+0.29−0.14 0.64+0.16−0.13 0.92+1.30−0.27 1.00+0.96−0.56 0.60+0.36−0.33 2.1+2.60−1.2 0.60+0.72−0.29 Source Stern et al. (2015) Sicardy et al. (2011) Stern et al. (2015) Braga-Ribas et al. (2013) Fornasier et al. (2013) Fornasier et al. (2013) Vilenius et al. (2014) Brown (2013b) Mommert et al. (2012) Vilenius et al. (2012) Vilenius et al. (2014) Vilenius et al. (2014) Stansberry et al. (2008), Spencer et al. (2006) Santos-Sanz et al. (2012) Vilenius et al. (2014) Vilenius et al. (2014) Vilenius et al. (2014) Vilenius et al. (2014) Stansberry et al. (2012) Notes. (a) Binary system properties; primary not individually measured through occultation or spacecraft encounter. Appendix B: Error propagation Applying the laws of error propagation, the error of the albedo estimate can be calculated as follows: ∆p = (cid:32) dp (cid:32) dp dHSun + daEl (cid:33)2 (cid:32) dp (cid:32) dp + dHTNO ∆bEl dbEl ∆HSun (cid:33)2 ∆aEl + (cid:33)2 ∆HTNO (cid:33)21/2 . (B.1) In our calculations we assumed ∆HS un = 0. The remaining dif- ferentials are dp dHTNO = 4 · (1 AU)2 aElbEl 100.4HSun ln 10 · 10−0.4HTNO (B.2) (B.3) (B.4) dp daEl dp dbEl = −4 · 100.4(HSun−HTNO) · (1 AU)2 a2 ElbEl = −4 · 100.4(HSun−HTNO) · (1 AU)2 aElb2 El Article number, page 16 of 16
1403.0403	2	1403	2015-03-04T11:01:48	Topological Classifications and Bifurcations of Periodic Orbits in the Potential Field of Highly Irregular-shaped Celestial Bodies	[ "astro-ph.EP" ]	This paper studies the distribution of characteristic multipliers, the structure of submanifolds, the phase diagram, bifurcations and chaotic motions in the potential field of rotating highly irregular-shaped celestial bodies (hereafter called irregular bodies). The topological structure of the submanifolds for the orbits in the potential field of an irregular body is shown to be classified into 34 different cases, including 6 ordinary cases, 3 collisional cases, 3 degenerate real saddle cases, 7 periodic cases, 7 period-doubling cases, 1 periodic and collisional case, 1 periodic and degenerate real saddle case, 1 period-doubling and collisional case, 1 period-doubling and degenerate real saddle case, and 4 periodic and period-doubling cases. The different distribution of the characteristic multipliers has been shown to fix the structure of the submanifolds, the type of orbits, the dynamical behaviour and the phase diagram of the motion. Classifications and properties for each case are presented. Moreover, tangent bifurcations, period-doubling bifurcations, Neimark-Sacker bifurcations and the real saddle bifurcations of periodic orbits in the potential field of an irregular body are discovered. Submanifolds appear to be Mobius strips and Klein bottles when the period-doubling bifurcation occurs.	astro-ph.EP	astro-ph	Topological Classifications and Bifurcations of Periodic Orbits in the Potential Field of Highly Irregular-shaped Celestial Bodies Yu Jiang1,2, Yang Yu2, Hexi Baoyin2 1. State Key Laboratory of Astronautic Dynamics, Xi’an Satellite Control Center, Xi’an 710043, China 2. School of Aerospace Engineering, Tsinghua University, Beijing 100084, China Y. Jiang () e-mail: [email protected] (corresponding author) Abstract. This paper studies the distribution of characteristic multipliers, the structure of submanifolds, the phase diagram, bifurcations and chaotic motions in the potential field of rotating highly irregular-shaped celestial bodies (hereafter called irregular bodies). The topological structure of the submanifolds for the orbits in the potential field of an irregular body is shown to be classified into 34 different cases, including 6 ordinary cases, 3 collisional cases, 3 degenerate real saddle cases, 7 periodic cases, 7 period-doubling cases, 1 periodic and collisional case, 1 periodic and degenerate real saddle case, 1 period-doubling and collisional case, 1 period-doubling and degenerate real saddle case, and 4 periodic and period-doubling cases. The different distribution of the characteristic multipliers has been shown to fix the structure of the submanifolds, the type of orbits, the dynamical behaviour and the phase diagram of the motion. Classifications and properties for each case are presented. Moreover, tangent bifurcations, period-doubling bifurcations, Neimark-Sacker bifurcations and the real saddle bifurcations of periodic orbits in the potential field of an irregular body are discovered. Submanifolds appear to be Mobius strips and Klein bottles when the period-doubling bifurcation occurs. 1 Keywords: Irregular-shaped Celestial Bodies; Periodic Orbits; Stability; Topological Classification; Bifurcations 1. Introduction The discoveries of large size ratio binary [1, 2] or triple asteroids [3-5] and space missions [6, 7] to minor bodies in the solar system have made the study of nonlinearly dynamical behaviour in the potential field of rotating highly irregular-shaped celestial bodies important [8-10]. This paper aims to discuss the nonlinearly dynamical behaviour of a massless particle in the potential field of irregular bodies, which can be asteroids, comets, or satellites of planets. The theory can be applied to the study of the dynamical behaviour of minor bodies, including a spacecraft around an irregular body and the large size ratio binary or triple asteroids, such as the binary asteroids 1862 Apollo [11], (162000) 1990OS [12], and 41 Daphne [13] and the triple asteroids 87 Sylvia [3] and 216 Kleopatra [5, 14]. Periodic orbits in the potential field of rotating highly irregular-shaped celestial bodies exist and can be found by analytic continuation with respect to certain parameters [9][15-18]. Periodic orbits can be classified into different families using the position characteristics [15], the geometrical appearances [19], and the topological characteristics [18]. Related to the concrete case of finding periodic orbits around rotating highly irregular-shaped celestial bodies, Scheeres et al. [15] calculated periodic orbits about asteroid 4769 Castalia using Poincaré maps and Newton-Raphson iteration and found three main families of periodic orbits close to 2 asteroid Castalia: quasi-equatorial direct, quasi-equatorial retrograde, and non-equatorial. Yu and Baoyin [19] developed a hierarchical searching method to compute periodic orbits around irregular bodies and classified periodic orbits near asteroid 216 Kleopatra into 29 families using the geometrical appearances and an advanced numerical method. Jiang et al. [18] classified periodic orbits near equilibrium points in the potential field of irregular bodies into several families on the basis of topological characteristics. The study of motion stability in the vicinity of minor irregular bodies includes the stabilities of periodic orbits, quasi-periodic orbits, and equilibrium points. Yu and Baoyin [20] discussed the stabilities of periodic orbits in the potential field of asteroid 216 Kleopatra. The stabilities of equilibrium points determine the motion stabilization near equilibrium points [18][21]. Elipe and Lara [21] modelled the asteroid 433 Eros as a finite straight segment. They found 4 equilibrium points around the segment and discussed their linear stabilities. Mondelo et al. [22] studied asteroid 4 Vesta and found 4 equilibrium points and determined their coordinates and stabilities, showing that two of them are stable, while the other two are unstable. Jiang et al. [18] established the theory about motions near equilibrium points in the potential field of a general rotating irregular body, which includes the linearized motion equation and characteristic equation near equilibrium points, two conditions of stability of the equilibrium points, and discovered that the stabilities of the equilibrium points are determined by the structure of submanifolds and subspaces near the equilibrium points. 3 Furthermore, bifurcations of motion will occur when the parameters are equal to certain special values. Riaguas et al. [23] discussed bifurcations when considering periodic orbits with parameter variations in the gravity field of a massive straight segment. The motions near 1:1 resonance in the gravity field of a massive straight segment with parameter variations can lead to bifurcations [21]; however, the resonances caused by the 1:1 ratio of the rotational period create equilibrium points [18]. In this paper, the nonlinearly dynamical behaviour of a massless particle in the potential field of an irregular body is studied, including the topological classifications and stabilities of periodic orbits and bifurcations of motion. It is shown that the topological structure of submanifolds for the orbits can be classified into 34 different cases. The different distribution of characteristic multipliers fixes the structure of submanifolds, the type of orbits, the dynamical behaviour and the phase diagram of the motion. The classifications and properties for each case are presented. Additionally, the theory reveals the mechanism of bifurcations for periodic orbits in the potential field of an irregular body. Period-doubling bifurcations, tangent bifurcations, Neimark-Sacker bifurcations and real saddle bifurcations of periodic orbits are discovered. Motions near bifurcation with parametric variation are sensitive to initial conditions. Additionally, it is found that submanifolds appear to be Mobius strips and Klein bottles when the period-doubling bifurcation occurs. The theory developed here is applied to the asteroids 216 Kleopatra and 6489 Golevka and the comet 1P/Halley to analyse the dynamical behaviour around these 4 irregular celestial bodies. For the asteroid 6489 Golevka, two periodic orbits, which have entirely similar shapes and belong to different cases, are presented as examples. This result implies that periodic orbits that belong to the different cases might have entirely similar shapes; the essential characteristic of periodic orbits is the topological type rather than shape. The number of different shapes is infinite, while the number of different topological types is finite. For the comet 1P/Halley, two periodic orbits are presented. These two periodic orbits belong to different periodic cases: one belongs to Case P2, the stable periodic case, while the other belongs to Case P4, the unstable periodic case. Dynamical behaviours and bifurcations around the large size ratio triple asteroid 216 Kleopatra are studied. All four types of bifurcations forecasted herein by the theory are discovered in the potential field of the asteroid 216 Kleopatra. Additionally, the attraction and exclusion effect of critical points in the vicinity of critical points around the asteroid 216 Kleopatra is found. 2. Topological Structures of Periodic Orbits In this section, motion equations, characteristic multipliers of the orbit, and the topological structure of submanifolds will be discussed. We will show that different distributions of characteristic multipliers fix the structure of submanifolds, the type of orbits, the dynamical behaviour, and the phase diagram of the motion. 2.1 Motion Equations Let be the body-fixed vector from the celestial body’s centre of mass to the particle, ω be the rotational angular velocity of the body relative to the inertial space, 5 rand be the gravitational potential of the body, which can be calculated by the polyhedral model [24, 25]) using radar observation data [26, 27]. The efficient potential can be defined as [15] . (1) The zero-velocity manifold for the particle satisfies [15] and the forbidden region satisfies , while the allowable region satisfies , (2) . Denoting the generalised momentum as , and the generalised coordinate as , the Lagrangian and Hamilton functions can then be given by and , (3) . (4) Using the efficient potential, the equation of motion can be written as Define . (5) . (6) Then, the dynamical equation can be expressed in the symplectic form , (7) where and are matrices, and is the gradient of 6 UrVr12VUrωrωrrVHrVHrVHrprωrqr122LUVppqqqqωqq2HUppqpq20Vrrω×rrTzpqH0IzFzzI0I033THHHzpq. Defining , where is a symplectic matrix and is the Hamiltonian vector field on the symplectic manifold, the dynamical equation in the symplectic form can be rewritten as 2.2 Characteristic Multipliers . (8) Let the orbit and be the solution of Eq. (8). Defining , the set generated by and is a one-parameter diffeomorphic group, and is a flow of the dynamical system of the particle in the potential field of a rotating celestial body. On the symplectic manifold , the orbit is denoted as . Here is a differential 2-form. Denote as the set of periodic orbits with period . For any periodic orbit , consider the matrix , which is a matrix. The state transition matrix is [19] . (9) The monodromy matrix of the periodic orbit is . (10) Eigenvalues of the monodromy matrix are characteristic multipliers of the periodic orbit, which are fixed by the periodic orbit . The state transition matrix is a symplectic matrix, which means that if is an eigenvalue of the monodromy matrix, then , , and are also eigenvalues of the state transition matrix; namely, all eigenvalues are likely to have the form 1, , , 7 Hz0IJI0JHJz0HJzz0,ttzfz000,fzz:,ttfzfztf1()tftf,S0,ttzfzppSTTppST:fzfz660ttpdfzppSMTp111sgnR,0,1e, and , where . Additionally, the periodic orbit has a characteristic multiplier equal to 1, and the multiple-number of the characteristic multipliers 1 is even. The Krein collision means that there are two pairs of eigenvalues that are coincident and in the form of . On the sympletic manifold, the periodic orbit can be calculated by the hierarchical grid searching method [19]. The hierarchical parameters are determined by the spherical coordinates of the normal vector of the grid plane. The residual of state [19] is examined for preliminary judgment of periodic orbit as the following equation stated , where xN and x0 are the initial state and end state, respectively. The gradient direction [19] is determined by the initial states of the former orbit, , where and are the location and velocity of the i-th step, . 2. 3 Distribution of Eigenvalues, Orbits, and Submanifolds In this section, we will discuss submanifolds of the symplectic manifold and subspace for the orbit in the potential field of an irregular body, which can help us to determine the motion state, classify the orbits, and analyse the bifurcations. Let 8 0,ie,R;0,0, ie1 if 0sgn1 if 00,ie0resNxxx1002iixxξ0ir0ir020iiUω×ω×rξr, the Jacobian constant be , and the eigenvector of the characteristic multiplier be . Denote the tangent space at the point as . Defining , is the tangent space of the symplectic manifold at , . is the non-degenerate anti-symmetric bilinear 2-form on the tangent space , and is a symplectic space. Denote the asymptotically stable subspace , the asymptotically unstable subspace , the central subspace , the periodic subspace , and the periodic doubling subspace as , , , , and . Denote the asymptotically stable subspaces and as well as the asymptotically unstable subspaces and as and . The asymptotically stable manifold , the asymptotically unstable 9 C is the characteristic multiplier of the orbit CpHcjjup,ppTTSSjjEpspanCuEp,SppTEpSppTS,,pppTEpSsEpuEpcEpeEpdEp,1sjjjEpspanCu,1cjjjEpspanCu,1ujjjEpspanCu,1ejjjEpspanCu,1djjjEpspanCusEpsEpuEpuEp,1,Im0,1,Im0sjjjjsjjjjEpspanCEpspanCuu,1,Im0,1,Im0ujjjjujjjjEpspanCEpspanCuusWSmanifold , the central manifold , the periodic manifold , and the periodic doubling manifold of the orbit (which is seen as a point in the 6-dimensional symplectic manifold) are tangent to the asymptotically stable subspace , the asymptotically unstable subspace , the central subspace , the periodic subspace , and the periodic doubling subspace , respectively. The asymptotically stable manifolds and are tangent to the asymptotically stable subspaces and , respectively. The asymptotically unstable manifolds and are tangent to the asymptotically unstable subspaces and , respectively. Denote as the topological homeomorphism, as the diffeomorphism, and as the direct sum. Then, and . The Krein collision of characteristic multipliers leads to the appearance of the collisional manifold and the collisional subspace. Denote as the collisional manifold, which is tangent to the collisional subspace . Thus, and . Denote as the uniform manifold, which is tangent to the uniform subspace , the manifolds and subspaces of and are uniform and the . When phase diagrams are coincident. Denote 10 as the strongly uWScWSeWSdWSpsEpuEpcEpeEpdEpsWSsWSsEpsEpuWSuWSuEpuEp,scuWWWSSSSscupTEpEpEpSrWS1,Im0,,..,rjjjkkjEpspanstjkurcEpEprcWWSSfWS,,..ReRe0,ImIm0,fjjkjkjkEpspanCstjkudim0fEpjk1,Im0,..,ljjjkkjEpspanstCustable space, where is the number of that satisfies , and is the set of integers. Denote as the strongly stable manifold, which is tangent to the strongly stable subspace and yields and . The topological structure of submanifolds fixes the phase diagram of the motion. Based on the discussion above, one can obtain the following theorem about the topological classification of submanifolds for the periodic orbits in the potential field of a rotating body. Theorem 1. Consider the topological structure of the manifolds for an orbit in the potential field of a rotating celestial body. There exist 6 ordinary cases, 3 purely collisional cases, 3 purely degenerate real saddle cases, 7 purely periodic cases, 7 purely period-doubling cases, 1 periodic and collisional case, 1 periodic and degenerate real saddle case, 1 period-doubling and collision case, 1 period-doubling and degenerate real saddle case, and 4 periodic and period-doubling cases. Distributions of eigenvalues on the complex plane for these cases are shown in Figure 1, and classifications and properties for these cases are shown in Table 1 in Appendix 1. □ The mixed case is a subtype of at least 2 different cases; for example, Case PK1 is a subtype of the periodic cases, meanwhile a subtype of the collisional cases. Consider the orbit as a point in the symplectic manifold. Then, the phase diagram on the asymptotically stable manifold is the motion approach to the point as a sink, while the phase diagram on the asymptotically stable manifold is the 11 Jj1Im0jjZlWScedrlEpEpEpEpEpcedrlWWWWWSSSSSsWSsWSmotion approach to the point as a spiral sink. The phase diagram on the asymptotically unstable manifold is the motion leaving the point as a source, while the phase diagram on the asymptotically unstable manifold is the motion leaving the point as a spiral source. The phase diagram on the central manifold is the motion around the point as a centre. Moreover, the phase diagram on the manifold is the motion around the point as a real saddle, while the phase diagram on the manifold is the motion around the point as a complex and spiral saddle. Fig. 1a. Ordinary cases Fig. 1b. Purely collisional cases 12 uWSuWScWSsuWWSSsuWWSSCase O1Case O2Case O3Case O6Case O4Case O5Case K2Case K3Case K1222233Fig. 1c. Purely degenerate real saddle cases Fig. 1d. Purely periodic cases Fig. 1e. Purely period-doubling cases 13 Case DRS3Case DRS1Case DRS2222233Case P4Case P3Case P1222Case P22Case P54Case P64Case P76Case PD4Case PD3Case PD1222Case PD22Case PD54Case PD64Case PD76 Fig. 1f. Periodic and Fig. 1g. Periodic and Fig. 1h. Fig. 1i. collisional cases degenerate real saddle Period-doubling and Period-doubling and cases collisional cases degenerate real saddle cases Fig. 1j. Period and period-doubling cases 3 Bifurcations of the Periodic Orbits In this section, we discuss bifurcations of quasi-periodic and periodic orbits in the potential field of an irregular body. The following lemma concerns the motion of characteristic multipliers. Lemma 1. Consider the characteristic multipliers of orbits for the motion in the potential field of an irregular body, that the characteristic multipliers on the unit circle will not leave the unit circle before colliding, and that the collisional point is at 1, or ; the characteristic multipliers on the real axis will not leave the real axis before colliding, and the collisional point is at 1, or . Proof. Let be the state transition matrix. Then, if is a characteristic 14 Case PK1222Case PDRS1222Case PDK1222Case PDDRS1222Case PPD142Case PPD242Case PPD322Case PPD42210,ie1sgnR,0,1eM0multiplier and it is the eigenvalue of , the values , , are also eigenvalues of . Denote , and is a sextic polynomial. Assume that there exists a single characteristic multiplier that will leave the unit circle before colliding when parameters change. Then, will also leave the unit circle before colliding when parameters change. After leaving the unit circle, the characteristic multipliers satisfy or , then . If are also eigenvalues of . In other words, 2 new characteristic multipliers appear with no collision, so the sextic polynomial has at least 8 roots, which contradicts the fundamental theorem of algebra. Thus, the assumption is incorrect, proving that the characteristic multipliers on the unit circle will not leave the unit circle before colliding. Likewise, the characteristic multipliers on the real axis will not leave the real axis before colliding. For characteristic multipliers on the unit circle, the collisional point is also on the unit circle; for characteristic multipliers on the real axis, the collisional point is also on the real axis. □ Bifurcations of periodic orbits can be analysed using this lemma. 3.1 The Existence and Continuity of Periodic Orbits with the Parameter Changes The continuation of periodic orbits depends on the existence and continuity of periodic orbits with the parameter changes. The motion in the potential field of an irregular-shaped celestial body is a dynamical system that is dependent on the parameters 15 M01001MdetqMIq00,ie100,ie0,R;0,0,ie0,R;0,0,ie0,R;0,0,ie, , ,R;0,0,iiieeeMqwhere and , (11) , , are parameters related to time . is the parameter for the rotational angular velocity , while is the parameter for the gravitational potential . The periodic orbit is dependent on the parameters and the Jacobian integral H. There are several causes that make the parameters and change, such as the YORP effect [28-29], the surface grain motion [30-31], the electrostatic and rotational ejection of dust particles[32], the disintegration of small celestial bodies [33], the energetic collision and disruption of rubble-pile asteroids [34], the continuation of periodic orbit families [15], etc. The YORP effect, which is the effect of solar radiation on the rotation of asteroids, leads to the change of the rotational speed and the rotational axis [28], i.e. varies; for instance, the asteroid (54509) 2000 PH5’s change in spin rate is (2.0 ± 0.2) × 10−4 deg/day2 [29]. The surface grain motion [30-31], the electrostatic and rotational ejection of dust particles[32], the disintegration of small celestial bodies [33] and the energetic collision and disruption of rubble-pile asteroids [34] make the parameters and synchronously vary. The continuation of periodic orbit families [15][19-20] leads to the change of the parameter . The surface grain motion as well as the electrostatic and rotational ejection of dust particles causes small change of the parameters and .The effect of solar radiation and planet’s gravitational force on the small celestial body’s surface grain leads to the motion of the grain on the surface and the change of the mass distribution, and the charge of the particles and the windmill effect caused by the solar radiation pressure are mainly occurring on the surface of the cometary nucleus. The disintegration of asteroids as well as the energetic collision and disruption of 16 ,20Vμrrωμ×rr1,,2UVUμrωμrωμrμrtμμUUtμμ,UtttμμμμttμμωUUtμμUrppSTμμUμμμUμUμμUμrubble-pile asteroids always cause great change of the parameters and , for example, the disintegration of the main-belt asteroid P/2013 R3 produced 10 or more distinct components and a comet-like dust tail, the rotational velocity and the shape of the asteroid have a substantial change [34]. The following theorem concerns the existence and continuity of the periodic orbit when the parameters change. Theorem 2. The periodic orbit belonging to Cases P1, P2, P3, P4, PK1, PDRS1, PPD2, PPD3 or PPD4 is existent and continuous when the parameter changes. In other words, if a periodic orbit belongs to one of the abovementioned cases with parameters , then there exists an open neighbourhood of , which can be denoted as , and exists a periodic orbit with parameters ; there . Proof. For the abovementioned cases, the periodic orbit has only two characteristic multipliers equal to 1. Consider the Poincaré surface of section, which is transverse with and the periodic orbit ; denote the Poincaré mapping as ; then, the section is a 4-dimensional surface. Eigenvalues of are characteristic multipliers of the orbit, which can be denoted as . Then, apply the implicit function theorem to the Poincaré mapping; for any , there exists a periodic orbit . Thus, the periodic orbit is existent and continuous when the parameter changes. □ Theorem 2 presents the condition of existence and continuity of periodic orbits with the parameter changes. 17 μUμ0ppST00,Hμ00,Hμ00,NGHμ1100,,NHGHμμ1ppSTT1100,,NHGHμμ0ppST0HH0ppSTPPz3,4,5,6;1jjj1100,,NHGHμμ1ppSTT3.2 Bifurcations of the Periodic Orbits Bifurcations of periodic orbits in the potential field of an irregular body include tangent bifurcations, period-doubling bifurcations and Neimark-Sacker bifurcations. Tangent bifurcations occur when the orbits have characteristic multipliers that cross 1; the multiple number of 1 is 4 or 6. Period-doubling bifurcations occur when the orbits have characteristic multipliers that cross ; the multiple number of is 2 or 4. Neimark-Sacker bifurcations occur when the orbits have two equal characteristic multipliers, which are or . Denote as the Poincaré surface of section. Then, . On the manifold , denote as the intersection point from the periodic orbit to the Poincaré surface of section . Then, the asymptotically stable manifold and the asymptotically unstable manifold of the periodic orbit can be expressed as . The invariant manifold and and satisfy and , which are the asymptotically stable manifold and the asymptotically unstable manifold of the fixed point . Let be the first intersection point from the periodic orbit to the Poincaré surface of section after that satisfies . Then, if the period orbit is periodic after rotating the celestial body by circles, we have . 3.2.1 Period-doubling Bifurcation: Mobius Strips and Klein bottles 18 110,ie0,iePdim4PHc00,0,0,0Q0ppSTP0sWp0uWp0p00:,,sWpptpptf00:,,sWpptpptf0sWQ0uWQ00ssWQWpP00uuWQWpP0Q10QgQppSTP0t10QQTk00kgQPQLet be a Poincaré surface of section. The following theorem gives a condition for the period-doubling bifurcation of periodic orbits of a massless test particle in the potential field of an irregular body. Theorem 3. Denote as the intersection point from the periodic orbit to the Poincaré surface of section at time . Let be the first intersection point from the periodic orbit to the Poincaré surface of section after that satisfies . The function satisfies the following conditions: a) ; b) The periodic orbit has characteristic multipliers equal to 1 and and has no other characteristic multipliers; c) . Let be a function that satisfies d) e) ; , where or is the diagonal matrix. Then, there exists an open neighbourhood of , which is denoted as , such that for any : if , there is no periodic orbit with minimal period ; if , then there exists a unique periodic orbit with period for any , which can be denoted as 19 P00,0,0,0QppSTP0t10QgQppSTP0t10QQgQ00QgQppST10330QQdggQdQ,gQparameter000,0,gQgQQgQgQ044,1QQgQQJ441,1,1,1diagJ1,1,1,1diag4400,QG,QG0330QQdggQdQ2T0330QQdggQdQ2T,QG2ppSTwith . Additionally, the period orbit is asymptotically stable if the periodic orbit is asymptotically unstable; the period orbit is asymptotically unstable if the periodic orbit is asymptotically stable. Proof. Here, we only prove the cases where the topological structure belongs to Case PPD1 or PPD2. For other cases, consider the restriction in the submanifold ; one can obtain the conclusion easily. The period orbit satisfies (12) is a root of Eq. (12), with period . The derivatives of have the form (13) and the topological structure belonging to Case PPD1 or PPD2 yields: (14) Then, the Taylor expansion of is where , . Consider the expression 20 (15) (16) ,QG2TpppST2TpppSTdeWWSS2T,,ggQggQQ0QQT,ggQ2,,,,,,,,,,,ggQggQgQggQggQgQggQgQ00000,10,0ggQggQggQggQ,ggQ22323000,126aaggQQQQQQQ200a3000aggQ2023000,226ggQQQaaQQQQQQIf , and have the same symbol. Eq. (16) has no solution, which means that there is no period orbit with the parameters . If , and have different symbols. Then, has two roots, corresponding to a period orbit , and is a solution corresponding to a period orbit with the parameters . In addition, has three solutions, which are , , and . Condition e yields that is not the period of this periodic orbit. If , then is unstable, and if , then the slope of is less than zero, so the period orbit is stable. Similarly, if , then is stable, and the period orbit is unstable. □ Remark 1. Assume that the period orbit exists; then, if the period orbit is periodic after rotating the celestial body on the circle, the period orbit is periodic after rotating the celestial body on the circle. Remark 2. The transfers about characteristic multipliers pass through , leading to the period-doubling bifurcations. For example, for Case PPD1, period-doubling bifurcations include , , , and . Figure 2a shows the appearance of the period-doubling bifurcation generated by . 21 0330QQdggQdQ23a2T,QG0330QQdggQdQ23a223002026aaQQQQ2T,,QgQ2T,QG,0ggQQ0QQ,gQT0ppST211,ggQQ2T0ppST2T2TTk2T2k1Case P5Case PPD1Case P6Case PD2Case PPD1Case PD4Case PD2Case PPD1Case PD3Case PPD3Case PPD1Case PPD4Case P2Case PPD3Case P4Denote as a characteristic multiplier for the periodic orbit . Denote as the submanifold of restricted by . Then, . Denote as the manifold expanded by , from which the expansion makes the periodic orbit from a point to a Jordan curve and causes the dimension of to become 2. Denote as the projective plane, as the Mobius strip, and as the boundary operator. Denote the manifold generated by as . Corollary 1. The manifolds and are both Mobius strips. These two Mobius strips satisfy and . Additionally, the boundary of the manifold is a Klein bottle, which satisfies and . Figure 2a shows the appearance of the period-doubling bifurcation: and the Mobius strip for the period-doubling bifurcation. As shown in the figure 2a, and are both Mobius strips. 22 110p1cWScWS11dim1cWS10cWp1dim1cWS0p10cWprPoM11110cWp10cWp110cWp11100ccoWpWpM11100dimdim2ccWpWp11100usWpWp1112002usrWpWpP11100dim3usWpWpCase P2Case PPD3Case P410cWp110cWpCase P2Case PPD3Case P42222Fig. 2a. Appearance of the period-doubling bifurcation: and the Mobius strip for the period-doubling bifurcation 3.2.2 Tangent Bifurcation The tangent bifurcation occurs when characteristic multipliers cross the point (1, 0) in the complex plane. The following theorem concerns the appearance of the tangent bifurcation for periodic orbits in the potential field of an irregular body. Theorem 4. Consider the motion of a massless test particle in the potential field of an irregular body. Denote as the intersection point from the periodic orbit to the Poincaré surface of section at time . Let be the first intersection point from the periodic orbit to the Poincaré surface of section after , satisfying . The function satisfies the following conditions: a) ; b) the periodic orbit has characteristic multipliers equal to 1, . Let be the function that satisfies 23 10uWp110sWpCase P2Case PPD3Case P400,0,0,0QppSTP0t10QgQppSTP0t10QQgQ00QgQppSTm4 or 6m,gQparameter; ; c) d) e) , where is the unit matrix. Then, there exists an open neighbourhood of , which is denoted by , such that for any , if , then there are no periodic orbits for , while there are two periodic orbits for ; if , then there are no periodic orbits for , while there are two periodic orbits for . Proof: The periodic orbit is a fixed point on the Poincaré surface of section . Denote , then is established if and only if is a fixed point for the function , and Using the implicit function theorem, there exists an open neighbourhood of , , and there exists a function that is defined in , such that for any , implies , while implies . Clearly, . Consider the function at . Then, 24 000,0,gQgQQgQgQ00,,0QQgQ044,1QQgQQI44I4400,QG,QG22,,0gQgQQ0022,,0gQgQQ00P,,QgQQ,0QQ,gQ0000,0,0,,,0QQQQQQgQ00,QGpG,QG,0QpQpQ,0Q00pQ,0pQQ0,0,QQ0000,0,,,'0QQQQQQQQpQQUsing conditions b and d yields and . Thus, . Additionally, Thus, Considering , using the Taylor expansion, Thus, if or , the above equation has no solution and there are no periodic orbits; if or , the above equation has two solutions and there are two periodic orbits. □ The following remark means that the motion state near the tangent bifurcation with parametric variation is sensitive to initial conditions. 25 000,0,,,0QQQQQgQ000,0,,,10QQQQQgQQQ0'0QQpQ0000002220,0,0,220,,,,',''0QQQQQQQQQQQQgQgQgQpQQQgQpQ00020,20,,'',QQQQQQgQQpQgQ00pQ2222,'',22gQQQQpQgQ220,,0gQgQQ220,,0gQgQQ220,,0gQgQQ220,,0gQgQQRemark 3. The transfers about characteristic multipliers pass through , leading to tangent bifurcations. For example, for Case P5, tangent bifurcations include and . Figure 2b shows the appearance of the tangent bifurcation generated by . Fig. 2b. Appearance of the tangent bifurcation: 3.2.3 Krein collision and Neimark-Sacker Bifurcation The following theorem gives a condition for the Neimark-Sacker bifurcation of periodic orbits around an irregular body. The Neimark-Sacker bifurcation occurs when the Krein collision appears. Theorem 5. Consider the motion of a massless test particle in the potential field of an irregular body. Denote as the intersection point from the periodic orbit to the Poincaré surface of section at time . Let be the first intersection point from the periodic orbit to the Poincaré surface of section after that satisfies . The function satisfies the following conditions: a) ; b) the periodic orbit has characteristic multipliers equal to 1, , and ; the multiple number of the 26 1Case P2Case P5Case P4Case P4Case P5Case P2Case P2Case P4Case P3Case P2Case P44Case P322Case P2Case P4Case P300,0,0,0QppSTP0t10QgQppSTP0t10QQgQ00QgQppST0,ie0,iecharacteristic multipliers is equal to 2; c) . Let be the function that satisfies the following: ; ; , where is the diagonal matrix in which . d) e) f) Then, 1) for any , there exists and a function that satisfies and corresponds to a periodic orbit; for any , there exists and a function that satisfies and corresponds to a periodic orbit; 2) there exists an open neighbourhood of , which is denoted as , such that for any , one of the following conditions is established: a) corresponds to an unstable orbit; b) corresponds to an stable quasi-periodic orbit, and the intersection point from the quasi-periodic orbit to the Poincaré surface is on a 2-dimensional tori ; or c) corresponds to an unstable and collisional quasi-periodic orbit, and the intersection point from the quasi-periodic orbit to the Poincaré surface is on a 1-dimensional tori . Proof: Condition e yields that the eigenvalues of the monodromy matrix of the orbit corresponding to are different from the eigenvalues of the monodromy 27 , 0,1ikkNe,gQparameter000,0,gQgQQgQgQ00,,0QQgQ044,1QQgQQK44,,,iiiidiageeeeK440,10,QG1,0Gp11pQ1,0G10,QGp11pQ00,QG,QG,gQ,gQ2T,gQ1T,gQmatrix , where is the monodromy matrix of the periodic orbit . The periodic orbit is a fixed point on the Poincaré surface of section . Let . Then, , is established if and only if is a fixed point for the function , and Using the implicit function theorem, there exists an open neighbourhood of , which is denoted by , and there exists a function that is defined in such that for any , . In addition, for any , there exists and a function that satisfies and corresponds to a periodic orbit; for any , there exists and a function that satisfies and corresponds to a periodic orbit, which leads to conclusion 1. Consider the topological classification of six characteristic multipliers for the orbit on the complex plane. If at least one characteristic multiplier lies outside the unit circle, corresponds to an unstable orbit. If all of the characteristic multipliers lie on the unit circle and the characteristic multipliers are in the form , the intersection point from the quasi-periodic orbit to the Poincaré surface is on a 1-dimensional tori and corresponds to an unstable and collisional quasi-periodic orbit. Additionally, if all of the characteristic multipliers lie on the unit circle and the characteristic multipliers are in the form , the intersection point from the quasi-periodic orbit to the Poincaré surface is on a 2-dimensional tori and 28 MMppSTP,,QgQQ00,0Q,0QQ,gQ0000,0,0,,,0QQQQQQgQ00,QGpG,QG,0QpQ10,QG1,0Gp11pQ1,0G10,QGp11pQ,gQ1,1,,,,0,iiiieeee1T,gQ1212121,1,,,,,0,iiiieeee2T corresponds to a stable quasi-periodic orbit, which leads to conclusion 2. □ Remark 4. The transfers about the characteristic multipliers pass through the unit circle without 1 or , leading to the Neimark-Sacker bifurcations. For example, for Case PK1, the Neimark-Sacker bifurcations include and . Figure 2c shows the appearance of the Neimark-Sacker bifurcation generated by . Fig. 2c. Appearance of the Krein collision and Neimark-Sacker bifurcation: 3.2.4 Real Saddle Bifurcation The following theorem gives a condition for the real saddle bifurcation of periodic orbits around an irregular body. Near the real saddle bifurcation, there exists motion of both types of saddle, namely the real saddle and the complex saddle. Theorem 5. Consider the motion of a massless test particle in the potential field of an irregular body. Denote as the intersection point from the periodic orbit to the Poincaré surface of section at time . Let be the first intersection point from the periodic orbit to the Poincaré surface of section after satisfying . The function satisfies the following conditions: a) ; b) the periodic orbit has characteristic multipliers equal to 1, 29 ,gQ1Case P2Case PK1Case P1Case P1Case PK1Case P2Case P2Case PK1Case P1Case P2Case PK1Case P122222Case P2Case PK1Case P100,0,0,0QppSTP0t10QgQppSTP0t10QQgQ00QgQppST, , where and and the multiple number of the characteristic multipliers is equal to 2. Let be the function that satisfies c) d) e) ; ; , where is the diagonal matrix in which . Then, we can conclude the following: 1) For any , there exists and a function that satisfies and corresponds to an unstable periodic orbit; for any , there exists and a function that satisfies and corresponds to an unstable periodic orbit; 2) there exists an open neighbourhood of , which is denoted by , such that for any , one of the following conditions is established: a) the topological structure of submanifolds is , and the dimensions of submanifolds satisfy ; b) the topological structure of the submanifold is , and the dimensions of submanifolds satisfy .□ 30 sgnR,0,1,1jjjjejsgnR,0,1,1jjjjej1 if 0sgn1 if 0,gQparameter000,0,gQgQQgQgQ00,,0QQgQ044,1QQgQQK44sgn,sgn,sgn,sgnjjjjjjjjdiageeeeK440,1j10,QG1,0Gp11pQ1,0G10,QGp11pQ00,QG,QG,esuWWWSSSSdimdimdim2esuWWWSSS,esuWWWSSSSdimdimdim2esuWWWSSSRemark 5. The transfers about characteristic multipliers pass through the real axis without 1 or , leading to the real saddle bifurcations. For example, for Case PDRS1, real saddle bifurcations include and . Figure 2d shows the appearance of the real saddle bifurcation generated by . Fig. 2d. Appearance of the real saddle bifurcation: 4. Applications to Irregular Celestial Bodies In this section, the theory developed in the previous sections is applied to asteroid 216 Kleopatra and 6489 Golevka, as well as the comet nucleus of 1P/Halley. Physical models of these celestial bodies were calculated using the polyhedral model [24, 25] using data from radar observations [26, 27]. Using the polyhedron method, the gravitational potential of the asteroid [24, 25] can be calculated by Besides[25], . (17) , (18) . (19) where G=6.67×10-11 m3kg-1s-2 is the gravitational constant, σ is the density of the body; 31 1Case P1Case PDRS1Case P3Case P3Case PDRS1Case P1Case P1Case PDRS1Case P3Case P1Case PDRS1Case P322222Case P1Case PDRS1Case P311GG22eeeeffffeedgesffacesULrErrFreeefffeedgesffacesUGLGErFr()eeffeedgesffacesUGLGEFre is a body-fixed vector from the field point to some fixed point on the edge e of face f, rf is a body-fixed vector from the field point to any point in the face plane; Ee and Ff are defined in terms of face- and edge-normal vectors, Ee is the geometric parameter of edges while Ff is the geometric parameter of faces; Le is the factor of integration that operates over the space between the field point and edges e of faces f, ωf is the signed solid angle subtended by planar region relative to the field point. 4.1 Periodic Orbits in the Potential Field of the Asteroid 6489 Golevka and the Comet 1P/ Halley 4.1.1 Asteroid 6489 Golevka The estimated bulk density of asteroid 6489 Golevka is [35], its rotational period is 6.026 h, and its overall dimensions are [35]. Figure 3 show two periodic orbits around the asteroid 6489 Golevka, while Figure 4 shows the topological distribution of characteristic multipliers for these two periodic orbits. Figure 4 shows that these two periodic orbits belong to the periodic cases―Cases P4 and P5. Clearly, the two orbits have the same shape but belong to different cases, which implies that periodic orbits that belong to different cases might have the same shape and that the essential characteristic of periodic orbits is the topological type rather than shape. In addition, the number of shapes is infinite, while the number of different topological types is finite. 32 2.73gcm0.350.250.25km 3a 3b Fig. 3. Two periodic orbits around the asteroid 6489 Golevka belong to Cases P4 and P5 4a 4b Fig. 4. The topological distributions of characteristic multipliers for two periodic orbits around the asteroid 6489 Golevka 4.1.2 Comet 1P/ Halley The nucleus of comet 1P/Halley is an irregular potato-shaped body [36]. The estimated bulk density of comet 1P/Halley’s nucleus is [37], its rotational period is 52.8 h, and its overall dimensions are [26][38]. Figure 5 show two periodic orbits around the comet 1P/Halley, while Figure 6 shows the topological distribution of characteristic multipliers for these two periodic orbits. Figure 6 shows that the periodic orbits in 5a belong to the stable periodic case, Case P2, while the periodic orbits in 5b belong to the unstable periodic case, Case P4, 33 0.63gcm16.8318.76747.7692 kmwhich means that different types of periodic orbits with different stabilities exist in the same highly irregular-shaped celestial body. 5a 5b Fig. 5. Two periodic orbits around the comet 1P/Halley belong to Cases P2 and P4 6a 6b Fig. 6. The topological distributions of characteristic multipliers for these two periodic orbits around the comet 1P/Halley 4.2 Bifurcations of Periodic Orbits around Asteroid 216 Kleopatra The estimated bulk density of asteroid 216 Kleopatra is [5], its rotational period is 5.385 h, and its overall dimensions are [4]. The asteroid 216 Kleopatra is a large size ratio triple asteroid with two moonlets, which are Alexhelios (S/2008 (216) 1) and Cleoselene (S/2008 (216) 2) [5]. Alexhelios is the outer moonlet, with an estimated size of 8.9 ± 1.6 km, while Cleoselene is the inner 34 3.63gcm2179481kmmoonlet, with an estimated size of 6.9 ± 1.6 km [5]. The study of dynamical behaviour around asteroid 216 Kleopatra can help explain the dynamical configuration of this large size ratio triple asteroid, including the orbital evolution, the motion stability, and the regular and chaotic motion of Alexhelios (S/2008 (216) 1) and Cleoselene (S/2008 (216) 2). Yu and Baoyin [20] found four equilibrium points in the potential of asteroid 216 Kleopatra, and all of these four equilibrium points are unstable [18][20]. Figure 7 shows the contour line of the zero-velocity manifold in the xy-plane, which shows that there are at least four equilibrium points outside the asteroid 216 Kleopatra. The hierarchical grid searching method can be used to search periodic orbits around irregular bodies; for 216 Kleopatra, this searching method was used to generate 29 basic periodic orbit families [19]. Collisions [39], perturbations from the sun [40], and the Yarkovsky effect [41] can lead to disruption of the asteroid [42] through bifurcation of the rotational stability and orbital motion [40][43]. Figure 7 clearly shows that there are 4 critical points outside this asteroid [18][20]. Additionally, from figure 7, one can see that there is a critical point near the origin of the body-fixed frame, which is unstable and implies that asteroid 216 Kleopatra will split near the unstable critical point inside the body. 35 Fig. 7. The contour line of the zero-velocity manifold for asteroid 216 Kleopatra in the xy-plane. (The unit of the effective potential is m/s.) 4.2.1 Period-doubling bifurcations of periodic orbits around asteroid 216 Kleopatra The period-doubling bifurcation of periodic orbits around asteroid 216 Kleopatra is discovered and shown in figure 8. Considering the Jacobi integral as the parameter, the periodic orbits change if the parameter changes; there are four periodic orbits around asteroid 216 Kleopatra presented in figure 8, with Jacobi integrals equal to 0.3135, 0.5635, 0.8135, and 1.0635. The unit of the Jacobi integral throughout this paper is . Figure 8b shows that the motion of characteristic multipliers belongs to the period-doubling bifurcation: . 36 32210 msCase P3Case PPD3Case P4Fig. 8a. Four periodic orbits around asteroid 216 Kleopatra when the Jacobi integral parameter changes. The Jacobi integral is equal to 0.3135, 0.5635, 0.8135, and 1.0635, with a unit of . Fig. 8b. Motion of characteristic multipliers that leads to the period-doubling bifurcation 4.2.2 Tangent bifurcations of periodic orbits around asteroid 216 Kleopatra The tangent bifurcation of periodic orbits around asteroid 216 Kleopatra is shown in figure 9. The Jacobi integral is also considered as the parameter. Three periodic orbits around asteroid 216 Kleopatra are presented in figure 9, with the Jacobi integral equal to 1.0635, 1.3135, and 1.5635. From figure 9b, one can see that the motion of characteristic multipliers belongs to the tangent bifurcation: . The periodic orbit with the Jacobi integral 1.0635 belongs to Case P4, while the periodic orbits with the Jacobi integrals 1.3135 and 1.5635 belong to Case P2. 37 32210 msCase P3Case PPD3Case P4Case P4Case P5Case P2Fig. 9a. Three periodic orbits around asteroid 216 Kleopatra when the Jacobi integral parameter changes. The Jacobi integral equals 1.0635, 1.3135, and 1.5635, with a unit of . Fig. 9b. Motion of characteristic multipliers that leads to the tangent bifurcation 4.2.3 Neimark-Sacker bifurcations of periodic orbits around asteroid 216 Kleopatra The Neimark-Sacker bifurcation of periodic orbits around asteroid 216 Kleopatra is shown in figure 10. Three periodic orbits around asteroid 216 Kleopatra are presented in figure 10, with the Jacobi integral equal to 0.8392569, 0.9392569, and 1.0392569. The motion of the characteristic multipliers belongs to the Neimark-Sacker bifurcation: . The periodic orbits with the Jacobi integrals 0.8392569 and 0.9392569 belong to Case P2, while the periodic orbit with the Jacobi integral 1.0392569 belongs to Case P1. 38 32210 msCase P4Case P5Case P2Case P2Case PK1Case P1Fig. 10a. Three periodic orbits around asteroid 216 Kleopatra when the Jacobi integral parameter changes. The Jacobi integral equals 0.8392569, 0.9392569, and 1.0392569, with a unit of . Fig. 10b. Motion of characteristic multipliers that leads to the Neimark-Sacker bifurcation 4.2.4 Real saddle bifurcations of periodic orbits around asteroid 216 Kleopatra The real saddle bifurcation of periodic orbits around asteroid 216 Kleopatra is shown in figure 11. Four periodic orbits around asteroid 216 Kleopatra are presented in figure 11, with the Jacobi integral equal to -1.880585, -1.780585, -1.680585, and -1.580585. The motion of characteristic multipliers belongs to the real saddle bifurcation: . The periodic orbits with the Jacobi integrals -1.880585 and -1.780585 belong to Case P2, while the periodic orbits with the Jacobi integrals -1.680585 and -1.580585 belong to Case P1. Figures 11a shows 39 32210 msCase P2Case PK1Case P1Case P2Case PDRS1Case P1the attraction and exclusion effect of the critical points (Jiang et al. 2014[19]). Fig. 11a. Four periodic orbits around asteroid 216 Kleopatra when the Jacobi integral parameter changes. The Jacobi integral equals -1.880585, -1.780585, -1.680585, and -1.580585, with a unit of . Fig. 11b. Motion of characteristic multipliers that leads to the real saddle bifurcation 4.3 Discussion of Applications The theory developed in the previous sections has been applied to asteroids 216 Kleopatra and 6489 Golevka, as well as the comet nucleus of 1P/Halley. Two periodic orbits that have similar shapes and belong to different cases around the asteroid 6489 Golevka are shown; this result implies that periodic orbits that belong to the different cases might have similar shapes and that the essential characteristic of periodic orbits 40 32210 msCase P2Case PDRS1Case P1is the topological type rather than shape. For the comet 1P/Halley, two periodic orbits are found. These two periodic orbits belong to different periodic cases: one belongs to Case P2, the stable periodic case, while the other belongs to Case P4, the unstable periodic case. Dynamical behaviours around the large size ratio triple asteroid 216 Kleopatra are presented in detail. There is an unstable critical point near the origin of the body-fixed frame for asteroid 216 Kleopatra, and asteroid 216 Kleopatra will split near this unstable critical point. All four types of bifurcations forecasted by the theory in this paper, specifically tangent bifurcations, period-doubling bifurcations, Neimark-Sacker bifurcations and real saddle bifurcations of periodic orbits, are discovered in the potential field of asteroid 216 Kleopatra. The parameter is the Jacobi integral. Each type of bifurcation is shown with the refinement of the parametric variation. 5. Conclusions Nonlinearly dynamical behaviours in the potential field of an irregular body are discussed in this paper. Different distributions of characteristic multipliers are shown to fix the structure of submanifolds, the type of orbits, the dynamical behaviour, and the phase diagram of the motion. The topological classifications and stabilities of orbits in the potential field of an irregular body are presented. The classification includes 34 cases: 6 ordinary cases, 3 collisional cases, 3 degenerate real saddle cases, 7 periodic cases, 7 period-doubling cases, 1 periodic and collisional case, 1 periodic and degenerate real saddle case, 1 period-doubling and collisional case, 1 41 period-doubling and degenerate real saddle case, and 4 periodic and period-doubling cases. For periodic orbits, the period-doubling bifurcation, the tangent bifurcation, the Neimark-Sacker bifurcation, and the real saddle bifurcation of periodic orbits are discovered with parametric variation. Motions near bifurcation are sensitive to initial conditions. It is found that submanifolds appear to be Mobius strips and Klein bottles when the period-doubling bifurcation occurs. As an application of the theory developed here, we study the relevant periodic orbits for the asteroids 216 Kleopatra and 6489 Golevka and the comet 1P/Halley. From the application to asteroid 6489 Golevka, the periodic orbits that belong to different cases can be shown to have potentially similar shapes, and the essential characteristic of periodic orbits is the topological type rather than shape. The application to comet 1P/Halley implies that different types of periodic orbits with different stabilities exist in the same highly irregular-shaped celestial body. All four types of bifurcations forecasted by the theory in this paper are discovered around asteroid 216 Kleopatra, which implies the complex dynamical behaviour in the potential field of this large size ratio triple asteroid. Acknowledgements This research was supported by the National Basic Research Program of China (973 Program, 2012CB720000), the State Key Laboratory Foundation of Astronautic Dynamics (No. 2014ADL0202), and the National Natural Science Foundation of China (No. 11372150). Appendix 1 42 Table 1.a Classifications and properties for the ordinary cases (C1: Cases; C2: Stability; C3: Phase diagram of motion; S: Stable; U: Unstable; K: Krein collision) Characteristic multipliers C2 S C3 Elliptic point U U Mixed point: Elliptic point in Exponential saddle in Mixed point: Elliptic point in Exponential saddle in ; ; U Mixed point: Elliptic point in ; Spiral saddle in U Mixed point: Exponential saddle in ; Spiral saddle in U Exponential saddle C1 O1 O2 O3 O4 O5 O6 Table 1.b Classifications and properties for the periodic cases (C1: Cases; C2: Stability; C3: Phase diagram of motion; S: Stable; U: Unstable; K: Krein collision) C2 U C3 Mixed point: Collisional in ; Spiral saddle in K Mixed point: Collisional in ; Elliptic point in K Mixed point: Collisional in ; Exponential saddle in Characteristic multipliers C1 P1 P2 P3 43 0,;1,2,3,1,2,3,.. jjikjjekjkstsgnR,0,1,1 jjjjej120,;1,2 jijejcWSsuWWSS12R,0,1;sgn1,2jjjjej0,,1 jijejcWSsuWWSS0,,1 jijej,R;0,0,iecWSsuWWSSsgnR,0,1,1 jjjjej,R;0,0,iesuWWSSsuWWSSR,0,1,1,2,3sgn,1,2,3,..jjjjkjjekjkst1;1,2 jjj,R;0,0,ieeWSsuWWSS1;1,2 jjj120,;1,2 jijejeWScWS1;1,2 jjj12R,0,1;sgn1,2jjjjejeWSsuWWSSP4 P5 P6 P7 K Mixed point: Collisional in Elliptic point in Exponential saddle in K Mixed point: Collisional in Elliptic point in ; ; ; K Mixed point: Collisional in ; Exponential saddle in K Collisional point Table 1.c Classifications and properties for the period-doubling cases (C1: Cases; C2: Stability; C3: Phase diagram of motion; S: Stable; U: Unstable; K: Krein collision) Characteristic multipliers C1 PD1 PD2 PD3 PD4 PD5 C2 U C3 Mixed point: Collisional in ; Spiral saddle in K Mixed point: Collisional in Elliptic point in K Mixed point: Collisional in ; ; Degenerate exponential saddle in K Mixed point: Collisional in ; Elliptic point in ; Exponential saddle in K Mixed point: Collisional in ; 44 1;1,20,,1sgnR,0,1,1jjjjijjjjjejejeWScWSsuWWSS1;1,2,3,4jjj0,,1 jijejeWScWS1;1,2,3,4jjjsgnR,0,1,1jjjjejeWSsuWWSS1;1,2,...,6jjj1;1,2 jjj,R;0,0,ieeWSsuWWSS1;1,2 jjj120,;1,2 jijejeWScWS1;1,2 jjj12R,0,1;sgn1,2jjjjejeWSsuWWSS1;1,20,,1sgnR,0,1,1jjjjijjjjjejejeWScWSsuWWSS1;1,2,3,4jjjeWSPD6 PD7 Elliptic point in Mixed point: Collisional in ; Exponential saddle in K K Collisional point Table 1.d Classifications and properties for the collisional cases (C1: Cases; C2: Stability; C3: Phase diagram of motion; S: Stable; U: Unstable; K: Krein collision) Characteristic multipliers C1 K1 K2 K3 C2 K C3 Collisional point K Collisional point K Collisional point Table 1.e Classifications and properties for the degenerate real saddle cases (C1: Cases; C2: Stability; C3: Phase diagram of motion; S: Stable; U: Unstable; K: Krein collision) C1 Characteristic multipliers DRS1 DRS2 DRS3 C2 U C3 Degenerate exponential saddle U Degenerate exponential saddle U Mixed point: Elliptic point in ; Degenerate exponential saddle in Table 1.f Classifications and properties for the mixed cases (C1: Cases; C2: Stability; C3: Phase diagram of motion; S: Stable; U: Unstable; K: Krein collision) C1 Characteristic multipliers C2 C3 45 0,,1 jijejcWS1;1,2,3,4jjjsgnR,0,1,1jjjjejeWSsuWWSS1;1,2,...,6jjj1230,,;1,2,3 jijej1230,,;1,2,3 jijejsgnR,0,1,1jjjjej120,,;1,2jijej123R,0,1,1,2,3sgnjjjjje123R,0,1,1,2,3sgnjjjjje0,;1jijej12R,0,1,1,2sgnjjjjjecWSsuWWSS S Mixed point: Fixed point in Elliptic point in U Mixed point: Fixed point in Degenerate exponential saddle in ; ; S U S S S Mixed point: Flipping point in ; Elliptic point in Mixed point: Flipping point in ; Degenerate exponential saddle in Mixed point: Fixed point in Flipping point in Mixed point: Fixed point in Flipping point in Mixed point: Fixed point in Flipping point in Elliptic point in ; ; ; ; U Mixed point: Fixed point in ; Flipping point in ; Degenerate exponential saddle in PK1 PDRS1 PDK1 PDDRS1 PPD1 PPD2 PPD3 PPD4 References 1. Lazzaro, D., Ferraz-Mello, S., Fernández, J. A.: Solar system binaries. Proceeding IAU symposium. 229, 301-318 (2006) 2. Taylor, P. A., Margot, J. L.: Binary asteroid systems: Tidal end states and estimates of material properties. Icarus 212 (2), 661-676 (2011) 3. Marchis, F., Descamps, P., Hestroffer, D., et al.: Discovery of the triple asteroidal system 87 Sylvia. Nature 436(7052), 822-824 (2005) 46 1;1,2 jjj120,,;1,2jijejeWScWS1;1,2 jjj12R,0,1,1,2sgnjjjjjeeWSsuWWSS1;1,2 jjj120,,;1,2jijejdWScWS1;1,2 jjj12R,0,1,1,2sgnjjjjjedWSsuWWSS1;1,2,3,41;5,6jjjjjjeWSdWS1;1,2,3,41;5,6jjjjjjeWSdWS1;1,21;3,4jjjjjj0,,1jijejeWSdWScWS1;1,21;3,4jjjjjjsgnR,0,1,1jjjjejeWSdWSsuWWSS4. Ostro, S. J., Hudson, R. S., Nolan, M. C., et al.: Radar observations of asteroid 216 Kleopatra. Science 288(5467), 836-839 (2000) 5. Descamps, P., Marchis, F., Berthier, J.: Triplicity and physical characteristics of Asteroid (216) Kleopatra. Icarus 211(2), 1022-1033 (2011) 6. Barucci, M. A., Fulchignoni, M., Fornasier, S., et al.: Asteroid target selection for the new Rosetta mission baseline 21 Lutetia and 2867 Steins. Astron. Astrophys. 430, 313-317 (2005) 7. Barucci, M. A., Cheng, A. F., Michel, P., et al.: MarcoPolo-R near earth asteroid sample return mission. Exp. Astron. 33(2-3), 645-684 (2012) 8. Scheeres, D. J.: Orbital mechanics about asteroids and comets. J. Guid. Control. Dynam. 35(3), 987-997 (2012) 9. Scheeres, D. J.: Orbital mechanics about small bodies. Acta. Astronaut. 7, 21-14 (2012) 10. Takahashi Y., Scheeres, D. J., Werner, R. A.: Surface gravity fields for asteroids and comets. J. Guid. Control. Dynam. 36(2), 362-374 (2013) 11. Kaasalainen, M., Ďurech, J., Warner, B. D., et al.: Acceleration of the rotation of asteroid 1862 Apollo by radiation torques. Nature 446(7134), 420-422 (2007) 12. Bennera, L. A. M., Ostroa, S. J., Magrib, C., et al.: Near-Earth asteroid surface roughness depends on compositional class. Icarus 198(2), 294-304 (2008) 13. Kaasalainen, M., Torppa, J., Piironen, J.: Binary structures among large asteroids. Astron. Astrophys. 383, 19-22 (2002) 14. Hartmann, W. K.: The Shape of Kleopatra. Science 288 (5467), 820-821 (2000) 15. Scheeres, D. J., Ostro, S. J., Hudson, R. S., et al.: Orbits close to asteroid 4769 Castalia, Icarus 121, 67-87 (1996) 16. Scheeres, D. J., Ostro, S. J., Hudson, R. S., et al.: Dynamics of orbits close to asteroid 4179 Toutatis. Icarus 132(1), 53-79 (1998) 17. Scheeres, D. J., Williams, B. G., Miller, J. K.: Evaluation of the dynamic environment of an asteroid: Applications to 433 Eros. J. Guid. Control. Dynam. 23(3), 466-475 (2000) 18. Jiang, Y., Baoyin, H., Li, J., et al.: Orbits and manifolds near the equilibrium points around a rotating asteroid. Astrophys. Space Sci. 349, 83-106 (2014) 19. Yu, Y., Baoyin, H.: Generating families of 3D periodic orbits about asteroids. Mon. Not. R. Astron. Soc. 427(1), 872-881(2012) 20. Yu, Y., Baoyin, H.: Orbital dynamics in the vicinity of asteroid 216 Kleopatra. Astron. J. 143(3), 62-70(2012) 21. Elipe, A., Lara, M.: A simple model for the chaotic motion around (433) Eros. J. Astron. Sci. 51(4), 391-404 (2003) 22. Mondelo, J. M., Broschart, S. B., Villac, B. F.: Dynamical Analysis of 1: 1 Resonances near Asteroids: Application to Vesta. Proceedings of the 2010 AIAA/AAS Astrodynamics Specialists Conference. Aug. 2-5, Toronto, Ontario, Canada (2010) 23. Riaguas, A., Elipe, A., Lara, M.: Periodic orbits around a massive straight segment. Celest. Mech. Dyn. Astron. 73(1/4), 169-178 (1999) 24. Werner, R. A.: The gravitational potential of a homogeneous polyhedron or don't cut corners. Celest. Mech. Dyn. Astron. 59(3), 253-278 (1994) 25. Werner, R. A., Scheeres, D. J.: Exterior gravitation of a polyhedron derived and compared with harmonic and mascon gravitation representations of asteroid 4769 Castalia. Celest. Mech. Dyn. Astron. 65(3), 313-344 (1997) 47 26. Stooke, P.: Small body shape models. EAR-A-5-DDR-STOOKE-SHAPE-MODELS-V1.0. NASA Planetary Data System, (2002) 27. Neese, C., E.: Small Body Radar Shape Models V2.0. EAR-A-5-DDR-RADARSHAPE-MODELS-V2.0, NASA Planetary Data System, (2004) 28. Scheeres, D. J.: The dynamical evolution of uniformly rotating asteroids subject to YORP. Icarus. 188, 430-450 (2007) 29. Taylor, P. A., Margot, J. L., Vokrouhlický, D., et al.: Spin rate of asteroid (54509) 2000 PH5 increasing due to the YORP effect. Science 316(5822) , 274-277(2007) 30. Bellerose, J., Girard, A., Scheeres, D. J.: Dynamics and control of surface exploration robots on asteroids. Optimization & Cooperative Ctrl. Strategies, LNCIS 381, 135-150 (2009) 31. Tardivel, S., Scheeres, D. J.: Contact motion on surface of asteroid. J. Spacecraft Rockets. 51(6),1857-1871 (2014) 32. Oberc, P.: Electrostatic and rotational ejection of dust particles from a disintegrating cometary aggregate. Planet. Space Sci. 45(2), 221-228 (1997) 33. Jewitt, D., Agarwal, J., Li, J., et al.: Disintegrating asteroid P/2013 R3. Astrophys. J. Lett. 784(1), L8 (2014) 34. Asphaug, S. J., Ostro, R. S., Hudson, D. J.: Disruption of kilometer-sized asteroids by energetic collisions. Nature 393, 437-440 (1998) 35. Mottola, S., Erikson, A., Harris, A. W., et al.: Physical model of near-Earth asteroid 6489 Golevka (1991 JX) from optical and infrared observations. Astron. J. 114(3), 1234-1245 (1997) 36. Sagdeev, R. Z., Szabó, F., Avanesov, G. A., et al.: Television observations of comet Halley from Vega spacecraft. Nature 321, 262-266 (1986) 37. Sagdeev, R. Z., Elyasberg, P. E., Moroz, V. I.: Is the nucleus of comet Halley a low density body? Nature 331, 240-242 (1988) 38. Peale, S. J., Lissauer, J. J.: Rotation of Halley's comet. Icarus 79(2), 396-430 (1989) 39. Michel, P., Benz, W., Tanga, P., et al.: Collisions and gravitational reaccumulation: Forming asteroid families and satellites. Science 294(5547), 1696-1700 (2001) 40. Scheeres, D. J., Fahnestock, E. G., Ostro, S. J., et al.: Dynamical configuration of binary near-Earth asteroid (66391) 1999 KW4. Science 314(5803), 1280-1283 (2006) 41. La Spina, A., Paolicchi, P., Kryszczyńska, A., et al.: Retrograde spins of near-Earth asteroids from the Yarkovsky effect. Nature 428(6981), 400-401 (2004) 42. Jewitt, D., Weaver, H., Agarwal, J., et al.: A recent disruption of the main-belt asteroid P/2010 A2. Nature 467(7317), 817-819 (2010) 43. Walsh, K. J., Richardson, D. C., Michel, P.: Rotational breakup as the origin of small binary asteroids. Nature 454 (7201), 188-191(2008) 48
1811.12724	1	1811	2018-11-30T11:05:54	High-speed photometry of the eclipsing polar UZ Fornacis	[ "astro-ph.EP", "astro-ph.SR" ]	We present 33 new mid-eclipse times spanning approximately eight years of the eclipsing polar UZ Fornacis. We have used our new observations to test the two-planet model previously proposed to explain the variations in its eclipse times measured over the past $\sim$35 years. We find that the proposed model does indeed follow the general trend of the new eclipse times, however, there are significant departures. In order to accommodate the new eclipse times, the two-planet model requires that one or both of the planets require highly eccentric orbits, that is, $e \geq$ 0.4. Such multiple planet orbits are considered to be unstable. Whilst our new observations are consistent with two cyclic variations as previously predicted, significant residuals remain. We conclude that either additional cyclic terms, possibly associated with more planets, or other mechanisms, such as the Applegate mechanism are contributing to the eclipse time variations. Further long-term monitoring is required.	astro-ph.EP	astro-ph	Astronomy & Astrophysics manuscript no. aanda December 3, 2018 c(cid:13)ESO 2018 High-speed photometry of the eclipsing polar UZ Fornacis Z. N. Khangale1, 2, S. B. Potter1, E. J. Kotze1, P. A. Woudt2 and H. Breytenbach1, 2 1 South African Astronomical Observatory (SAAO), PO Box 9, Observatory 7935, Cape Town, South Africa e-mail: [email protected] 2 Department of Astronomy, University of Cape Town, Private Bag X3, Rondebosch 7701, South Africa 8 1 0 2 v o N 0 3 . ] P E h p - o r t s a [ 1 v 4 2 7 2 1 . 1 1 8 1 : v i X r a Received 09 August 2018; accepted 21 November 2018 ABSTRACT We present 33 new mid-eclipse times spanning approximately eight years of the eclipsing polar UZ Fornacis. We have used our new observations to test the two-planet model previously proposed to explain the variations in its eclipse times measured over the past ∼35 years. We find that the proposed model does indeed follow the general trend of the new eclipse times, however, there are significant departures. In order to accommodate the new eclipse times, the two-planet model requires that one or both of the planets require highly eccentric orbits, that is, e ≥ 0.4. Such multiple planet orbits are considered to be unstable. Whilst our new observations are consistent with two cyclic variations as previously predicted, significant residuals remain. We conclude that either additional cyclic terms, possibly associated with more planets, or other mechanisms, such as the Applegate mechanism are contributing to the eclipse time variations. Further long-term monitoring is required. Key words. planets and satellites: detection -- planets and satellites: formation -- planetary systems -- novae, cataclysmic variables -- stars: individual: UZ Fornacis -- binaries: eclipsing 1. Introduction UZ Fornacis (hereafter UZ For) is an AM Herculis-type eclips- ing magnetic cataclysmic variable (CV) star discovered with the EXOSAT as an X-ray source, EXO 033319-2554.2 (Giommi et al. 1987). It has an orbital period of ∼126.5 min and spectral type of M4.5 (Bailey & Cropper 1991). UZ For has been stud- ied extensively on a wide range of wavelengths including optical (e.g. Allen et al. 1989; Schwope et al. 1990; Bailey & Cropper 1991), x-ray (e.g. Still & Mukai 2001, and references therein), ultra-violet (Stockman & Schmidt 1996) and extreme-ultraviolet (e.g. Warren et al. 1993; Sirk et al. 1994; Warren et al. 1995). UZ For displays one or two accretion spots depending on the accre- tion state, with magnetic fields of ∼53 and ∼48 MG (Beuermann et al. 1988; Ferrario et al. 1989; Bailey & Cropper 1991). The evolution of CVs is governed by the binary's angular momentum and stellar masses, and how these parameters change over time. It is generally understood that CVs evolve from longer to shorter orbital periods. After the common-envelope phase, when mass transfer begins, angular momentum is constantly be- ing exchanged between the white dwarf (WD) and red dwarf resulting in the shrinking of the orbital separation and thus re- ducing the orbital period of the binary. For systems with short orbital periods (≤ 3 h), the main angular momentum loss mech- anism is gravitational radiation (Faulkner 1971; Paczynski & Sienkiewicz 1981). At longer periods (≥3 h), the angular mo- mentum loss is driven by a mechanism called magnetic braking (Verbunt & Zwaan 1981; Rappaport et al. 1983). In one exam- ple, Brinkworth et al. (2006) argue that the period change in the post-common-envelope binary NN Ser can be explained by ei- ther a genuine angular momentum loss from the system or the presence of unseen companions around the binary system. Their observations showed that NN Ser was losing angular momentum at a rate predicted by Rappaport et al. (1983) but only if they assume that magnetic braking was not cut off as the secondary reaches 0.3 M(cid:12). However, Schreiber et al. (2010) present the re- sults of 670 post-common-envelope binaries and found strong evidence for disrupted magnetic braking at the fully convective boundary. The accretion rate in CVs is highly variable, and the majority of these systems change from a high state to a low state and back over timescales ranging from days to months and years. These variations range from both flickering typical of CVs on time- scales of minutes, and significant changes in the overall shape of the light curves. A number of factors can contribute to those vari- ations which include the activity and the shape of the secondary star. For example, the UZ For binary system has been reported to switch between the faint state (Bailey & Cropper 1991) and the bright state (Imamura & Steiman-Cameron 1998) on a timescale of years. The one constant in the light curve of eclipsing po- lars is the eclipse of the WD by the secondary star: generally, the ingress and egress last for ∼30 s. For UZ For the ingress and egress are rapid at ∼3 s enabling accurate determinations of mid-eclipse times, based on the midpoint between the ingress and egress. Recently, a number of systems have been found to show variations in their times of eclipse, for example, NN Ser (Qian et al. 2009; Beuermann et al. 2010), HU Aqr (Qian et al. 2011; Go´zdziewski et al. 2015) and DP Leo (Qian et al. 2010; Beuer- mann et al. 2011). Neither gravitational radiation nor magnetic braking is sufficient to explain the period changes observed. Sev- eral explanations for these eclipse time variations have been of- fered in the literature which include either solar-type magnetic cycles of the secondary star (Applegate mechanism: Applegate 1992) or the presence of circumbinary planets in an orbit around the binary, for example Brinkworth et al. (2006). Article number, page 1 of 11 A&A proofs: manuscript no. aanda According to Applegate's mechanism, the period variations result from quasi-periodic changes in the quadruple moment of the secondary star due to magnetic activity. In this model, it is assumed that a strong magnetic field is produced by a dynamo cycle resulting in the redistribution of the angular momentum within the star and hence a change in its quadruple moment. Hall (1989) found that for Algols there is a strong connection between the orbital period variations and the presence of mag- netic activity. However, Brinkworth et al. (2006) report that the orbital variation in NN Ser can not be explained by the Apple- gate mechanism. The original Applegate model linking magnetic activity to orbital period variations has been reviewed by different au- thors, for example Lanza et al. (1998). Recently, Völschow et al. (2016), presented an improved version of Applegate's mecha- nism which now includes the angular momentum exchange be- tween a finite shell and the core of the star to derive the general conditions under which the Applegate's mechanism can operate. They find that, out of the 16 systems that were analysed, only four systems (e.g. QS Vir, DP Leo, V471 Tau, BX Dra) could be explained by the improved Applegate's mechanism. For the remaining systems, more than the total energy generated by the secondary star is necessary to power the binary's period varia- tions. They note that for UZ For and three other systems, the ratio of energy required to power the improved Applegate mech- anism to the total energy generated by the secondary star is al- most unity. In the case of NN Ser, the ratio of energy required to power the improved Applegate mechanism to the total energy generated by the secondary star is greater than unity, implying that it can not be explained by magnetic activity. In the case of the period changes which are due to the pres- ence of a companion(s) in the binary, the observed minus cal- culated (O - C) time of eclipses vary as the binary orbits the centre of mass of the system. These small variations appear as periodic variations in the O - C diagram due to the light trav- eltime effect. However, high-eccentric and/or multi-planet so- lutions are required to fully explain the O - C variations (e.g. HU Aqr: Horner et al. 2011; Wittenmyer et al. 2012). These can be problematic for dynamically stable orbits. In addition, Hinse et al. (2012) re-analysed the eclipse times of HU Aqr and also found that the best-fitting model requires dynamically unstable solutions with high eccentricities for two companions. Never- theless, examples of dynamically stable solutions are possible if non-coplanar, high eccentric and even retrograde orbits are used (Hinse et al. 2012; Go´zdziewski et al. 2015). On the other hand, a two-planet solution for NN Ser has been shown to be dynamically stable and survived follow-up eclipse time measurements (Horner et al. 2012; Beuermann et al. 2013; Marsh et al. 2014). Furthermore, Bours et al. (2016) reported that the long-period quadratic term in the model of NN Ser is in the direction of lengthening period and can not be explained by natural processes that lead to angular momentum loss. This leaves the circumbinary planet hypothesis as an option to explain the periodicities in NN Ser. This is further supported by the ex- istence of the circumbinary disc around NN Ser (Hardy et al. 2016). Recently, Bours et al. (2016) carried out a long-term pro- gramme of eclipse measurements on 67 WDs in close binaries to detect the period variations. Their results show that systems with baselines exceeding ten years, and with companions of spectral types M5 or earlier, appear to show greater eclipse times varia- tions than systems with companions of spectral types later than M5. They found this to be consistent with an Applegate-type mechanism. However, they also considered it reasonable to as- Article number, page 2 of 11 sume that some planetary systems could exist around evolved WDs binaries, for example NN Ser (Beuermann et al. 2013; Hardy et al. 2016). A recent study by Pulley et al. (2018) agrees with the earlier conclusion by Bours et al. (2016) that higher val- ues of O - C residuals are found with secondary companions of spectral type M5/6 or possibly earlier as a result of an Applegate mechanism. The spectral type of UZ For is dM4.5 (Beuermann et al. 1988) suggesting that an Applegate-type mechanism could be significant in this system. Beuermann et al. (1988) used a quadratic ephemeris to fit eight eclipse times of UZ For which possibly indicated a de- crease in orbital period. Follow-up studies by Ramsay (1994) and Imamura & Steiman-Cameron (1998) pointed at an increas- ing orbital period of UZ For. Perryman et al. (2001) derived a linear ephemeris using their three eclipses combined with earlier six timings by Bailey & Cropper (1991). However, they noticed that their new ephemeris leaves residuals of order ±50 s when compared with historical data taken earlier. Dai et al. (2010) presented 44 mid-eclipses of UZ For and noticed a deviation from linear and quadratic trends in the O - C diagram of UZ For. They explained the deviations by adding a sinusoidal term to the ephemeris and attributed the cyclic variation as due to a planet with the period of ∼23(5) years. Potter et al. (2011) presented new mid-eclipse times, including those from literature and in- cluded in Dai et al. (2010), over the 28 year baseline and noticed a deviation from linear and quadratic trends with amplitudes of 60 s. They interpreted this as the result of two cyclic variations due to two extrasolar planets in orbit around the binary with pe- riods of ∼16(3) years and ∼5.25(25) years. However, they did not rule out the possible effect of a magnetic cycle mechanism. In this paper we present new photometric observations of the eclipsing system UZ For spanning an additional eight years with the aim of investigating the purported cyclic variations (Potter et al. 2011, hereafter Paper I) further. Sect. 2 gives an account of all our observations. In Sect. 3, we show the new eclipse O- C results as well as updated fitting parameters for the planetary model. We provide a general discussion in Sect. 4 and conclusion in Sect. 5. 2. Observations Photometric observations were made between 2011 March and 2018 February on the 1.0-m and 1.9-m telescopes located on the Sutherland site of the South African Astronomical Obser- vatory (SAAO), using either the HIgh-speed Photo-POlarimeter (HIPPO, Potter et al. 2010) or the Sutherland High-speed Opti- cal Camera (SHOC, Gulbis et al. 2011; Coppejans et al. 2013). The log of observations is shown in Table 1. All the observations were made in good seeing conditions. The HIPPO instrument was operated in its photo-polarimetry mode (all-Stokes) and the observations were clear filtered (3500 -- 9000 Å). Background sky measurements were taken at frequent intervals during the observations. All of our observa- tions were synchronized to GPS to better than a millisecond. Data reduction was carried out following the procedures de- scribed in Potter et al. (2010). A total of 16 eclipses were ob- tained in photometric conditions. The SHOC detector was used in a frame-transfer mode with a clear filter, a binning of 8 × 8 or 16 × 16 and exposure times of one second. Differential photometry was performed on the re- sulting data cubes using the SHOC-pipeline described in Coppe- jans et al. (2013). A total of 17 high-time resolution and high signal-to-noise ratio eclipses of the target were obtained. Z. N. Khangale et al.: High-speed photometry of the eclipsing polar UZ Fornacis Table 1. Observational log of UZ For. All observations were made with the HIPPO and SHOC instruments on the SAAO 1.0-m and 1.9-m telescope. Date of Observation Telescope SAAO Instrument used 2018-02-21 2017-11-16 2017-10-27 2016-11-14 2016-11-11 2015-09-04 2015-07-31 2015-03-21 2015-02-23 2015-02-18 2014-10-28 2014-10-26 2014-10-24 2014-10-23 2014-10-22 2014-03-02 2014-01-12 2013-11-28 2013-10-03 2013-03-17 2013-03-16 2013-03-13 2013-02-08 2012-12-17 2012-07-17 2012-02-26 2011-11-01 2011-10-31 2011-10-27 2011-03-08 1.9-m 1.9-m 1.0-m 1.0-m 1.0-m 1.9-m 1.9-m 1.9-m 1.0-m 1.0-m 1.9-m 1.9-m 1.9-m 1.9-m 1.9-m 1.9-m 1.9-m 1.9-m 1.0-m 1.9-m 1.9-m 1.9-m 1.9-m 1.0-m 1.9-m 1.9-m 1.9-m 1.9-m 1.9-m 1.9-m HIPPO HIPPO SHOC SHOC SHOC SHOC SHOC HIPPO SHOC SHOC SHOC SHOC SHOC SHOC SHOC HIPPO SHOC HIPPO SHOC HIPPO HIPPO HIPPO SHOC SHOC HIPPO HIPPO HIPPO HIPPO HIPPO HIPPO Length of observations (hours) 1.00 5.12 2.25 2.03 2.08 1.00 1.00 1.35 0.85 2.06 2.70 0.62 0.77 1.17 1.00 1.65 1.13 1.68 0.52 1.11 0.79 1.43 1.02 0.59 0.88 1.63 1.26 0.90 2.60 0.37 Binning Number of eclipse(s) - - 8 × 8 8 × 8 8 × 8 8 × 8 8 × 8 - 16 × 16 16 × 16 16 × 16 16 × 16 16 × 16 16 × 16 16 × 16 - 8 × 8 - 16 × 16 - - - 8 × 8 16 × 16 - - - - - - 1 3 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 A total of 33 new eclipses of UZ For were obtained. All the eclipse times were corrected for the light travel-time to the barycentre of the solar system, converted from Julian dates to Barycentric dynamical time (TDB) system as Barycentric Julian date (BJD, Eastman et al. 2010). This was done in order to re- move any timing systematics, particularly due to the unprece- dented accumulation of leap seconds with universal time cen- tral (utc) and the effects due to the influence of Jupiter and Sat- urn when heliocentric corrections only are applied. The times of mid-eclipse were determined as the midpoint between the steep drop to minimum (ingress) and the steep rise out of minimum (egress) of the main accretion spot of UZ For as shown in Figure 1. The ingress and egress are marked with blue dashed lines and the adopted time of eclipse is marked by the black dashed line in Figure 1. We estimated an error of approximately 0.00003 days for each of the new eclipses. The new mid-eclipse times were combined with the 42 mid-eclipse times presented in Paper I to give a new total of 75 and we present the results in Table 3. 3. Results 3.1. Eclipse profiles Figure 1 shows two of the eclipse profiles of UZ For obtained during a faint and bright state (left and right panels respectively). There was no flux calibration performed, however, the relative signal-to-noise of the two datasets is indicative of the relative brightness. The eclipse profiles can be understood in the frame- work of the standard polar model (Beuermann et al. 1988; Fer- rario et al. 1989; Bailey & Cropper 1991). UZ For shows either one- (Bailey & Cropper 1991; Imamura & Steiman-Cameron 1998) or two-pole (Perryman et al. 2001) accretion spots de- pending on the accretion states. Both panels show clearly defined ingresses and egresses of the main accretion spot lasting a few seconds, indicated by vertical blue dashed lines. The faint state eclipse profile (Fig. 1, left panel) appears flat- ter and noisier than the brighter state eclipse profile. It is appar- ent that there are two clear stages in both the ingress and egress. The ingress is defined by a fast drop (∼3 s) in counts followed by a more gradual decline to a minimum (∼50 s). The eclipse remains flat for (∼380 s) and the egress begins with a slow ini- tial rise (∼30 s) followed by a rapid rise (∼3 s). The shape of this eclipse profile is similar to those of the low-state of UZ For presented in Bailey & Cropper (1991). The bright state eclipse profile (Fig. 1, right panel) also shows a fast drop in counts during ingress (∼3 s) but now fol- lowed by a brighter and longer decline (∼260 s) to a minimum, compared to the fainter eclipse profile, resulting in a shorter time for the flat part (∼169 s). The egress begins with a two-step in- crease in counts (∼34 s), possibly indicating accretion at the sec- Article number, page 3 of 11 A&A proofs: manuscript no. aanda Fig. 1. Two example eclipses of UZ For obtained with HIPPO (left, 26 Feb 2012) and SHOC (right, 26 October 2014) instruments. The vertical blue dashed lines represent the times of ingress and egress of the main accretion spot, whereas the vertical black dashed line marks the time of mid-eclipse. ond pole, followed by a rapid increase (∼3 s) as the main pole egresses. The various stages are consistent with bright-state accre- tion with two accretion spots, and a brighter magnetic accretion stream contributing to the overall brightness of the system. Our bright-state eclipses are similar to those presented by Imamura & Steiman-Cameron (1998). In both cases, the faint and bright state, the eclipse width remained the same at ∼472 s consistent with the eclipse of the main accretion spot. 3.2. The new eclipse times Table 3 lists all of our new mid-eclipse times as well as those presented in Paper I. We used the epoch (T0) and the orbital pe- riod from Paper I to calculate the cycle number for each of the new mid-eclipse times. The orbital period is accurate enough to assign cycle numbers to the entire 35 years of eclipses. Figure 2 shows the O - C diagram of UZ For spanning ∼35 years. The crosses to the left of the vertical green line are those eclipse times presented in Paper I (up to 2010) and the crosses to the right are the new eclipse times amounting to an addi- tional eight years. Over-plotted is the solution from Paper I (red curves). The top panel shows the O - C after subtraction of the linear ephemeris and it is well known that a linear fit does not give a valid description of the data. The residuals on the top panel appears periodic and therefore, we need more terms in addition to the quadratic term in order to describe the O - C diagram of UZ For. The second panel shows the O - C after subtraction of the quadratic term with the first elliptical term overplotted (red curve), the third panel show the O - C residuals after subtraction of the first elliptical term with the second elliptical term over- plotted (red curve), and the bottom panel show the final O - C residuals after subtraction of the second elliptical term. The new mid-eclipse times were not included in the fit. We note that the original fit was not the formal best solution but instead included the constraint that the eccentricities of the elliptical terms were ≤ 0.1 in order to be consistent with a stable two-planet model. The mathematical function fitted to the old data (the crosses to the left of the green lines) predicted a maximum in the O - C residuals around ∼2010-2011 followed by a gradual decline in the residuals (Fig. 2, second and third panels) followed by a minimum in the O - C residuals. Our new eclipse times (the Article number, page 4 of 11 crosses to the right of the green lines) agree with the predictions in the sense that a maximum in the O - C residuals occurred around ∼2011-2012 followed by a gradual decline in the residu- als. However, the gradual decline appears steeper than predicted and occurs earlier and continues to diverge after ∼2013 until the minimum in the residuals is reached in 2016-2017. The mini- mum is followed by an upturn in the O - C residuals in agreement with predictions. 3.3. Calculating a new O - C The model from Paper I and shown in Fig. 2 consists of a com- bination of a quadratic and two elliptical terms, Eq. 1 below: T(BJDTDB) = T0 + PbinE + AE2 +Kbin,(3)sin(υ3 − 3) +Kbin,(4)sin(υ4 − 4) [1 − e2 3] [1 − e2 4] [1 + e3cos(υ3)] [1 + e4cos(υ4)] . (1) Substituting υ(3,4) = (E + T(3,4)) f(3,4) in the above equation result in the following: T(BJDTDB) = T0 + PbinE + AE2 +Kbin,(3)sin((E + T3) f3 − 3) +Kbin,(4)sin((E + T4) f4 − 4) [1 + e3cos((E + T3)f3)] [1 − e2 3] [1 − e2 4] [1 + e4cos((E + T4)f4)] , (2) where T0 is the time of epoch, Pbin is the orbital period of the binary in days, A is the quadratic parameter and E is the bi- nary cycle number which comprises the quadratic term of the ephemeris. The remaining ten parameters were introduced in the context that the variations in the eclipse times are due to the light travel-time effect caused by the gravitational influence of a third and fourth body orbiting the central binary system. Kbin,(3,4) are the amplitudes of the eclipse time variations as a result of the light travel-time effect of the two bodies. υ(3,4) are the true anomalies of the two bodies and, where υ(3,4) = (E + T(3,4)) f(3,4), which progresses through 2π over the orbital periods [P(3,4)] and are functions of E. T(3,4) are the times of periastron passages and 0.3220.3240.3260.3280.3300.332BJD(+2455984)Arbitrary counts0.4320.4340.4360.4380.4400.442BJD(+2456957)Arbitray countsZ. N. Khangale et al.: High-speed photometry of the eclipsing polar UZ Fornacis Fig. 2. O - C diagram of UZ For from Paper I but with new eclipses added. The vertical green lines separate the eclipse times from literature (to the left) and our new eclipse times (to the right). (See text for more details). (3,4) are the longitudes of periastron passage measured from the ascending node in the plane of the sky. Lastly, e(3,4) and f(3,4) are the eccentricities and orbital frequencies of the two bodies. We next investigated using Eq. 1 on all of the eclipse times. However Eq. 1 has thirteen variables for minimization, the first three are associated with the quadratic term and the remaining ten are associated with the two elliptical terms. Therefore we generated a starting grid, for minimizations, of 10360 starting points for the two elliptical frequencies evenly spaced with f3 (between 0.0000552 and 0.000138) and f4 (between 0.0001794 and 0.0004555) cycles per binary cycle. The remaining eight of the ten elliptical parameters were randomized to within reason- able values e.g. the eccentricities took random values between 0 and 1. We then performed simultaneous least-square fitting for each starting set of parameters. During minimization all the parameters were allowed to vary and as expected all the mini- mizations did not converge to a single solution but instead gave a range of solutions with final reduced χ2 between 2.06 and 397.27. Rerunning the minimizations with a new grid of ran- dom values for the parameters give the same general results. We explore the solutions in the next two sub-sections. 3.4. Distribution of the elliptical frequencies The distribution of the reduced χ2 (not shown) forms an asym- metric Gaussian profile peaking at approximately five. The Gaussian profile peaks rapidly from a reduced χ2 of ∼3.5 to 5-7 and declines gradually after the maximum and flattens out be- yond the reduced χ2 of ten. About 11% of all the minimized so- lutions have reduced χ2 less than 6.5. These solutions are shown in Fig. 3 as a function of the two elliptical frequencies. The so- lutions have been colour-coded to show increasing values of the reduced χ2 from blue to red. The figure is an expanded view of the parameter space focused on the region where most of the so- lutions were concentrated. We added an additional constraint to exclude the parabolic and/or hyperbolic solutions, i.e. where ei- ther e3 or e4 are ≥ 1.0. The overall distribution of the reduced χ2 as a function of the elliptical frequencies form an arc with chi- square increasing from the lower right through the centre to the upper right regions of Fig. 3. The grey star towards the centre of Fig. 3 marks the location of the solution presented in Paper I and lies in the region where the intermediate solutions are found. 3.5. Distribution of the eccentricities Figure 4 shows the range of eccentricities for the solutions pre- sented in Fig. 3, where e3 and e4 are the eccentricities of the Article number, page 5 of 11 10050050100O-C (sec)1983.71995.220072018Residuals402002040O-C (sec)First Elliptical TermResiduals402002040O-C (sec)Second Elliptical TermResiduals100000800006000040000200000200004000060000Cycle Number402002040O-C (sec)ResidualsA&A proofs: manuscript no. aanda Fig. 3. Reduced χ2 parameter space for the two frequencies. The points have been colour-coded with dark blue and dark red representing lower and higher reduced chi-square: 2.06 ≤ χ2 ≤ 6.50. The solutions with e ≥ 1.0 were excluded since their solutions will give parabolic and hyperbolic orbits. The grey star marks the location of the solution adopted in Paper I, whereas the red circle and magenta cross mark the solution with the lowest reduced χ2 adopted as the best-fit for this paper. first and second elliptical terms, respectively. The solutions have been colour-coded to show increasing values of the reduced χ2 from blue to red. The solutions with the lowest reduced χ2 values (dark blue circles) form a vertical ridge (first ridge) centred on e3 between ∼0.5-0.7 with a wide range of e4 eccentricities between ∼0.0-0.9. A significant number of solutions form another verti- cal ridge (second ridge) centred on e3 ≈0.4 and spanning from 0 to 1 for e4. This is associated with the large concentration of light blue and orange to dark red solutions shown in the top right region of Fig. 3. The scattered solutions between the two vertical ridges correspond to the lower density of blue and red circles to- wards the centre of Fig. 3. The solutions to the upper left of the second vertical ridge (e3 < 0.3 and e4 > 0.2) are associated with the solutions with lower f4 frequencies and not shown in Fig. 3. Immediately obvious is the absence of solutions in the lower left quadrant corresponding to low values of e3 and e4. There is also a cut-off in the e3 eccentricities at 0.85. 3.6. O - C diagram Figure 5 shows the formal best O - C using the solution indicated by a magenta cross and enclosed with a red circle in both Figs 3 and 4. This is the solution with the lowest reduced χ2 (2.06) and is the best-fitting solution. However, there are many other solutions with reduced χ2 values comparable to this one and in- dicated by the blue circles in Fig. 4. The parameter values of the best solution are listed in Table 2. This particular solution has the e4 eccentricity of ∼0.45. Similar to the other best solutions, it has an eccentricity of e3 ∼0.69. We conclude that the dataset is Article number, page 6 of 11 not consistent with low eccentricities for the two elliptical terms. The best solutions at minimum require the e3 eccentricity to be ∼0.7. The residuals in Fig. 5 are suggestive that by adding a third or fourth elliptical term would reduce the χ2 further. However, the dataset is currently under-constrained to warrant adding ad- ditional terms. We note that in Paper I, they concluded that the best mathematical solution also required high eccentricities. 4. Discussion In Paper I, they detected departures in the eclipse times of UZ For from a simple quadratic ephemeris of up to ∼60 s. They found that the departures also suggest the presence of two ellipti- cal terms with periodicities of ∼16(3) and ∼5.25(25) years. Sim- ilarly, our new results suggest that the deviations in the eclipse O - C shown in Fig. 5 continue to be best described with a com- bination of a quadratic term plus two elliptical terms with pe- riodicities of ∼14.67(1) and ∼5.82(3) years. The deviations are suggestive of both secular and/or periodic variations. Generally, period changes in binary systems are understood as being due to gravitational radiation, magnetic braking, the Applegate mecha- nism, and the presence of circumbinary planets in orbit around the binary. In UZ For, Paper I, they reasoned that the two favoured mechanisms to derive the periodicities are either two giant ex- trasolar planets as companions to the binary or the Applegate mechanism (Applegate 1992) due to magnetic cycle activity of the secondary star. However, in Paper I, they argued that Apple- gate's mechanism would require the entire radiant energy output 1.601.651.701.751.801.851.901.952.00 f3 (104 cycles/day)4.64.85.05.25.45.65.8 f4 (104 cycles/day)2.06 < 2 < 6.50 0.0 < e3 < 0.86 0.0 < e4 < 1.02.53.03.54.04.55.05.56.06.5Z. N. Khangale et al.: High-speed photometry of the eclipsing polar UZ Fornacis Fig. 4. Reduced χ2 parameter space for the two elliptical eccentricities. The points have been colour-coded with dark blue and dark red representing lower and higher reduced chi-square: 2.06 ≤ χ2 ≤ 6.50. The solutions with e ≥ 1.0 were excluded since their solutions will give parabolic and hyperbolic orbits. The red circle and magenta cross mark the solution adopted as the best-fit for this paper. of the secondary and they ruled this mechanism out and recom- mended a revision such as those described by Lanza et al. (1998). They also argued that a two-planet model was problematic given the quality of the data and a high eccentric orbit for one of the two planets was required in order to capture most of the eclipse times. In light of our new results and recent work in the literature we next discuss the two favoured mechanisms. 4.1. Circumbinary planets Repeating the calculations of Paper I using the solution above, we used the amplitudes of oscillations (Kbin,(3,4)) to calculate the projected distances a sin (i) from the centre of mass of the bi- nary to the centre of mass of each of the triple systems. The centres of mass were 0.064(1) and 0.011(1) au for the long and short periods, respectively. Setting the binary mass to be 0.84 M(cid:74) (i.e. 0.7 M(cid:74) + 0.14 M(cid:74), the total combined mass of f (m3,4) = 1.2116(3) ×10−6 and 4.045(7) ×10−8 M(cid:74). We used the WD and red dwarf) gives the corresponding mass functions the binary inclination (i) of 80◦ and found the respective mini- 0.00955(2) M(cid:74) and would therefore qualify as extrasolar gi- mum masses of the third and fourth body to be 0.00307(5) and ant planets [3.22(5) and 10.01(2) MJ]. The quoted errors are 1σ errors propagated based on the one solution presented and ex- cludes the range in periods shown in Fig. 3. These parameters are summarized in Table 2. We note that these two-planet parameters are specific to this one best-fitting solution only. Nevertheless, almost all of the solutions with similar reduced χ2 gave similar values for the = -3.29(3)× 10−12 s s−1. This cor- quadratic parameter Pbin = 2A Pbin responds to a rate of angular momentum loss of ∼ 8.3× 1034 erg using equation (5) of Brinkworth et al. (2006). The expected the- oretical rate of angular momentum loss due to gravitational radi- ation (Equ. (12) of Brinkworth et al. 2006, corrected from Equ. (2) of Andronov et al. 2003) and magnetic braking (Equ. (A6) of Knigge et al. 2011 (based on the magnetic braking law by Ver- bunt & Zwaan 1981)) amounts to ∼ 8.9 × 1033 erg. Therefore the implied rate of angular momentum loss using the Pbin term only, is ten times larger than the rates of angular momentum loss predicted by gravitational radiation plus magnetic braking alone. Within errors, the periodic variations of the formal best so- lution could be consistent with two planets locked in a 3:1 res- onance with orbital periods of 14.67(1) and 5.82(3) years, re- spectively. However the high eccentricity for both planets im- plies that such a two planet solution would not be stable. There are solutions of similar reduced χ2 in which the inner planet has a low eccentricity. Nevertheless all the best solutions require a high eccentricity for the outer planet. Perhaps indicating a one planet solution (see below). We certainly need an independent observational approach to shed some light on the existence of planets around UZ For. One such technique may be astrometric monitoring of the precession of the UZ For binary as it wobbles back and forth due to ex- trasolar companions. The ongoing GAIA mission provides mi- cro arc-seconds parallaxes to thousands of object and place UZ For at a distance of ∼240 pc and it might be possible to detect its companion after five years of observations. A simulation by Go´zdziewski et al. (2015) suggests that it should be possible to detect ∼30 micro arc-seconds for a 7 MJ planet in a 5 au orbit. They argued that for HU Aqr, a polar which shows variations in Article number, page 7 of 11 0.00.20.40.60.81.0 e30.00.20.40.60.81.0 e42.06 < 2 < 6.50 0.0 < e3 < 0.86 0.0 < e4 < 1.02.53.03.54.04.55.05.56.06.5A&A proofs: manuscript no. aanda Fig. 5. Formal best O-C diagram of UZ For, based on new parameters shown in Table 2, after successive subtraction of the three terms comprising our new eclipse ephemeris. Top: O - C after subtraction of the linear term. Second: O - C after subtraction of the quadratic term with the first elliptical term overplotted (solid red curve). Third: O - C after subtraction of the first elliptical term with the second elliptical term overplotted (solid red curve). Bottom: the final O - C residuals after subtraction of the second elliptical term. The vertical green line separate the eclipse times from literature (to the left) and our new eclipse times (to the right). its O - C diagram and lies at a distance of ∼200 pc, it should be possible to detect the outermost companion using parallaxes. 4.2. Revised Applegate Significant residuals remains in the O - C of UZ For that points to something more complicated than just the presence of two companions to the binary. In Paper I, they noted that the Ap- plegate mechanism would require more than the radiant energy output of the secondary to drive the period changes but sug- gested that this mechanism is unlikely to be causing the period changes in UZ For. Völschow et al. (2016) revision of Apple- gate's mechanism which includes angular momentum exchange between the finite shell and the core of the secondary star place UZ For amongst four systems that could be explained by mag- netic activity of the secondary star. We conclude that both extra- solar planets and some form of Applegate's mechanism should be considered when explaining the O - C diagram of UZ For. Given the smaller amplitude of the residuals after subtraction of the first elliptical term, corresponding to one highly eccentric extrasolar planet, the remaining residuals could be explained by some sort of Applegate mechanism. Article number, page 8 of 11 4.3. Other eclipsing mCVs UZ For is not the only post-common-envelope binary which pre- sumably host single- or multiple planetary systems, e.g. HU Aqr (Qian et al. 2011; Horner et al. 2011; Wittenmyer et al. 2012; Go´zdziewski et al. 2015) and DP Leo (Qian et al. 2010; Beuer- mann et al. 2011), are other magnetic CVs to show this be- haviour. For HU Aqr, Qian et al. (2011) reported that eccentricity of the outer planet as big as 0.5, and assumed circular orbit for the inner planet. However, revision by Go´zdziewski et al. (2015) suggests eccentricities ranging from 0.1-0.3 for both the inner and the outer planets. They also suggest that for stable orbits to exist in the HU Aqr system there must be a third companion or- biting with a very long orbital period and with the middle planet in retrograde orbit. Bours et al. (2016) found that for the non-magnetic CV, NN Ser, the remaining model has eccentricities ranging from 0.1-0.2 and constrained the period ratio to 2:1 resonance. If both Ap- plegate's mechanism and circumbinary planets are causing the period change in UZ For, this will complicate the process of modelling the eclipse times, since we do not understand both these mechanisms well and it is especially difficult to model the effect of the magnetic activity cycle. We can not say for sure that 10050050100O-C (sec)1983.71995.220072018Residuals402002040O-C (sec)First Elliptical TermResiduals402002040O-C (sec)Second Elliptical TermResiduals100000800006000040000200000200004000060000Cycle Number402002040O-C (sec)ResidualsZ. N. Khangale et al.: High-speed photometry of the eclipsing polar UZ Fornacis alternative models such as those of Applegate (1992) or angu- lar momentum loss due to gravitational radiation and magnetic braking are not operating on UZ For. Bours et al. (2016) had suggested that some form of Applegate's mechanism might be at work on NN Ser given its M4 spectral type of the secondary. 5. Summary and conclusion In this paper, we presented and analysed new photometric obser- vations of UZ For together with historical observations collected from literature. We used the new observations to test the two- planet model proposed in Paper I to explain the variations in its eclipse times. UZ For undergoes a change in mass transfer from faint to high state, this is captured in Fig. 1. The light curves of UZ For show variations in eclipse profiles from one epoch to the next which is consistent with what has been reported in the liter- ature. The eclipse widths remain unchanged and various stages are revealed depending on the accretion state. We find that the new mid-eclipse times follow the general trend predicted in Paper I (Fig. 2) but continues to diverge. In order to accommodate most of the eclipse times, we have re- calculated the fitting parameters, including the new data, in the similar manner as in Paper I. We adopted one of the solutions with the lowest reduced χ2 value as the best fit solution. The parameters for this solution are shown in Table 2 and overplot- ted in the top and middle panels of Fig. 5. The proposed model of the two planets requires the outer planet to have a relatively high eccentric orbit, i.e. e3 = 0.69. Significant residuals remain as indicated in the bottom panel of Fig. 5. Adding more elliptical terms (effectively adding more planets) would obviously lower the reduced χ2 value, however the data is of insufficient quantity to warrant this. Within errors, the departures in the O - C diagram (Fig. 5) are still consistent with the two cyclic variations (14.67(1) and 5.82(3) years) reported in Paper I. However a relatively large ec- centricity is required for the longer period and, in addition, seem- ingly random residuals still remain. This suggests that either the circumbinary planet solution is incorrect or requires extra plan- ets, or some form of cyclic magnetic activity is contributing an extra quasi-periodic term to the O - C variations. Further moni- toring of the eclipse times is recommended. In the the next 5-10 years, the GAIA space mission may be able to detect parallax variations that would be consistent with circumbinary planet so- lutions as also been suggested by Go´zdziewski et al. (2015) for HU Aqr. 1989, ApJ, 337, 832 1118 Meeting 2011, 1173 Coppejans, R., Gulbis, A. A. S., Kotze, M. M., et al. 2013, PASP, 125, 976 Dai, Z.-B., Qian, S.-B., Fernández Lajús, E., & Baume, G. L. 2010, MNRAS, 409, 1195 de Bruijne, J. H., Reynolds, A. P., Perryman, M. A., Favata, F., & Peacock, A. J. 2002, Optical Engineering, 41, 1158 Eastman, J., Siverd, R., & Gaudi, B. S. 2010, PASP, 122, 935 Faulkner, J. 1971, ApJ, 170, L99 Ferrario, L., Wickramasinghe, D. T., Bailey, J., Tuohy, I. R., & Hough, J. H. Giommi, P., Angelini, L., Osborne, J., et al. 1987, IAU Circ., 4486, 1 Go´zdziewski, K., Słowikowska, A., Dimitrov, D., et al. 2015, MNRAS, 448, Gulbis, A. A. S., O'Donoghue, D., Fourie, P., et al. 2011, in EPSC-DPS Joint Hall, D. S. 1989, Space Sci. Rev., 50, 219 Hardy, A., Schreiber, M. R., Parsons, S. G., et al. 2016, MNRAS, 459, 4518 Hinse, T. C., Lee, J. W., Go´zdziewski, K., et al. 2012, MNRAS, 420, 3609 Horner, J., Marshall, J. P., Wittenmyer, R. A., & Tinney, C. G. 2011, MNRAS, Horner, J., Wittenmyer, R. A., Hinse, T. C., & Tinney, C. G. 2012, MNRAS, Imamura, J. N. & Steiman-Cameron, T. Y. 1998, ApJ, 501, 830 Knigge, C., Baraffe, I., & Patterson, J. 2011, ApJS, 194, 28 Lanza, A. F., Rodono, M., & Rosner, R. 1998, MNRAS, 296, 893 Marsh, T. R., Parsons, S. G., Bours, M. C. P., et al. 2014, MNRAS, 437, 475 Osborne, J. P., Giommi, P., Angelini, L., Tagliaferri, G., & Stella, L. 1988, ApJ, 416, L11 425, 749 328, L45 1161 2202 611, A48 513, L7 587, A34 Paczynski, B. & Sienkiewicz, R. 1981, ApJ, 248, L27 Perryman, M. A. C., Cropper, M., Ramsay, G., et al. 2001, MNRAS, 324, 899 Potter, S. B., Buckley, D. A. H., O'Donoghue, D., et al. 2010, MNRAS, 402, Potter, S. B., Romero-Colmenero, E., Ramsay, G., et al. 2011, MNRAS, 416, Pulley, D., Faillace, G., Smith, D., Watkins, A., & von Harrach, S. 2018, A&A, Qian, S.-B., Dai, Z.-B., Liao, W.-P., et al. 2009, ApJ, 706, L96 Qian, S.-B., Liao, W.-P., Zhu, L.-Y., & Dai, Z.-B. 2010, ApJ, 708, L66 Qian, S.-B., Liu, L., Liao, W.-P., et al. 2011, MNRAS, 414, L16 Ramsay, G. 1994, Information Bulletin on Variable Stars, 4075, 1 Rappaport, S., Verbunt, F., & Joss, P. C. 1983, ApJ, 275, 713 Schreiber, M. R., Gänsicke, B. T., Rebassa-Mansergas, A., et al. 2010, A&A, Schwope, A. D., Beuermann, K., & Thomas, H.-C. 1990, A&A, 230, 120 Sirk, M. M., Warren, J. K., Vallerga, J. V., & Christian, C. A. 1994, in BAAS, Vol. 26, Bulletin of the American Astronomical Society, 793 Still, M. & Mukai, K. 2001, ApJ, 562, L71 Stockman, H. S. & Schmidt, G. D. 1996, ApJ, 468, 883 Verbunt, F. & Zwaan, C. 1981, A&A, 100, L7 Völschow, M., Schleicher, D. R. G., Perdelwitz, V., & Banerjee, R. 2016, A&A, Warren, J. K., Sirk, M. M., & Vallerga, J. V. 1995, ApJ, 445, 909 Warren, J. K., Vallerga, J. V., Mauche, C. W., Mukai, K., & Siegmund, O. H. W. 1993, in Bulletin of the American Astronomical Society, Vol. 25, American Astronomical Society Meeting Abstracts #182, 861 Wittenmyer, R. A., Horner, J., Marshall, J. P., Butters, O. W., & Tinney, C. G. 2012, MNRAS, 419, 3258 Acknowledgements We would like to thank the anonymous referee whose com- ments were helpful and improved our manuscript. This material is based upon work supported financially by the National Re- search Foundation. References Allen, R. G., Berriman, G., Smith, P. S., & Schmidt, G. D. 1989, ApJ, 347, 426 Andronov, N., Pinsonneault, M., & Sills, A. 2003, ApJ, 582, 358 Applegate, J. H. 1992, ApJ, 385, 621 Bailey, J. & Cropper, M. 1991, MNRAS, 253, 27 Beuermann, K., Buhlmann, J., Diese, J., et al. 2011, A&A, 526, A53 Beuermann, K., Dreizler, S., & Hessman, F. V. 2013, A&A, 555, A133 Beuermann, K., Hessman, F. V., Dreizler, S., et al. 2010, A&A, 521, L60 Beuermann, K., Thomas, H.-C., & Schwope, A. 1988, A&A, 195, L15 Bours, M. C. P., Marsh, T. R., Parsons, S. G., et al. 2016, MNRAS, 460, 3873 Brinkworth, C. S., Marsh, T. R., Dhillon, V. S., & Knigge, C. 2006, MNRAS, 365, 287 Article number, page 9 of 11 A&A proofs: manuscript no. aanda Table 2. Mid-eclipse ephemeris of the main accretion spot of UZ For and corresponding planet model parameters. The ephemeris are rounded off to the 1σ errors. The planet parameter errors are 1σ errors and were propagated from the one fitting solution and may underestimate true errors of range in parameter space of possible solutions. The minimum planet masses are listed assuming coplanearity and M3,4, f nc is the mass function. The combined mass of the primary and secondary stars is assumed to be 0.84M(cid:12). (Table reproduced from -- Potter et al. 2011). Quadratic term T0 = 2453405.300833(5) d Pbin = 0.087865421(1) d A = -14.5(2)×10−14 1st elliptical term υ3 = (E + T3) f3 T3 = 67198(145) (binary cycle) f3 = 0.0001030(1) (cycles per binary cycle) 3 = 2.10(1) Kbin,(3) = 0.000371(3) d e3 = 0.69(1) 2nd elliptical term υ4 = (E + T4) f4 T4 = 7444(219) (binary cycle) f4 = 0.000260(1) (cycles per binary cycle) 4 = -0.22(5) Kbin,(4) = -0.000065(3) d e4 = 0.45(6) Planet Parameters M3, f nc = 1.326(7)10−6M(cid:12) M3,Jup = 10.00(2) P3 = 14.67(1) years a3 = 5.7(1) au a1,3 = 0.064(1) au M4, f nc = 3.43(9)10−8M(cid:12) M4,Jup = 3.22(5) P4 = 5.82(3) years a4 = 3.0(2) au a1,4 = 0.011(1) au Table 3. Mid-eclipse times of the main accretion spot of UZ For. BJDTDB is the BJD in the TDB system. The ingress and egress times have the integer of BJD subtracted. All the times have been barycentrically corrected. Tegress Reference Continued on next page Cycle BJDTDB+2400000 58171.29654185 54242 58074.55667742 53141 58074.46884132 53140 53139 58074.38097010 58054.43550982 52912 57707.36706944 48962 57704.37964299 48928 57270.58825176 43991 43593 57235.61779679 57103.29264038 42087 57077.28447068 41791 57072.36399146 41735 40450 56959.45690562 56959.36909531 40449 56957.43605536 40427 56955.41517125 40404 40393 56954.44862704 56953.39425098 40381 56719.32090166 37717 56670.37988451 37160 36648 56625.39279285 56569.59832086 36013 56369.26522176 33733 56368.29859146 33722 56365.31135803 33688 33313 56332.36177629 56279.37896176 32710 56126.66890920 30972 55984.32701723 29352 28023 55867.55393431 55866.41166271 28010 ∆BJDTDB Width (s) 472(3) 0.00003 471(3) 0.00003 472(2) 0.00003 0.00002 471(2) 471(4) 0.00004 471(2) 0.00003 472(2) 0.00002 471(3) 0.00003 0.00004 470(3) 472(3) 0.00003 472(4) 0.00004 472(3) 0.00004 0.00004 470(3) 471(1) 0.00002 471(1) 0.00002 472(2) 0.00003 0.00002 472(1) 471(1) 0.00002 470(3) 0.00006 471(1) 0.00002 0.00004 471(3) 475(5) 0.00006 472(3) 0.00004 471(1) 0.00004 471(2) 0.00004 0.00003 470(2) 472(3) 0.00004 470(2) 0.00004 470(1) 0.00004 0.00003 472(2) 471(2) 0.00002 Tingress 0.293811(20) 0.553950(20) 0.466109(20) 0.378244(10) 0.432808(50) 0.364344(20) 0.376910(10) 0.585528(20) 0.615071(30) 0.289908(30) 0.281738(30) 0.361262(30) 0.454187(50) 0.366371(20) 0.433331(30) 0.412439(30) 0.445895(20) 0.391527(40) 0.317920(80) 0.377161(20) 0.390068(30) 0.595571(40) 0.262493(50) 0.295865(20) 0.308634(60) 0.359056(20) 0.376250(30) 0.666182(60) 0.324295(40) 0.551204(50) 0.408936(40) Article number, page 10 of 11 0.299227(20) 0.559405(20) 0.471574(20) 0.383696(10) 0.438234(60) 0.369795(20) 0.382376(10) 0.590976(20) 0.620522(20) 0.295372(20) 0.287203(30) 0.366721(30) 0.459624(80) 0.371819(30) 0.438779(30) 0.417903(40) 0.451359(30) 0.396975(20) 0.323621(30) 0.382608(40) 0.395518(30) 0.601070(40) 0.267951(60) 0.301318(30) 0.314082(40) 0.364497(20) 0.381696(30) 0.671642(50) 0.329740(20) 0.556664(40) 0.414390(30) znk znk znk znk znk znk znk znk znk znk znk znk znk znk znk znk znk znk znk znk znk znk znk znk znk znk znk znk znk znk znk Z. N. Khangale et al.: High-speed photometry of the eclipsing polar UZ Fornacis Table 3 -- Continued from previous page ∆BJDTDB Width (s) 0.00004 469(3) 470(1) 0.00002 468(2) 0.00001 468(2) 0.00001 0.00001 467(2) 469(2) 0.00001 469(1) 0.0000086 468(1) 0.0000086 469(1) 0.0000086 0.00001 469(2) 469(3) 0.000035 467(4) 0.00006 479(8) 0.000087 0.000035 469(3) 469(6) 0.00007 467(2) 0.00001 468(2) 0.00002 468(2) 0.00002 0.00002 468(2) 0.00005 0.00005 0.00005 0.00005 0.00005 0.00005 0.00004 0.00003 0.00004 0.00003 0.0001 0.00003 0.00003 0.00003 0.00003 0.0002 0.0002 0.0002 0.0002 0.0002 0.0002 0.0002 0.0002 0.00016 0.00016 463(4) 477(5) 467(4) 471(4) 466.5(2.5) Cycle BJDTDB+2400000 55862.54558562 27966 55629.26295762 25311 55506.42703435 23913 55478.48583116 23595 23277 55450.54462082 54857.36480850 16526 54857.36480517 16526 54417.33472170 11518 53408.28808581 34 23 53407.32157438 53405.30066303 0 53404.33404192 -11 52494.83919610 -10362 -10365 52494.57562568 52493.60905802 -10376 51821.70239393 -18023 51528.49543399 -21360 51528.40757990 -21361 -21429 51522.43272958 50021.77938800 -38508 50018.70410800 -38543 49755.63497800 -41537 -41538 49755.54714800 49753.61402800 -41560 49752.64756800 -41571 49733.40501704 -41790 49310.33259382 -46605 -46988 49276.68005500 48784.72141928 -52587 48482.72808573 -56024 47829.18486375 -63462 -63474 47828.13052000 47827.95478000 -63476 47437.91992000 -67915 47145.06433900 -71248 47127.22773900 -71451 -71452 47127.13943900 47097.79255900 -71786 47094.71735900 -71821 47091.55423900 -71857 -71868 47090.58778900 47088.74254900 -71889 46446.97380900 -79193 -89206 45567.17759700 Tingress 0.542871(60) 0.260238(20) Tegress 0.548301(60) 0.265677(30) Reference znk znk a a a a a a a a a a a a a b c c c d d d d d d a a e a f g g g f,h i i i j j j j j k k aPotter et al. (2011); bde Bruijne et al. (2002); cPerryman et al. (2001); dImamura & Steiman-Cameron (1998); eWarren et al. (1995); f Ramsay (1994); gBailey & Cropper (1991); hAllen et al. (1989); iFerrario et al. (1989); jBeuermann et al. (1988); kOsborne et al. (1988); znkThis Paper. Article number, page 11 of 11
1211.2095	3	1211	2013-01-09T11:06:35	Gravoturbulent Planetesimal Formation: The Positive Effect of long-lived Zonal Flows	[ "astro-ph.EP" ]	Recent numerical simulations have shown long-lived axisymmetric sub- and super-Keplerian flows in protoplanetary disks. These zonal flows are found in local as well as global simulations of disks unstable to the magnetorotational instability. This paper covers our study of the strength and lifetime of zonal flows and the resulting long-lived gas over- and underdensities as functions of the azimuthal and radial size of the local shearing box. We further investigate dust particle concentrations without feedback on the gas and without self-gravity. Strength and lifetime of zonal flows increase with the radial extent of the simulation box, but decrease with the azimuthal box size. Our simulations support earlier results that zonal flows have a natural radial length scale of 5-7 gas pressure scale heights. This is the first study that combines three-dimensional MHD simulations of zonal flows and dust particles feeling the gas pressure. The pressure bumps trap particles with $\textrm{St} = 1$ very efficiently. We show that $\textrm{St} = 0.1$ particles (of some centimeters in size if at $5\textrm{AU}$ in an MMSN) reach a hundred-fold higher density than initially. This opens the path for particles of $\textrm{St} = 0.1$ and dust-to-gas ratio of 0.01 or for particles of $\textrm{St} \geq 0.5$ and dust-to-gas ratio $10^{-4}$ to still reach densities that potentially trigger the streaming instability and thus gravoturbulent formation of planetesimals.	astro-ph.EP	astro-ph	Draft version April 3, 2018 Preprint typeset using LATEX style emulateapj v. 5/2/11 3 1 0 2 n a J 9 . ] P E h p - o r t s a [ 3 v 5 9 0 2 . 1 1 2 1 : v i X r a GRAVOTURBULENT PLANETESIMAL FORMATION: THE POSITIVE EFFECT OF LONG-LIVED ZONAL FLOWS K. Dittrich1, H. Klahr1, and A. Johansen2 1Max-Planck-Institut fur Astronomie, Konigstuhl 17, D-69117 Heidelberg, Germany and 2Lund Observatory, Department of Astronomy and Theoretical Physics, Box 43, SE-22100 Lund, Sweden Draft version April 3, 2018 ABSTRACT Recent numerical simulations have shown long-lived axisymmetric sub- and super-Keplerian flows in protoplanetary disks. These zonal flows are found in local as well as global simulations of disks unstable to the magnetorotational instability. This paper covers our study of the strength and lifetime of zonal flows and the resulting long-lived gas over- and underdensities as functions of the azimuthal and radial size of the local shearing box. We further investigate dust particle concentrations without feedback on the gas and without self-gravity. Strength and lifetime of zonal flows increases with the radial extent of the simulation box, but decreases with the azimuthal box size. Our simulations support earlier results that zonal flows have a natural radial length scale of 5 to 7 gas pressure scale heights. This is the first study that combines three-dimensional MHD simulations of zonal flows and dust particles feeling the gas pressure. The pressure bumps trap particles with St = 1 very efficiently. We show that St = 0.1 particles (of some centimeters in size if at 5 AU in an MMSN) reach a hundred- fold higher density than initially. This opens the path for particles of St = 0.1 and dust-to-gas ratio of 0.01 or for particles of St ≥ 0.5 and dust-to-gas ratio 10−4 to still reach densities that potentially trigger the streaming instability and thus gravoturbulent formation of planetesimals. Subject headings: magnetohydrodynamics (MHD) - planets and satellites: formation - protoplanetary disks 1. INTRODUCTION Planets form as a side product in star formation. The general understanding on how planets in our solar sys- tem form was detailed in Safronov (1969). Low-mass stars form out of molecular clouds which consist of 99% hydrogen and helium (further referred to as gas), and 1% dust and ices (Lodders 2003), i.e., everything that has a higher complexity than hydrogen molecules or he- lium. Those molecular clouds have cores between less than 0.1M⊙and more than 10M⊙(Krumholz et al. 2012) that are gravitationally unstable. Most parts of the mass will collapse into a newborn star. The remaining ∼ 1% of the total mass will form an accretion disk with pressure supported, sub-Keplerian gas (Weidenschilling 1977a; Cassen & Moosman 1981) around the young star. Those disks have lifetimes in the order of a few million years (Haisch et al. 2001; Fedele et al. 2010). Dust par- ticles grow due to coagulation (Weidenschilling 1997). However, coagulation models show that there are sev- eral barriers to overcome to grow dust large enough to become gravitationally bound in kilometer-sized plan- etesimals, such as the bouncing barrier (Zsom et al. 2010; Windmark et al. 2012a,b), the fragmentation bar- rier (e.g., Beitz et al. 2011; Birnstiel et al. 2012, and ref- erences therein), and the kilometer-size barrier (Ida et al. 2008; Cuzzi et al. 2008). Dust growth mechanisms are summarized in Dominik et al. (2007) and the review of Blum & Wurm (2008) gives an overview on the men- tioned barriers. This paper addresses the fragmentation barrier or meter-size barrier. Pebbles of several decimeters in size will drift very fast inward due to the headwind from the [email protected] sub-Keplerian gas (Weidenschilling 1977a). Thus, dust has to grow very quickly from some centimeters to sev- eral kilometers in size in order to avoid drifting into the inner region of the protoplanetary disk. Turbulence in protoplanetary disks around young stars provides promising mechanisms for rapid planetes- imal formation (Johansen et al. 2007, 2011). Shearing box simulations (Brandenburg et al. 1995) are a pow- erful tool for analyzing the magnetorotational instabil- ity (MRI; Balbus & Hawley 1991, 1998) as a source of turbulence. These simulations consider a local, coro- tating box, representing a small part of a Keplerian disk. Johansen et al. (2009a) reported long-lived ax- isymmetric sub- and super-Keplerian flows, zonal flows, in shearing box simulations of turbulence caused by the MRI. These zonal flows have been seen in several other local (Fromang & Stone 2009; Stone & Gardiner 2010; Simon et al. 2012) and global (Lyra et al. 2008; Dzyurkevich et al. 2010; Uribe et al. 2011; Flock et al. 2011, 2012) simulations using a wide variety of codes. Zonal flows are a product of large-scale variations in the magnetic field that transport momentum differ- entially, creating regions of slightly faster and slightly slower rotating gas. Large-scale pressure bumps are ex- cited through geostrophic balance. This creates long- lived over-densities that potentially trap dust particles. A more thorough description of zonal flows and their cre- ation put forward in Johansen et al. (2009a) found zonal flows always populating the largest radial mode avail- able in the local box approximation. Their largest box was simulating 10.56 pressure scale heights (H). More recently Simon et al. (2012) found a more complex struc- ture in their largest simulation with Lx = 16H. They further studied the autocorrelation function (Guan et al. 2 Dittrich, Klahr, & Johansen 2009) of the magnetic field and the gas density. Both have a two-component structure. The first is tilted with respect to the azimuthal axis and highly localized. The second component is seen at the largest scales and can be associated with the (predominantly toroidal) back- ground magnetic field. Simon et al. (2012) measure the radial length scale of the zonal flows to converge at 6H. In this paper we consider even larger physical extents for zonal flow structures. This gives us the opportunity to measure physical properties such as size and lifetime independent of the simulated domain. Further, we in- vestigate properties of the zonal flows in radially and azimuthally stretched boxes. We alter the radial and az- imuthal domain up to ∼20 gas pressure scale heights. Additionally, we study the behavior of dust in zonal flows. Whipple (1972) was the first to suggest that ax- isymmetric pressure bumps can trap gas. Pinilla et al. (2012) invoked zonal flows as a possibility to explain the submillimeter and millimeter-sized particles observed in protoplanetary disks. They used artificial static density bumps introduced as sinusoidal density perturbations with different amplitudes (e.g., A = 0.1 and A = 0.3) and different wavelengths (L = 0.3 . . . 3H). They found that a 30% density perturbation (with L = 1H) is neces- sary to stop the drift of the dust grains. The present work is the first three-dimensional MHD study that combines zonal flows and the reaction of dust particles on them. Our paper is organized as follows. In Section 2 we discuss the setup of the simulations in this paper. In Section 3 we study the zonal flow properties and their dependency on the physical box size. The behavior of dust particles in zonal flows is described in Section 4. A discussion and conclusions follows in Section 5 and Section 6 provides a summary and an outlook. 2. SIMULATION SETUP We use the Pencil Code,1 a sixth-order spatial and third-order temporal finite difference code, for our simu- lations. We simulated the standard ideal MHD equations in a local shearing box with vertical stratification. The simulation boxes are centered at an arbitrary distance r to the star. The radial direction is denoted by x, the azimuthal direction by y, and the vertical direction by z. The Keplerian frequency is Ω. We include dust particle dynamics, without back-reaction to the gas and without self-gravity. 2.1. Gas Dynamics The gas velocity u relative to the Keplerian shear is evolved via the equation of motion ∂u ∂y = y Ωux y + Ω2z z ∂u ∂t + (u · ∇) u + u(0) 2 Ωuy x − + 1 2 J × B − 1 ρ 1 ρ ∇P + f ν (u, ρ) . (1) On the left-hand side of the equation, the second and third terms are the advection terms by the perturbed velocity and by shear flow, respectively. The right-hand side contains the Coriolis force, the vertical component of the stellar gravity, the Lorentz force, the pressure gradi- ent, and the viscosity term. Here, u(0) y = −(3/2)Ωx is the Keplerian orbital velocity. The magnetic field B as well as the current density J are calculated from the vector potential A using B = ∇×A and J = µ−1 ∇×(∇ × A), respectively. Here, µ0 is the vacuum permeability. The viscosity term f ν is explained in Section 2.2.1. We evolve the magnetic potential with the uncurled 0 induction equation ∂A ∂t + u(0) y ∂A ∂y = u × B + 3 2 ΩAy x + f η (A) . (2) The terms on the right-hand side express the electromo- tive force, the stretching (creation of azimuthal magnetic field from radial field) by Keplerian shear and the resis- tivity f η (see Section 2.2.2). The gas density is evolved with the continuity equation ∂ρ ∂t + (u · ∇) ρ + u(0) y ∂ρ ∂y = −ρ∇ · u + fD (ρ) , (3) where the last term on the right-hand side describes mass diffusion (see Section 2.2.3). We use an isother- mal equation of state P = c2 s ρ, where the speed of sound is cs = HΩ; H is the gas pressure scale height. 2.2. Dissipation Maxwell and Reynolds stresses as well as the MRI re- lease kinetic and magnetic energy at large scales. This energy cascades down to small scales. Since numerical simulations have a finite resolution, this small-scale en- ergy needs to be dissipated. We use numerical dissipation in the form of hyper- and shock viscosity (Section 2.2.1), hyper-resistivity (Section 2.2.2), and hyper- and shock diffusion (Section 2.2.3). 2.2.1. Viscosity The viscosity term f ν in Equation (1) is expressed by f ν = ν3h∇6u +(cid:16)S(3) · ∇ ln ρ(cid:17)i + νsh [∇∇ · u + (∇ · u) (∇ · ln ρ)] + (∇νsh) ∇ · u . (4) We restricted our models to hyper- (ν3) and shock (νsh) viscosity. Thus, the regular Navier-Stokes viscosity term is neglected. The third-order rate-of-strain tensor S(3) is defined by S(3) ij = ∂5ui ∂x5 j . (5) The high-order Laplacian ∇6 in Equation (4) is ex- panded as ∇6 = ∂6/∂x6 + ∂6/∂y6 + ∂6/∂z6. Further- more, the shock viscosity is expressed by νsh = csh(cid:10)max [−∇ · u]+(cid:11) min (δx, δy, δz)2 . In the fashion of von Neumann & Richtmyer (1950) it is proportional to positive2 flow convergence. We take the (6) 1 Details on the Pencil Code and download information can be 2 Symbolized by the plus sign in Equation (6). We only apply found at http://www.nordita.org/software/pencil-code/. shock viscosity where the velocity flow is converging. Planetesimal Formation in Zonal Flows 3 maximum over five zones, and smoothed it to the sec- ond order. As suggested by von Neumann & Richtmyer (1950), we set the shock viscosity coefficient to csh = 1.0 to dissipate energy in shocks at high z above the mid- plane of the disk. 2.2.2. Resitivity The effects of resistivity are captured by the term (7) (8) f η = η3∇6A , where η3 is the hyper-resistivity. 2.2.3. Diffusion Mass diffusion is computed with fD = D3∇6ρ + Dsh∇2ρ + ∇Dsh · ∇ρ , where D3 is the hyper-diffusion parameter and Dsh is expanded as in Equation (6). 2.3. Dust Dynamics Dust particles are simulated as individual super- particles i with position xi and velocity vi. Each super- particle position is evolved with dx(i) dt = v(i) + u(0) y y . (9) The change of velocity for each particle is evolved through dv(i) dt = 2Ωv(i) y x − 1 2 Ωv(i) x y − Ω2z z 1 − τf hv(i) − u(x(i))i , (10) where the first and second terms are due to the Coriolis force. The third term corresponds to the vertical gravity of the star. Particles only feel the gas drag (the last term in Equation (10)) of nearby cells, but are not subjected to pressure or Lorentz forces. τf denotes the friction time, a measure for the size of the particles. measures are in units of the pressure scale height H = csΩ−1. Density is stated in units of the initial mid-plane gas density ρ0. Magnetic field strength is measured in units of cs(µ0ρ0)−1. Energy and stress are in units of the mean thermal pressure in the box hPi = c2 Since our simulations are dimensionless, they can be placed at any distance r to the star. Only by defining a global pressure gradient ∂Pglobal/∂r, which balances the Coriolis force in s hρi. 1 ρ ∂Pglobal ∂r = 2Ω∆v , (11) we restrict our simulations to a specific distance to the star where the chosen pressure gradient applies. The pa- rameter ∆v = u(0) y − uy is the difference to the azimuthal Keplerian velocity. We fix ∆v = 0.05cs (see also Sec- tion 2.6). Numerically, the global pressure gradient acts as an external force on gas and dust. 2.6. Initial Conditions The gas density is set to an isothermal hydrostatic equilibrium ρ(z) = ρ0 exp (−z2/2H 2). We start with random noise fluctuations in the gas velocity with δu = 10−3cs. The azimuthal component of the magnetic vector potential is initialized with Ay = A0 cos (kxx) cos (kyy) cos (kzz) where throughout kx = ky = kz = 4.76H −1 and A0 = 0.04cs(µ0ρ0)−1. Particles are released after the gas turbulence is satu- rated. We measured this to be after 20Torb for the largest runs. For convenience, we used the same saturation time for all our simulations. Particles have a Stokes number of St = τfΩ = 1, unless otherwise stated. The initial par- ticle distribution is Gaussian in z and uniform in x and y. The particle velocity is initialized with the station- ary solution (Nakagawa et al. 1986) for the radial and azimuthal velocity vx cs vy cs =− =− 2∆v τfΩ + (τfΩ)−1 ∆v 1 + (τfΩ)2 . (12) 2.4. Boundary Conditions We get ∆v from the solution of Equation (11) For our simulations, we use shearing box boundary conditions in radial (shear-periodic) and azimuthal (pe- riodic) directions. In the vertical direction we also use periodic boundary conditions. Although periodic bound- ary conditions in vertical direction are not physical, these boundary conditions conserve the average flux of the magnetic field. Simulations with outflow boundaries (not included in this paper) showed no considerable mass flux across the vertical boundary and did not change the av- erage properties of the zonal flow. 2.5. Dimensions We use the dimensionless unit system cs = Ω = µ0 = ρ0 = 1. Velocity is measured in units of the local sound speed cs. Gas velocities are always denoted by u whereas particle velocities are always denoted by v. All velocities are differences to the Keplerian orbital velocity vK = (0, u(0) y = −(3/2)Ωx. Time is measured in units of the local orbital time Torb = 2πΩ−1. Length y , 0), where u(0) ∆v cs = − 1 2(cid:18) H r (cid:19)2 ∂ ln P ∂ ln r . (13) We initialized ∆v = 0.05cs for our simulations. 2.7. Simulation Parameters set to Lz = 2.64H.3 The parameter space covered by our simulations The vertical extent is summarized in Figure 1. One simula- is always tion set (A) covers the boxes with a squared base, i.e., radial and azimuthal extent are kept the same: Lx = Ly = {1.32, 2.64, 5.28, 10.56, 21.12} H. These are marked with blue boxes in Figure 1 and are called runs S, M, L, XL, and XXL. The devia- tion to the global density profile in the largest box 3 L = 1.32 has been chosen as the basic box size, because Lx = 1.32 approximately marks the transition from subsonic to supersonic Keplerian shear flow (Johansen et al. 2009a). 4 Dittrich, Klahr, & Johansen Simulation Set Run Lx × Ly × Lz Nx × Ny × Nz ν3 = η3 = D3 Table 1 Run Parameters (2) (3) (4) (5) nparticles (6) 62,500 250,000 1,000,000 4,000,000 4,000,000 125,000 500,000 1,000,000 125,000 500,000 1,000,000 St (7) 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1,200,000 120,000,000 14,000,000 6,000,000 0.01 . . . 100 0.01 . . . 100 0.01 . . . 1.0 1.0 . . . 100 Shear ∆t (8) (9) FDA 121 FDA 121 FDA 121 FDA 121 FDA 121 FDA 121 FDA 121 FDA 121 FDA 121 FDA 121 FDA 121 FDA 121 FDA 121 FDA 121 FDA 223 (1) A A,B,C A,E A,E A B B B C C C D D D D E E S M L XL XXL x-S x-L x-XL y-S y-L y-XL LspecMR LspecHR LspecMRs MspecMRb 1.32 × 1.32 × 2.64 2.64 × 2.64 × 2.64 5.28 × 5.28 × 2.64 10.56 × 10.56 × 2.64 21.12 × 21.12 × 2.64 1.32 × 2.64 × 2.64 5.28 × 2.64 × 2.64 10.56 × 2.64 × 2.64 2.64 × 1.32 × 2.64 2.64 × 5.28 × 2.64 2.64 × 10.56 × 2.64 5.28 × 5.28 × 2.64 5.28 × 5.28 × 2.64 5.28 × 5.28 × 2.64 2.64 × 2.64 × 2.64 36 × 36 × 72 72 × 72 × 72 144 × 144 × 72 288 × 288 × 72 576 × 576 × 72 36 × 72 × 72 144 × 72 × 72 288 × 72 × 72 72 × 36 × 72 72 × 144 × 72 72 × 288 × 72 144 × 144 × 72 256 × 256 × 128 144 × 144 × 72 72 × 72 × 72 4.0 × 10−10 4.0 × 10−10 4.0 × 10−10 4.0 × 10−10 4.0 × 10−10 4.0 × 10−10 4.0 × 10−10 4.0 × 10−10 4.0 × 10−10 4.0 × 10−10 4.0 × 10−10 4.0 × 10−10 2.0 × 10−11 4.0 × 10−10 4.0 × 10−10 4.0 × 10−10 4.0 × 10−10 121 121 Notes. Column 1: simulation set. Column 2: name of run. Column 3: box size in units of pressure scale heights. Column 4: grid 144 × 144 × 72 288 × 288 × 72 L SAFI XL SAFI 5.28 × 5.28 × 2.64 10.56 × 10.56 × 2.64 SAFI SAFI 1,000,000 4,000,000 1.0 1.0 resolution. Column 5: dissipation coefficients. Column 6: number of particles in simulation. Column 7: Stokes number St = τfΩ. Column 8: shear advection scheme. Column 9: total run time in orbits Torb. XXL 576×576×72 Np=4×106 21.12 10.56 H / y L 5.28 2.64 1.32 x−S 36×72×72 Np=125,000 S 36×36×72 Np=62,500 XL 288×288×72 Np=4×106 x−XL 288×72×72 Np=106 L 144×144×72 Np=106 x−L 144×72×72 Np=5×105 y−XL 72×288×72 Np=106 y−L 72×144×72 Np=5×105 M 72×72×72 Np=250,000 y−S 72×36×72 Np=125,000 1.32 2.64 5.28 Lx/H 10.56 21.12 Figure 1. Parameter space of radial and azimuthal box sizes that was simulated for this paper. Every simulation has a vertical extent of 2.64H. The first line in each box states the name of the run, the second line the number of grid cells used, and the third line gives the number of simulated super-particles. More details on all simulations are found in Table 1. can be quite severe at the inner and outer bound- ary of the largest simulation. Thus, the results from run XXL have to be treated with caution. Another set of simulations (B ) varies the radial size of the box, Lx = {1.32, 2.64, 5.28, 10.56} H, with constant box size in azimuthal direction, Ly = 2.64H. This set is marked red in Figure 1 and includes runs x-S, M, x-L, and x-XL. The third set of simulations (C ) varies the azimuthal extent, Ly = {1.32, 2.64, 5.28, 10.56} H, while the radial extent is kept constant, Lx = 2.64H. This set includes runs y-S, M, y-L, and y-XL (marked yellow in Figure 1). All simulations are stratified and have dust particles with different couplings to the gas. The simulations displayed in Figure 1 have particles with a Stokes number of St = 1. Details on run parameters of those and six more sim- ulations are found in Table 1. The first set of simula- tions (A,B, and C ) in Table 1 are the simulations with medium resolution, i.e., 36 grid cells4 per 1.32 pressure scale heights. Simulation set D was carried out to in- vestigate the behavior of different particle sizes in the presence of zonal flows. Run LspecMR is very much like run L, but with 12 different particle Stokes numbers. The run LspecHR has a resolution of 64 grid cells per 1.32H. Runs LspecMR and LspecHR have 12 different parti- cle species, with Stokes numbers of St = 0.01 . . . 100.0. The runs LspecMRs and MspecMRb have particles with Stokes numbers of St = 0.01 . . . 1.0 and St = 1.0 . . . 100.0, respectively. These two simulations were carried out to study particle behavior with more particles per grid cell5 at medium resolution. The corresponding sizes for differ- ent protoplanetary disk models are found in Section 5.2. Simulation set E is a comparison of runs L and XL to the same runs (L SAFI and XL SAFI ) with the Shear Advection by Fourier Interpolation (SAFI) scheme. Here, all variables q(x, y, z) are transformed into Fourier space in the y-direction to get q(x, ky, z). Then each Fourier mode is multiplied by exp [ikyu(0) y (x)δt] to shift by u(0) y (x)δt in real space and is inverse Fourier transformed to real space. This method reduces the ad- vection error to the standard Finite Difference Advection (FDA) scheme in the Pencil Code (more details on FDA and SAFI are found in Johansen et al. 2009a). 4 We chose 36 grid cells instead of the usual 32 grid cells. That choice was done due to the architecture (12 CPUs per node) of the used cluster, THEO in the MPG computing center in Garching. 5 ∼ 9 and ∼ 16 particles per grid cell for runs LspecMRs and MspecMRb compared to ∼ 0.8 particles per grid cell for run LspecMR. Planetesimal Formation in Zonal Flows 5 Table 2 Turbulence Properties Run (1) h 1 2 u2 xi (2) h 1 2 u2 yi (3) h 1 2 u2 zi (4) h 1 2 B2 xi (5) h 1 2 B2 yi (6) h 1 2 B2 z i (7) hρuxuyi h−BxByi (8) (9) α (10) S M L XL XXL x-S x-L x-XL y-S y-L y-XL LspecMR LspecHR LspecMRs MspecMRb 2.1 × 10−3 3.9 × 10−3 5.0 × 10−3 5.2 × 10−3 5.1 × 10−3 4.0 × 10−3 3.8 × 10−3 3.8 × 10−3 2.3 × 10−3 5.2 × 10−3 5.4 × 10−3 4.8 × 10−3 3.0 × 10−3 5.3 × 10−3 4.3 × 10−3 5.1 × 10−3 5.4 × 10−3 2.8 × 10−3 5.5 × 10−3 5.9 × 10−3 5.2 × 10−3 4.9 × 10−3 5.8 × 10−3 5.1 × 10−3 5.2 × 10−3 3.1 × 10−3 6.3 × 10−3 5.6 × 10−3 5.6 × 10−3 3.5 × 10−3 6.5 × 10−3 6.0 × 10−3 6.1 × 10−3 L SAFI 5.6 × 10−3 XL SAFI Notes. Column 1: name of run. Columns 2-4: kinetic energy. Columns 5-7: magnetic energy. Column 8: Reynolds stress. Column 9: 3.3 × 10−3 5.2 × 10−3 5.6 × 10−3 5.0 × 10−3 4.6 × 10−3 5.2 × 10−3 5.2 × 10−3 5.0 × 10−3 3.7 × 10−3 5.7 × 10−3 5.0 × 10−3 5.3 × 10−3 4.2 × 10−3 6.2 × 10−3 5.7 × 10−3 5.8 × 10−3 5.3 × 10−3 3.5 × 10−4 7.6 × 10−4 8.0 × 10−4 6.7 × 10−4 6.2 × 10−4 8.4 × 10−4 6.9 × 10−4 7.1 × 10−4 4.1 × 10−4 8.8 × 10−4 7.2 × 10−4 7.5 × 10−4 5.4 × 10−4 9.2 × 10−4 8.7 × 10−4 8.4 × 10−4 7.3 × 10−4 8.8 × 10−4 1.9 × 10−3 2.0 × 10−3 1.7 × 10−3 1.6 × 10−3 2.0 × 10−3 1.7 × 10−3 1.8 × 10−3 1.0 × 10−3 2.2 × 10−3 1.8 × 10−3 1.9 × 10−3 1.3 × 10−3 2.2 × 10−3 2.1 × 10−3 2.1 × 10−3 1.8 × 10−3 1.5 × 10−3 2.2 × 10−3 2.3 × 10−3 2.1 × 10−3 2.0 × 10−3 2.4 × 10−3 2.1 × 10−3 2.2 × 10−3 1.6 × 10−3 2.5 × 10−3 2.2 × 10−3 2.2 × 10−3 1.4 × 10−3 2.5 × 10−3 2.5 × 10−3 2.4 × 10−3 2.2 × 10−3 7.7 × 10−4 1.6 × 10−3 1.9 × 10−3 1.8 × 10−3 1.7 × 10−3 1.7 × 10−3 1.5 × 10−3 1.5 × 10−3 8.6 × 10−4 2.0 × 10−3 1.9 × 10−3 1.8 × 10−3 1.0 × 10−3 2.1 × 10−3 1.8 × 10−3 2.0 × 10−3 1.9 × 10−3 3.4 × 10−3 6.6 × 10−3 6.9 × 10−3 6.1 × 10−3 5.7 × 10−3 7.0 × 10−3 6.1 × 10−3 6.2 × 10−3 3.8 × 10−3 7.5 × 10−3 6.4 × 10−3 6.6 × 10−3 4.2 × 10−3 7.7 × 10−3 7.2 × 10−3 7.2 × 10−3 6.5 × 10−3 6.4 × 10−3 1.2 × 10−2 1.3 × 10−2 1.1 × 10−2 1.0 × 10−2 1.3 × 10−2 1.1 × 10−2 1.2 × 10−2 7.1 × 10−3 1.4 × 10−2 1.2 × 10−2 1.2 × 10−2 7.5 × 10−3 1.4 × 10−2 1.4 × 10−2 1.3 × 10−2 1.2 × 10−2 Maxwell stress. Column 10: α-value, following Equation (14). Stresses and energies have been normalized to the mean thermal pressure in the box, hP i = c2 s hρi. Every simulation is run for 121 local orbits Torb = 2πΩ−1, except for LspecMRb which runs for 223Torb in order to follow the evolution of the slowly settling large particles. After 20Torb, when the initial conditions are sufficiently forgotten and the turbulence saturated, the particles are started. 3. ZONAL FLOW PROPERTIES Turbulence properties are summarized in Table 2. The kinetic and magnetic energy as well as the Reynolds and Maxwell stress almost doubles when increasing the box size from (1.32H)2×2.64H (run S ) to (2.64H)3 (run M ). Further increasing the box size does not change the re- sulting energies and stresses by much. The radially short box of run x-S with 1.32H × (2.64H)2 has similar results on these values. However, the azimuthally short box of run y-S has turbulent energies and stresses comparable to run S. These measurements show that the turbulence parameters are saturated for boxes with an azimuthal extent of at least 2.64H. This confirms the results from Fromang & Stone (2009) who found that the turbulence properties do not change when the box size is increased radially, if the azimuthal dimension is large enough. The α-value (Shakura & Sunyaev 1973) in Column 10 in Ta- ble 2 is calculated via α = 2 3 (hρuxuyi − hBxByi) hPi , (14) where hPi = c2 s hρi. The factor of 2/3 originates from the shear parameter q = −d ln Ω/d ln R. We use q = 3/2, appropriate for a Keplerian disk. For further details see Brandenburg et al. (1995, page 748). The Maxwell stress is around three times higher than the Reynolds stress and thus dominates the α-value. In order to verify that our numerical resolution is suf- ficient, we examined the quality factor as described in Simon et al. (2012): Qj = 2πva,j Ω∆xj , (15) where the Alfv´en speed is defined as va,j2 = hBji2/hρi. The notation hxi denotes volume averaging, x shows a time average. Sorathia et al. (2012) show that Qz & 10 − 15 for poorly resolved azimuthal quality factors (Qy ∼ 10) is required to resolve the MRI. Larger val- ues of the azimuthal quality factor (Qy & 25) allow for lower vertical quality factors. The azimuthal component of the magnetic field is very well resolved (Qy & 25) for all simulations, but runs S and y-S. The vertical com- ponent has values between 6 and 8. We thus conclude that all simulations, but runs S and y-S have sufficient resolution for the MRI. In Figure 2 a snapshot of the runs y-XL, XL, M, and x-XL are shown in scale, giving a real size comparison of high- and low-pressure regions. The large-scale sinu- soidal form of the dominant mode is observable in these plots. The higher modes are much shorter lived and seem to be non-axisymmetric density waves affected by the shear (Heinemann & Papaloizou 2009). The amplitudes of the pressure differences are higher in azimuthally large boxes. Only the axisymmetric density waves are long- lived and strong enough to make up a significant contri- bution to the pressure bump structure in an azimuthal as well as a temporal average over some local orbits. The azimuthal average of Figure 2 is seen in Figure 3. Here, the axisymmetric structure is clearly visible and a strong correlation between the particle location and a positive radial gradient of the gas density is seen. The black dots in Figures 2 and 3 show the radial and azimuthal posi- tion of every 100th particle. The particles are trapped by the axisymmetric pressure bumps. Also, the shapes of spiral density waves (Heinemann & Papaloizou 2009) can be seen in the structures. The particle distribution with respect to the gas flow will be discussed in more 6 Dittrich, Klahr, & Johansen y−XL XL t= 85.0Torb 1.30 5.3 3.9 2.6 1.3 H / y 0.0 −1.3 −2.6 −3.9 −5.3 1.3 1.24 1.18 1.12 1.06 ) 0 = t ( Σ 1.00 / Σ 0.94 0.88 0.82 0.76 0.70 M x−XL H / y 0.0 −1.3 −1.3 0.0 x/H 1.3 −5.3 −3.9 −2.6 −1.3 0.0 x/H 1.3 2.6 3.9 5.3 Figure 2. Collage of four gas surface density representations of the runs y-XL, XL, M, and x-XL. Each snapshot was taken after 85Torb. These plots show that gas overdensities are most pronounced in the largest box. The non-axisymmetric structures have very short lifetimes; less than a tenth of an orbit. The pressure bump structures are more visible when the density is averaged over the azimuthal direction (Figure 3). The black dots represent the position of every 100th particle, integrated in vertical direction. The particles are trapped both in axisymmetric pressure bumps and in spiral density waves as described in Heinemann & Papaloizou (2009). detail in Section 4. All of our simulations show signs of zonal flows. Strength and lifetime of the zonal flows and the asso- ciated pressure bumps differ very much with the physi- cal box size. Space–time plots of all different simulation sizes are shown in Figure 4 and the upper left panel of Figure 5. The pressure bumps are generally more pro- nounced in simulations with a larger radial extent. Sim- ulation set A strictly follows this general trend. Pressure bump features grow in strength and lifetime with the physical box size, always staying at the largest radial scale. This rule applies to all but the largest runs XL, x-XL, and XXL. There, instead of the formerly predom- inant kx = 1 (ωx = 2πkx/Lx in Fourier mode sin (ωxx)) mode, the mode kx = 2 (higher modes for run XXL) is occupied by the pressure bumps. In simulation set B, the strength and size of the pressure bumps converges for simulations with a radial extent larger than 5.28H. The lifetime of the pressure bumps even decreases for the largest simulation in this set. Simulation set C is qualita- tively different from the other simulation sets. Strength, size, and lifetime of the pressure bumps seem to be in- versely proportional to the azimuthal extent of the box, when the vertical and radial box sizes are kept constant. This effect was already seen in, e.g., Simon et al. (2012) and Flock et al. (2012). Both groups show that the mag- netic field consists of two components: a local turbulent component that is responsible for the zonal flows and a global azimuthal component. Since the total energy stays approximately constant, the local component gets weaker and consequently zonal flows as well as axisym- metric pressure bumps get weaker too. Figure 5 additionally shows space–time plots of the az- imuthal gas velocity and the radial gradients of gas den- sity and azimuthal gas velocity of run XXL. In the right panels, the position of the highest dust density are shown as dots. Particles get clearly slowed by the maxima of the azimuthal gas velocity, i.e., the large-scale maxima of the pressure gradients. The velocity has large-scale structures that are very similar to those of the density gradient, as expected in geostrophic balance. Thus, the structure of the velocity gradient can be approximated as the large-scale structure of the second derivative of the gas density. It is shown in the lower right panel that the particles get stopped at the minima of the radial deriva- tive of the azimuthal gas velocity (and thus at minima of the second derivative of the gas density) as analytically predicted (see, e.g., Klahr & Lin 2001). We calculated the correlation time of the pressure bumps and the zonal flows in the same way as it was calculated in Johansen et al. (2009a). We use the den- sity ρ, averaged over azimuthal and vertical directions, at a given time t. Then, we average over each point in radial direction the time it takes for the density at each point to change by a value corresponding to the standard devi- ation of the gas density. These measurements are taken for every local orbit. The measurements are averaged over the time between saturation of the turbulence and a time when the correlation does not extend the corre- lation time to the final time of the simulation. Finally, the averages are multiplied by two, in order to cover the full temporal extent of the correlated structures. The Planetesimal Formation in Zonal Flows 7 y−XL XL t= 85.0Torb 1.15 1.12 1.09 1.06 1.03 5.3 3.9 2.6 1.3 H / y 0.0 −1.3 −2.6 −3.9 −5.3 1.3 1.00 ) 0 = t ( Σ / y > Σ < 0.97 0.94 0.91 0.88 0.85 M x−XL H / y 0.0 −1.3 −1.3 0.0 x/H 1.3 −5.3 −3.9 −2.6 −1.3 0.0 x/H 1.3 2.6 3.9 5.3 Figure 3. Surface density distribution of Figure 2, averaged in azimuthal direction and averaged over the mean surface density. This reveals axisymmetric pressure bumps and valleys. Particles are trapped on the inner side of the density maxima, at places with a positive density gradient to overcome the negative global pressure gradient. These pressure bumps are stable for many orbits (compare Figures 4 and 5). correlation times measured in this fashion are in good agreement with the lifetime of the overdensities that is seen in Figures 4 and 5. However, a change of position of the structures, as seen in run XL (check Figure 4), is not accounted for. Thus, correlation times are more likely to be underestimated than overestimated. Also, we cannot be entirely sure whether this behavior is really drift or structure decay and reformation. The results of the correlation time determination are shown in Table 3 and in the upper panel of Figure 6. For the diagonal simulation set, (A), the correlation time in- creases with box size. It seems to saturate toward the largest box size. The trend to longer correlation times is also evident for simulation set B. Here only run x-XL has a shorter correlation time than expected. This might be an effect of the strongly stretched simulation box. The correlation time decreases slightly with an increasing az- imuthal box size in simulation set C (not shown in the figure). The lower panel in Figure 6 shows a measure- ment for the physical size of the zonal flow features. We Fourier-transformed the vertically and azimuthally av- eraged gas density and azimuthal gas velocity for each time step and averaged the amplitudes of the first four modes over the time of 20 . . . 120 local orbits. The length was normalized for the size of the simulation box, to get the physical size of the modes by λx = Lx/kx with the wave number kx. The turbulence is always strongest at the largest modes for simulations with Lx . 5H. The highest amplitude for both quantities in the largest sim- ulation domain is found between 5H and 7H (up to 10H for ρ). These measurements are also found in Table 3. The runs L SAFI and XL SAFI were carried out to compare the turbulence and zonal flow parameters with the runs L and XL. They were run to check that zonal flows are no effects from the shear advection scheme that was used in the Pencil Code. Comparing the values in Tables 2 and 3 shows that there is little change in the measured properties of the zonal flows and the associ- ated pressure bumps. However, the computation time increases if one uses the SAFI scheme. Thus, this scheme was only used to confirm our results. 4. PARTICLE BEHAVIOR IN ZONAL FLOWS Particle accumulations and planetesimal formation can occur in clumps and filaments of the overdensities in the dust. In our simulations, we do not include gravitational interaction between the particles. Thus, we only study the passively developed overdensities of the dust to see when and whether overdensities sufficient for the stream- ing instability can be reached. By not having explicit feedback one can retroactively study the concentration factor for various initial dust-to-gas ratios. Simulations including feedback will have to be done in future stud- ies. Figure 2 shows the position of every 100th parti- cle in selected simulations. These plots clearly show the trend for particles to accumulate in the downstream of high-pressure regions. Particles are pulled toward pres- sure gradient maxima (Klahr & Lin 2001). In the upper right panel in Figure 2, a snapshot of run XL after 85 local orbits is shown. The particles clump up at posi- tions just left of the maxima in the kx = 1 of the gas density; these are the locations of positive zonal flows, 8 Dittrich, Klahr, & Johansen Table 3 Zonal Flow Properties Run (1) S M L XL XXL x-S x-L x-XL y-S y-L y-XL LspecMR LspecHR LspecMRs MspecMRb L SAFI XL SAFI ρrms (2) 6.1 × 10−3 2.0 × 10−2 3.9 × 10−2 4.3 × 10−2 4.0 × 10−2 1.6 × 10−2 3.3 × 10−2 2.7 × 10−2 1.2 × 10−2 3.0 × 10−2 3.6 × 10−2 3.7 × 10−2 4.2 × 10−2 4.3 × 10−2 1.9 × 10−2 3.9 × 10−2 4.8 × 10−2 ρ(kx = 1) ρ(kx = 2) ρ(kx = 3) uy(kx = 1) uy(kx = 2) uy(kx = 3) (3) 4.1 × 10−3 1.0 × 10−2 2.1 × 10−2 1.5 × 10−2 5.0 × 10−3 3.9 × 10−3 2.3 × 10−2 1.2 × 10−2 9.9 × 10−3 6.9 × 10−3 5.6 × 10−3 1.8 × 10−2 9.8 × 10−3 2.4 × 10−2 8.6 × 10−3 2.1 × 10−2 2.4 × 10−2 (4) 7.5 × 10−4 2.5 × 10−3 5.6 × 10−3 1.1 × 10−2 7.9 × 10−3 1.7 × 10−3 7.3 × 10−3 1.0 × 10−2 2.4 × 10−3 3.8 × 10−3 4.1 × 10−3 5.5 × 10−3 3.2 × 10−3 5.4 × 10−3 2.6 × 10−3 5.3 × 10−3 1.0 × 10−2 (5) 4.0 × 10−4 1.6 × 10−3 3.2 × 10−3 5.0 × 10−3 7.9 × 10−3 1.2 × 10−3 2.7 × 10−3 6.4 × 10−3 1.1 × 10−3 3.1 × 10−3 3.3 × 10−3 3.1 × 10−3 2.0 × 10−3 3.2 × 10−3 1.6 × 10−3 3.2 × 10−3 5.4 × 10−3 (6) 9.9 × 10−3 1.2 × 10−2 1.3 × 10−2 4.6 × 10−3 7.8 × 10−4 8.0 × 10−3 1.4 × 10−2 3.5 × 10−3 1.2 × 10−2 7.8 × 10−3 5.7 × 10−3 1.1 × 10−2 5.8 × 10−3 1.4 × 10−2 1.0 × 10−2 1.2 × 10−2 7.0 × 10−3 (7) 3.4 × 10−3 5.2 × 10−3 6.5 × 10−3 6.6 × 10−3 2.4 × 10−3 3.2 × 10−3 8.8 × 10−3 6.2 × 10−3 5.9 × 10−3 3.7 × 10−3 2.6 × 10−3 6.1 × 10−3 3.6 × 10−3 6.1 × 10−3 5.3 × 10−3 6.1 × 10−3 6.2 × 10−3 (8) 2.1 × 10−3 3.1 × 10−3 3.7 × 10−3 4.3 × 10−3 3.5 × 10−3 2.0 × 10−3 4.7 × 10−3 5.6 × 10−3 4.1 × 10−3 2.3 × 10−3 1.6 × 10−3 3.6 × 10−3 2.0 × 10−3 3.9 × 10−3 3.1 × 10−3 3.9 × 10−3 4.8 × 10−3 τcorr (9) 7.6 11.2 23.2 43.2 47.3 4.4 37.6 20.2 14.4 10.8 10.3 21.8 23.4 10.9 26.4 25.6 48.6 Notes. Column 1: name of run. Column 2: root-mean-square density ρrms = ph(ρ − ρ)2i. Columns 3-5: Fourier amplitude of radial density modes kx = 1 . . . 3, normalized by mean density in the box. Columns 6-8: Fourier amplitude of azimuthal velocity modes kx = 1 . . . 3 with uy = uy − uy. Column 9: correlation time, in orbits T = 2πΩ−1, of the largest radial density mode. i.e., regions where the azimuthal gas velocity is higher than the pressure-supported Keplerian flow. 4.1. Particles in Zonal Flows In the upper right panel of Figure 5, the azimuthal gas velocity development of run XXL is shown, overplotted with the position of the most massive clump for each time step. The azimuthal gas velocity coincides with the derivative of the gas density, but it is much eas- ier to interpret. The speckled structure of the deriva- tives comes due to the high power in the smaller scales. However, the large-scale structure is still visible and the geostrophic correlation between the structures of uy(x, t) and d/dx[ρ(x, t)] is directly observed. Since they have the same large-scale structure, the particle position is much easier interpreted at the azimuthal gas velocity plot than on the density gradient plot. Sometimes the radial displacement from one orbit to the next is too large to be explained by radial drift. That happens when an- other clump becomes more massive than the previous one. These particles accumulate in regions with high az- imuthal gas velocities (see upper right panel in Figure 5). The only time when this is not true is at times from 80 to 100 local orbits. In this period, an inward-drifting clump stayed coherent during the time of its drift. The drift velocity of the most massive particle clump is indi- rectly encrypted in this plot. Particles are drifting much slower when they are trapped by a pressure gradient. As all particles drift inward this leads to accumulation of particles in regions where the perturbed pressure gradi- ent is positive. The maximal accumulation of particles for runs XL in Figure 7. and y-S are plotted in the top panel The second panel shows the evolution of the quantity x, a measure for the strength of the zonal flows. The third panel in Figure 7 shows the evolu- tion of the strength of the gas density enhancement as x. Comparing the second and third panels, phuy − huyiyzi2 phρ − hρiyzi2 one can see a clear correlation between the zonal flow strength and the gas density enhancements. The bottom panel in Figure 7 shows the evolution of the α-parameter, calculated as in Equation (14). The maximum of the dust overdensity that occurs dur- ing one simulation is plotted against the box size in the upper panel of Figure 8. The general trend shows that radially larger boxes have higher particle concentrations. An increased azimuthal extent does not have an effect on the particle concentrations. The most surprising result is in run y-S. It shows a very high particle concentration that occurs early in the simulation (compare Figure 7). This is most likely a stochastic coincidence. The lower panel in Figure 8 shows a plot of the maximum dust over- density against the correlation time of the zonal flows. The error margin are calculated with the standard de- viation of the temporal evolution of the two quantities. We see a clear trend that denser particle accumulations develop with longer correlation times. The distribution can be fitted by a power law. This gives an exponent of d log ρ/d log τcorr = 0.38 ± 0.05. The one point that does not overlap with the error margins of the fit is from run y- S. If we take the maximum of the top panel in Figure 7 after the two first maxima (i.e., after 45Torb) and plot this value again in the parameter space of Figure 8, we get the position marked with the blue square. It agrees well with the error margins of the fit. In isothermal geostrophic balance, 2ρΩuy = c2 s ∂ρ/∂x, the azimuthal gas velocity follows the radial density gra- dient. That this is true for large scales as shown in Fig- ure 9. The upper left panel shows the evolution of the azimuthally and vertically averaged azimuthal compo- nent of the gas velocity. Overplotted are the locations of the maxima in the dust density. In the upper right panel the dust density evolution of the same run L is plotted. In comparing the location and times of the maxima and minima on these two plots, one clearly sees that max- ima in the dust density occur often at times and loca- 120 100 80 60 40 20 −0.64 S b r o T / t 0.00 x/H 0.64 120 100 80 60 40 20 −0.64 y−L Planetesimal Formation in Zonal Flows x−S 0.00 x/H 0.64 b r o T / t 120 100 80 60 40 20 −1.30 −0.65 y−S 0.00 x/H 0.65 1.30 b r o T / t 120 100 80 60 40 20 −1.30 −0.65 9 M 0.00 x/H 0.65 1.30 L 120 100 80 60 40 20 −1.30 −0.65 120 100 80 60 40 20 −1.30 −0.65 0.00 x/H 0.65 1.30 y−XL 0.00 x/H 0.65 1.30 0.15 0.12 0.09 0.06 0.03 0.00 −0.03 −0.06 −0.09 −0.12 −0.15 > ) t ( ρ < / ] > ) t ( ρ < − ) t , x ( ρ [ b r o T / t b r o T / t 120 100 80 60 40 20 −2.62 −1.97 −1.31 −0.66 120 100 80 60 40 20 −2.62 −1.97 −1.31 −0.66 0.00 x/H 0.00 x/H 0.66 1.31 1.97 2.62 x−L 0.66 1.31 1.97 2.62 XL b r o T / t b r o T / t b r o T / t b r o T / t b r o T / t 120 100 80 60 40 20 −5.26 −4.60 −3.95 −3.29 −2.63 −1.97 −1.32 −0.66 120 100 80 60 40 20 −5.26 −4.60 −3.95 −3.29 −2.63 −1.97 −1.32 −0.66 0.00 x/H 0.00 x/H 0.66 1.32 1.97 2.63 3.29 3.95 4.60 5.26 x−XL 0.66 1.32 1.97 2.63 3.29 3.95 4.60 5.26 Figure 4. Evolution of the gas density perturbation of all runs from simulation sets A, B, and C. Run XXL is shown in Figure 5. The density is averaged in vertical and azimuthal direction and plotted in radial direction over time. The lifetime, the size, as well as the strength of the pressure bumps are clearly increasing with increasing box size in simulation set A, i.e., runs S, M, L, XL, and XXL. In simulation set B, i.e., runs x-S, M, x-L, and x-XL, we have the same increase of lifetime, size, and strength of the pressure bumps. Only for the very large simulation (Lx = 10.56H), there is no apparent difference in pressure bump size and strength to run x-L. For simulation set C, i.e., runs y-S, M, y-L, and y-XL, the strength of the pressure bumps is apparently constant throughout this set of simulations. Even the lifetime decreases slightly with increasing box size. tions where one finds maxima in the gas velocity. Two attempts to quantify this observation are shown in the lower row of Figure 9. In the left panel, the particle density and the azimuthal velocity from the two upper panels are plotted against each other, regardless of po- sition and time. In the right panel, a snapshot of the simulation (as in Figure 2) was taken at 85 local orbits, the time when the maximum dust density enhancement occurs. The particle density as well as the azimuthal gas velocity were integrated in vertical direction and plotted against each other, regardless of their radial or azimuthal position in the simulation, in this scatter plot. In order to visualize high densities of points in these plots, we computed a two-dimensional histogram of the scattered points. This is indicated by the color scale, showing the amount of points in each of the boxes in the scatter plot space. There is a clear trend for high dust density con- centrations to appear at high gas velocities. Without radial drift particles would concentrate where uy = 0, i.e., between the sub- and super-Keplerian flow. Due to the radial drift particles accumulate slightly down- stream at the formed pressure bumps. Those happen to be at the maxima of the azimuthal gas velocity. With the geostrophic balance, high velocities are also regions of a high radial density gradient. These plots prove that the particles in the simulations are trapped by the long-lived pressure gradients that occur due to stable zonal flows. If the dust-to-gas ratio increases to values larger than 10 b r o T / t 120 100 80 60 40 20 −10 −5 0 x/H 5 10 b r o T / t 120 100 80 60 40 20 Dittrich, Klahr, & Johansen > ) t ( ρ < / ] > ) t ( ρ < − ) t , x ( ρ [ ) > ) t ( ρ < / ) t , x ( ρ ( x d / d 0.15 0.12 0.09 0.06 0.03 0.00 −0.03 −0.06 −0.09 −0.12 −0.15 0.10 0.08 0.06 0.04 0.02 −0.00 −0.02 −0.04 −0.06 −0.08 −0.10 b r o T / t b r o T / t 120 100 80 60 40 20 −10 −5 0 x/H 5 10 120 100 80 60 40 20 0.060 0.048 0.036 0.024 0.012 0.000 −0.012 −0.024 −0.036 −0.048 −0.060 s c / ] v ∆ + ) t , x ( y u [ 0.10 0.08 0.06 0.04 0.02 −0.00 −0.02 −0.04 −0.06 −0.08 −0.10 ) s c / ) t , x ( y u ( x d / d −10 −5 0 x/H 5 10 −10 −5 0 x/H 5 10 Figure 5. Top panels show the evolution of the gas density perturbation and the azimuthal gas velocity of run XXL, whereas the bottom row shows the radial derivative of these quantities. The derivatives are very speckled, since small-scale fluctuations give stronger amplitudes to the derivatives. However, the underlying large-scale structure is still visible. The azimuthal gas velocity follows the radial gas density gradient, as expected for a geostrophic balance. Hence, it is possible to interpret the radial derivative of the azimuthal gas velocity as the second derivative of the gas density. In the upper right panel, the black dots represent the position of the most massive particle clump in the simulation at each time. It is clearly shown that particles get trapped in regions of positive zonal flow downstream of pressure bumps. unity, the streaming instability (Youdin & Goodman 2005; Johansen & Youdin 2007; Youdin & Johansen 2007) is triggered. This increases the dust density further on timescales shorter than an orbital period. To follow the streaming instability development, the back-reaction of the dust particles to the gas phase must be considered in future numerical simulation. This effect was neglected in this set of simulations. Otherwise the initial dust-to- gas ratio would have been an additional free parameter to be studied. 4.2. Radial Drift Radial drift velocities of the particles in the simula- tions with different box sizes are shown in Figure 10. The upper panel shows the measured and expected ra- dial drift of two simulations (M and XL). They show that particles drift slower in turbulent simulations than they would in a laminar disk. However, the size of the simulation has little effect on the actual drift velocity, as shown in the lower panel of Figure 10. It shows a time average of the particle drift velocity plotted against the box size. The uncertainties are too large to reveal a trend. Thus, the reduction of the radial drift velocity apparently only depends on the amplitude of the zonal flow, but not on the correlation time. Looking at the largest run XXL, we can estimate that the radial drift gets reduced by about 28% (drop of the absolute value from 0.05cs to (0.036 ± 0.003)cs). 4.3. Clustering The clustering degree of the particle distribution can be estimated with the distribution of the dust surface density Σp (Pan et al. 2011). The initial distribution is Planetesimal Formation in Zonal Flows 11 simulation set A simulation set B 102 101 b r o T / r r o c τ 10000 1000 ) 0 = t ( p ρ / ) p ρ ( x a m 1 Lx/H 10 simulation set A simulation set B ) y u ( s m r ) ρ ( s m r 0.04 0.03 0.02 0.01 0.00 0.04 0.03 0.02 0.01 0.00 α 0.010 0.001 XL y−S ) 0 = t ( ρ / ^ρ s c / y ^u 10−2 10−3 10−2 10−3 1 λ x/H 10 Figure 6. Upper plot shows the correlation times of the largest radial density mode against the radial box size. The lines corre- spond to the simulation sets A and B. The results from simulation set C are omitted for visibility. The correlation time τcorr grows for boxes with a larger radial extent. Only run x-XL does not fol- low this trend. The large ratio Lx/Ly may prohibit formation of stable zonal flows. The two lower plots show the first four am- plitudes of the radial Fourier modes of the gas density and the azimuthal gas velocity against their real size λx = Lx/kx; kx is the wave number of the corresponding Fourier mode, defined by ωx = 2πkx/Lx in the Fourier mode sin (ωxx). The lines connect the amplitudes of different Fourier modes for one simulation. Both quantities have most of their power in the largest modes. Only in the largest simulations, the power in the largest modes decreases. There the maximum is between 5H and 7H. represented by a Poisson distribution (see Figure 11).6 For this plot, we binned the measured dust surface den- sity of a snapshot. We then normalized them to the amount of grid cells. About three local orbits after the particles feel the gas drag, the shape of the distribution function is saturated. We averaged the distribution over the time of 23 . . . 121Torb. We see at the high density end of the distribution that higher densities develop in larger boxes due to the higher number of available particles. Thus, the clustering properties do not depend strongly on the strength or lifetime of the zonal flows (compare Figure 8, bottom). 4.4. Different Particle Sizes So far we only considered simulations with one parti- cle species, i.e., St = τfΩ = 1. We take the simulation 6 Run XXL was not included in this figure, because the number of particles per grid cell was different to the other runs. 0 20 40 60 t/Torb 80 100 120 Figure 7. Time series of runs XL and y-S. The plots show (from top to bottom) the maximum of the dust density, the root mean square of the azimuthal gas velocity and the gas density, and the α-value (Equation (14)). The dust overdensities of run XL have a higher base than those of run y-S. The latter has some spikes in the beginning, but is lower for most time of the simulation. The two panels in the middle show that the azimuthal gas velocity and the gas density are correlated. Both plots show maxima and minima at the same time, while α is rather stable with time. The time-averaged α-values for all simulations can be found in Table 2. size that simulates one fully extended zonal flow and in- vestigate 12 different particle species. The particle sizes range from St = 0.01 to St = 100. We choose run L with the dimensions 5.28H× 5.28H× 2.64H as simulation size for the last simulation set. For one simulation we used a smaller box, because the integration time had to be increased be a factor of two to give the particles with the high Stokes numbers the opportunity to react on the pressure differences. 4.4.1. Drift Velocity and Particle Densities The results are shown in Figure 12. The upper left panel shows the negative of the radial velocity of the par- ticles, averaged over all particles of a certain size and over time. The four different simulations match very well. The plot shows that particles with St = 1 drift fastest inward, also with turbulence in the simulations. On both sides the inward drift velocity decreases with simi- lar slopes. The key to the different colors and symbols is in the lower right panel. Overplotted, in a dashed gray line, we find the analytical prediction (following Equation (12)) for the radial drift in a laminar disk. The difference to the prediction is shown in the lower sub-panel. Large particles generally drift slower according to the steady- state solution and their coupling to the gas is also much 12 Dittrich, Klahr, & Johansen ) 0 = t ( p ρ / ) p ρ ( x a m ) 0 = t ( p ρ / ) p ρ ( x a m p,0 as a function of Li p)/ρ max(ρ simulation set A simulation set B simulation set C 1 10 Li/H (i=x,y) 104 103 104 103 10 τ corr/Torb 100 Figure 8. In the upper panel, the highest peak in the time series of the dust overdensity (the top panel in Figure 7) is plotted against the size of the simulation box. The diagonal simulation set A is marked by the blue line. The dust overdensity increases with box size, until it suddenly drops for the largest box. In simulation set B (red), the quantity saturates for boxes with a radial extent that is twice as large as the azimuthal extent or larger. When keeping the radial extent constant (simulation set C, yellow line), the maximum saturates for the cubic box case. Hence, the only in azimuthal direction extended boxes do not lead to an artificial enhancement of the dust overdensities. The very high overdensity for run y-S seems to be a stochastic coincidence (compare top panel in Figure 7). The lower panel shows the dust overdensity against the correlation time. The measured points can be approximated with a power law (shown as a red solid line) with an exponent of 0.38 ± 0.05. The shaded region, with the dashed red lines as edges, gives the uncertainty of the fit. The one cross off the fit shows the results for run y-S. The blue square marks its position, if we neglect the two maxima shown in Figure 7. It overlaps well with the fit region. weaker. Hence their radial drift velocity is almost not affected by the turbulence and they do not show strong concentrations. Small particles with low Stokes numbers are stronger coupled to the gas and, thus, also drift very slow. Particles with St ∼ 1 are concentrated most by the zonal flow and, thus, have a stronger decreased radial velocity. Thus, the accumulation of dust particles is ex- pected to be strongest for particles with Stokes numbers around unity. For St = 0.01 particles, the drift velocity is strongly determined by the gas flow. This explains the strong deviation from the expected drift velocity. The upper right panel shows the total particles over- density normalized to the initial particle number den- sity. For run LspecMR (black diamonds), the smallest particles have higher concentrations than in the other simulations. This resulted from the choice of too few particles per grid cell. There only 100,000 particles per size bin were simulated. This results in overestimation, because the number density is normalized with the ini- tial number density n0. For example, run LspecMRs (red squares) follows 2,000,000 particles per particle size bin. The highest concentrations were reached for particles of sizes St = 0.75 . . . 5, as expected. However, the exact peak has a stochastic factor to it. Thus, the simulations peak at different particle sizes. The overdensities are more investigated in the lower row of panels. The surface number density of the particles is shown in the lower left panel. Here, the particles were integrated in the vertical direction. The trend is similar to the upper right panel. We read from this plot that particles with St = 0.1 are concentrated about ten times the initial con- centration. Together with the vertical overdensity due to sedimentation (lower right panel), a total overdensity of about 100 is created for St = 0.1 particles. The peaks in the vertical density structure of the par- ticles are shown in the lower right panel of Figure 12. The Stokes number, St = τfΩ defines the timescale after which the particles are settled down to the mid-plane. Particles with a high Stokes number are not fully settled down to the mid- plane, not even in the long-integration run MspecMRb. The resolution also limits this measure- ment for particles that are very close to the mid-plane. Smaller particles are not that strongly stratified. Thus, the vertical (Gaussian) structure is wider and shallower. This results in a lower value in this plot. The points for Stokes numbers 0.01–1 follow a power law with the index of 0.58± 0.03. The measured power law index is slightly higher than the expected value of 0.5 (Dubrulle et al. 1995). Most of the particles with St & 1 sediment very close to the mid-plane. This prohibits a further increase in the vertical density. A higher resolution and a mea- surement of the dust scale height is achieved in the next section. 4.4.2. Dust Pressure Scale Height With a stratified particle distribution we can test the vertical diffusion model (see, e.g., Carballido et al. 2006). The dust pressure scale height can be directly calcu- lated from the vertical positions of the particles of the It is approximately proportional to St−0.5 same size. in agreement with Carballido et al. (2006, 2011) and Youdin & Lithwick (2007). The results are summarized in the upper panel of Figure 13. Since the analyti- cal value was calculated with the α-value, the vertical Schmidt number Scz = Hp, expected Hp, measured ∼(cid:18) α DT (∞)(cid:19) 1 2 (16) can be calculated. We measured the vertical Schmidt number to have a very weak dependence on the particle size. In the lower panel of Figure 13 we show that Scz = 3.4 · St0.11. 5. DISCUSSION AND CONCLUSIONS 5.1. Zonal Flows and Axisymmetric Pressure Bumps Our simulations have dimensionless units. This allows us to interpret our results manyfold. We can pick the distance to the star in a certain range. In Section 2.5, we defined the global pressure gradient to be ∆v = 0.05cs. In the minimum mass solar nebula (MMSN) model, we can choose the distance to the star to be between 0.35 and 40 AU (Hayashi 1981). For this discussion, we pick Planetesimal Formation in Zonal Flows 13 0.085 0.068 0.051 0.034 0.017 0.000 −0.017 −0.034 −0.051 −0.068 −0.085 s c / ] v ∆ + ) t , x ( y u [ b r o T / t 120 100 80 60 40 20 −2 −1 0 x/H 1 2 −2 −1 0 x/H 1 2 7.3 102 6.1 4.9 3.7 2.4 1.2 x o b g n i d n o p s e r r o c n i s t n i o p f o e g a t n e c r e p > ) 0 = t , y , x ( p ρ < / ) b r o T 5 8 = t , y , x ( p ρ 101 100 10−1 6.0 5.4 4.8 4.2 3.6 > ) t ( p ρ < / ) t , x ( p 3.0 2.4 ρ 1.8 1.2 0.6 0.0 1.3 1.1 0.9 0.6 0.4 0.2 x o b g n i d n o p s e r r o c n i s t n i o p f o e g a t n e c r e p b r o T / t > ) t ( p ρ < / ) t , x ( p ρ 120 100 80 60 40 20 102 101 100 10−1 10−2 −0.10 −0.05 0.00 <(uy+∆v)/cs>y,z 0.05 0.10 −0.10 −0.05 0.0 10−2 0.00 <(uy+∆v)/cs>z 0.05 0.10 0.0 Figure 9. Top row shows the evolution of the azimuthal gas velocity and the dust density evolution of run L, respectively. The quantities are averaged in vertical and azimuthal direction and plotted in radial direction over time. The black dots in the upper left panel show the position of the highest dust density at each orbit. This shows that the overdensities of the dust often appear at places and times where the azimuthal gas velocity is high. This relation shows that the zonal flows accumulate dust and are a possible venue of planetesimal formation. The bottom left panel shows a scatter plot of the dust density ρp against the azimuthal gas velocity, where both values are averaged in vertical and azimuthal direction, as in the upper panels. The bottom right panel shows a plot of the particle surface density ρp(x, y, t) in relation to the azimuthal gas velocity, averaged in vertical direction, computed from a snapshot taken at 85Torb, the time when the maximum in the dust density occurs. Both plots show that it is more likely to find a high dust density at a location where the azimuthal gas velocity is high. r = 5 AU. In a thin disk model, we get a ratio for H/r ∼ 0.033(5 AU/AU)1/4 ∼ 0.05; this defines us H = 0.25 AU. The isothermal sound speed is cs = HΩ ∼ 66,000 cm s−1. Thus, turbulent velocities (urms) are about 9000 cm s−1 (∼ 7000 cm s−1 for the high-resolution run LspecHR). Figure 14 shows the highest azimuthal velocity for all simulation sizes. We averaged over several maxima of uy(x, t) for every simulation to smooth over outliers. The zonal flows are super-Keplerian for all but runs XXL, x-S, and y-XL. In the largest box the flow only reaches slightly sub-Keplerian velocities. However, particles still get cap- tured in the resulting axisymmetric pressure bumps. The speeds measured in the largest simulation match those measured in Flock et al. (2011). We measured the radial size of the axisymmetric pres- sure bumps to be between 5 and 7H (see Figure 6). At a distance of 5 AU to the star, this size corresponds to ∼ 1.25 . . . 1.75 AU radial size for zonal flows, i.e., the dis- tance between peaks of hρiyz. This measurement agrees well with Simon et al. (2012) who measured the radial size of their zonal flows to be 6H. Further studies with varying box size in smaller steps could potentially narrow down the radial scale. We measured the lifetimes of the zonal flows up to 50Torb. This agrees well with earlier stated lifetimes (Johansen et al. 2009a; Uribe et al. 2011). The strength Dittrich, Klahr, & Johansen M (2.643) XL (10.562×2.64) measured vx smoothed vx stationary solution of vx r e b m u n d e z i l a m r o n 100 100 10−2 10−2 10−4 10−4 10−6 10−6 10−8 10−8 averaged distribution initial distribution poisson distribution S M L XL 20 40 60 t/Torb 80 100 120 10−1 10−1 100 100 101 101 p/Σ(t=0) Σ 102 102 simulation set A simulation set B stationary solution 14 s c / > x v < s c / > x v < 0.05 0.00 −0.05 0.05 0.00 −0.05 0.00 −0.01 −0.02 −0.03 −0.04 −0.05 t > s c / > x v < < Figure 11. Distribution of the dust-to-gas ratio of the surface densities for runs S, M, L, and XL. For comparison a Poisson dis- tribution is shown with crosses. The initial distribution of the nu- merical simulation (dotted line) fits very well to the normal distri- bution. The average strength of the clustering for the dust surface density does not depend on the simulation box size. tions we have to assume a distance to the star and pick a solar system model. This will allow us to discuss our results in context to recent experiments. By choosing a model for the solar system, we can con- vert the dimensionless Stokes number St = τfΩ to a real particle size. The friction time τf correlates to the parti- cle radius a with a = for Epstein drag and τ (Ep) f ΩΣgas √2πρ• , (17) a =s 9τ (St) f ΩµH 4ρ•σmol , (18) drag (see Here Σgas is Stokes supplementary for for Johansen et al. 2007). the column density of the gas, ρ• is the density of solid material, µ = 3.9 × 10−24 g is the mean molecular weight, and σmol = 2 × 10−15 cm2 is the molecular cross section of molecular hydrogen (Nakagawa et al. 1986; Chapman & Cowling 1970). info The Epstein regime applies, if the particle radius a does not exceed (9/4) (Weidenschilling 1977a) of the gas mean-free path λ = µ ρgasσmol √2πµH Σgasσmol . = (19) The gas density and hence also the particle size for a given Stokes number St = τfΩ depends very much on the used model. In Figure 15 we overview four different mod- els. The MMSN (Weidenschilling 1977b; Hayashi 1981) was calculated from the mass of the existing planets, ne- glecting migration. Because this model allows no mass loss through accretion, often 3·MMSN is used to account 1 10 Lx/H Figure 10. Radial drift velocity of the particles for two different simulations is shown in the upper panel. Particles in these simula- tions all have a Stokes number of St = τfΩ = 1. The orange lines are the exact measured radial velocities, averaged over all parti- cles. The black line represents the same value smoothed over the time of one local orbit. The blue dashed line shows the analytical result in a stationary box for particles of St = 1 following Equation (12). Particles in turbulent simulations generally drift slower than expected from the stationary solution, but the box size has little effect on the drift velocity. This is shown in the lower panel where the mean of the radial drift velocity is plotted against the box size. We omitted simulation set C for visibility. There is a minimum in drift speed for run L. However, this minimum is within the er- ror margins. The smallest errors are with run XXL; here the drift velocity drops by 28%. of the density bump reaches 15% and goes down to about 10% in the largest simulation. The lower amplitude is consistent with the results from global simulations (pri- vate communication with Mario Flock about the sim- ulations from Flock et al. 2011, 2012) who measured a density enhancement of slightly less than 10%. Some works (e.g., Uribe et al. 2011; Simon et al. 2012) measure stronger density enhancements. A possible explanation is that their α values are higher than in this work. Further studies on the dependence of volume average quantities to strength of zonal flows would be interesting. 5.2. Dust in Zonal Flows Particles get trapped downstream of pressure bumps and build up overdensities. To compare our dimension- less particle sizes with collision experiments and observa- Planetesimal Formation in Zonal Flows 15 −vx,stationary > ) 0 = t ( p n < / ) p n ( x a m 10000 1000 100 10 0.01 0.10 1.00 St=τ Ω f 10.00 100.00 0.01 0.10 1.00 St=τ Ω f 10.00 100.00 100 10 > y , x > ) 0 = t ( p n < < / ) y , x > p n < ( x a m LspecMR LspecHR LspecMRs MspecMRb ~St0.58 s c / > x v < − 0.010 0.001 y r a n o i t a t s , x v / ) y r a n o i t a t s , x v − > x v < ( 0.8 0.6 0.4 0.2 0.0 > ) 0 = t ( p Σ < / ) p Σ ( x a m 100 10 1 0.01 0.10 1.00 St=τ Ω f 10.00 100.00 0.01 0.10 1.00 St=τ Ω f 10.00 100.00 Figure 12. These panels show the behavior of particles with Stokes numbers from 0.01 to 100. The upper left panel shows the negative of the radial drift velocity and the relative difference between the measured and expected drift velocity. The dashed gray line shows the stationary solution for the radial drift, following Equation (12). The highest drift velocities are obtained for particles with St = 1, but they are also slowed down strongest by the MRI-turbulence. The upper right panel shows the highest overdensity that occurred for the specific particle size during the entire simulation. The slopes for the different simulations match very well, apart from a jump around St = 1 (for LspecMRs and MspecMRb) and an offset for run LspecMR at small particle sizes. The former can be explained with the usage of a smaller simulation box for run MspecMRb (2.643 with weaker zonal flows) than in the other simulations (5.282 × 2.64 with stronger zonal flows). The offset showed that the number of particles per particle size was not sufficient in run LspecMR (105 particles in ∼ 1.5 × 106 grid cells leads with five particles in one grid cell to a result of max (np)/hnp(t = 0)i = 75). The lower left panel shows the maximum of the column density for each particle size. It peaks at sizes of around St = 1. The lower right panel shows the maxima of the vertical distribution of particles. The curves (for St = 0.01 . . . 1) follow a power-law with the index of 0.58 ± 0.03. This is slightly steeper than the expected power law index of 0.5 (Dubrulle et al. 1995). In all four plots, the results of particles with St = 0.01 are to be interpreted with caution, because the simulations lacked sufficient amount of super-particles for these size bins. Further, large particles (St = 100) did not have enough time to sediment to the mid-plane. for some accretion. A low-density model was published by Brauer et al. (2008). This model is adopted from mea- surements that indicate a shallow surface density pro- file for protoplanetary disks (Andrews et al. 2010). The high-density model was adopted from Desch (2007), who introduced a "revised MMSN model" by using the start- ing positions in the Nice model of planetary dynamics (Tsiganis et al. 2005). This model also takes planetary migration into account. The equations used to calculate Dittrich, Klahr, & Johansen analytic (CFP06) fit: 0.039·St−0.56 s c / ] v ∆ + ) t , x ( y u [ x a m 100 St 101 102 LspecMR LspecHR LspecMRs MspecMRb 10−1 10−2 0.10 0.08 0.06 0.04 0.02 0.00 S M L X X L X x − x − x − y − y − y − L S L X S L X L L vertical Schmidt number fit: 3.4·St0.11 16 H / p H 100 10−1 10−2 10−3 101 z c S Figure 14. Measured highest azimuthal velocity of all box size simulated. The gray dashed line shows the threshold to Keplerian velocity. Only runs XXL and y-XL do never get super-Keplerian. Simulation x-S does get super-Keplerian at some time, but not often enough to be significant. The concentration factors of run LspecHR in the upper right panel of Figure 12 show us that with an initial dust- to-gas ratio of ǫ0 = 10−2 particles of sizes St = 0.5 . . . 10 (St = 0.1 . . . 0.25) reach a dust-to-gas ratio of 100 (& 1). These sizes translate to 30 . . . 400 cm (6 . . . 15 cm) in an MMSN at a 5 AU orbit using Figure 15. Considering back-reaction from the dust to the gas would allow the streaming instability to act. This will be subject of a future study. In our simulations, we see that the den- sity of 15 . . . 600 cm sized icy boulders increases several thousand times over the equilibrium density, even with- out streaming instability and self-gravity of the particles. Sedimentation to the mid-plane leads to overdensities of around 40, while the contribution from the turbulence concentrates the boulders several hundred times. Since we do not study the influence of the back-reaction from particles to the gas, we were able to study several particle sizes in one simulation. That also means that the initial dust-to-gas ratio (ǫ0) can be set arbitrary. We can interpret our results in the light of different metallic- ities. Particles with St ≥ 0.5 will trigger the streaming instability even with ǫ0 = 10−4, while St = 0.1 particles need ǫ0 = 10−2. At the assumed distance in this discussion, the result- ing rings of trapped dust are not observable with cur- rent telescopes. If zonal flows form at larger distances to the star and dust rings form at an observable size, they could potentially be observable with ALMA. For an analysis one would have to adjust the parameter ∆v to account for the steeper pressure gradient. A preliminary study showed that particles of about 10 cm in size can get capture for a short amount of time at 100 AU dis- tance. However, this question goes beyond the scope of this paper and should be addressed in a future study. 6. SUMMARY AND OUTLOOK We performed numerical simulations of MRI-driven turbulence in shearing boxes, covering the parameter space for radial and azimuthal box sizes up to 21.12H. Further, we followed the reaction of the dust particle den- sity to the turbulence. Our major findings are as follows. 1. Turbulent energy and stresses double when increas- ing the azimuthal size of the simulation from 1.32 to 2.64 pressure scale heights. Turbulence pa- 100 10−2 10−1 100 St 101 102 Figure 13. Dust scale height as a function of the Stokes number is shown in the upper panel. Different symbols depict the dif- ferent simulations. The expected dust scale height is calculated after Carballido et al. (2006) and compared with the fitted func- tion. This value is the vertical Schmidt number Scz, shown in the lower panel. Its dependence on the particle size is very weak. the particle sizes in Figure 15 are (20) Σgas =  1700 g 5100 g 683 g 51,000 g AU(cid:1)−1.5 cm2 (cid:0) r AU(cid:1)−1.5 cm2 (cid:0) r AU(cid:1)−0.9 cm2 (cid:0) r AU(cid:1)−2.2 cm2 (cid:0) r (MMSN) (3 · MMSN) (low density) (high density). Throughout the discussion, we assume the MMSN model at 5 AU distance to the star for size reference for our test particles. This choice affects only the translation from the Stokes number St to a size, not the dynamics in our models. If the local dust density exceeds the Roche density, a clump is gravitationally bound against shear. The Roche density can be approximated (Kopal 1989) by ρRoche(R = 5 AU) = 9 4π Ω2 G(R = 5 AU) ∼ 100ρ(R = 5 AU) , (21) the gravitational constant. for an MMSN. G is The streaming instability (Youdin & Johansen 2007; Johansen & Youdin 2007) starts to act at dust-to-gas ratios of order unity. We started all our simulations with ǫ0 = ρp(t = 0)/ρ = 0.01. Thus, a concentration of max(np)/hn0i = 100 corresponds to ǫstreaming = 1. The Roche density at 5 AU in an MMSN can be ex- pressed as ǫRoche = ρRoche/ρ ∼ 100. We can see that objects of several decimeters up to some meters reach ǫRoche, while pebbles of some centimeters up to a decime- ter reach ǫstreaming from combining Figures 12 and 15. Planetesimal Formation in Zonal Flows 17 ] m c [ a 103 102 101 100 10−1 MMSN 3·MMSN low high MMSN, this study 10−2 10−1 100 St 101 102 Figure 15. Particle sizes as a function of the dimensionless Stokes number for the four discussed models at 5 AU. The black squares show the used Stokes numbers and their corresponding size in the case of the minimum mass solar nebula model. rameters in radially small box sizes stay approx- imately constant. This confirms the results in Fromang & Stone (2009). In larger boxes, turbu- lent fluctuations and stresses are observed to re- main constant against changes in the box size (see also Johansen et al. 2009a). This rapid conver- gence was also observed in Simon et al. (2012). 2. Surface density fluctuations grow to large scales in the box and have lifetimes of up to 50 or- bits. The scales of these pressure bumps increase with increasing radial box size, until it saturates at approximately 5–7 pressure scale heights. The scales are decreased when the azimuthal box size is much more increased than the radial box size. The radial scales of the pressure bumps are con- sistent with the length scales measured in local (e.g., Johansen et al. 2009a; Simon et al. 2012) and global (e.g., Lyra et al. 2008; Uribe et al. 2011) simulations. This might be the natural size of these overdensities. The pressure bumps are in geostrophic balance with sub- and super-Keplerian zonal flows. At 5 AU distance to the star 6H corre- spond to ∼ 1.5 AU. The amplitude of the density bump reaches 15% and goes down to about 10% in the largest simulation. 3. Particles with St = τfΩ = 1 are getting trapped efficiently by the axisymmetric pressure bumps. They accumulate in regions of minima in the sec- ond derivative of the gas density as predicted ana- lytically (e.g., Klahr & Lin 2001). The concentra- tion factor correlates with the correlation time of the zonal flows. Hence, the first two steps of plan- etesimal formation7 in protoplanetary disk with an acting MRI are: vertical settling via sedimentation 7 After coagulation from µm-sized particles to St = 0.1, 1. and radial concentration by trapping of dust in ax- isymmetric pressure bumps. Further concentration comes likely from stochastic processes. Clustering properties do not depend strongly on strength or lifetime of the zonal flows. 4. We reach dust-to-gas ratios of 50–100. These den- sities are of the order of the Roche density at 5 AU in an MMSN. The dust overdensities scale with the lifetime of the zonal flow structures by a power law with an exponent of 0.38 ± 0.05 (see Figure 8). To what degree these high dust-to-gas ratios disturb the axisymmetric pressure bumps that developed in the zonal flows has to be investigated in further studies with back-reaction to the gas. 5. Particles of only a few centimeters in size (at 5 AU in an MMSN, St = 0.1) accumulate in overdensi- ties that are increased by a factor of ∼ 100, lead- ing to a dust-to-gas ratio of 1 in the mid-plane, thus triggering the streaming instability. With- out MRI and zonal flows St = 0.1 particles do not clump strongly and cannot trigger the stream- ing instability for solar metallicity Z = ǫ0 = 0.01 (Johansen et al. 2009b). This is the first work on the effect from large-scale zonal flows on dust particles in an MHD simulation. Dust gets trapped downstream of long-lived high-pressure re- gions and achieves overdensities that have the potential to generate streaming instability and to become gravita- tionally unstable. Planetesimal formation in large boxes will be further investigated in simulations with particle feedback on the gas and self-gravitating particles in a future study. In the future, we will focus on one model and study various initial dust-to-gas ratios and particle size distri- butions. We will probably use the already converged run L (5.28H × 5.28H × 2.64H). This choice is also a trade-off between simulation box size and computational expense. K.D. and H.K. have been supported by the Deutsche Forschungsgemeinschaft Schwerpunktprogramm (DFG SPP) 1385 "The first ten million years of the solar sys- tem"; K.D. received further support by the IMPRS for Astronomy & Cosmic Physics at the University of Hei- delberg. A.J. was partially funded by the European Research Council under ERC Starting Grant agreement 278675PEBBLE2PLANET and by the Swedish Research Council (grant 20103710). Computer simulations have been performed at the THEO cluster at Rechenzentrum Garching and at the JUGENE cluster in Forschungszen- trum Julich (project number HHD19). We thank an anonymous referee for a very thorough referee report that improved the quality of the paper greatly. REFERENCES Andrews, S. M., Wilner, D. J., Hughes, A. M., Qi, C., & Dullemond, C. P. 2010, ApJ, 723, 1241 Balbus, S. A., & Hawley, J. F. 1991, ApJ, 376, 214 Balbus, S. A., & Hawley, J. F. 1998, RvMP, 70, 1 Beitz, E., Guttler, C., Blum, J., et al. 2011, ApJ, 736, 34 Birnstiel, T., Klahr, H., & Ercolano, B. 2012, A&A, 539, A148 Blum, J., & Wurm, G. 2008, ARA&A, 46, 21 18 Dittrich, Klahr, & Johansen Brandenburg, A., Nordlund, A., Stein, R. F., & Torkelsson, U. 1995, ApJ, 446, 741 Brauer, F., Dullemond, C. P., & Henning, T. 2008, A&A, 480, 859 Carballido, A., Bai, X.-N., & Cuzzi, J. N. 2011, MNRAS, 415, 93 Carballido, A., Fromang, S., & Papaloizou, J. 2006, MNRAS, 373, 1633 Cassen, P., & Moosman, A. 1981, Icar, 48, 353 Chapman, S., & Cowling, T. G. 1970, The Mathematical Theory of Non-uniform Gases. An Account of the Kinetic Theory of Viscosity, Thermal Conduction and Diffusion in gases (3rd ed.; Cambridge: Univ. Press) Cuzzi, J. N., Hogan, R. C., & Shariff, K. 2008, ApJ, 687, 1432 Desch, S. J. 2007, ApJ, 671, 878 Dominik, C., Blum, J., Cuzzi, J. N., & Wurm, G. 2007 in Protostars and Planets V ed. B. Reipurth, D. Jewitt, & K. Keil (Tucson, AZ: Univ. Arizona Press), 783 Dubrulle, B., Morfill, G., & Sterzik, M. 1995, Icar, 114, 237 Dzyurkevich, N., Flock, M., Turner, N. J., Klahr, H., & Henning, T. 2010, A&A, 515, A70 Fedele, D., van den Ancker, M. E., Henning, T., Jayawardhana, R., & Oliveira, J. M. 2010, A&A, 510, A72 Flock, M., Dzyurkevich, N., Klahr, H., Turner, N. J., & Henning, T. 2011, ApJ, 735, 122 2012, ApJ, 744, 144 Kopal, Z. 1989, The Roche Problem and its Significance for Double-star Astronomy (Astrophysics and Space Science Library, Vol. 152; Dordrecht: Kluwer) Krumholz, M. R., Klein, R. I., & McKee, C. F. 2012, ApJ, 754, 71 Lodders, K. 2003, ApJ, 591, 1220 Lyra, W., Johansen, A., Klahr, H., & Piskunov, N. 2008, A&A, Nakagawa, Y., Sekiya, M., & Hayashi, C. 1986, Icar, 67, 375 Pan, L., Padoan, P., Scalo, J., Kritsuk, A. G., & Norman, M. L. 479, 883 2011, ApJ, 740, 6 Pinilla, P., Birnstiel, T., Ricci, L., et al. 2012, A&A, 538, A114 Safronov, V. S. (ed.) 1969, Evoliutsiia doplanetnogo oblaka (English transl.: Evolution of the Protoplanetary Cloud and Formation of Earth and the Planets, NASA Tech. Transl. F-677, 1972,; Jerusalem: Israel Sci. Transl.) Shakura, N. I., & Sunyaev, R. A. 1973, A&A, 24, 337 Simon, J. B., Beckwith, K., & Armitage, P. J. 2012, MNRAS, Sorathia, K. A., Reynolds, C. S., Stone, J. M., & Beckwith, K. Stone, J. M., & Gardiner, T. A. 2010, ApJS, 189, 142 Tsiganis, K., Gomes, R., Morbidelli, A., & Levison, H. F. 2005, 422, 2685 2012, ApJ, 749, 189 Natur, 435, 459 85 von Neumann, J., & Richtmyer, R. D. 1950, JAP, 21, 232 Weidenschilling, S. J. 1977a, MNRAS, 180, 57 Weidenschilling, S. J. 1977b, Ap&SS, 51, 153 Weidenschilling, S. J. 1997, Icar, 127, 290 Whipple, F. L. 1972, in Proc. Twenty-First Nobel Symp. on From Plasma to Planet, ed. A. Evlius (New York: Wiley), 211 Windmark, F., Birnstiel, T., Guttler, C., et al. 2012a, A&A, 540, A73 Windmark, F., Birnstiel, T., Ormel, C. W., & Dullemond, C. P. 2012b, A&A, 544, L16 Youdin, A., & Johansen, A. 2007, ApJ, 662, 613 Youdin, A. N., & Goodman, J. 2005, ApJ, 620, 459 Youdin, A. N., & Lithwick, Y. 2007, Icar, 192, 588 Zsom, A., Ormel, C. W., Guttler, C., Blum, J., & Dullemond, C. P. 2010, A&A, 513, A57 Flock, M., Dzyurkevich, N., Klahr, H., Turner, N., & Henning, T. Uribe, A. L., Klahr, H., Flock, M., & Henning, T. 2011, ApJ, 736, Fromang, S., & Stone, J. M. 2009, A&A, 507, 19 Guan, X., Gammie, C. F., Simon, J. B., & Johnson, B. M. 2009, ApJ, 694, 1010 153 Haisch, K. E., Jr., Lada, E. A., & Lada, C. J. 2001, ApJL, 553, Hayashi, C. 1981, PThPS, 70, 35 Heinemann, T., & Papaloizou, J. C. B. 2009, MNRAS, 397, 64 Ida, S., Guillot, T., & Morbidelli, A. 2008, ApJ, 686, 1292 Johansen, A., Klahr, H., & Henning, T. 2011, A&A, 529, A62 Johansen, A., Oishi, J. S., Mac Low, M.-M., et al. 2007, Natur, Johansen, A., & Youdin, A. 2007, ApJ, 662, 627 Johansen, A., Youdin, A., & Klahr, H. 2009a, ApJ, 697, 1269 Johansen, A., Youdin, A., & Mac Low, M.-M. 2009b, ApJL, 704, 448, 1022 75 Klahr, H. H., & Lin, D. N. C. 2001, ApJ, 554, 1095
1309.3119	1	1309	2013-09-12T11:30:50	A Survey of Low-Velocity Collisional Features in Saturn's F Ring	[ "astro-ph.EP" ]	Small (~50km scale), irregular features seen in Cassini images to be emanating from Saturn's F ring have been termed mini-jets by Attree et al. (2012). One particular mini-jet was tracked over half an orbital period, revealing its evolution with time and suggesting a collision with a local moonlet as its origin. In addition to these data we present here a much more detailed analysis of the full catalogue of over 800 F ring mini-jets, examining their distribution, morphology and lifetimes in order to place constraints on the underlying moonlet population. We find mini-jets randomly located in longitude around the ring, with little correlation to the moon Prometheus, and randomly distributed in time, over the full Cassini tour to date. They have a tendency to cluster together, forming complicated `multiple' structures, and have typical lifetimes of ~1d. Repeated observations of some features show significant evolution, including the creation of new mini-jets, implying repeated collisions by the same object. This suggests a population of <~1km radius objects with some internal strength and orbits spread over 100km in semi-major axis relative to the F ring but with the majority within 20km. These objects likely formed in the ring under, and were subsequently scattered onto differing orbits by, the perturbing action of Prometheus. This reinforces the idea of the F ring as a region with a complex balance between collisions, disruption and accretion.	astro-ph.EP	astro-ph	A Survey of Low-Velocity Collisional Features in Saturn's F Ring Nicholas O. Attree1,∗, Carl D. Murray1, Gareth A. Williams1, Nicholas J. Cooper1 School of Physics and Astronomy, Queen Mary University of London, Mile End Road, London E1 4NS, UK 3 1 0 2 p e S 2 1 . ] P E h p - o r t s a [ 1 v 9 1 1 3 . 9 0 3 1 : v i X r a Abstract Small (∼ 50km scale), irregular features seen in Cassini images to be emanating from Saturn's F ring have been termed mini-jets by Attree et al. (2012). One particular mini-jet was tracked over half an orbital period, revealing its evolution with time and suggesting a collision with a local moonlet as its origin. In addition to these data we present here a much more detailed analysis of the full catalogue of over 800 F ring mini-jets, examining their distribution, morphology and lifetimes in order to place constraints on the underlying moonlet population. We find mini-jets randomly located in longitude around the ring, with little correlation to the moon Prometheus, and randomly distributed in time, over the full Cassini tour to date. They have a tendency to cluster together, forming complicated 'multiple' structures, and have typical lifetimes of ∼ 1d. Repeated observations of some features show significant evolution, including the creation of new mini-jets, implying repeated collisions by the same object. This suggests a population of . 1km radius objects with some internal strength and orbits spread over ±100km in semi-major axis relative to the F ring but with the majority within 20km. These objects likely formed in the ring under, and were subsequently scattered onto differing orbits by, the perturbing action of Prometheus. This reinforces the idea of the F ring as a region with a complex balance between collisions, disruption and accretion. Keywords: Planetary rings; Saturn, rings; Saturn, satellites 1. Introduction The F ring region has long been thought home to a population of small bodies. Evidence ranges from the depletion of magnetospheric charged particles measured by Pioneer 11 (Cuzzi and Burns, 1988) to more recent stellar occultation results from Cassini VIMS and UVIS (Meinke et al., 2012) as well as direct imaging by both Cassini ISS (Beurle et al., 2010) and the Hubble Space Telescope (McGhee et al., 2001). At least some members of this population interact with the F ring core, with objects like S/2004 S 6 (hereafter referred to as S6) thought to produce large jets by physical collisions with the core (Charnoz et al., 2005; Murray et al., 2008; Charnoz, 2009). These jets extend over hundreds of kilometers in semi-major axis and persist for many days or months with new jets forming as S6 re-collides to form a sequence of features. These evolve through Keplerian shear, wrapping around through more than 360 degrees of longitude, to form the kinematic spiral strands seen on either side of the core (Charnoz et al., 2005). The core also contains many small, irregular features and remains changeable on timescales ranging from hours to years (Porco et al., 2005; French et al., 2012), suggesting additional processes because Keplerian shear should quickly smooth out any longitudinal variations. Attree et al. (2012) identified many of these 'mini-jet' features, examining the time evolution of one, and proposing low velocity (∼ 1ms−1) collisions with local objects as the formation mechanism. Here we present a much more detailed study of the full population of mini-jets observed to date in an attempt to investigate the collisional theory and put constraints on the underlying moonlet population. ∗Corresponding author Email address: [email protected] (Nicholas O. Attree) Preprint submitted to Elsevier June 15, 2021 Section 2 describes how the features are catalogued with measurements made from Cassini ISS images. In Section 3 we detail the basic orbital theory needed to explain the motion of mini-jets and outline the governing equations. Section 4 describes work done to analyse the distribution in space and time, proximity to Prometheus and evolution of the observed features. Some interesting examples are also discussed in more depth. Further analysis and the implications for the moonlet population are discussed in Section 5 before our concluding remarks are presented in Section 6. 2. Observations A search of all Cassini ISS sequences containing resolved images of the F ring has been performed. As before (Attree et al., 2012) we identified by eye small features (typically ∼ 10 -- 200 km in radial extent and ≪ 1◦ in longitude) emanating from the core, excluding bright kinks and clumps within the core itself and the larger jet features. After pointing the images using background stars we re-project a portion of each image containing a feature in an equal aspect radius-longitude plot for direct comparison and then assigned them to one of three classes according to their morphology: (i) classic mini-jets, (ii) objects and (iii) complex features; each is described in detail below. Representative examples of each are shown in Figs. 1, 2 and 3. All co-ordinates are given in the standard Saturn centric reference frame with the x-axis corresponding to the position of the ascending node of Saturns equatorial plane on the mean Earth equator at the J2000 epoch. Co-rotating longitudes are referenced to an epoch of 12:00 UTC, 1st January 2007 using the mean motion in section 3 below. As of the end of 2012 we have catalogued 889 discrete features, including 32 which are imaged more than once. Many are found clustered close together in 'multiple' structures and we have tried to identify individual mini-jets within them, finding 56 of these. In some cases, however, this proved impossible due to poor resolution or the number of individual features (sometimes dozens) and these were placed in the 'complex' category below. Some examples of multiple features are shown in Fig. 4. Both repeated and multiple features are of special interest and are discussed in more detail, in Section 4.3, below. The full catalogue is available online as supplementary material A. Entries are named by the Cassini image number that they appear in while multiple features in the same image are labelled alphabetically e.g. N1557026084a, N1557026084b etc. 2.1. Classic Mini-Jets Features with a clearly defined linear structure at a measurable angle from the longitudinal direction are defined as classic mini-jets like the original feature discussed in Attree et al. (2012). As shown in Fig. 1 they are seen at a variety of angles and lengths and a significant sub-category has been identified as 'bright-heads'; these are classic mini-jets with a particularly bright tip or head. Such a head may represent the colliding object itself or simply a concentration of ejecta at that location. 437 mini-jets, including 37 bright-heads have been found. 2.2. Objects 'Objects' have been identified as bright features separate from, but close to, the F ring core. As seen in Fig. 2 most are extended longitudinally in a similar way to S6 and some are connected to the core by a faint dust sheet or, in some cases, a mini-jet-like linear trail. The dividing line between objects with a faint trail and mini-jets with a bright head is therefore somewhat blurred and it is possible that these actually represent the same phenomenon imaged at different geometries or orbital phases. 207 objects have been identified. 2.3. Complex Features As shown in Fig. 3 some features are difficult to fit into either of the above classes and are instead described as 'complex'. Some are most likely mini-jets or objects imaged at poor resolution or difficult ge- ometries but we suggest that many are more complex features. A number of these appear in the very highest resolution images, for example there are 14 in one early sequence (ISS 029RI AZSCNLOPH001 PRIME) with 2 just over 1km radial resolution and 10−3 degrees longitudinal resolution. These may reveal a complicated substructure present in all mini-jets but not normally resolved. Several others have the appearance of a superposition of several classic mini-jets on top of one-another and resemble close multiple features. These are placed in the complex category when the resolution is insufficient to separate the individual components. Altogether we have found 245 complex mini-jets. 3. Theory We define the variables ∆a and ∆e as the differences in semi-major axis and eccentricity, respectively, from an orbit with a certain a and e. Mini-jets similar to the original feature consist of material with a range of ∆a and ∆e, with ∆a ≈ a∆e, ranging from 0 at the base up to some maximum at the tip (Attree et al., 2012). Epicyclic theory (see e.g. Murray and Dermott, 1999) predicts that, in a frame co-rotating with our start point (with a,e), particles in the mini-jet will each follow their own 2 : 1 centred ellipse with the addition of a linear drift due to Keplerian shear. All the particles move in phase to create the coherent linear structure. Radial, r, and longitudinal, l, positions in kilometers relative to the base will, to first order, follow r = ∆a − a∆e cos M, l = 2a∆e sin M − 3/2M ∆a (1) where we take a = 140221.3km as the semi-major axis of the F ring core (Albers et al., 2012), M = nt is orbital phase in radians, with n = 10.157 radians per day the mean motion, and t is time. The particles start off at periapse for mini-jets with positive ∆a and aposapse for those with negative ∆a. As shown in a simple schematic (Fig. 5) the tangent of the angle that the mini-jet makes with the horizontal is simply the ratio of these two lengths, i.e. θ = tan−1 1 − cos M 2 sin M − 3/2M (2) where, as noted above, we use the relationship ∆a ≈ a∆e. Equation (2) is plotted in Fig. 6 and describes the time evolution of the angle of any mini-jet, irrespective of size, as long as the above relation holds. Note that this is the same as the cant angle from Tiscareno et al. (2013). Also shown in Fig. 6 are the measured angles for the original feature (Attree et al., 2012) plotted against time, with the start time shifted to best fit the curve. The image numbers and measured co-ordinates of the mini-jet tip and base (from which the angles were calculated) can be found in supplementary information B. The scale of the mini-jet is determined by the speed of the collision which formed it; the maximum ∆a being therefore related, by the standard perturbation equations, to the velocity 'kick', ∆v, delivered to the particles at the tip. Under certain assumptions (dissipative collisions between F ring particles and a much more massive impactor, see Attree et al. (2012)) this is of the same order as the collision velocity, U, between the impacting object and the target (the F ring core or strand). The component of ∆v in, or opposite to, the direction of orbital motion dominates over any radial components so we can simplify the perturbation equations to (3) U ∼ ∆v ≈ n∆a 2 In essence the tip of the mini-jet is put onto an orbit approximating that of the colliding object while the rest of the material is dragged out between this and the F ring. It seems likely that the complex features are also collisional in nature but with more complicated ejecta patterns due to multiple objects or differing collision geometries. We have yet to develop a general theory for their time evolution and instead analyse them on an ad-hoc basis. 4. Analysis For classic features we measured the radial and longitudinal co-ordinates of the tip and base of the mini-jet and computed the lengths r and l and the angle θ from these. For complex features and objects 3 we measured the co-ordinates of the center of the feature (labeled base co-ordinates in supplement A) and the center of the object (labeled tip co-ordinates), respectively. Uncertainties in these measurements vary with image geometry and resolution, with a typical pixel being 5km in the radial and 0.01◦ in the longitudinal direction. This pixel error is carried forward when calculating the angle θ but the blurred, extended appearance of most features means it is an underestimate and we instead estimate typical errors by eye to be ∼ ±2◦. In the analysis below the derived quantities should then be taken as order of magnitude estimates. 4.1. Distribution A total of 857 unique features (see repeats in Section 4.3 below) have been found in 110 image sequences. The number in each sequence is highly variable, ranging from zero to 47, with a median of 5 and a mean and standard deviation of 8 ± 8.6 respectively. Features are found on both sides of the core and in both inner and outer strands (∼ 60 in the strands) and at a range of phase angles (13◦ − 163◦) and viewing geometries. Examining the numbers of features at each longitude in a frame co-rotating with the F ring revealed no noticeable trends (in an inertial frame the number is biased to where Cassini makes its observations). Likewise, when examining the Fourier transform of the distribution no traces or periodicity were detected. No particular trends were detected within the subclasses of feature and no particular differences between their distributions have been noted. Mini-jets and similar features, then, are distributed randomly around the F ring, as might be expected from a stochastic, collisional process. The average separation between adjacent features in the same sequence is 22◦ ± 23.0◦, where the error is the standard deviation which is large because the number of features in each sequence is so variable. We tested for clustering by taking the ratio of this to the expected 'degrees per feature' (the sequence coverage divided by number of features seen in it) and found that for nearly all sequences this was around one (median for the whole data set of 1.17). A value of one would be expected for no clustering as the observed average distance would match that of uniformly spaced features (e.g. following a Poisson distribution with a mean separation). On the large scale at least, the distribution is fairly uniform around the ring. On the small scale, however, there are over 50 'multiple' features where the separation between adjacent features is ≪ 1◦, much less than the average. This suggests that the components of 'multiples' are related to each other rather than being randomly distributed mini-jets that happened to be seen close together. For features with a mean separation of ∼ 20◦ the probability of finding two < 1◦ apart is given by Poisson statistics as negligible (∼ 10−9) even with over 800 features. Finally, in terms of spatial distribution, the longitude of each feature relative to Prometheus is plotted in Figure 7. The mean number of features in each 2◦ bin is 5 ± 2.66. The distribution is generally random and noisy but, with half of the bins within 10 degrees of Prometheus more than one standard deviation less than the mean, there is a slight dip in the number of features here. One might expect mini-jets to be more difficult to observe in the highly disturbed streamer-channel region and this may account for the dip, however there is some evidence that the decrease persists upstream of Prometheus, where the ring should be relatively undisturbed. In this case the feature would be real although the physical mechanism for a depletion in collisional features near Prometheus is, as yet, unexplained (one possibility is a phase-lag effect as in the 'predator-prey' model of Esposito et al., 2012). When considering the scatter in the rest of the data set we suggest that the decrease is around the level of the noise and any trend here is weak at best. The time distribution of the features was then examined by plotting the total number in each observation sequence, adjusted by longitude coverage, against date. This is shown in Fig. 8. Adjusting for the varying amount of the F ring covered in each sequence is done by dividing the number of features by the co-rotating longitude coverage and then multiplying by 360 degrees to get the expected number in the whole F ring at any one time. This is reasonable if features are distributed randomly so that all longitudes should contain a similar numbers of mini-jets. Only sequences with a co-rotating longitude coverage of > 50◦ are deemed representative of the whole ring and are included in this plot. A similarly shaped graph to Fig. 8, with the same trends, is seen when plotting just the raw number of features over time suggesting the validity of the adjustment method. The mean and standard deviation of the adjusted number of features in these sequences is 14 ± 13.1 respectively. This agrees with the median, adjusted number for the whole data set of 16 and is just within the limits of the mean of the raw count, quoted above, which is weighted downwards 4 by the large number of small longitude coverage observations with a single feature. In addition 16 features in the whole F ring would match the average of 22◦ per feature mentioned above. Thus we take ∼ 15 as our average for the number of features visible in the F ring at any one time. The symbols in Fig. 8 are coded by the range from the spacecraft to the ring as a measure of resolution, we find only a weak trend for more features at better resolution. Indeed the number of features is highly variable, as seen in the large standard deviation, and not particularly well correlated with any other observing statistic. This is partly because mini-jets are only visible at certain points in their orbit, depending on phase and geometry, and partly because of the difficulties of visual identification, especially when confronted with multiple features. The date of closest approach between the F ring and Prometheus (due to their eccentricities and differing precession rates) is also highlighted in Fig. 8 and there is a hint of an increase in the number of features in the year leading up to it. Unfortunately there is little coverage of the F ring at the actual closest approach and after it and any increase preceding it must be treated with caution because of the large number of observations here. The 2008/09 observations are among the best in terms of coverage, number and resolution and while we have tried to remove any biases by the methods above it should, perhaps, come as no surprise that the best observations produced the most entries in the catalogue. As above, when considering the large scatter, we suggest that the increase in 2008/09 is at the level of the noise (6 sequences above 2 standard deviations and none above 3) and any trend for greater numbers of mini-jets here is weak at best. 4.2. Morphology For the subset of 437 'classic' and 'bright-head' mini-jets we measured the lengths r and l and the angle θ. We then used a graphical method to find all the points where the curve from Eqn. (2) intersected this measured angle, each one representing a possible solution. Equations (1) and (3) were then used to find the ∆a and an approximate collision velocity for each solution. As can be seen in Fig. 6 for small angles (< 8.58◦) there can be many such possible solutions as the mini-jet could be collapsing on its first cycle or rising on its second or subsequent cycle. There is some degeneracy here e.g. a large, young mini-jet can appear similar to a small, old one. We computed an age and ∆a for each solution and then used the second part of Eqn. (1) to predict a length l for each. These predicted lengths were then compared to the measured length and the best fit was chosen as the preferred age/∆a solution. In effect this is a best guess at the real properties of the mini-jet and is the best that can be done from a single image. We checked for self-consistency by comparing the predicted lengths to the maximum and minimum lengths that a mini-jet of that ∆a and number of cycles could be and found all chosen solutions to fit. Nonetheless we still found a small number (18) of derived ∆a values to be very large and consider these outliers to be poorly fitted by the above method. These were all young, 'wrong-way' (against the direction of Keplerian shear) or large, jet-like, features which are difficult to measure but the fact that they are all much too large (hundreds of km) may mean we have misunderstood the early phases of mini-jet formation. Figure 9 shows the distribution of measured angles as well as a typical predicted distribution for com- parison. This was created by generating a population of mini-jets with random ages (weighted by number of cycles as found from the best fit ages) then using Eqn. (2) to produce a predicted angle for each. This serves to highlight the observational bias inherent in detecting mini-jets by eye: we find many more at medium angles (10◦ − 50◦) and many fewer at low angles (< 10◦) than predicted, simply because low angles are harder to spot in amongst the irregular F ring core. Apart from this discrepancy the measured angles are consistent with a population of mini-jets with ages ranging from a few hours to a few days although fewer young, 'wrong-way' mini-jets are detected than might be predicted. This is probably because they are very difficult to measure but again see the above comment on outliers. Figure 10 shows the best fit solutions for ∆a omitting these outliers. Nearly all the mini-jets derived with confidence lie within ∆a = 100km of the core with roughly equal numbers positive and negative and a mean absolute ∆a = 24 ± 17.2km (error is one standard deviation). This is a collision speed of U ∼ 1.2ms−1. The corresponding ages have a mean of 1.27 orbital periods or 0.79 days and a little over a quarter (125/437) of 5 mini-jets appear to have survived more than one cycle suggesting ∼ 3/4 are disrupted when re-entering the core at the end of their first cycle. 4.3. Multiple and Repeated Features The multiple mini-jets in Fig. 4a and c have similar angles, θ, and similar sizes implying that they formed at roughly the same time with the same collision velocities. A cluster of objects on very similar orbits, but stretched out in longitude by Keplerian shear, would impact the ring in a chain creating just such a series of mini-jets. We consider these to be evidence for groups of objects or objects disrupted into swarms. By contrast Fig. 4d to g have multiple mini-jets of different angles (ages) where the tips of each jet all seem to point to the same spot (Fig. 4d has three features showing both types). This would be expected if an object survives to collide with the core again at the next loop (see Fig. 5) creating a second mini-jet, in phase with, and following the tip of the first. Thus the objects in Fig. 4d to g may be re-colliding one orbital period later. Below we discuss repeated observations of the same features which further supports this idea, Nearly all of our catalogued features are seen in only 2 to 3 images as the F ring rotates through the ISS field of view over the course of a few minutes. However some observation sequences are taken within hours or days of each other and cover the same section of ring (i.e. the same range of co-rotating longitude) raising the possibility of repeated observations of the same feature. In principle this means more data points to fit to the curve in Fig. 6, allowing better constraints on mini-jet properties, whilst also directly measuring their lifetimes and any subsequent collisions. To this end we searched the catalogue by co-rotating longitude and date before comparing the morphology of features which are close in space and time. We have found 25 possible matches, imaged between ∼ 15 hours and ∼ 6.7 days apart, including 63 individual features which are listed in Table 1. There is no special distribution in time or longitude in those features which are repeated suggesting that it is only due to chance and good coverage that we managed to observe them. Visually the repeated features appear more sheared, as expected, being both longer (larger l) and at a lower angle (smaller θ), however they proved difficult to fit to the predicted evolution of θ. As noted in the table many are deemed complex with large uncertainties in their angles and morphologies and no particularly good fits were achieved, while classic mini-jets imaged more than one cycle apart also proved difficult to fit. This might imply that morphological changes occur when a mini-jet re-enters the core, e.g. the release of more ejecta or re-collision and displacement onto differing orbits. Another possibility is that the morphologically similar feature seen at a later date actually represents a new collision, at a similar geometry, after the first feature has dissipated. The extent to which colliding objects survive and go on to re-collide provides information about their physical properties and those of the core. Designation Date (year-DOY-time) θ(◦) Class Phase Exposure N1538204682 N1538283887 N1557026084a N1557026084b N1557080024 N1577828765a N1577828765c N1578409567a N1578409567b N1595326037 N1595501977 N1606006222 N1606033182 N1623224391a N1623224391b N1623332218a N1623332218b 2006-272-06:32:50.814 2006-273-04:32:55.309 2007-125-02:40:53.219 2007-125-02:40:53.219 2007-125-17:39:52.877 2007-365-21:09:57.025 2007-365-21:09:57.025 2008-007-14:29:55.038 2008-007-14:29:55.038 2008-203-09:29:05.444 2008-205-10:21:24.905 2008-327-00:10:55.278 2008-327-07:40:15.087 2009-160-06:58:20.565 2009-160-06:58:20.565 2009-161-12:55:27.199 2009-161-12:55:27.199 -- Extended Object -- Extended Object −46 Classic - bright head −18 Classic - bright head −11 Classic - bright head −1 Classic −4 Classic - bright head −2 Classic - bright head 0 Classic -- Extended Object −24 Classic −55 Classic −43 Classic - bright head −43 Classic −32 Classic −8 Classic −8 Classic 6 Angle(◦) 161 160 80 80 83 65 65 21 21 104 13 38 46 95 95 17 17 (ms) 680 680 1000 1000 1000 1200 1200 1000 1000 1500 100 1200 1200 1800 1800 1000 1000 Designation Date (year-DOY-time) θ(◦) Class Phase Exposure N1623224706a N1623224706b N1623332896a N1623332896b N1623225651a N1623333348a N1623225651b N1623360620 N1623226701 N1623335156b N1623329280d N1623355648 N1623330636b N1623357230a N1623330636c N1623357230b N1623342840 N1623369434 N1623345778b N1623372372 N1629353989 N1629517865 N1729028870 N1729057850 N1729030994 N1729059974a N1729059974b N1729035773 N1729064753 N1729037366a N1729037366b N1729259467a N1729259467b N1729259467c N1729043207 N1729072187 N1731113813 N1731140093 N1731126143 N1731152834 N1731129842a N1731129842b N1731157355 N1734573588 N1734595367a N1734595367b 2009-160-07:03:35.563 2009-160-07:03:35.563 2009-161-13:06:45.194 2009-161-13:06:45.194 2009-160-07:19:20.556 2009-161-13:14:17.190 2009-160-07:19:20.556 2009-161-20:48:48.597 2009-160-07:36:50.548 2009-161-13:44:25.178 2009-161-12:06:29.219 2009-161-19:25:56.632 2009-161-12:29:05.210 2009-161-19:52:18.621 2009-161-12:29:05.210 2009-161-19:52:18.621 2009-161-15:52:29.123 2009-161-23:15:42.534 2009-161-16:41:27.102 2009-162-00:04:40.513 2009-231-05:37:35.278 2009-233-03:08:50.432 2012-289-20:54:33.308 2012-290-04:57:33.424 2012-289-21:29:57.295 2012-290-05:32:57.411 2012-290-05:32:57.411 2012-289-22:49:36.264 2012-290-06:52:36.380 2012-289-23:16:09.254 2012-289-23:16:09.254 2012-292-12:57:49.144 2012-292-12:57:49.144 2012-292-12:57:49.144 2012-290-00:53:30.217 2012-290-08:56:30.333 2012-314-00:03:23.367 2012-314-07:21:23.200 2012-314-03:28:53.288 2012-314-10:53:44.119 2012-314-04:30:32.265 2012-314-04:30:32.265 2012-314-12:09:05.090 2012-354-01:05:56.394 2012-354-07:08:55.255 2012-354-07:08:55.255 Angle(◦) 95 95 17 17 94 17 94 34 93 17 17 34 17 34 17 34 19 34 19 35 119 126 138 155 138 155 155 139 156 139 139 71 71 71 140 158 152 145 155 148 156 156 149 96 91 91 (ms) 1800 1800 1000 1000 1800 1000 1800 1800 1800 1800 1000 1800 1800 1800 1000 1000 1000 1800 1800 1800 1200 560 1800 1200 1800 1200 1200 1800 1200 1800 1800 1200 1200 1200 1800 1200 1200 1200 1200 1200 1200 1200 1200 1200 1200 1200 −17 Classic −4 Classic −13 Classic −3 Classic -- Complex -- Complex -- Complex 26 Classic -- Complex -- Complex -- Extended Object -- Extended Object −40 Classic -- Complex -- Complex -- Complex -- Extended Object -- Complex -- Extended Object −23 Classic - bright head −13 Classic −5 Classic -- Extended Object -- Extended Object -- Complex -- Complex -- Complex −1 Classic −2 Classic −30 Classic −17 Classic −21 Classic −6 Classic −9 Classic -- Complex −14 Classic -- Complex -- Complex −1 Classic 0 Classic -- Complex −14 Classic -- Complex −54 Classic −12 Classic −3 Classic 7 Designation Date (year-DOY-time) θ(◦) Class Phase Exposure Angle(◦) (ms) Table 1: Repeated features, grouped together (each set separated by horizontal lines) and ordered by time of first appearance. Some new features have appeared on the second observation and these are given new designations, whilst others have disappeared or merged. The time between observations ranges from ∼ 15 hours to ∼ 6.7 days and is noted in Day Of Year format in the second column. Classes are determined by appearance as described in the observation section and the observation phase angle and camera exposure time are also noted. The angle θ is measured from the tip and base co-ordinates as described above and errors are generally ∼ 2◦. We now examine one particular repeated feature, seen in May 2007, which presents an interesting chal- lenge for mini-jet theory. Seen at first (top left of Figure 11) as a large double mini-jet with a bright head, it has evolved to a triple mini-jet by the time of the second image, approximately one orbital period later (bottom left of Fig. 11). Also visible in Fig. 11 are a number of dark, sheared channels in the strands either side of the core up- and down-stream of the mini-jets, two in the first image and four in the second. We believe all these features can be explained by multiple collisions with a single object if, on colliding with the dense, clumpy core, it creates a mini-jet but while moving through the more diffuse strands it merely sweeps up or scatters material leaving behind a dark channel. We note that this kind of behavior can also occur with the larger jets and S6 (Murray et al., 2013). The right-hand panels of Fig. 11 show corresponding frames from an animation of a simple model where mini-jets and channels are generated artificially to match the angles and locations of the observed features. Mini-jets are created by assigning a ∆a and a∆e to particles with a random number generator, up to a maximum of the object's orbit (below), which then evolve according to Eqns. (1) and (2). Channels are likewise generated, but without any a∆e, as particles which are coloured white. A single object will pass through these locations in the right sequence at roughly the correct time to form these features if it has ∆a ≈ 25km, a∆e ≈ 110km and ∆i ≈ 10−3 degrees, where ∆i is the inclination relative to the core. For the object to pass through strands both inside and outside of the core it must have a relative eccentricity greater than its ∆a, i.e. ∆a 6= a∆e contrary to what has been assumed for all mini-jets above. As far as we can tell this is not ruled out by the dynamics but could have implications for the derived mini-jet ages if some of them deviate from Eqn. (2). Indeed it may help explain the poorly fitted outliers mentioned above if these are, in fact, eccentric mini-jets on the 'wrong' side of the ring. This relative eccentricity also means that the object will pass through the core twice per cycle and a relative inclination is needed to explain why only one mini-jet is formed: the object loops above or below (i.e. out of the orbital plane) the core in one direction and collides with it in the other. The three dimensional nature has not been considered until now because inclinations are very small (i = 0.0067◦) in the F ring but collisions and gravitational interactions should induce small inclination differences between objects. To a first approximation one might expect Prometheus to induce inclinations comparable to its own of i = 0.008◦. An inclination difference of only ∆i = 8.17×10−4 degrees is needed to produce a vertical offset of (converting from radians) z ≈ 180a∆i/π = 2km at the F ring, enough for . 1km radius objects to 'miss' each-other and pass by without colliding. Even such an initially complex feature as this can then be accounted for by a single colliding object with a few reasonable assumptions. We present one final example of a mini-jet that has been imaged twice, one orbital period apart, in early 2013. This feature has not yet been included in the catalogue but is discussed here for comparison with the one above. The re-projected images are shown in Fig. 12 and the tip of the feature can be seen to have moved about 600km between frames. This, and the angles, are consistent with a ∆a ≈ 65km mini-jet, the theoretical path of which is superimposed on the figure. The tip and/or colliding object itself must have re-entered the F ring core in the time between images and one might expect a new mini-jet, in phase with the original, to have formed as above. Instead, visible in the second image is a dark, sheared channel in the 8 core located roughly at the point of re-collision. We suggest that this is a re-collision feature, as above, but one which happened to occur in an under-dense section of ring thus resulting in the sweep-up and scattering of fine grained material, leaving a depleted dark channel, rather than collision with a solid clump leaving a mini-jet. Finally we note that the appearance of the F ring before collision is uniform and bright so it is not immediately apparent that this is an under-dense region. This shows the difficulty in assessing the 'clumpiness' of the core from images. 5. Discussion The collisional theory proposed by Attree et al. (2012) remains consistent with the presented analysis and suggests a population of small bodies on orbits similar to the F ring's (∼ 10s of km difference in semi- major axis) interacting with it. These would have to be ∼ 1km or less in diameter in order not to be resolved in ISS imaging although the objects listed in the observations section may represent some of the larger ones or those with accompanying dust envelopes. Evidence exists for repeated collisions, suggesting that at least some of these objects survive passing through the F ring and so they must have some degree of strength, i.e. they are not just clumps of material on similar orbits. Repeated collisions do seem rare but whether this is due to weak objects or an unlikely combination of orbits in a dense region of the highly variable F ring is difficult to say. Multiple impact features are also relatively common, suggesting groups of objects moving together on related orbits. Similar complex structure is visible in the larger S6 collisions and is also highly suggestive of clusters of objects. These likely represent those objects that have been partially disrupted into a swarm of smaller objects, either by previous collisions with the core or by tidal forces. Keplerian shear swiftly stretches out an initial clump into a long, extended chain. Tiscareno et al. (2013) argue that elongated features observed in the A ring are dust clouds formed by impacts from cm- to m- sized meteoroids. They are fitted best as evolving purely by Keplerian sheer (rather than epicyclical motion combined with shear as in Eqn. (2)) and Tiscareno et al. (2013) suggest that this implies "the impact of a compact stream" of material from a broken up object in a similar way to our chains of mini-jets above. Though the dynamics are analogous we feel that the relative velocities and the presence of bright heads, objects and re-collision features presented in this paper suggest a local source for the F ring collisional population. We can can crudely estimate the number of colliding objects with a particle in a box method using N = 2πaHW ν σv , (4) where N is the required number of objects with average dispersion velocity v ≈ 1ms−1 and collisional cross section σ to result in a collision frequency of ν in a box of height H, width W and length 2πa (where a is semi-major axis as before), taken to be 50 × 50km. The frequency of mini-jet forming collisions must be ν ∼ 10 day−1 to sustain a visible population of that number, given their short lifespan of ∼ 1 day. The collisional cross section between a continuous ring of width x and a single spherical object of radius robj ∼ 500m is σ = πr2 obj + 2πax, assuming collisions are equally likely from all directions, i.e. taking the core as a flat 'ribbon' of width x all the way around its orbit. Taking x = 10km as the nominal core width results in N ≈ 30 objects. However the core is not uniform and simple, having dense clumps and variable width in longitude and time, not to mention a varying number of nearby strands. If we instead assume that significant features only form during collisions with a narrow, 1km, discontinuous inner core, containing the larger F ring particles (Murray et al., 2008) then N ≈ 290 objects are needed for the observed collision rate. If we further restrict mini-jet forming events to collisions between a finite number of discrete clumps or objects, rather than a continuous ring, we require more of these objects by several orders of magnitude; though the estimate will vary depending on how elongated the clumps are and how they move. Hence a more detailed knowledge of the collision process is needed to improve this estimate. Previous attempts by Charnoz (2009) to simulate the larger jet forming events have used two extremes: either a moonlet colliding with a continuous ring of massless particles or an unbound clump of massless 9 particles colliding with an embedded moonlet. Of these the second is preferred as it produces jets of material on either side of the core centered on the moonlet as is often observed (see Fig. 6 of Charnoz, 2009). However it cannot be fully correct as jets are also seen on only one side of the core at a time and this is nearly always the case with mini-jets. Furthermore, an unbound group of particles would disperse and not go on re-colliding as S6 and some mini-jet objects clearly do. The true situation may be somewhere between these two extremes with loosely bound aggregates colliding with a complex, clumpy core. For now, taking our number of features as a lower estimate because we may have missed low-angled mini-jets, we can say that the F ring region contains a steady population of at least ∼ 100 small objects interacting with the core. This would be a subset of the total number of objects implied by occultations which is estimated at several tens of thousands of objects greater than 100m in size (Meinke et al., 2012). Such a population would have to be continuously replenished as its members are eroded or broken down by collisions. This is further evidence for ongoing accretion in the F ring core as investigated by Canup and Esposito (1995) and Karjalainen (2007) among others. There is already evidence that Prometheus may aid in triggering clump formation (Beurle et al., 2010; Esposito et al., 2012) and perturb- ing objects onto their colliding orbits. Although collisional features are not correlated with Prometheus's location this does not rule out its influence as objects perturbed onto differing orbits would naturally spread around the ring. Because the tip of a mini-jet is placed on an orbit similar to the colliding object, Fig. 10 approximates the distribution of the object population in semi-major axis relative to the core. For comparison the maximum ∆a perturbation exerted by Prometheus is ∼ 20km when it and the F ring are at closest approach (Murray et al., 2008). This means multiple encounters are necessary to place some objects onto their mini- jet forming orbits, if indeed this is the mechanism. Subsequent encounters would be equally likely to perturb an object back towards the ring, depending on its exact orbit and the phase of closest approach, meaning objects could undergo a chaotic random walk in ∆a around the core. Likewise their solidity and size would evolve by collisions in a random way over time with only a lucky few aggregating into solid moonlets as described in Esposito et al. (2012). We conclude this to be the likely formation mechanism for S6 which must be a relatively young object. An overall picture then emerges of mini-jet forming objects, with some solidity and on orbits differing by tens of km from that of the core, being a subset of the total population of F ring objects. We consider these to be between temporary aggregates forming in the core and in the process of becoming larger more solid moonlets on more distant orbits like S6. Mini-jets and jets would then represent two ends of a continuum of collisional features formed by steadily increasing impact velocities from steadily increasing ∆a values. It appears that larger jets evolve purely under Keplerian shear rather than following Eqn. (2) and would therefore have ∆a 6= a∆e as would their progenitor objects. Prometheus's perturbations should preserve ∆a ≈ a∆e (Williams, 2009) so this might represent objects on more distant orbits freely precessing under Saturn's gravity rather than being locked to the F ring core. Further study of multiple Prometheus interactions and collisions on differing orbits will be needed to confirm this theory, 6. Conclusion Small, irregular features have been found all around the F ring, throughout the time that Cassini has observed it. Their numbers are highly variable, but average ∼ 15 in the ring at any one time. Based on our analysis of nearly 900 catalogued features they are randomly distributed but with a tendency to clump together in multiple structures. There may be fewer near the location of Prometheus and more in the years leading up to closest approach with this satellite but these trends seem weak at best. Those with a resolvable linear structure (mini-jets) have angles consistent with a range of ages, from a few hours to a few days, and with around a quarter being less than one orbital period (15 hours) old. This is also supported by the repeated detection of several features over the course of a few days, suggesting an average lifetime of ∼ 1 day. Many repeated features also show significant morphological changes between images, including the creation of new mini-jets, emphasising their extremely dynamic nature. Collisions with a local moonlet population present the most likely explanation and the lifetime and numbers mean that ∼ 15 mini-jet forming collisions must happen each day. Depending on assumptions 10 about the structure of the F ring core and the nature of the collisions this implies a population of order hundreds of . 1km radius objects. Some of these moonlets have enough strength to survive multiple collisions but others are disrupted into groups of smaller objects on similar orbits. We suggest that they are likely formed in the F ring itself, possibly due to the action of Prometheus, and are subsequently perturbed onto colliding orbits by further interactions. Those that survive may continue to grow and go on to form the larger visible objects such as S6. Mini-jets, therefore, represent one end of a continuum of collisional features with the other end being the large jets and spiral strands. Acknowledgments: This work was supported by the Science and Technology Facilities Council (grant number ST/F007566/1). References Albers, N., Sremcevic, M., Colwell, J.E., Esposito, L.W., 2012. Saturn's F ring as seen by Cassini UVIS: Kinematics and statistics. Icarus 217, 367 -- 388. Attree, N.O., Murray, C.D., Cooper, N.J., Williams, G.A., 2012. Detection of low-velocity collisions In Saturn's F ring. Astrophysical Journal Letters 755, L27. Beurle, K., Murray, C.D., Williams, G.A., Evans, M.W., Cooper, N.J., Agnor, C.B., 2010. Direct evidence for gravitational instability and moonlet formation in Saturn's rings. The Astrophysical Journal Letters 718, L176 -- L180. Canup, R.M., Esposito, L.W., 1995. Accretion in the Roche zone - Coexistence of rings and ringmoons. Icarus 113, 331 -- 352. Charnoz, S., 2009. Physical collisions of moonlets and clumps with the Saturn's F-ring core. Icarus 201, 191 -- 197. Charnoz, S., Porco, C.C., Deau, E., Brahic, A., Spitale, J.N., Bacques, G., Baillie, K., 2005. Cassini discovers a kinematic spiral ring around Saturn. Science 310, 1300 -- 1304. Cuzzi, J.N., Burns, J.A., 1988. Charged-particle depletion surrounding Saturns F-ring - evidence for a moonlet belt. Icarus 74, 284 -- 324. Esposito, L.W., Albers, N., Meinke, B.K., Sremcevic, M., Madhusudhanan, P., Colwell, J.E., Jerousek, R.G., 2012. A predator- prey model for moon-triggered clumping in Saturn's rings. Icarus 217, 103 -- 114. French, R.S., Showalter, M.R., Sfair, R., Argelles, C.A., Pajuelo, M., Becerra, P., Hedman, M.M., Nicholson, P.D., 2012. The brightening of Saturns F ring. Icarus 219, 181 -- 193. Karjalainen, R., 2007. Aggregate impacts in Saturn's rings. Icarus 189, 523 -- 537. McGhee, C.A., Nicholson, P.D., French, R.G., Hall, K.J., 2001. HST observations of saturnian satellites during the 1995 ring plane crossings. Icarus 152, 282 -- 315. Meinke, B.K., Esposito, L.W., Albers, N., Sremevi, M., 2012. Classification of F ring features observed in Cassini UVIS occultations. Icarus 218, 545 -- 554. Murray, C.D., Beurle, K., Cooper, N.J., Evans, M.W., Williams, G.A., Charnoz, S., 2008. The determination of the structure of Saturn's F ring by nearby moonlets. Nature 453, 739 -- 744. Murray, C.D., Cooper, N.J., Williams, G.A., Attree, N.O., 2013. The morphology and dynamics of Saturn's F ring. Icarus , in preparation. Murray, C.D., Dermott, S.F., 1999. Solar System Dynamics. Cambridge University Press. Porco, C.C., Baker, E., Barbara, J., Beurle, K., Brahic, A., Burns, J.A., Charnoz, S., Cooper, N., Dawson, D.D., Del Genio, A.D., Denk, T., Dones, L., Dyudina, U., Evans, M.W., Giese, B., Grazier, K., Helfenstein, P., Ingersoll, A.P., Jacobson, R.A., Johnson, T.V., McEwen, A., Murray, C.D., Neukurn, G., Owen, W.M., Perry, J., Roatsch, T., Spitale, J., Squyres, S., Thomas, P., Tiscareno, M., Turtle, E., Vasavada, A.R., Veverka, J., Wagner, R., West, R., 2005. Cassini imaging science: Initial results on Saturn's rings and small satellites. Science 307, 1226 -- 1236. Tiscareno, M.S., Mitchell, C.J., Murray, C.D., DiNino, D., Hedman, M.M., Schmidt, J., Burns, J.A., Porco, C.C., Beurle, K., Evans, M.W., 2013. Observations of ejecta clouds produced by impacts onto Saturns rings. Science (submitted) . Williams, G.A., 2009. The three body problem applied to close ring-satellite encounters. Ph.D. thesis. Queen Mary, University of London. 11 a g i o b c d e f h j k l m n p 100km Figure 1: Classic mini-jets re-projected in a radius/longitude frame to the same scale and ordered by age. Outwards mini-jets: (a) N1577813677a, (b) N1604028396, (c) N1733559846, (d) N1613003098, (e) N1597907705, (f) N1612005469, (g) N1623284964, (h) N1623331766c. Inwards mini-jets: (i) N1726901763a, (j) N1607629517, (k) N1605396128a, (l) N1616541581, (m) N1615488367, (n) N1610401148b, (o) N1727800548, (p) N1734593691. Contrast has been adjusted in each case to enhance visibility. 12 a d g b c e f h i 100km Figure 2: Objects re-projected in a radius/longitude frame to the same scale. (a) multiple objects N1616523599a,b, (b) N1501710723, (c) object which could be a bright head mini-jet N1623351880, (d) N1589120162, (e) N1549820347 (f) N1610593686 which could be a clump in the outer strand, (g) large, S6-like, object N1616506283 (artefact visible below this object), (h) N1589119327b, (i) N1618601283. Contrast has been adjusted in each case to enhance visibility. 13 a b d f c e g h 100km Figure 3: Complex features re-projected in a radius/longitude frame to the same scale. (a) numerous small mini-jets in N1537899083 and (b) N1537898708. Complex and poorly resolved mini-jets in (c) N1589620046 and (d) N1493639016 and possible object in (e) N1627640563. Complex mini-jets in (f) N1605531856, (g) N1623226701 and unknown feature in (h) N1601512734. Contrast has been adjusted in each case to enhance visibility. 14 a b e 100km c d f g Figure 4: Multiple features re-projected in a radius/longitude frame to the same scale. Mini-jets with similar ages (a) N1597902245a,b,c, (b) N1623224391a,b, (c) N1554046873a,b. Possible repeat collisions (c) N1729259467a,b, (d) N1615511698, (e) N1727132335a,b, (g) N1733524214a,b. Contrast has been adjusted in each case to enhance visibility. 50km l i s x A l r θ Longitudinal Axis i a d a R Figure 5: Schematic overlaid on a typical outwards mini-jet. The origin is centred on the point of collision and the white curve is the trajectory of the tip. The mini-jet starts out pointing the 'wrong way', against the direction of Keplerian shear, before quickly passing through the vertical and looping away down the ring. The colliding object has a similar orbit and will also roughly follow this path. 15 ) ) . . g g e e D D l l ( ( e e g g n n A A t t e e J J − − n n M M i i i i 100 100 50 50 0 0 −50 −50 −100 −100 0 0 1 1 2 2 3 3 Time (Orbital Periods) 4 4 Figure 6: Time evolution, in orbital periods (14.9 hours at the F ring), of θ, the angle in degrees that a mini-jet makes with the longitudinal axis. The angle passes through the vertical (denoted by the vertical dashed line) early in the first cycle before collapsing back into the core and then oscillating in and out of the core at ever smaller angles. The horizontal dashed line at θ = −8.56◦ marks the minimum of the second cycle. All mini-jets with angles less than this are on their first cycle while those with greater angles are degenerate and could be any number of cycles old. Crosses are the measured angles of the original feature over time (see Attree et al. (2012) and supplement B of this work for details) with errors ∼ 1◦. 16 s e r u a e F t f o r e b m u N 14 12 10 8 6 4 2 0 −100 0 100 Longitude Relative to Prometheus (Deg.) Figure 7: The number of features seen in 2◦ bins relative to Prometheus. There are no statistically significant trends. 17 s s e e r r u u a a e e F F t t f f o o r r e e b b m m u u N N d d e e t t s s u u d d A A j j 50 50 40 40 30 30 20 20 10 10 0 0 2004 2004 good resolution medium resolution poor resolution j e r u t c n u n o C s u e h e m o r P t f o e a D t 2006 2006 2008 2008 2010 2010 2012 2012 2014 2014 Year Figure 8: The number of features, per observation sequence with > 50◦ coverage, as a function of time. The number of features has been adjusted by longitude coverage as described in Section 4.1. There are no statistically significant trends. Good resolution refers to sequences imaged from a mean range of < 8 × 105km (radial resolution . 5km per pixel), medium is between 8 × 105 and 1.6 × 106km and poor is > 1.6 × 106km (radial resolution & 10km per pixel). 18 s t e J − n M i i f o r e b m u N 100 80 60 40 20 0 −50 0 Angle (Deg.) 50 Figure 9: Histogram of the measured mini-jet angles, θ, in 2◦ bins and a typical predicted distribution (shaded). See Section 4.2 for details. 19 s t e J − n M i i f o r e b m u N 40 30 20 10 0 −200 −100 0 100 200 Mini−Jet Δa (km) Figure 10: Histogram of the calculated mini-jet ∆a (semi-major axis of the mini-jet tip relative to its base) in 5km bins. 20 m k 0 0 1 1000km Figure 11: Multiple mini-jet feature imaged twice approximately one orbital period apart. Left: re-projections of Cassini NAC image N1557026084, taken at 2007-125-02:40:53.219, and N1557080024, taken at 2007-125-17:39:52.877. See Table 1 for further details. Right: corresponding frames from a formation model. A single object with a ∆a, ∆e and ∆i relative to the F ring core, loops along from left to right, creating a new mini-jet each time it moves through the core and sweeping out a dark channel on each pass through the strands. These subsequently shear to form the observed structure. 100km Channel Figure 12: Mini-jet imaged twice, one orbital period apart; the image numbers are top: N1739126746, taken at 2013-040- 17:51:25.327 and bottom: N1739180046, taken at 2013-041-08:39:44.988. The path of the tip/object is superimposed assuming ∆a ≈ 65km which matches the location of the bright head in both images and the dark channel in the core in the second. This then represents an object re-colliding with a sparsely populated piece of ring and sweeping up material rather than impacting to form a second mini-jet. Compare with Fig. 11 above. 21
1209.5397	2	1209	2012-12-15T05:20:16	From Dust to Planetesimals: Criteria for Gravitational Instability of Small Particles in Gas	[ "astro-ph.EP", "astro-ph.GA" ]	Dust particles sediment toward the midplanes of protoplanetary disks, forming dust-rich sublayers encased in gas. What densities must the particle sublayer attain before it can fragment by self-gravity? We describe various candidate threshold densities. One of these is the Roche density, which is that required for a strengthless satellite to resist tidal disruption by its primary. Another is the Toomre density, which is that required for de-stabilizing self-gravity to defeat the stabilizing influences of pressure and rotation. We show that for sublayers containing aerodynamically well-coupled dust, the Toomre density exceeds the Roche density by many (up to about 4) orders of magnitude. We present 3D shearing box simulations of self-gravitating, stratified, dust-gas mixtures to test which of the candidate thresholds is relevant for collapse. All our simulations indicate that the larger Toomre density is required for collapse. This result is sensible because sublayers are readily stabilized by pressure. Sound-crossing times for thin layers are easily shorter than free-fall times, and the effective sound speed in dust-gas suspensions decreases only weakly with the dust-to-gas ratio (as the inverse square root). Our findings assume that particles are small enough that their stopping times in gas are shorter than all other timescales. Relaxing this assumption may lower the threshold for gravitational collapse back down to the Roche criterion. In particular, if the particle stopping time becomes longer than the sound-crossing time, sublayers may lose pressure support and become gravitationally unstable.	astro-ph.EP	astro-ph	Criteria for Gravitational Instability of Small Particles in Gas From Dust to Planetesimals: Ji-Ming Shi1,2 and Eugene Chiang1,2,3 [email protected] ABSTRACT Dust particles sediment toward the midplanes of protoplanetary disks, forming dust-rich sub- layers encased in gas. What densities must the particle sublayer attain before it can fragment by self-gravity? We describe various candidate threshold densities. One of these is the Roche density, which is that required for a strengthless satellite to resist tidal disruption by its pri- mary. Another is the Toomre density, which is that required for de-stabilizing self-gravity to defeat the stabilizing influences of pressure and rotation. We show that for sublayers containing aerodynamically well-coupled dust, the Toomre density exceeds the Roche density by many (up to about 4) orders of magnitude. We present 3D shearing box simulations of self-gravitating, stratified, dust-gas mixtures to test which of the candidate thresholds is relevant for collapse. All our simulations indicate that the larger Toomre density is required for collapse. This result is sensible because sublayers are readily stabilized by pressure. Sound-crossing times for thin layers are easily shorter than free-fall times, and the effective sound speed in dust-gas suspen- sions decreases only weakly with the dust-to-gas ratio (as the inverse square root). Our findings assume that particles are small enough that their stopping times in gas are shorter than all other timescales. Relaxing this assumption may lower the threshold for gravitational collapse back down to the Roche criterion. In particular, if the particle stopping time becomes longer than the sound-crossing time, sublayers may lose pressure support and become gravitationally unstable. Subject headings: hydrodynamics -- instabilities -- planets and satellites: formation -- protoplanetary disks -- methods: numerical 1. INTRODUCTION Gravitational instability is an attractive mech- anism to form planetesimals, but how it is trig- gered in protoplanetary disks remains unclear. In one proposed sequence of events, most of the disk's solids first coagulate into particles 0.1 -- 1 m in size at orbital distances of a few AU. These "boulder"- sized bodies then further concentrate by the aero- dynamic streaming instability (Youdin & Good- man 2005; Johansen et al. 2007; Bai & Stone 1Department of Astronomy, UC Berkeley, Hearst Field Annex B-20, Berkeley, CA 94720-3411 2Center for Integrative Planetary Science, UC Berkeley, Hearst Field Annex B-20, Berkeley, CA 94720-3411 3Department of Earth and Planetary Science, UC Berkeley, 307 McCone Hall, Berkeley, CA 94720-4767 2010; and references therein). Local densities are so strongly enhanced by the streaming instability that they can exceed the Roche density (see §1.3 for a definition), whereupon collections of boul- ders may undergo gravitational collapse into more massive, bound structures. A weakness of this scenario is that it presumes that particle-particle sticking (i.e., chemical ad- hesion) can convert most of the disk's solids into boulders, or more accurately, particles whose mo- mentum stopping times in gas tstop ≡ mvrel Fdrag (1) are within a factor of 10 of the local dynamical time Ω−1, where Ω is the Kepler orbital frequency, m is the particle mass, vrel is the relative gas- particle velocity, and Fdrag is the drag force whose 1 form varies with disk environment (see, e.g., Wei- denschilling 1977). Figure 1 relates tstop to parti- cle radius s as a function of disk radius r in the minimum-mass solar nebula (MMSN). For r ≈ 1 -- 10 AU, the condition Ωtstop = 0.1 -- 1 corresponds to s ≈ 0.1 -- 1 m. Fig. 1. -- Stopping times of particles in the MMSN, normalized to the local dynamical time Ω−1. Disk parameters are taken from Chiang & Youdin (2010). Particles are assumed spherical with bulk density 1 g/cm3. The kinks in the curves are due to transitions between different drag force laws as taken from Wei- denschilling (1977; note that the transition between the Stokes and Epstein drag laws occurs when the gas mean free path equals 2/9 of the particle radius, not 4/9 as misprinted in that article). Marginally coupled particles (Ωtstop ∼ 1) correspond to meter-sized boul- ders at r ∼ 1 AU; decimeter-sized rocks at r ∼ 10 AU; and cm-sized pebbles at r ∼ 100 AU. The top panel plots Ωtstop at various fixed particle radii s; the bottom panel plots the same data but at fixed Ωtstop. In this paper we are interested in the small particle Ωtstop (cid:28) 1 limit. Unfortunately, particle-particle sticking might not produce boulders in sufficient numbers for the streaming instability to be significant. A 2 comprehensive study by Zsom et al. (2011; see also Birnstiel et al. 2010) found that for real- istic, experiment-based sticking models that in- clude both bouncing and fragmentation, particles no larger than ∼1 cm can form by sticking -- even when the disk is assumed to have zero turbulence. According to Table 1 of Zsom et al. (2011), coag- ulation models over most of parameter space pro- duce τs ∼ 10−4 -- 10−2. This range is too small for the streaming instability to concentrate parti- cles strongly -- see Bai & Stone (2010), who showed that when half or more of the disk's solid mass has Ωtstop < 0.1, densities enhanced by the streaming instability still fall short of the Roche density by more than a factor of 10. Even if particle-particle sticking could grow bodies with Ωtstop ∼ 0.1 -- 1 (e.g., Okuzumi et al. 2009, who neglected fragmen- tation), the disk's solids may not be transformed into such bodies all at once. Rather, boulders may initially comprise a minority on the extreme tail of the size distribution. Unless they can multiply from a minority to a majority within the time it takes for them to drift radially inward by gas drag (∼100 -- 1000 yr starting at 1 AU; Weidenschilling 1977), they threaten to be lost from the nebula by drag. We are therefore motivated to ask whether gravitational instability is practicable for parti- cles having realistically small sizes and concomi- tantly short stopping times, say Ωtstop (cid:46) 10−2. Smaller particles suffer the disadvantage that they are harder to concentrate; since they are well- entrained in gas, turbulence in the gas can loft particles above the midplane and prevent them from collecting into regions of higher density. The streaming instability provides one source of turbu- lence. Another driver of turbulence is the Kelvin- Helmholtz instability, caused by vertical velocity gradients which steepen as dust settles into a thin, dense "sublayer" at the disk midplane (Weiden- schilling 1980). Several recent studies (Chiang 2008, Lee et al. 2010a,b; see also Weidenschilling 2006, 2010) have measured the maximum sublayer densities permitted by the Kelvin-Helmholtz in- stability. Neglecting self-gravity, they found that dust-to-gas ratios between ∼2 -- 30 are possible in disks that are locally enriched in metallicity by fac- tors of 1 -- 4 above solar. Such local enrichment can be generated by radial drifts of particles relative to gas (see Chiang & Youdin 2010 for a review). -4-3-2-1012log Stopping time Ωtss = 100 cms = 10 cms = 1 cms = 0.1 cm0.1110100Disk radius r (AU)0.11.010.0100.0Particle size s (cm)Ωts = 1Ωts = 0.1Ωts = 0.01Ωts = 0.001For observational evidence of radial segregation of dust from gas, see Andrews et al. (2012). i.e., Are such enhancements in the local dust-to-gas ratio sufficient to spawn planetesimals? How high must dust + gas densities be before the effects of self-gravity manifest? Our paper addresses these questions in the limit Ωtstop (cid:28) 1, in the limit that particles are small enough to be well coupled to gas. In the next two subsections, we derive critical densities for gravitational instabil- ity in the cases of a pure gas disk (§1.1), and a disk composed of both gas and perfectly entrained (Ωtstop → 0) dust (§1.2). The two cases give re- markably different answers for dust-rich sublayers. In §1.3 we add two more densities from the liter- ature to the list of proposed criteria for gravita- tional collapse. Table 1 summarizes the various candidate threshold densities. In §2 -- §3, we present numerical simulations of 3D, self-gravitating, compressible flows of thin, dense sublayers of dust. We use these simulations to try to identify which of the proposed criteria (if any) is the most relevant for gravitational insta- bility. Section 4 summarizes our findings but also points out the limitations of our numerical simula- tions, which are restricted to the asymptotic limit Ωtstop → 0. We argue in §4.1 how finite but still small values of Ωtstop may lower the threshold for gravitational collapse. 1.1. Critical Density for Gravitational Instability in a Pure Gas Disk The usual criterion for gravitational instability in a razor-thin pure gas disk is expressed in terms of the dimensionless parameter Qg ≡ cgΩ πGΣg (2) where G is the gravitational constant, cg is the gas sound speed, and Σg is the gas surface den- sity (Goldreich & Lynden-Bell 1965; Toomre 1964; Toomre 1981). In (2), the Kepler orbital frequency Ω has been substituted for the radial epicyclic fre- quency. If Qg < Q∗ g = 1 , (3) the disk is gravitationally unstable to axisym- metric perturbations in the disk plane. The Q-criterion is a measure of the competition be- tween stabilizing pressure, stabilizing rotation, and de-stabilizing self-gravity (see, e.g., Binney & Tremaine 2008). When Qg > 1, horizontal perturbations having lengthscales < 2cg/GΣg are stabilized by pressure, while those having lengthscales > 2cg/GΣg are stabilized by rota- tion. When Qg equals 1, the first axisymmetric mode to become unstable to self-gravity has ra- dial wavelength 2cg/GΣg. And as Qg approaches 1 from above, the disk is increasingly susceptible to nonaxisymmetric perturbations which swing amplify (Goldreich & Lynden-Bell 1965). The criterion Qg (cid:46) Q∗ g for gravitational insta- bility can be translated into a criterion for the midplane density ρg0 (the subscript "0" denotes the initial midplane value). We define a disk half- thickness Hg using Σg ≡ 2ρg0Hg . We also define a half-thickness H† relation from vertical hydrostatic equilibrium: g using the usual (4) (5) g ≡ cg/Ω . H† Ordinarily Hg ≈ H† g and we would not bother to distinguish the two; however, we will later find cases where they differ by factors of several be- cause of the effects of dust, and thus we take care to separate the two lengths now. Upon substitu- tion of (4) and (5), the relation Qg (cid:46) Q∗ g is shown to be equivalent to ρg0 (cid:38) ρ∗ I = 1 2π 1 Q∗ g H† g Hg ρ† where we have defined a reference density1 ρ† ≡ M∗/r3 (6) (7) The ρ∗ with M∗ and r equal to the mass of the central star and the disk radius, respectively. g = 1, and H† I -criterion (6) is sometimes used (e.g., Lee et al. 2010a,b) to signal gravitational instability in dusty gas disks (with ρg0 replaced by the total dust + gas density ρd0 + ρg0, Q∗ g/Hg = 1). But using ρ∗ I for dust-gas mixtures is suspect be- cause the criterion does not account explicitly for the two-phase nature of such media. In the next subsection we make such an accounting to derive a substantially different criterion for gravitational collapse. 1In this paper, we will superscript critical threshold densities with ∗, and fiducial or reference quantities with †. 3 Candidate Critical Densities for Gravitational Collapse Table 1 Critical Density ρ∗ I ρ∗ Sekiya ρ∗ Roche ρ∗ II Value H† g Hg 1 2π 1 Q∗ g r3 ∼ 0.16 M∗ M∗ r3 (a) 0.60 M∗ r3 (cid:16) Σg Σd 1 2π Qg Q∗2 d 3.5 M∗ r3 (cid:17)2 H† r3 ∼ 2 × 104 M∗ M∗ r3 g Hg g for pure gas Comment Equivalent to Qg < Q∗ disks Required for the onset of an in- compressible, axisymmetric overstable mode Required by satellite to resist tidal dis- ruption by primary (b) Equivalent to Qd < Q∗ d for dust-rich sublayers in gas Reference This paper, equation (6) Sekiya (1983) Chandrasekhar (1987) This paper, equation (13) (a) Value is derived for Q∗ (b) Value is derived for Q∗ g = 1 and H † g /Hg = 1. d = 1, Qg = 30, Σd/Σg = 0.015, and H † g /Hg = 1. 1.2. Critical Density for Gravitational Inserting (10) into (8) and using Instability in a Dust-Rich Sublayer in the Limit Ωtstop → 0 For disks of gas and dust, gravitational instabil- ity should still be determined by the Q-criterion, except there is now the possibility that disk self- gravity is dominated by dust in a vertically thin sublayer at the midplane: ρg0 = ρ† 1 2πQg H† g Hg , (11) we solve for the total midplane density required for gravitational instability: ρ∗ II = ρ0 = ρd0 + ρg0 (cid:38) ρ∗ II 1 2π Qg Q∗2 (cid:18) Σg (cid:19)2 H† (cid:19)(cid:18) 1 ≈ 2 × 104ρ†(cid:18) Qg (cid:19)2(cid:32) (cid:33) (cid:18) 0.015 30 H† g Hg Σd d Q∗ d ρ† g/Hg 1 Σd/Σg (12) (13) (cid:19)2 . (14) local In (14), our normalizations for Qg and the bulk (height-integrated but to r) metallicity Σd/Σg derive from the MMSN at r = 1 AU (Chi- ang & Youdin 2010). For these parameter choices, the critical midplane density ρ∗ II is an astonish- ing five orders of magnitude greater than ρ∗ I . It is possible that real disks have masses and bulk metallicities enhanced over the MMSN by factors of a few, in which case ρ∗ II would be larger than ρ∗ by about three orders of magnitude. I Qd ≡ cdΩ πGΣd (cid:46) Q∗ d for instability. (8) where In using the dust surface density Σd in (8), we neglect the contribution of gas to the total sur- face density of the sublayer. Under typical cir- cumstances, the error accrued is small. In the limit Ωtstop → 0, the dust-gas mixture represents a colloidal suspension. In this suspen- sion, dust does not contribute to the pressure P -- which is still provided entirely by gas -- but instead adds to the inertia. In other words, P = ρgc2 g = (ρg + ρd)c2 d (9) by definition of cd, the speed of sound in the sus- pension: cd = cg√ 1 + µ (10) where ρg is the local gas density, ρd is the local dust density, and µ = ρd/ρg is the dust-to-gas ratio. In effect, dust increases the mean molecular weight of the gas. 4 1.3. Other Critical Densities 2. METHODS Another threshold density, already alluded to at the beginning of §1, is the Roche density: ρ∗ Roche = 3.5 M∗ r3 . (15) The Roche density is the density required for a strengthless, incompressible, fluid body in hydro- static equilibrium to resist tidal disruption, when in synchronous orbit at distance r about a star of mass M∗ (e.g., Chandrasekhar 1987). Yet another candidate threshold was proposed by Sekiya (1983), who found that when the mid- plane density exceeds ρ∗ Sekiya = 0.60 M∗ r3 , (16) the disk becomes susceptible to an unstable, in- compressible, axisymmetric mode in which in- plane motions generate out-of-plane bulges (i.e., an annulus that contracts radially becomes thicker vertically, and vice versa). The nonlinear outcome of this instability is not known, but Sekiya (1983) speculated that the dust sublayer might eventu- ally fragment on the scale of the wavelength of the overstable mode, and that dust particles might sediment toward the centers of fragments to form the first-generation planetesimals. Table 1 summarizes the four candidate thresh- old densities. For realistic parameters (Qg ∼ 10 -- 30; Σd/Σg ∼ 0.015 -- 0.15), the four densities obey (17) Roche (cid:28) ρ∗ II . ρ∗ I < ρ∗ Sekiya < ρ∗ The smallest three densities in this hierarchy are fixed multiples of the reference density ρ† = M∗/r3 (with coefficients ∼1/2π ≈ 0.16, 0.6, and 3.5, respectively). The last density ρ∗ in principle, be arbitrarily larger than ρ†; for typi- cal, astrophysically plausible parameters, it is 2 -- 4 orders of magnitude larger. II can, Which of the four densities in Table 1 is the most accurate predictor of gravitational collapse? In the next two sections, we describe numerical simulations performed in the Ωtstop → 0 limit that attempt to answer this question. We will find un- fortunately that the numerical expense of simulat- ing thin sublayers of dusty gas will force us into a parameter space where the difference between ρ∗ II and the other densities is not as large as it is in reality; we will have to make do with what we can. 5 2.1. Code We simulate hydrodynamic, self-gravitating, stratified flows in disks using Athena, configured for a shearing box, with no magnetic fields (Stone et al. 2008; Stone & Gardiner 2010). Dust is as- sumed to be perfectly aerodynamically coupled to gas so that they share the same velocity field v. The equations solved are: + ∇ · (ρv) = 0 , ∂ρ ∂t + ∇ · (ρgv) = 0 , ∂ρg ∂t + ∇ · (ρvv + P) = −ρ∇Φ −2ρ(Ωz) × v + 2qρΩ2xx − ρΩ2zz , ∇2Φ = 4πGρ , ∂ρv ∂t (18) (19) (20) (21) where ρ = ρg + ρd is the total density of the dust- gas suspension, P = P I is a diagonal tensor with components P = ρgc2 g as defined in equation (9) with constant cg (isothermal approximation), Ω is the mean (constant) orbital frequency, x points in the radial direction, z points in the vertical direc- tion, and Φ is the self-gravitational potential of the dust-gas mixture. We choose the shear parameter q = 3/2 for Keplerian flow. 2.1.1. Algorithms and boundary conditions Athena 4.0 provides several schemes for time integration and spatial reconstruction, and for solving the Riemann problem. Having experi- mented with various options, we adopted the van Leer algorithm for our dimensionally unsplit inte- grator (van Leer 2006; Stone & Gardiner 2009); a piecewise linear spatial reconstruction in the prim- itive variables; and the HLLC (Harten-Lax-van Leer-Contact) Riemann solver. To account for disk self-gravity, we use the routines written by Koyama & Ostriker (2009) and Kim et al. (2011) which solve Poisson's equation using fast Fourier transforms. Boundary conditions for our hydrodynamic flow variables (including density and velocity, but not the self-gravitational potential) are shearing- periodic in radius (x) and periodic in azimuth (y). For vertical height (z), we experimented with both periodic and outflow boundary conditions, and chose periodic boundary conditions to ensure strict mass conservation. When outflow boundary conditions were employed, mass was lost from the boundaries at early times and complicated the in- terpretation of our results. We verified that our results are insensitive to box height for sufficiently tall boxes; see §3 for explicit tests. The Poisson solver implements shear-periodic boundary conditions in x, periodic boundary con- ditions in y, and vacuum boundary conditions in z (Koyama & Ostriker 2009; Kim et al. 2011). In our simulations, self-gravity is dominated by dust, and our boxes are tall enough to contain the en- tire dust layer. Both vertical and radial stellar tidal gravity are included as source terms in the van Leer integrator. We further augmented the code to include dust in the limit of zero stopping time. In this limit, dust shares the same velocity field as gas, and con- tributes only to the mass density. In our modified version of Athena, two continuity equations are solved: one for the entire mixture (ρ = ρg + ρd, see equation 18), and one for the gas (ρg, see equa- tion 19). The dust density is given by the differ- ence (ρ− ρg). The remaining hydrodynamic equa- tions govern the dust-gas mixture (ρ), but with gas (ρg) contributing solely to the pressure P (see equation 20). For simplicity, we adopt an isother- mal equation of state so that P is related to ρg by equation (9) for constant cg. Isothermal flows are more prone to gravitational instability than adiabatic ones (Mamatsashvili & Rice 2010). We also modified the HLLC Riemann solver to accommodate our dust-gas mixture. Changes in- clude the following: (1) The speeds of the left, right, and contact waves are reduced by a factor (1 + µ)−1/2, where µ ≡ ρd/ρg is the local dust- to-gas ratio, to account for the added inertia from dust (see equation 10). (2) The pressure in the contact region is replaced by an equivalent but numerically more accurate form based on equa- tion (10.76) in Toro (1999). (3) When calculating left/right momentum fluxes, we ensure that only gas contributes to the pressure by using ρg and not ρ. (4) For the flux solver to predict the pres- sure and wave speeds, the left/right gas densities require specification. We therefore add a recon- struction process for the gas density which inter- polates cell-centered values to cell boundaries to second-order accuracy. Previous studies of dust in the perfectly coupled limit (Chiang 2008; Lee et al. 2010a,b) also intro- duced a static background radial pressure gradi- ent to mimic sub-Keplerian rotation of gas in a pressure-supported disk. We could also add the appropriate source term to the van Leer integra- tor. However, since our goal is to determine the minimum densities required for gravitational col- lapse and not to study vertical shear instabilities (i.e., the Kelvin-Helmholtz instability), we omit the background pressure gradient for simplicity. In many of our simulations, the dust layer at the midplane collapses vertically because it is gravita- tionally unstable. Because of our boundary condi- tions, "fresh" gas from outside the simulation box cannot enter into the box, and thus in the event of gravitational collapse toward the midplane, the topmost and bottommost regions of our simula- tion domain become evacuated. Low-density gas in those regions become increasingly easy to ac- celerate, and the code timestep shortens by orders of magnitude, effectively halting the simulation. The dramatic reduction in timestep is not a seri- ous limitation, as it usually occurs after the col- lapsing dust has attained some saturated state (see §3.2.1). In any case, we are more interested in the onset of gravitational instability than its nonlinear development. 2.1.2. Code tests The following test problems helped to validate our code. Linear wave propagation. -- We propagated a small-amplitude 1D wave in a medium with a uni- form background dust-to-gas ratio, with periodic boundary conditions, no background shear, and no gravity. We found the simulated wave speed matched the reduced sound speed calculated in (10). We chose our box to be one wavelength long, so that after one wave period, the wave crossed the boundaries and returned to its original posi- tion. With N = 128 grid cells and an initial (frac- tional) wave amplitude A = 10−4, we found the i ≈ 2 × 10−8, where deviation δq ≡ 1 q ∈ {ρd, ρg, ρ} and q0 represents the initial condi- tion. qi − q0 N N(cid:88) i=1 Dust cloud advection. -- We advected a Gaussian- shaped dust cloud in a 1D domain. The cloud oc- 6 cupied about half the size of the box and the code was run for one box-crossing time. With N = 256 grid cells, the root-mean-squared deviation in the shape of the cloud was < 1%. Hydrostatic equilibrium of a stratified but non- self-gravitating dusty disk. -- Omitting self-gravity but including stellar gravity (both radial and ver- tical), we set up 3D dust-gas mixtures in hydro- static equilibrium. A variety of vertical profiles for the dust-to-gas ratio were tested, ranging from uniform to linear to more complicated functional forms. All equilibria were found to be stable against small perturbations, even for dust-to-gas ratios as large as several hundred. Gravitational instability of 3D pure gas disks. -- We simulated isothermal, gravitationally unstable disks of pure gas in 3D. The gas was initialized in hydrostatic equilibrium (computed with verti- cal self-gravity), and box heights spanned approx- imately ±4 initial gas scale heights. We found that Qg = 1 did not trigger gravitational insta- bility, whereas Qg = 0.5 did. Our results are consistent with those of Goldreich & Lynden-Bell (1965), who found analytically that Q∗ g = 0.676 for a finite-thickness isothermal gas disk. 2.2. Initial Conditions and Run Parameters Initial conditions for our science simulations are of a dust-gas mixture with a pre-defined vertical profile for the dust-to-gas ratio µ(z) = ρd(z)/ρg(z). We choose the form µ(z) ≡ µ0 sech 2 , (22) (cid:18) z (cid:19) zd where µ0 is the midplane dust-to-gas ratio. The scale height zd can be thought of as the half- thickness of the dust layer insofar as ρg(z) is con- stant with z. The isothermal dust-gas mixture is initialized in vertical hydrostatic equilibrium, including both stellar tidal gravity and disk self-gravity: c2 g ρg + ρd dρg dz = −Ω2z − 4πG (ρg + ρd)dz. (23) (cid:90) z 0 We solve numerically the differential form of (23). (cid:21) (cid:21) (cid:20) (cid:20) Taking derivatives, we find d dz (1 + µ)−1 d ln ρg dz = − 1 † H g − 4πG c2 g ρg(1 + µ), (24) g ≡ cg/Ω is a fiducial (constant) gas scale where H† height, not to be confused with any actual disk scale height. A non-dimensional form of (24) is given by d dz (1 + µ)−1 d ln ρg dz = −1 − 2 hgQg ρg(1 + µ), (25) where we have defined the dimensionless variables z ≡ z/H† g, ρg ≡ ρg/ρg0 (where ρg0 is the midplane gas density), and hg ≡ Hg/H† g, with Hg ≡ Σg/(2ρg0) . (26) Upon insertion of (22), equation (25) can be solved numerically for ρg(z). But the solution must satisfy the following two constraints: hg = (cid:90) ∞ (cid:90) ∞ 0 Σd Σg = 1 hg 0 by definition of Hg, and ρg(z) dz (27) ρg(z)µ(z)dz (28) for a fixed height-integrated (i.e., bulk) metallicity Σd/Σg. Our procedure is as follows. We freely specify Qg, Σd/Σg, and µ0 as model input parameters. We then iteratively solve equations (25), (27) and (28) for the three unknowns ρg(z), hg, and zd. First we guess zd and hg, and integrate (25) to obtain ρg(z). If ρg so calculated fails (27), then we revise hg and re-integrate (25), repeating un- til (27) is satisfied. Next we check (28). If ρg(z) and hg fail (28), then we revise zd and repeat the procedure from the beginning, re-integrating (25) to obtain ρg, re-establishing (27), and so on. Typ- ically ∼100 iterations (∼10 for zd × ∼10 for hg) are required before all constraints are satisfied to ∼1% accuracy in zd and 10−6 accuracy in hg. Table 3 lists the parameters of our models. Note that these parameters do not describe plau- sible protoplanetary gas disks; in particular, our model metallicities Σd/Σg are orders of magnitude above the solar value of ∼0.015. Parameters are 7 instead chosen to yield disk flows that our code can adequately resolve while still testing equation (14). Unfortunately, more astrophysically realis- tic parameters correspond to dust sublayers that are too vertically thin for us to resolve numeri- cally; the code timestep, set by the sound-crossing time across a grid cell, becomes prohibitively short as thinner dust layers are considered. This diffi- culty means that the difference between ρ∗ II and the other candidate threshold densities is much less than what it is in reality, and our ability to distinguish between the candidates degrades as a result. Figure 2 plots the initial conditions for our stan- dard model (S = STD32), for which Qg = 24, Σd/Σg = 8, and µ0 = 35. For this specific case, we calculate that hg = 0.20 cg/Ω and zd = 0.083 cg/Ω. The top and bottom boundaries of our simulation box are indicated by dotted ver- tical lines; typically box heights span ±4zd (see §3 for box height tests). The right-hand panel of Figure 2 compares gas density profiles computed with and without self-gravity, and with and with- out dust, and shows that both the weight and self- gravity of the embedded dust layer force gas into a similarly thin layer. 2. -- Left: Fig. Initial dust and gas densities for standard run S = STD32. Right: Initial gas den- sity (dashed) on an expanded scale, together with the gas density computed without self-gravity but with dust (dash-dot) and without dust but with self-gravity (dash-double-dot). Vertical lines in both panels lines delimit the top and bottom of our computational box. Figure 3 plots the initial force densities within the upper half of the simulation box to demon- strate how well vertical hydrostatic equilibrium is satisfied. The sum of stellar gravity (blue dashed curve) and disk self-gravity (red solid curve com- puted via the integral in equation 23) should equal 8 Fig. 3. -- Vertical force balance for our initial con- ditions. The force density due to self-gravity is com- puted two ways: by direct integration of the density profile (red solid), and by using the code's Poisson solver (green dashed). The two methods agree. The force density due to the pressure gradient (black solid) should equal the sum of self-gravity and the static stel- lar potential (blue dashed). The horizontal gray dot- ted line shows the ratio of pressure to gravity. All force densities are shown in their absolute values. the pressure gradient (black solid curve). It does, as evidenced by the ratio of pressure to gravity (gray dotted line) which is practically constant at unity. We also overplot the self-gravitational force computed by our 3D Poisson solver (green dashed curve); the agreement with the exact solution is good. Every simulation listed in Table 3 is perturbed from its initial equilibrium by adding random cell- to-cell fluctuations of amplitude ∼10−3cg to the velocity field. The typical duration of a simula- tion is ∼20 Ω−1. Our rationales for box size and resolution are explained in §3. 3. RESULTS Results for 2D shearing sheets are described in §3.1, and those for 3D shearing boxes are in §3.2. 3.1. 2D Shearing Sheet For two-dimensional dusty disks, the criterion for gravitational instability reads Qd = cdΩ πG(Σd + Σg) = Qg (1 + µ0)3/2 < Q∗ d,2D . (29) −0.4−0.20.00.20.4z (cg/Ω)−4−3−2−1012log (ρ/ρg0)STD32dustgas0123z (cg/Ω)0.00.20.40.60.81.01.2ρg/ρg00.000.050.100.150.200.250.30z (cg/Ω)−0.50.00.51.01.5force density (cgΩ)Table 2 2D Simulation Parameters Name Qg S2D0 S2D1 S2D2 S2D3 S2D4 S2D5 S2D6 12 12 12 12 12 12 6 µ0 8 8 8 8 4.2 2.3 4.2 Qd λc(cg/Ω) 0.44 0.44 0.44 0.44 1.0 2.0 0.5 0.93 0.93 0.93 0.93 2.8 6.9 1.4 Lx × Ly (c2 10 × 10 10 × 10 0.5 × 0.5 0.5 × 0.5 30 × 30 70 × 70 15 × 15 g/Ω2) Resolution GI(a) Duration (Ω−1) 256 × 256 512 × 512 32 × 32 128 × 128 256 × 256 256 × 256 256 × 256 Y Y N N N N Y 5.9 6 100 100 100 100 6.2 (a) GI = Gravitational Instability. Y means max Σd increases by orders of magnitude over a few dynamical times, and N means it does not. We test this criterion by constructing a series of 2D shearing sheet simulations with various values of Qg and µ0, thereby seeing if we can converge on a unique value for Q∗ d,2D. Although total surface densities can change during the simulation, the dust-to-gas ratio stays fixed at its initial value be- cause of our perfect-coupling approximation. Ini- tial conditions are as follows: for a given domain size Lx and Ly, the flow velocity v = − 3 2 Ωxey and the surface density Σ = Σ0 + δΣ cos(k · x), with Σ0 = Σg0 + Σd0 = Σg0(1 + µ0), δΣ/Σ0 = 0.01, kx = −2(2π/Lx), and ky = 2π/Ly. In our 2D simulations, we choose cg = Ω = Σg0 = 1 as our units. Table 2 lists the parameters for our 2D runs. Our standard 2D run, labeled S2D0, has Qg = 12 and µ0 = 8.0 and therefore Qd = 0.44. For this run, the domain size is chosen large enough to eas- ily fit the critical wavelength λc for gravitational instability: Lx = Ly = 10cg/Ω (cid:38) 10λc, where λc ≡ 2c2 d GΣ0 (30) is the wavelength of the fastest growing mode ac- cording to the WKB dispersion relation for ax- isymmetric waves. It is also the wavelength of the first mode to become unstable when Qd just crosses Q∗ d,2D. The resolution of the standard run is Nx × Ny = 256 × 256 so that one critical wave- length is resolved across ∼10 grid cells. For S2D0, we find that the disk is indeed gravi- tationally unstable: density waves steepen quickly, and dense clumps of dusty gas form before one orbital period elapses. A simple way to portray instability is to track the maximum dust density max Σd versus time -- this is done in Figure 4, which shows that the maximum dust density in- creases by two orders of magnitude over a few dy- namical times for our standard run. Fig. 4. -- Time evolution of the maximum dust den- sity max Σd for our 2D shearing sheet simulations. The critical value Q∗ d,2D below which gravitational in- stability is triggered appears to be between 0.5 and 1.0. Also shown in Figure 4 are results for other runs. In S2D4, S2D5, and S2D6, either Qg or µ0 is varied relative to our standard run, so that Qd varies from 0.5 to 2.0. For all these runs, the domain size is ∼10λc in each direction and the res- olution is ∼ 10 cells per λc, just as in the standard case. Taken together, the results indicate that 0.5 < Q∗ d,2D < 1.0 . (31) Other runs explore the effects of varying resolu- tion and domain size. Doubling both Nx and Ny 9 02468t (Ω−1)1101001000max (Σd/Σd0)S2D0, Qd=0.4, 256 × 256S2D1, Qd=0.4, 512 × 512S2D6, Qd=0.5, 256 × 256S2D4, Qd=1.0, 256 × 256S2D5, Qd=2.0, 256 × 256relative to our standard run (S2D1) enables higher maximum densities to be achieved when the insta- bility saturates, but otherwise does not seem to alter the evolution. Reducing the size of the box so that it can no longer accommodate even a sin- gle critical wavelength (S2D2, S2D3) results in no instability, as expected (Gammie 2001; Johnson & Gammie 2003). 3.2. 3D Stratified Dusty Disks Equation (8; equivalently 29) gives the crite- rion for gravitational instability in a 2D razor-thin sheet. For a 3D, vertically stratified disk, there is some ambiguity as to how we evaluate cd in equa- tion (8) because its value varies with height. Here we simply take cd to be its value at the midplane, so that criterion (8) becomes Qd (cid:39) Qg 1 1 Σd/Σg (1 + µ0)1/2 (cid:46) Q∗ d . (32) An alternative is to calculate a vertically averaged, density-weighted sound speed. We found, how- ever, that such a procedure made little practical difference, since dust densities are much greater than gas densities near the midplane and drop steeply with height. 3.2.1. Standard run (S = STD32) To orient the reader, we present results for our standard 3D run (S, also labeled STD32 in §3.2.2), for which Qd = 0.5. The full set of model S pa- rameters are listed in Table 3, and the initial gas and dust density profiles are displayed in Figure 2. Our simulation box extends ±4zd vertically, and 14zd in either horizontal direction. Each horizon- tal length is about twice the critical wavelength (λc ≈ 6.3zd). The resolution is 32 × 32 × 32 so that one horizontal critical wavelength spans ∼16 cells, and one vertical scale length zd spans 4 cells. These choices for domain size and resolution are tested in §3.2.2. The simulation is terminated at ∼10.3Ω−1, at which point the timestep has be- come three orders of magnitude smaller than the initial timestep (see the final paragraph of §2.1.1). Figure 5 displays a time series of the volume- rendered dust density in the bottom half of the box. Over the course of several dynamical times, density waves shear and amplify, eventually con- centrating into a single azimuthally elongated fila- ment. This filament then fragments radially. The fragments gravitationally scatter and merge; by the end of the simulation, two clumps remain. A simple diagnostic that we use throughout this paper is the time evolution of the maximum dust density, shown in the left panel of Figure 6. Com- parison with Figure 5 reveals that max ρd grows exponentially when the filament fragments radi- ally. The maximum dust density ceases to rise once the clumps finish coalescing. At this point each clump is gravitationally bound, with a max- imum central density that depends on the simula- tion resolution (§3.2.2). Fig. 6. -- Left: Time evolution of maximum dust density for run STD32 (= S). Right: Time evolution of kinetic energies averaged horizontally and vertically over a thin slab subtending two grid cells at the mid- plane (red = x-component of kinetic energy; blue = y; green = z; black = total). The right panel of Figure 6 shows the time evolution of various kinetic energy densities, eval- uated in the three directions and excluding the background Keplerian shear. The energy densi- ties are averaged horizontally and vertically over a thin slab subtending two grid cells at the mid- plane (qualitatively similar results are obtained over larger vertical averages). The horizontal ki- netic energies grow exponentially from t = 2 -- 7 Ω−1, with an exponential growth rate of ∼1.5Ω. Radial motions dominate azimuthal motions until the end of the simulation when they become com- parable. Vertical motions develop immediately af- ter the beginning of the simulation because our discretized initial conditions cannot be in perfect hydrostatic balance; however the magnitude of the vertical motions is small and stays roughly con- stant for t (cid:46) 6 Ω−1. For t (cid:38) 6 Ω−1, vertical mo- tions amplify but for the most part remain smaller than horizontal motions. The in-plane motions of the dusty clumps are 10 02468101214t (Ω−1)101102103104105max (ρd/ρg0)024681012t (Ω−1)−15−10−505ln 〈ρ v2/(ρg0 cg2 )〉0Fig. 5. -- Time evolution of gravitational instability in our standard 3D stratified dusty disk (run S = STD32). Shown are volume renderings of dust density for the bottom half of the disk at t = 0, 5.0, 6.4, 7.3, 8.0, and 10.3Ω−1 (left to right, top to bottom). 11 3D Simulation Parameters ("Science Runs") Table 3 Name Qg µ0 Σd/Σg Hg zd (cg/Ω) (cg/Ω) λc (zd) Lx × Ly × Lz (z3 d) Resolution Duration (Ω−1) Qd ρ0 (ρ†) (a) ρ∗ (ρ†) I (b) ρ∗ (ρ†) II GI(c) S R1 R2 R3 R4 R5 SR Z Q M 24 35.0 24 165.0 12 93.0 12 143.0 12 322.0 12 322.0 24 35.0 35.0 24 35.0 48 24 8.0 8.0 2.0 0.67 0.54 0.56 1.33 8.0 4.0 8.0 8.0 0.20 0.91 1.04 1.07 1.07 0.87 0.20 0.52 0.35 0.24 0.083 0.011 0.008 0.004 0.002 0.004 0.083 0.077 0.12 0.71 14 × 14 × 8 90 × 90 × 8 32 × 32 × 32 6.3 256 × 256 × 32 42.6 150.3 256 × 256 × 8 256 × 256 × 32 255.4 400 × 400 × 8 400 × 400 × 32 220.7 400 × 400 × 8 400 × 400 × 32 400 × 400 × 8 400 × 400 × 32 44.7 400 × 400 × 8 400 × 400 × 32 6.3 32 × 32 × 32 30 × 30 × 8 13.1 32 × 32 × 32 20 × 20 × 8 8.6 8 × 8 × 8 32 × 32 × 32 3.2 10.3 11 30 30 30 3.6 10.0 30 30 30 0.5 1.20 0.93 1.21 1.86 1.20 1.86 1.77 4.0 1.2 4.9 0.5 0.5 1.20 0.46 1.0 0.34 1.0 1.0 0.25 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.30 Y 1.05 Y/N(d) 4.09 N 6.12 N 5.16 N 1.24 Y 0.30 Y 0.46 N 0.34 N 0.25 N (a) Values are derived using Q∗ (b) Values are derived using Q∗ (c) GI = Gravitational Instability. Y means max ρd increases by orders of magnitude over a few dynamical times, and N means d = 1. g = 1. it does not. (d) See Figure 11. Fig. 7. -- Snapshot of the midplane for run S = STD32 at t = 9.6Ω−1. The largest in-plane velocity shown is 2.16 cg. illustrated in Figure 7 with a snapshot of the mid- plane slice of STD32 at t = 9.6 Ω−1. The dust clumps are seen spinning about their centers of mass as a consequence of angular momentum con- servation. 3.2.2. Resolution and box size Table 4 lists the parameters of experiments de- signed to test our choices for resolution, box size, 12 and grid-cell aspect ratio. Figure 8 shows how varying the resolution changes the evolution of our standard, gravita- tionally unstable run (STD32 -- also labeled S in Table 3). We use again the simple metric of max ρd vs. t. Broadly speaking, the runs STD16, STD32, STD64 are all "acceptable" insofar as they all yield increases in max ρd by orders of magni- tude within several dynamical times (t (cid:46) 8 Ω−1). By contrast, the lowest resolution run, STD8, is unacceptable. Thus, the minimum acceptable res- olution appears to be ∼2 cells per scale length zd in the vertical direction (cf. Nelson 2006 who found that a minimum of four smoothing lengths per scale height is required for SPH simulations), and ∼8 cells per critical wavelength λc in the horizon- tal directions. Our standard choices for resolution -- as well as the resolutions characterizing all our "science" runs, listed in Table 3 and discussed in §3.2.3 -- satisfy these minimum requirements by a safety factor of 2. Examining Figure 8 more critically, we see that the maximum value attained by max ρd has not converged with resolution. Increasing the resolu- tion enables us to resolve ever higher densities in the collapsing clumps. Another point of concern is the non-uniform aspect ratios of individual grid cells, which ranges from x:y:z ≈ 2:2:1 to 4:4:1 over −6−4−20246x (zd)−6−4−20246y (zd) −101234t = 9.6Ω−1log [ρd(z=0)/ρg0]3D Simulation Parameters to Test Box Size and Resolution Table 4 Name STD32(S) STD8 STD16 STD64 U32 LZ2 LZ4 LZ6 LZ10 LZ14 LXY6 LXY10 LXY20 Lx × Ly × Lz(z3 d) 14 × 14 × 8 14 × 14 × 8 14 × 14 × 8 14 × 14 × 8 14 × 14 × 8 14 × 14 × 2 14 × 14 × 4 14 × 14 × 6 14 × 14 × 10 14 × 14 × 14 6 × 6 × 8 10 × 10 × 8 20 × 20 × 8 Resolution 32 × 32 × 32 8 × 8 × 8 16 × 16 × 16 64 × 64 × 64 56 × 56 × 32 32 × 32 × 8 32 × 32 × 16 32 × 32 × 24 32 × 32 × 40 32 × 32 × 56 16 × 16 × 32 24 × 24 × 32 48 × 48 × 32 GI(a) Duration (Ω−1) Y N Y Y Y N Y Y Y Y N Y Y 9.8 30.0 11.0 11.0 10.0 30.0 8.5 8.0 9.0 10.5 30.0 8.6 8.7 (a) GI = Gravitational Instability. Y means max ρd increases by orders of magnitude over a few dynamical times, and N means it does not. Fig. 8. -- Time evolution of the maximum dust den- sity for the resolution study (see Table 4). Fig. 9. -- Time evolution of the maximum dust den- sity for our box size tests (see Table 4). our set of science simulations (Table 3). The run U32 is characterized by perfectly cubical grid cells (1:1:1); the evolution is similar to STD32, but is characterized by an earlier onset of gravitational instability, and stronger density fluctuations. This comparison suggests that our science runs with non-cubical grid cells are biased slightly against gravitational instability. We next investigate how box size affects our re- sults. For all box size experiments, the spatial res- olution is kept at its standard value (32 grid cells per 14zd in either horizontal direction, and 4 grid cells per zd in the vertical direction). Runs LZ2 through LZ14 vary box height Lz while keeping Lx and Ly fixed at their standard (STD32 = S) values. As Figure 9 reveals, box heights of 4 -- 14zd yield comparable results, while a box height of 2zd is unacceptable. For the most part, increasing the box height seems to delay the onset of gravita- tional instability, with LZ4 being the exception to this rule. Our 2D simulations indicated that Lx and Ly must be large enough to encompass at least one critical wavelength λc. Our 3D simulations bear 13 02468101214t (Ω−1)101102103104105max (ρd/ρg0)STD8STD16STD32STD64U3202468101214t (Ω−1)101102103104105max (ρd/ρg0)STD32/SLZ2LZ4LZ6LZ10LZ14LXY6LXY10LXY20out this same requirement. Figure 9 shows that run LXY6, for which the box size is just under one critical wavelength, does not exhibit gravitational instability, unlike its bigger box counterparts. To summarize our findings in this subsection: (1) The simulation box should be at least 4zd tall (2zd above and below the midplane). (2) Each horizontal dimension must be longer than one critical wavelength λc as given by equation (30). (3) Simulations require a vertical resolu- tion of (cid:38) 2 grid cells per scale length zd, and a horizontal resolution of (cid:38) 8 grid cells per critical wavelength. (4) Individual grid cells that have in- creasingly non-uniform aspect ratios (squatter ver- tically than horizontally) tend to suppress gravita- tional instability, but the bias is minor and aspect ratios up to 4:4:1 appear acceptable. All of our science simulations (Table 3; §3.2.3) satisfy these requirements, in some cases by factors of 2. 3.2.3. Criteria for gravitational collapse Table 3 lists the simulations designed to test which of the various proposed criteria for gravita- tional instability is the best predictor of collapse. Figures 2 and 10 describe the initial dust and gas profiles, while Figure 11 displays the results using our simple diagnostic of max ρd vs. time. I or ρ∗ First consider runs S and R1 -- R5, and ask whether these runs favor ρ∗ II for the density required for gravitational collapse. Because dust is a major component of our disks, we do not ex- pect ρ∗ I -- which is strictly valid only for pure gas disks -- to be a good predictor. Indeed in all six of these runs, the midplane density ρ0 exceeds ρ∗ I , by factors of 7.5 -- 30, yet only runs S and R5, and to a much lesser extent R1, exhibit collapse. All six runs indicate instead that ρ∗ II -- equivalently, Qd -- is the better predictor, with the critical value 0.5 < Q∗ d < 0.9 . (33) There is some concern that the comparison be- tween runs R2 -- R5 and run S may not be fair be- cause runs R2 -- R5 have a factor of ∼2 poorer spa- tial resolution in x and y compared to run S. This concern is allayed by run SR, which has the same physical parameters as S but is run with the box size and resolution of R3, and which turns out to behave qualitatively similarly to S (see Figure 11). II is relevant and that Q∗ d obeys (33) is supported further by runs Z, Q, and Our conclusion that ρ∗ 10. -- Initial conditions for our science runs Fig. which explore parameter space. Solid lines denote dust, and dashed lines denote gas. The vertical lines delimit the vertical boundaries of our simulation box. M, each of which varies one of the three input parameters Σd/Σg, Qd, and µ0. Although runs R2 -- R4 do not exhibit the dra- matic growth in ρd shown by runs S, SR, and R5 -- a result that we interpret to mean that ρ∗ II gives the correct criterion for gravitational collapse -- 14 −0.20.00.2z (cg/Ω)−6−4−202log (ρ/ρg0)Z−0.6−0.4−0.20.00.20.40.6z (cg/Ω)−6−4−202log (ρ/ρg0)Q−202z (cg/Ω)−6−4−202log (ρ/ρg0)M−0.04−0.020.000.020.04z (cg/Ω)−6−4−202log (ρ/ρg0)R1−0.020.000.02z (cg/Ω)−6−4−202log (ρ/ρg0)R2−0.02−0.010.000.010.02z (cg/Ω)−6−4−202log (ρ/ρg0)R3−0.0050.0000.005z (cg/Ω)−6−4−202log (ρ/ρg0)R4−0.010.000.01z (cg/Ω)−6−4−202log (ρ/ρg0)R5R4 seem unlikely to lead to planetesimal forma- tion. In particular, the density concentrations in runs R2 -- R4 eventually disperse, unlike the density concentrations in runs S, SR, and R5 for which ρ0 > ρ∗ II. What evidence we have suggests that Sekiya's mode is not important for planetesimal formation, but higher resolution simulations that better separate ρ∗ II are needed for a more definitive assessment. Finally, what about ρ∗ II? Here runs R4 and R5 are the most telling. Both runs are characterized by the largest midplane densities ρ0 > ρ∗ Roche, but only R5, for which ρ0 > ρ∗ II, undergoes gravitational collapse (see Figure 11). Sekiya from ρ∗ Roche vs. ρ∗ Table 5 summarizes how the various candi- date critical densities relate to one another and to the midplane density for our science simulations. From Table 5, ρ∗ II emerges as the best predictor of collapse. 4. SUMMARY AND DISCUSSION Fig. 11. -- Time evolution of the maximum dust den- sity in our science simulations. Only for runs S, SR, and R5 does ρ0 > ρ∗ II, and indeed only those runs exhibit dramatic growth of the dust density due to gravitational instability. Fig. 12. -- Snapshots of the dust density at the mid- plane for run R3 (left panel) and run SR (right panel). In R3, the initial midplane density ρ0 > ρ∗ Sekiya, and there is some clumping, but it is much lower in ampli- tude compared to run SR, for which ρ0 > ρ∗ II. runs R2 -- R4 do show some clumping. Figure 12 compares snapshots of runs R3 and SR (performed with the same box size and resolution), taken at the same time t = 10Ω−1. Filaments do form in R3, although they are much weaker in density con- trast compared to the filaments in SR. The mild growth shown in runs R2 and R3 might simply reflect the fact that their values for Qd = 1.86 are still too close to Q∗ d to suppress instability en- tirely. An alternative (and not mutually exclusive) possibility is that because ρ0 > ρ∗ Sekiya = 0.60ρ† for runs R2 -- R4, the disk might be exhibiting the unstable (and formally incompressible) mode found by Sekiya (1983). Whatever the interpreta- tion, the modest growth factors exhibited by R2 -- Fig. 13. -- A tale of two particle sublayers, one of which is thinner and denser than the other. Dust den- sity is plotted as a solid line, and gas density as a dashed line. The disks have identical masses and bulk metallicities, enhanced over those of the minimum- mass solar nebula by factors of 3 -- 4. Left: Midplane density ρ0 = ρ∗ Roche = 3.5ρ† and Qd = 10.4. Right: Midplane density ρ0 ≈ ρ∗ Roche and Qd = 1. According to the results of our simulations, only the model in the right panel, having the thinner and denser sublayer, should be on the verge of gravitational col- lapse -- in the limit that particles are aerodynamically perfectly coupled to gas. We argue in §4.1 that when the perfect coupling approximation breaks down, it may be possible for the disk on the left to undergo gravitational instability. II ≈ 102ρ∗ Dust grains settle toward the midplanes of pro- toplanetary disks, forming a sublayer of solid par- 15 051015t (Ω−1)0.11.010.0100.01000.0max (ρd/ρd0)SSRZQMR1R2R3R4R5−7−6−5−4−3−2log (z cg−1Ω)−2024log (ρ/ρg0)Qg = 10Σd/Σg = 0.06µ0 = 255Qd = 10.4ρ0/(M∗/r3) = 3.5−7−6−5−4−3−2log (z cg−1Ω)−2024log (ρ/ρg0)Qg = 10Σd/Σg = 0.06µ0 = 2.8 × 104Qd = 1.0ρ0/(M∗/r3) = 381.3Comparison of Critical Densities and Actual Midplane Density for Science Simulations Table 5 Name S R1 R2 R3 R4 R5 Critical density relations ρ∗ II < ρ∗ I < ρ∗ Sekiya < ρ∗ ρ∗ I < ρ∗ ρ∗ I < ρ∗ ρ∗ I < ρ∗ ρ∗ I < ρ∗ Sekiya < ρ∗ ρ∗ I < ρ∗ Sekiya < ρ∗ II < ρ∗ Sekiya < ρ0 < ρ∗ II < ρ0 < ρ∗ Sekiya < ρ0 < ρ∗ Sekiya < ρ0 < ρ∗ Roche < ρ0 < ρ∗ Roche < ρ0 II Roche Roche Roche < ρ∗ Roche < ρ∗ II II GI(a) Y Y/N(b) N N N Y (a) GI = Gravitational Instability. Y means max ρd increases by orders of magnitude over a few dynamical times, and N means it does not. (b) See Figure 11. (cid:18) ρ∗ II ρ0 Qd ≈ (cid:19)1/2 (cid:46) 1 (cid:18) Σg (cid:19)2 Σd (34) . (35) ticles sandwiched from above and below by gas. Whether this sublayer can become thin enough and dense enough to undergo gravitational insta- bility and fragment into planetesimals is an out- standing question. We have found in this work that the density threshold for gravitational col- lapse can be extraordinarily high -- much higher even than the Roche density ρ∗ Roche = 3.5M∗/r3, where M∗ is the mass of the central star and r is the orbital radius. To trigger collapse in the limit that dust particles are small enough to be tightly coupled to gas, the density ρ0 in the sublayer must be such that the Toomre stability parameter where II ≈ M∗ ρ∗ 2πr3 Qg Q∗2 d (For more precise relations, see equations 8, 13, and 33.) Here Qg is the Toomre parameter for the ambient (and much thicker) gas disk, Σd/Σg is the ratio of surface densities of dust and gas (i.e., the height-integrated metallicity), and 0.5 < Q∗ d < 0.9 as measured from our simulations. For an astrophysically plausible disk having 3× the mass of the minimum-mass solar nebula (Qg ≈ 10) and a bulk metallicity enriched over solar by a factor of 4 (Σd/Σg ≈ 0.06), the critical density II ≈ 1.3 × 102Q∗−2 ρ∗ d ρ∗ Roche . (36) Figure 13 portrays two sublayers -- one for II ≈ 102ρ∗ which ρ0 = ρ∗ Roche and another, much thinner sub- layer for which ρ0 ≈ ρ∗ Roche (Qd = 1). The results of our simulations, performed in the limit of perfect aerodynamic coupling between particles and gas, indicate that only the latter, much denser disk is on the verge of fragmenting. Qualitatively, such extraordinary densities are required for gravitational instability because gas pressure renders the sublayer extremely stiff. Sound-crossing times for thin layers are easily shorter than free-fall times. We can examine the competition between stabilizing pressure, stabi- lizing rotation, and de-stabilizing self-gravity in both the horizontal (in-plane) and vertical direc- tions. Horizontal stability is controlled by Qd: d ∼ 1, all horizontal lengthscales when Qd > Q∗ λ (cid:46) 2c2 d/GΣd are stabilized by pressure, and all scales λ (cid:38) 2c2 d/GΣd are stabilized by rotation, where cd is the effective sound speed in the dust- gas mixture. At the same time, vertical stabil- ity is assured whenever the sound-crossing time across the vertical thickness of the sublayer 2Hd is shorter than the free-fall time: 2Hd cd 1√ Gρd < (37) which, after substituting Hd ≈ Σd/2ρd and cd ≈ cg (cid:112)ρg/ρd, translates to (cid:18) Σd (cid:19)2 1 < π 2 (38) Σg Qg which is easily satisfied for reasonable disk param- eters. 16 The severe obstacle that gas pressure presents to gravitational collapse of aerodynamically well- coupled particles is discussed by Cuzzi, Hogan, & Shariff (2008, see their section 3.1). Our 3D disk simulations support their 1D considerations. 4.1. Directions for Future Research Taken at face value, the higher density thresh- old ρ∗ II established by our work argues against using aerodynamically well-coupled particles to form planetesimals. The Kelvin-Helmholtz in- stability (KHI) may prevent dust from settling into the extraordinarily thin sublayers needed to cross the density threshold. One potential loop- hole is provided by Sekiya (1998) and Youdin & Shu (2002), who found in 1D that self-gravitating, non-rotating sublayers having constant Richard- son number Ri could develop cusps of infinite den- sity at the midplane. The presumption of these studies is that dust settles into a state that is marginally KH-stable and that this state is char- acterized by a constant Ri. Some evidence for a spatially constant Ri was found in the settling experiments of Lee et al. (2010b), but only near the top and bottom faces of the dust sublayer and not at the midplane. These numerical experiments suffered, however, from lack of spatial resolution toward the midplane, and moreover neglected self- gravity. Future simulations of cuspy dust profiles including self-gravity would be welcome. We have worked in the limit that the stopping times tstop of particles in gas are small compared to all other timescales. But in reality, finite particle sizes imply finite tstop (see Figure 1). When the assumption of infinitesimal stopping time breaks down, new effects may appear that might lower the threshold for gravitational instability. One such effect is as follows. Consider again the competition between stabilizing pressure and de-stabilizing self-gravity (in either the vertical or horizontal directions). A major reason why the sublayer so strongly resists collapse is that sound waves travel quickly across it. We have taken the sound speed for our dust-gas suspension to be cd = cg/(cid:112)1 + ρd/ρg ≈ cg (cid:112)ρg/ρd (equations 9 and 10). But this presumes that particles are per- fectly coupled to gas. If the sound-crossing time across some scale λ were to become shorter than λ cd the particle stopping time, i.e., if (cid:114) ρd (cid:112)ρg/ρd would be invalid. ≈ λ cg then our use of cd ≈ cg Particles on scales λ would lose support from gas pressure and become susceptible to gravitational instability. < tstop (39) ρg To get a sense of where in parameter space this instability may lie, we normalize λ to the full ver- tical thickness of the sublayer: λ ≡ 2Hd λ = λΣd ρd . (40) where λ can take any value (larger than or smaller than unity). Then equation (39) for the loss of pressure support translates to a midplane density (dominated by dust) of ρ0 ≈ ρd (cid:38) 2 π M∗ r3 1 (41) Σg Qg (Ωtstop)2 where Ω is the Kepler orbital frequency. For self- gravity to resist tidal disruption, ρd = ρ∗ Roche = 3.5M∗/r3. Substituting this requirement into (41), we find that (cid:18) Σd (cid:19)2 λ2 Ωtstop (cid:38) (cid:18) 2 3.5π (cid:19)1/2(cid:18) Σd (cid:19) λ (cid:19)(cid:32) λ (cid:18) Σd/Σg Q1/2 Σg g (cid:38) 8 × 10−3 (cid:33)(cid:18) 10 (cid:19)1/2 0.06 1 Qg (42) for particles on scales λ to decouple from sound waves. For λ = 1, requirement (42) could be ful- filled by particles having sizes of a few millimeters to a few centimeters at distances of 1 -- 10 AU (Fig- ure 1 -- but note that the curves in the figure need to be adjusted by factors of a few for mass-enriched nebulae). For λ < 1, even smaller particles could lose pressure support and collapse gravitationally. Future simulations that include finite particle stopping times could try to find such an instabil- ity. A complication would be that accounting for finite tstop would introduce the streaming insta- bility, which could prevent the dust density from attaining the Roche value -- see, e.g., runs R21-3D and R41-3D in Figure 5 of Bai & Stone (2010), for 17 which Ωtstop ≤ 0.1 and ρd < ρ∗ Roche. To find the instability that we are envisioning, one would have to restrict Ωtstop to small enough values to sup- press the streaming instability -- thereby permit- ting the setting of grains into sublayers for which ρd = ρ∗ Roche -- while at the same time keeping Ωtstop large enough to satisfy (42) and nullify pres- sure support. We are grateful to Eve Ostriker and Andrew Youdin for discussions. Section 4.1 was inspired by discussions with Eve that clarified the limitations of our study and pointed the way to a possible new route to gravitational instability. We thank Xuening Bai, Chang-Goo Kim, Eve Ostriker, Ian Parrish, and Jim Stone for help in augmenting Athena. The simulations were performed with the Berkeley cluster Henyey, which was made possi- ble by a National Science Foundation Major Re- search Instrumentation (NSF MRI) grant. Finan- cial support for the authors was provided by the Berkeley Center for Integrative Planetary Science, the Berkeley Theoretical Astrophysics Center, and grants from NSF (AST-0909210) and NASA Ori- gins. REFERENCES Andrews, S. M., Wilner, D. J., Hughes, A. M., Qi, C., Rosenfeld, K. A., Oberg, K. I., Birnstiel, T., Espaillat, C., Cieza, L. A., Williams, J. P., Lin, S.-Y., & Ho, P. T. P. 2012, ApJ, 744, 162 Cuzzi, J. N., Hogan, R. C., & Shariff, K. 2008, ApJ, 687, 1432 Gammie, C. F. 2001, ApJ, 553, 174 Goldreich, P. & Lynden-Bell, D. 1965, MNRAS, 130, 125 Johansen, A., Oishi, J. S., Low, M., Klahr, H., Henning, T., & Youdin, A. 2007, Nature, 448, 1022 Johnson, B. M. & Gammie, C. F. 2003, ApJ, 597, 131 Kim, C.-G., Kim, W.-T., & Ostriker, E. C. 2011, ApJ, 743, 25 Koyama, H. & Ostriker, E. C. 2009, ApJ, 693, 1316 Lee, A. T., Chiang, E., Asay-Davis, X., & Bar- ranco, J. 2010a, ApJ, 718, 1367 -- . 2010b, ApJ, 725, 1938 Mamatsashvili, G. R. & Rice, W. K. M. 2010, MN- RAS, 406, 2050 Nelson, A. F. 2006, MNRAS, 373, 1039 Okuzumi, S., Tanaka, H., & Sakagami, M.-a. 2009, ApJ, 707, 1247 Sekiya, M. 1983, Progress of Theoretical Physics, 69, 1116 Bai, X.-N. & Stone, J. M. 2010, ApJ, 722, 1437 -- . 1998, Icarus, 133, 298 Binney, J. & Tremaine, S. 2008, Galactic Dynam- ics: Second Edition, 2nd edn. (Princeton Uni- versity Press) Stone, J. M. & Gardiner, T. 2009, New A, 14, 139 Stone, J. M. & Gardiner, T. A. 2010, ApJS, 189, Birnstiel, T., Ricci, L., Trotta, F., Dullemond, C. P., Natta, A., Testi, L., Dominik, C., Hen- ning, T., Ormel, C. W., & Zsom, A. 2010, A&A, 516, L14 Chandrasekhar, S. 1987, Ellipsoidal Figures of Equilibrium, 1st edn. (Dover Publications, New York) Chiang, E. 2008, ApJ, 675, 1549 Chiang, E. & Youdin, A. 2010, Annual Reviews of Earth and Planetary Science, 38 142 Stone, J. M., Gardiner, T. A., Teuben, P., Hawley, J. F., & Simon, J. B. 2008, ApJS, 178, 137 Toomre, A. 1964, ApJ, 139, 1217 Toomre, A. 1981, in Structure and Evolution of Normal Galaxies, ed. S. M. Fall & D. Lynden- Bell, 111 -- 136 Toro, E. 1999, Riemann Solvers and Numerical Methods for Fluid Dynamics (Berlin: Springer) van Leer, B. 2006, Comm. Comput. Phys., 1, 192 18 Weidenschilling, S. J. 1977, MNRAS, 180, 57 -- . 1980, Icarus, 44, 172 -- . 2006, Icarus, 181, 572 -- . 2010, Meteoritics and Planetary Science, 45, 276 Youdin, A. N. & Goodman, J. 2005, ApJ, 620, 459 Youdin, A. N. & Shu, F. H. 2002, ApJ, 580, 494 Zsom, A., Ormel, C. W., Dullemond, C. P., & Henning, T. 2011, A&A, 534, A73 This 2-column preprint was prepared with the AAS LATEX macros v5.2. 19
1206.1235	1	1206	2012-06-06T14:13:39	Minimizing follow-up for space-based transit surveys using full lightcurve analysis	[ "astro-ph.EP" ]	One of the biggest challenges facing large transit surveys is the elimination of false-positives from the vast number of transit candidates. We investigate to what extent information from the lightcurves can identify blend scenarios and eliminate them as planet candidates, to significantly decrease the amount of follow-up observing time required to identify the true exoplanet systems. If a lightcurve has a sufficiently high signal-to-noise ratio, a distinction can be made between the lightcurve of a stellar binary blended with a third star and the lightcurve of a transiting exoplanet system. We perform simulations to determine what signal-to-noise level is required to make the distinction between blended and non-blended systems as function of transit depth and impact parameter. Subsequently we test our method on real data from the first IRa01 field observed by the CoRoT satellite, concentrating on the 51 candidates already identified by the CoRoT team. About 70% of the planet candidates in the CoRoT IRa01 field are best fit with an impact parameter of b>0.85, while less than 15% are expected in this range considering random orbital inclinations. By applying a cut at b<0.85, meaning that ~15% of the potential planet population would be missed, the candidate sample decreases from 41 to 11. The lightcurves of 6 of those are best fit with such low host star densities that the planet-to-star size ratii imply unrealistic planet radii of R>2RJup. Two of the five remaining systems, CoRoT1b and CoRoT4b, have been identified as planets by the CoRoT team, for which the lightcurves alone rule out blended light at 14% (2sigma) and 31% (2sigma). We propose to use this method on the Kepler database to study the fraction of real planets and to potentially increase the efficiency of follow-up.	astro-ph.EP	astro-ph	Astronomy&Astrophysicsmanuscript no. AA22042RevisedPaper14May November 27, 2017 c(cid:13) ESO 2017 2 1 0 2 n u J 6 . ] P E h p - o r t s a [ 1 v 5 3 2 1 . 6 0 2 1 : v i X r a Minimizing follow-up for space-based transit surveys using full lightcurve analysis. S.V. Nefs1, I.A.G. Snellen1 & E.J.W. de Mooij1 1 Leiden Observatory, Leiden University, P.O. Box 9513, 2300 RA Leiden, The Netherlands Revised Draft: 14 May 2012 ABSTRACT Context.One of the biggest challenges facing large transit surveys is the elimination of false-positives from the vast number of transit candidates. A large amount of expensive follow-up time is spent on verifying the nature of these systems. Aims. We investigate to what extent information from the lightcurves can identify blend scenarios and eliminate them as planet candidates, to significantly decrease the amount of follow-up observing time required to identify the true exoplanet systems. Methods. If a lightcurve has a sufficiently high signal-to-noise ratio, a distinction can be made between the lightcurve of a stellar binary blended with a third star and the lightcurve of a transiting exoplanet system. We first simulate lightcurves of stellar blends and transiting planet systems to determine what signal-to-noise level is required to make the distinction between blended and non-blended systems as function of transit depth and impact parameter. Subsequently we test our method on real data from the first IRa01 field observed by the CoRoT satellite, concentrating on the 51 candidates already identified by the CoRoT team. Results. Our simulations show that blend scenarios can be constrained for transiting systems at low impact parameters. At high impact parameter, blended and non-blended systems are indistinguishable from each other because they both produce V-shaped transits. About 70% of the planet candidates in the CoRoT IRa01 field are best fit with an impact parameter of b >0.85, while less than 15% are expected in this range considering random orbital inclinations. By applying a cut at b < 0.85, meaning that ∼15% of the potential planet population would be missed, the candidate sample decreases from 41 to 11. The lightcurves of 6 of those are best fit with such low host star densities that the planet-to-star size ratii imply unrealistic planet radii of R > 2RJup. Two of the five remaining systems, CoRoT1b and CoRoT4b, have been identified as planets by the CoRoT team, for which the lightcurves alone rule out blended light at 14% (2σ) and 31% (2σ). One system possesses a M-dwarf secondary, one a candidate Neptune. Conclusions. We show that in the first CoRoT field, IRa01, 85% of the planet candidates can be rejected from the lightcurves alone, if a cut in impact parameter of b < 0.85 is applied, at the cost of a < 15% loss in planet yield. We propose to use this method on the Kepler database to study the fraction of real planets and to potentially increase the efficiency of follow-up. Key words. 1. Introduction With the CoRoT and Kepler space observatories in full swing (Baglin et al. 2006, Borucki et al. 2003), which both deliver thousands of lightcurves with unprecedented photometric pre- cision and cadence, we have moved into an exciting new era of exoplanet research. Now, the characterisation of small, possi- bly rocky planets has finally become a realistic prospective (e.g. Corot-7b, Leger et al. 2009; Kepler-10b, Batalha et al. 2011). One of the biggest challenges is to seperate real planets from the significant fraction of (astrophysical) false-positives that can mimic a genuine transit signal (e.g. Batalha et al. 2010). Ground- based transit surveys have revealed that stellar eclipsing bina- ries (EBs) blended with light from a third star are the main source of contamination (e.g. Udalski et al. 2002). Also, for Super-Earth planet candidates blends with a background tran- siting Jupiter-sized planet system can be important. In these sys- tems the eclipse depth, shape and ellipsoidal light variations of an EB are diluted by the effects of chance alignment of a fore- ground or background star or associated companion inside a photometric aperture set by either the pixel scale or the point spread function. In addition, light from a third star in the pho- tometric aperture can bias the fitted parameters of a planet tran- sit system. High resolution, high signal-to-noise spectra are nor- Send offprint requests to: [email protected] mally required to exclude binary scenarios by excluding their large radial velocity or bi-sector variations, a process that can be very time-consuming. Stellar blends are common in space-based transit surveys as apertures are relatively large (e.g. 19"x21" for CoRoT), and tar- get fields are crowded since the number of target stars is maxi- mized in this way. To weed out false-positives, the CoRoT team relies on an extensive ground-based follow-up campaign for on- off photometry to identify the transited star in the CoRoT aper- ture (Deeg et al. 2009) and high resolution imaging observa- tions to identify possible stars that dilute the lightcurve of a planet candidate. Even so, many candidates remain unresolved and defy easy characterisation after such a campaign. Kepler uses its unique astrometric precision to minimise the number of blends, which can be identified by a position shift of the flux cen- troid during transit, but will still require enormous ground-based efforts on the remaining ∼1200 candidates (e.g. Borucki et al. 2011). Together with the new influx of planet candidates from current surveys, possible future missions (such as PLATO; e.g. Catala et al. 2011) and ground-based efforts to hunt for planets around low-mass stars, the telescope demand for full follow-up may grow enormously. Therefore, any new technique or strat- egy that can eliminate even a moderate fraction of all candidates from the discovery lightcurves, prior to follow-up, is extremely valuable. 1 Please give a shorter version with: \authorrunning and/or \titilerunning prior to \maketitle In this paper we investigate to what extent information from the lightcurves themselves can identify blend scenarios and eliminate them as planet candidates and on the other hand rule out blend scenarios in the case of true planet systems. Our key motivation is that the lightcurves of blended systems can not be perfectly fit by pure transit models and neither can genuine tran- sits be fit by blended light models. In section 2 we introduce our lightcurve fitting procedure and in section 3 we apply it to simu- lated data of a transiting hot Jupiter and Super-Earth. While such a procedure provides a natural tool to distinguish blends from genuine planetary systems by lightcurve fitting, it breaks down for transits with high impact parameters. We therefore only con- sider transiting systems with impact parameter b < 0.85, loosing potentially ∼15% of the planet catch, but significantly decreas- ing (by an order of magnitude) the required amount of follow-up observations. In section 4 we apply our method to the candidates of the CoRoT IRa01 field, whose candidates are almost com- pletely characterised through an extensive follow-up campaign, and discuss the results in section 5. 2. Method 2.1. Transitfitting Several methods have been presented in the literature to identify blended systems and to select the best planet candidates. Seager & Mallen-Ornellas (2003) proposed a diagnostic that involves fitting a trapezoid to the transit lightcurve to obtain estimates for the transit parameters and subsequently identify the best candi- dates. In this paper we use a method very similar to that used by Snellen et al. (2009) to reject blend scenarios for the tran- siting hot Jupiter OGLE2-TR-L9. It involves least-square fitting of a lightcurve using the standard transit parameters (see below) plus an additional parameter representing the extra light from a third light source. If the fit is significantly better with extra light, the lightcurve is from a blended system. If this is not the case, an upper limit to the third light fraction can be set to a degree depending on the signal-to-noise of the data. This pro- cedure is in essence similar to Blender, which is used by the Kepler team (e.g. Torres et al. 2011). However, Blender simu- lates physical systems involving so many parameters that it is impractical to run on a large number of candidates. Here we are not interested in the true nature of the second object (whether it is a background, foreground or physically related star), just in its possible influence on the transit lightcurve. We assume at this point that lightcurves with obvious signs of the presence of a stellar binary, such as ellipsoidal light vari- ations and/or secondary eclipses, have been excluded from the candidate list. Note that a useful upper limit to the amount of ellipsoidal light variation, and the likelihood of a genuine plan- etary secondary, can be obtained by taking a Fourier transform of the data with the transit signal removed. We therefore do not require EBOP (Popper and Etzel 1981) to model the complex binary effects in the lightcurve, but rather utilize an IDL rou- tine that incorporates the analytical transit model of Mandel and Agol (2002;M&A). Our system simply consists of a secondary transiting a primary with possible additional light from a tertiary. 2.2. Transitparameters We treat the transit mid-time T0 and the orbital period P as fixed parameters, resulting from the candidate selection process. For extra simplicity we keep the limb darkening parameters fixed at the tabulated solar values for CoRoT white light, assuming 2 quadratic parameters (a,b)=(0.44,0.23) from Sing et al. (2010). Although this gives a small bias (<0.06 in impact parameter) for primary stars of different stellar type, the method is not meant for precise planet characterization and does not influence the char- acterization of potential blended and non-blended systems. Our transit model has three free parameters; the ratio of secondary over primary radii (R2/R1), the impact parameter of the transit b, which is the smallest projected distance of the centre of the sec- ondary to that of the primary in units of R1, and the density of the primary star ρ1. This density can be converted to the scaled orbital radius (a/R1), assuming that M1>>M2, through R1!3 a = G 3π ρ1 P2 (1) The relative projected distances z between secondary and pri- mary are computed from the input orbital phases φ, z(φ) = s a R1!2 sin(φ)2 + b2cos(φ)2, (2) Together with (R2/R1), these are used as input to a custom-made IDL program, incorporating the routine from M&A, that com- putes the theoretical models. We introduce light to this transit system by adding the blended light fraction k, Ftotal(φ, b, R1/R2, ρ∗, k) = Feclipse · (1 − k) + k, (3) where Feclipse is the original transit lightcurve. We then devise the following chi-square statistic to compare the lightcurve to the data Fobs,i with uncertainty σobs,i, (Fobs,i − Ftotal,i)2 σ2 obs,i χ2 =Xi (4) Note that we assume circular orbits. This has no influence on the characterization of blended and non-blended systems, but it does affect the derived host star density, and is therefore important for the estimate of the radius of the secondary object. This is further discussed in section 5. 2.3. MCMC To obtain the best-matching system parameters, we use a Monte Carlo Markov Chain χ2 optimisation technique (MCMC, e.g. Tegmark et al. 1998) to map out the probability distribution for each lightcurve parameter. MCMC is found to be a more ro- bust technique to obtain a global parameter solution in multi- parameter space than (downhill) grid-based methods, due to the resolution inefficiency of the latter (e.g. Serra et al. 2011). In the MCMC algorithm, the parameters pi are perturbed by an amount drawn from a normal distribution N according to: pi+1 = pi + f ·N · σp, where f is the jump function and σp the standard deviation of the sampling distribution for each p. Subsequently χ2 is recalculated for these perturbed parameters and a Gaussian likelyhood L ∝ exp(−χ2/2) is determined. These random jumps in parameter space are accepted or rejected according to the Metropolis-Hastings rule (Metropolis et al. 1953;Hastings 1970) . If the perturbed parameter set has a higher likelyhood L′ than its progenitor, it will be accepted as a new chain point, otherwise it will be accepted with a probability of L′/L. We run the algo- rithm many times to build up a 'chain' of parameter values and tweak σp and f such that ∼40% of the jumps are accepted. After creating multiple chains from different starting conditions, we Please give a shorter version with: \authorrunning and/or \titilerunning prior to \maketitle check proper model convergence and mixing of the individual chains using the Gelman & Rubin R statistic (Gelman & Rubin 1992). To save time, first k is set to zero at the minimum χ2 de- termined with MCMC analysis. Subsequently k is increased in small steps (but always kept fixed during the MCMC) with the previously found parameters as starting values. In this way the parameter values (adopting the median of the distribution) and the uncertainties in the parameters are determined as function of k in an efficient way. 3. Tests on synthetic lightcurves In this section, we test our method on synthetic lightcurves to determine the required precision to detect or exclude third light in a particular transit system. We perform these simulations for two candidate systems: (i) a hot Jupiter orbiting a solar type star and (ii) a Super-Earth around a similar host. 3.1. TransitinghotJupiter We simulated a set of transit lightcurves for a hot Jupiter with R2 = 1RJup and P=2.5 days, orbiting a star with a solar den- sity, for a range of impact parameters. The lightcurve for an im- pact parameter of b = 0.2 is shown in Figure 1. As explained in the previous section, our method finds the best fit for a range in blended light fraction k. Of course, in this simulation a perfect fit is obtained for k=0. As can be seen in Figure 1, an increas- ingly worse fit is obtained for increasing k, most obviously seen by comparing the k=0.95 model to the synthetic data. This latter model fit assumes that 95% of the light is from a third object, meaning that the unblended transit is actually a factor 20 deeper, hence 20% deep instead of 1%. It implies that R2/R1 ∼ 0.45, resulting in a much longer duration transit unless it is grazing. This results in the best-fitting k=0.95 model being much more V-shaped than the synthetic lightcurve of the planet. We can now convert the differences between the synthetic lightcurves and model fits to χ2 values for each combination of b and k by assigning uncertainties to the synthetic data. In this way we can determine what photometric precision is required to exclude a certain blended light fraction in the lightcurves as a function of b. Figure 2 shows the precision per 5 minutes of data required to exclude a blended light fraction k at a 3σ level in a system with an impact parameter of b=0.2, 0.5, 0.8, and 0.95. The re- quired precision becomes more stringent for lower values of k and higher values of b. For b=0.2, 80% blended light (k=0.8) can be excluded in a lightcurve with a precision of only 2 × 10−3 per 5 minutes, while for b=0.8, 20% of blended light can only be rejected if the lightcurve has a precision of 4 × 10−5 per 5 minutes. 3.2. Transitingsuper-Earth or- We performed also tests on a Super-Earth with R2 = 2.5R⊕ biting a sun-like star, following the same procedure as described above. Since the transit itself is a factor ∼ 16 more shallow than for a Jupiter-size planet, the level of precision required to re- ject blend scenarios is also significantly higher, as can be seen in Figure 3. Note however that even for a blended light fraction of k=0.95, the radius of transiting object R2 is still in the Jupiter- size regime. Hence only if the blended light fraction is very high, k > 0.95, can an eclipsing binary mimic a Super-Earth transit. Blend models for a Jupiter/Sun system k=0.2 k=0.95 1.000 0.998 0.996 0.994 x u l F 0.992 0.990 0.988 -0.030 -0.025 -0.020 -0.015 -0.010 -0.005 0.000 Orbital phase Fig. 1: Simulated lightcurve for a transiting exoplanet system consist- ing of a hot Jupiter in a 2.5 day orbit around a solar type star with impact parameter b=0.2 (black dots). The solid curves show diluted bi- nary models with best-fit parameters determined by MCMC, for blended light fraction k=[0.2, 0.5, 0.8, 0.95]. . i n m 5 / n o i s i c e r p d e r i u q e R 10-2 10-3 10-4 10-5 0.0 Rp=1RJup b=0.2 CoRoT b=0.95 0.2 0.4 0.6 0.8 3rd-light fraction Fig. 2: The photometric precision per 5 minutes required to exclude a blended light fraction k at 3σ for a hot Jupiter around a solar type host star (R2/R1 = 0.1 and ρ∗ = ρ⊙ ), as a function of the system parameters b and k. The four solid curves are for impact parameters b=0.2, 0.5, 0.8, and 0.95. The upper and lower horizontal dotted lines indicate the range in precision for objects in the IRa01 field, determined by Aigrain et al. (2009). 4. Tests on candidates in the CoRoT IRa01 field 4.1. Thedataset In this section we test our method on real data, using the lightcurves of the candidates selected by the CoRoT team from CoRoT field IRa01 (Carpano et al. 2009). In this first field tar- geted by CoRoT, 3898 bright stars were observed in chromatic mode (with a blue, green and red channel) and another 5974 in a single monochromatic "white" band in a 66 day staring run towards the Galactic anti-center. From the 50 initial candidates, a subsample of 29 promising targets received extensive follow- up as discussed in Moutou et al. (2009). Two of these have so far been identified as genuine planets: CoRoT-1b, a low density Rp = 1.49RJup transiting hot Jupiter around a G0V host (Barge 3 Please give a shorter version with: \authorrunning and/or \titilerunning prior to \maketitle 2.0•10-4 1.5•10-4 e d u t i l p m A 1.0•10-4 5.0•10-5 0 0.1 Fourier spectrum 1.0 Frequency (1/Days) 10.0 Fig. 4: Fourier diagram of an example noise spectrum prior to lightcurve cleaning. Amplitude of the best-fitting sine curve on the ver- tical axis is plotted against frequency. Peaks around frequencies of 1.0 and ∼14 are due to remaining systematics related to the satellite orbit and Earth's rotational period. Fig. 5: The MCMC solution for all IRa01 candidates in (b, (Rp/R )) space. Note the strong parameter degeneracy at high b. Yellow CoRoT WinIDs are the two confirmed planets CoRoT-1b and CoRoT-4b, blue objects are confirmed blends from the follow-up work presented in Moutou et al. (2009), and red sources are either unsettled cases or con- firmed genuine binaries with non-planetary secondary masses from the radial velocity variations. ∗ rotational period, caused by ingress and egress of the spacecraft from Earth's shadow, variations in gravity and magnetic field and changes in the levels of thermal and reflected light from the Earth (e.g. Aigrain et al. 2009). By folding the out-of-eclipse data onto the dominant frequencies of the Fourier diagram, we then fit a sinusoidal function to the remaining systematics, fol- lowed by median averaging over all transits. We subsequently binned the lightcurves and assign errors, according to the stan- dard deviation divided by the square root of the number of points in each bin. Fig. 3: As for Fig. 2, but for a 2.5REarth SuperEarth planet around a solar type host in a 2.5 day orbital. We can exclude 80% blended light at the 3σ level at a moderate impact parameter of b=0.5. The horizontal dashed line refers to the precision reached in the discovery lightcurve of CoRoT 7b, the first rocky SuperEarth planet (Leger et al. 2009). et al. 2008) and CoRoT-4b, a Rp = 1.19RJup hot Jupiter around a F8V host (Aigrain et al. 2008). Seventeen additional systems were solved using the photometric and spectroscopic follow-up observations (Moutou et al. 2009). We choose to test our method on the 45 bright candidates with more than one transit observed, using the publicly available N2-level data. 4.2. Pre-cleaningofthelightcurves We first combine the multicolor lightcurves into one single 'white lightcurve' for each candidate under the assumption that the CoRoT analysis teams did not detect any significant vari- ation of eclipse depth with wavelength, which would already have been a clear sign of blending effects. We first clip each lightcurve by removing outliers at the 5σ level. These outliers are mostly associated with the epochs at which the satellite passes the South Atlantic Anommaly (SAA) or moves in/out of the Earth's shadow. We then iteratively refine the mid-times T0 and the orbital period P using the Kwee-van Woerden method (Kwee & van Woerden 1956) and cross-correlation with a the- oretical transit model (e.g. Rauer et al. 2009). Individual transit events that show temporary jumps in flux, caused by the impact of energetic particles (mainly protons) onto the CCD ("hot pix- els"), are excluded from our analysis. For 16 out of the initial 50 CoRoT IRa01 candidates (32%) we had to remove one or more transits from the lightcurve that were affected by such particle hits. Each individual lightcurve was then phasefolded around every transit. To normalise the data, we fit either a first order polynomial in a small range in phase (±0.1 from mid-transit) around each transit or a higher order polynomial (order n=13) in a larger phase range (typically ±0.4 in phase), depending on which approach delivers the lowest rms in and out of eclipse and the least red noise (Pont et al. 2006). Figure 4 shows a typical ex- ample of the dominant frequencies still remaining after the poly- nomial fit. For most objects, distinct peaks exist around periods of 103 minutes and at 24 hours. We identify these peaks with remaining systematics, related to the satellite's orbit and Earth's 4 Please give a shorter version with: \authorrunning and/or \titilerunning prior to \maketitle 4.3. Fittingthelightcurves Each lightcurve is first fitted with the method explained in section 2, assuming k=0, yielding the starting parameter sets (R2/R1, b, ρ∗ ) for our blend analysis. In Figure 5 we show the re- sulting MCMC distribution of impact parameter b versus R2/R1 for all the 45 candidates. CoRoT WinIDs (a shortcut of the CoRoT run identification number, e.g. IRa01-E1-2046) for each candidate are indicated, with yellow for the two confirmed plan- ets CoRoT1b and CoRoT4b, in blue those candidates that have been confirmed to be blended systems by Moutou et al. (2009), and in red unsettled cases (either suspected early type stars with only few or very broad spectral lines for further radial velocity follow-up observations with HARPS or confirmed genuine EBs with non-planetary secondary masses). As can be seen, a large fraction of the candidates are, assuming no blended light, best fit- ted with a very high (often larger than unity) impact parameter. This is even more clear in the distribution of fitted impact pa- rameters as shown in figure 6. For 32 out of the 45 (∼ 70%) can- didates b > 0.85, while from geometric arguments it is expected that ∼ 15% of planets would be found at such a high impact parameter. Assuming that all eleven candidates at b < 0.85 are non-blended systems only ∼1.6 objects are expected at b > 0.85. Since our tests in section 3 have shown that it is very difficult to distinguish blends from non-blended systems at high impact pa- rameters due to their V-shaped lightcurves, we apply a cut in the candidate list at b < 0.85, knowing that we will potentially re- move only a small fraction of the planet yield, in the case of the CoRoTa01 field < 0.3 planets. From this it can be seen that it is highly likely that all candidates with b > 0.85 are blended and/or grazing eclipsing binaries. For the eleven remaining candidates we used the transit parameters from the k=0 model to refit the lightcurve with an increasing value of k, as outlined in section 3. In this way we redetermine the best fit solution and χ2 as a func- tion of k. As an example we show the best fit transit models for a range of k and the χ2 as function of k for candidate E1-4617 in Figure 7. As can be seen, the lightcurve can only be well fit- ted by models with a low k. E.g. the χ2 of the best fitting k=0.5 model is ∼ 40% higher than that for k=0. The 2 sigma upper limit for the fraction of blended light (∆χ2) is k=0.20. We per- formed this same analysis for all eleven remaining candidates for which the χ2 versus k plots are shown in the Appendix, together with their best fit lightcurves. None of these candidates are better fitted by a high k model than a low k model, indicating that all blended systems have moved out of the remaining sample since they are all fitted with a high impact parameter. For six objects a significant fraction of blended light can be excluded from the lightcurve alone, including CoRoT-1b and CoRoT-4b. It would therefore not have been necessary to check whether the variabil- ity in these candidates came from the target star or not and the follow-up could have immediately concentrated on radial veloc- ity measurements. All parameters of the remaining candidates are shown in Table 1. An additional cut in the candidate list is made us- ing a combination of the best fit mean stellar densities ρ∗ and R2/R1, as shown in Figure 8. Six of the candidates have host stars with densities corresponding to A-stars, resulting in unre- alisticly large secondary radii of > 2.0RJup. Note that there is currently no consensus on the upper limit of planet size, mean- ing that by setting a hard limit on planet radius we may exclude very large or bloated (hot) Jupiters. However, there are currently only 4 out of 219 transiting exoplanets reported with radii larger than 1.8RJup (www.exoplanet.eu). Also, the probability that the secondary is a mid-type M-dwarf rather than a genuine planet increases when considering larger radii. This results in a remain- ing planet candidate sample of 5 objects instead of the original 45 using arguments based on the lightcurve alone. These five ob- jects have been marked with filled symbols in Figure 8. Details on each system are discussed in Appendix A. N y c n e u q e r F 10 8 6 4 2 0 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 Impact parameter b Fig. 6: Distribution of fitted impact parameter of the CoRoT IRa01 can- didates. The distribution is strongly peaked around b=1.0, indicating a significant population of (blended) EB contaminants. For a genuine planet distribution we would expect a flat histogram that falls off at high impact parameter. The dotted vertical line indicates the b=0.85 cutoff we have proposed in this paper. WinID+CoRoTID 1126 0102890318 0330 0102912369 0203 0102825481 1712 0102826302 4108 0102779966 (R2/R1) versus ρ∗ 4617 0102753331 2430 0102815260 4073 0102863810 1736 0102855534 3724 0102759638 P 1.51 9.20 5.17 2.77 7.37 19.76 3.59 15.00 21.72 12.33 (cid:16) R2 R1(cid:17) 0.14 0.10 0.18 0.05 0.07 0.19 0.10 0.18 0.11 0.10 b 0.43 0.18 0.62 0.60 0.80 0.10 0.24 0.36 0.43 0.50 Log(cid:16) ρ∗ ρ⊙(cid:17) -0.16 -0.13 0.04 -0.88 -0.06 -1.42 -0.81 -0.08 -1.24 -1.33 (cid:16) a R∗(cid:17) 4.93 16.96 13.09 4.27 15.41 10.47 5.36 24.40 12.77 8.17 2σ 14% 31% 30% 93% 95% 20% 44% 67% 62% 78% Table 1: The candidate sample that survives the impact param- eter cut. The last six sources are excluded using a second cut because the fitted host star density indicates a secondary radius R2 > 2RJup. The last column indicates the 2σ upper limit to the blended light fraction k. 5. Discussion In this paper we investigated to what extent we can use the high signal-to-noise lightcurves of space-based exoplanet transit sur- veys to identify blended light scenarios, and eliminate them as planet candidates. We concentrated on the 51 exoplanet candi- dates from the first CoRoT IRa01 field (Carpano et al. 2009). About 70% of the 51 planet candidates in the CoRoT IRa01 field are best fit with an impact parameter of b > 0.85, which at face value already indicates that the candidate distribution is 5 Please give a shorter version with: \authorrunning and/or \titilerunning prior to \maketitle R2/R1 vs Host star density 0.10Rj 0.50Rj 1.0Rj 2.0Rj M0 host star G0 host star F0 host star A0 host star 0.5 0.0 -0.5 -1.0 -1.5 -2.0 ) n u s ρ / ρ * y t i s n e d r a t s t s o H ( 0 1 g o L 0.1 0.2 k 0.3 0.4 -2.5 0.0 0.1 0.2 R2/R1 0.3 0.4 5σ 3σ3σ 2σ 1.14 1.12 d e 2 r χ 1.10 1.08 1.06 0.0 0.01 0.00 -0.01 x u l F -0.02 -0.03 -0.04 -0.05 -0.04 (a) CoRoT-4617 -0.02 0.00 Orbital phase 0.02 0.04 (b) Fig. 7: Panel a): The reduced χ2 as function of blended light fraction k. The horizontal dashed line shows the 2,3 and 5σ rejection criteria. Our lightcurve model directly indicates an early type main sequence stel- lar host, with a 2σ upper limit for blended light of k ∼ 20%. The low stellar density implies a large secondary radius, rejecting the planet hypothesis. Panel b): best fitting EB models with blended light frac- tion k=[0.2,0.5,0.90,0.95], clearly showing that solutions with low k are favoured. Note that an orbit with an eccentricity of e=0.5, orien- tated in the right way, could increase the estimated stellar density to that of the Sun, and decrease R2 to 2 RJup. This ambiguity can be easily removed by taking a single spectrum of the star, resolving its spectral type. strongly contaminated by blended and/or grazing systems. We find that by cutting a candidate sample such that those objects with high impact parameter are removed, at the cost of losing a small fraction of potential planets, a significant reduction in required follow-up observations can be achieved. Of all candi- dates, only 5 remain in the final sample of which two are genuine planet systems, one is a low mass transiting M dwarf and one is a candidate Neptune. The V-shaped lightcurves of near-grazing planet systems are strongly degenerate with blended eclipsing binary systems and can therefore not be distinguished from each other. How many planets are potentially missed by invoking the cut in impact pa- rameter? Of the known transiting exoplanets, ∼6% has an im- pact parameter larger than 0.85 and ∼16% an impact parameter 6 Fig. 8: The R2/R1 size ratio versus the log of the stellar density for the CoRoT candidates in the IRa01 field, assuming k=0. The dotted lines mark the densities of A to M type main-sequence hosts. The five filled dots are the candidates that survive both our cuts in impact pa- rameter and secondary size. The two confirmed transiting hot Jupiters CoRoT1b and CoRoT4b are shown as yellow filled dots. Open circles are the candidates we have excluded using our cuts. Blue circles indi- cate sources which have been identified as blended EBs by the CoRoT team follow-up, and red circles are either unsettled cases in the CoRoT follow-up or systems identified as genuine EBs through their radial ve- locities. The four solid curves indicate R2=[0.10,0.50,1.0,2.0]RJup, as- suming the main sequence mass-radius relation of Cox (2000) for the primary. of more than 0.751. The cumulative probability of a particular transit at a given impact parameter greater or equal to a cutoff value bX and transit depth ∆F is given by: Pc(b > bX) = 1 + √∆F − bX 1 + √∆F = 1 + R2/R1 − bX 1 + R2/R1 (5) Note that this expression is different from the equation presented in Seager and Mallen-Ornellas (2003), because the maximum impact parameter in their formula is determined by the grazing condition bmax = 1 − R2/R1, yielding a minus sign in equation 5. For a 1RJup planet around a solar type star ∼22% would po- tentially be missed by setting the cut in impact parameter (∼6% according to Seager and Mallen-Ornellas). However, extremely grazing systems will be very shallow and of short duration and will therefore provide very limited physical information. For ex- ample, a grazing 1RJup with impact parameter b =1.05, will show a transit with a duration of 30% and only 20% of the depth of a transit with b = 0. Therefore, the actual planet loss fraction will be closer to the predictions of Seager and Mallen-Ornellas (2003), i.e. < 15%. In this paper we have made the assumption of circular or- bits, but radial velocity surveys teach us that such an assumption is not valid for longer periods (e.g. Butler et al. 2006). In addi- tion, Barnes (2007) shows that a planet with an eccentric orbit is more likely to transit by a factor of (1 − e2)−1 than a planet in a circular orbit with the same semi-major axis. A significant pop- ulation of transiting exoplanets with an eccentric orbit is there- fore expected for long duration space-based surveys. Because 1+e Vcirc the planet orbital velocity varies from q 1+e 1−e Vcirc to q 1−e 1 www.exoplanet.eu Please give a shorter version with: \authorrunning and/or \titilerunning prior to \maketitle between periastron and apastron in an eccentric orbit, transit du- ration can vary as function of e and ω (the angle of pericenter). This leads to a wrong fit of the host star density (e.g. Kipping 2010a, Tingley et al. 2011), therefore directly affecting our es- timate of the secondary radius R2. We therefore can not reliable make the planet-to-star ratio versus host star density cut in the eccentric orbit case for longer period planets (P > 3.0days). Fortunately, the fitted impact parameter, R2/R1 and blended light fraction k are not affected by an eccentric orbit. This means that we can still first apply a cut in impact parameter b < 0.85 and remove likely blends. To subsequently determine the real host star density it is sufficient to take a single high-resolution spec- trum to determine ρ1 and estimate R2. Using this spectroscop- ically determined density an upper limit to ecos(ω) can be set. One particular case in our sample is CoRoT-4617 with an orbital period of P=19.76 days. Assuming a circular orbit, the host star is estimated to have a density only ∼4% of that of the Sun, in accordance with an early B-star. This would imply that the ra- dius of the secondary object has R2∼8RJup. However, an orbit with an eccentricity of e=0.5, orientated in the right way, could increase the estimated stellar density to that of the Sun, and de- crease R2 to 2 RJup. This ambiguity can be easily removed by taking a single spectrum of the star, resolving its spectral type. The method presented here is designed to remove false- positives in candidate lists through the identification of blended light. We do not assign a likelihood of planetary nature to the remaining candidates, meaning that we do not assess whether these are genuine planet systems, we just removed those sys- tems which are not (except for a small fraction of collateral dam- age). However, it is anyway interesting to link blended light frac- tions to the population of random background eclipsing binaries. Assuming that 1:300 of field stars are eclipsing binaries (Devor et al. 2008), and 1:1000 stars have a transiting hot Jupiter, we require an average of 0.3 background stars within the PSF, and within the magnitude range set by the limit of blended light, to have an equal likelihood for the two scenarios, and to end up with half of the remaining objects as eclipsing binaries. For a typical magnitude (V=14.0) for the candidate star, taking into account the size of the CoRoT PSF, this is reached at a ∆mag = ∼1.5, corresponding to k=0.8. For 8 out of the 10 remaining targets this level of blended light is excluded at the >3σ level. For a typical limit of k<0.6, the background eclipsing binary can at most be 0.5 magnitudes fainter than the target star, mak- ing this scenario a factor ∼5 less likely. Do note however that this does not take into account physical triple systems, for which radial velocity follow-up is required to exclude them. Recently, the Kepler team have announced the discovery of ∼1200 planet candidates (Burucki et al. 2011). We propose to use the method presented here on this candidate list, to identify clear blend sys- tems to obtain a better estimate on the fraction of planet systems in this sample. 6. Conclusions In this paper we have investigated to what extent informa- tion from lightcurves of a space-based exoplanet transit sur- vey can identify blended light scenarios and eliminate them as planet candidates, to significantly decrease the required amount of follow-up time. If a lightcurve has sufficiently high signal- to-noise, a distinction can be made between a blended eclips- ing binary and a transiting exoplanet. We first have simulated lightcurves of stellar blends and transiting planet systems to determine the required signal-to-noise as a function of impact parameter and transit depth. Our simulations show that blend scenarios can be distinguished from transiting systems at low impact parameter. At high impact parameter, blended and non- blended systems both produce V-shaped transits and are indis- tinguishable from each other. We have subsequently tested our method on real data from the first IRa01 field of CoRoT, con- centrating on the 51 candidates already identified by the CoRoT team (Carpano et al. 2009). We show that 70% of the planet can- didates in the CoRoT IRa01 field are best fit with an impact pa- rameter of b > 0.85, whereas ∼15% are expected assuming ran- dom orbital orientations. By applying a cut at b < 0.85, meaning that ∼15% of the potential planet population would be missed, the candidate sample decreases from 41 to 11. The lightcurves of 6 of those are best fit with such a low host star density that the planet-to-star size ratio implies an unrealistic planet radius of R2 > 2RJup. From the remaining five, two systems, CoRoT- 1b and CoRoT-4b, have been identified by the CoRoT team as planets, for which the lightcurves alone rule out blended light at a 14%(2σ) and 31%(2σ). One other candidate is also consistent with a non-blended system, but is a late M-dwarf, which will always require radial velocity follow-up for confirmation since M-dwarfs can have similar radii as Jupiter mass planets. One other system consists of a candidate Neptune around a M-dwarf according to Moutou et al. (2009). We have therefore shown that 85% of the planet candidates can be rejected for the IRa01 field from the lightcurves alone. We propose to use this method on the Kepler database to study the fraction of real planets and to potentially increase the efficiency of follow-up. For long period candidates, possible non-zero eccentricity will affect the cut in planet-to-star ratio versus host star density, effectively increas- ing the sample size. However a single high-resolution spectrum would be sufficient to determine the real host star density and estimate the size of transiting objects. References Aigrain, S., et al. 2008, A&A, 488, L43 Aigrain, S., et al. 2009, A&A, 506, 425 Baglin, A., et al. 2006, 36th COSPAR Scientific Assembly, 36, 3749 Barge, P., et al. 2008, A&A, 482, L17 Barnes, J. W. 2007, PASP, 119, 986 Batalha, N. M., et al. 2010, ApJ, 713, L103 Batalha, N. M., Borucki, W. J., Bryson, S. T., et al. 2011, ApJ, 729, 27 Borucki, W. J., et al. 2003, Proc. SPIE, 4854, 129 Borucki, W. J., et al. 2011, ApJ, 728, 117 Butler, R. P., et al. 2006, ApJ, 646, 505 Catala, C., Appourchaux, T.,& Plato Mission Consortium, t. 2011, Journal of Physics Conference Series, 271, 012084 Carpano, S., et al. 2009, A&A, 506, 491 Cox, A. N., & Pilachowski, C. A. 2000, Physics Today, 53, 100000 Deeg, H. J., et al. 2009, A&A, 506, 343 Devor J., Charbonneau D., O'Donovan F.T., Mandushev G., Torres G., 2008, ApJ 135, 850 Hastings, et al. 1970, Biometrika, Vol. 57, No. 1. (1 April 1970), pp. 97-109 Geman & Rubin 1992, Statistical Science, Vol. 7, No. 4. (Nov., 1992), pp. 457- 472 Kipping, D. M. 2010, MNRAS, 407, 301 Kwee, K. K., & van Woerden, H. 1956, Bull. Astron. Inst. Netherlands, 12, 327 L´eger, A., et al. 2009, A&A, 506, 287 Mandel, K., & Agol, E. 2002, ApJ, 580, L171 Metropolis, N., Rosenbluth, A. W., Rosenbluth, M. N., Teller, A. H., & Teller, E. 1953, J. Chem. Phys., 21, 1087 Moutou, C., et al. 2009, A&A, 506, 321 Pont, F., Zucker, S., & Queloz, D. 2006, MNRAS, 373, 231 Popper, D. M., & Etzel, P. B. 1981, AJ, 86, 102 Rauer, H., et al. 2009, A&A, 506, 281 Seager, S., & Mall´en-Ornelas, G. 2003, ApJ, 585, 1038 Serra, P., Amblard, A., Temi, P., Burgarella, D., Giovannoli, E., Buat, V., Noll, S., & Im, S. 2011, arXiv:1103.3269 Sing, D. K. 2010, A&A, 510, A21 Smalley, B., et al. 2011, A&A, 526, A130 7 Please give a shorter version with: \authorrunning and/or \titilerunning prior to \maketitle Snellen, I. A. G., et al. 2009, A&A, 497, 545 Tingley, B., Bonomo, A. S., & Deeg, H. J. 2011, ApJ, 726, 112 Torres, G., et al. 2011, ApJ, 727, 24 Udalski, A., Szewczyk, O., Zebrun, K., Pietrzynski, G., Szymanski, M., Kubiak, M., Soszynski, I., & Wyrzykowski, L. 2002, Acta Astron., 52, 317 8 Please give a shorter version with: \authorrunning and/or \titilerunning prior to \maketitle Appendix In this Appendix we discuss in detail the sample of 10 remaining CoRoT candidates, that were selected using the cut in impact pa- rameter and were presented in Section 4 and Table 1. In Figures 9-11, we show for each candidate the blended light fraction k versus reduced χ2, and the best fitting blended light models for k=0.2, 0.5, 0.9, and 0.95. In Table 2, we show best-matching sys- tem parameters for the full CoRoT IRa01, assuming no blended light. Commentsonindividualsources SELECTED PLANET CANDIDATES FROM THE LIGHTCURVES ALONE E2-1126-0102890318 We find a 2σ upper limit for blended light contribution of k < 0.14, therefore a blend scenario can be excluded for this source at high confidence using the lightcurve alone. In addition, by assuming that the host star is on the main sequence, its mean density points to a ∼1.5RJup radius, well in the range of known hot Jupiters. Of course, this source is exoplanet CoRoT-1b (Barge et al. 2008). E1-0330-0102912369 We find a 2σ upper limit for blended light contribution of k < 0.31 from its lightcurve, meaning that only a small con- tribution of blended light is tollerated. Assuming the host star is on the main sequence, its mean density points to a ∼1.2RJup radius for the secondary. This object is identified as exoplanet CoRoT-4b (Aigrain et al. 2008). Eventhough the CoRoT-4b host star is of similar brightness as CoRoT-1b, the significantly longer orbital period, the residual variability (caused by a spotted rotating stellar photosphere) and the 1.8 times smaller transit depth are the causes of the lower confidence on blended light. E2-0203-0102825481 The 2σ upper limit for blended light is k < 0.3 from its lightcurve. Radial velocity follow-up by the CoRoT team showed this to be an eclipsing binary of a low-mass M dwarf and a G-type primary (Morales et al., in prep). Assuming the host star is on the main sequence, its mean density points to a ∼1.7RJup radius. Although not a planet, it is consistent with a non-blended system as found from our lightcurve fitting. Such systems always require RV follow-up since late M dwarfs and Jupiter-mass planets can have similar radii. E2-1712-0102826302 We find a 2σ upper limit for blended light contribution of k < 0.93. We can therefore only exclude a high contribution of blended light for this shallow (2.4 mmag) transit. This means that at 2σ confidence the true eclipse depth is less than 2.4% in the presence of blended light. The fitted host star mean density points to an early type or evolved system. HARPS radial velocity follow-up has confirmed that the host star is an evolved fast rotator and Moutou et al. (2009) conclude that a triple system is the most probable scenario. E1-4108-0102779966 Because of the poor signal-to-noise of this transit and the relatively high impact parameter b =0.8, the 2σ upper limit for blended light is k < 0.95, therefore only a very high contribution of blended light can be excluded for this candidate. Assuming the host star is on the main sequence, its density is slightly lower compared to the solar value, indicating a stellar radius of . However, spectroscopic follow-up with HARPS R1 ∼ 1.2R⊙ suggested that the host is a low mass (∼ 0.8M⊙ ) star. No additional follow-up has thusfar been obtained by the CoRoT team. CANDIDATES REJECTED DUE TO THEIR LARGE SIZE E1-4617-0102753331 The 2σ upper limit for blended light is k < 0.20, therefore a blend scenario can be excluded at high confidence for this source. Assuming the host star is on the main sequence, its very low density points to an early B-type primary with a K dwarf secondary. The planet hypothesis is rejected and no additional follow-up is therefore required judging from the lightcurve alone. Note that an orbit with an eccentricity of e=0.5, orientated in the right way, could increase the estimated stellar density to that of the Sun, and decrease R2 to 2 RJup. This ambiguity can be easily removed by taking a single spectrum of the star, resolving its spectral type. E2-2430-0102815260 We find a 2σ upper limit for blended light contribution of k < 0.44. Again, only a small contribution of blended light is toller- ated. Assuming the host star is on the main sequence, its mean density, consistent with an A type or evolved star, points to a ra- dius R2 > 2.5RJup. Radial velocity follow-up by the CoRoT team showed this to be a single lined eclipsing binary of a fast rotating host star and an early type M dwarf (Moutou et al. 2009). E2-4073-0102863810 For this source, we find a 2σ upper limit for blended light of k < 0.67. This object shows ∼4% deep eclipses around a host star that is ∼20% less dense than the sun. This candidate was introduced in the original list of Carpano et al. (2009), but is not mentioned in the follow-up paper of Moutou et al. (2009). With an anticipated secondary radius of ∼2.1RJup this object could still belong to the rare group of low mass stars or brown dwarfs. In the case of a stellar M5 secondary, the secondary eclipse would be detectable at ∼3.5 mmag in depth. E2-1736-0102855534 The 2σ upper limit for blended light is k < 0.62. The low mean density of the host star, consistent with a very early main se- quence or evolved star, points to a > 2.0RJup radius. Analysis of the lightcurve reveals a secondary eclipse at the 9σ level, which indicates the secondary is in fact a low mass star. CoRoT radial velocity follow-up has confirmed that the host star is a fast ro- tating early type star and the system is a single lined eclipsing binary. E2-3724-0102759638 For this source, we find a 2σ upper limit for blended light of k < 0.78. Assuming the host star is on the main sequence, its very low density points to an A type primary, therefore R2 > 2.0RJup. This object is listed both as a planet candidate and a binary by Carpano et al. (2009). 9 Please give a shorter version with: \authorrunning and/or \titilerunning prior to \maketitle 1.30 1.25 d e 2 r χ 1.20 1.15 5σ 3σ3σ 2σ 0.0 0.1 0.2 0.3 0.4 0.5 k 0.005 0.000 -0.005 x u l F -0.010 -0.015 -0.020 -0.025 -0.04 (a) CoRoT-1b 1.14 1.12 1.10 d e 2 r χ 1.08 1.06 1.04 1.02 5σ 3σ3σ 2σ 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 k 0.005 0.000 x u l F -0.005 -0.010 -0.015 -0.02 CoRoT-1b -0.02 0.00 Orbital phase 0.02 0.04 CoRoT-4b -0.01 0.00 Orbital phase 0.01 0.02 1.35 1.30 1.25 d e 2 r χ 1.20 1.15 1.10 5σ 3σ3σ 2σ (b) CoRoT-4b CoRoT-0203 0.00 -0.01 x u l F -0.02 -0.03 1.05 0.0 5σ 0.1 0.2 0.3 0.4 0.5 0.6 0.7 k -0.02 -0.01 0.00 Orbital phase 0.01 0.02 (c) CoRoT-0203 3σ3σ 0.96 d e 2 r χ 0.95 2σ 0.94 0.93 0.0 0.2 0.4 0.6 0.8 1.0 k CoRoT-1712 x u l F 0.002 0.001 0.000 -0.001 -0.002 -0.003 -0.004 -0.005 -0.04 -0.02 0.00 Orbital phase 0.02 0.04 10 Fig. 9: For each CoRoT IRa01 candidate: the blended light fraction k versus reduced χ2 (left panels) and the best fitting blended light models for k=0.2, 0.5, 0.9, and 0.95. (d) CoRoT-1712 3σ3σ Please give a shorter version with: \authorrunning and/or \titilerunning prior to \maketitle 2σ 1.16 1.15 d e 2 r χ 1.14 1.13 1.12 CoRoT-4108 0.002 -0.000 -0.002 -0.004 -0.006 x u l F 0.0 0.2 0.4 0.6 0.8 1.0 k -0.008 -0.02 -0.01 0.00 Orbital phase 0.01 0.02 (a) CoRoT-4108 0.01 0.00 -0.01 x u l F -0.02 -0.03 -0.04 -0.05 -0.04 0.1 0.2 k 0.3 0.4 (b) CoRoT-4617 CoRoT-4617 -0.02 0.00 Orbital phase 0.02 0.04 1.14 1.12 d e 2 r χ 1.10 1.08 1.06 0.0 5σ 3σ3σ 2σ 5σ 1.10 1.09 3σ3σ d e 2 r χ 1.08 2σ 1.07 1.06 0.0 0.1 0.2 0.3 k 0.4 0.5 0.6 (c) CoRoT-2430 CoRoT-4073 1.22 1.20 1.18 d e 2 r χ 1.16 5σ 3σ3σ 2σ 1.14 1.12 1.10 0.00 -0.01 -0.02 x u l F -0.03 -0.04 0.0 0.2 0.4 k 0.6 0.8 -0.015 -0.010 -0.005 0.000 0.005 0.010 0.015 Orbital phase (d) CoRoT-4073 Fig. 10: Figure continued 11 Please give a shorter version with: \authorrunning and/or \titilerunning prior to \maketitle d e 2 r χ 1.02 1.00 0.98 0.96 0.94 0.92 5σ 3σ3σ 2σ 0.0 0.2 0.4 k 0.6 0.8 CoRoT-1736 0.005 0.000 x u l F -0.005 -0.010 -0.015 -0.03 -0.02 -0.01 0.00 0.01 0.02 0.03 (a) CoRoT-1736 Orbital phase CoRoT-3724 1.25 1.20 1.15 d e 2 r χ 1.10 1.05 5σ 3σ3σ 2σ 0.0 0.2 0.4 0.6 0.8 1.0 k 0.000 -0.005 x u l F -0.010 -0.015 k=0.2 k=0.5 k=0.9 k=0.95 -0.03 -0.02 -0.01 0.00 0.01 0.02 0.03 Orbital phase (b) CoRoT-3724 Fig. 11: Figure continued 12 Please give a shorter version with: \authorrunning and/or \titilerunning prior to \maketitle WinID+CoRoTID 1319 0102729260 1158 0102763847 0288 0102787048 3787 0102787204 1857 0102798247 4591 0102806520 1136 0102809071 2430 0102815260 0203 0102825481 1712 0102826302 0399 0102829121 1736 0102855534 0396 0102856307 1126 0102890318 0330 0102912369 2755 0102918586 4617 0102753331 3724 0102759638 4290 0102777119 4108 0102779966 1531 0102780627 2009 0102788073 2774 0102798429 3010 0102800106 4300 0102802430 2604 0102805893 2648 0102812861 2328 0102819021 4998 0102821773 3425 0102835817 3854 0102841669 3952 0102842120 1407 0102842459 2721 0102850921 0704 0102855472 4073 0102863810 2329 0102869286 3336 0102876631 4911 0102881832 4339 0102903238 4124 0102926194 3819 0102932089 4467 0102940315 3856 0102954464 P(days) 1.70 10.53 7.89 0.86 0.82 4.30 1.22 3.59 5.17 2.77 33.06 21.72 7.82 1.51 9.20 4.39 19.76 12.33 2.21 7.37 2.38 10.85 1.61 23.21 5.81 3.82 3.68 4.51 10.08 1.19 1.14 13.48 5.17 0.61 2.16 15.00 1.87 1.39 2.17 1.36 1.51 1.57 16.45 16.56 (cid:16) rp r∗(cid:17) 0.17 0.27 0.06 0.26 0.07 0.29 0.09 0.10 0.18 0.05 0.13 0.11 0.34 0.14 0.10 0.26 0.19 0.10 0.14 0.07 0.09 0.25 0.29 0.22 0.12 0.38 0.10 0.12 0.14 0.32 0.05 0.08 0.27 0.29 0.08 0.18 0.13 0.04 0.26 0.07 0.41 0.30 0.19 0.49 b 1.09(0.011) 1.10(0.017) 0.90(0.016) 1.22(0.040) 0.91(0.013) 1.21(0.085) 1.03(0.018) 0.24(0.107) 0.62(0.006) 0.60(0.287) 0.85(0.017) 0.43(0.119) 1.32(0.035) 0.43(0.017) 0.18(0.119) 1.01(0.005) 0.10(0.090) 0.50(0.105) 1.05(0.010) 0.80(0.085) 0.91(0.009) 1.17(0.432) 1.19(0.133) 1.00(0.127) 1.00(0.025) 1.33(0.052) 0.92(0.070) 0.97(0.037) 0.88(0.011) 1.25(0.024) 0.94(0.050) 0.85(0.356) 1.02(0.013) 1.18(0.017) 0.62(0.043) 0.36(0.036) 1.04(0.432) 0.84(0.110) 1.12(0.010) 1.00(0.126) 1.37(0.041) 1.07(0.035) 0.98(0.049) 1.31(0.047) Log(cid:16) ρ∗ ρ⊙(cid:17)(error in(cid:16) ρ∗ ρ⊙(cid:17)) -1.36(0.004) 0.13(0.044) -0.97(0.024) -1.55(0.002) -0.78(0.015) -0.83(0.062) -1.61(0.002) -0.81(0.014) 0.04(0.016) -0.88(0.074) 0.57(0.243) -1.24(0.009) -1.90(0.030) -0.16(0.017) -0.13(0.042) -0.21(0.006) -1.42(0.001) -1.33(0.008) -2.77(0.010) -0.06(0.492) -0.68(0.020) -1.44(0.045) -1.32(0.003) -0.17(0.091) -1.02(0.006) -1.60(0.009) -0.82(0.010) -1.66(0.008) -0.19(0.067) -1.65(0.008) -1.40(0.003) 1.47(0.068) 0.49(0.040) -0.97(0.004) -1.38(0.005) -0.08(0.047) -1.41(0.586) -0.69(0.121) -1.97(0.013) -1.63(0.062) -1.74(0.005) -0.97(0.012) -0.86(0.010) 0.55(0.347) Table 2: The fitting parameters for our blend models when applied to the CoRoT IRa01 sample, assuming k=0. (cid:16) a R∗(cid:17) 2.14 22.50 8.01 1.17 2.04 5.94 1.41 5.36 13.09 4.27 67.81 12.77 3.90 4.93 16.96 9.72 10.47 8.17 0.86 15.41 4.49 6.88 2.12 30.33 6.27 3.04 5.42 3.24 17.19 1.34 1.59 74.27 18.45 1.46 2.45 24.40 2.19 3.12 1.57 1.50 1.47 2.73 14.25 42.02 13
1210.4337	1	1210	2012-10-16T10:05:08	Benchmark experiments with global climate models applicable to extra-solar gas giant planets in the shallow atmosphere approximation	[ "astro-ph.EP" ]	The growing field of exoplanetary atmospheric modelling has seen little work on standardised benchmark tests for its models, limiting understanding of the dependence of results on specific models and conditions. With spatially resolved observations as yet difficult to obtain, such a test is invaluable. Although an intercomparison test for models of tidally locked gas giant planets has previously been suggested and carried out, the data provided were limited in terms of comparability. Here, the shallow PUMA model is subjected to such a test, and detailed statistics produced to facilitate comparison, with both time means and the associated standard deviations displayed, removing the time dependence and providing a measure of the variability. Model runs have been analysed to determine the variability between resolutions, and the effect of resolution on the energy spectra studied. Superrotation is a robust and reproducible feature at all resolutions.	astro-ph.EP	astro-ph	Mon. Not. R. Astron. Soc. 000, 000 -- 000 (0000) Printed 21 January 2021 (MN LATEX style file v2.2) Benchmark experiments with global climate models applicable to extra-solar gas giant planets in the shallow atmosphere approximation V. L. Bending,1 S. R. Lewis,1 and U. Kolb1 1 Department of Physical Sciences, The Open University, Milton Keynes MK7 6AA 21 January 2021 ABSTRACT The growing field of exoplanetary atmospheric modelling has seen little work on stan- dardised benchmark tests for its models, limiting understanding of the dependence of results on specific models and conditions. With spatially resolved observations as yet difficult to obtain, such a test is invaluable. Although an intercomparison test for models of tidally locked gas giant planets has previously been suggested and car- ried out, the data provided were limited in terms of comparability. Here, the shallow puma model is subjected to such a test, and detailed statistics produced to facilitate comparison, with both time means and the associated standard deviations displayed, removing the time dependence and providing a measure of the variability. Model runs have been analysed to determine the variability between resolutions, and the effect of resolution on the energy spectra studied. Superrotation is a robust and reproducible feature at all resolutions. Key words: methods: numerical - planets and satellites: atmospheres 1 INTRODUCTION Over the past several years, work has been carried out in the field of three-dimensional numerical simulation of exo- planetary atmospheres using large-scale global climate mod- els (GCMs), as reviewed in, e.g., Showman, Menou, & Cho (2008). In such modelling, it is important to establish clear model benchmarks, such as that set up for Earth by Held & Suarez (1994). Little benchmark work has been carried out for exoplanetary studies, however, with the exception of that initially described by Menou & Rauscher (2009), who laid out an intercomparison test and investigated the response of the spectral igcm dynamical core to it, and further studied by Heng, Menou & Phillipps (2011a), who investigated the response of both spectral and gridpoint cores available for implementation in the fms model to both this and additional tests. This study follows the lead provided by these two pa- pers in carrying out a clear and reproducible intercompar- ison test, and adds to it further statistics and information, including online data, to facilitate model comparisons. Although many modelling studies have been carried out on the atmospheres of 'hot Jupiters', gas giant planets less than about 0.1 AU from their host star (e.g. Johnson 2009), few direct comparisons of model responses to these condi- tions have been made. It is imperative that model-dependent responses be identified and understood. Without such inter- comparison studies, it cannot be determined which elements of a simulation may correspond to the planet under study, c(cid:13) 0000 RAS and which are simply artefacts of a specific model, or a re- sult of the high sensitivity of these complex, non-linear sys- tems to the precise input parameters and initial conditions. Thrastarson & Cho (2010) have studied the effects of initial flow on the final state reached by their model, at atmospheric depths from 1 bar down to 100 bar, and found the final state to be highly dependent on the initial conditions chosen. In contrast, Liu & Showman (2012) carried out a similar study down to 200 bar, and found almost no dependence on the ini- tial conditions, a result that Liu & Showman (2012) suggest may be due in part to the different vertical profiles of the thermal forcing and the restoration timescale between the two studies. With little observational information available, it becomes vitally important to have a reference simulation against which all models can be compared. Whereas models of Solar System planets may be compared to observations of the planet in question to determine their effectiveness, the modelling of extra-solar planets must proceed from first principles, though in some cases phase curves may be utilised to gain information on the conditions at the photosphere. A wide variety of models are utilised in simulating hot Jupiter atmospheres, from 'shallow-water' models such as that used in Showman & Polvani (2011) to three- dimensional GCMs of varying complexity and underlying assumptions. Atmospheric depths studied range from 1 bar ('shallow') to 100s of bars ('deep'), and model atmospheric equations range from the one-layer shallow-water equations 2 V. L. Bending, S. R. Lewis, and U. Kolb through the shallow, hydrostatic primitive equations to the full three-dimensional Navier-Stokes equations (e.g. Dobbs- Dixon & Lin 2008). The sparc/mitgcm of Showman et al. (2009) couples a correlated-k radiative transfer implementa- tion to their dynamical core, while other models may utilise dual-band 'grey' radiative transfer (e.g. Heng, Frierson & Phillipps 2011b), or omit radiative transfer and utilise in- stead Newtonian relaxation towards a predetermined tem- perature state. Large variability between models shows that some as- pects of the situation have not as yet been positively de- termined using such models in their present form. Since the input parameters of even simple GCMs are poorly con- strained by the available data on the planets to which they are applied, the choice of parameter adds a further degree of uncertainty to the already variable results. This again il- lustrates the importance of a single, fully-specified test case from which differing model-dependent responses may be de- termined. This paper builds on work suggested by Menou & Rauscher (2009) and further studied in Heng et al. (2011a), and adds the puma (Portable University Model of the Atmo- sphere) model (Fraedrich et al. 2005) to those to which the test has been applied. Since the precise state of the model atmosphere is highly time-dependent, long-term statistics are produced to allow a quantitative comparison for future tests. Time means of the output fields are produced to gain an understanding of the overall characteristics of the mod- elled atmosphere, and the associated standard deviations to gain a quantitative understanding of the degree of variability in each field. These statistical analyses of the model output fields will allow future work to be compared on a more solid footing. Section 2 outlines the model used and the description of the experiment, including a full list of model parameters. Section 3 displays the results of carrying out the intercom- parison test with puma, and provides time mean and vari- ability statistics. The data files required to produce these plots are provided online. Section 4 provides a discussion of these results and comparison to the previous results of Menou & Rauscher (2009) and Heng et al. (2011a). Section 5 covers the conclusions drawn, recommending the use of time mean and standard deviation statistics and a minimum sam- pling frequency, noting the degree of correspondence to the results of the previous authors, and suggesting additional plots to capture further aspects of the simulation not seen when studying only wind and temperature fields. In com- mon with previous studies, and as expected from the work of Showman & Polvani (2011), a strong equatorial superro- tation is found, resulting in an offset temperature hotspot. 2 A MODEL FOR GAS GIANT EXOPLANETS puma is a simple global circulation model of the atmosphere, developed in its current form at the University of Hamburg (Fraedrich et al. 2005) and descended from the spectral code of Hoskins & Simmons (1975). Though designed for use on Earth, the basic equations of the model are applicable to any system about which similar assumptions can be made, mod- elling an atmosphere in hydrostatic equilibrium that can be assumed to be a 'thin shell' with respect to the radius of the (spherical) planet. These assumptions, notably the hydro- static balance and shallow atmosphere approximations, re- sult in a set of equations typically referred to as the primitive equations of meteorology. Its long heritage means that the model's dynamical core is well known and has been exten- sively tested, rendering it particularly useful for benchmark tests. The model runs in spectral space in the horizontal using a triangular truncation, resolution being specified by the truncation coefficient, such as T21, which equates to a grid of approximately 64 longitudes and 32 latitudes. In general, if the truncation number is Tn, the number of latitudes is given by Nlat = (3Tn + 1)/2. Particular resolutions, such as T42 (128×64), T63 (192×96), or T85 (256×128), are widely used as standard resolutions owing to the resulting grid sizes being entirely or almost entirely powers of two, which allows for greater efficiency in the fast Fourier transform routines. A finite-difference method is used in the vertical, with vertical levels defined using the sigma-coordinate (Phillips 1957): σ = p(λ, φ, z, t) ps(λ, φ, t) (1) where p is the pressure at any point in the atmosphere, and ps is the 'surface' pressure at the same λ, φ, t, with λ, φ, z, and t being longitude, latitude, height above the model base, and time, respectively. σ is then always 1 at the base of the model atmosphere, decreasing upwards. Model levels σ = ln, on which the variables are calculated, take values in the range 0 < ln < 1, with boundary conditions imposed at σ = 0 and σ = 1. In this gas giant case, the 'surface' is flat, determined by a reference geopotential, and the boundary conditions equate to a rigid surface at the top and bottom of the atmosphere, at which the vertical velocity is set to zero. Any level distribution may be input; however, for the purposes of this experiment, only linear spacing in sigma was used. Choosing a horizontal and vertical resolution requires a compromise to be made between the degrees of freedom of the model and the computing power required to run it within a feasible time-scale. Too low a resolution may result in misleading results as features vital to the flow are not represented. The resolution dependence of this experiment is discussed further in Section 4. The particular resolutions used in this paper are chosen to correspond with those of Menou & Rauscher (2009) and Heng et al. (2011a). puma has been modified for use with gas giant plan- ets, permitting all planetary parameters to be changed, and also to permit the usage of three-dimensional, customised temperature restoration fields. Equations 2 to 5 were added as a selectable alternate option to the standard, Earth-like temperature restoration field. These modifications permit a tidally locked forcing scenario to be implemented, as is nec- essary for hot Jupiter type planets, with extreme tempera- ture differences between the star-facing dayside and shielded nightside. (Menou & Rauscher 2009). Teq(λ, φ, σ) = T vert eq (σ) + βtrop(σ)∆Tθ(λ, φ) where eq (z) = Tsurf − Γtrop T vert zstra + z − zstra 2 (cid:16) (2) (cid:17) c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 Benchmark experiments for exoplanetary GCMs 3 Figure 1. Latitude-longitude plot of forcing temperature for the T42L15 run at σ = 0.7, showing the dayside 'hotspot' and cool nightside. The substellar point is at the centre of the image, and plots for other resolutions are identical. (cid:114)(cid:16) 1 (cid:26) sin(cid:0) π 2 (σ − σstra)/(1 − σstra)(cid:1) σ ≥ σstra Γtrop[z − zstra] (cid:17)2 + δTstra 2 σ < σstra 2 + 0 βtrop(σ) = and (3) (4) (5) ∆Tθ(λ, φ) = cos(λ) cos(φ)∆TEP eq The temperature produced by radiative-convective equilibrium, without winds or other factors, is represented by Teq, with the purely vertical part of this structure rep- resented by T vert and the purely horizontal part by ∆Tθ. The vertical element of the profile was chosen by Menou & Rauscher (2009) to match that calculated by Iro, B´ezard & Guillot (2005) for HD 209458b. Such radiative-convective equilibrium temperatures may also be computed analytically from first principles using models such as that of Guillot (2010), which work has been extended to include the ef- fects of scattering by Heng et al. (2012). A scaling factor βtrop is applied to steadily decrease the temperature dif- ference between the dayside and nightside until it becomes zero above the tropopause, represented by σstra and zstra. Similarly, the temperature increment at the tropopause is given by δTstra (Menou & Rauscher 2009). The dry adia- batic lapse rate is represented by Γtrop, and the mean 'sur- face' temperature (temperature at the base of the model atmosphere) by Tsurf , with the equator-to-pole temperature difference denoted by ∆TEP. Examples of the resulting tem- perature forcing pattern can be seen in Fig. 1 and Fig. 2. A 'cold spot' is produced on the nightside of equal magnitude to the dayside hotspot. As a result, the zonal mean, or lon- gitudinally averaged, forcing temperature is the same from equator to pole on every level of the atmosphere, producing the vertical structure shown in Fig. 2. Runs were carried out at T42, T63, and T85 resolutions, with 15, 20, and 20 levels respectively, linearly spaced in σ. A full list of model, planetary, and run-specific parameters can be found in Tables 1, 2, and 3, respectively. All param- eters are taken from Menou & Rauscher (2009), who chose the planetary parameters to correspond to parameters for c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 Figure 2. Zonal mean plot of forcing temperature. The longitude averaging causes the day-night temperature difference to cancel out, so that the average equatorial temperature is equal to that at the poles. Table 1. Table of model parameters Parameter Symbol Value Ndel Dissipation time-scale / day Rayleigh friction time-scale / day Newtonian relaxation time /day tdiss tfrc trest 5 × 10−3 8 ∞ 0.5 HD209458b, a well-studied inflated hot Jupiter, and Heng et al. (2011a). The hyper-diffusion time-scale on the smallest resolved scale is given by tdiss, while Ndel gives the order of that diffu- sion. As a result of the conservation of energy built into these models, energy 'builds up' at the resolution limit, requiring artificial diffusion to be applied to damp down and remove this effect, which would otherwise introduce unwanted ef- fects into the results. It is important to note that the hyper- diffusion time-scale applies to the smallest resolved scale in each run, and thus results in a weaker dissipation on any given scale for a higher-resolution run as compared with a lower-resolution one, effectively producing weaker overall diffusion. The effects of this can clearly be seen in Fig. 16, where increasing resolution permits the model to represent an energy cascade to smaller and smaller scales before being cut off by diffusion. The order and magnitude of the hyper- diffusion values are non-physical to an extent and require tuning to the simulation being run. A more detailed discus- sion of the effects of forcing and dissipation, with particular relevance to simulations such as that carried out here, can be found in Thrastarson & Cho (2011). Rayleigh friction applies an effective frictional force to drag the winds towards zero velocity on each level on which it is set, using the time-scale tfrc; a time-scale of infinity thus indicates that no friction is applied. The Newtonian relaxation time, trest, determines the rate at which the model temperature is forced towards the input radiative-convective equilibrium state Teq. The 'surface', or base of the model, is defined as be- ing at the 1-bar pressure, P0. The gravitational constant throughout the model is given by gp, and the specific gas constant and heat capacity at constant pressure by Rs and cp respectively. 4 V. L. Bending, S. R. Lewis, and U. Kolb Table 2. Table of planetary parameters Parameter Symbol Value Planetary radius / m Rotation rate / 10−5 rad s−1 Gravity / m s−2 'Surface' pressure / bar 'Surface' temperature / K Equator-pole temperature difference / K Tropopause temperature increment / K Tropopause height / 106 m Adiabatic lapse rate / 10−4 K m−1 Specific gas constant / J kg−1 K−1 Heat capacity / J kg−1 K−1 a Ω gp P0 Tsurf ∆TEP δTstra zstra Γtrop Rs cp 108 2.1 8.0 1.0 1600 300 10 2 2 3779 13226.5 Table 3. Table of run-specific parameters Parameter Symbol T42L15 T63L20 T85L20 No. of latitudes No. of longitudes No. of vertical levels Nlat Nlon Nlev 64 128 15 96 192 20 128 256 20 Each run was carried out for a minimum of 350 plane- tary sidereal days following the determined spinup period, or approximately 1,000 Earth days (Heng et al. 2011a). Records were made ten times per day, equally spaced from one an- other, and all time averaging periods are over the 350 days unless otherwise stated. In this paper, the term 'day' shall always mean planetary sidereal days for consistency, noting that Menou & Rauscher (2009) and Heng et al. (2011a) differ in their use of the term 'day', the former to mean planetary sidereal, the latter to mean Earth solar. 3 RESULTS Fig. 3 shows the time mean temperature field in colour, with standard deviation contours overlaid. The model level clos- est to σ = 0.7 was chosen in each case, and all time means are taken over the 350-day period covering model days 30-380. It can be seen that the magnitude of both mean temperature and standard deviation is approximately consistent between runs, with standard deviation increasing slightly with res- olution, in particular noting that the T85L20 run in plot (c) shows no 15 K standard deviation contour towards the poles. The differences in the shape of the standard deviation contours between T42L15 and T63L20 may also be partially due to the different locations of the model levels, at σ = 0.7 and σ = 0.725 respectively, an unavoidable result of using linear sigma spacing with different numbers of levels. The highest variance is found at ±30◦ N and approximately 100◦ E of the substellar point, where the equatorial jet has car- ried warm air from the dayside to the cool nightside, making the fluctuations caused by its north-south movement most apparent. Though small differences are apparent, all runs demonstrate very similar mean temperature fields and stan- dard deviations. The maximum mean temperature is 1698 K at T42, 1705 K at T63, and 1709 K at T85, with corre- sponding maximum standard deviations of 67 K at T42, 64 K at T63, and 66 K at T85. Figure 3. Temperature mean and standard deviation for different resolution runs. Colours denote mean temperature, contour lines the standard deviation in Kelvin. The substellar point is at the centre of the image, or (0,0). (a) T42L15 at σ = 0.7, (b) T63L20 at σ = 0.725, (c) T85L20 at σ = 0.725. Versions of the T42L15 latitude-longitude instanta- neous 'snapshot' plots in fig. 3 of Menou & Rauscher (2009) are also shown for comparison in Fig. 4. While they do not precisely replicate the pattern of those images, the same gen- eral features can be seen, with the strong equatorial jet and weaker reverse flow beyond ±30◦ N producing the tempera- ture distribution seen in Fig. 4. While the winds appear sim- ilar, the lack of a scale in Menou & Rauscher (2009) means that their strength cannot be directly compared in this im- age. Instead, reference must be taken from Fig. 8, which shows very similar zonal wind speeds to fig. 4 of Menou & c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 Benchmark experiments for exoplanetary GCMs 5 Figure 4. 'Snapshot' of T42L15 temperature and wind vectors for comparison with Menou & Rauscher (2009). The substellar point is at the centre of the image. (a) σ = 0.7, (b) σ = 0.37. Rauscher (2009). Although the mean state is robust, the precise form of the equatorial jet and the temperature dis- tribution at a given moment is unpredictable, as the model is highly non-linear and thus its evolution is extremely sensi- tive, rendering such plots of limited use for the comparisons required in a benchmark test. The zonal mean temperature field, also averaged over the 350-day time period, is shown in Fig. 5. Note the differ- ent temperature scale to that of Fig. 3, due to the different temperature range encompassed. The most notable depar- ture from the forcing state's simple vertical profile can be seen in the centre of the figure, between ±40◦ N, where it can be seen that two 'hotspots' have developed, correspond- ing to the location of the temperature peaks seen either side of the equator in Fig. 3. These plots show the greatest dif- ference in standard deviation, with the location of the con- tours differing noticeably, particularly towards the top and bottom of the model atmosphere, although this may to some extent be due to the limited resolution in the vertical. All runs are noticeably cooler towards the poles below around σ = 0.8 than those in fig. 6 of Heng et al. (2011a), which show temperatures of almost equal magnitude to those at the equator. By subtracting the time mean temperature field from the known forcing state and dividing through by the restora- tion time-scale, plots of the heating rate resulting from the forcing can also be derived, as shown in Fig. 6. It can be more c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 Figure 5. Time-averaged zonal mean temperature plots at differ- ent resolutions. Colours denote mean temperature, contour lines the standard deviation in Kelvin. (a) T42L15, (b) T63L20, (c) T85L20. clearly seen from this figure that average net cooling is thus experienced in the equatorially symmetric regions between 0◦ and ±45◦ N, while net heating is experienced in regions poleward of ±45◦ N. Heating is seen in all locations below σ = 0.9, as well as between ±30◦ N above σ = 0.2. The regions of negative heating correspond to the two warm re- gions noted previously in Fig. 5. The cooling regions appear to grow broader and the heating regions stronger, particu- larly towards the base of the model, with increasing resolu- tion. The plots in Fig. 7 display the time-averaged, zon- ally averaged zonal (west-east) wind. Positive contours show wind directed west-to-east, 'out of the paper', negative con- tours the opposite. A strongly superrotating equatorial jet is 6 V. L. Bending, S. R. Lewis, and U. Kolb Figure 6. Time-averaged zonal mean heating rate plots at dif- ferent resolutions. Note the units of 10−4 K s−1. (a) T42L15, (b) T63L20, (c) T85L20. Figure 7. Time average of the zonal mean zonal wind. Positive contours and colours show wind directed 'out of the paper', neg- ative contours and colours the reverse. (a) T42L15, (b) T63L20, (c) T85L20. seen between 25◦ S and 25◦ N and below σ = 0.2, with peak windspeeds of approximately 1,200 m s−1 in all cases, and there is a corresponding broader and weaker return flow out- side this region. Maximum and minimum mean windspeeds are found to be 1220, -687 m s−1 for the T42L15 run, 1200, -693 m s−1 at T63L20, and 1179, -698 m s−1 at T85L20, comparable to the Heng et al. (2011a) 'shallow hot Jupiter' spectral model using the same parameters. From these re- sults, differences of order 1% are noted when varying the resolution and hyperdiffusion. Heng et al. (2011a) note that mean winds are uncertain at the level of approximately 10% in their 'deep hot Jupiter' simulation when changing pa- rameters. The jet is unbounded at the frictionless base of the model, but closes towards the top at around σ = 0.2, with weak reverse flow above. This pattern corresponds well to that seen in both Menou & Rauscher (2009) and the Heng et al. (2011a) paper for the 'shallow hot Jupiter' test case, although again, Menou & Rauscher (2009) display only snapshots. The corresponding zonal wind snapshot for the T42L15 run is shown in Fig. 8, and is broadly similar to that of Menou & Rauscher (2009) fig. 4, with the most notable difference being that the greatest zonally-averaged westward windspeed is approximately 830 as opposed to 730 m s−1 at this particular instant in time. The time evolution of the temperatures and wind speeds at various representative areas of the T42L15 run up to day 100, for comparison to fig. 5 of Menou & Rauscher (2009), is c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 Benchmark experiments for exoplanetary GCMs 7 Figure 8. T42L15 zonal wind 'snapshot' plot, for comparison to fig. 4 of Menou & Rauscher (2009). Positive colours and con- tours show wind directed 'out of the paper', negative colours and contours the reverse. Figure 9. Representative U plots for the T42L15 run, as in fig. 5 of Menou & Rauscher (2009). Solid line represents zonal mean wind at the equator, dashed line the equatorial maximum. shown in Fig. 9 and Fig. 10. Rather than use single points, representative areas were chosen, averaging around the lon- gitude circles at ±87.9◦ N for polar values, and over the latitude band between ±4.2◦ N for equatorial values, which were also averaged over the longitude points between ±2.8◦ of the specified location (e.g. the substellar point). This en- ables greater reproducibility, and also helps to avoid model- specific issues in choosing the points: puma, for example, has neither points at ±90◦ N nor exactly 0◦ N. For all runs, the data were sampled ten times per day. The plots shown demonstrated extremely high variability over small time- scales, and were smoothed using a simple boxcar smooth of width 1 day, or 11 records, to better display overall trends. Such high variability shows a requirement for high-frequency sampling in the creation of these plots, as daily sampling does not capture the full extent of the variability and can miss high-frequency features altogether, or produce spurious signals through aliasing. Fig. 11 (a), (c), and (e) display the streamfunction on the model level closest to σ = 0.7. This demonstrates the cir- culation of the flow, and its direction, with a large, positive streamfunction value indicating strong clockwise circulation. Its associated variance can be seen in Fig. 11 (b), (d), and (f). It can be seen that there are two major circulation fea- tures, one to either side of the equator, centred around 60◦ E c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 Figure 10. Representative T plots for the T42L15 run, as in fig. 5 of Menou & Rauscher (2009). Light and dark red lines repre- sent temperatures at the sub- and anti-stellar points, respectively, green and dark green dotted lines temperatures at the east and west limbs respectively, and blue and dark blue dashed lines tem- peratures at the north and south poles, respectively. of the substellar point, while the greatest variability is found on the equator around 60◦ W of the substellar point. The circulation at this level strengthens slightly with increasing resolution, with its maximum value just exceeding 4 × 1010 m2 s−1 at T85. Fig. 12 shows snapshots of vorticity, a measure of the local rotation at each point, on the level nearest σ = 0.7 for both T42L15 and T85L20 runs, demonstrating the impact of increasing resolution. Although a degree of structure is visi- ble at T42, much finer, smaller-scale structure can be seen in the T85 plot, with long 'streamers' of high-magnitude vor- ticity visible that are washed out at T42 resolution. A single time mean is shown as the time mean vorticity plots are almost identical between resolutions. Though fields such as temperature and wind appear quite similar to one another even in snapshot forms, the vorticity clearly shows the ben- efit of higher-resolution runs, enabling much finer detail to be resolved. The plot in Fig. 13 displays the mean meridional circu- lation, a measure of the mass of air circulating about a given point. Positive contours indicate clockwise circulation, neg- ative anticlockwise; a typical MMC plot for the Earth would show positive contours between 0◦N and 30◦N, illustrat- ing clockwise circulation in which air rises over the equator and descends further north, followed by the inverse between 30◦N and 60◦N, and a further clockwise circulation between there and the pole (Peixoto & Oort 1992). Similarly, the in- verse pattern is seen in the southern hemisphere. Here it can be seen that there are two main circulation features, with air descending rather than rising over the equator (as ex- pected from Showman & Polvani 2011), and rising between ±30◦N and ±60◦N, with a weak reverse circulation towards the poles. The equatorial circulation contracts and weak- ens slightly with increasing resolution, with the two distinct peaks at different altitudes becoming more obvious Fig. 14 shows the local superrotation index, which is a measure of the degree to which the angular momentum of each element of the atmosphere exceeds that which it would have in solid-body rotation. The local superrotation index s is defined by s = m/(Ωa2) − 1 (6) 8 V. L. Bending, S. R. Lewis, and U. Kolb Figure 11. Streamfunction mean (left) and standard deviation (right) for different resolution runs. Plots (a) and (b): T42L15 at σ = 0.7; (c) and (d): T63L20 at σ = 0.725; (e) and (f): T85L20 at σ = 0.725. The substellar point is at the centre of the image. Mean contours: 1010m2s−1; standard deviation contours: 109m2s−1 where m is the axial angular momentum per unit mass of the atmosphere derived from the zonal mean zonal wind u, the longitudinal average of the zonal (east-west) component u of the wind field (Lewis & Read 2003). In general, the axial angular momentum per unit mass is given by m = Ωa2 cos2(φ) + ua cos(φ) (7) A global superrotation index S can also be calculated by integrating over the whole atmosphere: (cid:18)(cid:90) (cid:90) (cid:90) (cid:19) S = (ma2 cos(φ)/g)dλdφdp /M0 − 1 (8) where M0 is the same integral for the atmosphere at rest, or u = 0. A westerly wind (blowing west-to-east) over the equa- tor cannot be created from an atmosphere initially at rest simply by moving air parcels from other regions of the at- mosphere, since the maximum angular momentum available is that at the equator. The existence of superrotation is thus a signature of eddy processes occurring in the atmosphere, transporting angular momentum equatorward. A detailed study of superrotation under hot Jupiter conditions can be found in Showman & Polvani (2011). Only the T85L20 run is displayed, as the results for each run are visually identical. The maximum value lies between 0.56 and 0.57 in each case, while the minimum is −1 at the poles. Detailed study of the angular momentum budget over the course of each run reveals that the global superrotation index begins at S = 0, as expected from the model's initial- isation state of zero wind. It then climbs over the first 15 days to a value of 0.033 ± 0.003 in each simulation, indicat- ing that angular momentum is not fully conserved, and ad- ditional energy has been imparted to the atmosphere. With no diurnal tides, surface friction, or topographical features to provide this extra momentum, it is likely to have been acquired through model dissipation. Fig. 15 shows the time evolution of the global statis- tics RMS vorticity and RMS divergence, which are output at each model timestep, over the initial 400-day period of the T42L15 run. These values are the root mean square of the entire global vorticity and divergence fields. A stable state is reached after 25-30 days, after which the observed degree of variability is unchanged; the same is true at the other resolutions, although the mean values reached differ. c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 Benchmark experiments for exoplanetary GCMs 9 Figure 12. Vorticity day-100 snapshots and time mean for T42L15 and T85L20 runs. (a): T42L15 snapshot at σ = 0.7; (b): T85L20 snapshot at σ = 0.725; (c): T85L20 time mean at σ = 0.725. The substellar point is at the centre of the image. Note the difference in scale of plot (c), at 10−5 s−1 rather than the 10−4 s−1 of plots (a) and (b) As with the representative area plots of Fig. 9 and Fig. 10, this demonstrates that the spin-up time for puma under these conditions is approximately 25-30 days, after which data may be taken without fear of compromising the results with spurious spin-up effects. Fig. 16 shows the kinetic energy spectra for three resolutions. Notably, low (large-scale), even (symmetric) wavenumbers have much higher amplitudes than their odd- valued counterparts, due to the highly equatorially sym- metric, large-scale nature of the thermal forcing and final state. The dotted line has a slope of −3, that expected from an enstrophy-cascading range in two-dimensional tur- bulence (Kraichnan 1967). The majority of the spectrum lies closely parallel to this line, demonstrating that this regime holds over most modelled scales. This quasi-two- dimensional regime is to be expected from the scales reach- able by these studies, as the smallest resolved scale even at maximum resolution, T85, is still on the order of 103 km, with flow on this scale strongly constrained by the ef- fects of planetary rotation and atmospheric depth, rather than fully three-dimaensional turbulence (Houghton 1986). Higher wavenumbers correspond to smaller scale features, and the greater kinetic energy present at higher wavenum- bers in the higher resolution runs thus results in the greater detail and higher extrema seen most clearly in the vorticity plots of Fig. 12. In each case, the spectrum tails off sharply towards the run's wavenumber cutoff, with a slope of around −15. This sharp decrease near the cutoff is not linked to physical expectations and is a result of the model diffusion. c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 While diffusive processes do naturally occur, the limitations of modelling require that they must be represented at pro- gressively larger scales (lower wavenumbers) as the model resolution decreases, to avoid an unphysical build-up of en- ergy at the smallest resolved scales. In a true system, this energy would continue to cascade down to ever smaller scales and eventually be dissipated; the model system, however, cannot reach such small scales, and since it conserves en- ergy efficiently, must have additional dissipation applied. 4 DISCUSSION As the results presented have shown, although 'snapshot' plots are useful for capturing a view of how the simulated atmosphere is behaving, they are effectively irreproducible (contrast, for example, Fig. 4 of this paper and fig. 3 of Menou & Rauscher (2009)), and can only be compared qual- itatively, rather than quantitatively. They may even be mis- leading, if the 'snapshot' time is chosen while the model is far from the mean state. Time mean plots, as produced by Heng et al. (2011a), provide directly comparable results, as the mean flow would be expected to be very similar between any two simulations reaching the same stable state. However, the variability of the model is then lost without the addition of variance or standard deviation information. Such small-scale variability is another important point of comparison, as it demonstrates potential planetary 'weather' and also shows the mean transport of such quantities as heat and momen- 10 V. L. Bending, S. R. Lewis, and U. Kolb Figure 14. Time-averaged local superrotation index for the T85L20 run. Figure 15. Global statistics for T42L15. (a) Global RMS vortic- ity, (b) global RMS divergence. Figure 13. Mean meridional circulation. Positive contours cor- respond to circulation in a clockwise sense, negative contours to anticlockwise circulation. (a) T42L15, (b) T63L20, (c) T85L20 tum by transient waves. Taken together, therefore, plots of the mean and associated standard deviation are found to be more useful for the purposes of model intercomparison work. Resolution is found to make little difference to many of the mean results studied, with similar mean temperatures and winds recorded in all cases. The most visible differences between runs are seen in the standard deviation contours, denoting different levels of variability in different regions. The persistence of the large-scale features between runs and models (Menou & Rauscher 2009; Heng et al. 2011a) sug- gests that such features, with a powerful superrotating equa- torial jet and correspondingly offset temperature maxima and minima, are likely to be observed on true extrasolar planets subject to such extreme forcing conditions. The ob- Figure 16. Kinetic energy spectra for all runs. The spectra have been averaged over all levels and over a period of ten days. The blue lines denote the spectrum of the T42L15 run, the red lines T63L20, and the green lines T85L20. The total energy spectrum is shown as a solid line, while the zonal component is dashed and the eddy dash-dotted. The dotted line is a reference line with a slope of −3. c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 Benchmark experiments for exoplanetary GCMs 11 servations available appear to support this conclusion, ev- idenced by the results of Knutson et al. (2007), who fit a simple model of planetary brightness temperature to the phase variation of the HD 189733 light curve to obtain a temperature map of HD 189733b, Majeau, Agol & Cowan (2012), who use two different methods to gain a temperature map of HD 189733b from its secondary eclipse, and Snellen et al. (2010), who detected a 2 km s−1 CO blueshift imply- ing strong winds flowing from the dayside to the nightside of HD 209458b. However, very few such observations have been made, and sufficient resolution to confirm or deny the presence of features at scales much below the global is still unattained. For the present, model results thus remain the only available avenue of study for the majority of features. The analysis of Fig. 15 as well as Fig. 9 and Fig. 10 shows that the approximate spin-up time for puma is around 25-30 days, after which a stable state is reached and the starting conditions are effectively erased. This determines the number of records it is necessary to discard before av- eraging the data to avoid introducing spurious information, and may vary from model to model. This is a relatively short period of time compared to typical Earth spinup times, and extremely short in comparison to the times required to spin up a model of Jupiter itself, which receives very little exter- nal heating. Sampling frequency can visibly alter the appearance of the representative value plots, with the daily sampling used by Menou & Rauscher (2009) hiding the high variability seen on smaller time-scales, and use of a time filter together with a high sampling frequency is recommended. The kinetic energy spectra at all resolutions largely fol- low the −3 law of Kraichnan (1967), but demonstrate a sharp fall in energy at scales approaching the truncation wavenumber, with a slope of −15 or steeper. Each increase in resolution studied produces an extension of the enstrophy- cascading range before the point at which dissipation takes over is reached. Thrastarson & Cho (2011) suggest that model calculations with these large forcing amplitudes and short restoration and dissipation time-scales are likely to be- come over- or under-damped, and the use of energy spectra and fields such as vorticity are recommended to determine whether this is the case. 5 CONCLUSIONS Benchmark tests of GCMs for hot Jupiters have been con- sidered, and a variety of diagnostics produced and analysed. 'Snapshot' plots are of limited use for the purposes of an intercomparison study because the precise phase of waves depends sensitively on the initial conditions and the evolu- tion of the flow. The use of mean and standard deviation plots is therefore encouraged. The addition of plots such as streamfunction, mean meridional circulation, and superro- tation index to the simple temperature and wind fields are also suggested, as these alone do not capture all aspects of the simulation, although more care is required in interpret- ing some such plots. The use of high sampling frequencies is also found to be important, with the degree of variability seen in the 'rep- resentative area' plots of Fig. 9 and Fig. 10 demonstrably dependent on the frequency at which the data is sampled. c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 A sampling frequency of at least ten records per day is sug- gested, and the data time-filtered. The resolution of the model runs is found to have little effect on the overall mean and variability of the data. How- ever, much finer detail is captured at high resolution, as illus- trated in the vorticity plots of Fig. 12. Smoother fields such as temperature show less visible effect of this increased res- olution. The energy spectra show that the higher-resolution models continue to follow the −3 law to smaller scales be- fore dissipation takes effect, demonstrating that worthwhile data can be gained by increasing the resolution. The large-scale features of the temperature and wind fields correspond in form and magnitude to those of Menou & Rauscher (2009) and Heng et al. (2011a), to the extent that those can be determined from the results provided. This reinforces the growing consensus that such features, with a 'hotspot' offset from the substellar point by a strongly su- perrotating equatorial jet, are likely to be found on a variety of hot Jupiter type exoplanets. 6 ACKNOWLEDGEMENTS The authors are grateful for the support of STFC. VLB addi- tionally thanks both STFC and the Open University's Char- ter Studentships scheme, and would also like to thank Yix- iong Wang at Oxford University, whose correspondence has been helpful. We also thank the anonymous referee, whose detailed comments helped to improve various aspects of this paper. REFERENCES Dobbs-Dixon I., Lin D.N.C., 2008, ApJ 673, 513 Fraedrich K., Kirk E., Luksch U., Lunkeit F., 2005, Mete- orologische Zeitschrift 14, 735 Guillot T., 2010, A&A 520, A27 Held I.M., Suarez M.J., 1994, Bull. Am. Meteorological Soc. 75, 1825 Heng K., Menou K., Phillipps P.J., 2011, MNRAS 413, 2380 Heng K., Frierson D.M.W., Phillipps P.J., 2011, MNRAS 418, 2669 Heng K., Hayek W., Pont F., Sing D.K., 2012, MNRAS 420, 20 Hoskins B.J., Simmons A.J., 1975, Quart. J. R. Met. Soc. 101, 637 Houghton J.T., 1986, The Physics of Atmospheres, 2nd ed., Cambridge University Press Iro N., B´ezard B., Guillot T., 2005, A&A 436, 719 Johnson J.A., 2009, PASP 121, 309 Knutson H.A., Charbonneau D., Allen L.E., Fortney J.J., Agol E., Cowan N.B., Showman A.P., Cooper C.S., Megeath T., 2007, Nature 447, 183 Kraichnan R.H., 1967, Phys. Fluids 10, 1417 Lewis S.R., Read P.L., 2003, J. Geophys. Res. 108, 5034 Liu B., Showman A.P., 2012, preprint (arXiv:1208.0126v1) Majeau C., Agol E., Cowan N.B., 2012, ApJ 747, L20 Menou K., Rauscher E., 2009, ApJ 700, 887 Peixoto J.P., Oort A.H., 1992, Physics of Climate. Springer-Verlag, New York 12 V. L. Bending, S. R. Lewis, and U. Kolb Phillips N.A., 1956, Journal of Meteorology 14, 184 Showman A.P., Polvani L.M., 2011, ApJ 738, 71 Showman A.P., Menou K., Cho J.Y-K., 2008, ASP Confer- ence Series 398, 419 Showman A.P., Fortney J.J., Lian Y., Marley M.S., Freed- man R.S., Knutson H.A., Charbonneau D., 2009, ApJ 699, 564 Snellen I.A.G., de Kok R.J., de Mooij E.J.W., Albrecht S., 2010, Nature 465, 1049 Thrastarson H.Th., Cho J.Y-K., 2010, ApJ 716, 144 Thrastarson H.Th., Cho J.Y-K., 2011, ApJ 729, 117 c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000
1007.4520	1	1007	2010-07-26T17:54:36	On the orbital evolution of a giant planet pair embedded in a gaseous disk. II. A Saturn-Jupiter configuration	[ "astro-ph.EP" ]	We carry out a series of high-resolution (1024 X 1024) hydrodynamic simulations to investigate the orbital evolution of a Saturn-Jupiter pair embedded in a gaseous disk. This work extends the results of our previous work by exploring a different orbital configuration---Jupiter lies outside Saturn (q<1, where q= M_i/M_o is the mass ratio of the inner planet and the outer one). We focus on the effects of different initial separations (d) between the two planets and the various surface density profiles of the disk, where \sigma \propto r^{-\alpha}. We also compare the results of different orbital configurations of the planet pair. Our results show that: (1) when the initial separation is relatively large(d>d_{iLr}, where d_{iLr} is the distance between Jupiter and its first inner Lindblad resonance), the two planets undergo divergent migration. However, the inward migration of Saturn could be halted when Jupiter compresses the inner disk in which Saturn is embedded. (2) Convergent migration occurs when the initial separation is smaller (d<d_{iLr}) and the density slope of the disk is nearly flat (\alpha<1/2). Saturn is then forced by Jupiter to migrate inward when the two planets are trapped into mean motion resonances (MMRs), and Saturn may get very close to the central star. (3) In the case of q<1, the eccentricity of Saturn could be excited to a very high value (e_{S}~0.4-0.5) by the MMRs and the system could maintain stability. These results explain the formation of MMRs in the exoplanet systems where the outer planet is more massive than the inner one. It also helps us to understand the origin of the "hot Jupiter/Saturn" undergoing high eccentric orbit.	astro-ph.EP	astro-ph	On the orbital evolution of a giant planet pair embedded in a gaseous disk. II. A Saturn-Jupiter configuration Hui Zhang and Ji-Lin Zhou ABSTRACT We carry out a series of high-resolution (1024 × 1024) hydrodynamic simu- lations to investigate the orbital evolution of a Saturn-Jupiter pair embedded in a gaseous disk. This work extends the results of our previous work by explor- ing a different orbital configuration -- Jupiter lies outside Saturn (q < 1, where q ≡ Mi/Mo is the mass ratio of the inner planet and the outer one). We focus on the effects of different initial separations (d) between the two planets and the various surface density profiles of the disk, where σ ∝ r−α. We also compare the results of different orbital configurations of the planet pair. Our results show that: (1) when the initial separation is relatively large(d > diLr, where diLr is the distance between Jupiter and its first inner Lindblad resonance), the two planets undergo divergent migration. However, the inward migration of Saturn could be halted when Jupiter compresses the inner disk in which Saturn is embedded. (2) Convergent migration occurs when the initial separation is smaller (d < diLr) and the density slope of the disk is nearly flat (α < 1/2). Saturn is then forced by Jupiter to migrate inward when the two planets are trapped into mean motion resonances (MMRs), and Saturn may get very close to the central star. (3) In the case of q < 1, the eccentricity of Saturn could be excited to a very high value (eS ∼ 0.4 − 0.5) by the MMRs and the system could maintain stability. These results explain the formation of MMRs in the exoplanet systems where the outer planet is more massive than the inner one. It also helps us to understand the origin of the "hot Jupiter/Saturn" undergoing high eccentric orbit. Subject headings: planet-disk interactions - protoplanetary disks 1. Introduction The existence of more than 40 multiple planet systems have been affirmed so far. The observational facts show that almost one-fourth of them contain two or more planets locked Department of Astronomy & Key Laboratory of Modern Astronomy and Astrophysics in Ministry of EducationNanjing University, Nanjing 210093,China ; [email protected] -- 2 -- in the mean motion resonances (MMRs). This ratio keeps growing as new detection methods are adopted, e.g. the transit time variation method which is particularly suitable for detecting low-mass planets locked in resonance. If we set q = Mi/Mo, where Mi is the mass of the inner planet and Mo is the mass of the outer one, the resonant systems could be simply divided into two types: (1) q > 1, e.g. 55 cnc(Fischer et al. 2003), and (2) q < 1, e.g. Gliese 876(Marcy et al. 2001). The establishments of MMRs when q > 1 could be explained by the convergent migration of the planets (Masset & Snellgrove 2001; Masset & Papaloizou 2003). According to the theory of disk-planet interaction, planets embedded in a gaseous disk may undergo various types of migration depending on their masses. Low(around several Earth masses) or moderate mass(around Saturn mass) planets usually undergoes fast type I or III migration, while massive planets (comparable to Jupiter) opens a gap on the gaseous disk and undergoes slow type II migration. In the case of 55 cnc, for example, the inner planet has a minimum mass of 0.87MJ and the outer one has a minimum mass of 0.17MJ , where MJ is the Jupiter mass. For this mass configuration, the outer planet probably migrates inward faster than the inner one and catches the inner one in the 3 : 1 MMR. These kinds of processes have been explored numerically by several classic works (Snellgrove et al. 2001; Nelson & Papaloizou 2002; Papaloizou & Szuszkiewicz 2005; Kley et al. 2004, 2005), as well as our previous research(Zhang & Zhou 2010, hereafter Paper I). Although the process is complicated, the convergent migration and the establishment of MMRs are robust outcomes. If the above-mentioned mechanism also dominates the orbital evolution when q < 1, then divergent migration should occur naturally and the establishments of MMRs would become ineffective. However, observations show that the MMRs exist in exoplanet systems of this orbital configuration as well. Table 1 shows a list of exoplanet systems that contain two planets probably locked in MMRs. In fact, one may find that more than half of the resonant systems have q < 1. How do these MMRs form? There should be a mechanism that suppresses or even halts the fast inward migration of the low-mass inner planet, so that the massive outer one could catch it before it is swallowed by the star. If such a mechanism exists, it may also help us to better understand the formation of a "hot Jupiter/Saturn" which survives the inward migration. For terrestrial planets (tens of Earth masses or below), their inward migration could be halted by the density jump on the disk (Masset et al. 2006; Morbidelli et al. 2008). Other- wise, they could be captured by the high-order MMRs of an inward migrating giants(Zhou et al. 2006). These processes explain the formation of terrestrial planets in "hot Jupiter/Saturn" systems well, but could not account for the low-order MMRs in the systems whose inner planet is as massive as Saturn or even Jupiter, e.g. Gliese 876, HD 160691 and HD 128311. 2005; Fogg & Nelson 2007; Raymond et al. -- 3 -- So far, few works have considered the orbital configuration in which the two giant planets have a mass ratio q < 1. Kley et al. (2005) studied the orbital evolution of Gliese 876, which has q ≈ 0.31. They simulated convergent migration for the two planets and successfully represented the their observed orbital configuration, especially for the right range of eccentricities. To obtain convergent migration, they assumed that there is a preformed cavity at the center of the disk. The two planets are set inside the cavity so that the outer planet keeps losing its angular momentum to the disk and migrates inward, while the inner planet barely migrates until it is trapped by MMRs with outer one. This assumption in fact implies that gas accretion onto the center star should be so vigorous that the inner part of the disk is depleted much earlier before the formation and migration of the giant planets. Other processes like the magnetic field may also account for the inner cavity at the center of the disk. However, the relatively small radii of the cavity limit this assumption to specific cases. A more general and self-consistent mechanism to bring about convergent migration is required in a whole and regular disk. To construct a full-region disk instead of a ring, we adopt the Cartesian computational domain. Details of these domain settings can be found in paper I. As shown in Table 1, we note that most of the planet pairs trapped in MMRs have q ≈ 0.3− 0.6. A familiar and typical planet pair in this range is a Jupiter-Saturn pair, which is most suitable for studying the dynamics of this mass configuration. Therefore, following Paper I, we continue to investigate the orbital evolution of a Saturn(inner)-Jupiter(outer) pair embedded in a gaseous protostellar disk. We focus on the effects of various surface density slopes(α = −d ln(σ)/d ln(r)) and various initial separations d between the two planets. We will show the following. (1) Although under divergent migration, when Jupiter migrates inward, it could halt the inward migration of Saturn by compressing the inner disk in which Saturn is embedded. (2) When the initial separation is smaller than the distance from Jupiter to its farthest inner Lindblad resonance diLr and the density slope of the disk is nearly flat, the two planets may undergo convergent migration and then be trapped into MMRs. (3) The eccentricity excitation of MMRs overwhelm the damping of the gas disk. Saturn may get very close to the central star(as < 1AU) preserving a very high eccentricity. (4) We also compare the results between different orbital configurations -- q > 1 and q < 1. These results help reveal the orbital architecture formation of some resonant exoplanet systems, e.g. HD 160691 and Gliese 876, as well as the formation of the 'hot Jupiter/Saturn' undergoing high eccentric orbit. This paper is organized as follows: the model including the numerical methods and the initial settings are introduced in Section 2; our results and the analysis are presented in Section 3; some discussions are made in Section 4, and the conclusions are summarized in Section 5. -- 4 -- 2. Numerical setup 2.1. Physical model We simulate the full dynamical interaction of a system including a solar-type star, a Saturn mass giant(inner planet), a Jupiter mass giant(outer planet) and a two-dimensional (2D) gas disk. The star is fixed at the origin of the system with both the planets and the disk surrounding it; thus, the whole system is accelerated by the gravity of the planets and disk. For efficiency, we ignore the self-gravitating effect of the gas. Therefore, the gravity exerted on the gas only comes from the central star, the two giant planets and the acceleration of the origin. For numerical convenience, the gravitational constant G is set to 1. The solar mass (M(cid:12)) and the initial semimajor axis of Saturn (Rs0 = 5.2 AU) are set in units of mass and length, respectively. We locate Saturn initially inside the orbit of Jupiter; thus, the mass ratio of the two planets is q ≈ 0.33. Our aim is to investigate the formation of MMRs and the consequential evolution of the system with this orbital configuration. Considering that Saturn and Jupiter would mostly undergo type II migration, we adopt a relatively large viscosity of ν = 5 × 10−5 to accelerate the evolution, and run simulations up to 2500 − 5000 initial orbital periods of the inner planet. We adopt a polytropic equation of state and set the disk aspect ratio to be H/r = 0.04. The evolution of the gas under the gravity of the star and planets is solved by the 2D Godunov coded Antares, which is based on the exact Riemann solution for isothermal or polytropic gas. While the dynamics of the two planets under the potential of the star and gas disk are integrated by an eighth-order Runge-Kutta integrator, the global time step is the minimum of the hydrodynamical part and the orbit integration part. Details of the numerical method as well as the computational configuration have been well described in our previous works(Paper I). A comparison with other well-studied codes has also been presented in paper I. 2.2. Initial condition One of the issues that we expect to figure out is the effect of the surface density profile on the disk. In this paper, we try several typical density profiles, which are only the function of disk radii σ = σ0rα. As shown in Table 2, the initial density distribution varies from flat to very steep: σ0, σ0r−1/2, σ0r−1 and σ0r−3/2. σ0 is set to be 0.0006 in our units, which corresponds to a height-integrated surface density ∼ 200g/cm2. The density slope on gas -- 5 -- disk results in a pressure gradient. To ensure that there is no radial flow at the beginning, we set the radial velocity ur0 to be 0 and adjust the initial angular velocity of gas uθ0 = rΩg to balance the pressure and central gravity, where Ω depends on α. Another issue is the role of the initial separation between Saturn and Jupiter. As shown in Table 2, we choose three separations: d ≡ aJ0 − aS0 = 1, 0.5, 0.25. For numerical convenience, we set the initial semi-major axis of the inner planet (Saturn) as the length unit (the initial location of Saturn is always aS0 = 1). Then we adjust the initial locations of the outer one (Jupiter) to obtain different separations. When d = 1, Jupiter initially locates at aJ0 = 2. At such a large distance, the mutual interaction due to gravity is negligible, so the two planets could migrate independent of each other at the very beginning. When d = 0.5, Jupiter is initially located at aJ0 = 1.5. The position of the PJ : PS = 2 : 1 MMR, where PJ and PS are the orbital periods of Jupiter and Saturn, respectively is now at r ≈ 0.94, which is a little bit inside the initial location of Saturn. By doing so, Saturn passes through the 2 : 1 resonance of Jupiter soon after their release if divergent migration occurs. Furthermore, the first inner Lindblad resonance (at Ω = ΩJ + κ/m, where κ is the epicycle frequency and m = −2) of Jupiter is located around ∼ 0.88 − 0.94 depending on the density slope α. Now the separation between the two planets is smaller than the distance between Jupiter and its first(also the furthest) inner Lindblad resonance, d (cid:46) diLr. Thus, the inner Lindblad torque of Jupiter should be reduced by the existence of Saturn at the beginning of evolution. Finally, when d = 0.25, Jupiter initially locates at aJ0 = 1.25. This location ensures that Saturn passes through the PJ : PS = 3 : 2 MMR of Jupiter, the location of which is at r = 0.95. More importantly, this small separation d ≤ 3rmH allows mutual scattering due to gravity between the two planets at the beginning of evolution, where rmH = 0.085 is the mutual Hill radius of Saturn and Jupiter at this configuration: rmH ≡ ( MJ + MS 3M(cid:12) 1 3 ( ) aJ + aS 2 ). (1) The initial settings of the disk do not take into account of the gravitational perturba- tion of the planets. Instead, we adopt the "quiet-start" prescription to setup a dynamical equilibrium to ensure that the streamlines of the gas are always closed when the planet is growing. Basically, we set the initial masses of the planets to be negligible, then we fix their orbits and increase their masses to Saturn and Jupiter masses adiabatically. At the end of their growth we release them at the same moment and start the evolution. Details of this prescription have also been described in Zhang et al. (2008). -- 6 -- 3. Results 3.1. Divergent migration and suppression of inward migration Our main results are summarized in Table 2. There are two main variables: the density slope α and the initial separation d between the two planets. To avoid confusion, we present the results mainly according to the sequence of the initial separations d. And for each d, we present the effects of the different density slopes first and then explain these effects together. First, we start with a relatively large initial separation (d = 1). As introduced in the previous section, the mutual interaction due to gravity between Saturn and Jupiter is now negligible (d > 5rmH). Our results show that, at large separation, Saturn and Jupiter will generally undergo divergent migration or achieve equilibrium state, depending on the surface density profile: (1) When the disk is very steep (α > 1/2), Jupiter digs a clear gap and migrates outward after the release. In the meantime, Saturn migrates inward very fast until it clears its coorbital region(type III migration). Then Saturn starts to follow the viscous evolution of the disk, and its fast inward migration is halted or even reversed. This result is consistent with our previous result that the massive planet follows the movement of the global disk and moves outward when α > 1/2. Figure 1 shows the orbital evolutions of the two giant planets embedded in different surface density slopes when d = 1. (2) When the disk is nearly flat (α ≤ 1/2), both planets are under inward migration. Since Saturn is still surrounded by gas at the moment of release, it migrates inward much faster than Jupiter during the first stage of evolution. As Saturn keeps clearing its vicinity, inward migration is then reduced gently. In fact, we observed an equilibrium state where Jupiter and Saturn both stop migrating inward and maintain their separation when α = 1/2, see Panel (b) in Figure 1. This is also consistent with the results of Paper I, which showed that the direction of viscous movement changes sign around α = 1/2: r = 1 ∂Γν/∂r 2πrσ d(r2Ω)/dr = −3ν( 1 2 − α)r−1. where Γν is the viscous torque, Γν = 2πr2νσ ∼ r1/2−α, rdΩ dr (2) (3) which stays constant over different radii r when α = 1/2, by assuming a constant viscosity ν across the disk. According to Figure 1, one may find that both the inward and the outward type II migration of Saturn are suppressed. After several hundred orbits evolution, the whole disk -- 7 -- has been well separated into two parts -- an inner disk and an outer disk -- by the gap opened by Jupiter. Saturn has also dug a gap in the inner disk. Although it is much weaker than that of Jupiter, this gap ensures that Saturn will undergo type II migration. Because of the large initial separation and divergent migration at the beginning stage, the gaps of the two planets will not overlap soon. Thus, the inner disk and Saturn could be treated as a sub-system which is shepherded by the tidal torque of Jupiter, see Figures 2 and 3. Then the surface density profile α makes some differences. If the disk was nearly flat (α ≤ 1/2), Jupiter migrates inward gently and pushes the gas of the inner disk toward the central star. Thus the local surface density distribution of the inner disk is changed. As shown in Figure 4, the surface density profile of the inner disk is changed from flat (α ≤ 1/2) to relatively steep (α ≥ 1). When the local disk mass exceeds the planet mass, πapσ (cid:38) Mp, the migration of the planet is then disk dominated. To exceed the mass of Saturn, the disk requires a minimum local density σmin ≥ 1.5 × 10−4 inside the orbit of Saturn when Saturn is located at aS = 0.8, which is much smaller than the present density of the inner disk (σ > 6 × 10−4). This ensures that the type II migration of Saturn is dominated by the disk evolution, which means Saturn follows the movement of the inner disk. On the one hand, the gas which flows across the orbit of Saturn from outside to inside exerts an additional positive corotation torque on Saturn, see Panel (b) and (d) in Figure 4. On the other hand, as the inner part becomes denser, the inner disk tends to spread outward. However the outward diffusion of the inner disk is suppressed by Jupiter(as well as the disk outside Jupiter's orbit), thus this local density profile is maintained(Panel (a) and (c) of Figure 4). As a result, the inward migration of Saturn is slowed down or even halted. If the disk is relatively steep (α > 1/2), Jupiter tends to migrate outward. The inner disk could now expand outward, and Saturn follows the movement of the gas. However, the Hill radius of Jupiter increases as it moves outward and the expansion of the inner disk is limited by the increasing width of the gap. Thus the outward migration of Saturn is also slower than that of Jupiter, see Panel (c) and (d) in Figure 1. The above results indicate that Saturn could migrate slower than Jupiter under some conditions, when they are both undergoing type II migration. If the surface density of disk is steep, both Saturn and Jupiter will migrate outward and thus, results in divergent migration, while in a nearly flat disk, the inward migration of Saturn is suppressed. Due to the large initial separation, the two planets would achieve an equilibrium state(Panel (a) and (b) in Figure 1). How could they then approach to each other further and get into MMRs? One possible way is reducing their initial separation to enlarge the interactions between them at the very beginning of the evolution. -- 8 -- 3.2. convergent migration and MMR captures Next, we try a moderate separation d = 0.5 by setting Jupiter at aJ0 = 1.5 and Saturn at aS0 = 1. At this distance, the initial mutual interaction due to the gravity of the two planets is still not important (d > 4rmH). However the indirect interaction becomes significant since Saturn initially locates near the position of the 2 : 1 MMR with Jupiter, and furthermore, it is also close to the location of the farthest inner Lindblad resonance (m = −2) of Jupiter. Our results show that the surface density profile also plays a great role in the final results: (1) When α > 1/2, we get divergent migrations which are similar to the cases with d = 1. The divergent rate is higher when a larger α is adopted. An important difference is that Jupiter migrates inward for a while right after the moment of release and turns back outward while Saturn had moved far away inward, see Panel (c) in Figure 5. This phenomenon indicates that, at small separation, the total inner Lindblad torque exerted on Jupiter is weakened by Saturn(at the beginning of evolution). (2) When α ≤ 1/2, we find that Jupiter migrates inward much faster than it does in the d = 1 cases, and the inward migration of Saturn is substantially slowed down. At this condition, we observe a gently convergent migration between Saturn and Jupiter. This slow convergent migration then results in the 2 : 1 MMR of the two planets and Saturn is forced to migrate further inward by Jupiter. Their eccentricities are greatly excited and maintained by the resonance, especially for that of Saturn (eS ∼ 0.4 − 0.5), see Panel (a) and (b) in Figure 5. These results can be understood as follows. At the beginning of the evolution, Saturn sweeps out the gas at the positions of the inner Lindblad resonances of Jupiter. Thus, the total inner Lindblad torque exerted on Jupiter is weakened. The outer Lindblad torque then dominates the migration of Jupiter for a while. When α > 1/2, their orbits become divergent because Saturn migrates inward faster than Jupiter at the first stage of the evolution. When their separation becomes large enough, Saturn's effect vanishes. Thus, the inner Lindblad torque overwhelms the outer one again and Jupiter start to migrate outward. Panel (b) in Figure 6 shows the torques exerted on Jupiter versus time when α = 3/2. It is clear that the inner torque decreases for a while around the moment of release. Then the inner torque recovers and overwhelms the outer one, while Saturn is quickly dragged inward by the corotation torque at the beginning (Panel (c) of Figure 6). Then, as its coorbital region is cleared, Saturn follows the expansion of the inner disk and moves outward. Figure 7 shows the semi-analysis of each mth Lindblad torque (inner and outer) exerted on Saturn(obtained by the same method in paper I). One may find that the total outer Lindblad torque decreases a lot after the release while the inner one increases a little bit, and thus the net torque becomes positive. -- 9 -- The situation is different when α ≤ 1/2. Although the decrease rate of the inner Lind- blad torque is slowed down when the separation between Saturn and Jupiter is increasing, the outer Lindblad torque still dominates the migration of Jupiter and pushes it inward slowly. In the meantime, the inward migration rate of Saturn decreases even more. This is mainly due to three reasons. First, as we have mentioned before, the inward migrating Jupiter compresses the inner disk and the steepened local surface density profile makes the net torque exerted on Saturn vanish there. Figure 8 shows each mth (m = 2−80) Lindblad torque exerted on Saturn at different time points. It is clear that during the first 300PS0 evolution, the total outer Lindblad torque is greatly reduced to the same value as the inner one. Second, the gas outside Saturn is pushed inward by the tidal torques of Jupiter. Parts of the gas flow across Saturn's orbit and generate an additional positive corotation torque on Saturn, which drives Saturn to move outward(see Figure 9). Third, the disk between Saturn and Jupiter is heavily weakened by the gap created by Jupiter. As Jupiter approaches Saturn, more higher order(m > 2) inner Lindblad torques of Jupiter and outer Lindblad torques of Saturn are cleared. Thus, both parts of the disk, outside Jupiter and inside Sat- urn, push the two planets toward each other further. Finally, when the disk between them is swept out gently and a common gap forms finally(see Figure 10). Since this convergent migration is slow, the 2 : 1 MMR is then a robust outcome. When Jupiter captures Saturn into MMR, Saturn is forced to migrate inward to the vicinity of the central star, preserving a high eccentricity eS ∼ 0.4−0.5(Figure 5). This result is consistent with the observations, e.g. Gliese 876 and HD 128311. Faster convergent migration and higher MMR may be achieved by reducing the initial separation between the two planets further. At last, we try a small separation d = 0.25 by setting Jupiter initially at aJ0 = 1.25 and Saturn initially at aS0 = 1. Since d < 3rmH, the mutual interaction due to gravity between the giant planets becomes important at the beginning of the evolution. Our results show that, although the steep surface density prevents fast inward migration, it does not determine the direction of the migration as in the previous cases. Now, the two planets migrate inward even when α = 3/2. The results are similar to the cases that have d = 0.5 and α < 1/2, but the inward migration of the planet pair becomes much faster and more unstable. As the result of this fast convergent migration, Saturn is trapped into the 3 : 2 MMR with Jupiter when the disk is flat α = 0. For the other surface density slopes α ≥ 1/2, the 2 : 1 MMR is still the preferred outcome. The eccentricity of Saturn never gets higher than 0.3 in the 3 : 2 MMR, see Figure 11. -- 10 -- 3.3. Comparison between q > 1 and q < 1 As we have shown in the previous section and in Paper I, both q > 1 and q < 1 configurations could result in convergent migration of the two planets and the the planets being trapped in MMRs. The observations show that more than half of the exoplanet pairs trapped in MMRs are of q < 1 configuration. It should be interesting to compare the results of these two configurations. First, the major difference between the cases of q > 1 and q < 1 is the direction of the common migration of the planet pair after they are trapped into MMRs. When q > 1, the common inward migration is halted and the planet pair starts to migrate outward. When q < 1, common inward migration is preferred, see Figures 5 and 11. Since Jupiter is much more massive, Saturn in fact follows the migration of Jupiter when they are locked in MMRs. The migration of Jupiter is dominated by the torque balance between the gas disk inside and outside its orbit. Thus, the surface density slope of the gas disk is an important issue(presented in paper I). For a density slope that could result in convergent migration, e.g. α = 0, the relative position of the two planets makes a difference: when Saturn lies outside Jupiter (q > 1), the total outer Lindblad torque exerted on Jupiter is weakened and the inner Lindblad torque becomes relatively stronger. Thus the inward migration of Jupiter will be slowed down, halted or even reversed; when Saturn lies inside Jupiter (q < 1), the situation is also similar: the total inner Lindblad torque exerted on Jupiter is reduced when Saturn sweeps out the gas inside Jupiter and the net inner Lindblad torque exerted on Jupiter decreases. So, Jupiter will migrate inward and push Saturn inward as well. Second, the eccentricity of Saturn could be excited to a higher value in the q < 1 configuration. When Saturn lies outside Jupiter (q > 1), its maximum eccentricity could be excited to eS ∼ 0.2 by the 2 : 1 MMR and to eS ∼ 0.15 by the 3 : 2 MMR. When eS > 0.15, the 3 : 2 MMR breaks, and the eccentricity is damped by gas disk quickly. Due to the damping effect of gas, the eccentricity of Saturn never increases to eS > 0.2 even in 2 : 1 MMR. However, when Saturn lies inside Jupiter, its maximum eccentricity could be excited to eS ∼ 0.4 − 0.5 by the 2 : 1 MMR and the system is still stable(Figure 5). This difference could be understood by considering the different structures of the heavily perturbed disk. When Saturn is located outside Jupiter(q > 1), the gas disk would be separated into two parts after the two gaps overlap. As the eccentricity keeps growing, Saturn gets closer and closer to the out edge of the common gap or even cuts into the disk. If the resonance is strong enough, e.g. the 2 : 1 MMR, the effect of damping and excitation will allow the system to achieve equilibrium. Otherwise the MMR would probably break -- 11 -- at high eccentricity, e.g. in the 3 : 2 MMR(see Paper I). The situation seems to be the same in the q < 1 configuration since Saturn will meet the inner edge of the common gap. However, since the two planets tend to migrate inward together after the common gap is formed, Saturn usually forms an inner cavity at the very center(Figure 10). With both Saturn and Jupiter located within the cavity, the damping effect of the gas is negligible. Thus the eccentricity of Saturn could be estimated by mutual dynamical analysis. Denoting eS0, the eccentricity of Saturn before p + 1 : p resonance, and aJ0, aJ the semimajor axes of Jupiter before and after resonance respectively, we can obtain the eccentricity of Saturn eS which is excited by the resonance through the following equation(Malhotra 1995): e2 S = e2 S0 + 1 p + 1 ln( aJ0 aJ ). (4) For the 2 : 1 MMR, when eS0 = 0.01, aJ0 = 1.05 at T ≈ 1000PS0 and aJ = 0.58 at T ≈ 5000PS0, both the above equation gives eS = 0.54, and our result gives eS = 0.5, which are in good agreement with each other(see Figure 5). The third difference is the frequency and stability of the planet pair in MMRs. As shown here and in Paper I, the two planets undergo convergent migration more easily in the q > 1 configuration. Convergent migration and MMRs are robust outcomes for all the surface density slopes when q > 1. However, when q < 1, the two planets are under divergent migration in almost half of the simulations. It seems that, most of the exoplanet pairs trapped in the MMRs should be of the q > 1 configuration(where the inner planet is more massive than the outer one). However, by considering the above-mentioned differences and the migration process of the two planets, we find that the MMRs in q < 1 are more stable than that in q > 1. In the case of q > 1, Saturn and Jupiter usually reverse their migration to outward when they are locked into MMRs. As Jupiter moves outward, Saturn is pushed outward further and their orbits are in fact divergent. As the separation between them increases, the mutual interaction due to gravity becomes weaker and Saturn is easily scattered away by the corotation torque, especially at a high eccentricity(unless the gas has already depleted). For q < 1, there is no such a problem since the two planets migrate inward and form an inner cavity at the center. Our long time evolution also proves this stability (T ≥ 8000PS0, see Figure 12). This explains why the relatively high frequency of exoplanet systems have the configuration of q < 1 contain MMRs; see Table 1. This relatively high frequency may also be a result of observational bias where the massive outermost planet is easier to detect by the radial velocity method. -- 12 -- 4. Discussions In our simulations, the onset of convergent migration is much earlier than the emergence of the inner cavity. Therefore, in our results, the cavity should not account for the suppression of the inward migration of Saturn. We would like to indicate that there are two essentials that make convergent migration happen: the steepened density slope on the inner disk(compressed by the pre-formed giant planet) and the proper separation between Saturn and Jupiter. The disk discussed here should be gas-full, otherwise it could not support the long range type II migration of massive planets. This usually happens at the early stage of the evolution of a protostellar disk. When a massive planet is forming, say Jupiter, it will substantially change the structure of the gas disk by digging a clear gap. The gas interior its orbit will be pushed inward and accumulates on the inner disk. If the timescale of gas accretion onto the central star(and other processes leading to loss of gas at the center) is longer than the timescale of this compression, the local density slope on the inner disk will steepen. The typical gas accretion rate of T Tauri star is around ∼ 10−7M(cid:12)yr−1, while the gap opening process is only ∼ 100 orbits if we drop a mature Jupiter in an unperturbed disk(in fact, we have already employed a 'quiet start' method in our simulations to simulate the growth of planets and avoid the unreal initial impact to the gas. The planets all start with a mass of 0.1M⊕ Earth masses, see section 2.2 and paper I). If we take into account the growth of the planet(core and gas accretion), the gap opening process will be prolonged. However, according to the core accretion model of giant planet formation, for the core mass stays below 15M⊕ most of the time. Since the lowest mass required to open a gap in a typical protostellar disk(H/r = 0.04, ν = 10−5) is around 30−50M⊕, the planet would not change the disk significantly during this stage. As soon as its core mass reaches the critical mass Mcore = 15M⊕ and equals to the mass of its gas envelope, the gas accretion steps into the runaway accretion phase(Pollack et al. 1996) and the total mass of planet would increase to 1 Jupiter mass within several hundred orbits(d'Angelo & Lubow 2008). That means, once the planet is massive enough to dig a gap on the disk, the gap will be opened and enlarged very quickly. So the gas accumulation on the inner disk should be a natural outcome when Jupiter is forming within the disk. Besides the surface density factor, a proper separation between the two giant plan- If the separation is small, the gas disk between Saturn and Jupiter ets is also required. would be weakened substantially when the two planets are digging gaps on the disk. For Jupiter(Saturn), the torques that come from the gas inside(outside) its orbit are reduced and it would be pushed inward(outward) by the disk outside(inside) its orbit. And, as the two planets get closer, the torque unbalance of each planet becomes more serious and the two planets would get closer. Thus, the total effect is that the gas disk tends to keep pushing -- 13 -- the two giant planets toward each other when the gas between their orbits becomes more and more tenuous. When the two planets get close enough(the gaps usually had already overlapped), the mutual gravitational interaction between the two planets would prevent them from further approaching, e.g. the MMRs. Our results show that the maximum separation leading to convergent migration is ap- proximately d ≤ 5rmH(when α = 0). According to the core accretion model, the embryos of planet will accrete all the solid mass within their vicinities and achieve their isolation masses. This isolation separation between embryos is also diso ∼ 5rmH. If we take into account of the migration of light planets(embryos) in the gas disk, the planets may get closer before they become massive enough to open gaps on the disk. Furthermore, the massive gas giants usually form in an area a little further outside the snow line, where the surface density of disk increases significantly due to the icing of water, and the collision timescale of solid grains is not long enough to prevent the effective growth of the planet core. So, it is reasonable to expect that some giant planet cores emerge with proper separations. This means that the initial conditions leading to convergent migration should not be rare, despite q > 1 or q < 1(q is the mass ratio of the inner planet and the outer planet). To focus on the effects of the surface density slope of the disk and the initial separa- tion between the two planets, we assume the two planets form simultaneously and the gas accretions onto planets(and the growth of planet) are not taken into account in this series of works(in both paper I and II). However, this does not mean that the accretion process is negligible. In fact, the gas accretion should become more vigorous on Saturn, whose gap formation process is well prolonged by Jupiter. Thus, we add some additional discussions about the accretion processes of the planets here. Many factors need to be considered when the mass growing of the planet is included, e.g. the formation sequence of the two planets, the time required to build a critical mass core, and the proper descriptions of accretion rate and range. We find that it becomes too complicated to concentrate on the dynamic evolu- tion of the planet pair if all the factors are taken into account. To avoid the complexity, one could take a reasonable assumption that Jupiter forms first while Saturn is still a planet core undergoing gas accretion. Although, in a multiple planet system, the accretion process of planet makes the orbit evolution more complicated by leading to mass of the planets grow- ing and various migration rates, we believe that the it may not change our main result -- the convergent migration -- qualitatively. First of all, the accretion onto Jupiter could be neglected since the clear gap prevents the effective gas accretion process. Then we only need to consider the accretion onto Saturn. One could further assume the core of Saturn is already around 15M⊕ and it undergoes fast type I migration initially. Since the direction of type I migration is always(or in most -- 14 -- cases) inward, the convergent migration should still be a robust outcome when the planet core(Saturn) lies outside Jupiter(q > 1). Comparing to the cases we studied in this paper, the growing Saturn may get closer to Jupiter and be trapped in MMRs with higher p(p+1 : p is the orbit ratio of the two planets). As the mass of Saturn core keeps increasing, the high-p MMRs would become unstable and lead to breaks or re-captures of MMRs(the resonances may overlap easily for massive planets at high p and leads to instability, see paper I). When the planet core lies inside Jupiter(q < 1), the two planets would undergo divergent migration at the beginning. However, the accretion process is fully runaway when the core mass is above 15M⊕. It will take only 200 − 400 orbits to achieve 1MJ for an accreting planet embedded in a disk as dense as the one we adopt here and its orbit decay is less than 20% during this process(d'Angelo & Lubow 2008). When the mass of the inner planet becomes massive enough to open a clear gap, its migration rate will decrease significantly. Furthermore, as it keeping accreting gas, the inner planet should become more massive(than Saturn) and dig a wider gap, which weakens the disk between the two planets further. Thus, the gas outside the planet pair(inside inner planet and outside outer planet) tends to push the two planets toward each other(the inner planet only gets angular momentum from inner disk and move outward, the outer planet only loses angular momentum to outer disk and move inward) and results in a more compacter orbital configuration. So, it may also lead to the convergent migration in the q < 1 configuration, when we consider the gas accretion onto the Saturn core. A more self-consistent simulation should include two accreting and interacting planet cores as well as the thermal evolution of the gas around planets, which would lead to much more complicated orbital evolution and thus stepping into a new aspect of orbital evolution investigation. The relational results are under preparation now. We also note that the characteristics of resonant systems with q < 1 are the relatively large eccentricity and the short orbit period of the inner giant planet, e.g. GJ 876, HD 128311, HD 45364 and HD 60532. These characteristic orbits are easier to detect by radial velocity methods. Thus, this observational bias also explains the relatively high frequency of the resonant system in the q < 1 configuration. The orbital eccentricities of the planets in these real systems are all less than 0.3, which indicates that there may exist other mecha- nisms that restrain the planets' eccentricities when the gas damping is absent. One possible mechanism which need to be addressed further is the interaction with planetesimal disk after the depletion of gas. The tidal dissipation that arise from the star will drive the resonant planet pair out of resonance, when they are close to the center(D.N.C. Lin et al. 2010, in preparation). The eccentricity of the inner planet could be damped by this tidal dissipation as well. So the 'hot -- 15 -- Jupiter/Saturn' would probably form in the q < 1 system as follows: at first, the fast inward migration of the inner planet is suppressed by the outer giant planet. Then the two planets are locked into MMRs and migrate inward together. They dig an inner cavity at the center of gas disk, and the eccentricity of inner planet is excited to a high value by the resonance. As the inner planet gets closer to the star, the tidal dissipation gently drives it out of the MMR with the outer one. After the depletion of gas, the inward migration of outer planet stops, while the inner one falls toward the center due to tidal interaction with the star. Finally, the two planets undergo divergent migration again and the inner one approaches the central star more closely with a moderate eccentricity which had been damped by the tidal dissipation of the star. To ensure the validity of this process, we need to carry out further N-body integration associated with tidal damping, by adopting the results of this work as the initial conditions. 5. Conclusions Following the paper I, we continue the investigation of the orbital evolution of Saturn and Jupiter embedded in a protostellar disk by running a series of 2D high-resolution hy- drodynamic simulations. The main aim of this work is to find out whether the convergent migration also happens in a system with q < 1(where the more massive giant planet is ini- tially located outside the lighter one). To do so, we switch the initial positions of Saturn and Jupiter to achieve q < 1 and focus on the effects of various surface density profiles of the gas disk and different initial separations between the two planets. From our results and analysis, we summarize our conclusions as follows: (1). The type II migration of Saturn could be suppressed by Jupiter when q < 1. As Jupiter digs a deep gap, the gas disk is cut into two parts -- an inner disk and an outer disk. Saturn also digs a gap on the inner disk and follows the viscous movement of gas. Being shepherded by the tidal torque of Jupiter, the expansion of inner disk is suppressed. When the surface density slope is steep α > 1/2, Jupiter migrates outward and the inner disk expands. The width of gap increases as Jupiter moves outward(aJ increases) and the expansion of inner disk is limited, thus the outward migration of Saturn is limited as well. When the surface density is nearly flat α ≤ 1/2, Jupiter tends to migrate inward. Being compressed by Jupiter, the inner disk becomes denser and the local surface density slope α increases. Thus the inner disk tends to spread outward and fight against the tidal compression of Jupiter. When the expansion and compression effects achieve equilibrium, the inward migration of Saturn is slowed down or even halted. In a system of q < 1, this mechanism provides a way to halt the inward migration of -- 16 -- the inner giant planet and does not require the overlapping of gaps. In fact, this suppression happens as early as the outer massive planet starts to compress the inner disk and makes the convergent migration possible. We also note that, the inner planet should also be massive enough to open a gap on the disk, otherwise it won't follow the viscous evolution of the inner disk and this mechanism won't be valid. The lowest mass required should be 30 − 50Me, depending on the scale height and viscosity of the disk. (2). Convergent migration could also happen in q < 1 configuration under some circum- stances. The two main factors that account for the occurrence of the convergent migration: are the nearly flat surface density profile and the relatively small separation between the two planets. On the one hand, as we have concluded above, when α ≤ 1/2, the inward migration of Saturn would be suppressed by the steepened local surface density slope on the inner disk. On the other hand, as the initial separation between the two planets is reduced, the total inner(outer) Lindblad torque exerted on Jupiter(Saturn) is weakened by the gap created by Saturn(Jupiter). Thus, the outer torque overwhelms the inner one and pushes Jupiter inward faster than the regular type II migration. In the meantime, the inward migration of Saturn is suppressed and it migrates more slowly than Jupiter. As a result, the net orbital movement of the two planets is convergent. The 2 : 1 MMR is a usual outcome when Jupiter and Saturn approach to each other adiabatically. If the initial separation becomes smaller, e.g. d ≤ 0.25 ∼ 3rmH, the inward migration of Jupiter is much faster and the two planets may achieve the 3 : 2 MMR. However, because of the mutual scattering due to gravity, the migrations of both planets are unstable when the initial separation is too small. (3). In the case of q < 1, after Saturn and Jupiter have been locked into MMRs, they will migrate inward together instead of migrating outward, and the eccentricity of Saturn could be excited much higher than that of the q > 1 configuration. After the stage of convergent migration, the gaps of the two planets overlap. As the disk between the two planets is cleared, Saturn is pushed outward by the inner disk and Jupiter is pushed inward by the outer disk. When the two approaching planets are locked into MMR, the mutual interaction due to gravity prevents them from getting close too quickly, and the separation between them becomes steady. Since Jupiter is much more massive, Saturn is then pushed inward by Jupiter. As they move toward the center, the inner disk is swept out and forms an inner cavity. This cavity plays a great role. First, without the damping effect of gas, the eccentricity of Saturn could be excited to eS ∼ 0.5(in the 2 : 1 MMR), which is much higher than that of q > 1 configuration and is consistent with the result of the analysis of two resonant planets. Second, without the scattering triggered by the massive gas within the coorbital zone of -- 17 -- planet, Saturn is able to maintain the high eccentricity and be pushed to the vicinity of the central star. Our long-time evolution shows that the inward migration of Saturn and Jupiter is stable at the high eccentricity for at least T ≥ 8000PS0. Thus, this cavity in fact ensures the stability of a highly eccentric system, e.g. 'Hot Jupiter/Saturn'. 6. Acknowledgement We thank D.N.C. Lin, A.Crida, W. Kley for their constructive conversations. Zhou is very grateful for the hospitality of Issac Newton institute during the Program 'Dynamics of Discs and Planets'. This work is supported by NSFC (Nos. 10925313,10833001,10778603), National Basic Research Program of China(2007CB814800) and Research Fund for the Doc- toral Program of Higher Education of China (20090091120025). REFERENCES Correia, A. C. M., et al. 2009, A&A, 496, 521 d'Angelo, G. & Lubow, S. H. 2008, ApJ, 685, 560 Fischer, D.A., et al. 2003, ApJ, 586, 1394 Fogg, M.J., & Nelson, R.P. 2007, A&A, 461, 1195 Gozdziewski, K., Maciejewski, A. J., & Migaszewski, C. 2007, ApJ, 657, 546 Guillem Anglada-Escud, Mercedes Lpez-Morales, and John E. Chambers 2010, ApJ, 709, 168-178 Guillot, T., Hueso, R. 2006. Monthly Notices of the Royal Astronomical Society 367, L47- L51. Kley, W., Lee, M. H., Murray, N., & Peale, S. J. 2005, A&A, 437, 727 Kley, W., Peitz, J., & Bryden, G. 2004, A&A, 414, 735 Laskar, J., & Correia, A. C. M. 2009, A&A, 496, L5 Lee, M. H., Butler, R. P., Fischer, D. A., Marcy, G. W., & Vogt, S. S. 2006, ApJ, 641, 1178 Malhotra, R. 1995, AJ, 110, 420 -- 18 -- Marcy, G. W., Butler, R. P., Fischer, D., Vogt, S. S., Lissauer, J. J., & Rivera, E. J. 2001, ApJ, 556, 296 Masset, F. S., Morbidelli, A., Crida, A., Ferreira, J., 2006, ApJ, 642, 478-487. Masset, F. S. & Papaloizou, J. C. B. 2003, ApJ, 588, 494 Masset, F. S. & Snellgrove, M. 2001. MNRAS, 320, L55 Morbidelli, A., Crida, A., Masset, F., Nelson, R. P., 2008, A&A, 478, 929-937 Nelson, R.P., & Papaloizou, J. C. B. 2002, MNRAS, 333, L26 Papaloizou, J. C. B., & Szuszkiewicz, E. 2005, MNRAS, 363, 153, Pollack, J. B., Hubickyj, O., Bodenheimer, P., Lissauer, J. J., Podolak, M. & Greenzweig, Y. 1996, Icarus, 124, 62-85. Raymond, S.N., Mandell, A.M., & Sigurdsson, S. 2006, Science, 313, 1413 Snellgrove, M.D., Papaloizou, J. C. B., & Nelson, R. P. 2001, A&A, 374, 1092 Tinney, C. G., Butler, R. P., Marcy, G. W., Jones, H. R. A., Laughlin, G., Carter, B. D., Bailey, J. A., & O'Toole, S. 2006, ApJ, 647, 594 Vogt, S. S., Butler, R. P., Marcy, G. W., et al. 2005, ApJ, 632, 638 Zhang, H., Yuan,C., Lin, D. N. C. & Yen, D. C. C. 2008, ApJ, 676, 639 Zhang, H., Zhou, J.-L, 2010, ApJ, 714, 532 Zhou, J. -L., Aarseth, S.J., Lin, D.N.C., Nagasawa, M., 2005, ApJL, 631, 85 This preprint was prepared with the AAS LATEX macros v5.2. -- 19 -- Table 1. An incomplete list of giant exoplanets probably locked in low order MMRs. System No. P [day] M sin i [MJ ] a [AU ] e q = Mi/Mo Refs. Gliese 876 (2:1) HD 73526 (2:1) HD 82943 (2:1) HD 128311 (2:1) HD 160691 (2:1) 47 Uma (2:1) HD 45364 (3:2) HD 60532 (3:1) 55 Cnc (3:1) c b b c c b b c d b b c b c b c b c 30.34 60.94 188 377 219 441 448 919 310 643 1083 2190 227 343 201 605 14.6 44.3 0.619 1.935 2.90 2.50 2.01 1.75 2.18 3.21 0.52 1.68 2.6 0.46 0.19 0.66 3.15 7.46 0.82 0.17 0.13 0.21 0.66 1.05 0.75 1.19 1.10 1.76 0.92 1.50 2.11 3.39 0.68 0.90 0.77 1.58 0.11 0.24 0.22 0.02 0.19 0.14 0.36 0.22 0.25 0.17 0.07 0.13 0.05 0.22 0.17 0.10 0.28 0.04 0.01 0.09 ∼ 0.31 ∼ 1.16 ∼ 1.14 ∼ 0.67 ∼ 0.31 ∼ 5.65 ∼ 0.29 ∼ 0.42 ∼ 4.94 1 2 3 4 5 6 7 8 9 Note. -- References. (1) Marcy et al. 2001;(2) Tinney et al. 2006; (3) Lee et al. 2006; (4) Vogt et al. 2005; (5) Gozdziewski et al. 2007; (6)Guillem et al. 2010; (7)Correia et al. 2009; (8) Laskar & Correia 2009; (9)Fischer et al. 2003. -- 20 -- Table 2. A summary of our simulations. Case Configuration Separation d σ Relative migration Resonance 1 2 3 4 5 6 7 8 9 10 11 12 S-J S-J S-J S-J S-J S-J S-J S-J S-J S-J S-J S-J d = 1 d = 1 d = 1 d = 1 d = 0.5 d = 0.5 d = 0.5 d = 0.5 d = 0.25 d = 0.25 d = 0.25 d = 0.25 σ ∼ r0 σ ∼ r−1/2 σ ∼ r−1 σ ∼ r−3/2 σ ∼ r0 σ ∼ r−1/2 σ ∼ r−1 σ ∼ r−3/2 σ ∼ r0 σ ∼ r−1/2 σ ∼ r−1 σ ∼ r−3/2 Divergent Equilibrium Divergent Divergent Convergent Convergent Divergent Divergent Convergent Convergent Convergent Convergent - - - - 2 : 1 2 : 1 - - 3 : 2 2 : 1 2 : 1 2 : 1 -- 21 -- Fig. 1. -- semi-major axis evolutions of Saturn and Jupiter embedded in a gas disk whose surface density slope varies from flat(σ ∝ r0) to very steep(σ ∝ r−3/2). The dash-dot curve in Panel (a) corresponds to the migration of single Saturn, which is faster than that in planet pair case. This indicates that the inward migration of Saturn is suppressed by the existence of Jupiter. Further more the inward migration may be halt(Panel b) or even reversed(Panel c, d) as the surface density becomes steeper. 0.511.522.5a 050010001500200025000.511.522.5T / PS0a 050010001500200025003000T / PS0 Single SaturnSaturnJupiter(a) σ ∝ r0(b) σ ∝ r−1/2(c) σ ∝ r−1(c) σ ∝ r−3/2 -- 22 -- Fig. 2. -- density map of the gas disk at different evolution stages, where the initial density slope α = 0 and the initial separation d = 1. Note that, after about 500PS0 evolution from release, the inner disk is compressed and shepherded by Jupiter while Saturn is digging a gap on it. This ensures Saturn moves with the inner disk. -- 23 -- Fig. 3. -- density map of the gas disk at different evolution stages, where the initial density profile is steeper than Figure 2, α = 1/2 and the initial separation is still d = 1. This steeper density slope prevents the inward migration of Jupiter, as well as that of Saturn. -- 24 -- Fig. 4. -- cross sections of the gas disk at different stages and the associated torque evolutions. The initial separations are d = 1. Panel (a): density cross sections at T = 0, 200PS0, 500PS0 and 1000PS0, where α = 0. The light doted line shows the distribution of σ ∝ r−1 as a reference. Panel (b): the evolution of torques exerted on Saturn, when α = 0. Panel (c): density cross sections at T = 0, 200PS0, 500PS0 and 1000PS0, where α = 1/2. The light doted line shows the distribution of σ ∝ r−1. Panel (d): the associated evolution of torques exerted on Saturn when α = 1/2. Note that the surface density increases at the inner disk and its local surface density becomes steeper. The outer Lindblad torque is initially much larger than the inner one, then it decreases very quickly while the inner one increases a bit after the release. And the corotation torque changes from negative to positive as some gas is forced to flow across Saturn's orbit from outside to inside. Thus the net torque is greatly negative at beginning and soon increases to near or even equal to zero. -- 25 -- Fig. 5. -- orbital evolutions of Saturn and Jupiter embedded in a gas disk whose surface density slope α is 0(Panel a), 1/2(Panel b) and 3/2(Panel c). The initial separation between the two planets is reduced to d = 0.5. Convergent migration happens when the disk is nearly flat α ≤ 1/2. -- 26 -- Fig. 6. -- Panel (a): the surface density evolution when α = 3/2. Panel (b): the evolution of torques exerted on Jupiter. Note that the jump around T = 250PS0 indicates the recovery of the inner Lindblad torque as Saturn migrates further away. Although the outer Lindblad torque is initially larger, the inner Lindblad torque decreases slower than the outer one does and makes the net torque positive soon after the release. Panel (c): the evolution of torques exerted on Saturn. Note that the corotation torque decreases to zero as Saturn digs a gap on the inner disk and the outer Lindblad torque decreases much faster than the inner one. -- 27 -- Fig. 7. -- semi-analytic results of the inner and outer Lindblad torques exerted on Saturn, for each integer m = 2 ∼ 80. The density slope α = 3/2 and the initial separation d = 0.5. The maximum of inner torque increases and moves slightly toward high value of m, this indicates the compression(surface density increases) and outward expands of the inner disk(the inner edge of gap which opened by Saturn moves toward Saturn). As the outer Lindblad torques decreases faster, the net torque becomes positive at around T = 500PS0. 01020304050607080012x 10−4Normallized Torque inner Lindblad (analytic)Outer Lindblad (analytic)inner LindbladOuter Lindblad010203040506070800123456x 10−4Normallized Torque010203040506070800123x 10−4mNormallized TorqueT=0 PS0T=200 PS0T=500 PS0 -- 28 -- Fig. 8. -- semi-analytic results of the inner and outer Lindblad torques exerted on Saturn, for each integer m = 2 ∼ 80. The density slope α = 0 and the initial separation d = 0.5. The outer Lindblad torques decrease while the inner ones maintain. Both the maximum of inner and outer torques move toward small m, which indicates a gap around Saturn is forming(surface density decreases around Saturn). 01020304050607080012x 10−4Normallized Torque 010203040506070800123456x 10−4Normallized Torque01020304050607080012x 10−4mNormallized TorqueInner Lindblad (analytic)Outer Lindblad (analytic)Outer LindbladInner LindbladT=0 PS0T=200 PS0T=500 PS0 -- 29 -- Fig. 9. -- Panel (a): surface density evolution when α = 0 and d = 0.5. The inner disk is swept out by the inward migrating planet pair and a cavity forms at T ≥ 1000PS0. Panel (b): the evolution of torques exerted on Jupiter. Panel (c): the evolution of torques exerted on Saturn. Note that the outer Lindblad torque decreases much faster than the inner one after the release moment. -- 30 -- Fig. 10. -- density map at different stages of evolution. The right Panels zoom in the inner disk in which Saturn is embedded. -- 31 -- Fig. 11. -- orbital evolutions of Saturn and Jupiter embedded in a gas disk whose surface density slope α is 0(Panel a), 1/2(Panel b) and 3/2(Panel c). The initial separation between the two planets is reduced to d = 0.25. The convergent migration becomes faster at small initial separation and the 3 : 2 MMR is reached when α = 0(Panel a). -- 32 -- Fig. 12. -- long time evolution of α = 0 and d = 0.5 case. This system is stable for at least 8000PS0 at high eccentricity of Saturn (eS ≤ 0.6) when the 2 : 1 MMR is preserving. Saturn get close to the central star at aS < 1AU . 00.511.5a 0 0.20.40.6e0 100020003000400050006000700080000123456T / PS02λ2−λ1−ϖ2SaturnJupiter σ ∝ r0
1004.0971	1	1004	2010-04-06T20:55:25	The Compositional Diversity of Extrasolar Terrestrial Planets: I. In-Situ Simulations	[ "astro-ph.EP" ]	Extrasolar planet host stars have been found to be enriched in key planet-building elements. These enrichments have the potential to drastically alter the composition of material available for terrestrial planet formation. Here we report on the combination of dynamical models of late-stage terrestrial planet formation within known extrasolar planetary systems with chemical equilibrium models of the composition of solid material within the disk. This allows us to determine the bulk elemental composition of simulated extrasolar terrestrial planets. A wide variety of resulting planetary compositions are found, ranging from those that are essentially "Earth-like", containing metallic Fe and Mg-silicates, to those that are dominated by graphite and SiC. This shows that a diverse range of terrestrial planets may exist within extrasolar planetary systems.	astro-ph.EP	astro-ph	The Compositional Diversity of Extrasolar Terrestrial Planets: I. In−Situ Simulations Jade C. Bond1,3 Lunar and Planetary Laboratory, University of Arizona, Tucson, AZ 85721 [email protected] and David P. O'Brien2 Planetary Science Institute, Tucson, AZ 85719 and Dante S. Lauretta1 Lunar and Planetary Laboratory, University of Arizona, Tucson, AZ 85721 Received ; accepted 0 1 0 2 r p A 6 . ] P E h p - o r t s a [ 1 v 1 7 9 0 . 4 0 0 1 : v i X r a 1Lunar and Planetary Laboratory, University of Arizona, 1629 East University Boulevard, Tucson, AZ 85721-0092. 2Planetary Science Institute, 1700 E. Fort Lowell, Tucson, AZ 85719. 3Now at Planetary Science Institute. – 2 – ABSTRACT Extrasolar planet host stars have been found to be enriched in key planet- building elements. These enrichments have the potential to drastically alter the composition of material available for terrestrial planet formation. Here we report on the combination of dynamical models of late-stage terrestrial planet formation within known extrasolar planetary systems with chemical equilibrium models of the composition of solid material within the disk. This allows us to determine the bulk elemental composition of simulated extrasolar terrestrial planets. A wide variety of resulting planetary compositions are found, ranging from those that are essentially "Earth-like", containing metallic Fe and Mg-silicates, to those that are dominated by graphite and SiC. This shows that a diverse range of terrestrial planets may exist within extrasolar planetary systems. Subject headings: planets and satellites: composition - planets and satellites: formation - planetary systems – 3 – 1. Introduction Extrasolar terrestrial planets are a tantalizing prospect. Given that the number of planets in the galaxy is expected to correlate inversely with planetary mass, it is expected that Earth-sized terrestrial planets are much more common than giant planets (Marcy et al. 2000). Although still undetectable by current exoplanet searches, the possibility of their existence in extrasolar planetary systems has been examined by several authors. Many such studies have focussed on the long term dynamical stability of regions within the planetary system where such planets could exist for geologic timescales (Barnes & Raymond 2004; Raymond & Barnes 2005; Asghari et al. 2004). Several systems have been found to posses such regions (e.g. Barnes & Raymond 2004), indicating that if they are able to form, terrestrial planets may still be present within extrasolar planetary systems. Analyses of this nature are of great interest to future planet search missions as they assist in constraining future planet search targets. However, they provide little insight into the formation mechanism and physical and chemical properties of such planets and do not necessarily indicate the presence of a terrestrial planetary companion. A few other studies have gone one step further and undertaken detailed N-body simulations of terrestrial planet formation. Raymond et al. (2005) considered terrestrial planet formation in a series of hypothetical 'hot Jupiter' simulations and found that terrestrial planets can indeed form in such systems (beyond the orbit of the giant planet) provided the 'hot Jupiter' is located within 0.5AU from the host star. Furthermore, such planets may even have water contents comparable to that of the Earth. Terrestrial planets have been found to form even in simulations of systems which have undergone large-scale migration of the giant planet (Raymond et al. 2006b; Mandell et al. 2007). Terrestrial planets were found to form both exterior and interior to the giant planet after migration has occurred and many were located within the habitable zone of the host – 4 – star. As many extrasolar planets are believed to have experienced such a migration, it is encouraging that terrestrial planets may still be able to form within these systems. To date, only Raymond et al. (2006a) have undertaken terrestrial planet formation simulations for specific planetary systems. They considered four known planetary systems and found that terrestrial planets could form in one of the systems (55Cancri). Small bodies comparable to asteroid sized objects would be stable in another (HD38529). An even more intriguing question beyond whether or not terrestrial planets could exist within these systems is their potential chemical composition. Extrasolar planetary host stars are already known to be chemically unusual (Gonzalez 1997, 1998; Butler et al. 2000; Gonzalez & Laws 2000; Gonzalez et al. 1999; Gonzalez & Vanture 1998; Santos et al. 2000, 2001, 2004; Gonzalez et al. 2001; Smith et al. 2001; Reid 2002; Fischer & Valenti 2005; Bond et al. 2006, 2008), displaying systematic enrichments in Fe and smaller, less statistically significant enrichments in other species such as C, Si, Mg and Al (Gonzalez & Vanture 1998; Gonzalez et al. 2001; Santos et al. 2000; Bodaghee et al. 2003; Fischer & Valenti 2005; Beirao et al. 2005; Bond et al. 2006). Given that these enrichments are likely primordial in origin (Santos et al. 2001, 2003a,b, 2005; Fischer & Valenti 2005; Bond et al. 2006), it is thus likely that the planet forming material within these systems will be similarly enriched. Hints of such a correlation between transiting giant planets and stellar metallicity have been observed (Guillot et al. 2006; Burrows et al. 2007). Consequently, it is likely that terrestrial extrasolar planets may have compositions reflecting the enrichments observed in the host stars. Furthermore, several known host stars have been found to have C/O values above 0.8 (Bond et al. 2008). Systems with high C/O ratios will contain large amounts of C phases (such as SiC, TiC and graphite), resulting in any terrestrial planets within these systems being enriched in C and potentially having compositions and mineralogies unlike any body observed within our Solar System. – 5 – Despite the likely chemical peculiarities of extrasolar planetary systems and the early successes of extrasolar terrestrial planet formation simulations, no studies of extrasolar terrestrial planet formation completed to date have considered both the dynamics of formation and the detailed chemical compositions of the final terrestrial planets produced. This study addresses this issue by simulating late-stage in-situ terrestrial planet formation within ten extrasolar planetary systems while simultaneously determining the bulk elemental compositions of the planets produced. This is the first such study to consider both the dynamical and chemical nature of potential extrasolar terrestrial planets and it represents a significant step towards understanding the diversity of potential extrasolar terrestrial planets. 2. System Composition The two most important elemental ratios for determining the mineralogy of extrasolar terrestrial planets are C/O and Mg/Si. Note that throughout this paper, Mg/Si and C/O refers to the elemental number ratios, not solar normalized logarithmic values often quoted in stellar spectroscopy (usually shown as [X/H] for the solar normalized logarithmic abundance of element X compared to H). The ratio of C/O controls the distribution of Si among carbide and oxide species. Under the assumption of equilibrium, if the C/O ratio is greater than 0.8 (for a pressure of 10−4 bar), Si exists in solid form primarily as SiC. Additionally, a significant amount of solid C is also present as a planet building material. For C/O values below 0.8, Si is present in rock-forming minerals as SiO4 4− (or SiO2), allowing for the formation of silicates. The silicate mineralogy is controlled by the Mg/Si value. For Mg/Si values less than 1, Mg is in pyroxene (MgSiO3) and the excess Si is present as other silicate species such as feldspars. For Mg/Si values ranging from 1 to 2, Mg is distributed between olivine (Mg2SiO4) and pyroxene. For Mg/Si values extending beyond – 6 – 2, all available Si is consumed to form olivine with excess Mg available to bond with other elements as MgO or MgS. Just as stellar C/O values are known to vary within the solar neighborhood (Gustafsson et al. 1999), the C/O values of extrasolar planetary systems also deviate from the solar value. The photospheric C/O vs. Mg/Si values for known extrasolar planetary host stars are shown in Figure 4, based on stellar abundances taken from Gilli et al. (2006) (Si and Mg), Beirao et al. (2005) (Mg), Ecuvillon et al. (2004a) (C) and Ecuvillon et al. (2006) (O). A conservative approach was taken in determining the average error shown in Figure 4. The errors published for each elemental abundance were taken as being the 2σ errors (as the method used to determine them naturally provides the 2σ error range) and were used to determine the maximum and minimum abundance values possible with 2σ confidence for each system. The elemental ratios produced by these extremum abundances were thus taken as the 2σ range in ratio values and are shown as errors in Figure 4. The mean values of Mg/Si and C/O for all extrasolar planetary systems for which reliable abundances are available are 1.32 and 0.77 respectively, which are above solar values (Mg/SiJ = 1.00 and C/OJ = 0.54) (Asplund et al. 2005). This non-solar average and observed variation implies that a wide variety of materials would be available to build terrestrial planets in those systems, and not all planets that form can be expected to be similar to that of Earth. Of the 60 systems shown, 21 have C/O values above 0.8, implying that carbide minerals are important planet building materials in potentially more than 30% of known planetary systems. This implies that a similar fraction of protoplanetary disks should contain high abundances of carbonaceous grains. As comets represent some of the most primitive material within our planetary system, it is likely that a similar mass fraction will apply to the protoplanetary nebula. Furthermore, infrared spectral features at 3.43 and 3.53 µm observed in 4% of protoplanetary disks have been identified as being produced by – 7 – nano-diamonds (Acke & van den Ancker 2006). Such high abundances of carbon-rich grains in nascent planetary systems is inconceivable if they have primary mineralogy similar to our Solar System, thus implying that C-rich planetary systems may be more common than previously thought. The idea of C-rich planets is not new (Kuchner & Seager 2005) but the potential prevalence of these bodies has not been previously recognized, nor have specific systems been identified as likely C-rich planetary hosts. These data clearly demonstrate that there are a significant number of systems in which terrestrial planets could have compositions vastly different to any body observed in our Solar System. Both host and non-host stars1 display the same distributions in C/O and Mg/Si values (see Figure 5). The mean, median and standard deviation for both the host and non-host stars is shown in Table 2. The values listed in Table 2 are based on the stellar abundances determined in Bond et al. (2008) as Gilli et al. (2006); Beirao et al. (2005); Ecuvillon et al. (2006, 2004a) provide abundances for all four elements for just three non-host stars, thus preventing host and non-host comparisons. It is essential to point out here that the values shown in Figure 5 and Table 2 are based on a different dataset than is used for the simulations presented in this paper and are meant for comparative purposes only. The [Mg/H] values in Bond et al. (2008) are known to be lower than for previously published common stars and the different spectral indicators were used to obtain O abundances (see Bond et al. (2008) for more details). However, as both host and non-host stars were examined in the same way, the use of this data to compare the two populations is still valid. Given the excellent agreement between host and non-host stars, we conclude that known planetary host stars are not preferentially biased towards higher C/O or Mg/Si values compared to stars not known to harbor a planetary companion. This in turn implies 1Throughout this paper, non-host stars refers to stars observed as part of a planet search program that are not currently known to harbor a planetary companion. – 8 – that the prevalence of C-rich planetary systems identified above is not statistically unusual (in terms of stellar composition). However, a high degree of uncertainty is associated with all stellar C/O and Mg/Si values. The primary source of this error is uncertainty in the stellar elemental abundances themselves. Spectrally determined abundances are sensitive to continuum placement and stellar atmospheric parameters such as stellar effective temperature (Teff), stellar surface gravity (log g), metallicity and microturbulence. These uncertainties result in an average 2-σ error of ±0.04 for [Mg/H], ±0.07 for [Si/H], ±0.08 for [C/H] and ±0.09 for [O/H] 2. These errors result in considerable percentage uncertainties for the Mg/Si and C/O values of up to 124%. Although large, the errors will not decrease until we are able to reduce the uncertainty on each individual stellar elemental abundance. Of the four key elements discussed thus far, O is by far the most controversial. Three different spectral lines are available for determining the photospheric O abundance - the forbidden OI lines located at 6300 and 6363 A, the OI triplet located between 7771 A and 7775 A and the OH lines located near 3100 A (Ecuvillon et al. 2006). Previous studies have found discrepancies between abundances obtained from different spectral lines for the same star of up to 1 dex (Israelian et al. 2004) (A dex is the logarithmic unit for elemental abundances). Each of these lines is subject to interferences from different stellar sources. The forbidden OI lines are weak and blended with Ni, the OI triplet is influenced by non-local thermodynamic equilibrium (non-LTE) effects and the OH lines are influenced by stellar surface features (Ecuvillon et al. 2006). Ecuvillon et al. (2006) undertook a detailed examination of the correlation between these different O indicators for a sample of 96 host and 59 non-host stars. Their study showed that the discrepancies 2[X/H] = log(X/H)⋆ - log(X/H)⊙ where X/H is the ratio of the abundance of a given element, X, to H. – 9 – in abundances determined by the three different indicators was less than 0.2dex for the majority of stars examined. The forbidden OI and OH lines were found to be in good agreement with each other while abundances obtained from the O triplet lines (with the appropriate NLTE corrections applied to them) were systematically lower. However, all three indicators produced abundances in keeping with the galactic evolutionary trends observed for lower metallicity (and thus younger) stars (Ecuvillon et al. 2006). The C/O ratios shown in Figure 4 (and used throughout this study) are based on the O abundances from Ecuvillon et al. (2006) obtained from the forbidden OI spectral line observed at 6300.3 A as abundances from this line are in agreement with the abundances obtained from the OH line and produce a marginally better fit to stellar evolutionary models. The solar O abundance itself has experienced a 'crisis' in recent years (Ayres et al. 2006) with several studies suggesting that a downward revision of the solar O value is required (e.g. Ayres et al. 2006; Socas-Navarro & Norton 2007). However, realistic errors for the suggested new O abundance are approximately 0.1 dex (Socas-Navarro & Norton 2007) and the abundances of the present study vary from [O/H] = 0.38 (HD 177830) to [O/H] = −0.10 (HD 108874), a range of 0.48 dex. As our present study range is almost five times larger than the errors of the revised O abundance, we feel confident in the compositional variations identified here as being caused by variations in the stellar O abundances of specific systems. Given the errors associated with each individual elemental abundance (and thus also ratio value), it is natural to consider the uncertainty associated with the dispersion seen in Figure 4. The observed dispersion is produced by both dispersion in the actual values and dispersion due to measurement errors. Propagating the errors of the individual abundances leads to a standard deviation of 0.22, slightly reduced from 0.27 without considering errors. Thus it is probable that the real range in elemental ratios is less than is shown in Figure – 10 – 4 and fewer planetary systems have C/O values above 0.8. Based on the errors described above and shown in Figure 4, only 5 of the 61 planetary systems shown (8% of the sample) can be said with 2σ confidence to have C/O values above 0.8. This increases to just 7 planetary systems (11% of the sample) when we reduce the confidence interval to 1σ. It should also be noted here that the O abundances utilized here are on average higher than those produced by other indicators. As such, the C/O values are lower and should be considered conservative values. Additionally, as the errors are bidirectional, it is also possible that our high C/O sample may be larger. Making the overly-optimistic assumption that all systems lie at the upper bounds of their error range, 47 of the 61 planetary systems (77% of the sample) may have C/O values above 0.8. Taken together, this suggests that while the number of systems with C/O values larger than 0.8 may be less than indicated in Figure 4, it is likely not as low as the most conservative estimates, and a significant number of planetary systems may have C-rich condensation sequences. Planets that form in these systems would contain significant amounts of carbide phases as major planet building materials, and would differ significantly from the silicate-dominated planets that form in systems with lower C/O values. 3. Simulations 3.1. Extrasolar Planetary Systems Ten known extrasolar planetary systems spanning the entire compositional spectrum of observed planetary systems were selected for this study. This wide range was purposefully chosen to explore the full diversity of possible extrasolar terrestrial planets. The dynamical and chemical characteristics of each system are shown in Tables 3, (orbital parameters of known planetary companions), 4 (stellar abundances in logarithmic units) and 5 (stellar – 11 – abundances normalized to 106 Si atoms), while the known giant planet architecture is shown in Figure 6 and the C/O and Mg/Si values were previously shown in Figure 4. Dynamical properties of the known giant planets were taken from the catalog of Butler et al. (2006) [these values were augmented with more recent values from http://exoplanets.org when available]. Note that the innermost planet of 55Cnc was neglected in our present simulations due to its location and low mass. 3.2. Chemical Simulations As in Bond et al. (2009), the chemical composition of material within the disk is assumed to be determined by equilibrium condensation within the primordial stellar nebula. Although the timescales for formation of planetesimals and embryos considered here can be comparable to the timescale for reaction kinetics (eg. less than 1×106yr for the Jovian circumplanetary disk (Mousis & Alibert 2006)), the hot inner mid-plane region (¡10AU from the host star) achieves thermodynamic equilibrium within approximately 100 to 10,000 years (Semenov et al. 2010). Therefore the assumption that the simulated planetesimals form in equilibrium with their surroundings is still valid for this type of study. This approach is further supported by evidence observed within our own Solar System. Primitive chondritic material has been found to be remarkably similar to that predicted by equilibrium condensation studies with a Solar composition (Ebel 2006), with bulk elemental abundances that are also functions of their equilibrium condensation temperature for relevant disk conditions (Davis 2006). Preserved equilibrium compositions can still be seen today in the observed thermal stratification of the asteroid belt (Gradie & Tedesco 1982). Although some uncertainty as to the precise cause of these features still lingers, the fact remains that we can utilize equilibrium condensation temperatures to determine the bulk elemental abundances of solid material within planetary systems. – 12 – Equilibrium condensation sequences for an identical list of elements as used in Bond et al. (2009) (H, He, C, N, O, Na, Mg, Al, Si, P, S, Ca, Ti, Cr, Fe and Ni) were obtained using the commercial software package HSC Chemistry (v. 5.1) using the same list of gaseous and solid species as in Bond et al. (2009) (listed here in Table 6). As these elements represent the major solid-forming species, no significant limitations are encountered by neglecting other species from these simulations. Observed stellar photospheric abundances were adopted as a proxy for the composition of the stellar nebula and were taken from Gilli et al. (2006); Beirao et al. (2005); Ecuvillon et al. (2004b, 2006). It should be noted that all abundances applied here were taken from the same research group and were obtained from the same spectra thus acting to limit any possible systematic differences in abundances due to instrument or methodological differences between various studies. The input values used in HSC Chemistry for each system are shown in Table 7. All species are initially assumed to be in their elemental and gaseous form and no other species or elements are assumed to be present within the system. Neither N or P abundances have been obtained for extrasolar planetary host stars, primarily due to the difficulty in finding unblended spectral lines to use within the visual spectral range (where most studies have been focussed). For this present study, we overcame this issue by obtaining approximate abundances for both elements based on the well known odd-even effect. Caused by the increased stability of even atomic number nuclei relative to odd-numbered nuclei, this effect produces the observed sawtooth pattern in the Solar elemental abundances. Extrasolar abundances were obtained by fitting a linear trend through the solar abundances for the same elements studied here and then applying this same fit to observed extrasolar host star abundances of Na and Al. This approach assumes that extrasolar host stars will display the same atomic sawtooth pattern, a valid assumption as host stars do not appear to have undergone any form of systematic processing (such as – 13 – pollution by a nearby supernova event) to cause a significant deviation (see Bond et al. 2008). The same approach was adopted for those systems without an observed [S/H] value. As for the Solar System simulations of Bond et al. (2009), "nominal" radial pressure and temperature profiles obtained from the Hersant et al. (2001) protoplanetary disk midplane model were used to correlate chemical compositions with a spatial location within the disk. The stellar mass accretion rate M , a primary input to the Hersant model, is known to vary with stellar mass as: M ∝ M3/2 (1) where M is the mass of the star in solar masses (ranging from 0.98 to 1.48MJ) and M is the mass accretion rate in solar masses per year. The input mass accretion rates were thus scaled for the stellar mass of the host star obtained from the Simbad database3. The resulting stellar accretion rates obtained are shown in Table 8. All other input parameters for the Hersant et al. (2001) models remained unchanged (α = 0.009, initial disk radius = 17AU). Minor differences in temperature and pressure profiles were produced for the various systems simulated. For example, the temperature and pressure values at a distance of 1AU from the host star for the systems with the highest and lowest mass accretion rates differed by just 51K and 1.6×10−6 bars respectively. It is quite possible that the mass accretion rate varies throughout the disk itself, potentially producing variations in midplane pressure and temperature values. The Hersant et al. (2001) models assumed a uniform mass accretion rate. However, the 3accessed at http://simbad.u-strasbg.fr/simbad/ – 14 – conditions at 1AU in the system with the highest mass accretion rate (HD177830) occur just 0.09AU closer to the host star in the system with the lowest mass accretion rate (HD72659). This indicates that solid material will still condense out at essentially the same radial location, despite variations in the mass accretion rate of the disk. As such, any modification of midplane conditions from those used here due to a varying mass accretion rate are likely to be small enough as to be neglected for the purposes of this study. It is also important to note that the current approach does not include variations in the midplane conditions produced by different chemical compositions (which would alter parameters such as disk viscosity and opacity), nor does it include the effects of stellar luminosity (which aren't incorporated into the Hersant et al. (2001) model). As such, the scaling applied here is a somewhat simplistic approach to a complex issue but is valid for the current aims of this study. Midplane pressure and temperature values were determined with an average radial separation of 0.01AU throughout the study region. As in Bond et al. (2009), an ensemble of disk compositions was determined based on Hersant et al. (2001) disk conditions at seven different evolutionary times (t = 2.5×105yr, 5×105yr, 1×106yr, 1.5×106yr, 2×106yr, 2.5×106yr, 3×106yr). However, as we previously found the best fit to known planetary values in the Solar System (specifically the compositions of Venus, Earth and Mars) to occur using disk conditions obtained for t = 5×105yr (Bond et al. 2009), our discussions will mostly focus on compositions produced by disk conditions at this time. Note that these times only refer to the timing of the thermodynamical conditions under which the planetesimals condensed within the disk. – 15 – 3.3. Dynamical Simulations In this study, we build upon our recent success in simulating terrestrial planet formation within the Solar System (Bond et al. 2009) and apply the same methodology here. Terrestrial planet formation in extrasolar planetary systems is a complex problem, especially when giant planet migration is taken into account, and modeling it in detail for even a single system is computationally expensive (Raymond et al. 2006b; Mandell et al. 2007). In the interest of exploring a wide range of systems, we use a more basic approach here which necessarily neglects some of the complexities of planet formation, but still allow us to demonstrate some of the potential implications of varying system compositions on final terrestrial planet compositions. More detailed dynamical simulations will be the subject of future work. N-body simulations of terrestrial planet accretion in each of the selected extrasolar planetary systems are run using the SyMBA n-body integrator (Duncan et al. 1998). The orbital parameters of the giant planets in each system are taken from the catalog of Butler et al. (2006), and updated with values from exoplanets.org as additional data on these systems were obtained. All values utilized are shown in Table 3. Inclinations of each of the giant planets are assumed to be zero since no such measurements have been obtained for these systems. Due to the computational time required, current simulations only contain an initial population of roughly Lunar- to Mars-mass embryos (i.e. no planetesimal swarm is included). This differs from the Solar System simulations of O'Brien et al. (2006) as used in Bond et al. (2009) as those simulations included a planetesimal swarm. As such, it is possible that differences between the dynamical simulations performed here and those of the Solar System as used in Bond et al. (2009) will occur, most likely in that the dynamical excitation of the resulting systems system will be somewhat larger than they would be if a planetesimal swarm were present. – 16 – For each extrasolar planetary system modeled, planetary embryos are distributed in the zone between the star and the giant planet (or in the case where there are one or more giant planets close to the host star such as 55Cnc and Gl777, in the region between the inner and outer giant planets) according to the relations between embryo mass, spacing, and orbital radius given by Kokubo & Ida (2000). No embryos are initially located interior to 0.3 AU. The timestep for the integration was set to at least 20× the orbital period of the innermost planet or planetary embryo, or the orbital period of a body at 0.1 AU, whichever is smaller (this corresponds to a half-day timestep for an inner radius of 0.1 AU), and the simulations are run for 100-250 Myr. Surface mass density profiles that vary as r−3/2, normalized to 10 gcm−2 at 1 AU, were assumed. For each system, a minimum of 4 accretion simulations were run, using different random number seeds for generating the initial embryo distribution. Migration of the giant planets is very likely to have occurred in many or all of these systems. However, if migration occurred very early, prior to planetesimal and embryo formation in the terrestrial planet region, then terrestrial planets could have potentially formed after migration, with the giant planets in their current configurations (e.g. Armitage 2003). Our simulations focus on this scenario (termed here "in-situ formation"). If giant planet migration occurred after planetesimals and embryos have formed, then our in-situ assumption does not apply and there is likely to be radial migration of planetary embryos, driven by giant planet migration (eg. Raymond et al. 2006a,b; Mandell et al. 2007). However, as there is currently no clear consensus as to the most common timing of planetary migration, and no evidence for the specific systems that we propose to study, each of our simulations begin with the gas giants already fully formed and located in their current positions. Simulations incorporating giant planet migration will be the focus of the second paper in this series. – 17 – 3.4. Combining Dynamics and Chemistry The dynamical and chemical simulations were combined together as outlined in Bond et al. (2009) whereby we assigned each embryo a composition based on its formation location (hence pressure and temperature) and assumed that it then contributed that same composition to the final terrestrial planet. The bulk compositions of the final planets are simply the sum of each object they accrete. Phase changes and outgassing were neglected and all collisions were assumed to result in perfect accretion (i.e. no mass loss occurred). 3.5. Stellar Pollution The issue of stellar pollution produced by terrestrial planet formation is of great interest in extrasolar planetary systems. Pollution of the stellar photosphere via accretion of a large amount of solid mass during planet formation and migration has been suggested as a possible explanation for the observed metallicity trend for known host stars (Laughlin 2000; Gonzalez et al. 2001; Murray et al. 2001). Thus we determined the amount of material accreted by the host star during the current terrestrial planet simulations and also determined the resulting change in spectroscopic photospheric abundance. As in Bond et al. (2009), any solid material migrating to within 0.1AU from the host star is assumed to have accreted onto the stellar photosphere. This material is then assumed to have been uniformly mixed throughout the stellar photosphere and convective zone. Decreasing convective zone mass with time, granulation within the photosphere and gravitational settling and turbulence within the convective zone are again neglected, resulting in the values determined here being the maximum expected enrichments. The mass of each element accreted onto the star was determined in the same way as for terrestrial planets, by summing the contributions of the individual embryos it accretes. – 18 – As a reminder, the resulting photospheric elemental abundance is given by: fX,J(cid:21) [X/H] = log(cid:20) fX (2) where [X/H] is the resulting abundance of element X after accretion of terrestrial planet material, fX is the mass abundance of element X in the stellar photosphere after accretion and fX,J is the mass abundance of element X in the Solar photosphere (from Murray et al. (2001)). Note that [X/H] is still dependant on fX,J as by definition it is taken relative to the Solar abundance. Explicitly, fX is given by: fX = (cid:20) mX mtotal(cid:21) (3) where mX is the mass of element X accreted during the simulation, mtotal is the total mass of the convective zone. Since the stellar composition is dominated by H, we can make the assumption that mtotal ∼ mH. Thus: fX = (cid:20) mX mtotal(cid:21) ∼ (cid:20) mX mH(cid:21) (4) The same holds true for the solar values. fX values for the extrasolar planetary host stars were obtained via the stellar abundances of Gilli et al. (2006); Beirao et al. (2005); Ecuvillon et al. (2004b, 2006), as these papers represent a comprehensive, internally consistent catalogue of photospheric abundances – 19 – for a large number of known planetary host stars. The mass of the convective zone of a star is known to vary with its mass, effective temperature (Tef f ) and, to some extent, its metallicity. Values for the masses of the convective zone for each of the target stars were thus obtained from Pinsonneault et al. (2001) using the Tef f values from Santos et al. (2004). The convective zone masses are shown in Table 9. fX,J values were obtained by utilizing the solar abundances of Asplund et al. (2005) and a current solar convective zone mass of 0.03MJ (Murray et al. 2001). 4. Results 4.1. Dynamical As the current simulations are only preliminary and are designed solely to illustrate potential compositional differences within the final terrestrial planets, a detailed examination of the simulation results is of minimal benefit. However, terrestrial planets were found to form in all simulations. 22 of the 40 simulations produced two or more terrestrial planets within one system. The general architecture of the resulting systems is shown in Figures 7 - 11. In several cases, the orbital elements of the giant planets change slightly from the values shown in Fig. 6 due to the ejection of embryos from the system, and mutual perturbations in systems with multiple giant planets. Several of these planets (e.g. namely those in the simulation for HD4203) are simply embryos that have survived for the duration of the simulation and have not accreted any additional material, but essentially all others are grown from collisions among multiple embryos. Of the ten planetary systems examined, only one (HD72659) is found to produce terrestrial planets with a median mass comparable to Earth (1.03 ML) with six of the 11 – 20 – planets produced having masses equal to or greater than 1 ML. All other systems have median planetary masses less than 1 ML, and, excluding those of HD72659, only four out of 51 terrestrial planets in our simulations have masses equal to or greater than 1 ML. With the exception of 55Cancri, terrestrial planets that form in our simulations are located inwards of 1 AU. This is primarily a selection effect as we are currently only simulating terrestrial planet formation interior to the known giant planets, and the majority of systems we study here contain giant planets orbiting within 2AU of their host star. 55Cancri is unusual in that its outermost giant planet has a periapse larger than 5AU, a significant increase over the other systems selected for study. That in turn dictated that the embryos in the 55Cancri simulations initially be located between 1 AU and 5AU, hence the larger semi-major axes of the planets that form there. The terrestrial planets in these systems generally accrete the vast majority of their mass from their immediately surrounding area without a large amount of radial mixing occurring. As such, the terrestrial planets in the simulations are expected to have compositions reflecting any radial trends within the disk. This amount of radial mixing, however, is likely to increase somewhat in future simulations that include a planetesimal swarm, and may greatly increase in simulations that include the effects of migrating giant planets. The above results are not a definitive determination of the likely terrestrial planet population in the systems we are studying, as we are currently only considering late stage, in-situ terrestrial planet formation. That is, we are only considering formation that has occurred after the known giant planets have formed and migrated to their current orbits. As previously discussed, this is a valid approach as there is no consensus on the timing or extent of migration within these systems and it provides us with a reasonable starting point to consider the chemical compositions of possible terrestrial planets in those systems. However, giant planet migration can have a strong influence on terrestrial planet – 21 – formation, in a worst-case scenario preventing terrestrial planets from forming, but it may also accelerate terrestrial planet formation and lead to significant radial mixing of material within the system, especially in systems with close-in giant planets (Raymond et al. 2006b; Mandell et al. 2007). Simulations addressing the formation and chemical/dynamical evolution of terrestrial planets under the influence of giant planet migration are currently running and will be the focus of a future paper. 4.2. Chemical The condensation sequences and abundances of solid species (normalized to the abundance of the least abundant species) for three representative systems are shown in Figure 12. Plots for all other systems in order of increasing C/O value are available online only (Figures 1 - 3). The 50% condensation temperatures (i.e. temperature at which half of the total amount of a given element has condensed) for each of the systems studied is shown in Table 10. The final elemental abundances for all simulated planets and for times studied are shown in Table 11 (available in full online). Two very distinct types of condensation sequence are produced for the systems studied here - those resembling the Solar condensation sequence (HD27442, HD72659 and HD213240) and those in which carbide phases are present within the disk (55Cnc, Gl777, HD4203, HD17051, HD19994, HD108874 and HD177830). The C enrichment can further be classified as being low (0.78<C/O<1.0), in which C and carbide phases are present within a spatially narrow region of the disk at temperatures below ∼1800 K (55Cnc, Gl777, HD17051 and HD177830), and high (C/O>1.0) where a broader carbide-dominated region is stable for temperatures below ∼2300 K and thus extends into the innermost reaches of the disk (HD4203, HD19994 and HD108874). The implications of these variations in the distribution of solid material are discussed in Section 5.1. The compositional differences – 22 – between these different classes of systems result in significantly different compositions of the terrestrial planets that form in those systems, and the characteristics of those planets will be discussed in the following sections. Unless otherwise stated, all compositions shown are produced by disk conditions at t = 5×105 years. As previously stated, compositions produced for disk conditions at this time were found by Bond et al. (2009) to produce the best fit to observed Solar System compositions (based on agreeing with the compositions of Venus, Earth and Mars). Compositional changes resulting from disk conditions at different times will be discussed in Section 5.2. 4.2.1. Earth-like Planets Before we can discuss the results produced by this work, we first need to define in detail what an "Earth-like" planet is. For the purposes of this study, it is taken to be at the most basic level a terrestrial planet whose composition is dominated by O, Fe, Mg and Si, with small amounts of Ca and Al and very little Carbon. Essentially, this refers to a terrestrial planet composed of Mg silicates and metallic Fe with other species present in relatively minor amounts. Water or other hydrous species may or may not be present. At a more detailed level, we have allowed for deviations of ±25% from the elemental abundances for the Earth listed in Kargel & Lewis (1993) for the major elements (O, Fe, Mg, Si). Larger abundances were permitted in the minor elements (with the exception of C) due to their lower relative abundance within the terrestrial planets. Given the variations permitted on the individual abundances, a larger range of variations of 35% was permitted in the geochemical ratios considered here (namely Mg/Si, Al/Si and Ca/Si). It is important to note that these accepted compositional variations span the full range of terrestrial planet compositions observed within the Solar System (i.e. under this taxonomy, both Venus and Mars would be classified based on their elemental composition as being Earth-like). – 23 – As such, the definition of Earth-like should not be taken to mean identical to Earth in composition. Rather the classification refers to an elemental composition merely similar to that of the Earth. Bearing that definition in mind, three systems (HD27442, HD72659 and HD213240) were found to produce condensation sequences (and thus also terrestrial planets) comparable to those of the Solar System. See Figure 12 for the condensation sequence for HD27442. Schematic representations of the abundances (for disk conditions at 5×105 years) are shown in Figure 13 while the final elemental abundances for all times studied are shown in Table 11. From these it can be seen that for HD27442, HD72659 and HD213240 the outermost terrestrial planets produced are grossly similar in composition to known terrestrial planets. Their compositions are dominated by O, Fe, Mg and Si with varying amounts of other elements. Upon closer examination, however, large and important differences emerge, primarily due to variations in the compositions of the host star and thus the initial system itself. Pronounced radial compositional variations can be seen in the simulated terrestrial planets of all three systems. Planets located within ∼0.5AU for all three systems contain large amounts of Al, Ca and O, indicating that these planets formed from bodies containing the high-temperature Al and Ca condensates (such as spinel and gehlenite) (see Figure 13). However, beyond ∼0.5AU, the refractory composition steadily decreases, producing planets with compositions more closely correlated with that of the Earth, dominated by O, Fe, Mg and Si, in the outer regions of the system. Planets produced within this transition zone between the two planetary types, however, are best described as being refractory-rich, Fe-poor silicate planets. As expected, this difference is reflected in the planetary geochemical ratios, as the planets located within the inner region have Mg/Si, Al/Si and Ca/Si ratios well above – 24 – the Solar System terrestrial planet values. However, for planets located beyond the compositional transitional point, this is not the case. In this region, planetary Mg/Si, Al/Si and Ca/Si values are comparable to Earth values. A steady transition in composition between these two regions is seen for all three systems (see Figure 14). This trend lies well above the observed Earth fractionation line and is a result of the condensation of refractory species in the innermost region and Mg-silicate species further out, with relatively little radial mixing between the two regions during the formation process. It is worth noting that while the Mg/Si values for the planets produced in the system of HD213240 are still just within the upper limits of the accepted Earth-like Mg/Si values, the higher modeled planetary values are a result of the increased Mg/Si value of HD213240 itself (Mg/SiJ = 1.05, Mg/SiGl777 = 1.48). This results in a system with olivine as the major Mg silicate condensate and little pyroxene present. The model terrestrial planets in the three systems with condensation sequences comparable to that of the Solar System (HD27442, HD72659 and HD213240) can thus be characterized as being essentially similar in composition to Ca- and Al-rich inclusions (CAI's) (for the innermost planets) and Earth-like (for the outermost planets). 4.2.2. C-rich Planets In systems with C/O values close to or above 0.8, the planets that form can begin to incorporate carbon as a significant planet-building material, in the form of graphite, SiC and TiC, which is a significant difference compared to what occurs in the Solar System. Seven such systems were selected for the current study: four with low carbon enrichment (0.8.C/O<1: 55Cnc, Gl777, HD17051 and HD177830) and three with a significantly higher C content (C/O>1: HD4203, HD19994 and HD108874). The final elemental abundances for all times studied are shown in Table 11. – 25 – 4.2.3. Low C-enriched Planets Gl777: For disk conditions at t = 5×105 years, Gl777 produces Earth-like terrestrial planets with minor C enrichment. See Figure 12 for the condensation sequence for Gl777 and Figure 15 for a schematic of the resulting planetary elemental abundances. The final elemental abundances of the refractory lithophile and siderophile elements are similar to those of Earth with the simulated planets displaying a marginal enrichment in Mg (∼3wt%) and a depletion in Si (∼1wt%) and Fe (∼2wt%) compared to Earth values. The most volatile species (P, Na and S) are enriched over Earth, which is likely a result of the fact that we do not consider volatile loss during accretion, which is expected to be significant as for the Solar System simulations. The geochemical ratios of Mg/Si and Al/Si are enriched compared to those of the Earth, although they are still in accepted range for an Earth-like planet in this study (see Figure 16). As for HD213240, this increase in the planetary Mg/Si values over that observed for the Earth is due to the slight increase the Mg/Si value of Gl777 itself (Mg/SiJ = 1.05, Mg/SiGl777 = 1.29). In turn, this produces a system containing nearly equal amounts of olivine and pyroxene (compared to the pyroxene dominated Solar disk) and results in Mg-enriched planets. Note that we do not see such a large spread in planetary Al/Si values as was observed in Figure 14 as Gl777 does not contain any terrestrial planets in the region dominated by Al-rich CAI-like material (i.e. within ∼0.5 from the host star). The Ca/Si values, however, are lower than those of Earth (Ca/SiGl777 = 0.07, Ca/SiL = 0.11), although once again falling within our accepted range of values. This variation is primarily due the fact that there is relatively less Ca within the system. Ca is one of the least enriched elements within Gl777 ([Ca/H] = 0.10 vs. [Al/H] = 0.34), resulting in a relative Ca depletion within the solid material. The variation in the abundances of the host star is reflected in the lower Ca/Si value of the final planets produced. This difference is – 26 – certainly no larger than that observed for the Solar System simulations previously discussed (Bond et al. 2009). Although the C/O ratio for Gl777 is slightly below 0.8 (C/OGl777=0.78), a carbon-rich region is still predicted to occur at lower pressures (10−5 bar and below), resulting in the production of a narrow carbide-dominated region (extending from 0.77 to 1.13AU at 5×105 years). As a result, the terrestrial planets produced in Gl777 are all slightly enriched in C compared to the Earth (containing up to 0.5 wt% C). This enrichment decreases for disk conditions at later times as the C region migrates inwards, interior to the feeding zones of the terrestrial planets. Given these simulated compositions, the terrestrial planets that could form around Gl777 are best characterized as being Earth-like with a minor C-enrichment in their chemical composition. It is also interesting to note Gl777 is very close to the average extrasolar planetary host star values of Mg/Si and C/O (1.29 and 0.78 respectively, compared to the average values of 1.32 and 0.77). This result implies that terrestrial planets in an "average" extrasolar planetary system would have compositions comparable to that of our own Solar System but moderately enriched in Mg, and with a potentially minor C-enrichment. 55Cnc: Like Gl777, 55Cancri produced terrestrial planets that are C-enriched, containing up to 9.28 wt% C (see Figure 15). They are similar to Earth in that they are expected to be dominated by Mg-silicate species, with metallic Fe also present. However, the planets of 55Cnc are also enriched in both S and Mg beyond our acceptable limits to be called an Earth-like planet. This results in modeled planets with Mg/Si values above that of the Earth and Ca/Si values well below. As for HD213240 ad Gl777, this high planetary Mg abundance is caused by the fact that 55Cancri is highly enriched in Mg ([Mg/H] = 0.48), resulting in olivine becoming the major silicate species present within the disk and thus producing the high Mg/Si value observed. – 27 – Although the disk of 55Cancri contains a C-rich zone, producing C-enriched terrestrial planets for disk conditions at t=5×105, for disk conditions at later times, none of the simulated planets are predicted to contain any C because the C-rich region moves interior to the terrestrial planet zone in that system. All of the terrestrial planets in 55Cnc are located between 1.5 and 4AU while the C-rich zone is located between ∼1375 and 713K, corresponding to a radial distance of 0.46 and 1.48AU (for disk conditions in the Hersant et al. (2001) model at 5×105 years). Thus the primary feeding zones for each of the planets are located on the outer edge of the C zone at 5×105 years, and well beyond it at later times. Thus for disk conditions at later times, planets best described as Mg-rich Earths are produced. The location of the C zone also implies that the four inner known giant planets of the 55Cnc system (located at 0.038AU, 0.115AU, 0.24AU and 0.781AU) should contain significant amounts of C, both in their solid cores and in their atmospheres if they formed at or in close proximity to their current orbital locations and also depending on the exact time of their formation. Given the variation in the location of the C rich zone with time, it is expected that terrestrial planets containing some C would also be produced in the current simulations for disk conditions at later times if a time varying equilibrium composition for the solid material was incorporated into the simulations rather than the "snapshot" approach taken here. HD17051: Although the disk of HD17051 does contain a C-rich zone, for the disk conditions at t = 5×105 years the simulated terrestrial planets do not contain any C, instead resembling the planets of Solar System-like systems previously discussed (see Figure 15). The innermost planets (within ∼0.5 AU from the host star) are dominated by Al and Ca, resembling CAI's, while the outer planets are more Earth-like, consisting of Mg-silicates and metallic Fe. They are still significantly enriched in Fe, mostly due to their primary feeding zone location (and subsequent composition) and the high metallicity of HD17051 itself ([Fe/H] = 0.26). – 28 – However, for planetesimals initially forming under disk conditions at later times the planetary composition for all simulated planets changes to more closely resemble a C-enriched Earth-like planet, with planets dominated by O, Fe, Mg and Si and a significant amount of C. Up to 4.37 wt% C is predicted to exist in the planets for the disk conditions at 3×106 years. These planets are essentially C-enriched Earths, containing the same major elements in geochemical ratios within limits to be considered Earth-like, but also an enhanced inventory of C, primarily accreted as solid graphite. As for 55Cnc, it is expected that if were we to incorporate time-varying equilibrium compositions into our models that we would see C occurring in the terrestrial planets for all simulation times. HD177830: HD177830 has the highest Mg/Si (and Al/Si) ratio of any system simulated. This enrichment alters the compositions of major silicate species present within the disk. While the Solar System should have condensed both olivine and pyroxene between 0.35 and 2.5 AU, HD177830 is dominated by olivine beyond 0.3 AU and contains only a small region where pyroxene is predicted to coexist. This unusual composition is reflected in the final planetary abundances as the planets contain large portions of Mg (up to 22.33 wt%) (see Figure 15) and have a mean Mg/Si value of 1.71, well above Earth values (Mg/SiL = 1.01). Al is also similarly enriched (up to 31.74 wt%), again because of the high Al abundance of the host star and thus the initial system itself. Other refractory and lithophile elemental abundances within the final planets are comparable to that of the Solar System. Of the 10 planets produced in our simulations, five also contain trace amounts of C (up to 2.96 wt%). This increases for disk conditions at later times as the C-rich region migrates inwards from its initial location at 0.63 - 1.66 AU (at t = 5×105 years), through the primary feeding zone, producing planets with increased C-abundances (up to 9.8 wt%). The planets of HD177830 can best be described as being Mg- and Al-rich silicate planets with some C-enrichment. – 29 – Such a Mg dominated planetary composition would undoubtedly alter the interior structure and processes of the planets themselves. Such considerations will be discussed further in Section 5. 4.2.4. High C-enriched Planets HD19994, HD108874 & HD4203: All three systems all have C/O values above 1.0 (1.26, 1.35 and 1.86 respectively). In all three systems, the inner regions of the disk are completely dominated by refractory species composed of C, SiC and TiC, as opposed to the Ca and Al-rich inclusions characteristic of the earliest solids within our Solar System (see Figure 12 for the condensation sequence for HD4203). Significant amounts of metallic Fe are also present within these systems. As all three systems produced terrestrial planets located within 0.7AU from their host star, these unusual inner disk compositions produced terrestrial planets primarily composed of C, Si and Fe. HD19994 produced terrestrial planets composed almost entirely of SiC and metallic Fe and containing up to 37 wt% Si and between 31 and 63 wt% C, over 100 times more C than is estimated for Earth (see Figure 17, upper left). The outermost terrestrial planet for HD19994 does contain significant amounts of O and Mg, primarily because its feeding zone, although still undoubtedly dominated by C, is also rich in pyroxene. This presence of a Mg silicate species produces a slightly more varied composition for a single simulated planet in one of the four simulations completed. More extreme deviations occur when we consider the planets formed for HD108874 and HD4203. Both of these systems have considerably wider graphite dominated regions, extending from 1.5AU to within 0.1AU (for disk conditions at 5×105 years). As a result, terrestrial planets are found to be composed almost entirely of C, Si and Fe (see Figure 17). Both HD108874 and HD4203 produced terrestrial planets composed almost entirely of C and SiC and containing between 67 and 74 wt% C (for midplane conditions at 5×105 – 30 – years). Only the outermost planet of HD108874 contained less than 50 wt% C (30.54 wt% C), with the remainder of the planet consisting of Fe and Si. For disk conditions at later times, Mg and O are also present in all planets produced for both systems, again due to the incorporation of pyroxene and olivine into the planetary feeding zones. It should be noted that most of the terrestrial planets formed in the HD4203 simulations are single embryos that survived for the duration of the simulation but did not accrete any other solid material. Terrestrial planets within these systems are unlikely to have compositions resembling that of any body we have previously observed. The possible implications of these types of planetary compositions will be discussed in Section 5. 4.3. Stellar Pollution The average change in stellar photospheric abundances produced by accretion of solids onto the star for disk conditions at 5×105 years is shown in Table 12 for each system studied. The majority of systems experienced minimal increases in photospheric abundances as a result of accretion during terrestrial planet formation. The largest elemental enrichments occurred for the most refractory elements (Al, Ca, Ti, Ni and Cr), primarily because refractory material is initially located closest to the star itself. The simulations for HD72659 produced many of the most significant enrichments for all elements examined. This is primarily due to the large amount of mass accreted (2.640 ML) and the lower estimated mass of the stellar convective zone (0.0112MJ). However, as previously discussed, mixing within the stellar radiative zone is not incorporated into the current approach. As such, these values are upper limits for those stars with low mass convective zones and large radiative zones. Furthermore, HD72659 (along with Gl777) has the highest degrees of radial mixing and subsequently also accreted the largest amount of solid material onto their host stars. This resulted in a larger change in the stellar photospheric abundances than for other – 31 – systems. All of the predicted abundance changes are below the errors of current spectroscopic surveys (e.g. ±0.03 for Fischer & Valenti (2005)), meaning that definitively observable elemental enrichments are not necessarily predicted by our terrestrial planet formation simulations. Of course, inclusion of a swarm of planetesimals within the formation simulations and migration of the giant planets is expected to increase the amount of material accreted by the host star and thus also the predicted stellar abundances. However, these increases are expected to be minor and would thus still result in only marginal increases in the observed elemental abundances 5. Implications and Discussion 5.1. Mass Distribution Radial midplane mass distributions based on the equilibrium condensation sequence were calculated for each system. As composition is correlated to a specific radial distance within the midplane (via the Hersant et al. (2001) model), the total mass of solid material present interior to a given radial distance within the disk is given by summing the masses contained within each annulus of the disk for which composition is calculated: Mass of solid material = Σi=5.0AU i=0.1AU2πridrMsolid, i (5) where Msolid, i is the mass of solid material determined by the chemical model to be located in an annulus of width dr at ri. The mass of solid material possible is determined by using – 32 – the minimum mass solar nebula with a gas surface mass density profile varying as r−3/2, normalized to 1700 gcm−2 at 1 AU. Note that this approach only considers equilibrium driven condensation and neglects other processes that may migrate material and alter the mass distribution. Based on this calculation, the most carbon-rich systems simulated have significant differences in their mass distributions compared to other systems. The combination of a broad zone of refractory carbon-bearing solids in the inner regions and the relatively small amount of water ice that condenses in the outer regions of these systems suggests that C-rich systems have significantly more solid mass located in the inner regions of the disk than for a Solar-composition disk. This can be seen in Figure 18 which shows the distribution of solid mass within each system normalized to a solar composition disk for disk conditions at t = 5×105 years. From Figure 18, the system with the highest C/O value (HD4203) clearly contains significantly more mass in the innermost regions of the disk than a disk of solar composition does. However, the total mass of solid material produced by the current approximations is only 18.4 ML. As such, it is not clear that a giant planet core composed of refractory C-rich species could be produced within several AU of the host star, allowing for giant planet formation to occur much closer to the host star than previously thought. If significantly more mass were present within the disk than is currently estimated, such a scenario may be possible and would obviously alter the extent and nature of planetary migration required within these systems as it would no longer be required that a planet located at 1-2AU originally formed at 5AU and migrated inwards. Alternatively, if insufficient mass is available for Jovian core formation, production of large terrestrial planets in this region may proceed faster and with greater ease, thus increasing the chance of forming detectable terrestrial planets. – 33 – It can be seen from Figure 18 that the planetary system with the most Solar-like composition (HD 72659) has a mass distribution similar to that of the Solar System. The enrichment observed over the solar mass distribution within ∼0.5AU from the host star is due to the higher Mg/Si value for HD72659 resulting in the condensation of more Mg silicates. This enhancement is more obvious for the system with the highest Mg/Si value (HD177830). Likewise, the refractory rich system HD27742 also contains more mass in the inner regions of the disk than for a Solar abundance. This is due to the high abundances of several refractory elements ([Al/H]=0.53, [Na/H]=0.41, [Ca/H]=0.12). The variation in mass observed for all systems at ∼4.8AU is due to the condensation of hydrous species in various amounts relative to the solar composition disk. These results imply that the use of an alternative initial mass distribution may be required for planetary systems with high C/O values. Although the planetary formation simulations presented here utilized the classical Solar-System based mass distribution, such an approach is still valid for the illustrative purposes of the current study. The full implications of these results need to be examined in future work by using alternative mass distributions for extrasolar planetary formation simulations for both gas giant planets and smaller terrestrial planets. 5.2. Timing of Formation Specific planetary compositions have been found to be highly dependent on the time selected for the disk conditions, especially for those systems containing C-dominated regions. This is primarily due to the low degree of radial mixing encountered within the simulations. As a result, as conditions within the area immediately surrounding the planet evolve, the composition of the solid material and thus the final planet itself drastically change. For disk conditions at later times,the simulated planetary compositions evolve to more – 34 – closely resemble those of the Solar System. They become dominated by Mg silicate species and metallic Fe. Terrestrial planets in Solar-like systems attain more hydrous material. Variations in composition with disk condition times are most noteworthy for those planets dominated by refractory compositions (such as the inner planets of HD72659 and HD27442). Under later disk conditions, these planets experience a complete shift in their composition, losing the majority of their refractory inventory to be composed primarily of Mg-silicates (olivine and pyroxene). Therefore if solid condensation and initial planetesimal formation occurred significantly later, we would expect to observe predominantly Mg-silicate and metallic Fe planets with enrichments in other elements (such as C) depending on the exact composition of the system. Although disk conditions at 5×105 years provide the "best fit" for Solar System simulations (based on fitting the compositions of Venus, Earth and Mars) and are thus utilized here, it remains to be seen whether or not disk conditions at this time provide an accurate description of the conditions under which planetesimals and embryos formed in other planetary systems. Therefore, we require a more detailed understanding of the timing of condensation and planetesimal and embryo formation within protoplanetary disks to be able to further constrain the predicted elemental abundances. Similarly, as the disk evolves, the various condensation fronts migrate closer to the host star. For example, the water ice line for Gl777 migrates from 7.29 AU for midplane conditions at 2.5×105 years to 1.48AU for midplane conditions at 3×106 years. Similar degrees of migration also occur for the condensation fronts of other species (such as the Mg silicates olivine and pyroxene). In effect, the change in location of the condensation fronts alters the mass distribution within the disk, increasing the mass present in both the very closest regions of the disk (< 1AU) as the refractory species are replaced by the more abundant Mg silicates and in the outer most regions (> 3AU) as water ices appear. The full effects of this change will require formation simulations to be run with alternative mass distributions but it is thought that such conditions will increase the efficiency of forming – 35 – close-in terrestrial planets and/or the mass of the resulting planets. Additionally, it will also allow for efficient terrestrial planet growth in the outer regions, possibly to the extent of forming gas giant cores. Given that the average disk lifetime is ∼3Myr, not all disks will reach the final chemical compositions modeled here. Thus it remains to be seen not only whether or not sufficient solid mass would be retained during the evolutionary process for Jovian cores to develop but also whether or not core formation can occur before the disk is cleared out. 5.3. Detection of Terrestrial Planets The results of this study are of great importance for the design of terrestrial planet finding surveys. Our simulations, while preliminary, suggest that terrestrial planets can be stable in a wide range of extrasolar planetary systems. Four distinct classes of planetary composition have been produced by the current simulations: Earth-like, Mg-rich Earth-like, refractory (compositions similar to CAI's) and C-rich. These planetary types are primarily a result of the compositional variations of the host stars and thus the system as a whole. Based on their observed photospheric elemental abundances, the majority of known extrasolar planetary systems are expected to produce terrestrial planets with compositions similar to those within our own Solar System. Therefore, systems with elemental abundances and ratios similar to these (e.g. HD72659) are ideal places to focus future "Earth-like" planet searches. Based on our dynamical simulations, the masses of the terrestrial planets produced are too low to be detected by current radial velocity surveys. However, many of the simulated planets are in orbits (assuming the simulated inclinations are correct) that would place them within the prime target space for detection by the Kepler mission. Designed to detect extrasolar planets via transit studies, Kepler is the first mission that has the potential of – 36 – detecting Earth-mass (and lower) extrasolar planets located within the habitable zone of a planetary system. It has the sensitivity to detect the transit of an Earth mass body within 2AU from the host star and a Mars mass body (0.1Me) within 0.4AU. Single transit events may not be sufficient to positively identify the presence of a planet, although planets at 1 AU should transit 3-4 times during the lifetime of the kepler mission, thus reducing the likelihood of contamination in the data. The vast majority of the terrestrial planets formed here (with the exception of the lowest mass, highest semimajor axis planets) are well within this range and thus may be detectable if they are indeed present within these systems (assuming the system is aligned such that transits can be observed from orbit). Only HD4203 produces no potentially detectable planets, based on their predicted masses. Thus it is likely that we will have an independent check of extrasolar terrestrial planet formation simulations within the next 5 years (but not necessarily for these systems, only for a range of systems which may be similar to these). Such information will be vital for further refinement of planetary formation models for both giant and terrestrial planets. Obtaining compositional information about such terrestrial planets, however, will be more difficult as the size and location of the predicted planets will prohibit direct spectroscopic studies. It is also unlikely that the terrestrial planets will contain atmospheres large enough to be detected and characterized by transit surveys. As such, detailed extrasolar terrestrial planetary chemical compositions will remain unknown for the foreseeable future. In addition to detection via transit surveys, attempts are also being made to obtain direct images of extrasolar planetary systems. One such example is Darwin, a proposed ESA space based mission that would utilize nulling interferometry in the infrared to directly search for terrestrial extrasolar planets. The compositional variations outlined here are likely to influence our ability to successfully detect these planets. Carbon-rich asteroids are known to be highly non-reflective. For example, 624 Hektor (D-type asteroid) has a geometric albedo of 0.025 while 10 Hygiea (C-type asteroid) has a geometric albedo of – 37 – 0.0717 (compared to a geometric albedo of 0.367 for the Earth and 0.113 for the moon). As both of these asteroids are assumed to be carbon-rich, it is likely that the carbon-rich planets identified here are similarly dark. Thus it is expected that searches for these planets in the visible spectrum will be difficult due to the small amount of light reflected by these bodies. However, a lower albedo results in greater thermal emission from a body, suggesting that the infrared signature from these planets would extend to shorter wavelengths than corresponding silicate planets. As a result, infrared searches (such as that of Darwin) are ideally suited to detect carbon-rich terrestrial planets and thus should be focused on stellar systems with compositions similar to that of the C-rich stars identified here to maximize results. Of course, this conclusion neglects any possible effects of a planetary atmosphere. 5.4. Hydrous Species As one would intuitively expect, hydrous material (water and serpentine in the current simulations) is primarily located in the outer, colder regions of the disks. This corresponds to beyond ∼7.3AU for disk conditions at 2.5×105 years and beyond ∼1.4AU for disk conditions at 3×106 years for all compositions examined. As a result of this distribution, terrestrial planets forming in the inner regions of a given planetary system are unlikely to directly accrete significant amounts of hydrated material. In the current in-situ simulations, none of the simulated terrestrial planets directly accrete any hydrous species for disk conditions at 5×105 years. As the composition of the planetesimals is dictated by the thermodynamic conditions of the disk at the time of condensation only, if any of the simulated planetesimals were to condense/form at later times, they would be more likely to – 38 – be water-rich given that the 'snow line'4 migrates inwards as the disk cools, producing a greater overlap between their feeding zones and the water-rich region of the disk. However, a far greater effect is expected to be produced by migration of giant planets within the system (Raymond et al. 2006b; Mandell et al. 2007). Migration of a giant planet has been shown to be capable of driving a large amount of material from the outer regions of a disk inwards. In the case of hydrous material, this has been found to result in water-rich terrestrial planets being formed both interior and exterior to the giant planet (Raymond et al. 2006b; Mandell et al. 2007). The full extent of this increase in water content will be examined by a suite of simulations incorporating giant migration that are currently running and will be the focus of future work. In the Solar System, it has been hypothesized that the Earth's water could have originated from hydrated material in the region where the asteroid belt now lies, or from comets beyond the orbit of Jupiter. While belts of debris resembling the asteroid belt could exist interior to the giant planets in several of the extrasolar systems and may be incorporated into the terrestrial planets, as noted above, none of that material is expected to be hydrated at early times in the disk. We can not address the issue of cometary delivery of water in these simulations, as bodies beyond the orbits of the giant planets are not presently included in our dynamical models. Water can also potentially be incorporated into a planetary body via adsorption onto solid grains within the disk (Drake 2005). While this process has not been considered in our current simulations, it is possible that there will be some water delivered during the formation process to the terrestrial planets produced in the Earth-like systems (HD27442, 4Note that here 'snow line' refers to the location within the disk where the thermodynamic conditions are such that water ice condenses (i.e. where T = 150K). – 39 – HD72659 and HD213240) as the solid grains are bathed in water vapor over the entire span of the disk. This same process will likely not be as effective at delivering water to the C-rich systems (55Cnc, Gl777, HD4203, HD17051, HD19994, HD108874 and HD177830) as they only have water vapor present at temperatures below ∼800 K. This temperature range corresponds to beyond a radial distance of ∼1.2AU for Hersant et al. (2001) midplane conditions at 2.5×105 years and ∼0.2AU for midplane conditions at 3×106 years. As few terrestrial planets in our simulations accrete material from beyond 1.2AU, it is expected that C-rich planets forming early in the lifetime of the disk will remain dry without additional water being incorporated into the planets via adsorption. Thus it appears that terrestrial planets are likely to obtain some amount of water (through giant planet migration mixing the disk, variations in composition with time i.e. heterogeneous accretion, adsorption and exogenous delivery), while those within Solar-like systems may receive more water and other hydrous species than terrestrial planets within C-rich systems. 5.5. Planetary Interiors and Processes Given the wide variety of predicted planetary compositions, a similarly diverse range of planetary interiors structures is also expected. To better quantify this, we examined three specific cases: a 1.03ML Earth-like planet located at 0.95AU (HD72659), a 1.22ML Mg-rich Earth-like planet located at 0.43AU (HD177830) and a 0.47ML C-rich planet located at 0.38AU (HD108874). Approximate interior structures for each were calculated using equilibrium mineralogy for a global magma ocean with P = 20GPa and T = 2000◦C. Equilibrium compositions at these conditions have been found to produce the best agreement between predicted and observed siderophile abundances within the primitive upper mantle of the Earth (Drake 2000). Elemental abundances were taken from the results discussed in Section 4.2. Resulting mineral assemblages were sorted by density to define – 40 – the compositional layers. Note that this assumes that a planet undergoes complete melting and differentiation. Approximate planetary radii were obtained from Sotin et al. (2007) based on planetary mass. These planetary radii are based on silicate planetary equations of state and as such are unlikely to completely describe the C-rich planets modeled here. However, no studies have considered such assemblages, forcing us to assume a silicate based approximate radius. Density variations at high pressures were not considered in defining the depths of various layers. Although important for planetary evolution processes such as mantle stripping, the effects of large impacts (such as the moon forming impact) are also neglected. The resulting interior structures (shown to scale) can be seen in Figure 19. The interior mineralogy and structure of one of our model planets orbiting HD72659 is similar to Earth. It contains a pyroxene and feldspathic dominated crust (∼133 km deep) overlying an olivine mantle (∼985 km deep) with an Fe-Ni-S core (radius ∼ 4930 km). The crust is thicker than seen on Earth as we are currently neglecting density and phase changes. Given its structure and comparable mineralogy, we would expect to observe planetary processes similar to those seen on Earth. The planets location within the habitable zone of the host star suggests that a liquid water ocean is feasible, provided sufficient hydrous material can be delivered. Melting conditions and magma compositions are expected to be comparable and it is feasible that a liquid core would develop, resulting in the production of a magnetic dynamo. In general, based on their mass and composition, the terrestrial planets of HD72659 are likely to have structures and mineral assemblages similar to those observed in our system. The simulated planet for HD177830 (the system with the highest Mg/Si value) is depleted in Si, relative to the Earth, resulting in high spinel and olivine content in the mantle (resembling that of type I kimberlites) and a thicker mellite and calcium dominated crust than found for HD72659 (∼309 km deep). The core would produce a considerable – 41 – amount of heat via potential energy release during differentiation, potentially producing melts with compositions similar to komatiite (dominated by olivine with trace amounts of pyroxene and plagioclase). Volcanic eruptions would be comparable to basaltic flows observed on Earth due to the low silica content of the melt. However, given the thickness of the crust, extrusive volcanism and plate tectonics are unlikely to occur as high stress levels would be required to fracture the crust. Producing and sustaining such stresses would be challenging. Therefore, it is questionable whether or not a planet with this composition and structure would be tectonically active for long periods of time. Given the similar composition and size of the core compared to Earth, a magnetic dynamo is still expected to be produced within the core. Finally, carbide planets are expected to form around HD108874. The resulting composition and structure is unlike any known planet. Its small size, refractory composition and possible lack of radioactive elements (due to the potential absence of phosphate species, common hosts for U and Th, and possible lack of feldspar and carbonates, the common host of K) will inhibit long-term geologic activity due to the difficulty of melting the mantle. Only large amounts of heat due to core formation and/or tidal heating would be able to provide the required mantle heating. Once all the primordial heat has been removed, it is unlikely that the mantle would remain molten on geologic timescales. Until that time, given the buoyancy of molten carbon, volcanic eruptions would be expected to be highly enriched in C. The core is also expected to be molten, thus making it likely that a magnetic dynamo would be produced (Gaidos & Selsis 2007). Note that this assumes that sufficient heat is initially available to melt the body and allow for differentiation and core formation to occur in the first place. In essence, although initially molten and probably active, old carbide planets of this type would be geologically dead. Incomplete mixing of material accreted at later times is likely to result in deviations – 42 – from the equilibrium picture presented here. For example, accretion of oxidized and water-rich material late in the formation process may result in a stratified redox state and water-rich crust as observed for the Earth. Unfortunately, it is not possible to determine these effects with current models as it requires a level of understanding of the impact and accretion process (e.g. mantle mixing, fragmentation) on small planetary bodies that we currently do not have. These results are also key for super Earth studies such as that of Valencia et al. (2007) and O'Neill et al. (2007). Previous simulations have assumed Earth-based compositions and structures. Based on the present simulations, a wide variety of both are possible and will need to be considered. 5.6. Planet Habitability The habitable zone of a planetary system is defined as being the range of orbital radii for which water may be present on the surface of a planet. For the stars considered here, that corresponds to radii from ∼0.7AU to ∼1.45AU. The vast majority of the planets produced by the current simulations orbit interior to this region (exterior in the case of 55Cnc) and thus are unlikely to be habitable in the classical sense. 10 planets are produced within the classical habitable zone, existing in orbits extending from 0.70AU to 1.19AU. Seven on these planets are formed in Solar-like systems (HD24772 and HD72659) and have compositions loosely comparable to that of Earth. As such, we feel that these systems (and others similar to them) are the ideal place to focus future astrobiological searches as they may not only contain planets with compositions similar to that of Earth but also exist in the biologically favorable region of the planetary system. Of the seven C-rich systems, only two produced planets within the habitable zone. – 43 – Gl777 formed two terrestrial planets within the habitable zone while HD19994 formed a terrestrial planet at 0.70AU, just at the inner edge of the habitable zone. All other planets are located well inside the required radii. Both of the habitable planets around Gl777 are C-enriched Earth-like planets, making them potential sites for the development of life. The single habitable planet produced by the HD19994 simulations is dominated by C, along with O, Fe, Si and Mg. It is unclear whether such a composition would be favorable to life. Additionally, the low planet mass (0.06ML) further makes it unlikely that this particular simulated planet could ever actually host life. As such, under the current definition of habitable, we conclude that of the seven C-rich systems currently simulated, Gl777 has the best chance of supporting life, but this is by no means guaranteed. 5.7. Biologically Important Elements In addition to water, complex life (as we know it) also requires several key elements to exist. The six essential elements are H, C, N, O, P and S. As was the case for the Solar System simulations discussed in Bond et al. (2009) none of the planets accreted any N and are also lacking in H. The terrestrial planets formed in the Solar-like systems contained various amounts of O and P but some were deficient in S and all were laking C, as for the Solar System simulations. The most C-rich systems, on the other hand, were lacking in O, P and S. Thus it is clear that for life as we know it to develop on any of the terrestrial planets formed in the current simulations, significant amounts of several elements must be supplied from exogenous sources within the system. All elements may be supplied from the outer, cooler regions of the disk. Thus it is possible that migration or the radial mixing of cometary-type material into the terrestrial planet region may produce planets with the necessary elements for life to develop. As for the Solar System simulations, all biologically – 44 – required elements would be introduced in a form that could potentially be utilized by early life. This is especially intriguing for those planets located within the habitable zone. On the other hand, alternative pathways could potentially develop for the formation of an alternative biologic cycle without requiring the same six elements. 5.8. Host Star Enrichment As previously discussed, stellar photospheric pollution has been suggested as a possible explanation for the observed high metallicity of extrasolar planetary host stars (Laughlin 2000; Gonzalez et al. 2001; Murray et al. 2001). The current simulations, though, do not support this hypothesis. Enrichments are produced primarily in Al, Ca and Ti, not Fe as is required by the pollution theory. Furthermore, relatively small masses of solid material are accreted by the host stars during planet formation, suggesting that insufficient material is accreted to produce the observed enrichments. Thus unless migration of the giant planets can systematically result in accretion of giant planets by the host star, our results agree with with previous authors (e.g Santos et al. 2001, 2003a,b, 2005; Fischer & Valenti 2005) in finding that the observed host star enrichment is primordial in origin. Our simulations also imply that enrichments due to stellar pollution are most likely to be observed for the refractory elements in high mass stars with low convective zone masses. This suggests that surveys for pollution effects caused by terrestrial planet formation should focus on Ti, Al and Ca abundances in A-type and high mass F-type stars as they are expected to have the lowest convective zone masses. However, more detailed simulations of the fate of material accreted into radiative zones need to be undertaken to support this hypothesis. – 45 – 6. Summary Terrestrial planet formation simulations have been undertaken for ten different extrasolar planetary systems. Terrestrial planets were found to form in all systems studied, with half of the simulations producing multiple terrestrial planets. The simulated planets are possibly detectable by Kepler (if the system orientation is favorable), thus potentially allowing for future independent verification of formation simulations. The compositions of these planets are found to vary greatly, from those comparable to Earth and CAIs to other planets highly enriched in carbide phases. These compositional variations are produced by variations in the elemental abundances of the host star and thus the system as a whole, with the Mg/Si and C/O ratios being the most important for determining the final planetary compositions. Based on this, it is expected that C-rich planets will comprise a sizeable portion of extrasolar terrestrial planets and need to be considered in significantly more detail. These compositions are highly dependant on the disk conditions selected for study, requiring us to develop a more detailed understanding of the timing of planetary formation within these systems. Given the wide variety of compositions predicted, it is also likely that planetary mineralogies and processes within these planets will be different from those of our own Solar System. Compositions range from planets dominated by Fe and Mg-silicate species to those composed almost entirely of Fe and C. These compositional variations are likely to generate differences in detectability with C-rich planets being easier to detect via infrared surveys such as the Darwin mission due to their lower albedo and hence larger IR emission. The most habitable planets are expected to be those forming in systems with compositions similar to Solar. The simulated compositions make these planets ideal targets for future astrobiological surveys and studies. Planets in the C-rich systems that we model are likely to be lacking water and generally located interior to the habitable zone, making – 46 – such planets unfavorable for the development of life as we know it. Finally, pollution of the host star by the planetary formation process appears to be negligible for the majority of systems. Enrichments are produced only for those stars with the least massive convective zones and even then only in the most refractory elements (Ti, Al and Ca). Therefore, it is unlikely that pollution by accreted terrestrial material is a viable explanation for the currently observed host star metallicity trend. This also implies that pollution studies should be undertaken for A-type and massive F-type stars as they are more likely to display the preferential enrichment in Ti, Al and Ca that appears to be indicative of terrestrial planet formation. J. C. Bond and D. P. O'Brien were funded by grant NNX09AB91G from NASA's Origins of Solar Systems Program. J. C. Bond and D. S. Lauretta were funded by grant NNX07AF96G from NASA's Cosmochemistry Program. Thanks to the anonymous reviewer for their helpful comments in the production of this paper. A. Online Material: Tables Table 1. Predicted bulk elemental abundances for all simulated extrasolar terrestrial planets. All values are in wt% of the final predicted planet for all seven sets of disk conditions examined. Planet number increases with increasing distance from the host star. System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C 55Cnc t=2.5×105 years 55Cnc 1-4 0.00 16.59 26.02 2.58 20.92 1.40 0.65 0.51 1.27 0.24 0.16 0.06 11.54 0.00 18.06 55Cnc 2-4 0.00 16.50 25.88 2.93 20.79 1.39 0.65 0.51 1.26 0.24 0.16 0.06 11.47 0.00 18.18 – 4 7 – 55Cnc 3-4 0.00 16.80 26.21 2.20 21.14 1.41 0.71 0.52 1.28 0.24 0.16 0.06 11.73 0.00 17.53 55Cnc 3-5 3.71 12.59 53.11 3.09 15.88 1.06 0.50 0.00 0.96 0.18 0.12 0.05 8.76 0.00 0.00 55Cnc 4-4 1.79 15.74 43.90 3.87 19.87 1.32 0.62 0.31 1.20 0.23 0.15 0.06 10.95 0.00 0.00 t=5×105 years 55Cnc 1-4 0.00 19.43 30.61 4.75 24.49 1.63 0.77 0.62 1.48 0.28 0.18 0.07 13.51 0.00 2.18 55Cnc 2-4 0.00 19.85 31.29 4.87 25.03 1.67 0.78 0.63 1.51 0.29 0.19 0.08 13.81 0.00 0.00 55Cnc 3-4 0.00 19.06 31.26 4.63 24.04 1.60 0.75 0.60 1.45 0.28 0.18 0.07 13.26 0.00 2.83 55Cnc 3-5 0.00 18.67 35.33 4.58 23.58 1.57 0.74 0.60 1.42 0.27 0.18 0.07 12.99 0.00 0.00 55Cnc 4-4 0.00 17.57 30.44 3.78 22.18 1.48 0.69 0.55 1.34 0.25 0.17 0.07 12.21 0.00 9.28 Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C 55Cnc 1-4 0.00 18.68 35.30 4.59 23.59 1.57 0.74 0.60 1.42 0.27 0.18 0.07 12.99 0.00 0.00 t=1×106 years 55Cnc 2-4 0.00 18.68 35.33 4.59 23.57 1.57 0.74 0.60 1.42 0.27 0.18 0.07 12.99 0.00 0.00 55Cnc 3-4 0.00 18.71 35.25 4.59 23.58 1.57 0.74 0.60 1.43 0.27 0.18 0.07 13.01 0.00 0.00 55Cnc 3-5 0.71 17.64 38.81 4.33 22.24 1.48 0.70 0.00 1.34 0.25 0.17 0.07 12.27 0.00 0.00 55Cnc 4-4 0.03 19.23 33.40 4.72 24.26 1.62 0.76 0.61 1.47 0.28 0.18 0.07 13.37 0.00 0.00 t=1.5×106 years 55Cnc 1-4 0.00 18.70 35.33 4.59 23.62 1.54 0.71 0.59 1.43 0.27 0.18 0.07 12.97 0.00 0.00 55Cnc 2-4 0.00 18.67 35.33 4.58 23.58 1.57 0.74 0.60 1.42 0.27 0.18 0.07 12.99 0.00 0.00 55Cnc 3-4 0.21 18.36 36.37 4.51 23.18 1.54 0.72 0.42 1.40 0.27 0.17 0.07 12.77 0.00 0.00 55Cnc 3-5 3.71 12.59 53.11 3.09 15.88 1.06 0.50 0.00 0.96 0.18 0.12 0.05 8.76 0.00 0.00 55Cnc 4-4 1.79 15.77 43.83 3.87 19.87 1.33 0.62 0.31 1.20 0.23 0.15 0.06 10.97 0.00 0.00 t=2×106 years – 4 8 – Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C 55Cnc 1-4 1.24 16.70 41.35 4.10 21.06 1.40 0.66 0.13 1.27 0.24 0.16 0.06 11.61 0.00 0.00 55Cnc 2-4 0.49 17.96 37.72 4.41 22.66 1.51 0.71 0.19 1.37 0.26 0.17 0.07 12.49 0.00 0.00 55Cnc 3-4 1.11 16.86 40.64 4.14 21.28 1.42 0.66 0.42 1.28 0.24 0.16 0.06 11.72 0.00 0.00 55Cnc 3-5 0.00 19.85 31.28 4.87 25.03 1.67 0.78 0.64 1.51 0.29 0.19 0.08 13.81 0.00 0.00 55Cnc 4-4 0.00 18.36 28.22 3.15 23.20 1.54 0.72 0.31 1.40 0.27 0.17 0.07 12.76 0.01 9.82 t=2.5×106 years 55Cnc 1-4 2.71 14.23 48.35 3.50 17.95 1.20 0.56 0.13 1.08 0.21 0.13 0.05 9.90 0.00 0.00 – 4 9 – 55Cnc 2-4 2.73 14.23 48.48 3.50 17.94 1.20 0.56 0.00 1.08 0.21 0.13 0.05 9.89 0.00 0.00 55Cnc 3-4 1.45 16.35 42.35 4.02 20.62 1.37 0.64 0.13 1.25 0.24 0.15 0.06 11.37 0.00 0.00 55Cnc 3-5 3.71 12.59 53.11 3.09 15.88 1.06 0.50 0.00 0.96 0.18 0.12 0.05 8.76 0.00 0.00 55Cnc 4-4 1.79 15.74 43.90 3.87 19.87 1.32 0.62 0.31 1.20 0.23 0.15 0.06 10.95 0.00 0.00 t=3×106 years 55Cnc 1-4 3.05 13.70 49.98 3.37 17.27 1.15 0.54 0.00 1.04 0.20 0.13 0.05 9.53 0.00 0.00 Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C 55Cnc 2-4 3.70 12.60 53.09 3.10 15.89 1.06 0.50 0.00 0.96 0.18 0.12 0.05 8.76 0.00 0.00 55Cnc 3-4 2.93 13.90 49.40 3.42 17.53 1.17 0.55 0.00 1.06 0.20 0.13 0.05 9.67 0.00 0.00 55Cnc 3-5 3.71 12.59 53.11 3.09 15.88 1.06 0.50 0.00 0.96 0.18 0.12 0.05 8.76 0.00 0.00 55Cnc 4-4 1.79 15.76 43.89 3.87 19.89 1.31 0.62 0.30 1.20 0.23 0.15 0.06 10.93 0.00 0.00 Gl777 t=2.5×105 years Gl777 1-4 0.00 16.61 33.23 1.58 22.48 5.64 3.79 0.27 1.46 0.28 0.15 0.30 13.96 0.01 0.24 Gl777 2-4 0.00 15.09 34.73 0.30 19.48 9.98 6.69 0.00 1.41 0.19 0.12 0.55 11.46 0.01 0.00 Gl777 2-5 0.00 12.90 36.83 0.48 13.81 13.74 9.30 0.12 1.00 0.15 0.08 0.76 10.62 0.00 0.20 Gl777 3-4 0.00 11.92 37.57 0.76 11.59 15.79 10.62 0.00 0.75 0.15 0.08 0.88 9.89 0.01 0.00 Gl777 3-5 0.00 16.15 33.83 0.76 21.45 7.32 4.96 0.08 1.54 0.23 0.13 0.40 13.00 0.01 0.13 Gl777 4-4 0.00 13.97 35.90 0.43 16.43 11.77 7.94 0.11 1.09 0.19 0.11 0.65 11.21 0.01 0.18 t=5×105 years – 5 0 – Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C Gl777 1-4 0.00 18.02 32.19 2.08 26.24 1.60 1.09 0.67 1.70 0.33 0.18 0.08 15.47 0.00 0.34 Gl777 2-4 0.00 18.62 31.84 0.62 27.28 1.67 1.13 0.37 1.77 0.34 0.19 0.07 15.61 0.01 0.50 Gl777 2-5 0.00 18.38 31.43 2.39 26.68 1.62 1.10 0.29 1.73 0.33 0.18 0.09 15.65 0.02 0.10 Gl777 3-4 0.00 18.65 31.78 0.95 27.32 1.67 1.13 0.35 1.77 0.34 0.19 0.08 15.49 0.00 0.29 Gl777 3-5 0.00 18.17 31.59 2.38 26.35 1.60 1.09 0.46 1.71 0.33 0.18 0.08 15.77 0.01 0.27 Gl777 4-4 0.00 18.45 31.82 1.50 26.84 1.64 1.11 0.41 1.74 0.33 0.18 0.08 15.68 0.01 0.19 t=1×106 years Gl777 1-4 0.00 17.21 32.61 4.69 24.89 1.52 1.03 0.77 1.61 0.31 0.17 0.08 15.05 0.00 0.04 Gl777 2-4 0.00 17.68 31.66 3.51 25.63 1.56 1.13 0.79 1.66 0.32 0.21 0.08 15.51 0.00 0.25 Gl777 2-5 0.00 17.61 32.47 3.24 25.49 1.55 1.05 0.80 1.65 0.32 0.17 0.08 15.41 0.00 0.16 Gl777 3-4 0.00 17.72 32.10 2.96 25.69 1.56 1.06 0.80 1.66 0.32 0.18 0.08 15.54 0.00 0.33 Gl777 3-5 0.00 17.40 32.09 4.39 25.20 1.54 1.07 0.79 1.63 0.32 0.19 0.09 15.24 0.00 0.06 Gl777 4-4 0.00 17.57 32.32 3.57 25.44 1.55 1.05 0.79 1.65 0.32 0.17 0.08 15.38 0.00 0.10 – 5 1 – Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C Gl777 1-4 0.00 16.94 33.13 5.23 24.50 1.49 1.01 0.76 1.58 0.31 0.17 0.08 14.78 0.00 0.02 t=1.5×106 years Gl777 2-4 0.00 17.21 32.33 4.87 24.97 1.52 1.03 0.78 1.62 0.31 0.17 0.09 15.10 0.00 0.00 Gl777 2-5 0.00 17.15 32.42 5.23 24.76 1.51 1.02 0.77 1.60 0.31 0.17 0.08 14.98 0.00 0.00 Gl777 3-4 0.00 17.30 31.95 4.79 25.22 1.53 1.04 0.78 1.62 0.32 0.17 0.09 15.17 0.00 0.03 Gl777 3-5 0.00 16.98 32.86 5.43 24.49 1.49 1.01 0.76 1.59 0.31 0.17 0.08 14.82 0.00 0.00 – 5 2 – Gl777 4-4 0.00 17.13 32.36 5.18 24.85 1.51 1.02 0.77 1.60 0.31 0.17 0.08 14.99 0.00 0.01 t=2×106 years Gl777 1-4 0.14 17.00 34.99 5.20 23.10 1.41 0.95 0.69 1.50 0.29 0.16 0.08 14.50 0.00 0.00 Gl777 2-4 0.00 17.07 32.64 5.43 24.57 1.50 1.01 0.77 1.59 0.31 0.17 0.08 14.86 0.00 0.00 Gl777 2-5 0.07 16.79 33.37 5.48 24.26 1.48 1.00 0.76 1.57 0.30 0.17 0.08 14.67 0.00 0.00 Gl777 3-4 0.00 17.01 32.87 5.40 24.49 1.49 1.01 0.76 1.59 0.31 0.17 0.08 14.82 0.00 0.00 Gl777 3-5 0.00 16.85 33.47 5.51 24.19 1.47 1.00 0.76 1.57 0.30 0.17 0.08 14.64 0.00 0.00 Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C Gl777 4-4 0.06 16.83 33.41 5.45 24.23 1.48 1.00 0.76 1.57 0.30 0.17 0.08 14.66 0.00 0.00 t=2.5×106 years Gl777 1-4 0.62 15.73 37.22 5.15 22.62 1.38 0.93 0.68 1.46 0.28 0.15 0.08 13.68 0.00 0.00 Gl777 2-4 0.00 16.80 33.60 5.48 24.15 1.47 1.00 0.75 1.56 0.30 0.17 0.08 14.62 0.00 0.00 Gl777 2-5 0.10 16.98 34.25 5.36 23.51 1.43 0.97 0.71 1.52 0.30 0.16 0.08 14.62 0.00 0.00 Gl777 3-4 0.00 16.88 33.24 5.52 24.29 1.48 1.00 0.76 1.57 0.31 0.17 0.08 14.70 0.00 0.00 Gl777 3-5 0.05 16.62 34.31 5.45 23.86 1.45 0.99 0.74 1.54 0.30 0.16 0.08 14.44 0.00 0.00 – 5 3 – Gl777 4-4 0.07 16.42 33.79 5.57 24.46 1.52 0.92 0.76 1.58 0.31 0.17 0.08 14.33 0.00 0.00 t=3×106 years Gl777 1-4 1.12 14.96 39.74 4.91 21.52 1.31 0.89 0.65 1.39 0.27 0.15 0.07 13.01 0.00 0.00 Gl777 2-4 0.00 16.76 33.90 5.48 24.02 1.46 0.99 0.75 1.56 0.30 0.16 0.08 14.54 0.00 0.00 Gl777 2-5 0.12 16.51 34.53 5.43 23.79 1.45 0.98 0.72 1.54 0.30 0.16 0.08 14.39 0.00 0.00 Gl777 3-4 0.00 16.81 33.69 5.50 24.09 1.47 1.00 0.75 1.56 0.30 0.16 0.08 14.58 0.00 0.00 Gl777 3-5 0.11 16.22 34.71 5.49 24.05 1.49 0.92 0.75 1.56 0.30 0.16 0.08 14.15 0.00 0.00 Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C Gl777 4-4 0.39 16.13 35.87 5.29 23.19 1.41 0.96 0.70 1.50 0.29 0.16 0.08 14.03 0.00 0.00 HD4203 t=2.5×105 years HD4203 1-3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.16 26.18 0.00 73.66 HD4203 2-3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.16 26.13 0.00 73.71 – 5 4 – HD4203 3-3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.16 26.20 0.00 73.64 HD4203 4-3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.16 26.23 0.00 73.61 t=5×105 years HD4203 1-3 0.00 0.00 0.00 0.00 0.63 0.00 0.00 0.00 0.00 0.00 0.12 0.16 26.07 0.00 73.02 HD4203 2-3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.16 26.26 0.00 73.58 HD4203 3-3 0.00 0.00 0.00 0.00 0.74 0.00 0.00 0.00 0.00 0.00 0.14 0.16 26.03 0.00 72.93 HD4203 4-3 0.00 0.00 0.00 0.00 7.95 0.00 0.00 0.00 0.78 0.02 0.30 0.14 23.90 0.00 66.89 Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD4203 1-3 0.00 0.00 0.00 0.77 31.36 1.95 0.96 0.00 2.13 0.35 0.23 0.09 16.58 1.01 44.56 t=1×106 years HD4203 2-3 0.00 0.00 0.00 0.16 30.45 1.83 0.20 0.00 2.16 0.26 0.24 0.10 16.83 0.95 46.81 HD4203 3-3 0.00 0.00 0.00 0.98 31.73 1.99 1.23 0.00 2.12 0.38 0.22 0.09 16.48 1.03 43.75 HD4203 4-3 0.00 0.09 0.00 1.10 31.84 1.98 1.23 0.00 2.11 0.40 0.22 0.09 16.51 1.03 43.40 t=1.5×106 years HD4203 1-3 0.00 11.43 18.62 2.13 16.94 1.06 0.65 0.00 1.12 0.21 0.12 0.05 10.61 0.01 37.03 – 5 5 – HD4203 2-3 0.00 12.48 16.02 1.67 18.51 1.15 0.71 0.00 1.23 0.23 0.13 0.05 10.57 0.00 37.23 HD4203 3-3 0.00 11.09 19.41 2.38 16.43 1.02 0.63 0.00 1.09 0.21 0.11 0.05 10.65 0.01 36.91 HD4203 4-3 0.00 11.03 20.08 1.36 16.34 1.02 0.63 0.00 1.08 0.21 0.11 0.05 10.83 0.01 37.23 t=2×106 years HD4203 1-3 0.00 11.00 20.63 0.46 16.29 1.01 0.63 0.44 1.08 0.20 0.11 0.04 10.80 0.01 37.28 Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD4203 2-3 0.00 11.00 20.43 0.74 16.29 1.01 0.63 0.29 1.08 0.20 0.11 0.05 10.80 0.01 37.35 HD4203 3-3 0.00 11.00 20.59 0.55 16.27 1.01 0.63 0.49 1.08 0.20 0.11 0.05 10.80 0.01 37.21 HD4203 4-3 0.00 11.06 21.00 0.00 16.36 1.02 0.63 0.51 1.09 0.21 0.11 0.03 10.85 0.00 37.13 t=2.5×106 years HD4203 1-3 0.00 11.05 21.06 0.66 16.35 1.02 0.63 0.51 1.09 0.21 0.11 0.04 10.84 0.00 36.43 HD4203 2-3 0.00 11.06 21.01 0.00 16.37 1.02 0.63 0.51 1.09 0.21 0.11 0.03 10.86 0.00 37.10 HD4203 3-3 0.00 11.02 21.05 0.78 16.31 1.02 0.63 0.51 1.08 0.20 0.11 0.05 10.82 0.00 36.41 HD4203 4-3 0.00 11.10 21.22 1.90 16.43 1.02 0.63 0.51 1.09 0.21 0.11 0.05 10.90 0.00 34.82 t=3×106 years HD4203 1-3 0.00 11.21 21.44 2.12 16.60 1.03 0.64 0.52 1.10 0.21 0.12 0.05 11.01 0.00 33.95 HD4203 2-3 0.00 11.06 21.14 1.66 16.37 1.02 0.63 0.51 1.09 0.21 0.11 0.05 10.86 0.00 35.30 HD4203 3-3 0.00 11.19 21.40 2.19 16.57 1.03 0.64 0.52 1.10 0.21 0.12 0.05 10.99 0.00 34.00 – 5 6 – Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD4203 4-3 0.00 11.72 22.61 2.79 17.21 1.07 0.66 0.54 1.14 0.22 0.12 0.05 11.70 0.00 30.18 HD17051 t=2.5×105 years HD17051 1-3 0.00 3.00 41.33 0.00 0.00 22.04 24.11 0.00 0.00 0.00 0.00 1.51 8.01 0.00 0.00 HD17051 1-4 0.00 0.00 42.22 0.00 0.00 25.66 22.86 0.00 0.00 0.00 0.00 1.77 7.49 0.00 0.00 HD17051 2-3 0.00 0.00 47.02 0.00 0.00 46.02 3.52 0.00 0.00 0.00 0.00 3.13 0.32 0.00 0.00 HD17051 2-4 0.00 0.00 42.34 0.00 0.00 26.16 22.39 0.00 0.00 0.00 0.00 1.80 7.32 0.00 0.00 – 5 7 – HD17051 3-3 0.00 0.00 46.00 0.00 0.00 41.69 7.62 0.00 0.00 0.00 0.00 2.84 1.84 0.00 0.00 HD17051 4-3 0.00 0.00 44.45 0.00 0.00 35.12 13.87 0.00 0.00 0.00 0.00 2.40 4.16 0.00 0.00 HD17051 4-4 0.00 0.00 42.11 0.00 0.00 25.20 23.30 0.00 0.00 0.00 0.00 1.74 7.65 0.00 0.00 t=5×105 years HD17051 1-3 0.00 0.00 40.61 0.00 0.00 18.85 29.24 0.00 0.00 0.00 0.00 1.49 9.81 0.00 0.00 HD17051 1-4 0.00 11.52 22.22 0.00 48.02 1.98 2.18 0.00 2.59 0.56 0.22 0.14 10.53 0.04 0.00 HD17051 2-3 0.00 0.00 41.80 0.00 0.00 23.70 24.71 0.00 0.00 0.00 0.00 1.49 8.30 0.00 0.00 Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD17051 2-4 0.00 0.00 41.76 0.00 0.00 23.55 24.85 0.00 0.00 0.00 0.00 1.50 8.34 0.00 0.00 HD17051 3-3 0.00 0.00 41.83 0.00 0.00 23.62 24.77 0.00 0.00 0.00 0.00 1.33 8.45 0.00 0.00 HD17051 4-3 0.00 0.00 41.85 0.00 0.00 23.78 24.62 0.00 0.00 0.00 0.00 1.42 8.33 0.00 0.00 HD17051 4-4 0.00 10.43 20.09 0.00 51.86 2.37 2.60 0.00 2.95 0.48 0.25 0.13 8.81 0.02 0.00 t=1×106 years HD17051 1-3 0.00 16.98 22.80 1.72 35.78 1.47 1.62 0.00 1.92 0.42 0.17 0.10 14.56 0.00 2.46 HD17051 1-4 0.00 14.01 28.38 0.00 29.37 1.21 1.33 0.83 1.58 0.34 0.14 0.05 15.12 0.00 7.64 – 5 8 – HD17051 2-3 0.00 10.46 19.86 0.00 52.27 2.40 2.64 0.00 2.98 0.48 0.25 0.12 8.53 0.01 0.00 HD17051 2-4 0.00 13.66 25.62 1.73 33.97 1.40 1.54 0.00 1.83 0.40 0.16 0.10 13.72 0.02 5.86 HD17051 3-3 0.00 10.86 20.68 0.00 50.78 2.25 2.48 0.00 2.85 0.50 0.24 0.13 9.21 0.02 0.00 HD17051 4-3 0.00 12.43 22.34 0.81 43.61 1.93 2.36 0.00 2.44 0.45 0.20 0.12 11.35 0.02 1.95 HD17051 4-4 0.00 13.93 27.87 0.56 29.22 1.20 1.32 0.65 1.57 0.34 0.13 0.07 15.04 0.01 8.08 t=1.5×106 years HD17051 1-3 0.00 13.92 28.19 0.00 29.20 1.20 1.32 0.81 1.57 0.34 0.13 0.05 15.02 0.00 8.24 Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD17051 1-4 0.00 13.50 27.57 3.15 28.29 1.16 1.28 0.81 1.52 0.33 0.13 0.08 14.57 0.00 7.60 HD17051 2-3 0.00 14.68 26.11 2.13 30.78 1.27 1.39 0.00 1.66 0.36 0.14 0.09 14.48 0.01 6.90 HD17051 2-4 0.00 14.02 28.22 0.35 29.42 1.21 1.33 0.65 1.58 0.34 0.14 0.08 15.13 0.01 7.52 HD17051 3-3 0.00 14.43 26.49 1.93 30.27 1.25 1.37 0.00 1.63 0.35 0.14 0.09 14.69 0.01 7.35 HD17051 4-3 0.00 14.35 27.02 1.14 30.09 1.24 1.36 0.30 1.62 0.35 0.14 0.08 14.82 0.01 7.49 HD17051 4-4 0.00 14.59 29.78 2.04 30.53 1.26 1.38 0.87 1.64 0.36 0.14 0.09 15.74 0.00 1.59 t=2×106 years HD17051 1-3 0.00 14.43 29.48 0.66 30.27 1.24 1.37 0.86 1.63 0.35 0.14 0.09 15.58 0.00 3.90 HD17051 1-4 0.00 14.03 28.84 4.75 29.41 1.21 1.33 0.84 1.58 0.34 0.14 0.08 15.15 0.00 2.30 HD17051 2-3 0.00 13.84 27.68 0.82 29.01 1.19 1.64 0.65 1.56 0.34 0.13 0.07 14.93 0.01 8.12 HD17051 2-4 0.00 14.51 29.59 1.78 30.39 1.25 1.38 0.87 1.64 0.36 0.14 0.09 15.66 0.00 2.35 HD17051 3-3 0.00 13.90 27.94 0.53 29.15 1.20 1.53 0.71 1.57 0.34 0.13 0.07 15.00 0.01 7.91 HD17051 4-3 0.00 14.14 28.61 0.54 29.64 1.22 1.34 0.75 1.60 0.35 0.14 0.07 15.26 0.00 6.34 – 5 9 – Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD17051 4-4 0.00 13.71 28.08 4.23 28.71 1.18 1.30 0.82 1.54 0.34 0.13 0.08 14.79 0.00 5.10 t=2.5×106 years HD17051 1-3 0.00 14.51 29.90 3.57 30.40 1.25 1.65 0.87 1.63 0.36 0.14 0.09 15.64 0.00 0.00 HD17051 1-4 0.00 13.85 31.78 4.99 29.00 1.19 1.31 0.83 1.56 0.34 0.13 0.08 14.94 0.00 0.00 HD17051 2-3 0.00 14.19 28.92 0.00 29.75 1.22 1.35 0.85 1.60 0.35 0.14 0.08 15.32 0.00 6.24 HD17051 2-4 0.00 14.31 29.41 3.59 29.80 1.23 1.50 0.85 1.61 0.35 0.14 0.09 15.44 0.00 1.68 HD17051 3-3 0.00 14.29 29.16 0.41 29.98 1.23 1.36 0.85 1.61 0.35 0.14 0.09 15.44 0.00 5.08 HD17051 4-3 0.00 14.43 29.47 1.62 30.24 1.24 1.37 0.86 1.63 0.35 0.14 0.09 15.57 0.00 2.99 HD17051 4-4 0.00 14.08 30.66 4.97 29.50 1.21 1.34 0.84 1.59 0.35 0.14 0.08 15.20 0.00 0.05 t=3×106 years HD17051 1-3 0.00 13.70 28.24 4.38 28.76 1.18 1.56 0.82 1.54 0.34 0.13 0.08 14.77 0.00 4.50 HD17051 1-4 0.00 13.84 31.77 5.02 28.99 1.19 1.31 0.83 1.56 0.34 0.13 0.08 14.93 0.00 0.00 HD17051 2-3 0.00 14.64 29.89 1.99 30.65 1.26 1.39 0.87 1.65 0.36 0.14 0.09 15.80 0.00 1.28 HD17051 2-4 0.00 14.19 29.18 4.57 29.76 1.22 1.35 0.85 1.60 0.35 0.14 0.08 15.31 0.00 1.40 – 6 0 – Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD17051 3-3 0.00 14.62 29.88 2.51 30.62 1.26 1.39 0.87 1.65 0.36 0.14 0.09 15.77 0.00 0.83 HD17051 4-3 0.00 14.40 29.49 3.28 30.19 1.24 1.37 0.86 1.62 0.35 0.14 0.09 15.54 0.00 1.43 HD17051 4-4 0.00 13.85 31.78 5.00 28.99 1.19 1.31 0.83 1.56 0.34 0.13 0.08 14.93 0.00 0.00 HD19994 t=2.5×105 years HD19994 1-3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.21 36.61 0.00 63.18 – 6 1 – HD19994 2-3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.21 36.62 0.00 63.17 HD19994 3-3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.21 36.64 0.00 63.15 HD19994 3-4 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.21 36.59 0.00 63.20 HD19994 3-5 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.21 36.67 0.00 63.13 HD19994 4-3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.21 36.52 0.00 63.27 HD19994 4-4 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.21 36.68 0.00 63.11 t=5×105 years HD19994 1-3 0.00 0.00 0.00 0.00 3.49 0.00 0.00 0.00 0.27 0.02 0.11 0.20 35.25 0.00 60.66 Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD19994 2-3 0.00 0.00 0.00 0.00 4.05 0.00 0.00 0.00 0.36 0.02 0.11 0.20 35.00 0.00 60.27 HD19994 3-3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.21 36.68 0.00 63.11 HD19994 3-4 0.00 0.00 0.00 0.88 22.83 1.34 1.10 0.00 1.57 0.28 0.15 0.14 26.50 0.70 44.50 HD19994 3-5 0.00 10.85 20.26 1.81 20.82 1.21 0.99 0.00 1.41 0.28 0.14 0.05 11.02 0.00 31.16 HD19994 4-3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.21 36.68 0.00 63.11 HD19994 4-4 0.00 0.00 0.00 0.41 15.58 0.55 0.51 0.00 1.08 0.18 0.13 0.17 29.93 0.29 51.17 – 6 2 – t=1×106 years HD19994 1-3 0.00 4.06 7.42 1.11 29.28 1.48 1.17 0.01 2.07 0.35 0.21 0.09 18.10 0.36 34.29 HD19994 2-3 0.00 4.52 8.92 1.05 25.31 1.31 1.02 0.03 1.81 0.30 0.23 0.10 18.81 0.33 36.25 HD19994 3-3 0.00 4.78 7.26 1.47 29.66 1.65 1.33 0.00 2.08 0.37 0.23 0.09 18.57 0.43 32.08 HD19994 3-4 0.00 6.06 12.79 0.63 28.18 1.64 1.34 0.57 1.92 0.38 0.18 0.07 14.84 0.51 30.89 HD19994 3-5 0.00 10.72 22.87 0.14 20.53 1.19 0.97 1.02 1.39 0.28 0.13 0.05 12.12 0.00 28.59 HD19994 4-3 0.00 0.26 0.41 0.78 28.40 1.06 0.81 0.00 2.16 0.28 0.31 0.13 24.22 0.37 40.81 HD19994 4-4 0.00 9.05 16.44 1.04 24.60 1.51 1.17 0.33 1.67 0.33 0.16 0.06 14.02 0.23 29.39 Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD19994 1-3 0.00 9.89 20.15 0.79 21.49 1.30 0.93 0.40 1.46 0.29 0.14 0.05 12.02 0.01 31.09 t=1.5×106 years HD19994 2-3 0.00 10.21 19.89 0.87 21.55 1.36 1.02 0.51 1.46 0.29 0.14 0.05 12.24 0.00 30.40 HD19994 3-3 0.00 10.53 20.77 0.79 20.53 1.51 0.83 0.56 1.39 0.28 0.13 0.05 11.56 0.01 31.05 HD19994 3-4 0.00 10.75 22.60 1.11 20.49 1.25 0.97 0.66 1.39 0.28 0.13 0.05 12.16 0.01 28.16 HD19994 3-5 0.00 10.42 22.21 2.81 19.89 1.16 0.94 0.99 1.35 0.27 0.13 0.05 11.76 0.00 28.03 – 6 3 – HD19994 4-3 0.00 9.72 17.77 1.65 22.90 1.53 1.09 0.17 1.55 0.31 0.15 0.06 12.47 0.01 30.63 HD19994 4-4 0.00 10.34 21.79 0.81 20.04 1.30 0.85 0.82 1.36 0.27 0.13 0.05 11.62 0.00 30.62 t=2×106 years HD19994 1-3 0.00 10.55 22.27 0.60 20.09 1.27 0.95 0.89 1.36 0.27 0.13 0.05 11.86 0.00 29.71 HD19994 2-3 0.00 10.67 22.49 0.92 20.37 1.25 0.97 0.82 1.38 0.27 0.13 0.05 12.02 0.00 28.65 HD19994 3-3 0.00 10.56 22.34 0.22 20.20 1.19 0.96 0.90 1.37 0.27 0.13 0.05 11.93 0.00 29.89 HD19994 3-4 0.00 10.40 22.10 1.60 19.89 1.16 0.94 0.99 1.35 0.27 0.13 0.04 11.75 0.00 29.39 HD19994 3-5 0.00 14.06 32.41 4.29 26.79 1.56 1.27 1.33 1.82 0.36 0.17 0.07 15.87 0.00 0.00 Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD19994 4-3 0.00 10.50 21.89 0.52 19.89 1.38 0.94 0.64 1.35 0.27 0.13 0.05 11.75 0.01 30.69 HD19994 4-4 0.00 10.61 22.63 1.31 20.35 1.18 0.96 1.01 1.38 0.27 0.13 0.05 11.97 0.00 28.13 t=2.5×106 years HD19994 1-3 0.00 10.80 23.00 1.16 20.70 1.20 0.98 1.03 1.40 0.28 0.13 0.05 12.19 0.00 27.06 HD19994 2-3 0.00 10.66 22.69 1.30 20.42 1.19 0.97 1.01 1.39 0.27 0.13 0.05 12.03 0.00 27.89 – 6 4 – HD19994 3-3 0.00 11.07 23.59 1.13 21.18 1.23 1.00 1.05 1.44 0.29 0.14 0.05 12.51 0.00 25.34 HD19994 3-4 0.00 11.73 25.18 2.22 22.41 1.30 1.06 1.11 1.52 0.30 0.15 0.06 13.25 0.00 19.71 HD19994 3-5 0.00 14.05 32.40 4.32 26.78 1.56 1.27 1.33 1.82 0.36 0.17 0.07 15.86 0.00 0.00 HD19994 4-3 0.00 10.59 22.48 0.25 20.26 1.18 0.96 1.00 1.37 0.27 0.13 0.05 11.96 0.00 29.49 HD19994 4-4 0.00 11.86 25.96 2.58 22.67 1.32 1.08 1.12 1.54 0.31 0.15 0.06 13.39 0.00 17.99 t=3×106 years HD19994 1-3 0.00 11.33 24.41 2.14 21.74 1.26 1.03 1.08 1.48 0.29 0.14 0.05 12.75 0.00 22.29 HD19994 2-3 0.00 11.50 24.80 2.19 22.05 1.28 1.05 1.09 1.50 0.30 0.14 0.05 12.97 0.00 21.06 Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD19994 3-3 0.00 11.00 23.49 2.12 21.09 1.23 1.00 1.04 1.43 0.28 0.14 0.05 12.41 0.00 24.72 HD19994 3-4 0.00 12.91 28.96 3.36 24.69 1.43 1.17 1.22 1.67 0.33 0.16 0.06 14.59 0.00 9.44 HD19994 3-5 0.00 14.04 32.37 4.35 26.81 1.56 1.27 1.33 1.82 0.36 0.17 0.07 15.85 0.00 0.00 HD19994 4-3 0.00 11.13 23.78 0.98 21.31 1.24 1.01 1.06 1.45 0.29 0.14 0.05 12.58 0.00 25.00 HD19994 4-4 0.00 12.05 26.65 3.30 23.08 1.34 1.10 1.14 1.57 0.31 0.15 0.06 13.57 0.00 15.68 HD27442 t=2.5×105 years – 6 5 – HD27442 1-3 0.00 9.56 42.97 0.00 0.00 29.22 12.88 0.00 0.00 0.00 0.00 1.21 4.16 0.00 0.00 HD27442 1-4 0.00 9.02 43.25 0.00 0.00 29.02 12.66 0.00 0.00 0.00 0.00 1.23 4.82 0.00 0.00 HD27442 1-5 0.00 9.46 42.94 0.00 0.00 29.10 13.05 0.00 0.00 0.00 0.00 1.24 4.21 0.00 0.00 HD27442 2-3 0.00 13.86 43.62 0.00 0.00 23.36 10.33 0.00 0.00 0.00 0.00 0.97 7.85 0.00 0.00 HD27442 2-4 0.00 9.89 43.08 0.00 0.00 28.32 12.66 0.00 0.00 0.00 0.00 1.20 4.86 0.00 0.00 HD27442 3-3 0.00 7.64 43.34 0.00 0.00 30.69 12.90 0.00 0.00 0.00 0.00 1.29 4.14 0.00 0.00 HD27442 3-4 0.00 8.89 43.40 0.00 0.00 28.84 12.44 0.00 0.00 0.00 0.00 1.22 5.21 0.00 0.00 HD27442 3-5 0.00 28.40 45.18 0.00 0.00 6.02 2.70 0.00 0.00 0.00 0.00 0.26 17.45 0.00 0.00 Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD27442 4-3 0.00 9.54 42.96 0.00 0.00 29.19 12.92 0.00 0.00 0.00 0.00 1.22 4.17 0.00 0.00 HD27442 4-4 0.00 9.46 42.94 0.00 0.00 29.10 13.05 0.00 0.00 0.00 0.00 1.24 4.21 0.00 0.00 HD27442 4-5 0.00 9.61 42.78 0.00 2.66 25.78 10.68 0.00 0.16 0.03 0.02 1.10 7.19 0.00 0.00 t=5×105 years HD27442 1-3 0.00 12.21 43.57 0.00 0.00 24.67 11.06 0.00 0.00 0.00 0.00 1.05 7.44 0.00 0.00 HD27442 1-4 0.00 19.05 39.82 0.00 11.17 9.63 4.32 0.00 0.79 0.08 0.05 0.41 14.67 0.00 0.00 HD27442 1-5 0.00 18.78 34.93 0.00 23.78 1.76 0.79 0.00 1.49 0.17 0.12 0.08 18.10 0.00 0.00 HD27442 2-3 0.00 16.19 40.29 0.00 9.30 14.18 6.36 0.00 0.60 0.08 0.06 0.60 12.33 0.00 0.00 HD27442 2-4 0.00 17.39 40.30 0.00 9.62 12.48 5.60 0.00 0.66 0.07 0.04 0.53 13.30 0.00 0.00 – 6 6 – HD27442 3-3 0.00 14.57 40.66 0.00 7.53 17.67 7.93 0.00 0.54 0.05 0.04 0.75 10.26 0.00 0.00 HD27442 3-4 0.00 19.05 39.48 0.00 12.60 8.06 3.61 0.00 0.82 0.10 0.07 0.34 15.89 0.00 0.00 HD27442 3-5 0.00 18.29 34.56 0.00 24.75 1.65 0.74 0.00 1.46 0.24 0.19 0.07 18.05 0.00 0.00 HD27442 4-3 0.00 13.27 43.43 0.00 0.63 23.00 10.31 0.00 0.07 0.00 0.00 0.98 8.31 0.00 0.00 HD27442 4-4 0.00 18.32 34.47 0.00 24.77 1.65 0.74 0.00 1.46 0.24 0.19 0.07 18.08 0.00 0.00 HD27442 4-5 0.00 19.15 37.54 0.00 17.22 5.38 2.41 0.08 1.08 0.14 0.10 0.23 16.65 0.00 0.00 t=1×106 years Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD27442 1-3 0.00 19.47 35.20 0.00 22.91 1.90 0.85 0.00 1.47 0.19 0.13 0.08 17.80 0.00 0.00 HD27442 1-4 0.00 18.16 34.40 0.26 24.56 1.64 0.74 0.39 1.45 0.24 0.18 0.07 17.91 0.00 0.00 HD27442 1-5 0.00 18.10 34.47 0.00 24.51 1.63 0.73 0.76 1.45 0.24 0.18 0.07 17.86 0.00 0.00 HD27442 2-3 0.00 18.34 34.39 1.24 23.65 1.71 0.77 0.25 1.46 0.20 0.14 0.07 17.78 0.00 0.00 HD27442 2-4 0.00 18.06 34.22 0.90 24.38 1.63 0.73 0.35 1.44 0.24 0.18 0.07 17.81 0.00 0.00 HD27442 3-3 0.00 18.55 34.76 0.00 24.05 1.73 0.77 0.22 1.47 0.21 0.15 0.07 18.02 0.00 0.00 HD27442 3-4 0.00 18.02 34.20 0.92 24.37 1.63 0.73 0.44 1.44 0.24 0.18 0.07 17.77 0.00 0.00 HD27442 3-5 0.00 16.90 32.40 6.47 22.84 1.52 0.68 0.71 1.35 0.22 0.17 0.06 16.67 0.00 0.00 – 6 7 – HD27442 4-3 0.00 18.56 34.72 0.00 24.22 1.72 0.77 0.00 1.48 0.21 0.15 0.07 18.08 0.00 0.00 HD27442 4-4 0.00 17.23 33.04 4.60 23.30 1.56 0.70 0.73 1.38 0.23 0.17 0.07 17.00 0.00 0.00 HD27442 4-5 0.00 17.86 34.00 1.50 24.16 1.61 0.72 0.61 1.43 0.24 0.18 0.07 17.62 0.00 0.00 t=1.5×106 years HD27442 1-3 0.00 18.24 34.51 0.00 24.67 1.65 0.74 0.25 1.46 0.24 0.18 0.07 17.99 0.00 0.00 HD27442 1-4 0.00 17.49 33.41 3.28 23.67 1.58 0.71 0.74 1.40 0.23 0.18 0.07 17.26 0.00 0.00 HD27442 1-5 0.00 17.01 32.63 5.77 23.03 1.54 0.69 0.72 1.36 0.23 0.17 0.07 16.79 0.00 0.00 Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD27442 2-3 0.00 17.74 33.71 2.45 23.99 1.60 0.72 0.40 1.42 0.23 0.18 0.07 17.50 0.00 0.00 HD27442 2-4 0.00 17.56 33.50 3.02 23.75 1.58 0.71 0.68 1.40 0.23 0.18 0.07 17.32 0.00 0.00 HD27442 3-3 0.00 17.85 33.93 1.74 24.14 1.61 0.72 0.49 1.43 0.24 0.18 0.07 17.61 0.00 0.00 HD27442 3-4 0.00 17.39 33.22 3.88 23.53 1.57 0.70 0.70 1.39 0.23 0.18 0.07 17.15 0.00 0.00 HD27442 3-5 0.00 16.65 31.95 7.82 22.51 1.50 0.67 0.70 1.33 0.22 0.17 0.06 16.42 0.00 0.00 HD27442 4-3 0.00 18.16 34.44 0.19 24.58 1.64 0.74 0.39 1.45 0.24 0.18 0.07 17.92 0.00 0.00 HD27442 4-4 0.00 16.71 32.05 7.46 22.62 1.51 0.68 0.71 1.34 0.22 0.17 0.06 16.48 0.00 0.00 HD27442 4-5 0.00 17.08 32.90 5.27 23.10 1.54 0.69 0.72 1.36 0.23 0.17 0.07 16.85 0.00 0.00 t=2×106 years HD27442 1-3 0.00 17.93 34.17 0.95 24.25 1.62 0.73 0.75 1.43 0.24 0.18 0.07 17.69 0.00 0.00 HD27442 1-4 0.00 16.94 32.47 6.21 22.92 1.53 0.69 0.72 1.35 0.22 0.17 0.07 16.71 0.00 0.00 HD27442 1-5 0.00 16.72 32.07 7.39 22.63 1.51 0.68 0.71 1.34 0.22 0.17 0.06 16.50 0.00 0.00 HD27442 2-3 0.00 17.47 33.73 3.06 23.62 1.58 0.71 0.73 1.40 0.23 0.18 0.07 17.23 0.00 0.00 HD27442 2-4 0.00 17.01 32.71 5.74 23.00 1.53 0.69 0.72 1.36 0.22 0.17 0.07 16.77 0.00 0.00 HD27442 3-3 0.00 17.42 33.29 3.68 23.56 1.57 0.70 0.73 1.39 0.23 0.18 0.07 17.18 0.00 0.00 – 6 8 – Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD27442 3-4 0.00 16.95 32.57 6.08 22.93 1.53 0.69 0.72 1.35 0.22 0.17 0.07 16.72 0.00 0.00 HD27442 3-5 0.00 16.29 33.27 7.79 22.02 1.47 0.66 0.69 1.30 0.22 0.17 0.06 16.07 0.00 0.00 HD27442 4-3 0.00 17.74 33.85 1.94 23.99 1.60 0.72 0.75 1.42 0.23 0.18 0.07 17.50 0.00 0.00 HD27442 4-4 0.00 16.58 32.10 7.89 22.43 1.50 0.67 0.70 1.32 0.22 0.17 0.06 16.35 0.00 0.00 HD27442 4-5 0.00 16.68 32.40 7.23 22.56 1.51 0.67 0.71 1.33 0.22 0.17 0.06 16.45 0.00 0.00 t=2.5×106 years HD27442 1-3 0.00 17.35 33.17 4.02 23.48 1.57 0.70 0.73 1.39 0.23 0.18 0.07 17.12 0.00 0.00 HD27442 1-4 0.00 16.68 32.25 7.37 22.57 1.51 0.67 0.71 1.33 0.22 0.17 0.06 16.46 0.00 0.00 HD27442 1-5 0.00 16.63 31.95 7.85 22.50 1.50 0.67 0.70 1.33 0.22 0.17 0.06 16.40 0.00 0.00 – 6 9 – HD27442 2-3 0.00 17.04 33.07 5.25 23.05 1.54 0.69 0.72 1.36 0.23 0.17 0.07 16.81 0.00 0.00 HD27442 2-4 0.00 16.71 32.42 7.11 22.60 1.51 0.68 0.71 1.34 0.22 0.17 0.06 16.48 0.00 0.00 HD27442 3-3 0.00 17.03 32.70 5.66 23.04 1.54 0.69 0.72 1.36 0.22 0.17 0.07 16.80 0.00 0.00 HD27442 3-4 0.00 16.65 32.43 7.31 22.52 1.50 0.67 0.70 1.33 0.22 0.17 0.06 16.42 0.00 0.00 HD27442 3-5 0.00 16.23 33.45 7.78 21.98 1.47 0.66 0.69 1.30 0.21 0.16 0.06 16.01 0.00 0.00 HD27442 4-3 0.00 17.17 32.84 5.02 23.22 1.55 0.69 0.73 1.37 0.23 0.17 0.07 16.94 0.00 0.00 Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD27442 4-4 0.00 16.28 33.36 7.79 21.96 1.47 0.66 0.69 1.30 0.22 0.16 0.06 16.05 0.00 0.00 HD27442 4-5 0.00 16.54 32.41 7.72 22.38 1.49 0.67 0.70 1.32 0.22 0.17 0.06 16.32 0.00 0.00 t=3×106 years HD27442 1-3 0.00 16.93 32.46 6.29 22.88 1.53 0.68 0.72 1.35 0.22 0.17 0.07 16.70 0.00 0.00 HD27442 1-4 0.00 16.55 32.42 7.70 22.38 1.49 0.67 0.70 1.32 0.22 0.17 0.06 16.32 0.00 0.00 HD27442 1-5 0.00 16.38 32.87 7.84 22.17 1.48 0.66 0.69 1.31 0.22 0.17 0.06 16.15 0.00 0.00 HD27442 2-3 0.00 16.70 32.77 6.78 22.59 1.51 0.68 0.71 1.34 0.22 0.17 0.06 16.48 0.00 0.00 HD27442 2-4 0.00 16.58 32.41 7.60 22.42 1.50 0.67 0.70 1.32 0.22 0.17 0.06 16.35 0.00 0.00 – 7 0 – HD27442 3-3 0.00 16.72 32.52 6.98 22.60 1.51 0.68 0.71 1.34 0.22 0.17 0.06 16.49 0.00 0.00 HD27442 3-4 0.00 16.51 32.58 7.66 22.34 1.49 0.67 0.70 1.32 0.22 0.17 0.06 16.29 0.00 0.00 HD27442 3-5 0.00 16.24 33.48 7.78 21.94 1.47 0.66 0.69 1.30 0.21 0.16 0.06 16.01 0.00 0.00 HD27442 4-3 0.00 16.84 32.31 6.73 22.78 1.52 0.68 0.71 1.35 0.22 0.17 0.06 16.62 0.00 0.00 HD27442 4-4 0.00 16.24 33.46 7.78 21.95 1.46 0.66 0.69 1.30 0.21 0.16 0.06 16.02 0.00 0.00 HD27442 4-5 0.00 16.41 32.81 7.81 22.19 1.48 0.66 0.69 1.31 0.22 0.17 0.06 16.19 0.00 0.00 HD72659 t=2.5×105 years Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD72659 1-3 0.00 5.36 41.06 0.00 4.61 20.79 16.65 0.07 0.30 0.06 0.02 1.28 9.81 0.00 0.00 HD72659 1-4 0.00 17.89 35.13 0.09 22.28 2.79 2.43 0.37 1.36 0.29 0.13 0.16 17.07 0.00 0.00 HD726592-3 0.00 1.76 42.51 0.00 0.00 26.96 20.40 0.00 0.00 0.00 0.00 1.73 6.64 0.00 0.00 HD726592-4 0.00 9.32 41.28 0.00 5.46 15.86 13.67 0.04 0.36 0.06 0.03 0.90 13.01 0.00 0.00 HD726592-5 0.00 16.68 33.80 0.73 24.46 2.79 2.44 0.55 1.48 0.33 0.15 0.14 16.46 0.00 0.00 HD72659 3-3 0.00 6.11 40.13 0.24 7.11 19.61 14.79 0.13 0.43 0.10 0.04 1.23 10.08 0.00 0.00 HD72659 3-4 0.00 12.68 39.74 0.00 10.00 11.45 9.97 0.00 0.65 0.11 0.05 0.64 14.70 0.00 0.00 HD72659 3-5 0.00 13.57 35.93 0.00 20.08 6.62 5.43 0.29 1.22 0.27 0.12 0.28 16.20 0.00 0.00 – 7 1 – HD72659 4-3 0.00 7.09 41.67 0.00 3.73 18.49 15.98 0.00 0.23 0.05 0.02 1.07 11.67 0.00 0.00 HD72659 4-4 0.00 0.91 43.14 0.00 0.00 27.83 19.03 0.00 0.00 0.00 0.00 1.75 7.34 0.00 0.00 HD72659 4-5 0.00 18.10 33.51 0.00 26.03 1.43 1.31 0.44 1.61 0.33 0.15 0.10 16.99 0.00 0.00 t=5×105 years HD72659 1-3 0.00 11.47 38.56 0.35 11.80 11.96 10.39 0.16 0.74 0.15 0.07 0.69 13.66 0.00 0.00 HD72659 1-4 0.00 17.56 32.99 1.40 25.99 1.38 1.27 0.76 1.57 0.35 0.16 0.09 16.48 0.00 0.00 HD726592-3 0.00 5.46 42.61 0.00 0.00 21.14 19.43 0.00 0.00 0.00 0.00 1.31 10.05 0.00 0.00 Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD726592-4 0.00 15.82 35.73 0.04 20.59 5.34 4.02 0.25 1.29 0.26 0.12 0.26 16.28 0.00 0.00 HD726592-5 0.00 17.19 33.38 2.40 25.47 1.29 1.24 0.77 1.54 0.34 0.15 0.09 16.13 0.00 0.00 HD72659 3-3 0.00 9.63 39.34 0.54 9.73 14.04 12.14 0.23 0.59 0.13 0.06 0.81 12.76 0.00 0.00 HD72659 3-4 0.00 18.39 33.81 0.00 25.33 1.47 1.35 0.41 1.57 0.33 0.15 0.10 17.10 0.00 0.00 HD72659 3-5 0.00 17.80 33.12 0.50 26.34 1.40 1.28 0.65 1.60 0.36 0.16 0.09 16.71 0.00 0.00 HD72659 4-3 0.00 14.10 37.45 0.00 15.99 8.11 6.91 0.16 0.99 0.21 0.09 0.41 15.56 0.00 0.00 HD72659 4-4 0.00 7.63 39.39 0.00 7.58 17.73 15.94 0.00 0.46 0.10 0.04 1.19 9.94 0.00 0.00 HD72659 4-5 0.00 17.48 32.68 1.85 25.91 1.38 1.26 0.85 1.57 0.35 0.15 0.09 16.43 0.00 0.00 t=1×106 years HD72659 1-3 0.00 18.19 34.52 0.70 23.59 2.14 1.64 0.44 1.48 0.30 0.13 0.13 16.74 0.00 0.00 HD72659 1-4 0.00 16.87 33.37 3.48 25.03 1.29 1.22 0.82 1.51 0.34 0.15 0.09 15.84 0.00 0.00 – 7 2 – HD726592-3 0.00 17.68 36.48 0.00 19.04 4.28 3.71 0.06 1.23 0.23 0.10 0.21 16.97 0.00 0.00 HD726592-4 0.00 17.69 33.16 0.89 26.10 1.39 1.28 0.71 1.58 0.35 0.15 0.09 16.60 0.00 0.00 HD726592-5 0.10 16.36 34.87 3.82 24.28 1.20 1.18 0.79 1.47 0.33 0.14 0.09 15.38 0.00 0.00 HD72659 3-3 0.00 17.29 34.76 1.13 23.08 2.81 2.10 0.35 1.45 0.29 0.13 0.15 16.49 0.00 0.00 Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD72659 3-4 0.00 17.57 32.85 1.40 26.01 1.38 1.27 0.85 1.58 0.35 0.16 0.09 16.51 0.00 0.00 HD72659 3-5 0.00 17.07 33.00 3.13 25.27 1.34 1.23 0.83 1.53 0.34 0.15 0.09 16.02 0.00 0.00 HD72659 4-3 0.00 17.80 33.10 0.68 26.24 1.40 1.28 0.58 1.59 0.35 0.16 0.09 16.72 0.00 0.00 HD72659 4-4 0.00 17.56 37.58 0.00 16.76 5.59 3.78 0.25 1.12 0.19 0.08 0.25 16.84 0.00 0.00 HD72659 4-5 0.00 16.70 33.73 3.77 24.80 1.25 1.20 0.81 1.50 0.33 0.15 0.09 15.67 0.00 0.00 t=1.5×106 years HD72659 1-3 0.00 17.44 33.17 1.15 26.46 1.34 1.26 0.65 1.56 0.35 0.15 0.09 16.37 0.00 0.00 HD72659 1-4 0.00 16.51 34.22 4.06 24.49 1.18 1.19 0.80 1.48 0.33 0.15 0.09 15.50 0.00 0.00 – 7 3 – HD726592-3 0.00 17.96 33.24 0.00 26.58 1.41 1.29 0.43 1.61 0.36 0.16 0.10 16.86 0.00 0.00 HD726592-4 0.00 17.26 33.14 2.28 25.57 1.36 1.24 0.80 1.55 0.35 0.15 0.09 16.20 0.00 0.00 HD726592-5 0.14 16.04 35.26 4.07 24.38 1.14 1.15 0.78 1.44 0.32 0.14 0.09 15.05 0.00 0.00 HD72659 3-3 0.07 17.30 34.00 1.43 25.61 1.36 1.25 0.60 1.55 0.35 0.15 0.09 16.24 0.00 0.00 HD72659 3-4 0.00 17.12 32.95 2.99 25.37 1.33 1.23 0.83 1.54 0.34 0.15 0.09 16.07 0.00 0.00 HD72659 3-5 0.00 16.66 34.13 3.66 24.67 1.17 1.20 0.81 1.49 0.33 0.15 0.09 15.64 0.00 0.00 HD72659 4-3 0.00 17.46 32.97 1.76 25.85 1.37 1.26 0.78 1.57 0.35 0.15 0.09 16.38 0.00 0.00 Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD72659 4-4 0.00 17.88 33.10 0.46 26.49 1.41 1.29 0.37 1.60 0.36 0.16 0.09 16.79 0.00 0.00 HD72659 4-5 0.00 16.18 34.08 4.14 25.39 1.07 1.17 0.78 1.45 0.32 0.14 0.09 15.18 0.00 0.00 t=2×106 years HD72659 1-3 0.05 17.22 33.62 1.91 25.53 1.33 1.24 0.79 1.55 0.34 0.15 0.09 16.18 0.00 0.00 HD72659 1-4 0.09 16.17 35.41 4.13 23.94 1.14 1.17 0.78 1.45 0.32 0.14 0.09 15.17 0.00 0.00 HD726592-3 0.00 16.97 33.26 3.27 25.14 1.30 1.22 0.82 1.52 0.34 0.15 0.09 15.92 0.00 0.00 HD726592-4 0.00 16.97 33.26 3.27 25.14 1.30 1.22 0.82 1.52 0.34 0.15 0.09 15.92 0.00 0.00 HD726592-5 1.26 14.29 41.61 3.68 21.20 1.02 1.03 0.69 1.28 0.29 0.13 0.08 13.43 0.00 0.00 – 7 4 – HD72659 3-3 0.09 17.12 34.32 1.66 25.34 1.29 1.23 0.75 1.53 0.34 0.15 0.09 16.07 0.00 0.00 HD72659 3-4 0.00 16.75 33.36 3.93 24.84 1.32 1.21 0.81 1.50 0.33 0.15 0.09 15.71 0.00 0.00 HD72659 3-5 0.00 16.45 34.63 4.12 24.33 1.02 1.18 0.80 1.47 0.33 0.15 0.09 15.43 0.00 0.00 HD72659 4-3 0.00 17.12 33.41 2.60 25.34 1.30 1.23 0.83 1.53 0.34 0.15 0.09 16.06 0.00 0.00 HD72659 4-4 0.00 17.67 32.92 1.09 26.17 1.39 1.27 0.70 1.58 0.35 0.16 0.09 16.60 0.00 0.00 HD72659 4-5 0.10 16.08 35.49 4.14 23.96 1.16 1.16 0.78 1.45 0.32 0.14 0.09 15.12 0.00 0.00 t=2.5×106 years HD72659 1-3 0.07 16.99 33.78 2.49 25.26 1.31 1.22 0.82 1.52 0.34 0.15 0.09 15.95 0.00 0.00 Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD72659 1-4 0.19 15.91 36.33 4.08 23.58 1.08 1.15 0.77 1.43 0.32 0.14 0.08 14.93 0.00 0.00 HD726592-3 0.00 17.72 33.10 0.64 26.21 1.39 1.28 0.86 1.59 0.35 0.16 0.09 16.62 0.00 0.00 HD726592-4 0.00 16.73 33.58 3.82 24.82 1.26 1.21 0.81 1.50 0.33 0.15 0.09 15.71 0.00 0.00 HD726592-5 1.39 14.12 42.28 3.63 20.92 0.98 1.02 0.68 1.27 0.28 0.12 0.07 13.25 0.00 0.00 HD72659 3-3 0.72 16.09 37.20 1.95 23.80 1.21 1.16 0.78 1.44 0.32 0.14 0.09 15.11 0.00 0.00 HD72659 3-4 0.00 16.53 34.03 4.17 24.54 1.18 1.19 0.80 1.48 0.33 0.15 0.09 15.51 0.00 0.00 HD72659 3-5 0.00 16.32 35.07 4.18 24.14 1.02 1.17 0.79 1.46 0.33 0.14 0.09 15.31 0.00 0.00 – 7 5 – HD72659 4-3 0.00 16.93 33.34 3.30 25.12 1.28 1.22 0.82 1.52 0.34 0.15 0.09 15.88 0.00 0.00 HD72659 4-4 0.00 17.43 33.29 1.18 26.11 1.37 1.25 0.84 1.56 0.35 0.15 0.09 16.38 0.00 0.00 HD72659 4-5 0.25 15.74 36.95 4.04 23.31 1.10 1.13 0.76 1.41 0.31 0.14 0.08 14.77 0.00 0.00 t=3×106 years HD72659 1-3 0.07 16.85 33.87 3.00 24.99 1.29 1.21 0.82 1.51 0.34 0.15 0.09 15.82 0.00 0.00 HD72659 1-4 0.91 14.85 39.97 3.81 21.99 0.99 1.03 0.72 1.33 0.30 0.13 0.08 13.91 0.00 0.00 HD726592-3 0.00 17.53 32.81 1.59 25.96 1.38 1.26 0.85 1.57 0.35 0.15 0.09 16.45 0.00 0.00 HD726592-4 0.00 16.72 34.20 4.04 24.06 1.22 1.20 0.81 1.50 0.33 0.15 0.09 15.69 0.00 0.00 Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD726592-5 2.21 12.91 46.29 3.31 19.10 0.92 0.93 0.62 1.16 0.26 0.11 0.07 12.11 0.00 0.00 HD72659 3-3 0.74 16.07 37.23 2.55 23.17 1.24 1.16 0.78 1.44 0.32 0.14 0.09 15.08 0.00 0.00 HD72659 3-4 0.00 16.37 34.76 4.19 24.23 1.09 1.18 0.79 1.47 0.33 0.14 0.09 15.36 0.00 0.00 HD72659 3-5 0.23 15.87 36.32 4.08 23.61 1.13 1.14 0.77 1.42 0.32 0.14 0.08 14.89 0.00 0.00 HD72659 4-3 0.00 16.74 33.69 3.75 24.81 1.23 1.21 0.81 1.50 0.33 0.15 0.09 15.71 0.00 0.00 HD72659 4-4 0.00 17.39 33.27 1.57 25.82 1.37 1.25 0.84 1.56 0.35 0.15 0.09 16.32 0.00 0.00 HD72659 4-5 0.38 15.55 37.64 3.99 23.00 1.05 1.12 0.75 1.39 0.31 0.14 0.08 14.59 0.00 0.00 – 7 6 – HD108874 t=2.5×105 years HD108874 1-4 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.30 28.73 0.00 70.98 HD108874 2-4 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.30 28.77 0.00 70.93 HD108874 3-4 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.30 28.69 0.00 71.01 HD108874 4-4 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.30 28.74 0.00 70.96 HD108874 4-5 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.29 28.77 0.00 70.94 Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD108874 1-4 0.00 0.00 0.00 0.00 0.65 0.00 0.00 0.00 0.00 0.00 0.12 0.29 28.62 0.00 70.32 t=5×105 years HD108874 2-4 0.00 0.00 0.00 0.00 6.03 0.05 0.00 0.00 0.40 0.03 0.12 0.27 26.90 0.02 66.18 HD108874 3-4 0.00 0.00 0.00 0.00 0.46 0.00 0.00 0.00 0.00 0.00 0.09 0.29 28.66 0.00 70.50 HD108874 4-4 0.00 0.00 0.00 0.00 0.22 0.00 0.00 0.00 0.00 0.00 0.04 0.29 28.77 0.00 70.68 HD108874 4-5 0.00 0.01 0.00 1.19 45.78 2.61 1.47 0.00 2.62 0.52 0.30 0.12 13.48 1.36 30.54 t=1×106 years HD108874 1-4 0.00 4.65 5.98 1.74 38.67 2.20 1.29 0.00 2.19 0.46 0.25 0.10 11.90 0.74 29.82 – 7 7 – HD108874 2-4 0.00 10.38 14.20 2.88 27.92 1.59 0.94 0.00 1.58 0.34 0.18 0.07 10.85 0.31 28.78 HD108874 3-4 0.00 3.26 3.20 2.28 40.73 2.29 1.18 0.00 2.34 0.46 0.27 0.11 13.15 0.97 29.76 HD108874 4-4 0.00 7.41 10.08 2.43 33.15 1.87 1.02 0.00 1.89 0.39 0.22 0.09 11.32 0.62 29.52 HD108874 4-5 0.00 13.03 21.80 0.33 21.20 1.20 0.71 0.44 1.20 0.26 0.14 0.03 10.42 0.01 29.24 t=1.5×106 years Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD108874 1-4 0.00 13.01 20.64 2.24 21.17 1.20 0.71 0.15 1.19 0.26 0.13 0.05 10.35 0.01 28.88 HD108874 2-4 0.00 13.06 21.50 1.26 21.26 1.21 0.71 0.38 1.20 0.26 0.14 0.04 10.44 0.01 28.53 HD108874 3-4 0.00 13.22 20.29 2.30 21.51 1.22 0.72 0.11 1.21 0.26 0.14 0.06 9.94 0.01 29.00 HD108874 4-4 0.00 13.14 20.97 1.54 21.38 1.21 0.72 0.27 1.21 0.26 0.14 0.03 10.21 0.01 28.92 HD108874 4-5 0.00 12.93 22.00 3.37 21.03 1.20 0.71 0.50 1.19 0.26 0.13 0.06 10.34 0.00 26.30 – 7 8 – t=2×106 years HD108874 1-4 0.00 13.11 22.00 0.88 21.43 1.22 0.71 0.50 1.21 0.26 0.14 0.04 10.40 0.00 28.10 HD108874 2-4 0.00 13.61 23.01 2.56 22.22 1.26 0.75 0.52 1.25 0.27 0.14 0.05 10.81 0.00 23.52 HD108874 3-4 0.00 13.05 21.72 0.63 21.41 1.22 0.71 0.46 1.21 0.26 0.14 0.04 10.27 0.01 28.86 HD108874 4-4 0.00 13.23 22.26 1.70 21.61 1.23 0.72 0.49 1.22 0.27 0.14 0.04 10.50 0.00 26.58 HD108874 4-5 0.00 13.01 22.24 4.98 21.15 1.20 0.71 0.50 1.19 0.26 0.13 0.06 10.41 0.00 24.15 t=2.5×106 years HD108874 1-4 0.00 13.75 23.40 2.84 22.35 1.27 0.75 0.53 1.26 0.27 0.14 0.06 11.00 0.00 22.36 Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD108874 2-4 0.00 13.59 23.16 4.09 22.10 1.26 0.74 0.52 1.25 0.27 0.14 0.06 10.87 0.00 21.96 HD108874 3-4 0.00 13.43 22.81 2.11 21.85 1.24 0.73 0.52 1.23 0.27 0.14 0.06 10.74 0.00 24.88 HD108874 4-4 0.00 13.97 23.79 3.41 22.71 1.29 0.76 0.54 1.28 0.28 0.14 0.06 11.17 0.00 20.60 HD108874 4-5 0.00 16.43 28.23 6.77 24.64 1.52 0.89 0.63 1.51 0.33 0.17 0.07 13.14 0.00 5.68 t=3×106 years HD108874 1-4 0.00 13.87 23.69 4.51 22.55 1.28 0.75 0.53 1.27 0.28 0.14 0.06 11.09 0.00 19.97 – 7 9 – HD108874 2-4 0.00 14.85 25.44 5.55 23.55 1.37 0.81 0.57 1.36 0.30 0.15 0.06 11.87 0.00 14.12 HD108874 3-4 0.00 14.81 25.28 4.51 24.08 1.37 0.81 0.57 1.36 0.30 0.15 0.06 11.85 0.00 14.85 HD108874 4-4 0.00 13.64 23.31 4.65 22.17 1.26 0.74 0.52 1.25 0.27 0.14 0.06 10.91 0.00 21.07 HD108874 4-5 0.00 17.36 30.33 7.15 25.97 1.59 0.94 0.66 1.58 0.34 0.18 0.07 13.84 0.00 0.00 HD177830 1-3 0.00 10.03 44.23 0.00 0.00 34.53 7.69 0.00 0.00 0.00 0.00 1.19 2.35 0.00 0.00 HD177830 t=2.5×105 years Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD177830 1-4 0.00 6.45 45.25 0.00 0.00 38.68 6.44 0.00 0.00 0.00 0.00 1.31 1.87 0.00 0.00 HD177830 2-3 0.00 1.56 47.00 0.00 0.00 45.84 3.33 0.00 0.00 0.00 0.00 1.55 0.71 0.00 0.00 HD177830 2-4 0.00 9.38 44.20 0.00 0.00 34.39 8.29 0.00 0.00 0.00 0.00 1.18 2.56 0.00 0.00 HD177830 2-5 0.00 18.06 35.12 1.20 17.14 14.42 4.00 0.00 1.24 0.16 0.15 0.50 7.99 0.01 0.00 HD177830 3-3 0.00 11.90 43.55 0.00 0.00 31.78 8.88 0.00 0.00 0.00 0.00 1.10 2.79 0.00 0.00 HD177830 3-4 0.00 8.55 43.96 0.00 1.83 34.92 6.66 0.00 0.16 0.01 0.01 1.19 2.71 0.00 0.00 – 8 0 – HD177830 4-3 0.00 2.23 46.60 0.00 0.00 44.18 4.38 0.00 0.00 0.00 0.00 1.50 1.10 0.00 0.00 HD177830 4-4 0.00 5.39 45.59 0.00 0.00 40.06 5.93 0.00 0.00 0.00 0.00 1.36 1.68 0.00 0.00 HD177830 4-5 0.00 10.43 43.94 0.00 0.00 33.33 8.51 0.00 0.00 0.00 0.00 1.15 2.65 0.00 0.00 t=5×105 years HD177830 1-3 0.00 15.70 37.27 0.23 13.62 18.03 5.94 0.00 1.05 0.10 0.12 0.81 7.12 0.01 0.00 HD177830 1-4 0.00 14.61 39.62 0.46 7.93 23.36 6.58 0.00 0.70 0.06 0.06 0.81 5.52 0.00 0.30 HD177830 2-3 0.00 11.59 43.46 0.00 0.00 31.39 9.41 0.00 0.00 0.00 0.00 1.21 2.94 0.00 0.00 HD177830 2-4 0.00 16.18 36.74 0.67 14.00 17.80 5.03 0.00 1.14 0.10 0.11 0.61 7.18 0.00 0.44 HD177830 2-5 0.00 21.96 30.76 0.62 25.79 2.02 0.57 0.16 1.88 0.24 0.23 0.07 12.73 0.01 2.96 Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD177830 3-3 0.00 22.33 29.23 2.03 29.10 2.28 0.65 0.00 2.11 0.27 0.26 0.08 11.64 0.02 0.00 HD177830 3-4 0.00 15.95 37.75 0.39 12.20 19.08 5.54 0.00 1.00 0.09 0.09 0.69 6.85 0.00 0.35 HD177830 4-3 0.00 11.86 43.55 0.00 0.00 31.74 8.94 0.00 0.00 0.00 0.00 1.10 2.81 0.00 0.00 HD177830 4-4 0.00 14.72 38.29 0.54 10.68 21.50 6.62 0.00 0.86 0.09 0.08 0.87 5.74 0.00 0.00 HD177830 4-5 0.00 14.57 40.46 0.00 7.15 23.83 6.73 0.00 0.70 0.04 0.04 0.82 5.65 0.00 0.00 t=1×106 years HD177830 1-3 0.00 22.02 30.02 1.96 24.55 1.99 0.58 0.13 1.82 0.21 0.21 0.07 12.68 0.00 3.75 HD177830 1-4 0.00 20.86 31.14 1.81 23.47 4.95 1.43 0.06 1.77 0.20 0.20 0.16 11.51 0.01 2.42 – 8 1 – HD177830 2-3 0.00 20.46 32.53 0.74 23.63 6.70 1.98 0.00 2.05 0.16 0.17 0.22 11.02 0.00 0.33 HD177830 2-4 0.00 21.97 29.87 2.44 24.45 1.92 0.54 0.11 1.78 0.23 0.21 0.06 12.72 0.01 3.69 HD177830 2-5 0.00 21.16 30.56 3.85 22.22 1.74 0.49 0.26 1.62 0.21 0.20 0.06 12.83 0.00 4.80 HD177830 3-3 0.00 22.00 31.47 0.38 23.13 1.81 0.51 0.28 1.68 0.22 0.20 0.06 13.35 0.00 4.92 HD177830 3-4 0.00 21.30 31.06 1.89 23.25 3.98 1.13 0.10 1.75 0.20 0.20 0.13 12.01 0.01 3.00 HD177830 4-3 0.00 21.36 29.63 1.07 29.57 2.83 0.80 0.00 2.35 0.22 0.24 0.07 11.78 0.00 0.07 Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD177830 4-4 0.00 21.07 31.98 1.38 22.59 5.63 1.60 0.09 1.78 0.19 0.18 0.19 11.36 0.00 1.96 HD177830 4-5 0.00 22.12 29.67 2.71 24.00 1.88 0.53 0.04 1.75 0.22 0.21 0.07 12.78 0.01 4.00 t=1.5×106 years HD177830 1-3 0.00 21.78 30.60 2.84 22.89 1.79 0.51 0.20 1.67 0.21 0.20 0.06 12.86 0.00 4.39 HD177830 1-4 0.00 21.88 30.39 2.79 23.59 1.85 0.52 0.18 1.72 0.22 0.21 0.06 12.86 0.01 3.71 HD177830 2-3 0.00 22.68 29.02 3.57 24.82 1.78 0.55 0.04 1.81 0.23 0.22 0.06 12.34 0.00 2.89 HD177830 2-4 0.00 21.59 31.18 2.60 22.85 1.79 0.51 0.26 1.66 0.21 0.20 0.06 13.19 0.01 3.89 HD177830 2-5 0.00 21.22 30.77 6.16 22.31 1.75 0.49 0.21 1.62 0.21 0.20 0.06 12.87 0.00 2.13 – 8 2 – HD177830 3-3 0.00 20.76 29.94 5.46 21.85 1.71 0.48 0.15 1.59 0.20 0.19 0.06 12.59 0.00 5.01 HD177830 3-4 0.00 21.81 30.56 3.11 23.36 1.81 0.52 0.19 1.70 0.22 0.21 0.06 12.89 0.00 3.56 HD177830 4-3 0.00 22.36 29.51 3.08 23.50 1.84 0.52 0.05 1.71 0.22 0.21 0.06 12.69 0.01 4.22 HD177830 4-4 0.00 21.90 30.12 3.11 24.02 1.80 0.53 0.20 1.75 0.22 0.21 0.06 12.69 0.00 3.38 HD177830 4-5 0.00 21.86 31.33 1.47 22.96 1.80 0.51 0.24 1.67 0.21 0.20 0.06 13.25 0.00 4.43 t=2×106 years HD177830 1-3 0.00 21.26 30.74 4.50 22.33 1.75 0.50 0.27 1.63 0.21 0.20 0.06 12.89 0.00 3.68 HD177830 1-4 0.00 21.67 31.09 3.30 22.76 1.78 0.50 0.23 1.66 0.21 0.20 0.06 13.12 0.00 3.41 Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD177830 2-3 0.00 21.87 31.10 0.96 22.97 1.80 0.51 0.19 1.67 0.21 0.20 0.05 13.24 0.01 5.22 HD177830 2-4 0.00 21.41 30.98 4.53 22.50 1.76 0.50 0.24 1.64 0.21 0.20 0.06 12.98 0.00 2.99 HD177830 2-5 0.00 20.69 32.33 6.44 21.72 1.70 0.48 0.25 1.58 0.20 0.19 0.06 12.55 0.00 1.81 HD177830 3-3 0.00 20.41 29.71 6.28 21.43 1.68 0.48 0.24 1.56 0.20 0.19 0.06 12.38 0.00 5.40 HD177830 3-4 0.00 21.36 30.79 3.93 22.44 1.76 0.50 0.22 1.63 0.21 0.20 0.06 12.95 0.00 3.95 HD177830 4-3 0.00 21.90 31.17 0.88 22.99 1.80 0.51 0.26 1.67 0.21 0.20 0.04 13.20 0.01 5.15 HD177830 4-4 0.00 21.35 30.65 2.97 22.42 1.76 0.50 0.21 1.63 0.21 0.20 0.06 12.95 0.01 5.09 HD177830 4-5 0.00 21.65 31.39 4.56 22.74 1.78 0.50 0.25 1.66 0.21 0.20 0.06 13.13 0.00 1.88 t=2.5×106 years HD177830 1-3 0.00 21.46 31.11 5.42 22.54 1.77 0.50 0.24 1.64 0.21 0.20 0.06 13.02 0.00 1.84 HD177830 1-4 0.00 21.31 30.84 4.36 22.37 1.75 0.50 0.26 1.63 0.21 0.20 0.06 12.93 0.00 3.60 HD177830 2-3 0.00 21.84 31.39 1.94 22.93 1.80 0.51 0.27 1.67 0.21 0.20 0.06 13.25 0.00 3.92 HD177830 2-4 0.00 21.06 30.64 5.87 22.11 1.73 0.50 0.25 1.61 0.21 0.19 0.06 12.78 0.00 3.00 HD177830 2-5 0.00 21.03 32.92 6.65 22.10 1.73 0.49 0.26 1.61 0.21 0.19 0.06 12.75 0.00 0.00 – 8 3 – Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD177830 3-3 0.00 21.54 31.35 6.79 22.62 1.77 0.50 0.26 1.65 0.21 0.20 0.06 13.06 0.00 0.00 HD177830 3-4 0.00 21.31 31.14 5.00 22.37 1.75 0.50 0.26 1.63 0.21 0.20 0.06 12.93 0.00 2.65 HD177830 4-3 0.00 21.80 31.42 2.65 22.88 1.79 0.52 0.28 1.67 0.21 0.20 0.06 13.23 0.00 3.28 HD177830 4-4 0.00 21.48 31.10 4.17 22.56 1.77 0.50 0.27 1.64 0.21 0.20 0.06 13.03 0.00 3.02 HD177830 4-5 0.00 21.00 30.52 5.81 22.05 1.73 0.49 0.24 1.61 0.20 0.19 0.06 12.74 0.00 3.36 t=3×106 years HD177830 1-3 0.00 21.20 30.86 6.18 22.27 1.74 0.49 0.26 1.62 0.21 0.20 0.06 12.86 0.00 2.05 HD177830 1-4 0.00 21.22 31.17 5.45 22.29 1.75 0.50 0.26 1.62 0.21 0.20 0.06 12.87 0.00 2.40 – 8 4 – HD177830 2-3 0.00 21.59 31.29 4.26 22.69 1.78 0.51 0.27 1.65 0.21 0.20 0.06 13.10 0.00 2.39 HD177830 2-4 0.00 21.20 31.28 6.40 22.25 1.74 0.49 0.26 1.62 0.21 0.20 0.06 12.86 0.00 1.44 HD177830 2-5 0.00 21.03 32.92 6.66 22.10 1.73 0.49 0.26 1.61 0.21 0.19 0.06 12.75 0.00 0.00 HD177830 3-3 0.00 21.53 31.34 6.81 22.61 1.77 0.50 0.26 1.65 0.21 0.20 0.06 13.06 0.00 0.00 HD177830 3-4 0.00 21.26 31.15 5.85 22.33 1.75 0.50 0.26 1.63 0.21 0.20 0.06 12.90 0.00 1.91 HD177830 4-3 0.00 21.74 31.64 4.95 22.83 1.79 0.51 0.27 1.66 0.21 0.20 0.06 13.19 0.00 0.94 HD177830 4-4 0.00 21.52 31.24 5.40 22.60 1.77 0.50 0.26 1.65 0.21 0.20 0.06 13.05 0.00 1.54 Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD177830 4-5 0.00 20.66 30.07 6.25 21.68 1.70 0.48 0.25 1.58 0.20 0.19 0.06 12.53 0.00 4.37 HD177830 t=2.5×105 years HD213240 1-3 0.00 4.57 42.18 0.00 0.00 25.65 19.79 0.00 0.00 0.00 0.00 1.24 6.57 0.00 0.00 HD213240 2-3 0.00 4.11 42.25 0.00 0.00 25.94 19.85 0.00 0.00 0.00 0.00 1.26 6.59 0.00 0.00 HD213240 3-3 0.00 3.92 42.29 0.00 0.00 26.07 19.87 0.00 0.00 0.00 0.00 1.26 6.59 0.00 0.00 – 8 5 – HD213240 4-3 0.00 3.64 42.33 0.00 0.00 26.26 19.89 0.00 0.00 0.00 0.00 1.27 6.60 0.00 0.00 HD213240 4-4 0.00 5.51 41.99 0.00 0.00 24.89 19.81 0.00 0.00 0.00 0.00 1.21 6.59 0.00 0.00 t=5×105 years HD213240 1-3 0.00 13.00 43.08 0.00 0.00 19.89 13.62 0.00 0.00 0.00 0.00 0.97 9.44 0.00 0.00 HD213240 2-3 0.00 14.37 39.87 0.00 7.87 13.81 11.25 0.00 0.48 0.07 0.04 0.67 11.57 0.00 0.00 HD213240 3-3 0.00 13.68 42.59 0.00 1.43 18.03 12.89 0.00 0.12 0.01 0.00 0.88 10.37 0.00 0.00 HD213240 4-3 0.00 13.04 42.42 0.00 1.68 18.70 13.16 0.00 0.14 0.01 0.00 0.91 9.95 0.00 0.00 Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD213240 4-4 0.00 19.02 31.55 0.00 29.53 1.53 1.22 0.00 1.71 0.33 0.17 0.07 14.86 0.00 0.00 t=1×106 years HD213240 1-3 0.00 19.89 32.68 0.00 27.02 1.67 1.33 0.09 1.56 0.30 0.16 0.08 15.22 0.00 0.00 HD213240 2-3 0.00 19.47 32.25 0.00 27.89 1.58 1.26 0.30 1.67 0.29 0.14 0.08 15.07 0.00 0.00 HD213240 3-3 0.00 19.43 32.18 0.00 28.05 1.58 1.26 0.18 1.68 0.29 0.15 0.08 15.12 0.00 0.00 – 8 6 – HD213240 4-3 0.00 19.56 32.26 0.00 27.88 1.59 1.26 0.15 1.67 0.28 0.14 0.08 15.13 0.00 0.00 HD213240 4-4 0.00 18.77 31.65 0.00 29.10 1.51 1.20 0.85 1.68 0.33 0.17 0.07 14.67 0.00 0.00 t=1.5×106 years HD213240 1-3 0.00 18.80 31.61 0.00 29.19 1.51 1.21 0.74 1.69 0.33 0.17 0.07 14.69 0.00 0.00 HD213240 2-3 0.00 18.78 31.53 0.29 29.14 1.51 1.20 0.63 1.68 0.33 0.17 0.07 14.67 0.00 0.00 HD213240 3-3 0.00 18.83 31.61 0.00 29.23 1.51 1.21 0.63 1.69 0.33 0.17 0.07 14.72 0.00 0.00 HD213240 4-3 0.00 18.85 31.60 0.00 29.26 1.51 1.21 0.59 1.69 0.33 0.17 0.07 14.73 0.00 0.00 HD213240 4-4 0.00 18.04 30.43 3.86 27.98 1.45 1.16 0.82 1.62 0.31 0.16 0.07 14.10 0.00 0.00 Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD213240 1-3 0.00 18.69 31.50 0.41 29.00 1.50 1.20 0.84 1.67 0.32 0.17 0.07 14.61 0.00 0.00 t=2×106 years HD213240 2-3 0.00 18.52 31.23 1.33 28.73 1.49 1.19 0.84 1.66 0.32 0.17 0.07 14.47 0.00 0.00 HD213240 3-3 0.00 18.62 31.40 0.79 28.88 1.50 1.19 0.84 1.67 0.32 0.17 0.07 14.55 0.00 0.00 HD213240 4-3 0.00 18.65 31.46 0.60 28.93 1.50 1.20 0.84 1.67 0.32 0.17 0.07 14.58 0.00 0.00 HD213240 4-4 0.00 17.87 30.16 4.63 27.75 1.44 1.15 0.88 1.60 0.31 0.16 0.07 13.97 0.00 0.00 t=2.5×106 years HD213240 1-3 0.00 18.42 31.07 1.82 28.59 1.48 1.18 0.83 1.65 0.32 0.17 0.07 14.40 0.00 0.00 – 8 7 – HD213240 2-3 0.00 18.31 30.89 2.37 28.42 1.47 1.17 0.85 1.64 0.32 0.17 0.07 14.31 0.00 0.00 HD213240 3-3 0.00 18.40 31.03 1.95 28.55 1.48 1.18 0.83 1.65 0.32 0.17 0.07 14.38 0.00 0.00 HD213240 4-3 0.00 18.45 31.12 1.66 28.63 1.48 1.18 0.83 1.65 0.32 0.17 0.07 14.42 0.00 0.00 HD213240 4-4 0.00 17.85 30.12 4.70 27.78 1.43 1.14 0.88 1.60 0.31 0.16 0.07 13.95 0.00 0.00 t=3×106 years Table 1-Continued System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C HD213240 1-3 0.00 18.15 30.61 3.25 28.18 1.46 1.16 0.83 1.63 0.31 0.16 0.07 14.19 0.00 0.00 HD213240 2-3 0.00 18.15 30.62 3.22 28.18 1.46 1.16 0.85 1.63 0.31 0.16 0.07 14.19 0.00 0.00 HD213240 3-3 0.00 18.19 30.69 3.01 28.24 1.46 1.17 0.84 1.63 0.32 0.16 0.07 14.22 0.00 0.00 HD213240 4-3 0.00 18.23 30.75 2.83 28.29 1.47 1.17 0.84 1.63 0.32 0.16 0.07 14.25 0.00 0.00 HD213240 4-4 0.00 16.72 34.55 4.41 26.00 1.34 1.07 0.83 1.50 0.29 0.15 0.07 13.07 0.00 0.00 – 8 8 – – 89 – B. Online Material: Chemical Condensation Sequence Plots – 90 – HD72659 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 HD213240 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 HD177830 C CaMgSi 2O6 CaS Fe Fe2SiO4 Fe3O4 FeS FeSiO3 H2O Mg2SiO4 Mg3Si2O5(OH)4 MgAl2O4 MgS MgSiO3 NaAlSi 3O8 Ni SiC l ) e o m ( e c n a d n u b A d e z i l a m r o N 10000 1000 100 10 10000 1000 100 10 10000 1000 100 10 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 T (K) Fig. 1.- Schematic of the output obtained from HSC Chemistry for HD72659 (top), HD213240 (middle) and HD177830 (bottom) at a pressure of 10−4 bar. Only solid species present within the system are shown. All abundances are normalized to the least abundant species present. Input elemental abundances are shown in Table 7. – 91 – HD17051 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 55Cnc 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 10000 1000 100 10 10000 1000 100 10 l ) e o m ( e c n a d n u b A d e z i l a m r o N HD19994 10000 1000 100 10 C CaMgSi 2O6 CaS Fe Fe2SiO4 Fe3O4 FeS FeSiO3 H2O Mg2SiO4 Mg3Si2O5(OH)4 MgAl2O4 MgS MgSiO3 NaAlSi 3O8 Ni SiC 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 T (K) Fig. 2.- Schematic of the output obtained from HSC Chemistry for HD17051 (top), HD55Cnc (middle) and HD19994 (bottom) at a pressure of 10−4 bar. Only solid species present within the system are shown. All abundances are normalized to the least abundant species present. Input elemental abundances are shown in Table 7. – 92 – 10000 l ) e o m 1000 HD108874 ( e c n a d n u b A d e z i l a m r o N 100 10 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 T (K) C CaMgSi 2O6 CaS Fe Fe2SiO4 Fe3O4 FeS FeSiO3 H2O Mg2SiO4 Mg3Si2O5(OH)4 MgAl2O4 MgS MgSiO3 NaAlSi 3O8 Ni SiC Fig. 3.- Schematic of the output obtained from HSC Chemistry for HD108874 at a pressure of 10−4 bar. Only solid species present within the system are shown. All abundances are normalized to the least abundant species present. Input elemental abundances are shown in Table 7. – 93 – REFERENCES Acke, B., & van den Ancker, M. E. 2006, Astronomy & Astrophysics, 457, 171 Armitage, P. J. 2003, Astrophysical Journal Letters, 582, L47 Asghari, N., et al. 2004, Astronomy & Astrophysics, 426, 353 Asplund, M., Grevesse, N., & Sauval, A. J. 2005, in Astronomical Society of the Pacific Conference Series, Vol. 336, Cosmic Abundances as Records of Stellar Evolution and Nucleosynthesis, ed. T. G. Barnes, III & F. N. Bash, 25–38 Ayres, T. R., Plymate, C., & Keller, C. U. 2006, Astrophysical Journal Supplemental Series, 165, 618 Barnes, R., & Raymond, S. N. 2004, Astrophysical Journal, 617, 569 Beirao, P., Santos, N. C., Israelian, G., & Mayor, M. 2005, Astronomy & Astrophysics, 438, 251 Bodaghee, A., Santos, N. C., Israelian, G., & Mayor, M. 2003, Astronomy & Astrophysics, 404, 715 Bond, J. C., Lauretta, D. S., & O'Brien, D. P. 2009, Icarus Bond, J. C., Tinney, C. G., Butler, R. P., Jones, H. R. A., Marcy, G. W., Penny, A. J., & Carter, B. D. 2006, Monthly Notices of the Royal Astronomical Society, 370, 163 Bond, J. C., et al. 2008, Astrophysical Journal, 682, 1234 Burrows, A., Hubeny, I., Budaj, J., & Hubbard, W. B. 2007, Astrophysical Journal, 661, 502 – 94 – Butler, R. P., Vogt, S. S., Marcy, G. W., Fischer, D. A., Henry, G. W., & Apps, K. 2000, Astrophysical Journal, 545, 504 Butler, R. P., et al. 2006, Astrophysical Journal, 646, 505 Davis, A. M. 2006, in Meteorites and the Early Solar System II, ed. D. S. Lauretta & H. Y. McSween Jr. (Tucson, AZ USA: University of Arizona Press), 295–307 Drake, M. J. 2000, Geochimica et Cosmochimica Acta, 64, 2363 -. 2005, Meteoritics and Planetary Science, 40, 519 Duncan, M. J., Levison, H. F., & Lee, M. H. 1998, Astronomical Journal, 116, 2067 Ebel, D. S. 2006, in Meteorites and the Early Solar System II, ed. D. S. Lauretta & H. Y. McSween Jr. (Tucson, AZ USA: University of Arizona Press), 253–277 Ecuvillon, A., Israelian, G., Santos, N. C., Mayor, M., Villar, V., & Bihain, G. 2004a, Astronomy & Astrophysics, 426, 619 -. 2004b, Astronomy & Astrophysics, 426, 619 Ecuvillon, A., Israelian, G., Santos, N. C., Shchukina, N. G., Mayor, M., & Rebolo, R. 2006, Astronomy & Astrophysics, 445, 633 Fischer, D. A., & Valenti, J. 2005, Astrophysical Journal, 622, 1102 Gaidos, E., & Selsis, F. 2007, in Protostars and Planets V, ed. B. Reipurth, D. Jewitt, & K. Keil (Tucson, AZ USA: University of Arizona Press), 929–944 Gilli, G., Israelian, G., Ecuvillon, A., Santos, N. C., & Mayor, M. 2006, Astronomy & Astrophysics, 449, 723 Gonzalez, G. 1997, Monthly Notices of the Royal Astronomical Society, 285, 403 – 95 – -. 1998, Astronomy & Astrophysics, 334, 221 Gonzalez, G., & Laws, C. 2000, Astrophysical Journal, 119, 390 Gonzalez, G., Laws, C., Tyagi, S., & Reddy, B. E. 2001, Astronomical Journal, 121, 432 Gonzalez, G., & Vanture, A. D. 1998, Astronomy & Astrophysics, 339, L29 Gonzalez, G., Wallerstein, G., & Saar, S. H. 1999, Astrophysical Journal, 511, L111 Gradie, J., & Tedesco, E. 1982, Science, 216, 1405 Guillot, T., Santos, N. C., Pont, F., Iro, N., Melo, C., & Ribas, I. 2006, Astronomy & Astrophysics, 453, L21 Gustafsson, B., Karlsson, T., Olsson, E., Edvardsson, B., & Ryde, N. 1999, A&A, 342, 426 Hersant, F., Gautier, D., & Hur´e, J.-M. 2001, Astrophysical Journal, 554, 391 Israelian, G., Shchukina, N., Rebolo, R., Basri, G., Gonz´alez Hern´andez, J. I., & Kajino, T. 2004, Astronomy & Astrophysics, 419, 1095 Kargel, J. S., & Lewis, J. S. 1993, Icarus, 105, 1 Kokubo, E., & Ida, S. 2000, Icarus, 143, 15 Kuchner, M. J., & Seager, S. 2005, ArXiv Astrophysics e-prints Laughlin, G. 2000, Astrophysical Journal, 545, 1064 Lodders, K., & Fegley, B. 1997, Icarus, 126, 373 Mandell, A. M., Raymond, S. N., & Sigurdsson, S. 2007, Astrophysical Journal, 660, 823 – 96 – Marcy, G., Butler, P. R., Fischer, D. A., & Vogt, S. S. 2000, in Astronomical Society of the Pacific Conference Series, Vol. 213, Bioastronomy 99, ed. G. Lemarchand & K. Meech, 85–94 McDonough, W. F., & Sun, S. 1995, Chemical Geology, 120, 223 Morgan, J. W., & Anders, E. 1980, Proceedings of the National Academy of Science, 77, 6973 Mousis, O., & Alibert, Y. 2006, A&A, 448, 771 Murray, N., Chaboyer, B., Arras, P., Hansen, B., & Noyes, R. W. 2001, Astrophysical Journal, 555, 801 O'Brien, D. P., Morbidelli, A., & Levison, H. F. 2006, Icarus, 184, 39 O'Neill, C., Jellinek, A. M., & Lenardic, A. 2007, Earth and Planetary Science Letters, 261, 20 Pinsonneault, M. H., DePoy, D. L., & Coffee, M. 2001, Astrophysical Journal Letters, 556, L59 Raymond, S. N., & Barnes, R. 2005, Astrophysical Journal, 619, 549 Raymond, S. N., Barnes, R., & Kaib, N. A. 2006a, Astrophysical Journal, 644, 1223 Raymond, S. N., Mandell, A. M., & Sigurdsson, S. 2006b, Science, 313, 1413 Raymond, S. N., Quinn, T., & Lunine, J. I. 2005, Icarus, 177, 256 Reid, I. N. 2002, Publications of the Astronomical Society of the Pacific, 114, 306 Santos, N. C., Israelian, G., & Mayor, M. 2000, Astronomy & Astrophysics, 363, 228 -. 2001, Astronomy & Astrophysics, 373, 1019 – 97 – -. 2004, Astronomy & Astrophysics, 415, 1153 Santos, N. C., Israelian, G., Mayor, M., Bento, J. P., Almeida, P. C., Sousa, S. G., & Ecuvillon, A. 2005, Astronomy & Astrophysics, 437, 1127 Santos, N. C., Israelian, G., Mayor, M., Rebolo, R., & Udry, S. 2003a, Astronomy & Astrophysics, 398, 363 -. 2003b, Astronomy & Astrophysics, 398, 363 Semenov, D., Chakraborty, S., & Thiemens, M. 2010, in Protoplanetary Dust, ed. D. Apai & D. S. Lauretta (New York, NY USA: Cambridge; New York: Cambridge University Press, c2010), 97–127 Smith, V. V., Cunha, K., & Lazzaro, D. 2001, Astronomical Journal, 121, 3207 Socas-Navarro, H., & Norton, A. A. 2007, Astrophysical Journal Letters, 660, L153 Sotin, C., Grasset, O., & Mocquet, A. 2007, Icarus, 191, 337 Valencia, D., O'Connell, R. J., & Sasselov, D. D. 2007, Astrophysical Journal Letters, 670, L45 This manuscript was prepared with the AAS LATEX macros v5.2. – 98 – 2.0 1.5 O C / 1.0 0.5 0.0 0.5 1.0 1.5 Mg/Si 2.0 2.5 Fig. 4.- Mg/Si vs. C/O for known planetary host stars with reliable stellar abundances. Filled circles represent those systems selected for this study. Stellar photospheric values were taken from Gilli et al. (2006) (Si, Mg), Beirao et al. (2005) (Mg), Ecuvillon et al. (2004a) (C) and Ecuvillon et al. (2006) (O). Solar values are shown by the black star and were taken from Asplund et al. (2005). The dashed line indicates a C/O value of 0.8 and marks the transitions between a silicate-dominated composition and a carbide-dominated composition at 10−4 bar. Average 2-σ error bars shown in upper right. All ratios are elemental number ratios, not solar normalized logarithmic values. – 99 – Host stars 0.0 0.5 1.0 1.5 2.0 2.5 Non-host stars Host stars 0.0 0.5 1.0 1.5 2.0 2.5 Non-host stars 10 5 0 30 25 20 15 10 5 0 s r a t s f o r e b m u N s r a t s f o r e b m u N 10 5 0 30 25 20 15 10 5 0 s r a t s f o r e b m u N s r a t s f o r e b m u N 0.0 0.5 1.0 1.5 2.0 2.5 0.0 0.5 1.0 1.5 2.0 2.5 C/O Ratio Mg/Si Ratio Fig. 5.- C/O and Mg/Si distributions for host and non-host stars based on the abundances determined in Bond et al. (2008). Left: C/O distributions for host (top) and non-host (bottom) stars. Right: Mg/Si distributions for host (top) and non-host (bottom) stars. All ratios are elemental number ratios, not solar normalized logarithmic values. Note that these values are for a different dataset to that shown in Figure 4 and utilized in this work. – 100 – Extrasolar Planetary Systems Extrasolar Planetary Systems 55Cnc 55Cnc Gl777A Gl777A HD4203 HD4203 HD17051 HD17051 HD19994 HD19994 HD27442 HD27442 HD72659 HD72659 HD108874 HD108874 HD177830 HD177830 HD213240 HD213240 1 MJ 1 MJ 0 0 1 1 2 2 3 3 4 4 5 5 6 6 7 7 Semimajor Axis (AU) Semimajor Axis (AU) Fig. 6.- Location of known giant planets in the systems selected for study. The horizontal lines indicate the variation from periastron and apastron. The size of the circles scales with the planetary Msini value. All planets are assumed to have zero inclination. All values taken from the Butler et al. (2006) catalog. – 101 – Final Planetary Systems - 55 Cnc Final Planetary Systems - 55 Cnc Final Planetary Systems - 55 Cnc Final Planetary Systems - 55 Cnc Final Planetary Systems - 55 Cnc Final Planetary Systems - 55 Cnc Final Planetary Systems - 55 Cnc Final Planetary Systems - 55 Cnc 0.66 0.66 0.66 0.66 0.50 0.50 0.50 0.50 0.64 0.64 0.64 0.64 0.25 0.25 0.25 0.25 0.8 0.2 0.1 MJ 0.8 0.2 0.1 MJ 0.8 0.2 0.1 MJ 0.8 0.2 0.1 MJ 0.47 0.47 0.47 0.47 3.84 MJ 3.84 MJ 3.84 MJ 3.84 MJ 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 4 4 4 4 4 4 4 4 5 5 5 5 5 5 5 5 6 6 6 6 6 6 6 6 7 7 7 7 7 7 7 7 Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Final Planetary Systems - Gl 777 Final Planetary Systems - Gl 777 Final Planetary Systems - Gl 777 Final Planetary Systems - Gl 777 Final Planetary Systems - Gl 777 Final Planetary Systems - Gl 777 Final Planetary Systems - Gl 777 Final Planetary Systems - Gl 777 0.80 0.80 0.80 0.80 0.94 0.94 0.94 0.94 0.40 0.40 0.40 0.40 0.47 0.47 0.47 0.47 1.10 1.10 1.10 1.10 0.06 MJ 0.06 MJ 0.06 MJ 0.06 MJ 1.03 1.03 1.03 1.03 1.55 MJ 1.55 MJ 1.55 MJ 1.55 MJ 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 4 4 4 4 4 4 4 4 5 5 5 5 5 5 5 5 Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Fig. 7.- Schematic of the results of the dynamical simulations for 55Cancri (top panel) and Gl777 (bottom panel). Known giant planets are also shown with their masses in Jupiter masses (MJ ). The horizontal lines indicate the range in distance from apastron to periastron. The vertical lines indicate variation in height above the midplane due to orbital inclination. Numerical values represent the mass of the planet in Earth masses. – 102 – Final Planetary Systems - HD4203 Final Planetary Systems - HD4203 Final Planetary Systems - HD4203 Final Planetary Systems - HD4203 Final Planetary Systems - HD4203 Final Planetary Systems - HD4203 Final Planetary Systems - HD4203 Final Planetary Systems - HD4203 0.17 0.17 0.17 0.17 0.04 0.04 0.04 0.04 0.04 0.04 0.04 0.04 0.04 0.04 0.04 0.04 2.1 MJ 2.1 MJ 2.1 MJ 2.1 MJ 0 0 0 0 0 0 0 0 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 1 1 1 1 1 1 1 1 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 2 2 2 2 2 2 2 2 Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Final Planetary Systems - HD17051 Final Planetary Systems - HD17051 Final Planetary Systems - HD17051 Final Planetary Systems - HD17051 Final Planetary Systems - HD17051 Final Planetary Systems - HD17051 Final Planetary Systems - HD17051 Final Planetary Systems - HD17051 0.05 0.05 0.05 0.05 0.06 0.06 0.06 0.06 0.17 0.17 0.17 0.17 0.20 0.20 0.20 0.20 0.26 0.26 0.26 0.26 0.27 0.27 0.27 0.27 0.12 0.12 0.12 0.12 2.08 MJ 2.08 MJ 2.08 MJ 2.08 MJ 0 0 0 0 0 0 0 0 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.4 0.4 0.4 0.4 0.4 0.4 0.4 0.4 0.6 0.6 0.6 0.6 0.6 0.6 0.6 0.6 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 1 1 1 1 1 1 1 1 1.2 1.2 1.2 1.2 1.2 1.2 1.2 1.2 Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Fig. 8.- Schematic of the results of the dynamical simulations for HD4203 (top panel) and HD17051 (bottom panel). Known giant planets are also shown with their masses in Jupiter masses (MJ ). The horizontal lines indicate the range in distance from apastron to periastron. The vertical lines indicate variation in height above the midplane due to orbital inclination. Numerical values represent the mass of the planet in Earth masses. – 103 – Final Planetary Systems - HD19994 Final Planetary Systems - HD19994 Final Planetary Systems - HD19994 Final Planetary Systems - HD19994 Final Planetary Systems - HD19994 Final Planetary Systems - HD19994 Final Planetary Systems - HD19994 Final Planetary Systems - HD19994 0.57 0.57 0.57 0.57 0.62 0.62 0.62 0.62 0.35 0.35 0.35 0.35 0.10 0.10 0.10 0.10 0.06 0.06 0.06 0.06 0.28 0.28 0.28 0.28 0.46 0.46 0.46 0.46 0 0 0 0 0 0 0 0 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 1 1 1 1 1 1 1 1 Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) 1.7 MJ 1.7 MJ 1.7 MJ 1.7 MJ 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 2 2 2 2 2 2 2 2 Final Planetary Systems - HD27442 Final Planetary Systems - HD27442 Final Planetary Systems - HD27442 Final Planetary Systems - HD27442 Final Planetary Systems - HD27442 Final Planetary Systems - HD27442 Final Planetary Systems - HD27442 Final Planetary Systems - HD27442 0.31 0.63 0.05 0.31 0.63 0.05 0.31 0.63 0.05 0.31 0.63 0.05 0.35 0.72 0.35 0.72 0.35 0.72 0.35 0.72 0.56 0.56 0.56 0.56 0.45 0.45 0.45 0.45 0.07 0.07 0.07 0.07 0.44 0.44 0.44 0.44 0.71 0.71 0.71 0.71 0.06 0.06 0.06 0.06 1.56 MJ 1.56 MJ 1.56 MJ 1.56 MJ 0 0 0 0 0 0 0 0 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.4 0.4 0.4 0.4 0.4 0.4 0.4 0.4 0.6 0.6 0.6 0.6 0.6 0.6 0.6 0.6 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 1 1 1 1 1 1 1 1 1.2 1.2 1.2 1.2 1.2 1.2 1.2 1.2 1.4 1.4 1.4 1.4 1.4 1.4 1.4 1.4 Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Fig. 9.- Schematic of the results of the dynamical simulations for HD19994 (top panel) and HD27442 (bottom panel). Known giant planets are also shown with their masses in Jupiter masses (MJ ). The horizontal lines indicate the range in distance from apastron to periastron. The vertical lines indicate variation in height above the midplane due to orbital inclination. Numerical values represent the mass of the planet in Earth masses. – 104 – Final Planetary Systems - HD72659 Final Planetary Systems - HD72659 Final Planetary Systems - HD72659 Final Planetary Systems - HD72659 Final Planetary Systems - HD72659 Final Planetary Systems - HD72659 Final Planetary Systems - HD72659 Final Planetary Systems - HD72659 1.53 1.53 1.53 1.53 1.35 1.35 1.35 1.35 0.60 1.10 0.60 1.10 0.60 1.10 0.60 1.10 1.03 1.03 1.03 1.03 1.28 0.99 0.26 1.28 0.99 0.26 1.28 0.99 0.26 1.28 0.99 0.26 0.44 0.44 0.44 0.44 1.32 1.32 1.32 1.32 0.71 0.71 0.71 0.71 3.3 MJ 3.3 MJ 3.3 MJ 3.3 MJ 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 4 4 4 4 4 4 4 4 5 5 5 5 5 5 5 5 Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Final Planetary Systems - HD108874 Final Planetary Systems - HD108874 Final Planetary Systems - HD108874 Final Planetary Systems - HD108874 Final Planetary Systems - HD108874 Final Planetary Systems - HD108874 Final Planetary Systems - HD108874 Final Planetary Systems - HD108874 0.34 0.34 0.34 0.34 0.46 0.46 0.46 0.46 0.18 0.18 0.18 0.18 0.40 0.13 0.40 0.13 0.40 0.13 0.40 0.13 1.3 MJ 1.3 MJ 1.3 MJ 1.3 MJ 1.07 MJ 1.07 MJ 1.07 MJ 1.07 MJ 0 0 0 0 0 0 0 0 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 1 1 1 1 1 1 1 1 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 2 2 2 2 2 2 2 2 2.5 2.5 2.5 2.5 2.5 2.5 2.5 2.5 3 3 3 3 3 3 3 3 Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Fig. 10.- Schematic of the results of the dynamical simulations for HD72659 (top panel) and HD108874 (bottom panel). Known giant planets are also shown with their masses in Jupiter masses (MJ ). The horizontal lines indicate the range in distance from apastron to periastron. The vertical lines indicate variation in height above the midplane due to orbital inclination. Numerical values represent the mass of the planet in Earth masses. – 105 – Final Planetary Systems - HD177830 Final Planetary Systems - HD177830 Final Planetary Systems - HD177830 Final Planetary Systems - HD177830 Final Planetary Systems - HD177830 Final Planetary Systems - HD177830 Final Planetary Systems - HD177830 Final Planetary Systems - HD177830 0.78 0.78 0.78 0.78 0.24 0.24 0.24 0.24 0.35 0.35 0.35 0.35 0.61 0.61 0.61 0.61 0.14 0.14 0.14 0.14 1.22 1.22 1.22 1.22 0.06 0.06 0.06 0.06 0.36 0.36 0.36 0.36 0.24 0.24 0.24 0.24 0.34 0.34 0.34 0.34 1.4 MJ 1.4 MJ 1.4 MJ 1.4 MJ 0 0 0 0 0 0 0 0 0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.75 0.75 0.75 0.75 0.75 0.75 0.75 0.75 1 1 1 1 1 1 1 1 1.25 1.25 1.25 1.25 1.25 1.25 1.25 1.25 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Final Planetary Systems - HD213240 Final Planetary Systems - HD213240 Final Planetary Systems - HD213240 Final Planetary Systems - HD213240 Final Planetary Systems - HD213240 Final Planetary Systems - HD213240 Final Planetary Systems - HD213240 Final Planetary Systems - HD213240 0.47 0.47 0.47 0.47 0.63 0.63 0.63 0.63 0.71 0.71 0.71 0.71 0.60 0.60 0.60 0.60 0.07 0.07 0.07 0.07 4.72 MJ 4.72 MJ 4.72 MJ 4.72 MJ 0 0 0 0 0 0 0 0 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 1 1 1 1 1 1 1 1 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 2 2 2 2 2 2 2 2 Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Semimajor Axis (AU) Fig. 11.- Schematic of the results of the dynamical simulations for HD177830 (top panel) and HD213240 (bottom panel). Known giant planets are also shown with their masses in Jupiter masses (MJ ). The horizontal lines indicate the range in distance from apastron to periastron. The vertical lines indicate variation in height above the midplane due to orbital inclination. Numerical values represent the mass of the planet in Earth masses. – 106 – HD27442 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 Gl777 10000 1000 100 10 10000 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 HD4203 1000 100 10 10000 1000 100 10 l ) e o m ( e c n a d n u b A d e z i l a m r o N C CaMgSi 2O6 CaS Fe Fe2SiO4 Fe3O4 FeS FeSiO3 H2O Mg2SiO4 Mg3Si2O5(OH)4 MgAl2O4 MgS MgSiO3 NaAlSi 3O8 Ni SiC 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 T (K) Fig. 12.- Schematic of the output obtained from HSC Chemistry for 3 representative sys- tems: HD27442 (top), Gl777 (middle) and HD4203 (bottom). All simulations were run with a system pressure of 10−4 bar. Only solid species present within the system are shown. All abundances are normalized to the least abundant species present. Input elemental abun- dances are shown in Table 7.Note that although Gl777 is described as a low-C enriched systems, C and other carbide phases only appear for pressures at and below 10−5 bar and are thus absent from the current figure. – 107 – Final Composition - HD27442 (0.5Myr) Final Composition - HD72659 (0.5Myr) Sim. 1 Sim. 2 Sim. 3 Sim. 4 Sim. 1 Sim. 2 Sim. 3 Sim. 4 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 Semimajor Axis (AU) Semimajor Axis (AU) Final Composition - HD213240 (0.5Myr) Sim. 1 Sim. 2 Sim. 3 Sim. 4 O Fe Mg Si C S Al Ca Other 0.0 0.2 0.4 0.6 0.8 Semimajor Axis (AU) Fig. 13.- Schematic of the bulk elemental planetary composition for the Earth-like planetary systems HD27442 (top left), HD72659 (top right) and HD213240 (bottom). All values are wt% of the final simulated planet. Values are shown for the terrestrial planets produced in each of the four simulations run for the system. Size of bodies is not to scale. Earth values taken from Kargel & Lewis (1993) are shown in the upper right of each panel for comparison. – 108 – ) o i t a r t h g i e w ( i / S g M 1.8 1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 Earth fractionation line Martian fractionation line Mars Earth Venus HD213240 HD72659 HD27442 0.0 0.2 0.4 0.6 0.8 Al/Si (weight ratio) Fig. 14.- Al/Si v. Mg/Si for the planets of HD27442 (triangles), HD72659 (squares) and HD213240 (circles). Values are for disk conditions at 5×105 years. Earth values are shown as filled circles and are taken from Kargel & Lewis (1993) and McDonough & Sun (1995). Martian values are shown as filled diamonds and are taken from Lodders & Fegley (1997). Venus values are shown as filled squares and are taken from Morgan & Anders (1980). Note that the values for HD27442, HD72659 and HD213240 all extend off to the right, reaching Mg/Si values of up to 3.5. – 109 – Final Composition - 55 Cnc (0.5 Myr) Final Composition - Gl777 (0.5Myr) Sim. 1 Sim. 2 Sim. 3 Sim. 4 0 1 2 3 4 0.0 0.2 0.4 0.6 0.8 1.0 Semimajor Axis (AU) Semimajor Axis (AU) Final Composition - HD17051 (0.5Myr) Final Composition - HD177830 (0.5Myr) O Fe Mg Si C S Al Ca Other Sim. 1 Sim. 2 Sim. 3 Sim. 4 Sim. 1 Sim. 2 Sim. 3 Sim. 4 Sim. 1 Sim. 2 Sim. 3 Sim. 4 0.0 0.2 0.4 0.6 0.8 0.0 0.2 0.4 0.6 Semimajor Axis (AU) Semimajor Axis (AU) Fig. 15.- Schematic of the bulk elemental planetary composition for the low C-enrichment systems 55Cnc (top left), Gl777 (top right), HD17051 (bottom left) and HD177830 (bottom right). All values are wt% of the final simulated planet. Values are shown for the terrestrial planets produced in each of the four simulations run for the system. Size of bodies is not to scale. Earth values taken from Kargel & Lewis (1993) are shown in the upper right of each panel for comparison. – 110 – Earth fractionation line Mars Earth Venus Gl777 Mars fractionation line 0.05 0.10 0.15 0.20 Al/Si (weight ratio) 1.4 1.2 1.0 0.8 0.6 ) o i t a r t h g i e w ( i / S g M 0.4 0.00 Fig. 16.- Al/Si v. Mg/Si for planets of Gl777. Open circles indicate values for simulated terrestrial planets produced with disk conditions at t = 5×105 years. Values at all other times are concentrated at the 5×105 years values and omitted for clarity. Earth values are shown as filled circles and are taken from Kargel & Lewis (1993) and McDonough & Sun (1995). Martian values are shown as filled diamonds and are taken from Lodders & Fegley (1997). Venus values are shown as filled squares and are taken from Morgan & Anders (1980). – 111 – Final Composition - HD19994 (0.5Myr) Final Composition - HD108874 (0.5Myr) Sim. 1 Sim. 2 Sim. 3 Sim. 4 Sim. 1 Sim. 2 Sim. 3 Sim. 4 0.0 0.2 0.4 0.6 0.8 0.0 0.1 0.2 0.3 0.4 0.5 0.6 Semimajor Axis (AU) Semimajor Axis (AU) Final Composition - HD4203 (0.5Myr) Sim. 1 Sim. 2 Sim. 3 Sim. 4 O Fe Mg Si C S Al Ca Other 0.0 0.1 0.2 0.3 0.4 0.5 Semimajor Axis (AU) Fig. 17.- Schematic of the bulk elemental planetary composition for the high C-enrichment systems HD19994 (top left), HD108874 (top right) and HD4203 (bottom). All values are wt% of the final simulated planet. Values are shown for the terrestrial planets produced in each of the four simulations run for the system. Size of bodies is not to scale. Earth values taken from Kargel & Lewis (1993) are shown in the upper right of each panel for comparison. – 112 – HD 4203 HD27442 HD177830 HD 72659 1 2 3 4 5 r (AU) 1000 100 10 s s a M d e z i l a m r o N r a o S l 1 0 Fig. 18.- Solid mass distribution within the disk for four known extrasolar planetary systems. All distributions are normalized to the Solar distribution. Mass distributions are shown for HD4203 (solid) (Mg/Si= 1.29, C/O=1.86), HD27442 (dashed-dotted) (Mg/Si= 1.17, C/O=0.63), HD177830 (long dash) (Mg/Si= 1.91, C/O=0.83) and HD72659 (short dash) (Mg/Si= 1.23, C/O=0.40). – 113 – Fig. 19.- Schematic of notional interior models are based on calculations of bulk planetary compositions for disk conditions at t = 5×105 years resulting from three different planetary systems: Gl777 (HD190360) (top), HD177830 (middle) and HD108874 (bottom). Figures are to scale for planet and layer sizes and planet location. – 114 – Table 2: Statistical analysis of the host and non-host star distributions of Mg/Si and C/O. All values are based on the abundances determined in Bond et al. (2008). The quoted uncertainty is the standard error in the mean. All ratios are elemental number ratios, not solar normalized logarithmic values. Mean Median Standard Deviation Mg/Si: Host Stars 0.83± 0.04 Non-Host Stars 0.80± 0.03 C/O: Host Stars 0.67± 0.03 Non-Host Stars 0.67± 0.03 0.80 0.79 0.68 0.69 0.22 0.16 0.23 0.23 – 115 – Table 3: Orbital parameters of known extrasolar planets for the systems studied. Values taken from the University of California catalog located at www.exoplanets.org Planet M a e (MJupiter) (AU) 55Cnc-b 55Cnc-c 55Cnc-d 55Cnc-e 55Cnc-f Gl777-b Gl777-c 0.82 0.17 3.84 0.02 0.14 1.55 0.06 0.11 0.02 0.24 0.05 5.84 0.08 0.04 0.09 0.70 0.20 4.02 0.35 0.13 0.07 HD4203-b 2.07 1.16 0.52 HD17051-b 2.08 0.93 0.14 HD19994-b 1.69 1.43 0.30 HD27442-b 1.56 1.27 0.06 HD72659-b 3.30 4.76 0.26 HD108874-b HD108874-c 1.30 1.07 1.05 0.21 2.75 0.16 HD177830-b 1.43 1.22 0.03 HD213240-b 4.72 1.92 0.42 Table 4. Target star elemental abundances in standard logarithmic units, normalized to H and Solar values. A − indicates that a value was not available for a given element or star. See text for references. 55Cnc Gl777A HD4203 HD17051 HD19994 HD27442 HD72659 HD108874 HD177830 HD213240 [Fe/H] [C/H] [O/H] [Na/H] [Mg/H] [Al/H] [Si/H] [S/H] [Ca/H] [Ti/H] [Cr/H] [Ni/H] 0.33 0.31 0.13 0.26 0.48 0.47 0.29 0.12 0.08 0.36 0.22 0.31 0.24 0.29 0.22 0.26 0.33 0.34 0.24 0.10 0.11 0.32 0.17 0.25 0.40 0.45 0.00 0.42 0.48 0.51 0.44 0.20 0.24 0.41 0.33 0.42 0.24 0.39 0.11 0.48 0.21 0.32 0.23 −0.05 0.17 0.18 0.20 0.27 0.03 −0.11 0.11 0.07 0.11 0.07 0.05 −0.23 −0.03 0.13 −0.01 0.01 0.23 0.21 −0.10 0.13 0.27 0.30 0.14 − 0.01 0.20 0.15 0.18 0.33 0.48 0.38 0.37 0.56 0.54 0.31 − −0.07 0.31 0.13 0.39 0.26 0.28 0.16 0.24 0.19 0.19 0.19 0.00 0.17 0.26 0.16 0.19 0.17 0.17 0.35 0.22 0.23 0.20 0.09 −0.10 0.04 0.12 0.05 0.13 0.39 0.34 0.36 0.41 0.51 0.53 0.47 − 0.12 0.39 0.21 0.36 – 1 1 6 – Table 5. Target star elemental abundances as number of atoms and normalized to 106Si atoms. A − indicates that a value was not available for a given element or star. See text for references. Element Solar 55Cnc Gl777A HD4203 HD17051 HD19994 HD27442 HD72659 HD108874 HD177830 HD213240 Fe C O Na Mg Al Si S Ca Ti Cr Ni 8.32 ×105 9.12 ×105 8.32 ×105 7.59 ×105 9.77 ×105 8.51 ×105 6.92 ×105 7.94 ×105 10.23 ×106 8.71 ×105 10.00 ×106 1.02 ×107 1.07 ×107 1.15 ×107 1.05 ×107 1.26 ×107 1.48 ×107 7.59 ×106 7.08 ×106 1.20 ×107 1.51 ×107 1.23 ×107 1.55 ×107 1.07 ×107 1.48 ×107 5.62 ×106 1.45 ×107 1.17 ×107 1.20 ×107 1.78 ×107 8.91 ×106 1.82 ×107 2.82 ×107 6.03 ×104 5.62 ×104 6.31 ×104 5.75 ×104 6.76 ×104 1.07 ×105 5.25 ×104 6.31 ×104 5.89 ×104 6.92 ×104 8.13 ×104 1.07 ×106 1.66 ×106 1.32 ×106 1.17 ×106 1.07 ×106 1.02 ×106 1.17 ×106 1.23 ×106 1.45 ×106 1.91 ×106 1.48 ×106 8.32 ×104 1.26 ×105 1.05 ×105 9.77 ×104 8.32 ×104 1.02 ×105 9.55 ×104 8.71 ×104 1.20 ×105 1.41 ×105 1.07 ×105 1.00 ×106 1.00 ×106 1.00 ×106 1.00 ×106 1.00 ×106 1.00 ×106 1.00 ×106 1.00 ×106 1.00 ×106 1.00 ×106 1.00 ×106 4.57 ×105 3.09 ×105 3.31 ×105 2.63 ×105 2.95 ×105 2.40 ×105 − − 4.57 ×105 4.57 ×105 − 6.46 ×104 3.98 ×104 4.79 ×104 4.07 ×104 6.17 ×104 5.62 ×104 2.88 ×104 5.37 ×104 4.79 ×104 2.69 ×104 5.75 ×104 2.75 ×103 3.24 ×103 3.31 ×103 2.57 ×103 3.24 ×103 2.45 ×103 2.29 ×103 3.31 ×103 3.16 ×103 2.75 ×103 2.95 ×103 1.32 ×104 1.12 ×104 1.12 ×104 1.02 ×104 1.23 ×104 1.23 ×104 7.24 ×103 1.15 ×104 1.35 ×104 0.87 ×103 1.20 ×104 5.01 ×104 5.25 ×104 5.13 ×104 4.79 ×104 5.01 ×104 5.50 ×104 3.89 ×104 4.57 ×104 5.50 ×104 6.03 ×104 5.50 ×104 – 1 1 7 – – 118 – Table 6. Chemical species included in the equilibrium calculations of HSC Chemistry. Gaseous Species CrO MgOH CrOH MgS CrS Fe FeH FeO FeOH FeS H H2 HCN HCO H2O HPO HS H2S He Mg MgH MgN MgO N N2 NH3 NO NS Na Na2 NaH NaO NaOH Ni NiH NiO NiOH NiS O O2 P PH Al AlH AlO Al2O AlOH AlS C CH4 CN CO CO2 CP CS Ca CaH CaO CaOH CaS Cr CrH CrN PN PO PS S S2 SN SO SO2 Si SiC SiH SiN SiO SiP SiP2 SiS Ti TiN TiO TiO2 TiS Solid Species Al2O3 FeSiO3 CaAl2Si2O8 C CaAl12O19 Fe3P NaAlSi3O8 SiC – 119 – Table 6-Continued Gaseous Species Fe3C Cr2FeO4 Ca3(PO4)2 FeS Fe3O4 TiC TiN AlN CaS Mg3Si2O5(OH)4 MgS H2O Ti2O3 CaTiO3 Ca2Al2SiO7 MgAl2O4 Mg2SiO4 MgSiO3 Fe Ni P Si Cr Fe2SiO4 CaMgSi2O6 Table 7. HSC Chemistry input values for the extrasolar planetary systems studied. All inputs are entered into the simulations of HSC Chemistry their elemental and gaseous state. All values are in moles and are based on the stellar abundance values listed in Tables 4 and 5. Element System 55Cnc Gl777A HD4203 HD17051 HD19994 HD27442 HD72659 HD108874 HD177830 HD213240 H He C N O Na Mg Al Si P S Ca Ti Cr Fe Ni 1.0 ×1012 1.0 ×1012 1.0 ×1012 1.0 ×1012 1.0 ×1012 1.0 ×1012 1.0 ×1012 1.0 ×1012 1.0 ×1012 1.0 ×1012 8.5 ×1010 8.5 ×1010 8.5 ×1010 8.5 ×1010 8.5 ×1010 8.5 ×1010 8.5 ×1010 8.5 ×1010 8.5 ×1010 8.5 ×1010 7.4 ×108 7.1 ×108 1.0 ×109 6.9 ×108 8.9 ×108 7.9 ×108 2.8 ×108 5.9 ×108 1.1 ×109 5.4 ×108 1.8 ×108 1.7 ×108 2.5 ×108 1.7 ×108 2.2 ×108 2.0 ×108 0.7 ×108 1.5 ×108 2.7 ×108 1.3 ×108 7.4 ×108 9.1 ×108 5.5 ×108 7.9 ×108 7.1 ×108 1.3 ×109 7.1 ×108 4.4 ×108 1.3 ×109 1.2 ×109 3.9 ×106 3.9 ×106 5.6 ×106 3.7 ×106 6.5 ×106 5.5 ×106 2.5 ×106 2.9 ×106 5.0 ×106 3.6 ×106 – 1 2 0 – 1.2 ×108 8.1 ×107 1.2 ×108 5.9 ×107 6.2 ×107 1.2 ×108 4.9 ×107 7.1 ×107 1.4 ×108 6.5 ×107 8.7 ×106 6.5 ×106 9.6 ×106 4.6 ×106 6.2 ×106 1.0 ×107 3.5 ×106 5.9 ×106 1.0 ×107 4.7 ×106 6.9 ×107 6.2 ×107 9.8 ×107 5.5 ×107 6.0 ×107 1.1 ×108 4.0 ×107 4.9 ×107 7.2 ×107 4.4 ×107 8.5 ×105 6.3 ×105 9.3 ×105 4.5 ×105 6.0 ×105 9.8 ×105 3.4 ×105 5.8 ×105 1.0 ×106 4.6 ×105 2.1 ×107 2.0 ×107 2.6 ×107 1.6 ×107 1.4 ×107 4.5 ×107 1.0 ×107 2.2 ×107 3.3 ×107 1.3 ×107 2.8 ×106 3.0 ×106 4.0 ×106 3.4 ×106 3.4 ×106 3.0 ×106 2.1 ×106 2.3 ×106 1.9 ×106 2.5 ×106 2.2 ×105 2.0 ×105 2.5 ×105 1.8 ×105 1.5 ×105 2.4 ×105 1.3 ×105 1.6 ×105 2.0 ×105 1.3 ×105 7.8 ×105 6.9 ×105 1.0 ×106 6.8 ×105 7.4 ×105 7.6 ×105 4.6 ×105 6.6 ×105 6.3 ×105 5.2 ×105 6.3 ×107 5.1 ×107 7.4 ×107 5.4 ×107 5.1 ×107 7.2 ×107 3.2 ×107 5.0 ×107 6.3 ×107 4.4 ×107 3.6 ×106 3.2 ×106 4.7 ×106 2.8 ×106 3.3 ×106 4.1 ×106 1.8 ×106 2.7 ×106 4.4 ×106 2.4 ×106 – 121 – – 122 – Table 8: Stellar accretion rates for the extrasolar host stars studied. Stellar masses were obtained from the Simbad database. Solar values are the nominal model determined by Hersant et al. (2001). See text for details on the scaling relations applied. System Stellar Mass M (MJ) (MJ/year) Solar 1.00 5.00×10−6 55Cnc 1.03 5.23×10−6 Gl777 1.04 5.30×10−6 HD4203 1.06 5.46×10−6 HD17051 1.11 5.50×10−6 HD19994 1.35 7.84×10−6 HD27442 1.20 6.00×10−6 HD72659 0.95 4.63×10−6 HD108874 1.00 5.00×10−6 HD177830 1.48 9.00×10−6 HD213240 1.22 6.10×10−6 – 123 – Table 9: Convective zone masses for each of the target stars. Tef f values taken from Santos et al. (2004). See text for details on the determination of MCZ. System Tef f MCZ (K) (MJ) 55Cnc 5279 0.0398 Gl777 5584 0.0316 HD4203 5636 0.0288 HD17051 6253 0.0035 HD19994 6190 0.0045 HD27442 4825 0.0562 HD72659 5995 0.0112 HD108874 5596 0.0321 HD177830 4804 0.0501 HD213240 5984 0.0100 Table 10. T50% condensation, C/O and Mg/Si for extrasolar planetary systems studied in order of increasing C/O value. Solar values are also shown for comparison and are taken from Bond et al. (2009). Where less than 50% of an element condensed within a system, the maximum amount of condensation that occurred is shown in parentheses. All values are in K. System HD72659 HD213240 HD27442 Solar Gl777A HD177830 HD17051 55Cnc HD19994 HD108874 HD4203 – 1 2 4 – 1665 <150 1698 <150 1688 1639 1636 <150 <150 <150 1541 1297 1334 1359 927 1346 180 1312 625 1349 1593 0.40 1.23 1562 1309 1346 1381 944 1358 183 1307 641 1374 1610 0.44 1.48 1613 1527 1327 1301 1365 1339 1377 1339 935 941 1377 1351 183 180 1327 1309 709 658 1360 1329 1590 1580 0.63 1.17 0.66 1.07 1562 1314 1353 1328 870 1364 181 1346 664 1306 1574 0.78 1.32 1647 976 (6%) 1556 1322 1361 1321 801 1373 184 1353 691 1315 1567 0.83 1.91 1551 1311 955 920 1320 954 (9%) (34%) (39%) 1484 1316 1354 1224 866 1248 1322 1361 1081 864 1365 1372 180 207 1359 1375 652 1125 1548 667 1169 1754 0.87 1.07 1.00 1.66 1287 1313 1353 1075 873 1364 180 1392 646 1626 1810 1.26 1.02 1315 946 1289 1306 1352 1066 854 1350 331 1391 669 1520 1810 1.35 1.45 1333 1049 1300 1324 1367 1074 864 1378 1001 1407 677 1645 1832 1.86 1.29 Element Al C Ca Cr Fe Mg Na Ni O P S Si Ti C/O Mg/Si – 125 – Table 11. Predicted bulk elemental abundances for all simulated extrasolar terrestrial planets. All values are in wt% of the final predicted planet for all seven sets of disk conditions examined. Planet number increases with increasing distance from the host star. System H Mg O S Fe Al Ca Na Ni Cr P Ti Si C 55Cnc t=2.5×105 years 55Cnc 1-4 0.00 16.59 26.02 2.58 20.92 1.40 0.65 0.51 1.27 0.24 0.16 0.06 11.54 0.00 18.06 55Cnc 2-4 0.00 16.50 25.88 2.93 20.79 1.39 0.65 0.51 1.26 0.24 0.16 0.06 11.47 0.00 18.18 55Cnc 3-4 0.00 16.80 26.21 2.20 21.14 1.41 0.71 0.52 1.28 0.24 0.16 0.06 11.73 0.00 17.53 55Cnc 3-5 3.71 12.59 53.11 3.09 15.88 1.06 0.50 0.00 0.96 0.18 0.12 0.05 8.76 0.00 0.00 – 1 2 6 – 55Cnc 4-4 1.79 15.74 43.90 3.87 19.87 1.32 0.62 0.31 1.20 0.23 0.15 0.06 10.95 0.00 0.00 Note. - Table 11 is published in its entirety in the electronic edition of the Astrophysical Journal. A portion is shown here for guidance regarding its form and content. Table 12. Average change in host star photospheric abundances produced by terrestrial planet formation. Values are based on the results of four separate simulations. System Mass Accreted Change in Abundance (M L ) 0.145 2.100 0.631 0.727 0.898 0.342 2.640 0.755 0.237 1.310 55Cnc Gl777 HD4203 HD17051 HD19994 HD27442 HD72659 HD108874 HD177830 HD213240 Mg O S Fe Al Ca Na Ni Cr P Ti Si C 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.011 0.009 0.009 0.009 0.011 0.011 0.011 0.011 0.011 0.004 0.012 0.011 0.009 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.002 0.001 0.001 0.004 0.004 0.004 0.004 0.013 0.016 0.004 0.005 0.005 0.000 0.020 0.005 0.004 0.004 0.003 0.003 0.003 0.005 0.005 0.003 0.005 0.005 0.000 0.006 0.006 0.005 0.001 0.001 0.000 0.001 0.001 0.001 0.001 0.001 0.001 0.000 0.001 0.001 0.001 0.021 0.017 0.001 0.016 0.020 0.030 0.023 0.021 0.025 0.007 0.026 0.023 0.016 – 1 2 7 – 0.001 0.001 0.001 0.001 0.002 0.001 0.001 0.002 0.002 0.000 0.003 0.002 0.001 0.001 0.001 0.001 0.001 0.001 0.001 0.001 0.001 0.001 0.000 0.001 0.001 0.001 0.015 0.012 0.005 0.011 0.019 0.024 0.014 0.015 0.016 0.011 0.021 0.014 0.011
1010.1753	1	1010	2010-10-08T18:05:50	A new look at NICMOS transmission spectroscopy of HD189733, GJ-436 and XO-1: no conclusive evidence for molecular features	[ "astro-ph.EP" ]	We present a re-analysis of archival HST/NICMOS transmission spectroscopy of three exoplanet systems; HD 189733, GJ-436 and XO-1. Detections of several molecules, including H20, CH4 and CO2, have been claimed for HD 189733 and XO-1, but similarly sized features are attributed to systematic noise for GJ-436. The data consist of time-series grism spectra covering a planetary transit. After extracting light curves in independent wavelength channels, we use a linear decorrelation technique account for instrumental systematics (which is becoming standard in the field), and measure the planet-to-star radius ratio as a function of wavelength. For HD 189733, the uncertainties in the transmission spectrum are significantly larger than those previously reported. We also find the transmission spectrum is considerably altered when using different out-of-transit orbits to remove the systematics, when some parameters are left out of the decorrelation procedure, or when we perform the decorrelation with quadratic functions rather than linear functions. Given that there is no physical reason to believe the baseline flux should be modelled as a linear function of any particular set of parameters, we interpret this as evidence that the linear decorrelation technique is not a robust method to remove systematic effects from the light curves for each wavelength channel. For XO-1, the parameters measured to decorrelate the light curves would require extrapolation to the in-transit orbit to remove the systematics, and we cannot reproduce the previously reported results. We conclude that the resulting NICMOS transmission spectra are too dependent on the method used to remove systematics to be considered robust detections of molecular species in planetary atmospheres, although the presence of these molecules is not ruled out.	astro-ph.EP	astro-ph	Mon. Not. R. Astron. Soc. 000, 1 -- 17 (2002) Printed 7 October 2018 (MN LATEX style file v2.2) A new look at NICMOS transmission spectroscopy of HD189733, GJ-436 and XO-1: no conclusive evidence for molecular features N. P. Gibson1,2(cid:63), F. Pont2 and S. Aigrain1,2 1Department of Physics, University of Oxford, Denys Wilkinson Building, Keble Road, Oxford OX1 3RH, UK 2School of Physics, University of Exeter, Exeter, EX4 4QL, UK Submitted July 8th 2010 ABSTRACT We present a re-analysis of archival HST/NICMOS transmission spectroscopy of three exoplanet systems; HD 189733, GJ-436 and XO-1. Detections of several molecules, including H20, CH4 and CO2, have been claimed for HD 189733 and XO-1, but similarly sized features are attributed to systematic noise for GJ-436. The data consist of time-series grism spectra covering a planetary transit. After extracting light curves in independent wavelength channels, we use a linear decorrelation technique account for instrumental systematics (which is becoming standard in the field), and measure the planet-to-star radius ratio as a function of wavelength. We use a resid- ual permutation algorithm to calculate the uncertainties, in an effort to evaluate the effects of systematic noise on the resulting transmission spectra. For HD 189733, the uncertainties in the transmission spectrum are significantly larger than those previ- ously reported. We also find the transmission spectrum is considerably altered when using different out-of-transit orbits to remove the systematics, when some parame- ters are left out of the decorrelation procedure, or when we perform the decorrelation with quadratic functions rather than linear functions. Given that there is no physical reason to believe the baseline flux should be modelled as a linear function of any par- ticular set of parameters, we interpret this as evidence that the linear decorrelation technique is not a robust method to remove systematic effects from the light curves for each wavelength channel. For XO-1, the parameters measured to decorrelate the light curves would require extrapolation to the in-transit orbit to remove the system- atics, and we cannot reproduce the previously reported results. We conclude that the resulting NICMOS transmission spectra are too dependent on the method used to re- move systematics to be considered robust detections of molecular species in planetary atmospheres, although the presence of these molecules is not ruled out. Key words: methods: data analysis, stars: individual (HD 189733), stars: individual (GJ-436), stars: individual (XO-1), planetary systems, techniques: spectroscopic 0 1 0 2 t c O 8 . ] P E h p - o r t s a [ 1 v 3 5 7 1 . 0 1 0 1 : v i X r a 1 INTRODUCTION The number of known exoplanets is increasing rapidly, re- vealing a diverse range of systems with vastly different prop- erties. Transiting exoplanets offer a unique opportunity to study the structure and composition of planets other than those in our own solar system, as transit light curves allow a measurement of the radius of the planet, and the comple- mentary radial velocity technique measures the mass. From (cid:63) E-mail: [email protected] c(cid:13) 2002 RAS the derived density, the bulk composition of the planet may be inferred. Transiting planets also provide the opportunity to mea- sure the composition of planets' atmospheres. Typically, transit light curves are modelled assuming the planet as an opaque disk, whose size is defined by the altitude at which the atmosphere becomes opaque to starlight. However, the optical depth in the atmosphere is wavelength dependent, being sensitive to atomic and molecular absorption. There- fore, the transit depth and measured radius of a planet are wavelength dependent, and measuring the planetary radius 2 N. P. Gibson et al. as a function of wavelength may allow the detection of ab- sorption features in the atmospheres, and thus enable obser- vations to infer the presence of atomic and molecular species (e.g. Seager & Sasselov 2000; Brown 2001). Hot Jupiters are a class of planets with masses similar to Jupiter and on very short-period orbits, therefore they are intensely irradiated resulting in very hot atmospheres. Consequently, their large atmospheric scale heights make them particularly good tar- gets for this type of measurement. There are two approaches to measuring the wavelength dependence of a planet's radius. Transmission spectroscopy consists of monitoring the transit with spectroscopic mea- surements, which can be subsequently split into separate light curves for each wavelength channel, allowing a mea- surement of transit depth at each wavelength. This has been used to detect various species in the atmospheres of hot-Jupiters, including HD 209458 (e.g. Charbonneau et al. 2002, Na) and HD 189733 (e.g. Swain et al. 2008, H2O, CH4), and has also suggested a haze in the upper atmosphere of HD 189733 (Pont et al. 2008) at optical wavelengths. An- other method to measure the wavelength dependence of a planet's radius, is to make multi-colour photometric obser- vations of the transit light curve, and measure the transit depth for each. Spitzer observations of this type were used to infer the presence of H2O in the atmosphere of HD 189733 (Tinetti et al. 2007), although this is disputed by Ehrenreich et al. (2007) and D´esert et al. (2009). These measurements have provided a wealth of infor- mation which feeds into theories of planetary atmospheres. However, both methods have advantages and disadvantages, and have been the subject of some controversy. For exam- ple, photometric measurements are much more straight for- ward, and the data reduction and analysis use established methods, and are therefore relatively robust against instru- mental systematic effects. However, observations are (typ- ically) taken during different transits, and stellar activity may cause variations of the measured planetary radius due to variable spots on the surface of the star, which change as the star rotates and reveals different hemispheres, and also evolve on timescales comparable to the period of typi- cal transiting planets (Mosser et al. 2009). Transmission spectroscopy on the other hand, avoids this problem by simultaneously monitoring the transit light curve at different wavelengths. Whilst there is still a chro- matic variance due to any spots on the stellar surface dur- ing a particular transit, this is a much smaller effect than if the surface spot distribution has changed. However, trans- mission spectroscopy is usually affected by systematic noise from the telescope and instrument (e.g. Pont et al. 2008; Swain et al. 2008), often larger than the signal we are try- ing to measure. This is commonly treated by constructing a multi-linear model of the baseline function, which describes how the measured light curve of the star would behave in the absence of a planetary transit due to changes in the state of the optics and detector. This is built from a linear combination of parameters that describe the optical state of the system, such as the position of the spectral trace on the detector, the temperature of the detector, and in the case of Hubble Space Telescope (HST) observations, the or- bital phase of the observatory. The baseline function is de- termined from the out-of-transit observations, and then used to decorrelate the light curve by projecting the function to the in-transit observations; however, the physical origins of these linear decorrelation models are poorly understood, and as we will see later, the choice of model has significant effects on the output transmission spectra. HST transmission spectroscopy with NICMOS (Near Infrared Camera and Multi-Object Spectrometer) has led to some of the most detailed studies of exoplanet systems, but it in particular suffers from these systematics effects (e.g. Swain et al. 2008; Pont et al. 2009), which are comparable to if not larger than the expected signal due to molecu- lar absorption. Of the NICMOS observations currently in the literature, one group have claimed the detection of sev- eral molecules in the atmospheres of various hot Jupiters (e.g. Swain et al. 2008, HD 189733; Tinetti et al. 2010, XO- 1). However, other analyses of NICMOS transmission spec- troscopy, have been unable to untangle the instrumental sys- tematic effects from wavelength dependent absorption from the planet (e.g. Pont et al. 2009; Carter et al. 2009). Fur- thermore, for HD 189733, there have been significant dis- agreements with wavelength dependent photometric mea- surements of the light curve and transmission spectroscopy, with Sing et al. (2009) failing to detect water in the atmo- sphere of HD 189733 previously reported by Swain et al. (2008), using narrow band photometric measurements with NICMOS. Tinetti et al. (2010) attribute this to variable spot distributions for each of transit, arguing multi-colour photo- metric observations cannot reach the required accuracy for molecular spectroscopy, despite Sing et al. (2009) account- ing for this using continuous ground-based monitoring of the stellar flux. As these results are vital to our understanding of plan- etary atmospheres, we must explore the possibility that the wavelength dependance of the planetary radius is no more than systematic noise in the detector. As the HST orbits with a period of ∼96 minutes, transit light curves are split into discrete orbits, typically sampling several out-of-transit portions of the light curve, used to establish the baseline function, and 1 -- 2 in-transit portions. Trends are typically seen in each orbit, related to the orbital phase of the HST and its effects on the position of the spectral trace. Fur- thermore, as the optical system cannot be reset to exactly the same configuration for every orbit, systematic offsets be- tween the flux levels in each orbit may also result. Efforts to gain a deeper understanding of NICMOS systematics and obtain photon limited photometry are ongoing (e.g. Burke et al. 2010). However, the offsets are not satisfactorily ad- dressed in much of the literature, as any systematic offset may not be corrected for or indeed may be over-corrected for when the baseline function is projected to the in-transit portion of the light curve in the case of an imperfect baseline model. These in-transit offsets will be hidden from the light curve residuals when fitting the transit depth, and therefore not taken into account in the error analysis. Fig. 1 shows a plot of the NICMOS spectra of transit- ing planets reported in the literature to date1; transmission spectra for HD 189733 (Swain et al. 2008), GJ-436 (Pont et al. 2009), and XO-1 (Tinetti et al. 2010), and dayside 1 For XO-1 and HD 209458 secondary, only every 5th point is plotted for comparison, because a boxcar smoothing was applied to the published data. c(cid:13) 2002 RAS, MNRAS 000, 1 -- 17 A new look at NICMOS transmission spectroscopy 3 2 HD 189733 2.1 Observations A transit of HD 189733 was monitored on 25 May 2007 with HST/NICMOS, using the G206 grism covering the wave- length range 1.4 -- 2.5 µm. Analyses of these data were first reported in Swain et al. (2008, hereafter S08). As HD 189733 is not in the continuous viewing zone of the HST, the tran- sit was observed over five half-orbits (∼ 48 minutes each), consisting of 638 spectra in total, all with exposure times of 1.624 seconds. The first, second, fourth and fifth orbits cover the out-of-transit part of the light curve and consist of 119, 128, 131 and 130 observations, respectively. These data are required to determine the photometric baseline. Only the third orbit was taken in-transit, and consists of the remaining 130 images. In addition to the spectra, some exposures were taken at the beginning of the first orbit to enable wavelength calibration. The calibrated images, one of which is displayed in Fig. 2, include all basic calibrations except for flat-fielding. Correctly flat fielding spectroscopic data requires taking into account the wavelength dispersion and position of the source, and therefore a different flat-field correction would be required at each point along the spectral trace of the tar- get. S08 argue that flat-fielding is not required for differential spectroscopy, as each light curve will be normalised before measurement of parameters from the light curves. However, strictly speaking flat-fielding is required to accurately de- termine the background value in each wavelength channel. This is particularly important, as underestimating or overes- timating the background results in a variable transit depth, which is exactly what we are trying to measure. Further- more, the G206 grism has a relatively high background com- pared with NICMOS's other two grisms (about 800 electrons per pixel per exposure), due to thermal background emission from HST. Whilst the background is very stable temporally, unfortunately, it does significantly vary spatially over the detector by as much as 200 electrons per pixel per expo- sure. This will particularly effect the light curves extracted from the edge of the spectrum, where the overall counts are much lower. We did try flat-fielding each image using a flat-field taken with the G206 grism in place. This cor- rects for diffuse light over the whole detector. This will not effect the depth of the final normalised light curves except through the background value. This process does not pro- vide a satisfactory background correction, and the analyses described below were attempted with and without flat-field corrections. We also note that the background varies quite smoothly, and therefore is unlikely to be responsible for any narrow 'features' seen in the transmission spectrum, but can certainly affect the overall depth and shape. Prior to extracting the spectra, we identified the defec- tive pixels flagged by the calibration pipeline, and additional significant outliers, and corrected for them by replacing the pixel value with an interpolation of the surrounding eight pixels. A relatively small number of pixels required correc- tions, and in fact this process has very little effect on the output spectra. A 1D spectrum was then extracted as follows for each image. For each column along the spectral trace in the dispersion axis (x), a centroid along the spatial axis (y) was calculated to determine the position of the spectrum. A sum of 35 pixels along each column centred on this position Figure 1. Plot of some of the NICMOS transmission and emis- sion spectra reported in the literature; transmission spectra for HD 189733 (Swain et al. 2008), GJ-436 (Pont et al. 2009), and XO-1 (Tinetti et al. 2010), and dayside emission spectra for HD 189733 (Swain et al. 2009b) and HD 209458 (Swain et al. 2009a). Each spectrum is plotted as % absorption as a function of wave- length, but offset for clarity. The peak-to-peak amplitudes of all reported spectra are approximately equal, and also show similar features. emission spectra for HD 189733 (Swain et al. 2009b) and HD 209458 (Swain et al. 2009a). The emission spectra are ob- tained in much the same way as the transmission spectra, by observing an eclipse of the planet, and using the same linear decorrelation methods to correct for the baseline function. It is clear that all five spectra show approximately the same peak-to-peak amplitude, and have wavelength dependance on similar scales. This is perhaps surprising, given the plan- ets have vastly different masses, radii and irradiation levels. Transmission and emission spectra also show the same am- plitudes. Whilst this is by no means evidence that the spec- tra are the result of systematic noise in the detector, and the similarity could be coincidental, it is nonetheless wise to explore this possibility. It is therefore important that the re- sults reported from NICMOS transmission spectroscopy be re-analysed to confirm previous detections of molecules, in particular because the results are dependent on a complex decorrelation procedure, with an arguably ad hoc model. Here, we present a re-analysis of the three NICMOS transmission spectra shown in Fig. 1. Sects. 2, 3 and 4 de- scribe the data reduction, analyses and results of the HD 189733, GJ-436, and XO-1 data, respectively. Finally, Sect. 5 summarises and presents our conclusions. c(cid:13) 2002 RAS, MNRAS 000, 1 -- 17 1.01.52.02.5wavelength (µm)0.40.30.20.10.00.10.20.30.4Delta Absorption / %HD 189733 transitGJ-436 transitXO-1 transitHD 209458 secondaryHD 189733 secondary4 N. P. Gibson et al. Figure 2. NICMOS image of HD 189733 taken with the G206 grism. The blue line marks the centroids of the first-order spec- trum of HD 189733, with the green lines flanking representing the extent of the extraction region. Immediately left is the second- order spectrum, and below is a companion star to HD 189733. The bright band at the right of the detector is a feature of G206 grism spectra, caused by the warm edge of the aperture mask. was used to extract the flux for each wavelength channel, after subtraction of a background value for each pixel. A width of 35 pixels was chosen to minimise the RMS in the white light curve of orbits 2, 4 and 5. This was repeated for 90 pixel columns along the dispersion axis. The extraction regions used are marked in Fig. 2. For the background sub- traction, we experimented with various techniques. First, a global background subtraction was used, similarly to S08. The background was taken to be the average of a large un- illuminated region above the spectral trace (we tested using different regions). This is not ideal, as previously mentioned the background varies spatially over the detector. As a first order correction, we instead calculated the background sep- arately for each pixel column, as the average value of the un-illuminated region above the spectral trace along the col- umn. Again we note that this does not provide a satisfac- tory correction, as the background varies along both x and y. Both global and wavelength dependent corrections were used for the subsequent analysis. For the remainder of this paper, we will display results from data that is not flat-fielded, and using separate columns for background subtraction, but our conclusions remain the same for each case. Extracted spectra from a typical in-transit and out-of- transit observation are shown in Fig. 3, giving approximately 430 000 electrons in the brightest pixel channel, and approx- imately 120 000 electrons in the faintest channel. The fea- tures in the spectra do not correspond to real stellar fea- tures, but result in variation of the sensitivity of the grism with wavelength. Each wavelength channel in the 1D spec- tra is then used to construct a time-series, after binning in 5 pixels along the dispersion direction, resulting in 18 light Figure 3. Extracted 1D spectra of a typical in-transit (green) and out-of-transit (blue) observation of HD 189733 showing the number of electrons collected per pixel channel. The wavelength decreases with increasing x position. Figure 4. Raw 'white' light curve of HD 189733 obtained by in- tegrating the flux from each spectra over all wavelengths, showing the sampling of the transit, alongside its best-fit model. While all orbits are seen to suffer from systematics, orbit one exhibits by far the largest, commonly attributed to spacecraft 'settling', and is excluded from subsequent analysis. curves. A 'white' light curve was constructed by integrating the flux over the entire wavelength range for each image. This was extended to include 110 pixel columns, which min- imises systematics arising from small changes in the position of the spectral trace. This is plotted in Fig. 4, and shows the sampling of the light curve over the five orbits. Systemat- ics are evident in each orbit, but particularly for orbit one. This is commonly found in similar NICMOS data (see e.g. Sects. 3.1 and 4.1). It is attributed to spacecraft 'settling' (e.g. S08; Pont et al. 2009), and orbit one is excluded for the remainder of this work. The raw light curves are shown in Fig. 5 for each of the 18 wavelength channels, after normalising by fitting a line through orbits 2, 4 and 5, and dividing the light curves by this function. The light curves are clearly seen to exhibit strong time-correlated noise, which needs to be removed if c(cid:13) 2002 RAS, MNRAS 000, 1 -- 17 050100150200250x pixel050100150200250y pixel100120140160180x position (pixels)100150200250300350400450counts (electrons×103)246.00246.05246.10246.15246.20246.25246.30HJD - 24540000.9750.9800.9850.9900.9951.000relative fluxA new look at NICMOS transmission spectroscopy 5 Figure 5. Raw light curves of HD 189733, for each of the 18 wavelength channels from 2.50 µm (top) to 1.48 µm (bottom), af- ter normalising by fitting a linear function through orbits 2, 4 and 5. Orbit 1 is not plotted as it is discarded for the analysis. Clearly, time-correlated noise is present in each light curve, and must be removed to measure the transmission spectrum. The dashed grey lines show a transit model generated for HD 189733, used to guide the eye. we are to measure the transmission spectrum at the level re- quired to detect molecular features. The systematics may be understood to arise from small motions of the spectra across the detector, related to the orbital motion of the HST. Refer- ring back to Fig. 3, the wavelike features in the spectra will move into different wavelength channels and cause features on short wavelength intervals, even if the stellar spectrum and flat-fielding are smooth. The optical state parameters, described by S08, were therefore measured in an effort to model and remove these systematic effects. We extracted the shift of the spectral trace along the x axis (∆X) by cross-correlation of the 1D spectra, the shift of the position of the y axis by averaging the centroids determined earlier (∆Y ), and the angle the spectral trace makes with respect to the x-axis (θ) by fitting a line to the centroids of the spectra. The width (W ) of each spectrum was also measured by fitting gaussian functions along each extraction column, and taking the average. In the absence of a direct measurement of the temperature (T ) of the detector, or a proxy calculated from the bias levels (Gilliland & Arribas 2003), a proxy for this was taken as the temperature of the NIC1 mounting cup (S08). However, this is not monitored at a high enough precision to accurately monitor the temperature. The temperature is an important c(cid:13) 2002 RAS, MNRAS 000, 1 -- 17 Figure 6. Extracted state parameters for HD 189733 observa- tions for orbits 2 -- 5. For orbit 3 we reconstruct the model baseline as a linear function of all state parameters, plus a constant flux level, and the orbital phase of the HST and its square. The co- efficients for each parameter of the model are determined from orbits 2, 4 and 5. parameter to describe the state of a NIR detector, and this is likely an important limitation of NICMOS analysis. A plot of the optical state parameters and the detector temperature is shown in Fig. 6. It is important to note that we extract our parameters slightly differently from S08 in some cases. In particular we used an averaged value for the shifts in position ∆X and ∆Y per image, in order to monitor the overall optical state of the detector, and therefore had identical parameters for all wavelength channels, unlike S08. S08 determined x and y positions by fitting spectra along diagonals and rotating back along the two nominal axis, resulting in different pa- rameters for each wavelength channel, and in some cases different trends. In either case, the parameters show a sim- ilar dispersion and amplitude, and should allow a similar removal of the systematics. 2.2 Analysis Each light curve was decorrelated using the multi-linear baseline model from S08. This assumes that the baseline function can be constructed by a linear combination of the optical state vectors ∆X, ∆Y , W and θ, the proxy for the detector temperature T , and also the orbital phase of the HST (φH ) and its square (φ2 H ). In addition, we used a con- stant vector to represent the out-of-transit flux level (f0), 246.05246.10246.15246.20246.25246.30[HJD] - 24540000.750.800.850.900.951.00relative flux-0.08-0.06-0.04-0.020.000.02∆X (pix)64.9665.0065.0465.0865.12∆Y (pix)2.802.842.882.92W (pix)1.881.901.921.941.961.982.00θ (deg)246.05246.10246.15246.20246.25246.30[HJD] - 245400077.0877.1077.1277.1477.16T (K)6 N. P. Gibson et al. as clearly the overall flux level is not given by a linear func- tion of the state variables, but rather the state variables are used to correct for variations from the expected out- of-transit flux. These parameters are hereafter collectively referred to as the 'decorrelation parameters'. For each wave- length channel, the out-of-transit baseline flux is given by N(cid:88) yi = βkXi,k + i k=1 at time i, where βk are the coefficients for each of the N decorrelation parameters (in this case; f0, ∆X, ∆Y , W , θ, T , φH and φ2 H ), Xi,k is the value of each decorrelation pa- rameter k for time i, and i the corresponding residual. This may be easily written in matrix form as (cid:126)y = X(cid:126)β + (cid:126), where X is the state matrix containing the measured de- correlation parameters, (cid:126)β is a vector containing the coeffi- cients of each decorrelation parameter, and (cid:126) are the residu- als. The best-fit coefficients β are then found by linear least squares β = (XT X) −1XT (cid:126)y, for orbits 2, 4 and 5 only. The model baseline function (cid:126)b can then be reconstructed for all orbits from a linear combina- tion of β and X; (cid:126)b = X β. Each light curve is then decorrelated by dividing through by the model baseline function. This technique can only be expected to work if we interpolate for the in-transit orbit, and if the baseline function is well represented by the linear model over the range of the decorrelation parameters. For this analysis, each light curve is treated indepen- dently, with β calculated separately for each. An example of this decorrelation process on one of the wavelength channels is shown in Fig. 7, and all of the decorrelated light curves are shown in Fig. 8. S08, who use a similar procedure to this, conclude that it satisfactorily removes time-correlated systematics from the light curves. Whilst our results agree with this conclusion for the out-of-transit orbits, residual systematics are clearly visible in the in-transit orbit, which is the important one for determining the radius ratio. In order to obtain the transmission spectrum from the light curves, a transit model must be fitted to each decorre- lated light curve to measure the planetary radius. We used the transit model described in Gibson et al. (2008), which assumes a circular orbit to calculate the normalised sepa- ration of the star and planet centres (z) as a function of time, using the orbital parameters and the masses and radii of the star and planet. The analytic models of Mandel & Agol (2002) were then used to calculate the light curves from z, the ratio of the planet-to-star radii ρ, and the limb darkening parameters. The limb darkening parameters were calculated in each of the wavelength channels for a quadratic limb darkening law (D. Sing, private communication) using the methods described in Sing (2010). The orbital parame- ters of the system and the stellar mass and radius were held fixed at the values given by Pont et al. (2008), and the cen- tral transit time was determined from the ephemeris. Each Figure 7. Example of the decorrelation procedure on one of the wavelength channels. The top plot is of the raw light curve in elec- tron counts. The green points represent the reconstructed baseline function ((cid:126)b) used to decorrelate the light curve. The bottom plot shows the decorrelated light curve along with its best-fit transit model, used to determine the planet-to-star radius ratio. Whilst the out-of-transit orbits are usually whitened, residual system- atic noise is often seen in the in-transit orbit, showing the linear baseline model is not robust. decorrelated light curve was then fitted for ρ, to obtain the transmission spectrum. To calculate the uncertainty in ρ for each wavelength channel, we used a residual permutation (or 'prayer bead') algorithm (see e.g. Gillon et al. 2007; Southworth 2008), sim- ilar to that used in Gibson et al. (2009, 2010). This method accounts for the correlated noise in the light curve by re- constructing light curves by combining the best-fit model and its residuals, but shifting the residuals before combin- ing them to determine the effects of the correlated noise on the determined parameters. This method preserves both the correlated and random noise in the resampled light curves, and therefore both are taken into account when determining uncertainties. The residual permutation was applied to the raw light curves prior to performing the decorrelation procedure, in order to take into account uncertainties from both the linear decorrelation and the light curve fitting. The best fit baseline function (cid:126)b and transit model (cid:126)mbf were determined as before, and the residuals from the fit (cid:126)r are used to reconstruct the light curve, but each time the residuals are shifted by a random offset to give (cid:126)rp. Any residuals that fall off the 'edge' are looped back to the beginning. The new light curve (cid:126)yp is then reconstructed by adding the shifted residuals to the best-fit model, followed by a pointwise multiplication of the best-fit baseline function; (cid:126)yp = ( (cid:126)mbf + (cid:126)rp) · (X β) The decorrelation and light curve fitting is done as before, to determine ρ. The procedure is repeated 1000 times with c(cid:13) 2002 RAS, MNRAS 000, 1 -- 17 23402360238024002420counts (electrons×103)246.05246.10246.15246.20246.25246.30HJD - 24540000.950.960.970.980.991.001.01relative fluxA new look at NICMOS transmission spectroscopy 7 Figure 9. Transmission spectrum of HD 189733 generated by determining the planet to star radius ratio from the decorrelated light curves for the 18 wavelengths channels. Our results are the red points, and those from S08 are shown in grey for comparison. The horizontal dashed line is the radius ratio measured for the white light curve. larger than those given in S08, particularly at the edge of the spectrum. However, we do not believe the residual permutation method fully accounts for the uncertainties in ρ. Fig. 10 shows a plot of the in-transit residuals for three wavelength channels. The residuals are of decorrelated light curves, af- ter subtracting a model generated from the planet-to-star radius ratio measured from the white light curve, and us- ing limb darkening parameters specific to each wavelength channel. The difference between models generated for the best-fit planet-to-star radius ratio for each channel (i.e. the value in the transmission spectrum), and the planet-to-star radius ratio from the white light curve (i.e. a constant ra- dius model) are also shown. As is clear for all three channels, systematic noise is present at a level comparable to or larger than the difference between the two models. The deviations from a constant transmission spectrum may therefore arise from systematics not removed from the in-transit orbit. A similar level of systematics is visible in the residuals for all wavelength channels. We are yet to address one final correction on each wave- length channel carried out by S08, the 'channel-to-channel' corrections. This involves taking the weighted average of the light curve residuals for all the wavelength channels, and subtracting them from individual channels after decorrelat- ing the light curves. This should remove any common time- correlated systematics from the light curves. To check that this did not affect our results significantly, we carried out this step. Fig. 11 shows the same light curves as Fig. 10 af- ter the channel-to-channel correction has been applied. This process does marginally reduce the RMS of the residuals, but does not completely remove the systematics. This is not surprising given all wavelength channels do not show the same structure of systematics. We would also not expect this to change the transmission spectrum, as applying the same correction to all light curves will only shift the depths uniformly. However, reducing the RMS may have an impact on the uncertainties. The resulting transmission spectrum Figure 8. Decorrelated light curves for each of the 18 wavelength channels. Much of the time-correlated systematic noise is removed from the light curves, as seen from a direct comparison with Fig. 5. However, some of the correlated noise remains particularly in the in-transit orbit, and must be taken into account in the error anal- ysis of the transmission spectrum. The dashed grey lines again show a transit model generated for HD 189733. random perturbations to the shift in residuals, each time varying the starting value for ρ to ensure the starting pa- rameters do not affect the results. The resulting distribution of ρ is then used to estimate its uncertainty in each wave- length channel. 2.3 Results The resulting NICMOS transmission spectrum for HD 189733 is shown in Fig. 9. The data from S08 are also plot- ted for comparison, after converting from transit depth to ρ. For the most part, the spectra show the same basic shape. It is not clear exactly where the discrepancies in a few of the wavelength channels arise, but they are likely explained by one or a combination of the following; different pixel columns and widths used for the wavelength channels, different meth- ods used to determine the background, the fact that we fit for the light curve rather than just taking an average of the in-transit orbit, that the decorrelation parameters are extracted slightly differently, and finally the corrections ap- plied by S08 for limb darkening and star spots. We were unable to reproduce exactly the same results as S08 using a global background correction. Another obvious difference between the two spectra is that the uncertainties we calcu- late using the residual permutation method are significantly c(cid:13) 2002 RAS, MNRAS 000, 1 -- 17 246.05246.10246.15246.20246.25246.30[HJD] - 24540000.750.800.850.900.951.00relative flux1.41.61.82.02.22.42.6wavelength µm0.1510.1520.1530.1540.1550.1560.1570.1580.159ρ8 N. P. Gibson et al. Figure 10. Residuals of the in-transit orbit from a 'constant ra- dius' model generated from the white light curve planet-to-star radius ratio for three wavelength channels; 2.41, 2.07 and 1.66 µm from top to bottom, respectively. The limb darkening coeffi- cients used were specific to each wavelength channel. The hori- zontal black dashed line shows the constant radius model, and the dashed red line the difference between it and the best-fit model determined when fitting ρ for each wavelength channel. It is clear that systematic noise is present in the in-transit orbit for the three channels, at a level comparable to or even larger than the devia- tions from the constant radius. A similar level of systematics can be seen in all wavelength channels. is shown at the top of Fig. 12. This does not cause signif- icant changes in the transmission spectrum, or even in the calculated uncertainties. This indicates that the uncertain- ties in the transmission spectrum are probably dominated by determination of the baseline function. Recently, Burke et al. (2010) identified systematic effects that should be ac- counted for with NICMOS grism spectroscopy, in particular 'gain-like' variations that arise from seven states of the de- tector electronics. Importantly, they noted that the channel- to-channel correction would likely account for this. Significant uncertainties in the depth will also come from offsets in the baseline flux level between the orbits. Off- sets in the flux levels are seen in the raw light curves, and could easily be induced by the decorrelation model (as well as corrected for), as the decorrelation parameters themselves are seen to have offsets between orbits, therefore clearly a linear baseline model can produce 'artificial' offsets. These flux offsets were also identified by Carter et al. (2009), who tried to fit decorrelation models based on state parameters to attempt to remove the flux offsets, and concluded the procedure was unreliable with no physical justification for a Figure 11. Same as Fig. 10 but after the channel-to-channel cor- rection has been applied. In some cases the RMS of the residuals is reduced, but there are still systematics present that are compa- rable to or even larger than the deviations from a constant radius model. linear baseline model. As variations in the flux level for the in-transit orbit are fitted for when determining the plane- tary radius, the residual permutation algorithm cannot fully take this additional uncertainty into account, as the resid- uals will not contain the signal of any induced offset. To test for the possibility that the decorrelation methods may not fully correct for the flux level, we repeated the decor- relation process only using orbits 2 and 4 to determine the decorrelation coefficients. Referring to Fig. 6, we are still in- terpolating for the in-transit orbit for all the decorrelation parameters. Therefore, if the model for the baseline flux is correct, we should still see similar results. Fig. 13 shows an example of this decorrelation process. It is clear that orbit 5 has a significant offset from the baseline flux of orbits 2 and 4. This shows that the linear decorrelation method is not ro- bust when extrapolating over the decorrelation parameters. If the functions describing the baseline function are not lin- ear in nature (which is by no means clear, as even assuming the baseline flux can realistically be written in terms of the optical state parameters is a strong assumption), a spurious offset could just as easily be induced in the in-transit orbit, leading to channel dependent transit depths, and resulting in a variable transmission spectrum. The transmission spectrum obtained when decorrelat- ing only using orbits 2 and 4 is shown in Fig. 12. As expected, the uncertainties are now significantly larger than before (or- bit 5 is not used in the fitting process). However, the trans- c(cid:13) 2002 RAS, MNRAS 000, 1 -- 17 246.14246.15246.16246.17[HJD] - 2454000-0.0050.0000.0050.0100.0150.0200.025relative flux246.14246.15246.16246.17[HJD] - 2454000-0.0050.0000.0050.0100.0150.0200.025relative fluxA new look at NICMOS transmission spectroscopy 9 Figure 13. Example of the decorrelation procedure on one of the HD 189733 wavelength channels, similar to Fig. 7, but showing a different wavelength channel, and now decorrelating only using orbits 2 and 4. It is clear that the decorrelation does not prop- erly correct for orbit 5, and an artificial offset is introduced by the linear baseline model. Note that this offset is larger than the features in the reported transmission spectrum. We argue that the decorrelation process may introduce similar spurious offsets to the in-transit orbit, indistinguishable from a real atmospheric signature. We also conducted additional tests to see how de- pendent the transmission spectrum is on the decorrelation model. A 'quadratic' decorrelation was tried by adding the terms ∆X 2, ∆Y 2, W 2, θ2 and T 2, to the state matrix (φH already has a quadratic term). We also made tests decorre- lating the spectra after removing each vector from the state matrix. We found that the most important decorrelation parameter to be θ, and to a lesser extent ∆Y . The other decorrelation parameters play a much less important role. This is consistent with the findings in S08, and is rather obvious after closer inspection of the decorrelation parame- ters in Fig. 6. θ and ∆Y show significant offsets between the in-transit orbit and the out-of-transit orbits. Thus, decorre- lating with these state parameters causes larger shifts in the in-transit orbit than the others, and consequently the most significant corrections to the transmission spectrum. The spectra produced using the 'quadratic' decorrela- tion, and after removing θ from the state matrix are shown in Fig. 12. Note all of the spectra plotted in Fig. 12 are af- ter the channel-to-channel correction has been applied, but its exclusion does not alter the results significantly for any case. It is clear from these plots that the various techniques used to decorrelate the light curves produce quite different output transmission spectra. Each of the features reported in S08 vanishes in at least one of the cases. As θ was pre- viously shown to be the most important parameter in the decorrelation, the interpretation of molecular features in the spectrum is heavily dependant on a linear dependance of the flux level with the angle the spectrum makes on the detec- Figure 12. Transmission spectra of HD 189733 generated by de- termining the planet radius ρ for the 18 wavelengths channels. The top plot shows the transmission spectrum obtained after ap- plying the channel-to-channel correction. The second plot shows the spectra by decorrelating the light curves only using orbits 2 and 4. The third plot shows the result of a 'quadratic' decorrela- tion, and the bottom plot shows a linear decorrelation but with the angle vector (θ) removed from the state matrix (X). Again, our results are the red points, and those from S08 are shown in grey for comparison. The horizontal dashed grey line marks the level of the planet-to-star radius ratio of the white light curve. For the last three cases the spectra are altered considerably from the first, and none of the features reported in S08 are present in all spectra using the different methods. mission spectrum shows a significantly different structure than before (c.f. Fig 9). In fact the only 'feature' common to both, is the dip at around 2.1 µm, although its depth relative to the uncertainties varies. The edges of the spectrum show quite different behaviour. We interpret the significant differ- ences in the two spectra as evidence that the linear baseline model is not sufficient to correct the light curves, and can induce unwanted offsets in the in-transit orbit, which are indistinguishable from a real atmospheric signal. c(cid:13) 2002 RAS, MNRAS 000, 1 -- 17 0.1520.1540.1560.1580.160ρ0.1520.1540.1560.1580.160ρ0.1520.1540.1560.1580.160ρ1.41.61.82.02.22.42.6wavelength µm0.1520.1540.1560.1580.160ρ16401650166016701680169017001710counts (electrons×103)246.05246.10246.15246.20246.25246.30HJD - 24540000.950.960.970.980.991.001.01relative flux10 N. P. Gibson et al. tor. We do not argue that any of these methods are better than the others or the method used by S08, but they give incompatible results and there is no way to distinguish be- tween them. Whilst most of the systematic effects in the light curves are caused by intra-pixel sensitivity changes as the spectral trace drifts and rotates on the detector, it is not clear that a linear decorrelation will remove these systemat- ics. In fact, we have shown that the decorrelation can cause spurious offsets in the in-transit flux level. , B F0 Furthermore, the largest disagreements between the spectra occur at the edges, where the background is expected to significantly affect the transit depth. The ratio of the true transit depth (D) is related to the measured transit depth (D(cid:48)) by D D(cid:48) = 1 + where B is the (uncorrected) background counts, and F0 the out-of-transit counts. Therefore, an underestimated or over- estimated background will lead to incorrect transit depths. This effect scales with the ratio of the background to the out-of-transit flux. Thus where the flux count is lower, i.e. at the fainter channels (the edges) of the spectrum, this ef- fect will be much greater. For the faintest channels, we have approximately 150 000 × 5 electrons (Fig. 3) collected over 35 × 5 pixels. If we assume a background correction error of 50 electrons per pixel (a conservative estimate considering the background can vary by up to 200 electrons per pixel), the ratio of true depth to measured depth is about 1.012, more than enough to significantly affect the measured tran- sit depth. Clearly we cannot trust the edges of the spectra to give us an accurate measurement of the planet-to-star radius ratio. 3 GJ-436 3.1 Observations A transit of GJ-436 was monitored on 10-11th November 2007 with HST/NICMOS, using the G141 grism covering the wavelength range 1.1 -- 1.9 µm. Analyses of these data were first reported in Pont et al. (2009, hereafter P09). We only used the first of two light curves reported in P09 for this work, as they concluded that the noise and systematic effects were lower during the first visit. GJ-436 is again not in the continuous viewing zone of the HST, and the transit was observed over four half-orbits, consisting of 935 spec- tra in total, all with exposure times of 1.993 seconds. The first, second and fourth orbits cover the out-of-transit part of the light curve and consist of 180, 252 and 251 observa- tions, respectively, and are used to determine the photomet- ric baseline. The third orbit was taken in-transit, covering mid-transit and egress, and consists of the remaining 252 images. Similarly to HD 189733, some exposures were taken at the beginning of the first orbit to set the wavelength cal- ibration. Again, calibrated images include all basic calibrations except for flat-fielding. The background with the G141 grism is much lower than with the G206 grism, and is therefore the background correction is less of an issue for this data set. The background was again very stable over the duration of the observations, with a typical level of only 25 − 30 electrons Figure 14. Extracted 1D spectra for GJ-436 of a typical in- transit (green) and out-of-transit (blue) observation showing the number of electrons collected per pixel channel. The wavelength decreases along the x-axis. per pixel. The spatial variation over the detector is about 10 electrons per pixel. We again tried flat-fielding using a flat-field taken with the G141 grism in place, but this did not provide a satisfactory correction and we did not use the flat-field corrected images for this analysis. The spectra and light curves were extracted using the same technique as for HD 189733. The spectra were ex- tracted along 90 pixel columns, with a width of 20 pix- els. The width of 20 pixels was selected to minimise the RMS in orbits 2 and 4 of the white light curve. Again we experimented using both global and column-specific back- ground corrections, but it made insignificant difference due to the low background. Fig.14 shows a 1D spectrum for an in-transit and out-of-transit observation. About 420 000 electrons were collected per exposure in the brightest pixel channel, and approximately 170 000 electrons in the faintest pixel channel. The raw light curves were then constructed as for HD 189733 by binning the spectra into 5 pixel bins, and are shown in Fig. 15. They show similar systematic effects to HD 189733. The white light curve was extracted using 110 pixel columns and is shown in Fig. 16, excluding orbit 1 (left out of the remaining analysis due to large systematics), and the optical state parameters are shown in Fig. 17. 3.2 Analysis Each light curve was decorrelated using the procedure de- scribed in Sect. 2.2. The decorrelation coefficients were de- termined for orbits 2 and 4. Referring to Fig. 17, it is clear that we are extrapolating for the angle of the spectrum when correcting the in-transit orbit. We therefore must exclude this parameter from the multi-linear decorrelation, other- wise this results in a spectrum with very large features caused by spurious offsets in the in-transit baseline func- tion. An example of the decorrelation procedure is shown in Fig. 18. The decorrelated light curves are shown in Fig. 19. Each of the light curves was then fitted using the transit models described in Sect. 2.2 to determine ρ as a function of wave- length, fixing the stellar and orbital parameters at those de- c(cid:13) 2002 RAS, MNRAS 000, 1 -- 17 80100120140160x position (pixels)150200250300350400450counts (electrons×103)A new look at NICMOS transmission spectroscopy 11 Figure 16. Raw 'white' light curve of GJ-436 found by integrat- ing the flux from each spectra over all wavelengths, showing the sampling of the GJ-436 transit, excluding the first orbit. Figure 15. Raw light curves for each of the 18 wavelength chan- nels for GJ-436, from 1.87 µm (top) to 1.19 µm (bottom), after normalising each one by fitting a linear function through orbits 2 and 4. Similarly to HD 189733, correlated noise is seen in each light curve. termined by P09, who undertook a thorough analysis of the integrated light curve. The central transit time was deter- mined from the white light curve (as a well sampled egress allows an accurate determination of the central transit time), and fixed at this value when fitting for each wavelength chan- nel. The limb darkening parameters were calculated using the method of Sing (2010) as for HD 189733, and the uncer- tainties were calculated using the residual permutation algo- rithm described previously, with 1000 light curves generated with random shifts applied to the light curve residuals. 3.3 Results The resulting transmission spectrum for GJ-436 is shown in Fig. 20. There seems to be a significant variation of the radius ratio with wavelength. Unfortunately, with only two out-of-transit orbits to work with, we do not have enough data to test the decorrelation model with differ- ent orbits. The planet-to-star radius ratio is calculated as ρ = 0.0830 ± 0.0005 for the white light curve2, consistent with P09. The transmission spectrum does not appear con- sistent with this value, as nearly all wavelength channels 2 This uncertainty is obtained when only fitting for the transit depth and a linear baseline function, and therefore is likely un- derestimated. c(cid:13) 2002 RAS, MNRAS 000, 1 -- 17 Figure 17. Extracted de-correlation parameters for GJ-436 dur- ing orbits 2 -- 4 as a function of time. For the angle of the spectral trace θ, the baseline function must be extrapolated for the in- transit orbit, hence this parameter cannot be used in the decor- relation process, without introducing large offsets in the transit light curve. 415.55415.60415.65415.70[HJD] - 24540000.750.800.850.900.951.00relative flux415.55415.60415.65415.70HJD - 24540000.9880.9900.9920.9940.9960.9981.0001.0021.004relative flux-0.12-0.08-0.040.00∆X (pix)64.4064.4464.48∆Y (pix)2.722.762.80W (pix)1.361.401.441.481.52θ (deg)415.55415.60415.65415.70[HJD] - 245400077.1277.1677.20T (K)12 N. P. Gibson et al. Figure 18. Example of the decorrelation procedure on one of the GJ-436 wavelength channels, showing the raw and decorrelated light curve at the top and bottom, respectively. The green points represent the baseline function. Figure 19. De-correlated transit light curves of GJ-436 for each of the 18 wavelength channels. Again, much of the time-correlated systematic noise is removed from the light curves, but some re- mains. Figure 20. Transmission spectrum of GJ-436 generated by deter- mining the planet-to-star radius ratio from the decorrelated light curves for the 18 wavelengths channels. The horizontal dashed line shows the planet-to-star radius ratio calculated from the white light curve. in the transmission spectrum give a larger depth. However, the transmission spectrum is consistent with the white light curve if we decorrelate only using the orbital phase parame- ters. A likely explanation for this is that the linear decorre- lation process induces false offsets in the in-transit flux level, as was suspected for HD 189733. It is interesting that the amplitude of the measured transit depth ((cid:39) ρ2) shows a similar amplitude to that of HD 189733. Any variations in the transmission spectrum is likely due to systematic noise in the detector that we have not accounted for in our error analysis, rather than any real physical effect. Perhaps this represents the limit of NIC- MOS transmission spectroscopy observations due to corre- lated noise, without a more robust way to remove or control the systematics. 4 XO-1 4.1 Observations A transit of XO-1 was observed on 21 February 2008 with HST/NICMOS, using the G141 grism. These data were first reported in Tinetti et al. (2010, hereafter T10). XO-1 is a similar spectral type to HD 189733, but it is considerably fainter, with a J-band magnitude of 9.94 (c.f. 6.07 for HD 189733). The transit was observed over five half orbits, con- sisting of 279 spectra in total, each with an exposure time of 40 seconds. The first, second and fifth orbit cover the out- of-transit part of the light curve, and consist of 56, 56 and 55 spectra, respectively. The third orbit covers the ingress with 56 spectra, and the fourth covers mid-transit, consist- ing of 55 spectra. XO-1 has a longer transit duration than HD 189733 or GJ-436, hence two consecutive orbits cover in-transit data. Again, some exposures were taken prior to the first orbit to set the position for wavelength calibration. The images were treated and the light curves extracted using the techniques described in Sects. 2.2 and 3.2. The 1D spectra were extracted along 90 pixel columns of width 16 c(cid:13) 2002 RAS, MNRAS 000, 1 -- 17 19251930193519401945195019551960counts (electrons×103)415.55415.60415.65415.70HJD - 24540000.9700.9750.9800.9850.9900.9951.0001.0051.010relative flux415.55415.60415.65415.70[HJD] - 24540000.750.800.850.900.951.00relative flux1.11.21.31.41.51.61.71.81.9µm0.0800.0820.0840.0860.0880.0900.092ρA new look at NICMOS transmission spectroscopy 13 Figure 21. Extracted 1D spectra for XO-1 of a typical in-transit (green) and out-of-transit (blue) observation showing the number of electrons collected per pixel channel. Figure 23. Raw light curves of XO-1 for each of the 18 wave- length channels from 1.87 µm (top) to 1.19 µm (bottom) excluding orbit 1, and after normalising each one by fitting a linear function through orbits 2 and 5. 4.2 Analysis Given the large amount of systematics, one would like to apply the decorrelation used for HD 189733 and GJ-436 on these data. However, it is clear from Fig. 24, that for the in-transit orbits (3 and 4), we must extrapolate if we are to use corrections for ∆X, ∆Y and θ. We therefore must exclude these decorrelation parameters from the procedure. An example of the decorrelation procedure using only W , T , φH and φ2 H , is shown in Fig. 25. The large residuals show the decorrelation process does not provide a more satisfactory correction. The decorrelated light curves are shown in Fig. 26. Time-correlated noise still remains in the light curves, but no combination of decorrelation parameters produces a satisfac- tory result. The light curves were then fitted using the tran- sit models described previously, fixing the orbital and stel- lar parameters at those determined by Torres et al. (2008). The central transit time was again set by fitting the white light curve with a transit model, and was then fixed when fitting the light curve for each wavelength channel. The un- certainties were determined using the residual permutation algorithm described, with 1000 realisations. 4.3 Results Fig. 27 shows the resulting transmission spectrum for XO-1. The results from T09 are also plotted for comparison. The amplitude in terms of flux is >0.3 %, and the corresponding Figure 22. Raw 'white' light curve of XO-1 found by integrat- ing the flux from each spectra over all wavelengths, showing the sampling of the XO-1 transit, excluding the first orbit. pixels. The background was estimated from a column strip above each wavelength channel. We also experimented with flat-fielding and global background corrections, but neither significantly affects the final results. Fig.21 shows typical 1D spectra extracted for in-transit and out-of-transit obser- vations, with the number of electrons collected per image per pixel column ranging from approximately 160 000 to 400 000. Fig. 23 shows the light curves extracted for each of the 18 wavelength channels, after binning into 5 pixel bins. Fig. 22 shows the resulting white light curve, extracted over 110 pixel columns. The first few points in each orbit exhibit strong variation in flux, and are thus excluded from the sub- sequent analysis. The white light curve shows much larger systematics than either the HD 189733 and GJ-436 light curves. Conse- quently, the light curves for each wavelength channel suffer from significant levels of correlated noise. The decorrelation parameters were extracted as for HD 189733 and GJ-436, and are plotted in Fig. 24. c(cid:13) 2002 RAS, MNRAS 000, 1 -- 17 20406080100120x position (pixels)150200250300350400450counts (electrons×103)518.25518.30518.35518.40518.45518.50HJD - 24540000.9750.9800.9850.9900.9951.0001.005relative flux518.25518.30518.35518.40518.45518.50[HJD] - 24540000.750.800.850.900.951.00relative flux14 N. P. Gibson et al. Figure 24. Extracted optical state parameters for XO-1 during orbits 2 -- 5. For the in-transit orbits (3 and 4), we have to ex- trapolate for ∆X, ∆Y and θ, making the decorrelation process extremely difficult. Figure 25. Example of the decorrelation procedure on one of the XO-1 wavelength channels, showing the raw and decorrelated light curve at the top and bottom, repectively. The green points represent the baseline function. It is clear this does not provide a satisfactory correction for the systematics, which is obvious in the residuals for all orbits. Figure 26. De-correlated light curves of XO-1 for each of the 18 wavelength channels. Some of the time-correlated systematic noise is removed from the light curves; however, a significant amount remains which could not be removed in the decorrelation procedure with any linear combination of decorrelation parame- ters. uncertainties are about 0.05%. Both of these are significantly larger than those reported in T10, and obviously we have not reached the precision required to detect any molecular species. The reason for this is clearly related to the decor- relation parameters. Unfortunately, T10 did not provide a plot of their decorrelation parameters, or plots of the light curves. This makes it difficult to check where our reductions differ. As T10 did not mention the need to extrapolate for any of the decorrelation parameters, we must presume the difference between our analysis and theirs resides chiefly in the extraction of the decorrelation parameters. However, recently Burke et al. (2010) reported a broad- band analysis of this light curve, and provided plots of ∆X, ∆Y and θ. The slight differences in ∆X and ∆Y are probably related to the extraction technique as discussed in Sect. 2.1. The values reported for θ agree with those in this paper. This confirms the need to extrapolate for the in-transit orbits, in particular for θ, which earlier we con- cluded was the most important decorrelation parameter for HD 189733. 5 SUMMARY AND DISCUSSION We report a re-analysis of transmission spectra observed with NICMOS for HD 189733, GJ-436 and XO-1. We used a linear decorrelation model, similar to that used in other HST c(cid:13) 2002 RAS, MNRAS 000, 1 -- 17 -0.04-0.020.000.020.04∆X (pix)-0.08-0.040.000.040.08∆Y (pix)2.802.842.88W (pix)0.520.560.600.640.680.720.76θ (deg)518.25518.30518.35518.40518.45518.50[HJD] - 245400077.0877.12T (K)1475148014851490149515001505151015151520counts (electrons×103)518.25518.30518.35518.40518.45518.50HJD - 24540000.950.960.970.980.991.001.01relative flux518.25518.30518.35518.40518.45518.50[HJD] - 24540000.750.800.850.900.951.00relative fluxA new look at NICMOS transmission spectroscopy 15 sorption in the transmission spectrum, which is similar to that seen in P09; however, the overall levels are inconsis- tent. We also find the transmission spectrum is inconsistent with the transit depth of the white light curve. The latter is much more robust given that the integrated white light curve suffers from fewer systemics. The transmission spectrum is consistent with the white light curve when only decorrelat- ing using the orbital phase parameters. This suggests that decorrelating with more parameters, particularly those with offsets in the parameter values, induces spurious offsets in the transmission spectrum. Generally, we conclude that it is unstable to decorrelate HST data using decorrelation param- eters that contain offsets between each orbit, as these offsets in easily be transferred to spurious offsets in flux levels of the in-transit orbits. The fact that we can produce a vari- able but unphysical transmission spectrum with the linear decorrelation model, suggests that we cannot extract use- ful spectral information below about 0.1% using NICMOS transmission spectroscopy, at least using current methods. For XO-1 we are unable to reproduce the results from T10, because the decorrelation parameters we measure re- quire an extrapolation for the in-transit orbits to determine the baseline function, and introduce large offsets in the flux levels for different orbits. The best transmission spectrum that we can produce show uncertainties much larger than those reported in T10. We do not have enough informa- tion to identify the source of the discrepancy, but we ex- pect that their uncertainties are underestimated, in a similar way to HD 189733. This is all the more puzzling given that Burke et al. (2010) also produce decorrelation parameters that would require extrapolation for the in-transit orbits. We note the importance of reporting in detail the methods used to correct systematics when they dominate the error budget. Our re-analysis of all three datasets shows that the sys- tematics in NICMOS transmission spectra cannot reliably be corrected for - at the level needed to detect molecular ab- sorption in hot Jupiters - using the multi-linear decorrelation techniques described in S08 and repeatedly used since. In- deed, there is no physical reason why the systematics should be described by a linear combination of the selected state parameters. In the absence of a better systematics removal technique, there appears to be a ∼0.1 % floor below which real absorption variation cannot be distinguished from sys- tematics with NICMOS. All claimed detections lie near this level, regardless of object (Fig. 1). Whilst this could be coin- cidental, more evidence is required to support these results, particularly when Pont et al. (2008) and Sing et al. (2009) suggest otherwise for HD 189733. ACKNOWLEDGMENTS All of the data presented in this paper were obtained from the Multimission Archive at the Space Telescope Sci- ence Institute (MAST). STScI is operated by the Associa- tion of Universities for Research in Astronomy, Inc., under NASA contract NAS5-26555. Support for MAST for non- HST data is provided by the NASA Office of Space Science via grant NNX09AF08G and by other grants and contracts. N. P. G and S. A. acknowledge support from STFC grant ST/G002266/2. We are very grateful to D. Sing for pro- Figure 27. Transmission spectrum of XO-1 generated by deter- mining the planet radius from the decorrelated light curves ρ for the 18 wavelengths channels. Out results are the red points, and those from T10 are shown in grey for comparison. The variation is not caused by the atmosphere of the planet, but rather the sys- tematics in the data and the decorrelation method used to correct for these. analyses (e.g. Pont et al. 2008, S08, P09, T10), where pos- sible, to decorrelate the light curves as they all suffer from considerable systematics. We used a residual permutation algorithm in an effort to obtain more realistic uncertainties than in the literature, although this does not account for uncertainties which arise from orbit-to-orbit flux offsets. For HD 189733, we find a similar transmission spectrum to S08 when following their procedure as closely as possi- ble. However, using residual permutation, the uncertainties we calculate are often significantly larger, thus reducing the significance of the spectral features. We show that signifi- cant systematic noise remains in the in-transit orbits, which is not fully removed by the linear decorrelation model, or the channel-to-channel correction. The shape and amplitude of the spectral features are also strongly dependent on the choice of orbits, decorrelation parameters, and the nature of the model assumed for the baseline function. Given there is no physical reason to assume the baseline flux should follow a linear function of a particular set of optical state param- eters, there is no reason to prefer one model over another. We take this as evidence that the linear decorrelation model used is not a robust method to remove systematic effects from the light curves, and that the baseline function in the case of NICMOS grism data cannot reliably be given by a linear function of the optical state vectors from S08. Fig. 28 shows a plot of the overall transmission spec- trum of HD 189733, including the ACS optical data from Pont et al. (2008), the NICMOS photometric measurements from Sing et al. (2009), and the reduction using only orbits 2 and 4 from this paper. This shows that we can interpret the NICMOS transmission spectrum as consistent with the optical haze, if we adopt this reduction. We emphasise that our analysis does not rule out the presence of molecules in the NIR transmission spectrum of HD 189733, but merely shows that the detection of molecular species is highly de- pendent on the decorrelation method used. For GJ-436 we find an amplitude of about 0.07 % ab- c(cid:13) 2002 RAS, MNRAS 000, 1 -- 17 1.11.21.31.41.51.61.71.81.9µm0.1200.1250.1300.135ρ16 N. P. Gibson et al. Figure 28. Overall transmission spectrum of HD 189733, including the ACS optical data from Pont et al. (2008, green circles), the NICMOS photometric measurements from Sing et al. (2009, blue squares), and a re-reduction of G206 NICMOS grism data using only orbits 2 and 4 from this paper (red circles). The dotted-dashed line is the haze given in Lecavelier Des Etangs et al. (2008), which the Pont et al. (2008) and Sing et al. (2009) results are consistent with. The NICMOS grism data is consistent with the haze using this reduction. We emphasise that this case is in no way special, rather we want to illustrate that the NICMOS transmission spectroscopy data may still be consistent with the haze interpretation rather than the reported molecular species, given the difficulty in propagating the effects of systemics to the final uncertainties. viding limb darkening co-efficients for the NICMOS wave- length channels. Finally, we thank the referee, whose insight- ful comments helped improve the clarity of this paper. Gilliland, R. L. & Arribas, S. 2003, Instrument Science Re- port NICMOS 2003-001 Gillon, M., Demory, B., Barman, T., et al. 2007, A&A, 471, REFERENCES Brown, T. M. 2001, ApJ, 553, 1006 Burke, C. J., McCullough, P. R., Bergeron, L. E., et al. 2010, ApJ, 719, 1796 L51 Lecavelier Des Etangs, A., Pont, F., Vidal-Madjar, A., & Sing, D. 2008, A&A, 481, L83 Mandel, K. & Agol, E. 2002, ApJL, 580, L171 Mosser, B., Baudin, F., Lanza, A. F., et al. 2009, A&A, 506, 245 Pont, F., Gilliland, R. L., Knutson, H., Holman, M., & Carter, J. A., Winn, J. N., Gilliland, R., & Holman, M. J. Charbonneau, D. 2009, MNRAS, 393, L6 2009, ApJ, 696, 241 Pont, F., Knutson, H., Gilliland, R. L., Moutou, C., & Charbonneau, D., Brown, T. M., Noyes, R. W., & Gilliland, Charbonneau, D. 2008, MNRAS, 385, 109 R. L. 2002, ApJ, 568, 377 D´esert, J., Lecavelier des Etangs, A., H´ebrard, G., et al. 2009, ApJ, 699, 478 Seager, S. & Sasselov, D. D. 2000, ApJ, 537, 916 Sing, D. K. 2010, A&A, 510, A21+ Sing, D. K., D´esert, J., Lecavelier Des Etangs, A., et al. Ehrenreich, D., H´ebrard, G., Lecavelier des Etangs, A., 2009, A&A, 505, 891 et al. 2007, ApJL, 668, L179 Gibson, N. P., Pollacco, D., Simpson, E. K., et al. 2009, Southworth, J. 2008, MNRAS, 386, 1644 Swain, M. R., Tinetti, G., Vasisht, G., et al. 2009a, ApJ, ApJ, 700, 1078 704, 1616 Gibson, N. P., Pollacco, D., Simpson, E. K., et al. 2008, Swain, M. R., Vasisht, G., & Tinetti, G. 2008, Nature, 452, A&A, 492, 603 329 Gibson, N. P., Pollacco, D. L., Barros, S., et al. 2010, MN- Swain, M. R., Vasisht, G., Tinetti, G., et al. 2009b, ApJL, RAS, 401, 1917 690, L114 c(cid:13) 2002 RAS, MNRAS 000, 1 -- 17 0.51.01.52.02.5wavelength (µm)0.1520.1530.1540.1550.1560.157ρA new look at NICMOS transmission spectroscopy 17 Tinetti, G., Deroo, P., Swain, M. R., et al. 2010, ApJL, 712, L139 Tinetti, G., Vidal-Madjar, A., Liang, M., et al. 2007, Na- ture, 448, 169 Torres, G., Winn, J. N., & Holman, M. J. 2008, ApJ, 677, 1324 c(cid:13) 2002 RAS, MNRAS 000, 1 -- 17
1508.06331	1	1508	2015-08-26T00:00:33	Precise Distances for Main-Belt Asteroids in Only Two Nights	[ "astro-ph.EP" ]	We present a method for calculating precise distances to asteroids using only two nights of data from a single location --- far too little for an orbit --- by exploiting the angular reflex motion of the asteroids due to Earth's axial rotation. We refer to this as the rotational reflex velocity method. While the concept is simple and well-known, it has not been previously exploited for surveys of main-belt asteroids. We offer a mathematical development, estimates of the errors of the approximation, and a demonstration using a sample of 197 asteroids observed for two nights with a small, 0.9-meter telescope. This demonstration used digital tracking to enhance detection sensitivity for faint asteroids, but our distance determination works with any detection method. Forty-eight asteroids in our sample had known orbits prior to our observations, and for these we demonstrate a mean fractional error of only 1.6% between the distances we calculate and those given in ephemerides from the Minor Planet Center. In contrast to our two-night results, distance determination by fitting approximate orbits requires observations spanning 7--10 nights. Once an asteroid's distance is known, its absolute magnitude and size (given a statistically-estimated albedo) may immediately be calculated. Our method will therefore greatly enhance the efficiency with which 4-meter and larger telescopes can probe the size distribution of small (e.g. 100 meter) main belt asteroids. This distribution remains poorly known, yet encodes information about the collisional evolution of the asteroid belt --- and hence the history of the Solar System.	astro-ph.EP	astro-ph	Precise Distances for Main-Belt Asteroids in Only Two Nights Aren N. Heinze1,2 and Stanimir Metchev3,4,1 ABSTRACT We present a method for calculating precise distances to asteroids using only two nights of data from a single location — far too little for an orbit — by exploiting the angular reflex motion of the asteroids due to Earth’s axial rotation. We refer to this as the rotational reflex velocity method. While the concept is simple and well-known, it has not been previously exploited for surveys of main-belt asteroids. We offer a mathematical development, estimates of the errors of the approximation, and a demon- stration using a sample of 197 asteroids observed for two nights with a small, 0.9-meter telescope. This demonstration used digital tracking to enhance detection sensitivity for faint asteroids, but our distance determination works with any detection method. Forty-eight asteroids in our sample had known orbits prior to our observations, and for these we demonstrate a mean fractional error of only 1.6% between the distances we calculate and those given in ephemerides from the Minor Planet Center. In contrast to our two-night results, distance determination by fitting approximate orbits requires observations spanning 7–10 nights. Once an asteroid’s distance is known, its absolute magnitude and size (given a statistically-estimated albedo) may immediately be calcu- lated. Our method will therefore greatly enhance the efficiency with which 4-meter and larger telescopes can probe the size distribution of small (e.g. 100 meter) main belt asteroids. This distribution remains poorly known, yet encodes information about the collisional evolution of the asteroid belt — and hence the history of the Solar System. Subject headings: astrometry, celestial mechanics, ephemerides, minor planets, aster- oids: general 1Physics & Astronomy Department, Stony Brook University, Stony Brook, NY 11794-3800, USA; [email protected] 2Visiting astronomer, Kitt Peak National Observatory, National Optical Astronomy Observatory, which is operated by the Association of Universities for Research in Astronomy (AURA) under a cooperative agreement with the National Science Foundation. 3Physics & Astronomy Department, The University of Western Ontario, London, ON N6A 3K7, Canada; [email protected] 4Centre for Planetary and Space Exploration, The University of Western Ontario, London, ON N6A 3K7, Canada; [email protected] – 2 – 1. Introduction The main asteroid belt is a relic from the formation of the Solar System. Although much of its mass has been lost, it retains a great deal of information about Solar System history and presents us with a laboratory in which we can study collisional processes that once operated throughout the circumsolar disk in which Earth and the other planets were formed. One of the most straightforward observables constraining such processes is the asteroid belt’s size-frequency distribution (SFD; Bottke et al. 2005a). The current main belt’s SFD can be successfully modeled as the result of 4.5 billion years of collisional evolution (Bottke et al. 2005a; de El´ıa & Brunini 2007). While such models fit the ‘collisional wave’ set up by 100 km asteroids able to survive unshattered through the age of the Solar System, they cannot be observationally tested in the 100 meter size range. Objects in this size range are very interesting, because they supply most near-Earth asteroids and meteorites by shattering one another and/or migrating inward via Yarkovsky and resonance ef- fects (Farinella et al. 1994; Bottke et al. 2005b). Modern 8-10 meter telescopes can detect them, but monitoring them over many nights to determine an orbit requires a prohibitively large time invest- ment for such powerful telescopes (e.g., 7–10 nights; Gladman et al. 2009). Thus their distances and sizes remain unknown, and detailed analyses are confined to larger objects (Gladman et al. 2009) or use only rough statistical distances (Yoshida et al. 2003; Yoshida & Nakamura 2007). We present a method to obtain precise distances to main belt asteroids (MBAs) using only two nights of observations. Distances translate directly into absolute magnitudes and hence to sizes given a reasonable assumption for the albedo distribution. This method, which we refer to as rotational reflex velocity (RRV), will greatly increase the efficiency of surveys aimed at probing collisional evolution in the Solar System by measuring the SFDs for extremely small MBAs. We demonstrate RRV distance determination using a data set from the 0.9-meter WIYN telescope1, which we have analyzed using digital tracking (Heinze et al. 2015) in order to enhance our sensitivity to faint asteroids. Digital tracking is a method for detecting faint moving objects that was first applied to the Kuiper Belt (e.g. Bernstein et al. 2004), and very recently has begun to be applied to asteroids (Zhai et al. 2014; Heinze et al. 2015). Although the RRV distances we calculate herein are all based on our digital tracking analysis, the RRV method is equally useful for asteroids detected by more conventional means, or by other specialized methods such as those of Milani et al. (1996) and Gural et al. (2005). 1The WIYN Observatory is a joint facility of the University of Wisconsin-Madison, Indiana University, the National Optical Astronomy Observatory and the University of Missouri. – 3 – 2. Distances from Rotational Reflex Velocity Suppose that at a given instant, an asteroid located a distance d from an Earth-based observer is moving with velocity va, while the observer is moving with velocity vo (e.g., the orbital velocity of the Earth). The angular velocity at which the observer sees the asteroid move relative to distant stars is given by: ω = va⊥ − vo⊥ d (1) where the ⊥ subscript indicates the vector component perpendicular to the line of sight, so that va⊥ − vo⊥ is the projection of the asteroid’s relative velocity onto the plane of the sky. Although vo can be accurately calculated for any Earth-based observation, the velocity va of a newly discovered asteroid is always unknown initially, and hence the distance cannot be calculated by simply plugging the measured value of ω into Equation 1. Given appropriate measurements, however, we can isolate the component of ω that reflects the observer’s motion around the geocenter due to Earth’s rotation, and from this calculate the distance. This is the essence of the RRV method for distance determination. The velocity vo of an observer on the surface of the Earth can be expressed as the sum of Earth’s orbital velocity vorb and the velocity vrot with which the Earth’s axial rotation carries the observer around the geocenter. Neglecting the slight asphericity of the Earth, vrot = veq cos θ, where θ is the observer’s terrestrial latitude and veq is the Earth’s equatorial rotation velocity of 1674.4 km/hr. For convenience, we define vrel as the asteroid’s velocity relative to the geocenter: vrel = va − vorb. The angular velocity ωg that would be measured by an ideal observer located at the geocenter then depends only on vrel and the distance, but the angular velocity ωo that is measured by a real observer based on the Earth’s surface depends also on vrot. The two angular velocities are given by: ωg = vrel⊥ d ωo = vrel⊥ − vrot⊥ d If we could measure ωg, we could therefore calculate the distance: d = vrot ωg − ωo (2) (3) (4) where we have dropped the ⊥ subscript, because it will henceforward apply to all physical velocities in our calculations. – 4 – Now suppose that the asteroid is observed near midnight on two different nights, that the two observations are separated by exactly one sidereal day, and that the position and angular velocity ωo are recorded for each observation. The angular distance the asteroid moved between the two observations will thus be accurately known; call this φ. Because exactly one full rotation of the Earth elapsed between the two observations, the observer’s position relative to the geocenter is the same for both of them. Thus, the average geocentric angular velocity of the asteroid in between the two measurements is ωg = φ/∆t, where ∆t is the elapsed time between the observations: one sidereal day. Let the measured values of ωo on the first and second nights be ωo1 and ωo2, and similarly let the perpendicular rotational velocities (which are obtained by calculation, not measurement) be vrot1 and vrot2. We can then evaluate the difference between geocentric and observer angular velocities twice: the average of ωg − ωo1 and ωg − ωo2 will be a factor of √2 more precise than a single measurement if the uncertainty on ωg measurement is much smaller than on ωo1 and ωo2. This is likely to be the case, since ωg is based on a longer temporal baseline. The distance is then given by: d = vrot1 + vrot2 (ωg − ωo1) + (ωg − ωo2) (5) So far we have assumed that vrel and d show no appreciable change over the 24 hour period of the measurements, so that ωg − ωo1 and ωg − ωo2 are effectively two measurements of the same quantity. We will now determine the errors that result when this assumption is violated. We will first consider changes in vrel and d that are linear in time: that is, when the first time derivatives vrel and d are significant but the second derivatives vrel and d are not. We will parameterize the change in vrel by ∆v = (vrel2 − vrel1)/2, and the change in d by ǫ = (d2−d1)/(2d). For convenience, we will also use δ = (vrot2−vrot1)/(2vrot) since the observations do not have to be taken exactly one sidereal day apart, and hence the Earth’s projected rotational velocity may vary. For well-optimized observations δ should always be close to zero. Note that while ǫ and δ are unit-less, fractional changes, ∆v still has units of linear velocity. We then obtain: ωg − ωo1 = vrel ωg − ωo2 = vrel (1−ǫ)d d − vrel−∆v−(1−δ)vrot d − vrel+∆v−(1+δ)vrot (1+ǫ)d (6) We substitute these into Equation 5. Retaining terms only to second order in the variational quantities ∆v, ǫ, and δ, after a good deal of algebra we find that the fractional error on the calculated distance is: dcalc − dtrue dtrue = ǫ2 vrel vrot − ǫ ∆v vrot + ǫδ − ǫ2 (7) – 5 – For main belt asteroids ǫ is always very small (i.e. < 10−3). This renders δ innocuous even though it can reach 0.3 for poorly-planned observations (Section 3.2). The velocity change ∆v approaches zero for asteroids at opposition, where the accuracy of Equation 5 can be better than 10−5, but even months from opposition the error remains only of order 10−3. Equation 5 is therefore very accurate under all conditions where the required input data could reasonably be obtained, and the uncertainties even of highly accurate measurements (Section 3.3) will generally be larger than the equation’s intrinsic error. We caution the reader against a much less accurate alternative to Equation 5 that may appear more attractive because it is obtained simply by applying Equation 4 individually to each night and then averaging the resulting two distance measurements: d = 0.5 vrot1 ωg − ωo1 + 0.5 vrot2 ωg − ωo2 (8) Despite its intuitive simplicity, this equation should never be used. Its fractional error expansion contains the term ∆v2/v2 rot, which has very large values for objects more than a few days from opposition, as illustrated by Figure 1. By contrast, for Equation 5 the precursors of the ∆v2/v2 rot term and several other error terms cancel in the denominator, leaving only the very much smaller error terms given by Equation 7. Equation 8 should therefore be avoided, and Equation 5 (or its generalized form, Equation 18) should always be used for RRV distances. 3. Asteroid Angular Velocities from Positions Measured at Discrete Times 3.1. Basic Equations and Methodology For simplicity, the discussion in Section 2 assumed the observer could make instantaneous angular velocity measurements. In practice, of course, angular velocity measurements must be made over a period of time. This has very little effect on our mathematical derivation. The only change is that the projected velocities of the observer relative to the geocenter, vrot1 and vrot2, should not be considered as instantaneous but rather as averages over the same period of time in which the angular velocities ωo1 and ωo2 were measured. Positions and angular velocities of asteroids are measured in terms of right ascension (RA) and declination (DEC) on the celestial sphere, and thus are two dimensional. For example, an observer could measure an asteroid as having an angular velocity of -37.36 arcsec/hr in RA (the negative sign indicating westward rather than eastward motion) and 10.32 arcsec/hr in DEC, with the positive sign indicating the object is moving toward celestial north. In principle, independent versions of Equation 5 could be constructed for each dimension, with angular velocities in RA corresponding exactly to projected east-west motion of the observer and DEC velocities corresponding to projected north-south motion. However, the true rotational velocity of the observer is always strictly east- – 6 – Fig. 1.— Errors of RRV distances calculated from JPL Horizons ephemerides for representative known MBAs. Black points are for Equation 18, a version of Equation 5 that is generalized for non- optimal timing of observations (Section 3.2), while gray points are for a similarly generalized form of Equation 8. Left: Near optimally timed observations with χ (Equation 17) equal to -0.0007, corresponding to t2m − t1m equal to one minute less than a sidereal day. Right: Non-optimally timed observations with χ = 0.052, corresponding to t2m − t1m equal to 1.3 hours more than a sidereal day. Equation 18 yields accurate distances even with poorly timed observations, while Equation 8 should never be used. ‘Noise’ in the plots is due to roundoff error in the ephemerides rather than intrinsic error in the equations. It is more severe in the right-hand case because of the shorter temporal baseline on each night (1.7 hr vs. 2.9 hr). – 7 – west, and although its projection vrot can have a north-south component, this component is usually small — i.e., less than 10% of the east-west component for observations with mean hour angle in the range ±1.0 targeting asteroids within 20◦ of the celestial equator. As we will find below, for our test observations only the RA velocity components are large enough to supply useful distances. We expect this will generally be the case. In most asteroid surveys, many fields are observed, and each field is visited at least three times per night. Each visit yields a single celestial position of each detected asteroid. Two observations of each asteroid on each night will suffice for our purposes. Using the traditional symbols α and δ for RA and DEC, suppose that on night 1 we obtain observations α1a, δ1a at time t1a and α1b, δ1b at time t1b, and similarly on night 2. Then we can also define mean positions and the mean time α1m = (α1a + α1b)/2, δ1m = (δ1a + δ1b)/2, and t1m = (t1a + t1b)/2, with the analogous quantities being calculated for night 2. Note that the times here are measured on a continuous sequence and not reset when the date changes. For simplicity, at present we will suppose further that the time of each measurement on night 2 is exactly one sidereal day later than the time of the corresponding measurement on night 1: thus t2a−t1a = t2b−t1b = one sidereal day. The required angular velocities are then given by the following equations, where for simplicity we show only the RA component: ωo1α = ωo2α = α1b − α1a t1b − t1a α2b − α2a t2b − t2a α2m − α1m t2m − t1m ωgα = (9) (10) (11) where the denominator of the last equation is of course also equal to one sidereal day. The only other quantities needed for the distance calculation are vrot1 and vrot2, the observer’s projected rotational velocities relative to the geocenter. As illustrated by Figure 2, the observer’s projected motion relative to the geocenter traces out a segment of an ellipse whose axis ratio is the sine of the asteroid’s declination. The projected distance of the observer from the geocenter is given by: x(t) = R⊕ cos(θlat) sin(cid:16) 2π(t−tz ) τS (cid:17) y(t) = −R⊕ cos(θlat) sin(δ) cos (cid:16) 2π(t−tz ) τS (cid:17) + R⊕ sin(θlat) (12) where x and y are positive toward the east and north; θlat is the observer’s latitude on the Earth; δ is the declination of the asteroid; t is the time of the observation; tz is the time when the asteroid stands at zero hour angle for the observer; and τS is the length of a sidereal day. Note that the – 8 – Fig. 2.— Example illustration of an observer’s motion due to the Earth’s rotation, as viewed from an asteroid. Observation A and Observation B are both made from the same location on Earth’s surface (at 20◦ N, roughly the latitude of Mauna Kea), but they are separated by an elapsed time of 3.989 hr, corresponding to 60◦ of sidereal rotation. The rotational velocity of Earth’s surface at 20◦ N lat is 1573.4 km/hr. The effect on the asteroid’s measured positions is determined by the projected distance moved between the observations, which has eastward and northward components as labeled. The projected rotational velocity vrot that should be used in the RRV distance calculation (Equation 5 or Equation 18) is therefore 5625.7 km/3.989 hr = 1410.3 km/hr eastward and 423.8 km/3.989 hr = 106.2 km/hr northward. – 9 – argument 2π(t − tz)/τS is the hour angle at which the observer sees the asteroid: at t = tz the asteroid crosses the meridian and stands closest to the observer’s zenith and in the best sky position for observing. For observations at t1a and t1b, the projected linear velocity vrot1 in the east-west (x) and north-south (y) directions is: vrot1x = vrot1y = x(t1b) − x(t1a) t1b − t1a y(t1b) − y(t1a) t1b − t1a (13) (14) Figure 2 illustrates a specific example of two observations of an asteroid at -12◦ DEC, made from a single observing site at 20◦ N lat on the Earth’s surface. The first observation is made when the asteroid is rising, and the second about four hours later, shortly after it has crossed the meridian. The rotational velocity of Earth at 20◦ N lat is 1573.4 km/hr and is of course strictly eastward, but projecting the velocity into a plane perpendicular to the line of sight reduces the absolute magnitude and introduces a north-south component (except for asteroids exactly at 0◦ DEC). In this example the projected velocities east and north are 1410.3 km/hr and 106.2 km/hr. Since celestial RA and DEC correspond exactly to terrestrial longitude and latitude, the east-west physical velocity of Equation 13 can be combined with the RA components of the angular velocities from Equations 9–11 and plugged directly into Equation 5 to yield the distance. We will now provide a simplified but basically realistic example. Suppose the MBA illustrated by Figure 2 is being observed at opposition. It will be in its retrograde loop, and hence its motion in RA will be negative (westward). The angular velocity ωo measured by the observer will be faster to the west than the geocentric angular velocity ωg, because on the night side of the Earth the rotational velocity adds to the orbital velocity, making the asteroid fall back westward at a faster rate. Suppose ωgα = −35 arcsec/hr and ωo1α = −36 arcsec/hr. The rotational reflex velocity, given by the difference ωgα − ωo1α, is then 1 arcsec/hr or 4.85 × 10−6 rad/hr. Taking the physical and angular velocities to be exactly the same on the second night for simplicity, Equation 5 reduces to: d = vrotx ωgα − ωoα = 1410.3 km/hr 4.85 × 10−6 rad/hr = 2.91 × 108 km = 1.94 AU (15) With Equation 15 we have calculated the distance to an asteroid based on the known prop- erties of Earth’s rotation. Interestingly, Bernstein & Kushalani (2000) have developed a closely analogous method for determining the distance to a Kuiper Belt object using the known properties of Earth’s orbit. Just as our derivation does not (explicitly) include the effects of Earth’s orbit, that of Bernstein & Kushalani (2000) does not include the effects of Earth’s rotation. Neglecting Earth’s orbit works in our case because, as we have demonstrated in Section 2, the dominant errors – 10 – due to orbital acceleration of both the Earth and the asteroid cancel when the correct form of the distance equation is used. Similarly, Bernstein & Kushalani (2000) can safely neglect Earth’s rota- tion because the Kuiper Belt objects whose distances they aim to measure are sufficiently far away that their RRV signature is negligible. It follows that the method of Bernstein & Kushalani (2000) should not be directly applied to measuring asteroid distances without some form of correction for the RRV signal, although (as they point out) it is useful for distinguishing true Kuiper Belt objects from asteroids that have similar angular velocities for a few days near their turnaround points. 3.2. The Case of Observations that are Not Optimally Timed Up to now we have assumed for simplicity that the observations obtained on successive nights are exactly one sidereal day apart. While it is not difficult to plan and execute observations that satisfy this criterion to within a few minutes, clouds or instrument failures could intervene, or one could be processing an archival data set that was not taken with RRV distances in mind. In such cases the mean times t1m and t2m may be separated by up to a couple of hours less or more than one sidereal day. A direct calculation of the geocentric angular velocity ωg then becomes impossible. In its place, we must use instead the mean angular velocity across the two nights, defined by: ωµα = α2m − α1m t2m − t1m (16) We will define three additional quantities. The time ts is the moment exactly one sidereal day before t2m (in the ideal case, t1m would have been equal to ts). The projected rotational velocity vrot,s is the observer’s mean projected rotational velocity over the time interval between t1m and ts. Finally, the fractional parameter χ is given by: χ = ts − t1m t2m − t1m Note that χ goes to zero in the case of optimally-timed measurements. The distance formula analogous to Equation 5 becomes: d = vrot1 + vrot2 − 2χvrot,s (ωµ − ωo1) + (ωµ − ωo2) (17) (18) The error terms for Equation 18 are the same as those given in Equation 7, with the addition of an ǫ2χ term. For main belt asteroids, this will always be negligible. In closing our discussion of error in the case of non-optimal observation timing, we note that δ, the fractional change in projected rotational velocity from night 1 to night 2, is likely to be considerably larger in this case, up to 0.3. However, in the error terms for Equation 18 it is always multiplied by ǫ, which is – 11 – always of order 10−3 or smaller for MBAs, and thus should not constitute a major source of error. Equation 18 is therefore the central formula for asteroid distance determination with RRV. In addition to our analytical error calculation (Equation 7), we have probed the errors of our approximations numerically using ephemerides of known objects from the JPL Horizons ephemeris generator. Note that no actual observations are involved in this test, only ephemeris positions and velocities for objects with well-known orbits. In Figure 1 we plot the error of Equation 18 as determined by this test. In contrast to our analytical error calculations, this test intrinsically includes the contributions of second and higher-order time derivatives of d and vrel, since it uses JPL Horizons ephemerides for real objects. The plotted errors are affected by roundoff error in the ephemerides (accurate only to 0.15 arcsec in RA), which introduces pseudo-noise such that only the lower envelope of the plotted points can be meaningfully compared with our analytical error estimates. Nevertheless, the plots show that Equation 18 produces accurate distances out to at least 60 days from opposition; that accurate distances can be obtained from observations that are not optimally timed; and that the second-order time derivatives vrel and d do not introduce significant error. 3.3. Measurement Precision Required for Good Distances Figure 1, although based on calculated ephemerides and not actual data, shows that relatively small position errors can create distance inaccuracies at the level of a few percent. We now consider the measurement precision that is necessary to yield distances to various levels of accuracy. In Equation 15, the difference between the night-to-night mean angular velocity in RA (ωµα) and the angular velocity in RA observed within a given night (ωoα) was one arcsec/hr, and this rotational reflex velocity corresponded to a distance of about 2 AU. This is typical for observations obtained near opposition at terrestrial latitudes of 20–30◦. The fractional uncertainty of the distance is equal to that of ωµα − ωoα. This will generally be dominated by the uncertainty in ωoα, since it is calculated within a single night and is thus based on a smaller temporal baseline than ωµα. Thus, if ωoα is measured with an accuracy of 0.1 arcsec/hr, we should expect the distance to have an accuracy of only 10% at 2 AU. At 1 AU, the RRV signature is twice as large: ωµα − ωoα will therefore be about 2 arcsec/hr, and the calculation will be accurate to 5%. As we will now illustrate, one can usually obtain angular velocity measurements with better than 0.1 arcsec/hr uncertainties in practice. In our current paradigm, ωoα is based on two measure- ments of celestial positions separated by a time tb − ta (i.e., Equation 9 or 10). If the uncertainty on RA for these measurements is σα, the uncertainty on ωoα will be: σω = σα√2 tb − ta (19) – 12 – Based on our experience, well-sampled (e.g. < 0.5 arcsec/pixel) images taken in ∼ 1.5 arc- second seeing can yield σα = 0.03–0.05 arcsec for bright objects (e.g., those at least several times brighter than the 10σ detection threshold). For faint objects near the detection limit, uncertainties of σα ∼ 0.1 arcsec are more typical2 A temporal baseline of 5 hr is easy to obtain near opposition. Thus, we can expect σω = 0.008–0.014 arcsec/hr for bright objects and σω = 0.03 arcsec/hr for fainter objects. At 2 AU, distances will therefore be accurate to about 3% for faint objects and 1–1.5% for bright objects, while at a distance of 1 AU, the accuracy should be about 1.5% even for faint objects. These accuracies are sufficient for good analyses of the size statistics of small MBAs, especially given that their albedos are unknown and must be treated as a statistical distribution spanning a factor of ∼10 (Pravec et al. 2012; Masiero et al. 2013). Where the objects are faint or distant (e.g. 3–5 AU), or where extremely accurate distances are desired, it may be possible to obtain smaller values of σα by obtaining each RA measurement not from a single image but from the average over a set of images acquired close together in time. Extending the same principle further, we can use the technique of digital tracking to obtain accurate measurements for objects too faint even to be detected in individual images, as we describe below. 4. Measurements Using Digital Tracking Data 4.1. A Brief Introduction to Digital Tracking The observations we use to demonstrate the usefulness of rotational reflex distances for as- teroids were obtained using digital tracking, as described in our companion paper Heinze et al. (2015). These data consist of 126 two-minute exposures of a single starfield that we acquired using the WIYN 0.9-meter telescope at Kitt Peak on the night of April 19, and 130 identically acquired images from the following night. Very briefly, a digital tracking search involves shifting and stacking such sets of images to reveal moving objects too faint to be detected on any individual frame. A separate trial stack is produced for each angular velocity vector in a finely-sampled grid that spans the full range of possible sky velocities for the target population. For MBAs, a digital tracking data set can span only one night (Heinze et al. 2015), hence we analyze our April 19 and April 20 observations independently. Digital tracking has been used to great effect for Kuiper Belt objects (e.g. Allen et al. 2001, Fraser & Kavelaars 2009, and many others), but has not typically been used for faster-moving asteroids where a larger numbers of trial vectors must be probed. It is now computationally tractable even for asteroids, however, and produces a factor of ∼ 10 increase in sensitivity over conventional methods. This sensitivity increase enables us to detect 215 asteroids within the 1-degree field of our test observations, despite using only a small, 0.9-meter telescope. We obtained precise angular velocity 2Note that the distance uncertainty must be evaluated differently for digital tracking observations, where the angular velocity is directly measured (see Section 4.2) rather than being obtained from two positions. – 13 – measurements on both nights for 197 of these asteroids, including 48 previously known objects with accurate orbits. These last allow us to test the accuracy of RRV distance measurements using real data. 4.2. Angular Velocity Measurement with Digital Tracking The geocentric angular velocity ωg of an asteroid detected on two subsequent nights is deter- mined in exactly the same way with digital tracking observations as with conventional data. The determination of the observed angular velocities ωo on each individual night is different, however — particularly because digital tracking enables the accurate measurement of asteroids that cannot even be detected on an individual image. Any asteroid detection in a digital tracking search occurs on a particular trial image stack, which corresponds to a particular angular velocity. Thus, for example, our automated digital tracking routine might search for objects with angular velocities between -50 and -20 arcsec/hr in RA and -10 and +20 arcsec/hr in DEC, using a grid spacing of 0.2 arcsec/hr, and might detect a particular asteroid as a bright point source on a trial stack whose shifts correspond to an angular velocity of -41.2 arcsec/hr in RA and 11.8 arcsec/hr in DEC. We improve on the relatively crude angular velocity measurement that is implicit in such a detection by probing a new, much more finely spaced grid of angular velocity vectors, using small postage-stamp images centered on the detected asteroid to make the search computationally tractable. For each trial stack in this finely sampled grid, we calculate the measured flux of the asteroid within a small aperture of radius roughly equal to the half-width of the point spread function (PSF) of our image. We then perform a 2-D quadratic fit to the measured flux as a function of the trial angular velocity. The angular velocity at which the quadratic fit reaches its peak value then constitutes an accurate measurement of ωo for that asteroid, and the uncertainty on ωo is derived from the uncertainty of the quadratic fit, which is naturally larger for faint objects that are more affected by sky background noise. All the inputs required to calculate the distance using Equation 18 are now available except vrot. As in the case of discrete observations, Equation 12 indicates how vrot should be calculated, but there is one subtle aspect. The appropriate value of vrot is not a projected distance divided by an elapsed time as it was in the discrete case (Figure 2), but neither is it the time average of the projected velocities x(t) and y(t). Instead, the value of vrot that corresponds to the angular velocity measured by digital tracking is given by the slopes of the best linear fits to x and y as functions of time. Figure 3 illustrates the difference between linear fit velocities and average velocities. For our data, the linear fit velocities of the Kitt Peak observer were 1340.0 km/hr eastward and -4.8 km/hr northward on April 19; and 1337.3 km/hr eastward and -16.3 km/hr northward on April 20; by contrast the average velocities were 1286.6 km/hr eastward and -6.9 km/hr northward on April 19 and 1283.0 km/hr eastward and -14.6 km/hr northward on April 20. In this case, the north-south velocities are negligible for practical purposes. – 14 – 5. Results from our Test Data 5.1. Initial Distance Determinations Our observations on 2013 April 19 span a temporal range of 5.72 hr. Those obtained on April 20 have the same range, and are centered 23.968 hr later in time. As this interval is only 2 minutes longer than a sidereal day, the timing of our observations is almost exactly optimal as defined in Section 3.1: the parameter χ from Equation 17 is only 0.0014. We detected a total of 215 asteroids through digital tracking analysis of our observations in a field of view only slightly larger than 1 square degree. We reported positions and brightnesses for all of these objects to the Minor Planet Center. Of these 215 MBAs, 202 were detected on both April 19 and April 20, and 197 of these two- night objects had sufficiently accurate measurements on both nights for good distance calculations. Of these, 48 were previously discovered objects with accurately known orbits, nine were previously discovered objects with poorly known orbits prior to our observations, and the remaining 140 were new discoveries for which we received designations from the Minor Planet Center. Heinze et al. (2015) provides further details. We have used Equation 18 to calculate distances for all 197 objects with sufficient measure- ments. The 48 objects in this sample with well-known orbits allow us to test the accuracy of rotational reflex distances using real data. The mean absolute error of our calculated distances for these objects is 0.042 AU, and the mean absolute fractional error is 2.2%. As expected, errors are smaller for more nearby objects. Distances with 2% accuracy are more than sufficient to analyze the size statistics of MBAs. Nevertheless, even better results are possible, as we will see below. 5.2. Correcting for Bias from Track Curvature Close investigation reveals some evidence of systematic error in the distance calculations de- scribed in Section 5.1. The weighted average signed fractional error for the 48 known objects is −0.0179± 0.0017: i.e., the calculated distances are 1.8% too small on average, and the offset is 10σ significant. It might be suspected that this bias is due to using the wrong value for vrot — i.e., that we were wrong to use linear fit velocities rather than time-averaged velocities (Section 4.2). This is easily demonstrated not to be the case, however. Substituting averaged values in place of linear fit values for vrot changes the result too much and in the wrong direction: the mean systematic offset goes from -1.8% to -5.2%. The true cause of the bias is more subtle, and points to a further interesting measurement that can be extracted from our data. – 15 – Fig. 3.— The projected motion of a ground-based observer due to the Earth’s rotation, which affects the measured positions and angular velocities of detected asteroids. Left: Schematic diagram of Earth viewed from the perspective of asteroids in our April 19, 2013 field, which was near -12◦ declination. The heavy black curve is composed of 126 X’s marking the projected position (x, y from Equation 12) of the Kitt Peak (31.9583◦ N lat) observer at the moment each of our images was taken. Top Right: x(t) and y(t) for our April 19 images, showing the linear fit velocities that should be used for comparison to linear, constant-velocity digital tracking results. Bottom Right: The velocities x(t) and y(t) for our data, showing how the mean velocities are significantly different from the best-fit velocities shown in the top panel. – 16 – With digital tracking, our measurement of the single-night observed angular velocity ωo is made over a period of time equal to the temporal span of our observations on each night. The instantaneous angular velocity over this interval is continuously changing, with the rotational reflex velocity itself being the dominant source of change. For example, the rotation of the Earth causes the westward sky motions of the asteroids to be slightly faster near midnight than they are at the beginning and end of each night’s observations (see Figure 3). As we stated in Section 4.2 and illustrated by Figure 3, we can account for the acceleration and curvature of an asteroid’s observed motion by using a value for vrot that is obtained by a linear fit to the observer’s projected coordinates x(t) and y(t). This assumes, however, that the angular velocity measurement is effectively an unweighted fit to angular position as a function of time. Such is not necessarily the case. For example, the images taken near midnight, when the westward sky motion is at its fastest, may be more sensitive due to better seeing or lower atmospheric extinction. The digital tracking stack will then effectively weight them more highly and produce a mean angular velocity that is systematically too fast. This effect would produce a systematic underestimation of the asteroids’ geocentric distances: exactly what we observe for known asteroids in our data. The effective weighting of different images will not necessarily be the same for different as- teroids. An asteroid that passes too close to a bright star near midnight will only be measured near the beginning and end of the night, resulting in a measured angular velocity slower than the true average value. Asteroids can also exhibit significant brightness changes during the night due to their own rotation, which will change the effective weighting of different images in the final measurement of their angular velocities. If asteroids actually moved in straight lines at constant velocity, none of these weighting effects would introduce error into our angular velocity determinations. If we could predict the curvature and acceleration of each asteroid’s track due to the Earth’s rotation, and apply small shifts to each image in our digital tracking stacks in order to linearize the motion, we would therefore remove the bias in our angular velocity determinations. We now describe how to do this. The angular distances between the position an observer measures for an asteroid and the position it would have if measured from the geocenter are simply a reflection of Equation 12, scaled by the distance: ∆α(t) = − ∆δ(t) = − x(t) d y(t) d (20) Averaged over the span of a digital tracking integration, ∆α(t) and ∆δ(t) contribute a constant ∆α(t) and ∆δ(t) offset to the measured position of an asteroid, and their first time derivatives – 17 – produce a constant offset (i.e. the RRV signal itself) to its measured angular velocity. Neither of these constant offsets concerns us here: we wish to isolate and remove only the component that represents curvature and acceleration. Thus, we create new functions ξ(t) and ζ(t) that behave exactly like ∆α(t) and ∆δ(t) except that over the span of our digital tracking integrations, their means and the means of their first time derivatives are zero: ξ(t) = ∆α(t) − h ∆α(t)i(t − hti) − h∆α(t)i ζ(t) = ∆δ(t) − h ∆δ(t)i(t − hti) − h∆δ(t)i (21) where the notation hi denotes the time-average of the enclosed quantity over the span of a digital tracking integration. The left panel of Figure 4 presents Equation 21 in graphical form, plotting the curved track of an asteroid in ξ, ζ space over the course of our April 19 observations. The calculation of ξ and ζ of course requires knowledge of the asteroid’s distance d, but the 2% accurate values from Section 5.1 are more than sufficient for this purpose. Thus, we adjust the methodology described in Section 4.2 for measuring precise angular ve- locities from digital tracking data: we apply additional shifts equal to −ξ(t),−ζ(t) to each postage stamp image before the final stack. This straightens the asteroid’s track so that curvature can no longer influence the calculated angular velocity. The distance can be recalculated based on the resulting new, de-biased angular velocities. Note that because Equation 21 subtracts average reflex velocities, we must now use averaged rather than linear-fit values for vrot (Section 4.2 and Figure 3 illustrate how the velocities differ). Compared to the calculation in Section 5.1, our curvature correction reduces the mean absolute error of the distances for known objects from 0.042 AU to 0.032 AU, and the mean absolute fractional error from 2.2% to 1.6%. More significantly, the new calculation changes the weighted average signed fractional error from −0.0179 ± 0.0017 to −0.0028 ± 0.0017. The systematic offset is thus reduced by a factor of six, and the new value of -0.3% is statistically consistent with zero. The left panel of Figure 5 illustrates the precision of our measured distances for known objects. We emphasize that these distance measurements were entirely calculated from first principles and known quantities describing the Earth. We have used the known asteroids in our data to quantify our errors, but not to determine any of the parameters used in the distance calculation. 5.3. Measurements of Track Curvature within a Single Night Beyond removing systematic errors in our measurements of asteroids’ mean angular velocities, we can in principle use Equation 21 to measure the curvatures of asteroid tracks within just a single night’s data. This is an extremely challenging measurement because the curvatures are so small: the RMS curvature amplitude, defined as the root mean of ξ2(t) + ζ 2(t), is less than 0.1 arcsec – 18 – Fig. 4.— Left: The curvature terms ξ(t), ζ(t) from Equation 21 for an asteroid at a distance of 2 AU in our April 19 data. The projected rotational velocity of the observer peaks at zero hour angle (near the center of the observing sequence); hence the asteroid’s retrograde (westward) angular velocity peaks at that time. Right: Measured rms curvature amplitudes (that is, the root mean of ξ2(t)+ ζ 2(t)) for asteroids brighter than R = 20.5 in our April 19 data, plotted against distance and compared to the predicted (not fit) relationship from Equations 20 and 21. Despite large scatter and a small systematic offset, the predicted 1/d dependence of curvature amplitude on distance is clearly seen, and the RMS error from this relationship is only 0.03 arcsec. In principle, curvature measurements with digital tracking data enable a measurement of an asteroid’s geocentric distance based on only one night’s data. – 19 – Fig. 5.— Left: The precision of our final RRV distance measurements for known objects. The line is the 1:1 correspondence and is not a fit: our distances are calculated from basic geometry and known quantities describing the Earth, not based on a fit to known objects in our data. As expected given the 1/d dependence of the RRV signal, our measurements are most precise for the nearest objects. Right: Histograms of the apparent magnitudes (and corresponding physical diameters) for all asteroids (solid line) and for previously known objects (dashed line) measured in our data. The current census of the main belt becomes substantially incomplete at a diameter of about 2 km. By contrast, we have detected dozens of new asteroids in the 200-500 meter size range. – 20 – for an asteroid at 2 AU. We attempt to measure it using our curvature-corrected postage-stamp analysis as described in Section 5.2, but instead of calculating ξ(t) and ζ(t) based on a previously- calculated approximate distance, we probe a range of RMS curvature amplitudes for each asteroid. For each trial curvature amplitude we calculate ξ(t) and ζ(t), apply the appropriate shifts to the postage stamp images, and perform a quadratic fit to flux as a function of angular velocity as described in Section 4.2. We compare the peak asteroid flux values produced by the quadratic fits for the different trial curvature amplitudes, and identify the curvature amplitude that yields the highest peak asteroid flux. This should be the curvature value that resulted in all of the individual asteroid images being most accurately registered. As the curvature amplitudes map uniquely to distances (Equations 20 and 21), our identification of the best-fit amplitude constitutes a distance measurement, in principle. However, the uncertainty of curvature amplitudes measured from the current data set is large enough that we do not choose to calculate individual curvature-based distances. Instead, we plot the best-fit curvature amplitudes as a function of the known RRV distances calculated in Section 5.2. A clear inverse relationship emerges, as predicted by Equation 20. The relationship gets tighter when we reject the minority of asteroids with out-of-range best-fit curvature values and all asteroids fainter than R = 20.5 mag, which cannot be measured as accurately. The 48 objects meeting these criteria in our April 19 observations are plotted in the right-hand panel of Figure 4, which shows the RMS curvature amplitude as a function of geocentric distance in AU, and compares the data with the predicted curve from Equations 20 and 21. Although there is a mild systematic offset, the slope of the best linear fit to curvature as a function of d−1 for these 48 asteroids is 9σ significant, matches the predicted value to within 0.4σ, and has an intercept only 1.4σ away from the predicted value of zero. The RMS error of the measured curvature amplitudes relative to their predicted values is only 0.03 arcsec. Thus we have effectively performed astrometry of moving, 20th magnitude objects with a precision of 30 milli-arcsec using only an 0.9m telescope. In principle, our success at measuring asteroid curvatures within a single night’s data means that digital tracking can be used to measure the geocentric distances to asteroids based on obser- vations from only one night. However, our current curvature measurements have uncertainties that are too large to yield useful distances for individual objects: we have been able to demonstrate the efficacy of our curvature-measuring methodology only by recourse to an ensemble of 48 asteroids with distances known by other means. With a larger telescope, more accurate curvature measure- ments of fainter asteroids would be possible, especially in good seeing. These could conceivably yield meaningful distance measurements for unknown main-belt asteroids over just a single night, albeit only for objects considerably brighter than the digital tracking detection limit. By contrast, the two-night RRV calculations of Sections 5.1 – 5.2 yield precise distances even for faint objects just above the detection threshold. There is therefore little reason to use the curvature method for asteroids measured on more than one night. However, where only one night’s data is available, distances based on digital-tracking curvature measurements could be very valuable. This is particularly true in at least two cases that may arise – 21 – in future surveys. The first is the case of an NEO survey where some objects — perhaps due to their fast motion — have inadvertently been measured on only one night. Being much closer to the Earth than MBAs, NEOs will show much larger and more easily measurable curvature, which is likely to yield useful distances. Such distances could enable the inclusion in a statistical analysis of objects that were otherwise unusable due to their one-night status. The second case is the one- night detection of a nearby NEO moving at nearly the same space velocity as Earth, such that its slow angular velocity mimics that of a much more distant object. This scenario is statistically rare but troubling, because such an object would likely be overlooked in an NEO survey and yet could be an incoming Earth-impactor. If the discovery survey used digital tracking, a curvature analysis would reveal the object’s actual, very small geocentric distance. We note in closing that curvature measurements can in principle be performed on asteroids detected with conventional methods rather than digital tracking, but they will normally be less accurate because of the smaller number of images available. 5.4. Distances and Sizes of Detected Asteroids While our current data only suggest the possibility of curvature-based distance measurements from a single night, they allow precise two-night distance measurements from rotational reflex velocity (Section 5.2). These precise distance measurements enable us to calculate the absolute magnitudes of each of our detected asteroids, and hence their physical diameters modulo the un- certainty in albedo. The histograms of these values are shown in the right-hand panel of Figure 5. At least half of the asteroids we have measured are smaller than 1 km. By contrast, only a small fraction of known MBAs are in this size range. The smallest objects we have detected are 300 meters in diameter under the assumption of low, 5% albedo, but could be as small as 150 meters if their albedo is 25%. The statistics of MBAs in this size range are already known (Gladman et al. 2009) based on a sample several times larger than we present herein. Given this, the detailed com- pleteness analysis that would be required to convert our detections into a measurement of the SFD is not worthwhile. The aims of the current work and our companion paper Heinze et al. (2015) are instead the validation of the RRV technique and of digital tracking for asteroids, respectively. Similar observations using a larger telescope, however, would extend to much smaller objects and accurately measure the SFD of a previously unexplored size regime in the main belt. 6. Conclusion We have described how the reflex angular velocity of asteroids due to Earth’s rotation can be used to determine the distances to main belt asteroids based on only two nights of observations. We refer to this as the rotational reflex velocity (RRV) method for measuring asteroid distances (Equa- tion 18). The required approximations are accurate to about 10−3 (Equation 7), and measurement uncertainties are typically 1-3%. Such distances can be used to calculate precise size statistics of – 22 – small main belt asteroids using a much smaller investment of time on a large telescope than has previously been required. Accurate RRV distances may be calculated either from conventional asteroid-search observations, from data analyzed by the technique of digital tracking (Zhai et al. 2014; Heinze et al. 2015), or from any other specialized asteroid-observation technique (e.g. Milani et al. 1996 or Gural et al. 2005) that accurately measures the celestial coordinates of the asteroids. We have tested RRV distance determination with a data set acquired on April 19 and 20, 2013 using the WIYN 0.9-meter telescope at Kitt Peak. Using measurements based on digital tracking, we have calculated distances to 197 asteroids in this data set. While the majority of these were new discoveries in our data, 48 of them are previously known objects with accurate orbits. These allowed us to test the accuracy of rotational reflex distances. A preliminary analysis yielded distances with a mean fractional error of only 2.2%. Without digital tracking, our detections would be confined to substantially brighter asteroids (or, the observations would require a larger telescope), but the accuracy of RRV distances for detected objects would likely be at least this good (see Section 3.3). We have identified a 1.8% systematic error in our preliminary RRV distances. We have linked this error to the curvature of the asteroids’ observed tracks that results from Earth’s rotation. This curvature, combined with the non-uniform sensitivity of images obtained under conditions of different seeing and airmass, causes a slight bias in the angular velocity measurements obtained from our digital tracking image-stacks. Finding our initial distance measurements more than sufficient to predict curvature amplitudes, we have calculated the form and amplitude of the curvature for each of our asteroids. We have applied a curvature correction to our image stacks to linearize the asteroid motions. Re-calculating the distances using angular velocities from curvature-corrected image stacks reduces the systematic bias from 1.8% to 0.3 ± 0.2%, and thus eliminates it as a significant effect. The corrected distances have a mean fractional error of only 1.6% for the 48 known objects in our data. Note that such a curvature correction would not be needed for RRV distances based on asteroids detected on discrete images. The calculation of accurate RRV distances would therefore be simpler for conventional asteroid detections, although digital tracking can detect much fainter objects. Besides correcting our digital tracking angular velocity measurements for the curvature due to Earth’s rotation, we have attempted to measure this curvature by creating several different image stacks for each asteroid at a range of different curvature amplitudes. As the curvature amplitude is proportional to 1/d, this constitutes a rotational reflex velocity measurement of an asteroid’s distance based on just a single night’s data. We find that we can indeed measure the curvature amplitudes for asteroids brighter than R = 20.5 mag with a precision of about 30 milli-arcsec. The measured curvature values for 48 asteroids brighter than this limit show the expected 1/d dependence at 9σ significance, and the best-fit slope is within 0.4σ of the predicted value. While distance measurements based on single-night curvature amplitudes are too noisy to be useful for asteroids in our current data set (and are not needed given our highly precise two-night values), single-night distances could be valuable in specific cases for future surveys. In particular, curvature measurements can identify nearby objects moving at slow angular velocities characteristic of a much – 23 – more distant population. Our precise distances from two-night RRV measurements allow us to calculate absolute mag- nitudes and hence approximate diameters for our newly discovered asteroids. Our faintest objects have HR ∼ 21.5 mag and hence diameters of 130–300 meters depending on their albedo. While the current census of the main belt becomes substantially incomplete at a diameter of about 2 km, we have detected dozens of new asteroids in the 200–500 meter size range with an 0.9m telescope. 7. Acknowledgments Based on observations at Kitt Peak National Observatory, National Optical Astronomy Obser- vatory (NOAO Prop. ID: 2013A-0501; PI: Aren Heinze), which is operated by the Association of Universities for Research in Astronomy (AURA) under a cooperative agreement with the National Science Foundation. This publication makes use of the SIMBAD online database, operated at CDS, Strasbourg, France, and the VizieR online database (see Ochsenbein et al. (2000)). We have also made extensive use of information and code from Press et al. (1992). We have used digitized images from the Palomar Sky Survey (available from http://stdatu.stsci.edu/cgi-bin/dss_form), which were produced at the Space Telescope Science Institute under U.S. Government grant NAG W-2166. The images of these surveys are based on photographic data obtained using the Oschin Schmidt Telescope on Palomar Mountain and the UK Schmidt Telescope. Facilities: 0.9m WIYN REFERENCES Allen, R. L., Bernstein, G. M., & Malhotra, R. 2001, ApJ, 549, L241 Bernstein, G. & Kushalani, B. 2000, AJ, 120, 3323 Bernstein, G. M., Trilling, D. E., Allen, R. L., Brown, M. E., Holman, M., & Malhotra, R. 2004, AJ, 128, 1364 Bottke, W. F. Jr., Durda, D. D., Nesvorn´y, D., Jedicke, R., Morbidelli, A., Vokrouhlick´y, D., & Levison, H. 2005, Icarus, 175, 111 Bottke, W. F. Jr., Durda, D. D., Nesvorn´y, D., Jedicke, R., Morbidelli, A., Vokrouhlick´y, D., & Levison, H. F. 2005, Icarus, 179, 64 de El´ıa, G. C. & Brunini, A. 2007, A&A, 466, 1159 Farinella, P., Froeschl´e, Chr., Froeschl´e, Cl., Gonczi, R., Hahn, G., Morbidelli, A., & Valsecchi, G. B. 1994, Nature, 371, 314 – 24 – Fraser, W. C. & Kavelaars, J. J. 2009, AJ, 137, 72 Gladman, B. J., Davis, D. R., Neese, C., Jedicke, R., Williams, G., Kavelaars, J. J., Petit, J-M., Scholl, H., Holman, M., Warrington, B., Esquerdo, G., & Tricarico, P. 2009, Icarus, 202, 104 Gural, P. S., Larsen, J. A., & Gleason, A. E. 2005, AJ, 130, 1951 Heinze, A. N., Metchev, S., & Trollo, J. 2015, AJ, submitted Masiero, J. R., Mainzer, A. K., Bauer, J. M., Grav, T., Nugent, C. R., & Stevenson, R. 2013, ApJ, 770, 7 Milani, A., Villani, A., & Stiavelli, M. 1996, Earth, Moon, and Planets, 72, 257 Ochsenbein, F., Bauer, P. & Marcout, J. 2000, ApJS, 143, 23O Pravec, P., Harris, A. W., Kusnir´ak, P., Galad, A., & Hornoch, K. 2012, Icarus, 221, 365 Press, W. H., Teukolsky, S.A., Vetterling, W. T., & Flannery, B. P. 1992, Numerical Recipes in C (Second Edition; New York, NY: Cambridge University Press) Yoshida, F., Nakamura, T., Watanabe, J. 2003, PASJ, 55, 701 Yoshida, F. & Nakamura, T. 2007, P&SS, 55, 1113 Zhai, C., Shao, M., Nemati, B., Werne, T., Zhou, H., Turyshev, S. G., Sandhu, J., Hallinan, G., & Harding, L. K. 2014, ApJ, 792, 60 This preprint was prepared with the AAS LATEX macros v5.2.
1710.05276	1	1710	2017-10-15T05:33:37	High-Resolution Spectroscopic Detection of TiO and Stratosphere in the Day-side of WASP-33b	[ "astro-ph.EP" ]	We report high-resolution spectroscopic detection of TiO molecular signature in the day-side spectra of WASP-33 b, the second hottest known hot Jupiter. We used High-Dispersion Spectrograph (HDS; R $\sim$ 165,000) in the wavelength range of 0.62 -- 0.88 $\mu$m with the Subaru telescope to obtain the day-side spectra of WASP-33 b. We suppress and correct the systematic effects of the instrument, the telluric and stellar lines by using SYSREM algorithm after the selection of good orders based on Barnard star and other M-type stars. We detect a 4.8-$\sigma$ signal at an orbital velocity of $K_{p}$= +237.5 $^{+13.0}_{-5.0}$ km s$^{-1}$ and systemic velocity $V_{sys}$= -1.5 $^{+4.0} _{-10.5}$ km s$^{-1}$, which agree with the derived values from the previous analysis of primary transit. Our detection with the temperature inversion model implies the existence of stratosphere in its atmosphere, however, we were unable to constrain the volume-mixing ratio of the detected TiO. We also measure the stellar radial velocity and use it to obtain a more stringent constraint on the orbital velocity, $K_{p} = 239.0^{+2.0}_{-1.0}$ km s$^{-1}$. Our results demonstrate that high-dispersion spectroscopy is a powerful tool to characterize the atmosphere of an exoplanet, even in the optical wavelength range, and show a promising potential in using and developing similar techniques with high-dispersion spectrograph on current 10m-class and future extremely large telescopes.	astro-ph.EP	astro-ph	DRAFT VERSION OCTOBER 17, 2017 Typeset using LATEX twocolumn style in AASTeX61 HIGH-RESOLUTION SPECTROSCOPIC DETECTION OF TIO AND STRATOSPHERE IN THE DAY-SIDE OF WASP-33B STEVANUS K. NUGROHO,1, 2 , ∗ HAJIME KAWAHARA,2, 3 KENTO MASUDA,4, 5 TERUYUKI HIRANO,6 TAKAYUKI KOTANI,7, 8 AND AKITO TAJITSU9 1Astronomical Institute, Tohoku University, Sendai 980-8578, Japan 2Department of Earth and Planetary Science, The University of Tokyo, Tokyo 113-0033, Japan 3Research Center for the Early Universe, School of Science, The University of Tokyo, Tokyo 113-0033, Japan 4Department of Astrophysical Sciences, Princeton University, Princeton, NJ 08544, U.S.A 5NASA Sagan Fellow 6Department of Earth and Planetary Sciences, Tokyo Institute of Technology, Tokyo 152-8551, Japan 7National Astronomical Observatory of Japan, Tokyo 181-8588, Japan 8Astrobiology Center, National Institutes of Natural Sciences, Tokyo, Japan 9Subaru Telescope, 650 N. Aohoku Place, Hilo, HI 96720, U.S.A ABSTRACT We report high-resolution spectroscopic detection of TiO molecular signature in the day-side spectra of WASP-33 b, the second hottest known hot Jupiter. We used High-Dispersion Spectrograph (HDS; R ∼ 165,000) in the wavelength range of 0.62 – 0.88 µm with the Subaru telescope to obtain the day-side spectra of WASP-33 b. We suppress and correct the systematic effects of the instrument, the telluric and stellar lines by using SYSREM algorithm after the selection of good orders based on Barnard star and other M-type stars. We detect a 4.8-σ signal at an orbital velocity of Kp= +237.5 +13.0−5.0 km s−1 and systemic velocity Vsys= -1.5 +4.0−10.5 km s−1, which agree with the derived values from the previous analysis of primary transit. Our detection with the temperature inversion model implies the existence of stratosphere in its atmosphere, however, we were unable to constrain the volume-mixing ratio of the detected TiO. We also measure the stellar radial velocity and use it to obtain a more stringent constraint on the orbital velocity, Kp = 239.0+2.0−1.0 km s−1. Our results demonstrate that high-dispersion spectroscopy is a powerful tool to characterize the atmosphere of an exoplanet, even in the optical wavelength range, and show a promising potential in using and developing similar techniques with high-dispersion spectrograph on current 10m-class and future extremely large telescopes. Keywords: techniques: spectroscopic - planets and satellites: atmospheres - planets and satellites: composition - planets and satellites: individual (WASP-33 b) 7 1 0 2 t c O 5 1 . ] P E h p - o r t s a [ 1 v 6 7 2 5 0 . 0 1 7 1 : v i X r a Corresponding author: Stevanus K. Nugroho [email protected] ∗ Indonesia Endowment Fund for Education Scholar 2 NUGROHO ET AL. 1. INTRODUCTION Thermal inversion in exoplanetary atmosphere is still a problem not completely understood in the theory of exo- planetary atmosphere. This inversion layer was predicted by Hubeny et al. (2003) and Fortney et al. (2008) in a highly irradiated planet, which is caused by high-temperature ab- sorber molecules such as TiO or VO that absorb UV and vis- ible radiation from the incoming stellar radiation, and heat up the upper atmosphere. By measuring shallower eclipse depths at 4.5 µm and 5.6 µm, caused mainly by H2O and CO, than the thermal continuum at the nearby band (3.6 and 9 µm) using the Spitzer Space Telescope (Spitzer), the first evidence of inversion layer was reported by Knutson et al. (2008) in the atmosphere of HD 209458 b. Although there have been similar observations of secondary eclipse depth by the same telescope suggesting the detection of an inver- sion layer, it was shown by Hansen et al. (2014) that several previous single-transit observations using Spitzer have sig- nificantly higher uncertainties than previously known, which made the measured emission-like features doubtful. Using the CRyogenic high-resolution InfraRed Echelle SPectro- graph (CRIRES Kaeufl et al. 2004) in the Very Large Tele- scope (VLT), Schwarz et al. (2015) observed the day-side of HD 209458 b and reported the non-detection of CO at 2.3 µm and constrained the nonexistence of the inversion layer in the pressure range of 10−1–10−3 bar, which is consis- tent with the results by Zellem et al. (2014) and Diamond- Lowe et al. (2014). Despite the non-detection of CO in the day-side of HD 209458 b, Snellen et al. (2010) detected a CO absorption feature at 5.6-σ using a similar instrument in the transmission spectrum. One of the possible causes of the non-detection is that the average day-side atmosphere of HD 209458 b is near-isothermal in the pressure range that they probe, which makes the day-side spectrum almost featureless (Schwarz et al. 2015). Evans et al. (2017) reported the detection of H2O emission features in a super-hot Jupiter, WASP-121b (Teq ∼ 2500 K), using the combination of Hubble Space Telescope (HST) and Spitzer. The presence of water was resolved and detected at 5-σ confidence level, strengthening the previous conclu- sion by Evans et al. (2016), which made WASP-121b the first exoplanet with resolved emission features and detected stratosphere layer. Their findings also include the possible detection of VO in its atmosphere at the level of 1000x solar abundance, and a thermal inversion-like atmospheric struc- ture, even though they only used a 1D atmospheric model structure and did not take the non-equilibrium chemistry into account. Sedaghati et al. (2017) observed WASP-19b dur- ing transit using a low-dispersion spectrograph, FORS2, on VLT. They confirmed the presence of water (7.9-σ) and si- multaneously revealed the presence of TiO (7.7-σ), strongly scattering haze (7.4-σ) and sodium (3.4-σ). They also con- strained the relative abundance of those molecules, however, the presence of a thermal inversion remains unproved be- cause a transmission spectrum has less information on tem- perature structure of the atmosphere. Other evidence of ther- mal inversion has also been reported in the atmosphere of WASP-33 b (Teq ∼ 2700 K), which is the second hottest hot Jupiter, by Haynes et al. (2015), who observed the day-side of the exoplanet using HST, and revealed a thermal excess at about 1.2 µm, which is consistent with the TiO spectral fea- ture. Other than these, there has been no significant evidence, nor direct detection of TiO and/or VO in the atmosphere of hot Jupiters. The non-detection of TiO or VO can be caused by sev- eral factors. Hubeny et al. (2003) and Spiegel et al. (2009) suggested that gravitational settling could drag TiO/VO from the upper atmosphere to the colder layer in the deeper atmo- sphere. Meanwhile, owing to high-speed winds of the tidally locked hot Jupiter, condensed molecules can also be brought into the colder night side of the planet. If the temperature of the atmosphere is below the TiO/VO condensation level, it can condensate and if the vertical mixing rate, which is re- lated to the temperature of the planet atmosphere is not high enough, it cannot be redistributed to the upper atmosphere. This effect is called the cold-trap effect. Knutson et al. (2010) found a possible connection between the UV chromospheric stellar activity and the existence of an inversion layer in the atmosphere of known hot Jupiters. Hot Jupiters orbiting ac- tive stars tend to have no inversion layer and those orbiting quiet stars showed evidences of the existence of an inver- sion layer in their atmosphere. Knutson et al. (2010) sug- gested that the increased UV intensity can probably destroy the compounds that are responsible of creating an inversion layer. Meanwhile, using high-dispersion spectroscopy, Hoei- jmakers et al. (2015) reported the inaccuracy of the TiO line list that was used at wavelengths shorter than 6300 A. The accuracy level, however, tends to increase at longer wave- lengths, which was shown by the cross-correlation result be- tween the spectrum of Barnard's Star and the model spectrum created by using the corresponding TiO line list (see Figure 9 of Hoeijmakers et al. 2015). Recently, direct detection of molecular signature in exo- planet atmosphere using high-dispersion spectroscopy is one of the most widespread approach to the attempt of exoplanet characterization (e.g Snellen et al. 2010; Crossfield et al. 2011; Birkby et al. 2013; de Kok et al. 2013; Hoeijmak- ers et al. 2015; Schwarz et al. 2015; Birkby et al. 2017; Esteves et al. 2017). Unlike low-dispersion spectroscopy, high-dispersion spectroscopy can resolve molecular bands into individual absorption lines. The variation of Doppler shifts during observation caused by the orbital movement en- ables to distinguish absorption lines in the exoplanet spec- trum from telluric lines, and ensures the unambiguous de- TIO WASP-33B DRAFT 3 tection of specific molecules. Owing to these resolved indi- vidual lines it is also possible to investigate several physical parameters of the exoplanet such as the axial tilt (Kawahara 2012), the projected equatorial rotational velocity (Snellen et al. 2014), wind speed (Brogi et al. 2016), and the thermal inversion layer (Schwarz et al. 2015) of the planet. Obtain- ing the exoplanet spectrum can be done by cross-correlating the data with the exoplanet atmosphere spectrum model, after removing telluric and stellar lines using a specific method. We observed WASP-33 b (Smith et al. 2011), which is the second hottest known hot Jupiter (wavelength dependent brightness temperature of 3620 K), orbiting quiet δ Scuti stars. It has a retrograde orbit with a period of ∼ 1.22 days. It is an ideal choice for transmission spectroscopy measure- ments: owing to its unusually large radius (Collier Cameron et al. 2010) and its high temperature making it not only suit- able for secondary eclipse spectroscopy measurements, but also as the main target to find TiO/VO in its atmosphere, as the cold-trap effect is unlikely at this temperature level. In this paper, we report a direct detection of TiO molecules and a stratosphere layer of WASP-33b, based on ground- based observation of the day-side emission spectrum in the visible wavelength range (6170-8817 A). The observation, data reduction, and systematic effect removal (including the correction of the blaze function variation, common wave- length grid, and telluric and stellar line removal) are de- scribed in §2. We also examine our methods to create the model spectrum and to cross-correlate the data with the model spectrum. Here, we also explain the method to con- firm the radial velocity of WASP-33 and the accuracy of TiO line list used by us. In §3, the possibility of signal detection of TiO and the order-by-order (order-based) optimization of the SYSREM algorithm are explored. Here, we also show the final result after the optimization and the statistical test. This is followed by the discussion and the conclusions our findings in §4. 2. OBSERVATION AND DATA REDUCTION 2.1. Subaru Observation of WASP 33 We observed WASP-33 on UT October 26 2015 (see Table 1 for stellar and exoplanet physical and dynamical param- eters) using High Dispersion Spectrograph (HDS Noguchi et al. 2002) at f/12.71 optical Nasmyth focus of the the Sub- aru 8.2 m telescope (proposal ID: S15B-090, PI: H. Kawa- hara). The observation was conducted in a standard NIRc set up (without Iodine cell) with Messia5, 2x1 binning set- ting, and without image rotator. Image slicer 3 (Tajitsu et al. 2012), each with slit width= 0.(cid:48)(cid:48)2 was used, resulting in the highest spectral resolution of R= 165,000 (1.8 km s−1 reso- lution), which were sampled by two detectors (blue and red CCDs) with 4100 × 2048 pixels (0.9 km s−1 per pixel) con- taining 18 orders covering 6170 – 7402 A, and 12 orders cov- Figure 1. The coverage of WASP-33 b orbital phase during our ob- servation (showed by bold red line). The vertical dotted line shows the ingress and egress phase ering 7537 – 8817 A, respectively. We obtained 52 spectra of WASP 33, each with an exposure time of 600 s, from air mass 2.97 to 1.96 on the other meridian covering 0.23 – 0.56 exo- planet orbital phases (see Table 1 for ephemeris and derived the orbital period, the orbital coverage of our observation is shown by bold red line in Figure 1) with a typical seeing of ∼ 0.(cid:48)(cid:48)6 – 0.(cid:48)(cid:48)7. 2.2. M-dwarfs Spectra Hoeijmakers et al. (2015) showed that the TiO line list used by them is not accurate at short wavelengths (< 6000 A) however, it tends to be more accurate at longer wavelengths. For a robustness of analysis, we checked the accuracy of the TiO line list we use in this paper, by comparing it with the spectra of Barnard's Star (M4V) and HD 95735 (M1.5V), as TiO is commonly found in the atmosphere of M-type stars and brown dwarfs (Burrows & Sharp 1999; Burrows et al. 2001). In 2016, we observed Gl752A (M3V), and HD173739 (M3V) with the Subaru telescope using the same instrumen- tal configuration as in 2015 (proposal ID: S16A-107, PI: H. Kawahara). To check the robustness of our cross-correlation we also downloaded and used the calibrated spectrum of Proxima Centauri from the ESO Science Archive Facility. The spectrum was obtained by XSHOOTER at VLT between 5336.6 A and 10200 A with a resolution of R= 18,000 (pro- posal ID: 092.D-0300(A), PI: Neves). 2.3. Standard Reduction The data were reduced using IRAF tools1 and a custom- built application written in Python 2.7. We corrected the 1 The Image Reduction and Analysis Facility (IRAF) is distributed by the US National Optical Astronomy Observatories, operated by the Associ- 4 NUGROHO ET AL. (a) (b) Figure 2. Relative shifts of telluric lines compared to those in the first frame in the blue CCD (1) and red CCD (2). Noticeably the general trend of the shift follows the trend of temperature variation inside the dome (solid red line) over-scan and non-linearity using CL scripts that can be ob- tained from the HDS website2 before debiasing the frames. During the analysis of Narita et al. (2005) data, Snellen et al. (2008) noticed the non-linearity effect in HDS. After fixing this problem empirically, the analysis gave the detection of sodium in the atmosphere of HD 209458b. The CCD re- sponse (gain) for high signal-to-noise (S/N) spectra is higher than those in a lower S/N spectra, which would make the correction of telluric and stellar lines difficult. This prob- lem was solved by Tajitsu et al. (2010), who provided a CL script to correct this effect after applying the over-scan cor- rection. The scattered light in the inter-order area was fitted with a cubic spline function along the slit and dispersion di- rection individually, using apscatter for the arc lamp frame, smoothed and subtracted from all science frames. A median flat frame was calculated from 71 dome-flat frames to cor- rect the pixel-to-pixel sensitivity variation, and normalized using the apnormalize task in IRAF in order to conserve the fringe pattern along the slit direction. Using the CL script hdsis ecf.cl 3, we flat-fielded and extracted 1-dimensional spectra of total 30 orders. Using hdsis ecf.cl the science spec- tra were sliced along the slit from −12 to +8 pixels and from −11 to +8 pixels relative to the center of the order tracer of each order for the blue and red CCD respectively, divided them by the slice of the normalized median flat frame and sum combined. If a larger width was used, a high-frequency noise would appear along the wavelength in the extracted ation of Universities for Research in Astronomy, Inc., under a cooperative agreement with the National Science Foundation. 2 http://www.subarutelescope.org/Observing/ Instruments/HDS/index.html 3 https://www.naoj.org/Observing/Instruments/HDS/ hdsql/hdsql-cl-20170807.tar.gz spectra, caused by the edge of the aperture, which was de- fined earlier in apnormalize by using a median flat frame. We derived the wavelength solution by identifying 133 emission lines of Thorium-Argon arc lamp frames taken at the begin- ning and the end of the observation, and fitted with a fifth and third order Chebyshev function corresponding to the dis- persion and slit directions respectively, using ecidentify. The pixel RMS value of the residual fitting was ∼ 0.0012 for both CCDs. Then, the spectra were assigned by interpolating be- tween the preceding and the following Thorium-Argon arc lamp frames in respect of the observation time, using ref- spec with the relative distance of the spectra to the reference spectra as the weight of the interpolation. Then, the wave- length solution was applied using dispcor. These steps were applied to the non-normalized median flat frame to create the corrector for the blaze function. All science frames were then divided by the blaze function corrector to create the final re- duced frames. Then the next step is correcting the variation of blaze function along the time which is given in detailed in Appendix A. 2.4. Common Wavelength Grid We measured the shifts of the spectrum during the observa- tion, by measuring the relative shifts of strong telluric lines to the reference spectrum. First, we detected telluric lines using peakutils4 (15 – 20 lines on average per selected or- der) in the order that contain well-spread telluric lines in the wavelength direction. We then fitted a Gaussian function to determine the precise centroid of each line and compared it with the centroid of the same lines in the first frame. The relative shift of each order was calculated by taking the me- 4 https://bitbucket.org/lucashnegri Table 1. System parameters of WASP-33 Parameter Value WASP-33 M(cid:63) (M(cid:74)) R(cid:63) (R(cid:74)) L(cid:63) (L(cid:74)) Spectral type Tef f (K) log g [Fe/H] [M/H] Centre-of-mass velocity (km s−1) vrot sin i(cid:63) (km s−1) d (pc) WASP-33b To-2450000 (BJD) P (days) Ttransit (days) Tingress (days) a/R(cid:63) i (◦) 2015 MP (MJ) RP (RJ) log gP [CGS] Kp (km s−1) a 1.561+0.045 −0.079 1.512 ± 0.04 d 1.495 ± 0.031 b 1.509+0.016 −0.027 6.17 ± 0.43 A5 7400 ± 200b 4.3 ± 0.2 0.1 ± 0.2b 0.1 ± 0.2a -2.19 ± 0.09b -3.69 ± 0.09b -2.11 ± 0.05b 86.63+0.37−0.32 117 ± 2 c 6934.77146 ± 0.00059c 1.2198709a 0.1143 ± 0.0002a 0.0124 ± 0.0002a c a 3.69 ± 0.01a 88.695+0.031 −0.029 3.266 ± 0.726a 1.679+0.019 −0.030 a 3.46+0.08−0.12 231.11 +2.20−3.97 228.67 +2.00−2.04 227.81 +1.56−31.59 b d a aAdopted from Kov´acs et al. 2013 b Adopted from Collier Cameron et al. (2010) c Adopted from Johnson et al. (2015) dAdopted from Smith et al. (2011) dian combination of the shifts in that order. There are order dependent shifts of about 0.05 pixel, but because the infor- mation for the order that does not has strong telluric lines is not available and the value is too small to affect the result TIO WASP-33B DRAFT 5 (0.05 pixels corresponding to ∼0.04 km s−1), we decided to ignore it. The results of all selected order were combined to estimate the shift of the corresponding frame, by calculating its weighted median combined. The amplitude of relative shifts is about 0.13 and 0.08 pixel (∼0.12 and 0.07 km s−1) for blue and red CCD, respec- tively. Noticeably the general trend of the shifts are cor- related with the shifts of the temperature inside the dome (∆T ∼1.75 K) during the observation (see Figure 2 a & b). This wavelength-temperature shift relation trend is relatively weaker than those for the CRIRES in VLT reported by Brogi et al. (2013) (1.5 K change in temperatures corresponding to 1.5 pixels shifts), which gave a non-detection of the 51 Peg b signal due to the odd-even effect for detector 4. Then, the spectra were shifted with spline interpolation, based on the weighted median combined shift of each frame, and then re-sampled into a common wavelength bin. The spectra of each order were then stacked in a matrix with wavelength bin values in its column, and spectrum num- bers (or time/orbital phases) in the rows. 5-σ clipping was performed for each wavelength bin to identify any bad pix- els/cosmic rays, which are replaced by the mean value of the bin. We excluded these bad region for the rest of the analy- sis, which is 5.30% and 7.19% of all pixels in blue and red CCDs, respectively, mostly because they are on the edge of the CCD. 2.5. Removal of Telluric and Stellar Absorption Lines In the studied wavelength range, telluric and stellar ab- sorption lines are dominating the spectrum, while the exo- planet to stellar flux contrast is expected at the level of 10−3 (when extrapolated from the best-fitted spectrum in Haynes et al. 2015). Removal of telluric and stellar lines is critical in order to detect the TiO signature by cross-correlation, as the exoplanet spectrum is buried under sharp noises due to those lines. The the spectrum of WASP-33b is expected to be Doppler-shifted from +230 km s−1 (at its maximum) to −54 km s−1 (at the end of the observation), while telluric and stellar lines remain relatively stationary during the course of the observation. Telluric lines, which are dominated mostly by water and oxygen lines, are varied in strength due to the changes of geometric air mass and water vapor column of the atmosphere. The quasi-static telluric and stellar absorption lines were removed by implementing the SYSREM algorithm to re- move systematic effects (variation of atmospheric condition, the changing of CCD efficiency, variation of point spread function, etc.) without any priors in a large set of light curves (Tamuz et al. 2005; Mazeh et al. 2007). It has been used either in the detrending/systematic effects removal of transit surveys (e.g SuperWASP, CoRoT light curves Collier Cameron et al. 2006; Ofir et al. 2010), or in the removal of 6 NUGROHO ET AL. (a) (b) Figure 3. Reduction process for each order of both blue (a) and red CCDs (b). The first row ([1]) shows an example of the normalized 1D spectrum of WASP-33 following the correction of the blaze function variation and the common wavelength grid iteration. The next row ([2]) shows the 2D spectrum with the wavelength as the horizontal axis, each label representing the median wavelength of the order, while the vertical axis is the frame number. The third row ([3]) shows the final reduced spectra after the correction of the blaze function in common wavelength grid. The variation of brightness along the frames for all orders is due to the blaze function variation. The fourth row ([4]) shows the mean subtracted spectra as the input to SYSREM. The fifth and sixth rows ([5] & [6]) show the residual spectra after running SYSREM with 1 and 4 iteration and through the double high-pass filter, at the latter stage almost all telluric lines have been removed. The masked bad regions are shown by grey areas in all rows except the first one. TIO WASP-33B DRAFT 7 quasi-static telluric lines in the similar analysis to this work (e.g. Birkby et al. 2013, 2017; Esteves et al. 2017). Each wavelength bin (4100 wavelength bin per order on average) is treated as a "light-curve" consisting of 52 frames (includ- ing the frames when WASP-33 b in the secondary eclipse phase). Then each "light-curve" was subtracted from their mean value before applying the SYSREM algorithm order- by-order (fourth row in Figure 3). The detailed description of the SYSREM algorithm is given in Appendix B. The step-by-step removal of telluric can be seen in Fig- ure 3. It can be noticed that there are curve shaped features, which can be seen in the blue CCD spectra matrix, and the degree of the curvature decreases as increasing wavelength. These features can be caused by the imperfection of the cor- rection of the blaze function variation. As the residual was cross-correlated with the high-frequency TiO features, this feature does not affect the results and these were roughly re- moved after a double high-pass filter was applied (see Figure 3). Any linear systematic variation along all wavelength bins in each order can be found and removed from the spectrum starting with the most significant one, such as air mass vari- ation. Ideally, the expected residual is the Doppler-shifted exoplanet spectrum only, with additional noise. We perform two cases of the SYSREM reduction: One is that we re- move the first ten systematics (henceforth SYSREM itera- tion, Nsys = 10) for all of the orders (SYSREM with com- mon iteration number in §3.1). This simple procedure gives a conservative estimate of the signal detection. Another at- tempt is that we determine the optimum number of subse- quent SYSREM iterations for each order that gives the opti- mum systematic removal (Order-based SYSREM optimiza- tion in §3.2). The latter is based on the fact that SYSREM tends to remove the exoplanet spectrum after the Doppler- shifted variation dominate the systematic change in the spec- trum. Before performing cross-correlation for each residual, we applied a double high-pass filter similarly to the case of M- dwarfs star spectrum, using a smoothing function with a 25- pixel width for the first filter and 51-pixel width for the sec- ond filter, to remove any low-frequency variations along the spectrum, before cross-correlating with a grid of the Doppler- shifted WASP-33b model spectrum. Then, the values of each wavelength bin were weighted by their noise, which is de- fined by the standard deviation of the bin as a function of time. The results of this removal process are 10 sets of smoothed weighted SYSREM residuals for each CCD. 2.6. Spectral Template To constrain the existence of TiO molecules in WASP- 33 b emission spectra by cross-correlation, model spectra was generated with various profiles and volume mixing ra- tio (VMR) of TiO, assuming several atmospheric models as described in Table 2. Three different temperature–pressure (T/P) profiles were adopted, which described the average ver- tical temperature structure of the planetary day-side atmo- sphere: full inversion (FI), non-inversion (NI) as shown in the right panels of Figure 4, and T/P profile from Haynes et al. (2015), known as the H-model, as shown in the right panel of Figure 5. For the NI model, it was assumed that the temper- ature decreases with a constant lapse rate from P0= 102 bar with T0= 3700 K to P1= 10−5 bar with T1= 2700 K. For the FI model, the temperature increases from P0= 102 with T0= 2700 K to P1= 10−5 with T1= 3700 K. For both models, a constant TiO concentration at solar abundance level was as- sumed, that is, VMR= 10−7.2 and no other molecule in the atmosphere. For the H-model, seven constant VMRs of TiO from sub-solar to super-solar abundance level were assumed, log VMRT iO= [−4, −5, −6, −7, −8, −9, −10]. The atmo- sphere was divided into 50 layers, which were evenly spaced in log pressure between 102 to 10−5 bar, and the altitudes were calculated by assuming hydro-static H2 dominant at- mosphere and using the derived mass and a radius of WASP- 33b from the reference (Table 1). We also modeled the non- inverted atmosphere (henceforth M-dwarf model) with T0= 2600 K at P0= 0.01 bar and T1= 4000 K at P1= 1 bar assum- ing a constant rate of temperature change with log pressure, resulting in TiO absorption features only for log VMR= −7. Table 2. Models of planetary atmosphere Name FI NI H T/P Profile Full inversion No inversion log VMR −7 −7 Realistic inversion (Haynes et al. (2015)) −4 – −10 The cross section of molecules was calculated using Python scripts for Computational Atmospheric Spectroscopy (Py4CATS5). The procedure of computing the cross sec- tion is described in Appendix C. Then, the cross-sections were combined into continuum absorption coefficients and integrated along the line-of-sight through the atmosphere, resulting delta optical depths per layer. The Schwarzschild equation was solved by integrating the Planck function ver- sus the monochromatic transmission, along the line-of-sight and convolved with a Gaussian kernel to match with the HDS resolution. In total, we produced 9 WASP-33b mock spectra and 1 non-inverted model for TiO line list accuracy analysis; 1 full inversion (FI-spec), 1 no inversion (NI-spec), 7 spectra for Haynes (H-spec), and 1 M-dwarf model. See Figure 4 for 5 see http://www.libradtran.org for details 8 NUGROHO ET AL. FI-spec and NI-spec, Figure 5 for H-spec, and Figure 6 for 5 M-dwarfs spectra+1 M-dwarf model spectrum. In this analysis, the systemic velocity (Vsys) is one of the important parameters to confirm the detection (if there is any); thus by measuring radial-velocity (RV) of WASP-33 and comparing it with the previous results, the confidence and the robustness of our analysis are improved. Instead of creating the WASP-33 comparison spectra by our self, we used the stellar model spectrum from Coelho (2014) for Teff= 7500 K, log g= 4.5, [Fe/H]= +0.2 and [α/Fe]= 0. The stellar model spectrum was convolved to the HDS resolution and rotationally broadened to resemble the Doppler broadening caused by its fast rotation (henceforth Coelho model spec- trum). 2.7. System Velocity from Stellar Spectra To measure the radial velocity of the WASP-33 system, we analyzed the standard reduced WASP-33 spectra and masked the telluric lines of the selected order that contain signifi- cant stellar absorption lines. Then, the spectra were cross- correlated order-by-order with the Doppler-shifted Coelho model spectrum from −100 km s−1 to +100 km s−1 with 0.1 km s−1 intervals. In this cross-correlation, only the spec- tra in blue CCD was used, because there are a lot of telluric lines and less stellar absorption lines in the red CCD spec- trum. For each selected order, the RV value with the highest cross-correlation signal was extracted, and median combined to calculate the RV of WASP-33 for the corresponding frame. Then, the weighted median combine of the RV of all frames was calculated, following the correction of the barycentric ra- dial velocity difference to have the final WASP-33 RV value. The results can be seen in Figure 7b, and the median of the measured RV is −3.02 ± 0.42 km s−1. Although this value is different from the value −9.2 ± 2.8 km s−1 (Gontcharov 2006) in the SIMBAD database, our result is consistent with Collier Cameron et al. (2010), which used Doppler tomog- raphy technique to confirm the existence of the planet and measured the center-of-mass velocity (see Table 1). There- fore we used this value as the radial velocity of the WASP-33 system, which is taken into consideration when evaluating the result of the WASP-33b vs TiO model cross-correlation analysis. 2.7.1. Line List Accuracy and Excluding Bad Orders To check the TiO line list accuracy, five standard reduced M-dwarf spectra were cross-correlated order-by-order with the M-dwarf model. The model spectra were Doppler-shifted from −80 km s−1 to +80 km s−1 relative to the RV of the expected target, from SIMBAD database with 1 km s−1 in- tervals. The object and model spectrum was divided by their continuum profile, calculated by applying a double high-pass filter with 501 pixels and 1001 pixels of smoothing function before the cross-correlation, in order to maintain the cross- correlation scale (from −1 to 1). The cross-correlation results of five M-dwarf spectra ver- sus the M-dwarf model is shown in Figure 8. As Hoeijmak- ers et al. (2015) expected, the accuracy of the TiO line list improved in the longer wavelength, although there are cross- correlation functions (CCFs) of several order, which blue- shifted from their expected radial velocity, and several others do not have any significant peak. For the other of the order, the peaks are located at their expected radial-velocity. The shifts are most likely caused by the inaccuracy of the TiO line list itself, because the first three orders of the blue CCD show similar blue-shifted CCF for the five different M-dwarf spec- tra. The shift is about ∼ 21 km s−1 (shown by blue dashed line in Figure 8). There are several orders that have no signif- icant peak, which can be caused by the imperfection of our simple atmosphere modeling or the inaccuracy of the line list. Note that all orders that have no significant CCF peak, have a large bad region except the last order of the red CCD (see Figure 3 and/or Figure 8). These bad orders are excluded for the rest of the analysis including order 2 in red CCD due to heavily contaminated by strong telluric lines. 3. TIO SIGNAL DETECTION 3.1. Results from SYSREM with Common Iteration Number As described in §2.5, we first analyze the spectra with a common SYSREM iteration number for all orders. A grid of Doppler-shifted WASP-33b model spectrum was cross- correlated with weighted SYSREM residuals. The Doppler- shifted model spectrum covers planet radial velocity (RVp) between −169.69 km s−1 ≤ RVp ≤ +393.30 km s−1 with 0.5 km s−1 intervals, corresponding to half of the HDS sampling resolution. For each detector, the CCFs of every good orders (all orders excluding the bad orders explained in §2.7.1) were summed. This value is listed in the CCF ma- trix with a dimension of 1127 (RV as column) × 52 (orbital phase as row). Then, the CCF matrix of blue and red CCDs was summed to calculate the final CCF matrix. These steps were done for all SYSREM residuals. We then calculate the CCF map in the Kp – Vsys plane for all spectrum models by doing the following steps. The CCF of the frames (40 frames in total, excluding the frames when WASP-33b was expected in the secondary eclipse phase) was integrated along the expected RVp curve: RVp(t) = Kp sin(2πφ(t)) + Vsys + vbary(t) with φ(t) = t − T0 P (1) (2) where vbary(t) is the barycentric correction, φ(t) is the planet orbital phase, t is the mid observation time in BJD, T0 is the ephemeris, and P is the orbital period of the planet, Kp is the TIO WASP-33B DRAFT 9 Figure 4. WASP-33b spectrum model using full inversion (FI) and no inversion (NI) of the TP profile. The left panel shows the WASP-33b model spectrum for both FI-spec and NI-spec. The color shade represents the wavelength range of the observed order, blue, and orange color represent the wavelength range in the blue and red CCDs, respectively. The vertical axis shows the planet to star flux contrast in the level of 10−3. The middle panel shows the weight function along the wavelength for the pressure range considered. The right panel shows our adopted temperature profile of the WASP-33b atmosphere. Figure 5. WASP-33b spectrum model using TP profile of Haynes et al. (2015). The left panel shows the WASP-33b model spectrum of various VMR. The vertical axis shows the planet to star flux contrast in the level of 10−3. The right panel shows our adopted temperature profile of the WASP-33b atmosphere. 10 NUGROHO ET AL. Figure 6. Observed spectrum of five M-dwarf stars that were used for the TiO line list accuracy comparison. The bumps on the 4 first spectra are caused by the uncorrected blaze function. Before order-by-order cross-correlation, the spectrum was normalized by its continuum profile. The color shade represents the wavelength range of the observed order, the blue color is in the blue CCD and orange color is in the red CCD. (a) (b) Figure 7. (a) The colored line is an example of blue CCD spectrum of WASP-33 after standard reduction. Different colors mean different order. The black line is Coelho model spectrum. (b) The barycentric corrected RV of WASP-33 during the observation (red star), as a result of combining the RV of each order that has the highest cross-correlation signal. The blue line is the weighted median combine of all frames that was taken as the RV of WASP-33.The error bar is the 1-σ scatter across the observation time. semi-amplitude of the radial velocity of the planet, and Vsys is the systemic velocity 6. We consider the radial velocity of the planet semi-amplitude between +150 km s−1 ≤ Kp ≤ +310 km s−1, and systemic velocity between −80 km s−1 ≤ Vsys ≤ +80 km s−1 with 0.5 km s−1 steps. The result is a 321 (Vsys as column)× 320 (Kp as row) CCF matrix. Then this matrix was divided by its the standard deviation to make Kp – Vsys S/N map. except for NI-spec, we detected positive peaks with > 4-σ, while the map for NI-spec exhibits a negative peak at the same place. The maps for H-spec and FI-spec exhibits the strong peak in almost all SYSREM iteration number at Kp= +237.0 km s−1 and Vsys= −1.5 km s−1 (henceforth peak A). Among all H-spec, the strongest signal was found in log VMR= −8. This peak is also the strongest one in the Kp – Vsys c.c. maps of both FI-spec and NI-spec for most of the SYSREM iteration number, although for NI-spec the value is negative. The positive value of the peak A in H-spec and FI- spec, and the negative value in NI-spec suggested that non- inversion atmosphere is unlikely for WASP-33b if peak A is the real signal. The second strongest peak was detected at Kp= +192.0 km s−1 and Vsys= +19.5 km s−1 (henceforth peak B) in the SYSREM iteration number= 2 only. Figure 9 shows the Kp – Vsys S/N map for Nsys = 4 (H- spec) and 10 (H-, FI-, and NI-spec models). In all the cases 6 The nodal precession of WASP-33b caused the orbital inclination to evolve from ∼ 86.61◦ in 2008 to 88.70◦ in 2014 (Johnson et al. 2015), and based on Smith et al. (2011) analysis on its orbital eccentricity, we can safely assume a circular orbit with orbital inclination ∼ 90◦ to calculate RVp TIO WASP-33B DRAFT 11 Figure 8. Cross-correlation results between M-dwarf model and five M-dwarf spectra, Barnard's Star (black line), HD 95735 (blue line), Gl752A (green line), HD173739 (red line), and Proxima Centauri (brown line). The color of the label of the wavelength range also represents the order to which the CCD belongs. Note that the y-axis scale is not uniform, thus the orders cannot be compared. The black dash line is the rest-frame RV of each M-dwarf stars, while the vertical blue dashed line in the first three orders of the blue CCD is the position of CCF peaks on those orders. The blue and red shaded panel show the bad shape CCF that are masked for the rest of the analysis (bad). The grey shaded panel shows the masked order in the later analysis due to heavily contaminated by strong telluric lines (ht). Figure 10 shows the S/N of peaks A and B as a function of the SYSREM iteration, although peak B is unlikely to be a real signal, due to its physically non-realistic Kp and Vsys values compared with the expected value from previous studies, it was chosen as a representative of the noise/false- positive signal. The S/N of peak B decreases as increasing of the SYSREM iteration number. The S/N of peak A increases from SYSREM iteration= 1 until SYSREM iteration= 4 then decreases until SYSREM iteration= 6 before begin to in- crease again until SYSREM iteration= 10. The increasing of S/N of peak A after SYSREM iteration= 6 is most likely due to the difference level of telluric and/or stellar lines removal in each order. It can be seen in the Figure 9 a for SYSREM iteration= 4 where multiple peaks, 5 peaks, within 1σ value from the strongest peak can be found (white color). While in the Figure 9 b there are only 2 peaks, peak A and the other one at about Kp ∼ +160.0 km s−1 and Vsys ∼ −20 km s−1. From the curve of Figure 10, it is natural to adopt Nsys = 10 as a fiducial iteration number. Because the common iteration number for all orders does not fully optimize the procedure of the systematics removal, 4.3-σ of the significance level for H-spec (VMR= −8) we obtained should be regarded as a conservative estimate of detection level. 3.2. Order-based SYSREM Optimization The level of systematics (telluric lines) should be differ- ent for each order, thus to find the optimized SYSREM iter- ation number of each order, following steps were also per- formed. A scaled artificial signal was injected at the detected RV (Kp= +237 km s−1 and Vsys= −1.5 km s−1) in the spec- tra before telluric and stellar lines removal. The scaling was performed according to the equation (cid:18) Rp (cid:19)2 Fscaledpm(λ) = sc × Fpm(λ) Fstar(λ) Rstar , (3) where sc is the scaling constant, Fscaledpm is the scaled ar- tificial signal, Fpm(λ) is the planet model spectrum (H-spec with log VMR= −8) from the integration of Planck function versus monochromatic transmission along the line-of-sight (see §2.6), Fstar(λ) is the black body flux of T= 7400 K rep- resenting the continuum level of the WASP-33 flux, Rp and Rstar are the planet and star radius, respectively. 12 NUGROHO ET AL. (a) (c) (b) (d) Figure 9. (a) and (b) are the Kp – Vsys cross-correlation map of H-spec (log VMR= -8) for SYSREM iteration number= 4 and 10 respectively. (c) and (d) are the Kp – Vsys cross-correlation map of FI-spec and NI-spec for SYSREM iteration number= 10 respectively. The white dashed line is the expected Kp and Vsys from the previous studies. The black line is the maximum S/N peak in the map. The top and right panels show the 1-dimensional cross section of the CCF peak along the Vsys and Kp respectively. A and B are the peak A and peak B respectively. To check the dependence on the strength of the injected signal, we adopt 5 different injected signals with sc = [0.2, 0.4, 0.6, 0.8, 1.0]. The injected signals are broadened by a rotation kernel, using fastRotBroad from PyAstronomy with v. sin i= 0.4 times the expected projected velocity of WASP- 33b (calculated by assuming a tidally locked condition with the parameter in the Table 1 referring to the typical broaden- ing width of tidally locked exoplanet as shown in Kawahara 2012). We convolve with a box-function to take into account the change of Doppler shift during the 600 s exposures per frame. The signals were injected into the spectra, before per- forming telluric and/or stellar lines removal using SYSREM algorithm. The residuals for each SYSREM iteration were cross-correlated with the artificial signal itself. TIO WASP-33B DRAFT 13 to the rest-frame of the planet (middle panel). The TiO sig- nal can be seen in the left panels as a positive (dark) sig- nal with an arc shape for H- and FI- spec and as a negative (bright) one for NI-spec. In the middle panels of the fig- ures, the planet signal was aligned such that it can be seen as a vertical dark/bright trail at the Vsys= −1.5 km s−1. The right panels show the mean CCF with the width of 6 pixels centered at Vsys= −1.5 km s−1 (CCFexo). For FI- and H- models, the CCF curve exhibits a positive offset during the visible phase while there is no offset when the planet is be- hind the star. This feature supports the atmospheric origin of the CCF signal. The exposure time for a single frame is 600 s, which corresponds to ∼ 7 km s−1 at the near-eclipse phase. This long exposure time should have smeared out the day- side spectrum by the change in the radial component of the orbital velocity of the planet, especially in the phases near the secondary eclipse. This smearing effect may account for the fact that the signal is only visible at the phase (cid:46) 0.35, while it should be stronger at the phase (cid:38) 0.35 because larger part of the dayside of the WASP-33b is visible. Figure 14 shows a Kp – Vsys S/N map for H-spec with log VMRT iO= −8. The noise level in the Kp – Vsys S/N map after optimization is significantly suppressed. The strongest peak was found with 4.8-σ detection significance at Kp= +237.5 +13.0−5.0 km s−1 and Vsys= −1.5 +4.0−10.5 km s−1 within an elliptic region of 1 sigma. The latter is consistent with the estimates from the stellar RV (§2.7 and Collier Cameron et al. 2010). Because we already have the strong constraint on Vsys = −3.02 ± 0.42 km s−1 from the stellar spectrum (§2.7), we may assume this value as the prior of Vsys on the Kp – Vsys S/N map. Then we obtain a stronger constraint, Kp = 239.0+2.0−1.0 km s−1. This is the first time that the or- bital velocity of WASP-33b was dynamically measured. The orbital velocity of a planet depends on the orbital period and the mass of the host star. Since the measurement of orbital period is very precise for a transiting exoplanet, our obser- vation provides the first model-independent measurement of the mass of the host star. Using the period of WASP-33b from the previous study (see Table 1) the measured mass of the host star is M(cid:63) = 1.73+0.04−0.02 M(cid:74), which is larger than values previously reported. 3.3. Statistical Tests The in-trail CCF (henceforth in-trail signal) was compared with the out-of-trail CCF (henceforth out-of-trail signal) by performing Welch's t-test to check if the mean is same as- suming that two distributions were drawn from the same par- ent distribution, using the SciPy module in Python 2.7. We used equation (1) to calculate the expected RVp for the same range to ones for the Kp – Vsys CCF maps, and took the 1 pixel mean CCF value at the closest RV to RVp. The out-of- Figure 10. The S/N of peak A and B along the SYSREM iteration number, Nsys, for H-spec (log VMR= −8). The blue diamond-lined is the S/N of peak A and the red circle-dotted line is the S/N of peak B. The CCF of each order of each frame was aligned to the planet rest-frame according to the injected Kp – Vsys (e.g. the middle panel of Figure 13). Then, the aligned CCFs of all frames except the frames corresponding to the secondary eclipse were summed to create the total mean CCF. To exam- ine the exoplanet signal for various combinations of SYS- REM iteration numbers, the peak value of the total mean CCF in ± 2.5 km s−1 was taken between the expected rest- frame radial velocity of the (0 km s−1) planet and divided by the its outside standard deviation (non-detection CCF value) to include the "noise" (henceforth Smax/n). The results for five injected signals are shown in Figure 11. The signal strength (Smax/n) behaves differently in ev- ery order, depending on the amount of telluric and/or stel- lar lines contamination and the strength of the injected sig- nal (represented by sc value). There are several jumps in Smax/n curves, which may be caused by spurious signal, thus by following the general trend of the recovered Smax/n curves of 5 different injected signals the optimum SYSREM iteration number we chose the optimum SYSREM iteration numbers. For blue and red CCDs the optimum SYSREM it- eration numbers are [2, 2, 8, 2, 1, 1, 5, 3, 1, 3, 5, 2, 7, 2, 8, 2, 2, 2] and [2, 3, 5, 2, 3, 4, 2, 8, 3, 9, 3, 5], respectively, for each of its orders. Figure 12 shows the level of telluric and/or stel- lar lines contamination of each order, which is represented by the standard deviation value of the standard reduced spec- trum in that order, and its corresponding optimum SYSREM iteration number. One can see the tendency that the optimum SYSREM iteration number increases as exhibiting more tel- luric and/or stellar lines in the spectrum, except for one in the bottom right of the figure ( the order having many O2 lines). Using the optimum SYSREM iteration number, the mean CCF map was calculated for all of the spectrum model as shown in Figure 13 (left panel). Then the CCF was aligned 14 NUGROHO ET AL. Figure 11. Smax/n of all orders for various SYSREM iteration numbers. The vertical lines mark the SYSREM iteration number that made the signal strength optimum. The blue circle, yellow asterisk, green diamond, red square, and purple cross are the Smax/n of the recovered injected signal with sn= [0.2, 0.4, 0.6, 0.8, 1.0] respectively. The number in the white box at the bottom left of each panel represents the order number and the color represents the CCD (blue or red). The out-of-trail and in-trail signal histograms (width of the in-trail signal= ± 3 pixels) were plotted. The in-trail signal distribution shifted further from the out-of-trail signal distri- bution. The distribution of the in-trail signal is clearly shifted from the zero values. From the Q-Q plot (see Figure 17) it can be seen that the out-of-trail signal distribution is a Gaus- sian distribution until 4-σ, therefore we can safely convert the half p-value to σ value of the detection significance using an error function7. The Welch's t test shows that the in-trail distribution is deviated from the out-of-trail signal distribu- tion by 5.0-σ (see Figure 15) in line with the S/N of this peak detection. 4. DISCUSSION AND CONCLUSIONS We confirmed the previously claimed of the inaccuracy of the TiO line list for wavelengths shorter than 6300 A (see Hoeijmakers et al. 2015). We also showed that the line list accuracy is enough high for longer wavelengths, which was considered in processing order-based optimization, therefore our analysis no longer suffers from the inaccuracy issue. By 7 The output from the scipy module is the two-tailed p-value. Figure 12. Standard deviation of the spectrum of each order with their corresponding optimized SYSREM iteration number. The blue cross and the red circle indicate the spectrum in the blue and red CCDs, respectively. The value of the standard deviation represents the amount of telluric and/or stellar lines contamination for every order. trail signal contains all mean CCF values except the in-trail signal. TIO WASP-33B DRAFT 15 Figure 13. Order-based optimized version of cross-correlation map for H-spec with log VMR= −8 (first row), FI-spec (second row), NI-spec (third row), and the injected signal (fourth row). The left panel shows the mean CCF map. The middle panel shows the aligned mean CCF map at Kp= +237.5 km s−1 and Vsys= −1.5 km s−1. The planet signal can be seen as a dark trail in the ex- pected rest-frame of the radial velocity of the planet from the first observed phase until the appearance of the secondary eclipse phase (red dashed). The right panel shows the mean CCF for a 6 pixel column bin (CCFexo) centered on Vsys= −1.5 km s−1 along the or- bital phase. The solid black line shows the smoothed CCFexo with a 3 frame smoothing window. The blue dashed line shows the zero value of CCF. measuring the radial velocity of the host star we also con- firmed the measurement by Collier Cameron et al. (2010), which differs by ∼ 4 km s−1 from the SIMBAD database. This confirmation gives us a narrow Vsys search space to find the possible exoplanet signal. We reported a TiO molecule signature detection in the day-side spectra of WASP-33b by 4.8-σ confidence level, which provided direct evidence of the existence of TiO in the atmosphere of the hot Jupiter. The detection levels for VMR= -8, -9, and -10 (H-spec) lie within 1-σ from the high- est one (VMR= -8); moreover, in our analysis we did not use a self-consistent atmospheric model, thus the constraint on the VMR cannot be obtained directly from our result only. Our TiO molecular detection for H-spec and FI-spec confirmed the existence of stratosphere (thermal inversion layer) in the day-side of WASP-33 b, as previously has been claimed by several studies (e.g. von Essen et al. 2015; Haynes Figure 14. Figure Kp – Vsys S/N map of H-spec with maximum peak at Kp= +237.5 +13.0−5.0 km s−1 and Vsys= −1.5 +4.0−10.5 km s−1, which gives 4.8-σ detection. The white dashed line is the expected Kp and Vsys from the previous studies. The top and right panels show the 1-dimensional cross section of the CCF peak along the Vsys and Kp respectively. The black dash lines show the most sig- nificant signal, the white dash lines show the expected Kp and Vsys, and the color bar grid interval is 1-σ, the white area also represents the 1-σ error of the detected signal. Figure 15. Significance map after converting from the p-value from Welch's t-test using erf function for the most significant detected signal with the H-spec model spectrum of log VMRT iO= −8. The black dashes show the most significant signal, the white dashes show the expected Kp and Vsys, and the color bar grid interval is 1- σ, the white area also represents the 1-σ error of the detected signal. et al. 2015). The full inversion T/P profile has also been re- ported for another super hot Jupiter, WASP-121b by Evans et al. (2017), which resolved the emission spectral feature of H2O at near-infrared wavelength, using HST. Our result 16 NUGROHO ET AL. spectroscopy in WASP-19b by Sedaghati et al. (2017), who were able to constrain the relative abundance of TiO (0.12 p.p.b) but could not provide information about T/P profile as they were only able to measure the transmission level of the molecules in the atmosphere. The constraint on the rela- tive abundance level of TiO in WASP-33b can be obtained by analyzing it using a self-consistent atmospheric model and/or by combining low- and high-dispersion spectroscopy in order to introduce a more precise constraint of the relative abun- dance of each detected molecules and the T/P profile of the exoplanet atmosphere (Brogi et al. 2017). By observing for about nine hours only using the 8.2 m Subaru telescope, we are able to detect significant signature of TiO in the atmosphere of WASP-33b. Our results demon- strate that high-dispersion spectroscopy is a powerful tool to characterize the atmosphere of an exoplanet, and show a promising potential of developing similar/more advanced techniques for the Infra Red Doppler instrument (Kotani et al. 2014) in the Subaru telescope, and for extra large telescopes facilities in the future, as suggested by several authors (e.g. Kawahara & Hirano 2014; Snellen et al. 2015). 5. ACKNOWLEDGEMENT This work was based in part on data collected on HDS at the Subaru Telescope, which is operated by the National Astronomical Observatory of Japan. We thank our anony- mous referee, whose insightful comments improved the manuscript. S.K.N. acknowledges support from Indonesia Endowment Fund for Education Scholarship. S.K.N also acknowledges Toru Yamada for fruitful discussions during the analysis of the data. H.K. is supported by Grant-in- Aid for Young Scientist (B) from Japan Society for Pro- motion of Science (JSPS), No. 17K14246 and the Astro- biology Center from NINS, No. AB291003. T.H. is sup- ported by Grant-in-Aid for Young Scientist (B) from Japan Society for Promotion of Science (JSPS), No. 16K17660. Figure 16. Histogram of the in-trail and out-trail mean CCF dis- tribution, the in-trail distribution is slightly shifted from the out-of- trail distribution. Figure 17. Q-Q plot of the out-of-trail distribution which shows that the distribution is Gaussian until 4-σ. is complementary to the TiO detection using low-dispersion REFERENCES Barklem, P. S., & Collet, R. 2016, A&A, 588, A96 Birkby, J. L., de Kok, R. J., Brogi, M., et al. 2013, MNRAS, 436, L35 Birkby, J. L., de Kok, R. J., Brogi, M., Schwarz, H., & Snellen, I. A. G. 2017, AJ, 153, 138 Brogi, M., de Kok, R. J., Albrecht, S., et al. 2016, ApJ, 817, 106 Brogi, M., Line, M., Bean, J., D´esert, J.-M., & Schwarz, H. 2017, ApJL, 839, L2 Brogi, M., Snellen, I. A. G., de Kok, R. J., et al. 2013, ApJ, 767, 27 Burrows, A., Hubbard, W. B., Lunine, J. I., & Liebert, J. 2001, Reviews of Modern Physics, 73, 719 Burrows, A., & Sharp, C. M. 1999, ApJ, 512, 843 Coelho, P. R. T. 2014, MNRAS, 440, 1027 Collier Cameron, A., Pollacco, D., Street, R. A., et al. 2006, MNRAS, 373, 799 Collier Cameron, A., Guenther, E., Smalley, B., et al. 2010, MNRAS, 407, 507 Crossfield, I. J. M., Barman, T., & Hansen, B. M. S. 2011, ApJ, 736, 132 de Kok, R. J., Brogi, M., Snellen, I. A. G., et al. 2013, A&A, 554, A82 TIO WASP-33B DRAFT 17 Diamond-Lowe, H., Stevenson, K. B., Bean, J. L., Line, M. R., & Fortney, J. J. 2014, ApJ, 796, 66 Esteves, L. J., de Mooij, E. J. W., Jayawardhana, R., Watson, C., & de Kok, R. 2017, AJ, 153, 268 Evans, T. M., Sing, D. K., Wakeford, H. R., et al. 2016, ApJL, 822, L4 Evans, T. M., Sing, D. K., Kataria, T., et al. 2017, Nature, 548, 58 Fortney, J. J., Lodders, K., Marley, M. S., & Freedman, R. S. 2008, ApJ, 678, 1419 Gontcharov, G. A. 2006, Astronomical and Astrophysical Transactions, 25, 145 Hansen, C. J., Schwartz, J. C., & Cowan, N. B. 2014, MNRAS, 444, 3632 Haynes, K., Mandell, A. M., Madhusudhan, N., Deming, D., & Knutson, H. 2015, ApJ, 806, 146 Hoeijmakers, H. J., de Kok, R. J., Snellen, I. A. G., et al. 2015, A&A, 575, A20 Hubeny, I., Burrows, A., & Sudarsky, D. 2003, ApJ, 594, 1011 Johnson, M. C., Cochran, W. D., Collier Cameron, A., & Bayliss, D. 2015, ApJL, 810, L23 Kaeufl, H.-U., Ballester, P., Biereichel, P., et al. 2004, in Proc. SPIE, Vol. 5492, Ground-based Instrumentation for Astronomy, ed. A. F. M. Moorwood & M. Iye, 1218–1227 Kawahara, H. 2012, ApJL, 760, L13 Kawahara, H., & Hirano, T. 2014, ArXiv e-prints, arXiv:1409.5740 Knutson, H. A., Charbonneau, D., Allen, L. E., Burrows, A., & Megeath, S. T. 2008, ApJ, 673, 526 Knutson, H. A., Howard, A. W., & Isaacson, H. 2010, ApJ, 720, 1569 Kotani, T., Tamura, M., Suto, H., et al. 2014, in Proc. SPIE, Vol. 9147, Ground-based and Airborne Instrumentation for Astronomy V, 914714 Kov´acs, G. and Kov´acs, T. and Hartman, J. D. 2013, A&A, 553, A44 Mazeh, T., Tamuz, O., & Zucker, S. 2007, in Astronomical Society of the Pacific Conference Series, Vol. 366, Transiting Extrapolar Planets Workshop, ed. C. Afonso, D. Weldrake, & T. Henning, 119 Narita, N., Suto, Y., Winn, J. N., et al. 2005, PASJ, 57, 471 Noguchi, K., Aoki, W., Kawanomoto, S., et al. 2002, PASJ, 54, 855 Ofir, A., Alonso, R., Bonomo, A. S., et al. 2010, MNRAS, 404, L99 Plez, B. 1998, A&A, 337, 495 Schwarz, H., Brogi, M., de Kok, R., Birkby, J., & Snellen, I. 2015, A&A, 576, A111 Sedaghati, E., Boffin, H. M. J., MacDonald, R. J., et al. 2017, Nature, 549, 238 Sharp, C. M., & Burrows, A. 2007, ApJS, 168, 140 Smith, A. M. S., Anderson, D. R., Skillen, I., Collier Cameron, A., & Smalley, B. 2011, MNRAS, 416, 2096 Snellen, I., de Kok, R., Birkby, J. L., et al. 2015, A&A, 576, A59 Snellen, I. A. G., Albrecht, S., de Mooij, E. J. W., & Le Poole, R. S. 2008, A&A, 487, 357 Snellen, I. A. G., Brandl, B. R., de Kok, R. J., et al. 2014, Nature, 509, 63 Snellen, I. A. G., de Kok, R. J., de Mooij, E. J. W., & Albrecht, S. 2010, Nature, 465, 1049 Spiegel, D. S., Silverio, K., & Burrows, A. 2009, ApJ, 699, 1487 Tajitsu, A., Aoki, W., Kawanomoto, S., & Narita, N. 2010, Publications of the National Astronomical Observatory of Japan, 13, 1 Tajitsu, A., Aoki, W., & Yamamuro, T. 2012, PASJ, 64, 77 Tamuz, O., Mazeh, T., & Zucker, S. 2005, MNRAS, 356, 1466 von Essen, C., Mallonn, M., Albrecht, S., et al. 2015, A&A, 584, A75 Winn, J. N., Suto, Y., Turner, E. L., et al. 2004, PASJ, 56, 655 Zellem, R. T., Lewis, N. K., Knutson, H. A., et al. 2014, ApJ, 790, 53 18 NUGROHO ET AL. APPENDIX A. CORRECTION FOR BLAZE FUNCTION VARIATION AND NORMALIZATION OF SPECTRA As Winn et al. (2004) showed, observation using HDS suffered from variation of the blaze function during the observation. The continuum profile is not important in our analysis, but it is useful to correct this variation as a method of subtracting the continuum of each spectrum. We used the first frame as a reference spectrum and compared the rest of spectra by calculating the ratio of both spectra. By looking at the ratio of all spectra, we found wavelength variations with similar patterns for all orders, but these vary along the time of the observation (see Figure 18). The ratio spectra were then clipped to remove any outlier. In order to avoid any residual broad stellar absorption lines mismatch in the ratio spectra, we selected a 301-pixels smoothing window and applied smoothing, using PyAstronomy.pyasl.smooth8 with a flat window function. Each of the compared spectra was then divided by its smoothed ratio spectra, resulting in a spectra with shared blaze function with the reference spectrum, while conserving stellar lines and telluric line strength variations. A spline function was fitted to the reference spectrum using continuum manually and divided all spectra by it to get the normalized spectra. Figure 18. Top panel: Spectra from two different epoch for order 93 (left panel), 92 (middle panel), 91 (right panel), which were taken at the beginning (blue box) and the end (orange triangle) of the observation. The continuum profile of the second spectrum after correction (red circle) matched with the first spectrum. Bottom panel: ratio of the spectra before (blue circle) and after (black star) correction. The orange line is the smoothed ratio profile used to correct the variation. B. SYSREM ALGORITHM As in Tamuz et al. (2005), the aim is to find the two sets of effective extinction coefficients (ci) and air mass (aj) that optimally describe the atmospheric absorption in each wavelength bin (i) of each frame (j). By taking the observed air mass as the first input of a, we search for the optimal ci that minimizes (B1) (cid:88) R2 i = (rij − ciaj)2 j σ2 ij 8 http://www.hs.uni-hamburg.de/DE/Ins/Per/Czesla/PyA/PyA/index.html (cid:1) j j j ij j /σ2 ij (cid:0)rijaj/σ2 (cid:1) (cid:0)a2 (cid:80) (cid:80) (cid:88) (cid:0)rijci/σ2 (cid:80) (cid:1) (cid:0)c2 (cid:80) σ2 ij i /σ2 ij ij i i i (cid:1) ci = a(1) j = ij = rij − c(1) r(1) i a(1) j (cid:16) (cid:88) j R2 i = ij − c(2) i a(2) r(1) σ2 ij j (cid:17)2 (B2) (B3) (B4) i (B5) , a(2) j that minimize (B6) where σ is the uncertainty of pixels ij calculated by taking the root sum square of the standard deviations of its frame, and the wavelength bin. Then, ci can be estimated by TIO WASP-33B DRAFT 19 Then, by using the estimated ci, the "optimized air mass" (a(1) by calculating R2 j = ) can also be found that minimizes (rij − ciaj)2 By using the "optimized air mass", the "optimized coefficient" c(1) the stable value of c(1) can be found. The residual can be calculated as , a(1) i i j can be calculated and by performing this "criss-cross" iteration At this point, the first systematic effect has been removed; then, by performing a similar calculation to find c(2) the second and subsequent systematics can be calculated and removed. Note that ciaj does not actually represent the real extinction coefficient and air mass even for the first iteration, but a linear systematic effect that varies as a function of wavelength and time (or frame number). C. CROSS SECTION OF MOLECULES The TiO line list from Plez (1998) was used, which include five different isotopes with nine electronics systems, and the partition function that was published by Barklem & Collet (2016). As Py4CATS only supports HITRAN- and GEISA-like databases, instead of using its extract module, we used our custom build Python script to extract the TiO line list and wrote them in HITRAN-like format, which then were used to calculate line-by-line cross-sections. The cross-section (k) of each line is a product of the line strength (S) and a normalized line profile function (g) k(ν; ν, S, γ) = S(T ) · g(ν; ν, γ) with g dν = 1 (C7) (cid:90) ∞ −∞ where γ is the line broadening half width at half maximum (HFWHM), ν is the frequency, and ν is the line centroid position. We modified lbl2xs.py at the adjusting line parameter of the p and T section to enable it to calculate line strength using other partition function databases, and to include thermal (Doppler), natural, and van der Waals broadening for TiO in the line profile calculations. The line strength at temperature T (S(T )) is calculated by using Equation (1) in Sharp & Burrows (2007). A Voigt function was used for the line profile, which is defined as (cid:90) ∞ −∞ K(x, y) = y π e−t2 (x − t)2 + y2 dt (C8) where x, y are defined as dimensionless variables in terms of distance from the line centroid position, ν − ν, and the ratio of Lorentz and Gaussian HWHM γL, γD: √ ln 2 x = ν − ν γD and y = √ ln 2 γL γD (C9) 20 NUGROHO ET AL. The thermal broadening is expressed with a Gaussian profile (gD): gD(ν) = 1 γD (cid:18)ln 2 (cid:19)1/2 π (cid:34) −ln 2 exp (cid:19)2(cid:35) (cid:18) ν − ν (cid:114) γD with γD = ν 2 ln 2kT mc2 while natural and van der Waals broadening is expressed with a Lorentz profile (gL) gL(ν) = γL/π (ν − ν)2 + γ2 L (C10) (C11) (C13) where k is Boltzmann's constant, T is temperature, m is the mass of the molecular absorber, and c is the speed of light. γL is a sum of van der Waals (γLW ) and natural (γLN ) line broadening HWHM, γL = γLW + γLN (C12) As there was no information available for the van der Waals line broadening width of both molecules, we calculated it by following Sharp & Burrows (2007) as with (cid:18) T0 (cid:19)n T γLW = γ0 × p p0 (C14) where J(cid:48)(cid:48) is the lower rotational quantum number of each energy transition, w0 is the FWHM of a transition at 1 atm (p0) when J(cid:48)(cid:48) = 0, w1 is a scale factor of the dependency of the broadening on J(cid:48)(cid:48), and as suggested in the paper we used w0 = 0.1 cm−1, w1 = 0.002 cm−1, and n = 0. The minimum criterion in Equation C14 means that the line broadening at J(cid:48)(cid:48) = 30 is used for larger J(cid:48)(cid:48) values. We used the formalism by Gray (1976) for the natural broadening width: γ0 = [w0 − min(J(cid:48)(cid:48), 30)w1] 1 2 γLN = 0.222 ν2 4πc (C15)
1701.02994	1	1701	2017-01-11T14:53:10	Terrestrial Planets Formation under Migration: the Systems near 4:2:1 Mean Motion Resonance	[ "astro-ph.EP" ]	In this work, we extensively investigate the formation of near 4:2:1 mean motion resonances (MMRs) configuration by performing two sets of N-body simulations. We model the eccentricity damping, gas drag, type I and type II planetary migration of planetesimals, planetary embryos and giant planets in the first sets. For the simulations of giant planets with type II migration, the massive terrestrial planets, with a mass up to several Earth masses, are likely produced in the systems. We further show that by shepherding and/or scattering mechanisms through Jovian planet's type II migration, the terrestrial planets and giant planets in the systems can be evolved into a near 4:2:1 MMRs. Moreover, the models are applicable to the formation of Kepler-238 and 302 systems. In the second set, we study the 4:2:1 MMRs formation in the terrestrial planetary systems, where the planets undergo type I migration and eccentricity damping. By considering type I migration, $\sim$ 17.1\% of the simulations indicate that terrestrial planets are evolved into 4:2:1 MMRs. However, this probability should depend on the initial conditions of planets. Hence, we conclude that both type I and type II migration can play a crucial role in close-in terrestrial planet formation.	astro-ph.EP	astro-ph	Mon. Not. R. Astron. Soc. 000, 1 -- ?? (2015) Printed 17 September 2018 (MN LATEX style file v2.2) Terrestrial Planets Formation under Migration: the Systems near 4:2:1 Mean Motion Resonance Zhao Sun1,2, Jianghui Ji1⋆, Su Wang1, Sheng Jin1 1CAS Key Laboratory of Planetary Sciences, Purple Mountain Observatory, Chinese Academy of Sciences, Nanjing 210008, China 2University of Chinese Academy of Sciences, Beijing 100049, China Accepted. Received; in original form ABSTRACT In this work, we extensively investigate the formation of near 4:2:1 mean motion resonances (MMRs) configuration by performing two sets of N-body simulations. We model the eccen- tricity damping, gas drag, type I and type II planetary migration of planetesimals, planetary embryos and giant planets in the first sets. For the simulations of giant planets with type II migration, the massive terrestrial planets, with a mass up to several Earth masses, are likely produced in the systems. We further show that by shepherding and/or scattering mechanisms through Jovian planet's type II migration, the terrestrial planets and giant planets in the sys- tems can be evolved into a near 4:2:1 MMRs. Moreover, the models are applicable to the formation of Kepler-238 and 302 systems. In the second set, we study the 4:2:1 MMRs for- mation in the terrestrial planetary systems, where the planets undergo type I migration and eccentricity damping. By considering type I migration, ∼ 17.1% of the simulations indicate that terrestrial planets are evolved into 4:2:1 MMRs. However, this probability should depend on the initial conditions of planets. Hence, we conclude that both type I and type II migration can play a crucial role in close-in terrestrial planet formation. Key words: planetary systems -- methods: numerical -- planets and satellites: formation. 1 INTRODUCTION The number of exoplanets has substantially increased in re- cent years. At the time of writing, over 3500 exoplanets (see www.exoplanet.eu) are discovered, mostly through radial veloc- ity and transiting surveys. In particular, as of Nov. 10, 2015, Ke- pler mission released approximately 1030 confirmed planets , along with over 4696 transiting planetary candidates (Fressin et al. 2013). This may imply that the planets appear to be very common, orbiting other stars beyond our own solar system. The observations show that close-in (short-period) terrestrial planets are typically common in the exoplanetary systems. Roughly speaking, one third to half of solar-type (FGK) stars can host at least one planet with a mass less than 10 M⊕ and an or- bital period ranging from 50 to 100 days (Howard et al. 2010; Mayor et al. 2011). The frequency of short-period terrestrial plan- ets is at least as high around M stars as around FGK stars, or even higher (Howard et al. 2012; Bonfils et al. 2013; Fressin et al. 2013). These terrestrial planets are usually found in multi-planet systems on compact but non-resonant orbits (Udry et al. 2007; Lovis et al. 2011; Lissauer et al. 2011). However, from the statis- tical results of Kepler release data, a great many of planet pairs in the systems are observed to be in near-resonant configurations (Zhang et al. 2010; Lee et al. 2013; Mart´ı et al. 2013; Wang & Ji ⋆ E-mail: [email protected] c(cid:13) 2015 RAS 2014; Zhang et al. 2014), such as the 4:2:1 MMRs. For example, Kepler-238 (Rowe et al. 2014) and Kepler-302 (Rowe et al. 2014) systems harbor close-in giant planets in near mean motion reso- nance (MMR) configuration with other planets. Therefore, these exciting observations motivate us to explore and understand the formation and evolution of the planetary systems, especially to in- vestigate the short-period terrestrial planet formation and the con- figuration formation of the 4:2:1 MMRs. Several models have been proposed to explain the formation of close-in terrestrial planets (Raymond et al. 2008), and the for- mation scenarios include in situ accretion, orbital migration arising from planetary embryos (type I migration (Goldreich & Tremaine 1980)) and gas giant planets (type II migration (Lin & Papaloizou 1986)), dynamical instabilities in systems of multiple gas gi- ant planets, tidal circularization of eccentric terrestrial planets, and photo-evaporation of close-in giant planets. For the systems that are composed of short-period terrestrial planets, their pro- toplanetary disks are suggestive of very massive (Raymond et al. 2008; Hansen & Murray 2012, 2013; Chiang & Laughlin 2013; Raymond & Cossou 2014), thus the planets can accrete in situ from a large number of planetesimals and planetary embryos in the disk. The formation and final configuration of close-in terres- trial planets are closely related to the strength of type-I migra- tion, and the suppression of type I migration is required in case of in-situ super-Earths formation (Ogihara & Ida 2009; Ogihara et al. 2015). The observed systems of hot super-Earths mostly can con- 2 Z. Sun et al. tain 20 - 40 M⊕ in mass within a fraction of an AU of the host star (Batalha et al. 2013). Star-planet tidal interactions may play a role in circularizing planets in highly-eccentric orbits and therefore reduce their semi-major axis in the evolution (Ford & Rasio 2006; Fabrycky & Tremaine 2007; Beaug´e & Nesvorn´y 2012; Dong & Ji 2013). However, Raymond et al. (2008) suggested that tidal effect can produce hot super-Earths, but only for those relatively massive planets (& 5M⊕) with very small perihelion distances (. 0.025 AU), and even then the inward movement in orbital distance is only 0.1-0.15 AU at most, therefore tides are not strong enough to move many of the Kepler planets to the nowadays observed separations, and additional dissipative processes are at play (Lee et al. 2013). Alternatively, terrestrial planet may form inside a migrat- ing giant planet (Raymond et al. 2006). Gas-giants can shepherd planetesimals and embryos interior to their orbits as they mi- grate inward, which can further collide and merge into Earth- like planets (Zhou et al. 2005). Mandell et al. (2007) showed ma- terials that have been shepherded interior to the migrating gi- ant planet by moving MMRs can accrete into close-in terres- trial planets. In addition, Izidoro et al. (2014) indicated that fast- migrating super-Earths only weakly perturb the planetesimals disk and planetary embryos, whereas slowly migrating super-Earths shepherd rocky material interior to their orbits in resonances and push toward the star. Moreover, the orbital migration and planet- planet scattering play a vital role in producing short-period ter- restrial planets (Brunini & Cionco 2005; Terquem & Papaloizou 2007; Raymond et al. 2008; Cossou et al. 2014). According to core-accretion model, the planetary embryos in the terrestrial planet formation region, and the solid cores of giant planets, are both formed within ∼ 1 Myr from kilometer-sized planetesimals (Safronov 1969; Wetherill 1980). Subsequently, the massive solid cores can further accrete gas from the protoplanetary disk to form gas-giants (Kokubo & Ida 2002; Ida & Lin 2004) at Myr timescale, before the disk disperses (Haisch et al. 2001). In the late stage of planet formation, after the gas disk clears, the giant planets ceases migrating, the numerous planetesimals and planetary embryos in the disk become turbulent due to dynamical stirring by gas-giants over hundreds of Myrs or even longer. Consequently, frequent or- bital crossings and giant impacts are likely to occur, which may eventually yield short-period Earth-like planets (Chambers 2001; Raymond et al. 2004; Zhang & Ji 2009; Ji et al. 2011). type II migration and in the model Raymond et al. (2006) investigated terrestrial planets for- mation under they in- cluded a giant planet and gas drag. Based on Raymond et al's model, Mandell et al. (2007) further considered an additional non- migrating giant planet. In our earlier study, we have investigated the 3:2 and 2:1 MMRs configuration formation (Wang & Ji 2014) in the system observed by Kepler mission, and simply considered the planets with their nominal masses that are over a few Earth masses without any growth (Wang et al. 2012; Wang & Ji 2014). In the present work, we aim to explore terrestrial planet formation especially in the system with 4:2:1 configuration under migration by performing N-body simulations. In short, we have carried out two kinds of simulations, where in the first model we include the presence of giant planet and as a comparison, in the second model we consider the configuration formation in the system with only terrestrial planets. The paper is structured as follows. In Section 2, we de- scribe the adopted models of planetary formation, including the disk model and the planetary migration scenarios. In Section 3 we present numerical setup and major results of our investigation. Fi- nally, we summarize the outcomes and give a brief discussion in Section 4. 2 MODELS Following the empirical minimum-mass solar nebula (MMSN, Hayashi 1981) model, the surface density of solid disk at stellar distance a is described as Σd = 10fdγice(cid:16) a 1AU(cid:17)−3/2 gcm−2, (1) where fd and γice are the solid and the volatile enhancement factor, respectively. γice is 4.2 exterior to the snow line or 1 interior to snow line. The profile of gas density is given as Σg = 2.4 × 103fgfdep(cid:16) a 1AU(cid:17)−3/2 gcm−2, (2) where fg and fdep are the gas enhancement factor and gas depletion factor, respectively. fdep=exp(−t/τdep), where t is the time and the timescale τdep is about few million years (Haisch et al. 2001). Herein we adopt τdep = 106 yr. The inner edge of the gas disk locates at 0.1 AU. As they are considered small bodies in our simulations, the planetesimals will be affected by the aerodynamical drag of the gas around them (Adachi et al. 1976; Tanaka & Ida 1999). The force that a planetesimal with a mass m will suffer from aerodynamical drag is written as Faero = − 1 2 CDπS2ρgUU, (3) where U = Vk − Vg is the relative velocity between the planetesi- mal's Keplerian motion Vk and the gas motion Vg. CD is the drag coefficient. If an object has a large Reynold number, CD can be taken as 0.5. S and ρg are the planetesimal's radius and gas den- sity, respectively. For a planetary embryo embedding in the gaseous disk, their mutual interactions will lead to an eccentricity damping of the em- bryo with a timescale τdamp written as (Cresswell & Nelson 2006) τdamp1 =(cid:16) e e(cid:17) = Qe r(cid:19)4 Σga2(cid:19)(cid:18) h m (cid:19)(cid:18) M∗ 0.78 (cid:18) M∗ h(cid:17)3(cid:21) , ×(cid:20)1 + 4 (cid:16)e 1 r Ω−1 (4) where h, r, Ω, and e are disk scale height, distance from central star, Keplerian angular velocity, and the eccentricity of the embryo, respectively. Qe = 0.1 is a normalized factor in association with hydrodynamical simulation results. Moreover, the angular momen- tum exchange between the embryos and the gas disk will result in orbital migration of planets. When the planets are less massive, they will undergo type I migration, whereas they grow large enough, they will experience type II migration (Ida & Lin 2004). The timescale of type I migration can be assessed using linear model and the net loss on embryo (Goldreich & Tremaine 1979; Ward 1997; Tanaka et al. 2002) is expressed as c(cid:13) 2015 RAS, MNRAS 000, 1 -- ?? τmigI = a a = 1 (2.7 + 1.1β) (cid:18) M∗ a(cid:19)2" 1 + ( er ×(cid:18) h Σga2(cid:19) m (cid:19)(cid:18) M∗ 1.1h )4# Ω−1, 1 − ( er 1.3h )5 Terrestrial Planets Formation under Migration 3 where ri and Vi represent the position and velocity vectors of planet mi and all vectors are given in stellar-centric coordinates. (5) where e, r, h, and Ω are the same meaning as given in Equation 4. Using the gas density profile in Equation (2), we can achieve β = −d ln Σg/d ln a = 1.5. Recent work showed that type I migration could be directed either inward or outward depending on different planetary and gas disk properties (Cossou et al. 2013), but in general, outward type I migration can simply at work when the mass of the planets can be several M⊕ (Cossou et al. 2014; Lega et al. 2014), which are much larger than the embryos and planetesimals in our model, so we do not consider outward type I migration in this work. The planet will experience type II migration when a planet grows to a massive one (M > Mcrit) in the viscous disk (Lin & Papaloizou 1993). The timescale of type II migration is de- scribed as (Ida & Lin 2008) τmigII1 ≃ 5 × 105f −1 g ×(cid:18) C2α 10−4(cid:19)−1(cid:18) m = 0.7 × 105(cid:16) α MJ(cid:19)(cid:16) a 10−3(cid:17)−1(cid:16) a M⊙(cid:19)−1/2 1AU(cid:17)(cid:18) M∗ 1AU(cid:17)1/2(cid:18) M∗ M⊙(cid:19)−1/2 , τmigII2 = a a , (6) where α and C2 are efficiency factor of angular momentum trans- port and reduce factor. Herein C2α = 10−4. τmigII1 is fit to the case that the planet mass is comparable to the entire mass of the gas disk. While the planet mass is lower than the mass of the gas disk, τmigII2 is the right fit. We adopt an empirical formula for the eccentricity damping from the gas disk, (τdamp2)−1 = ( e/e) = −K a/a (Lee & Peale 2002). Herein, we choose K=10. The criti- cal mass of planet that can produce a gap is given as Mcrit ≃ 30(cid:16) α 10−3(cid:17)(cid:16) a 1AU(cid:17)1/2(cid:18) M∗ M⊙(cid:19) M⊕. (7) In this work, besides mutual gravitational interaction among the objects in the system, we also consider the orbital migration, eccentricity damping and aerodynamical drag of planetary embryo (planetesimal). The acceleration of the planet or planetary embryo (planetesimal) with mass mi is expressed as N d dt 2 ri Vi = − G(M∗ + mi) (cid:18) ri ri(cid:19) + Gmj(cid:20) (rj − ri) j(cid:21) rj − ri3 − +(cid:26) Fdamp1 + Faero + FmigI (for planetesimals/embryos) Fdamp2 + FmigII (for giant planets) Xj6=i rj r3 (8) where Fdamp1 2 = −2 , (Vi · ri)ri r2 i τdamp1 , 2 , FmigI = − Vi 2τmigI , FmigII = − Vi 2τmigII , c(cid:13) 2015 RAS, MNRAS 000, 1 -- ?? (9) 3 NUMERICAL SIMULATIONS AND RESULTS 3.1 Initial conditions In our simulations, all bodies are assumed to be initially in copla- nar and near-circular trajectory orbiting the central star, where the argument of pericenter, mean anomaly, and longitude of ascending node are randomly distributed between 0◦ to 360◦, respectively. In total, five runs were carried out for the investigation of planetary formation. In simulation S1-S5, the planetary system is composed of the host star, one or two giant planets, and 2000 equal-mass planetes- imals. As shown in Table 1 and Equation (8), each planetesimal performs aerodynamical drag, eccentricity damping, Type I migra- tion and the gravitational perturbations from Jupiter or Saturn and other planetesimals in the system. The giant planets are modeled to suffer from eccentricity damping, Type II migration, and the inter- action from the planets and small bodies. In several runs, the giant planets are assumed to be fixed about the original region without migration. In the following, we will briefly summarize each case. In these simulations, we mainly explore the terrestrial plan- etary formation under the circumstance of the existence of giant planets. Thus, in each simulation we start with a swarm of planetes- imals, together with one or two giant planets in the system, orbiting the central star. The initial positions for Jupiter and Saturn are set to be 5.0 AU and 9.54 AU, respectively. Originally, the inner region of the system consists of 2000 planetesimals that each owns a mass of 5×10−3 M⊕, thereby leading to an entire mass of the planetesimal disk of 10 M⊕. Herein for all cases, the planetesimals are initially distributed in the region [0.5, 3.78] AU. S1: In this scenario, the planetary system comprises the host star, the planetesimals, and one Jupiter-mass planet. The model as- sumes that the giant planet suffers from type II migration over the dynamical evolution. S2: The initials parameters for Jupiter and planetesimals are the same as those given in S1. As a comparison, in this simulation the model does not account for type II migration for Jupiter. S3, S4 and S5: in this three runs, the initial system consists of the host star, the planetesimals, two giant planets -- Jupiter and Saturn. In simulation S3, both Jupiter and Saturn do not migrate in- ward but remain stable orbits. In simulation S4 and S5, both Jupiter and Saturn are allowed to migrate inward in the dynamical evolu- tion. We integrate Equation (8) using a hybrid symmetric algorithm in MERCURY package (Chambers 1999). However, we have mod- ified codes to incorporate gas drag and orbital migration scenarios for our simulation. In these runs, mutual interactions of all bodies are fully taken into account. Two bodies are considered to be in collision stage whenever their distance is less than the sum of two physical radii (Chambers 1999). If two objects collide each other, they can merge and form a single larger body without any fragmen- tation. In our simulation, each run is integrated for 2 - 100 Myr with a time step of 2 days and a Bulirsch-Stoer tolerance of 10−12. As usual, when the simulation ends up, the variations of energy and angular momenta are 10−3 and 10−11, respectively. 4 Z. Sun et al. Table 1. The initial conditions of the runs. The meaning of Force is given by Equation 8. Name S1 S2 S3 S4 S5 Mass of planetesimal M⊕ 5 × 10−3 5 × 10−3 5 × 10−3 5 × 10−3 5 × 10−3 No. Planetesimals Jupiter Yes or Not Saturn Yes or Not 2000 2000 2000 2000 2000 Y Y Y Y Y N N Y Y Y Table 2. The statistics of the final destination of the planetesimals in each runs. Name S1 S2 S3 S4 S5 Time Myr 5 3.9 3.7 5 5 Percentage of mass (inside 0.1 AU) Percentage of Mass (beyond 5 AU) Saturn Yes or Not 79.5% 14.1% 17.9% 50.5% 36.9% 8% 0 0 14.8% 5.9% N N Y Y Y Force on giant planets FmigII No migration No migration FmigII FmigII Force on giant planets FmigII No migration No migration FmigII FmigII 3.2 Terrestrial planets formation with giant planets The simulations of S1-S5 exhibit classical terrestrial planetary ac- cretion scenarios in their late stage formation (Chambers 2001; Raymond et al. 2004, 2006; Fogg & Nelson 2005, 2009). 3.2.1 Terrestrial planets formation with one giant planet Figure 1 shows the orbital evolution of the planetesimals and the Jupiter-mass planet in the first 5 Myr for simulation S1. At an early time, the Jupiter-mass planet migrates inward in the disk, thereby giving rise to the planetesimals nearby the giant planet to either be scattered outward into high-eccentric orbits (Mandell & Sigurdsson 2003) or shepherded inward by the gi- ant planet's moving mean motion resonances (Tanaka & Ida 1999; Fogg & Nelson 2005). The buildup of inner material induces rapid growth of two close-in planets within a few Myr. At 5 Myr, five terrestrial planets form inside 0.1 AU. The total mass of these five planets is 7.95 M⊕, corresponding to ∼ 80% of the initial building materials set in the simulation. There are 16 planetesimals that were scattered to orbits beyond 5 AU, the total mass of them is ∼ 0.08 M⊕, the final destination of the planetesimals is shown in Table 2. The simulation S1 continues evolving for 100 Myr. By the end of 100 Myr, most of the planetesimal in the initial disk have been nearly cleared up by ejection or collision scenarios, arising from frequent orbital crossings over the chaotic evolution. We observe that there also exist a couple of moderate-eccentric planetesimals which are not involved in accretion process beyond ∼ 5 AU for the remaining disk. Moreover, we find that two terrestrial planets with a mass of 5.48 M⊕ and 1.62 M⊕ are eventual survivors in the system, locat- ing at 0.069 AU and 0.104 AU, respectively. These two terrestrial planets and the Jupiter form a 4:2:1 MMR orbital configuration at ∼ 10 Myr, as shown in Figure 2. The discovery of the exoplanets shows that the planetary systems in the universe are quite diverse. Our simulations present very interesting results, which may provide some clues to future observations for such systems. Our simulation S1 is similar with the work in Raymond et al. (2006) that investigated habitable terrestrial planet formation un- der a migrating giant planet. In their model, they adopted a higher mass of 17 M⊕ and a more extended planetesimal disk that ar- ranges from 0.25 to 10 AU. They found that hot Earths can form interior to a migrating giant due to the shepherding effect, and in some cases water-rich earth-mass planet can form outside the mi- grating giant, located inside the habitable zone. Simulation S1 also shows similar shepherd mechanism, as shown in Figure 1. Interest- ingly, the two inner terrestrial planets, as well as the migrating giant planet, are discovered to finally form a 4:2:1 MMR orbital config- uration. However, we do not observe material around the habitable zone. In simulation S1, the massive giant planet sweeps away or accrete most of the materials along its pathway when it migrates inward. Furthermore, we point out that such material depletion of migrating giant planet can be observed in our simulation S5 (Fig- ure 7), in which the system contains two migrating giants. Initials in simulation S2 are the same as those in S1, however in this model we simply do not let Jupiter undergo type II migra- tion. Figure 3 shows the orbital evolution of the planetesimals and the Jupiter-mass planet for simulation S2. In similar case, the plan- etesimals are quickly excited due to their mutual gravitation, along with that of the Jupiter-mass planet locating at 5.0 AU. The objects, involved in the 3:1 MMR at 2.50 AU, 2:1 MMR at 3.28 AU, and the 5:3 MMR at 3.70 AU, are stirred within 0.1 Myr. This trend can be clearly seen by the rise of the eccentricities of the planetesimals at these locations as shown in Figure 3. All stuff in the planetesimal disk exterior to the 3:2 resonance at 3.97 AU is quickly removed from the system via collision and ejection resulting from the giant planet. In the time evolution, the planetesimals' eccentricities can increase and the system becomes chaotic when the eccentricities are larger enough. The bodies in the inner disk begin to grow via accretionary collisions within 1 Myr. The larger bodies tend to have smaller eccentricities and inclinations, due to the dissipative effects of dynamical friction. At 3.9 Myr, there are four less massive ter- restrial planets (as compared to S1) with a mass of 0.15 − 0.51M⊕ formed in the region [0.07, 0.095] AU. The entire mass of four ter- restrial planets is 1.41 M⊕, corresponding to ∼ 14.1% of the total initial materials as shown in Table 2. In comparison, we can see that an inward-migrating giant planet can significantly increase the accretion rate in the inner part of planetesimal disk by the shepherd effect. Meanwhile, it can scat- ter the bodies in the inner disk to the outer disk. c(cid:13) 2015 RAS, MNRAS 000, 1 -- ?? Terrestrial Planets Formation under Migration 5 Table 3. Physical parameters of the 3 runs that form 4:2:1 MMR at 100 Myr and comparison with exoplanet systems. Name S1 Planet 1 S1 Planet 2 S1 Jupiter S4 Saturn S4 Planet 1 S4 Jupiter S5 Planet 1 S5 Jupiter S5 Saturn Mass M⊕ 5.48 1.62 333.55 95.69 5.05 333.46 3.69 333.02 95.12 Radius R⊕ 2.47 1.65 9.72 6.41 2.41 9.72 2.16 9.71 6.40 Period day 6.62 12.25 24.91 5.64 11.72 24.91 10.02 20.15 40.79 a AU 0.069 0.104 0.167 0.062 0.101 0.167 0.0911 0.1449 0.2315 e Name Kepler-238 c Kepler-238 d Kepler-238 e Kepler-302 b - Kepler-302 c 0.001 0.038 0.000 0.001 0.012 0.001 0.142 0.025 0.014 a Kepler planetary masses estimated with Eq. (1) in (Lissauer et al. 2011b). Mass a M⊕ 6.0 10 77 18 - 180 Radius R⊕ 2.39 3.07 8.26 4.06 - 12.45 Period day 6.16 13.23 23.65 30.18 - 127.28 a AU 0.069 0.115 0.169 0.193 - 0.503 e - - - - - - 3.2.2 Terrestrial planets formation with two giant planets In the following, we will present the simulation outcomes of the terrestrial planetary formation co-existence with two giant planets in the planetary systems. In simulation S3, two giant planets, which bear a Jupiter or Saturn mass, respectively, do not perform type II migration in the simulation. Therefore, similar to simulation S2, the planetesimals in the disk can be swiftly excited due to gravitational perturbations from two giant planets. In particular, the bodies, which are involved in the MMR with the Jupiter-mass planet, then acquire moderate eccentricities within 0.1 Myr. As compared to Jupiter, the Saturn- mass planet plays a less dominant role in the mass accretion of plan- etesimals in forming terrestrial planets. In the meanwhile, owing to the co-existence of two giant planets, the terrestrial formation pro- cess can be speeded up although Jupiter and Saturn do not deviate much from their initial orbits in this run. Figure 4 shows tempo- ral orbital evolution of all the planetesimals and the giant planets in simulation S3. At ∼ 2 Myr, three terrestrial planets are yielded with a mass in the range 0.10 − 0.57M⊕, and they orbit in the broad re- gion from 0.07 AU to 1 AU. At the time of 3.7 Myr, the entire mass of four terrestrial planets formed inside 0.1 AU is 1.79 M⊕, occu- pying 17.9% of the initial mass of the planetesimal disk. Similar with the simulation S2, there is no planetesimal beyond 5 AU. In order to investigate the role of type II migration, we have performed two additional runs for the systems composed of two giant planets, Simulation S4 and S5. In this two runs, both Jupiter and Saturn undergo type II migration. Our results provide evidence that terrestrial formation may take place in the inner region of the planetary systems (Fogg & Nelson (2005); Mandell et al. (2007); Fogg & Nelson (2009)). However, new runs further show some in- teresting outcomes - the 4:2:1 MMR configurations are formed for these systems. Figure 5 shows temporal orbital evolution of all the planetes- imals and giant planets. According to Equation 6, τmigII1 is pro- portional to the planetary mass, thus we learn that Saturn migrates faster than Jupiter does, indicating that there would be a possibil- ity for two giant planets to rendezvous in the system. At 0.02 Myr, Jupiter's eccentricity is gradually pumped up to 0.30 due to gravi- tational interaction from Saturn when it moves inward closer to the Jovian planet. At 0.036 Myr, Jupiter and Saturn reach the orbit at 5.75 AU and 3.58 AU, respectively, indicating that they pass through a 1:2 mean motion resonance. This resonance crossings may have ex- cited the orbital eccentricities of the planets which cross the res- onance (e.g., Tsiganis et al. 2005; Morbidelli et al. 2007). Thus, c(cid:13) 2015 RAS, MNRAS 000, 1 -- ?? the orbital crossing between two giant planets happens within sev- eral thousand years and brings about fairly chaotic behaviors for them. Hence, a sudden jump in the Jupiter's eccentricity occurs up to ∼ 0.72 at 0.037 Myr, whereas Saturn may obtain a moder- ate eccentricity right after the close encounter but its eccentricity is quickly damped by the gas disk. From the simulations, we find that Jupiter experienced close encounters with Saturn at the max- imum star-centric distance at ∼ 2.41 AU, which leads to strong interaction between them. From ∼ 0.037 Myr to 0.04 Myr, note that the semi-major axis of Jupiter drops down from 3.58 AU to 1.63 AU, whereas that of Saturn dramatically decreases from 5.75 AU to 0.45 AU. Furthermore, such chaotic motions for Jupiter and Saturn trigger catastrophic fate for the planetesimals in the system, where most of their orbits are severely excited and scattered. As Figure 6 shown, the eccentricity of the planetesimal (Planet 1) is first kicked to be 0.88 then falls to 0.13, while its semi-major axis can suddenly increase amount up to 5.18 AU, then soon drop down to 0.26 AU over the time span. As the two giant planets continue migrating inward, the 2:1 MMR between them is broken up. Subsequently, Saturn is kicked into the region ∼ 0.1 AU due to dynamical instabilities. By 10 Myr, the giant planets have moved very close to their central star. In the late stage, the eccentricity of Jupiter is gradually damped by dy- namical dissipation. However, an inner terrestrial planet (Planet 1) is excited onto a highly eccentric orbit exterior to the regime of two giant planets, and soon crosses Jupiter's orbit, then finally captured inside the orbits of Jupiter and Saturn. The mass growth for Planet 1 starts from one planetesimal at ∼ 0.06 Myr to a super Earth with a mass of 5.05 M⊕ at ∼ 0.40 Myr. Finally, Planet 1 remains at 0.101 AU, whereas Saturn and Jupiter cease migrating at 0.062 AU and 0.167 AU, respectively. This planetary system shows a very interesting configuration that a super Earth locates between two gi- ant planets' orbits, and three planets are trapped into a near 4:2:1 MMR (see Figure 8). Accordingly, the residual materials of disk are either removed out of the system or scattered into distant orbits. Their eccentricities rise along with their semi-major axes. The scenario of S4 shows a resemblance to NICE Model (e.g., Tsiganis et al. 2005; Morbidelli et al. 2007), which proposes that the Late Heavy Bombardment (LHB) -- a spike in the impact rate on multiple solar system bodies that lasted from roughly 400 un- til 700 Myr after the start of planet formation (Tera et al. 1974; Cohen et al. 2000; Chapman et al. 2007) -- was triggered by an in- stability in the giant planets' orbits. In the Nice Model, the orbits of the giant planets would have been in a more compact configu- ration, with Jupiter and Saturn interior to the 2:1 resonance. In our model of simulation S4, the Jupiter-mass and Saturn-mass planets 6 Z. Sun et al. can switch each place, similar to the case of Uranus and Neptune in the NICE Model. Raymond et al. (2008b) showed that planet- planet scattering could create MMRs, most of these resonances are indefinitely stable. The previous investigations reported that in a convergent migration scenario for two giant planets in the so- lar system, the 2:1 MMR could be formed and then disintegrated in the evolution (Morbidelli & Crida 2007b; Zhang & Zhou 2010). Zhang & Zhou (2010b) stated that when Jupiter lies outside Saturn, convergent migration could occur, Saturn is then forced to migrate inward by Jupiter where the two planets are trapped into MMR, and Saturn may move on its tracks of approaching the central star. Furthermore, to apply our model to explain the planetary forma- tion of the systems, we extensively examine the published data by Kepler mission and find a relevant planetary system (Kepler-302) similar to this simulation. Kepler-302 is a system consisting of two planets orbiting a star with a mass of 0.97 M⊙, as shown in Ta- ble 3. The outer planet of Kepler-302c is more massive than the inner companion Kepler-302b, and the two planets have approx- imate sizes as compared with those of the giant planets in simu- lation S4, whereas their orbital distances from the host star are a bit farther than those of the planets in the present run. Interest- ingly, the two giant planets of Kepler-302 are nearly close to a 4:1 MMR (Rowe et al. 2014), suggesting that the system may have gone through similar migration process as in our simulation. There- fore, our model herein presents a likely formation scenario for two planets that are close to a 4:1 commensurability, and also provides evidence that terrestrial planet formation under the influence of two giant planets in the compact system. In simulation S5, although Saturn migrates faster than Jupiter, there is no rendezvous between Saturn and Jupiter. Instead, when Saturn enters the 2:1 MMR orbit of Jupiter at 0.02 Myr, these two giant planets were locked in the 2:1 MMR configuration and they migrate inward together. Figure 7 shows temporal orbital evolution of the simulation S5 for the first 5 Myr. Jupiter's eccentricity is gradually pumped up to 0.30 due to gravitational interaction from Finally, a terrestrial planet of 3.69 M⊕ form inside 0.1 AU. This new terrestrial planet and the two giant planets form a 4:2:1 MMR orbital configuration at ∼ 5 Myr. Although Jupiter keeps migrating until 10 Myr in S1, Jupiter ceases its migration at 1.5 Myr and 3.5 Myr in S4 and S5, respectively. This is probably due to the torque from Planet 1, and the torque depends on the edge. Unlike in simulation S3, in simulation S4 and S5 we can see planetesimals beyond 5 AU due to scattering of inward-migrating giant planets. At 5 Myr, there are about 1.48 M⊕ materials beyond 5 AU in simulation S4, corresponding to 14.8% of the initial total mass. While in simulation S5, there are ∼ 0.59 M⊕ beyond 5 AU. Figure 8 shows the final configuration in the inner disk for the three runs that have turned on type II migration. Interestingly, all these runs yield 4:2:1 MMR. In simulation S1, two terrestrial plan- ets formed interior to the Jupiter mass planet. The migrating giant increased the accretion rate in the inner disk and cleared all mate- rial in its pathway. As a result, there is nothing left between 0.2 to 4.5 AU. In simulation S4, there was an orbital crossing between the migrating Jupiter mass and Saturn mass planets, and such a chaotic event changed the orbital configuration of the two giant planets. Hence, we note that a close-in terrestrial planet formed between two giant planets. There are some residual materials between 0.2 to 5 AU in this run, and this could be a consequence of the quickly orbital change of the two giants at 0.1 Myr, as shown in Figure 6. While in simulation S5, there is no chaotic orbital change of the Jupiter mass and Saturn mass planets. Similar to simulation S1, we see that only one terrestrial planet formed interior to two migrating giants, and in this case the material of the region between 0.2 and 4.5 AU of the disk was fully cleared by two giant planets. 4 THE EMERGENCE OF 4:2:1 MMRS ONLY WITH TERRESTRIAL PLANETS The investigations imply that the planetary configurations in asso- ciation with a near 4:2:1 MMRs would be common in the universe. Herein, we consider two planets with a period ratio in the range of [1.83, 2.18] as a pair near 2:1 MMR (Lissauer et al. 2011b). For example, Jupiter's Galilean moons are known to be locked into a so-called three-body Laplacian resonance (i.e., 4:2:1 MMR) in our solar system. On the other hand, other evidences are found in the exoplanetary systems, in which three planets of the GJ 876 system (Marcy et al. 2001; Rivera et al. 2010; Batygin et al. 2015; Nelson et al. 2016) and Kepler-79 (KOI-152) (Steffen et al. 2010; Wang et al. 2012) are reported to be close to 4:2:1 MMR. In this work, we have provided substantial evidences on this from the sim- ulations. For example, in simulation S1, S4 and S5 of Section 3, we show that the systems initially composed of one or two giant planets can finally produce 4:2:1 MMRs, in which the formed ter- restrial planets are involved in. Naturally, a question arises - what if the systems only contain terrestrial planets? As mentioned previously, the released Kepler outcomes re- port that a great number of planetary systems harboring terrestrial planets involved in a near 2:1 resonance. Furthermore, a portion of the terrestrial planets discovered by Kepler are in near 4:2:1 res- onance. Herein, we further investigate the configuration formation of those systems only with terrestrial planets to understand whether the 4:2:1 MMRs can be produced in such systems. With these pur- pose, we perform additional runs to explore the systems that simply consist of terrestrial planets through simulations. In our additional simulations, there are three terrestrial plan- ets initially with masses less than 30 M⊕ around central star, and they are assumed to migrate from the outer region of the system under the effect of the gas disk. The gas density profile is propor- tional to r−1, and an empty hole located in the inner region of the system. During the formation process, the terrestrial planets suffer from type I migration and gas damping. The timescales of type I migration and gas damping are shown in Equation (4) and (5), re- spectively. In total, we carry out seven groups of simulations, and each group contains five runs by considering different speed of type I migration. We add an enhance factor η to the timescale of type I migration. The five runs correspond to η times of typical timescales of type I migration as shown in Equation 5. The values of η are 100, 33.3, 10, 3.3, and 1, respectively. Figure 9 shows one of the runs with typical outcomes.Considering various speed of type I mi- gration, three planets are located at 140, 500, and 1450 d when η is larger than 33.3, otherwise they are located at 100, 250, and 600 d, respectively. The masses of the planets in five groups are 5, 10, and 15 M⊕, respectively, while in other two groups they are set to be 5, 5, and 5 M⊕, respectively. In Figure 9, the formation of the innermost, the middle and the outermost planets are marked by the black, red and blue lines, respectively. Panel (a), (b) and (c) show the evolution of the orbital periods, eccentricities, and the period ratios, respectively. Panel (d) and (e) display the 2:1 mean motion resonance angles. In the case with terrestrial planets in the system, they all undergo type I migration. Due to the short timescales for the first two planets, they are trapped into 2:1 MMR first. From Panel (c) in Figure 9, we find that the first and second planets are involved into 2:1 MMR at ∼ 0.5 Myr, and then three terrestrial planets are c(cid:13) 2015 RAS, MNRAS 000, 1 -- ?? trapped into 4:2:1 MMRs at ∼ 1.3 Myr. From Panel (b), we can learn that when the planet pairs enter into MMRs, their eccentrici- ties are excited, but due to the damping of gas disk, they cannot be stirred up to high values. The 4:2:1 MMR is usually produced when the first planet is trapped in the edge of the holes in the gas disk. The formation process is similar to that mentioned in Pierens & Nelson (2008), the MMR formed when the planet is captured in the gap of the other planet. In addition, we investigate the resonance angles of each pair which are shown in Panel (d) and (e) of Figure 9 and find that θ21 = 2λ1 − λ2 − 1 librates at ∼ 0◦, θ32 = 2λ2 − λ3 − 2 librates at ∼ 0◦, and θ33 = 2λ2 − λ3 − 3 librates at ∼ 180◦, where the subscript 1 and 2 represent the orbital elements of Planet 1 and 2, respectively. At the end of the simulation, three planets locate at about 11.52, 23.26, 47.17 days, respectively. This con- figuration is similar to the system Kepler-92 which contains three planets in the orbital periods of 13.75, 26.72, and 49.36 days, re- spectively. In final, by examining all runs from seven groups, we find that 6 out of 35 runs (17.1%) show three planets are finally involved in 4:2:1 MMRs, whereas 10 runs report that only the two inner planets enter into 2:1 MMR, and two runs for two inner plan- ets in 3:2 MMR. About half of the simulations (48.6%) are not in MMR. Based on our simulations, the proportion that planets pairs trapped near 4:2:1 MMRs appears to be high. There are several rea- sons that could lead to such results. Firstly, we set the initial orbital separation of planets at ∼ 20 Hill radius, which is a little larger than the empirical value of 12 Hill radius (Ida & Lin 2004). In such cases, most of the planets are first trapped by the 2:1 MMRs dur- ing their evolution. However, as a comparison, if they are initially in a compact configuration, the probability of the forming planets in 2:1 MMRs will significantly decrease. As shown in the work of Wang & Ji (2014), they found that if the planets pairs were initially in compact configuration at the separations of ∼ 15 Hill radius, the probability of 2:1 MMRs decreased to 10% in 50 runs and there was no simulation that three planets were evolved into 4:2:1 MMRs. Secondly, in our simulations we slow down the speed of type I mi- gration for low-mass planets, and this could further explain why two planets are fairly easier to be trapped into 2:1 MMRs. Finally, the final orbital periods of the planets in all simulations are less than 50 days which may be affected by the tidal force that arises from the central star. The tidal effect may lead to the deviation from 2:1 MMRs (Lee et al. 2013). In summary, the above-mentioned factors may play a major role in causing the planets pairs evolved in near 4:2:1 MMRs at a relatively moderate possibility. The simulations in Raymond et al. (2006) utilized the typical speed of type I migration which was very fast, and the embryos were initially located close to each other in compact configuration. These may be the leading differences producing diverse results between our simulations and those of Raymond et al. (2006). By comparing S1, S4 and S5 simulation containing giant plan- ets and the runs that simply consist of terrestrial planets in this sec- tion, we find that the timescales of the planet pairs trapped into MMRs are various for different cases. For the systems that only contain terrestrial planets, the two inner planets are found to be trapped into 2:1 MMRs within 1 Myrs, and the two outer pairs are in resonance at approximately 2-3 Myrs. Therefore, we may conclude that three terrestrial planets are in 4:2:1 MMRs within 3 Myrs on average. In comparison, for the systems consisting of one or two giant planets, the 4:2:1 MMRs emerges at over 2 Myrs when the terrestrial planets complete its mass accretion. Additionally, the eccentricities of planets which is in MMR with giant planets are always excited to be high values, while the eccentricities of planets in systems without giant planets will be suppressed to low values. c(cid:13) 2015 RAS, MNRAS 000, 1 -- ?? Terrestrial Planets Formation under Migration 7 5 CONCLUSIONS AND DISCUSSIONS In this work, we have investigated terrestrial planetary formation , especially the configuration formation of a near 4:2:1 MMRs by carrying out two sets of simulations with and without giant planets. From the results we find that 4:2:1 MMRs configuration can be formed in the systems both with and without giant planets. Herein we summarize the major results as follows. (i) In the first sets, we place one or two giant planets into the initial systems as well as the planetesimals, to understand terrestrial formation in such cases in S1-S5. In simulations S2 and S3, the giant planets do not undergo mi- grating in the disk, and the terrestrial formation procedure can be expedited because of the existence of Jupiter-mass and Saturn-mass planets. In final, several low-mass close-in terrestrial planets can be formed as a result of radial mixing and type I migration. In the two runs, there is no 4:2:1 MMRs in the evolution. In simulation S1, S4, and S5, we place one or two giant planets in the system , which undergo type II migration during the evo- lution. Thus the planetesimals are shepherded inward, cleared or ejected along with migration of giant planets. The close-in terres- trial planets are yielded with a mass up to several Earth-masses. For simulation S1, we find that two short-period terrestrial planets and one giant planet are finally involved in 4:2:1 MMR, whereas in the runs of S4, and S5, we show that one terrestrial planet and two giant planets enter into the 4:2:1 MMR. The resultant results of S1 and S4 are similar to Kepler-238 and 302 systems in orbital spac- ing. We examine the mass concentration statistic Sc and the orbital spacing statistic Ss (Chambers 2001; Hansen & Murray 2012) for these systems. Comparing the system S1 and Kepler 238, we can note that the orbital spacing statistics are close to each other. Their values for two systems are 1.8 and 2.76, respectively, indicating that S1 and Kepler 238 are similar in the orbital distribution. However, the mass concentration statistics of them differ from each other, and their values are 125.7 and 30.3, respectively. One reason for this difference in mass concentration statistic is that the masses of planets in our simulations are quite uncertain. Herein we adopted an estimated masses from Eq. (1) in Lissauer et al. (2011b). If we can obtain the true masses of planets in S1, the difference of Sc could be narrowed down. Moreover, we further compare Ss and Sc of the system S4 with those of Kepler 302, where Ss for each sys- tem is 5.43 and 4.0, respectively, and Sc for two systems is 10.4 and 23.3, respectively. This may imply that the system S4 and Kepler 302 have similar orbital distribution. Therefore, we can conclude that S1 and Kepler 238, S4 and Kepler 302 have similar orbital configurations. The present observations show that there is a lack of companion planets in the observed hot Jupiter systems. However, from our sim- ulation results, we can learn that terrestrial planets can survive in- side the inner region of hot Jupiter systems based on our formation scenario. As shown in the system S1, the final formed configuration is similar to that in Raymond et al. (2006). Moreover, if the gas disk is small enough, the hot Jupiter could have terrestrial companions formed in the system (Ogihara et al. 2013). Nevertheless, currently such terrestrial companions in the hot Jupiter systems have not dis- covered yet, this is because they may be under the detection limit. In the forthcoming, we hope that the hot Jupiter systems with low- mass companions could be discovered with the assistance of the improvement of observational precision and techniques. (ii) Furthermore, in the second sets, we also explore the 4:2:1 MMRs formation in the case of three terrestrial planets. Under type I migration, ∼ 17.1% of all runs, indicate that terrestrial planets 8 Z. Sun et al. are evolved into 4:2:1 MMRs. However, as aforementioned, the probability that planets are trapped into MMRs depends on their initial separation and the speed of type I migration. Comparing the processes with and without giant planets, the timescales that three planet are in 4:2:1 MMRs in systems without giant planets are shorter than systems with giant planets. Additionally, eccen- tricities are exited to higher values in systems with giant planets than without giant planets. However, in the treatment of accretion scenario in MER- CURY package (Chambers 1999), the collisions between two bodies are assumed to be perfect gravitational aggregations. Leinhardt & Stewart (2012) pointed out that real collisions could suffer from partially accreting collision, hit-and-run impacts, or graze-and-merge events, rather than a complete merger. Recently, Chambers (2013) showed that fragmentation does have a notable effect on accretion under consideration of fragmentation and hit- and-run collisions (Leinhardt & Stewart 2012; Genda et al. 2012). The final stages of accretion are lengthened by the sweep up of collision fragments. The planets that formed with fragmentation in the simulations have smaller masses and lower eccentricities when compared to those simulations without fragmentation. In this sense, the scenarios that include imperfect merger collision will not alter the conclusions of this study. In forthcoming work, we will take into account hit-and-run and fragmentation process in our model for further investigation. ACKNOWLEDGMENTS We thank the anonymous referee for insightful comments and good suggestions that helped to improve the contents. This work is fi- nancially supported by National Natural Science Foundation of China (Grants No. 11273068,11473073,11503092,11573073), the Strategic Priority Research Program-The Emergence of Cosmolog- ical Structures of the Chinese Academy of Sciences (Grant No. XDB09000000), the innovative and interdisciplinary program by CAS (Grant No. KJZD-EW-Z001), the Strategic Priority Research Program on Space Science, CAS (Grant No. XDA04060901), the Natural Science Foundation of Jiangsu Province (Grant No. BK20141509), and the Foundation of Minor Planets of Purple Mountain Observatory. REFERENCES Adachi, I., Hayashi, C., & Nakazawa, K., 1976, Prog. Theo. Phys. 56, 1756 Batalha, N. M., et al., 2013, ApJS, 204, 24 Batygin, K., Deck, K., & Holman, M. J., 2015, AJ, 149, 167 Beaug´e, C., & Nesvorn´y, D., 2012, ApJ, 751, 119 Bonfils, X., et al., 2013, A&A, 549, A109 Brunini, A., & Cionco, R. G., 2005, Icarus, 177, 264 Cresswell, P., & Nelson, R. P., 2006, A&A, 450, 833 Chambers, J. E., 1999, MNRAS, 304, 793 Chambers, J. E., 2001, Icarus 152, 205 Chambers, J. E., 2013, Icarus, 224, 43 Chapman, C. R., et al., 2007, Icarus, 189, 233 Chiang, E., & Laughlin, G., 2013, MNRAS, 431, 3444 Cohen, B. A., et al., 2000, Science, 290, 1754 Cossou, C., et al., 2013, A&A, 553, L2 Cossou, C., et al., 2014, A&A, 569, A56 Dong, Y., & Ji, J. H., 2013, MNRAS, 430, 951 Fabrycky, D. & Tremaine, S., 2007, ApJ, 669, 1298 Fogg, M. J., & Nelson, R. P. 2005, A&A, 441, 791 Fogg, M. J., & Nelson, R. P. 2009, A&A, 498, 575 Ford, E. B., & Rasio, F. A., 2006, ApJL, 638, L45 Fressin, F., et al., 2013, ApJ, 766, 81 Genda, H., Kokubo, E., & Ida, S., 2012, ApJ, 744, 137 Goldreich, P., & Tremaine, S., 1979, ApJ, 233, 857 Goldreich, P., & Tremaine, S., 1980, ApJ, 241, 425 Haisch, K. E., Jr., Lada, E. A., & Lada, C. J., 2001, ApJL, 553, L153 Hansen, B. M. S., & Murray, N., 2012, ApJ, 751, 158 Hansen, B. M. S., & Murray, N., 2013, ApJ, 775, 53 Hayashi, C., 1981, Prog. THeor. Phys. Suppl., 70, 35 Howard, A. W., et al., 2010, Science, 330, 653 Howard, A. W., et al., 2012, ApJS, 201, 15 Ida, S., & Lin, D. N. C., 2004, ApJ, 604, 388 Ida, S., & Lin, D. N. C., 2008, ApJ, 673, 487 Ida, S., & Makino, J., 1992, Icarus, 96, 107 Izidoro, A., Morbidelli, A., & Ramond, S. N., 2014, ApJ, 794, 11 Ji, J. H., Jin, S., & Tinney, C. G., 2011, ApJL, 727, L5 Jin, S., & Ji, J. H., 2011, MNRAS, 418, 1335 Kokubo, E., & Ida, S., 1998, Icarus, 131, 171 Kokubo, E., & Ida, S., 2002, ApJ, 581, 666 Lee M. H., Peale S. J., 2002, ApJ, 567, 596 Lee, M. H., Fabrycky, D., & Lin, D. N. C., 2013, ApJL, 774, L52 Lega, E., et. al., 2014, MNRAS, 440, 683 Leinhardt, Z.M., & Stewart, S.T., 2012, ApJ, 745,79 Lin, D. N. C., & Papaloizou, J. C. B., 1986, ApJ, 309, 846 Lin, D. N. C., & Papaloizou, J. C. B., 1993, in: E.H. Levy & J.I. Lunine (eds.), Protostars and Planets III, (Tucson: Unv. Arizona) Lissauer, J. J., et al., 2011, Nature, 470, 53 Lissauer, J. J., et al., 2011, ApJS, 197, 8 Lovis, C., et al., 2011, A&A, 528, A112 Mandell, A. M., & Sigurdsson, S., 2003, ApJ, 591, L111 Mandell, A. M., Ramond, S. N., & Sigurdsson, S., 2007, ApJ, 660, 823 Marcy, G. W., et al., 2001, ApJ, 556, 296 Marcy, G. W., et al., 2014, ApJS, 210, 20 Mart´ı, J. G., Giuppone, C. A., & Beaug´e, C., 2013, MNRAS, 433, 928 Mayor, M., et al., 2011, arXiv:1109.2497 Morbidelli, A., et al., 2007, AJ, 134, 1790 Morbidelli, A., & Crida, A., 2007, Icarus, 191, 158 Nelson, B. E., et al., 2016, MNRAS, 455, 2484 Ogihara, M., & Ida, S. 2009, ApJ, 699, 824 Ogihara, M., Inutsuka, S. & Kobayashi, H. 2013, ApJL, 778, 9 Ogihara, M., Morbidelli, A., & Guillot, T. 2015, A&A, 578, A36 Pierens, A., & Nelson, R. P., 2008, A&A, 482, 333 Raymond S. N., Quinn T., Lunine J. I., 2004, Icarus, 168, 1 Raymond, S. N., et al., 2006, Science, 313, 1413 Raymond, S. N., et al., 2006, Icarus, 183, 265 Raymond, S. N., et al., 2008, MNRAS, 384, 663 Raymond, S. N., et al., 2008, ApJ, 687, L107 Raymond, S. N., & Cossou, C., 2014, MNRAS, 440, L11 Rivera, E. J., et al., 2010, ApJ, 719, 890 Rowe, J. F., et al., 2014, ApJ, 784, 45 Safronov,V.S., 1969, Evolution of the Protoplanetary Cloud and Formation of the Earth and the Planets (Moscow:Nauka) Schneider J. et al., 2011, A&A, 532, A79 Steffen, J. H., Batalha, N. M., Borucki, W. J., et al., 2010, ApJ, 725, 1226 Tanaka, H., & Ida, S., 1999, Icarus, 139, 350 c(cid:13) 2015 RAS, MNRAS 000, 1 -- ?? Terrestrial Planets Formation under Migration 9 Tanaka, H., Takeuchi, T., & Ward, W. R., 2002, ApJ, 565, 1257 Tera, F., et al., 1974, Earth and Planetary Science Letters, 22, 1 Terquem, C., & Papaloizou, J. C. B., 2007, ApJ, 654, 1110 Tsiganis, K., et al., 2005, Nature, 435, 459 Udry, S., et al., 2007, A&A, 469, L43 Wang, J., Xie, J. W., Barclay, T., & Fischer, D. A., 2014, ApJ, 783, 4 Wang, S., Ji, J. H., & Zhou, J. L., 2012, ApJ, 753, 170 Wang, S., & Ji, J. H., 2014, ApJ, 795, 85 Ward, W. R., 1997, Icarus, 126, 261 Wetherill, G. W., 1980, ARAA, 18, 77 Wetherill, G. W., & Stewart, G. R., 1989, Icarus, 77, 330 Zhang, H., & Zhou, J. L., 2010, ApJ, 714, 532 Zhang, H., & Zhou, J. L., 2010, ApJ, 719, 671 Zhang, N., & Ji, J. H., 2009, Science in China Series G, 52(5), 794 Zhang, N., Ji, J. H., & Sun, Z., 2010, MNRAS, 405, 2016 Zhang, X.-J., Li, H., Li, S.-T., & Lin, D. N. C., 2014, ApJL, 789, L23 Zhou, J. L., Aarseth, S. J., et al., 2005, ApJ, 631, L85 c(cid:13) 2015 RAS, MNRAS 000, 1 -- ?? 10 Z. Sun et al. ] U A [ s x A i j r o a m - i m e S 10.0 5.0 2.0 1.0 0.5 0.2 0.1 0.05 1.00 0.10 y t i c i r t n e c c E 0 1x106 2x106 3x106 4x106 5x106 Time [ Year ] 0.01 Figure 1. Orbital evolution of the planetesimals and Jupiter mass planet in simulation S1. The solid points show the planetesimals. The radii of the planetesimals are proportional to their mass. The color of each point shows the orbital eccentricity. The open circle shows the giant planet. As the Jupiter mass planet migrating inward, the planetesimals at the MMR orbits were excited to high eccentric orbits. Due to the inward-migrating giant planet, planetesimals were shepherded inward and hence the accretion rate in the inner part of the planetesimal disk increases. There are also some planetesimals were scattered to the outer part of the disk. Several terrestrial planets form at 5 Myr. c(cid:13) 2015 RAS, MNRAS 000, 1 -- ?? 102 100 10−2 5 4 3 2 1 0 0.2 0.15 0.1 0.05 0 100 Jupiter Planet 2 Planet 1 PP2/PP1 PJ/PP2 Jupiter Planet 2 Planet 1 Planet 1 Planet 2 Terrestrial Planets Formation under Migration 11 (a) (b) (c) (d) ) U A ( s i x a r o j a m − m e S i o i t a R d o i r e P y t i c i r t n e c c E ) ⊕ M ( M 100 101 102 103 104 Time (yr) 105 106 107 108 Figure 2. Results of the evolution of semi-major axis (upper panel) and eccentricities (lower panel) for simulation S1. Two inner planets undergo type I migration while the outer one under the influence of type II migration. Note that three planets are close to a 4:2:1 MMR. c(cid:13) 2015 RAS, MNRAS 000, 1 -- ?? 12 Z. Sun et al. ] U A [ s x A i j r o a m - i m e S 10.0 5.0 2.0 1.0 0.5 0.2 0.1 0.05 1.00 0.10 y t i c i r t n e c c E 0 1x106 2x106 Time [ Year ] 3x106 4x106 0.01 Figure 3. Orbital evolution of the planetesimals and Jupiter mass planet in simulation S2. The Jupiter mass planet in this simulation was fixed at 5 AU. The radii of the planetesimals are proportional to their mass. The color of each point shows the orbital eccentricity. We can see the planetesimals at its MMR orbits were excited to high eccentric orbits. The accretion rate in this simulation is much lower than in simulation S1. c(cid:13) 2015 RAS, MNRAS 000, 1 -- ?? Terrestrial Planets Formation under Migration 13 1.00 0.10 y t i c i r t n e c c E 0 1x106 2x106 Time [ Year ] 3x106 4x106 0.01 ] U A [ s x A i j r o a m - i m e S 10.0 5.0 2.0 1.0 0.5 0.2 0.1 0.05 Figure 4. Orbital evolution of the planetesimals and two giant planets in simulation S3. The Jupiter and Saturn in this simulation was fixed at 5 and 9.2 AU. The radii of the planetesimals are proportional to their mass. The color of each point shows the orbital eccentricity. The planetesimals at the MMR orbits of the Jupiter mass planet were excited to high eccentric orbits. c(cid:13) 2015 RAS, MNRAS 000, 1 -- ?? 14 Z. Sun et al. ] U A [ s x A i j r o a m - i m e S 10.0 5.0 2.0 1.0 0.5 0.2 0.1 0.05 1.00 0.10 y t i c i r t n e c c E 0 1x106 2x106 3x106 4x106 5x106 Time [ Year ] 0.01 Figure 5. Orbital evolution of the planetesimals and two giant planets in simulation S4. The Jupiter and Saturn in this simulation underwent type II migration. The radii of the planetesimals are proportional to their mass. The color of each point shows the orbital eccentricity. Note that three planets are close to a 4:2:1 MMR. c(cid:13) 2015 RAS, MNRAS 000, 1 -- ?? 102 100 10−2 4 3 2 1 0 0.8 0.6 0.4 0.2 0 100 Jupiter Planet 1 Saturn PP1/PPS PJ/PP1 Jupiter Planet 1 Saturn Planet 1 Terrestrial Planets Formation under Migration 15 (a) (b) (c) (d) ) U A ( s i x a r o j a m − m e S i o i t a R d o i r e P y t i c i r t n e c c E ) ⊕ M ( M 100 101 102 103 104 Time (yr) 105 106 107 108 Figure 6. Results of the evolution of semi-major axis (upper panel) and eccentricities (lower panel) for simulation S4. One inner planet undergo type I migration while the outer two under the influence of type II migration. Note that Saturn is kicked inward and three planets are evolved into 4:2:1 MMR. c(cid:13) 2015 RAS, MNRAS 000, 1 -- ?? 16 Z. Sun et al. ] U A [ s x A i j r o a m - i m e S 10.0 5.0 2.0 1.0 0.5 0.2 0.1 0.05 1.00 0.10 y t i c i r t n e c c E 0 1x106 2x106 3x106 4x106 5x106 Time [ Year ] 0.01 Figure 7. Orbital evolution of the planetesimals and two giant planets in simulation S5. The Jupiter and Saturn in this simulation underwent type II migration. The radii of the planetesimals are proportional to their mass. The color of each point shows the orbital eccentricity. Note that three planets are evolved into 4:2:1 MMR. c(cid:13) 2015 RAS, MNRAS 000, 1 -- ?? 0.2 0.1 e 0 0.2 0.1 e 0 0.2 0.1 e 0 0 Terrestrial Planets Formation under Migration 17 S1 S4 S5 101 100 10-1 10-2 10-3 ) s s a M h t r a E ( s s a M 0.1 0.2 a (AU) 0.3 0.4 Figure 8. Final configuration of the 3 runs, S1, S4 and S5, that have formed 4:2:1 MMR at 100 Myr. The open circle show the giant planets. The solid point show the formed terrestrial planets. The color of the terrestrial planets show their masses. c(cid:13) 2015 RAS, MNRAS 000, 1 -- ?? P1 P2 P3 (a) (b) (c) P2/P1 P3/P2 (d) (e) 5 5 5 18 Z. Sun et al. ) y a d ( d o i r e p l a t i b r O y t i c i r t n e c c E o i t a r d o i r e P 2 2 1 2 3 3 2 3 600 400 200 0 0.003 0.002 0.001 0.000 2.8 2.4 2.0 0.8 0.8 0.4 0.4 0.0 0.0 0.8 0.8 0 0.4 0.4 0.0 0.0 0 0 1 1 1 2 2 2 Time (Myr) 3 3 3 4 4 4 Figure 9. A typical run of simulations for terrestrial planets. Panel (a) shows the evolution of the orbital periods. Panel (b) displays the evolution of eccentricities and Panel (d) means the period ratio changed with the time. In panel (a) and (b), the black, red and blue lines represent the innermost (P1), middle (P2) and the outermost (P3) planets, respectively. In panel (c), the green line and pink line represent the period ratio of P2/P1 and P3/P2, respectively. Panel (d) and (e) display the resonance angles of each pairs of planets, where θ21 = 2λ1 −λ2 −1, θ22 = 2λ1 −λ2 −2, θ32 = 2λ2 −λ3 −2, and θ33 = 2λ2 −λ3 −3. c(cid:13) 2015 RAS, MNRAS 000, 1 -- ??
1603.01721	2	1603	2016-06-15T15:28:33	An independent discovery of two hot Jupiters from the K2 mission	[ "astro-ph.EP" ]	We report the discovery of two hot Jupiters using photometry from Campaigns 4 and 5 of the two-wheeled Kepler (K2) mission. K2-30b has a mass of $ 0.65 \pm 0.14 M_J$, a radius of $1.070 \pm 0.018 R_J$ and transits its G dwarf ($T_{eff} = 5675 \pm 50$ K), slightly metal rich ([Fe/H]$=+0.06\pm0.04$ dex) host star in a 4.1 days circular orbit. K2-34b has a mass of $ 1.63 \pm 0.12 M_J$, a radius of $1.38 \pm 0.014 R_J$ and has an orbital period of 3.0 days in which it orbits a late F dwarf ($T_{eff} = 6149 \pm 55$ K) solar metallicity star. Both planets were validated probabilistically and confirmed via precision radial velocity (RV) measurements. They have physical and orbital properties similar to the ones of the already uncovered population of hot Jupiters and are well-suited candidates for further orbital and atmospheric characterization via detailed follow-up observations. Given that the discovery of both systems was recently reported by other groups we take the opportunity of refining the planetary parameters by including the RVs obtained by these independent studies in our global analysis.	astro-ph.EP	astro-ph	Draft version September 15, 2018 Preprint typeset using LATEX style emulateapj v. 12/16/11 AN INDEPENDENT DISCOVERY OF TWO HOT JUPITERS FROM THE K2 MISSION Rafael Brahm1,2, Mat´ıas Jones1, N´estor Espinoza1,2, Andr´es Jord´an1,2, Markus Rabus1, Felipe Rojas1, James S. Jenkins3, Cristi´an Cort´es4,2, Holger Drass1, Blake Pantoja3, Maritza G. Soto3, Maja Vuckovi´c5 Draft version September 15, 2018 ABSTRACT We report the discovery of two hot Jupiters using photometry from Campaigns 4 and 5 of the two- wheeled Kepler (K2) mission. K2-30b has a mass of 0.589± 0.023MJ , a radius of 1.069± 0.021RJ and transits its G dwarf (Teff = 5675 ± 50 K), slightly metal rich ([Fe/H]= +0.06 ± 0.04 dex) host star in a 4.1 days circular orbit. K2-34b has a mass of 1.698 ± 0.055MJ , a radius of 1.377 ± 0.014RJ and has an orbital period of 3.0 days in which it orbits a late F dwarf (Teff = 6149 ± 55 K) solar metallicity star. Both planets were confirmed via precision radial velocity (RV) measurements obtained with three spectrographs from the southern hemisphere. They have physical and orbital properties similar to the ones of the already uncovered population of hot Jupiters and are well-suited candidates for further orbital and atmospheric characterization via detailed follow-up observations. Given that the discovery of both systems was recently reported by other groups we take the opportunity of refining the planetary parameters by including the RVs obtained by these independent studies in our global analysis. Keywords: kepler, exoplanets 1. INTRODUCTION Extrasolar planets with structural properties similar to Jupiter, orbiting at close separations from their host stars (a < 0.05 AU, P < 8 days) are known as hot Jupiters. Nowadays, ∼250 transiting hot Jupiters have been discovered, mostly thanks to the existence of ded- icated ground-based photometric surveys like HATNet (Bakos et al., 2004), SuperWasp (Pollacco et al., 2006) and HATSouth (Bakos et al., 2013). The brightness of the host stars of the majority of these transiting planets, coupled with the relatively strong observational signa- tures (e.g. transit depth, radial velocity semi-amplitude) have allowed the determination of both, the radii and the masses of most of the discovered transiting hot Jupiters, which has been used to directly compute their bulk den- sities. Moreover, by comparing this information with theoretical models (e.g. Fortney et al., 2007; Burrows et al., 2007), the inner structure and composition of these planets can be inferred. In addition to the estimation of the physical parameters, if the hosts stars are bright enough, the execution of detailed photometric and spec- troscopic follow-up observations on these systems permit to characterize their atmospheric structure and compo- sition via transmission spectroscopy and/or secondary eclipses (see, e.g., Seager & Deming, 2010; Crossfield, 2015); to refine the geometry of the orbit via the mea- surement of the Rossiter-McLaughlin effect (McLaugh- 1 Instituto de Astrof´ısica, Facultad de F´ısica, Pontificia Uni- versidad Cat´olica de Chile, Av. Vicuna Mackenna 4860, 782-0436 Macul, Santiago, Chile 2 Millennium Institute of Astrophysics, Av. Vicuna Mackenna 4860, 782-0436 Macul, Santiago, Chile 3 Departamento de Astronom´ıa, Universidad de Chile, Camino al Observatorio 1515, Cerro Cal´an, Santiago, Chile 4 Departamento de F´ısica, Facultad de Ciencias B´asicas, Uni- Jos´e Pedro versidad Metropolitana de la Educaci´on, Av. Alessandri 774, 7760197, Nunoa, Santiago, Chile 5 Instituto de F´ısica y Astronom´ıa, Facultad de Ciencias, Uni- versidad de Valpara´ıso, Gran Bretana 1111, Playa Ancha, Val- para´ıso 2360102, Chile lin, 1924; Rossiter, 1924); and to discover additional planetary companions by performing long term RV mon- itoring (e.g. Neveu-VanMalle et al., 2016), TTV analysis (e.g. Steffen et al., 2012) or searching for additional transits in the light curve (e.g. Becker et al., 2015). Even though hot Jupiters are arguably the most char- acterized type of extrasolar planet, there are several the- oretical problems about their existence that remain to be solved. For example, there is no consensus about how these massive planets reached their current short orbital semi-major axes. In situ formation has proven to be unlikely (Rafikov, 2006), but the current observa- tional evidence is not able to discriminate between gen- tle migration by gravitational interactions with the pro- toplanetary disk (Lin et al., 1996) and high eccentric- ity migration mechanisms(Rasio & Ford, 1996). On the other hand, the mass and radius determination of tran- siting hot Jupiters have revealed a wide diversity regard- ing their internal structure. In particular, an important fraction of these systems present radii that are too large to be explained with current theoretical models of plane- tary structure (e.g. Anderson et al., 2010; Hartman et al., 2011). The inflated radii of these planets has been shown to be correlated with the degree of insolation from their parent stars (Guilliot, 2005), but the main responsible mechanism is still unknown. The detection of more hot Jupiters, particularly those transiting bright stars, can be used to test theories about their structure and evolution. We report the discovery of two new systems by using data from the two-wheeled Kepler K2 mission. Unlike the original Kepler mission, K2 is currently observing fields that are located close the ecliptic plane, and ground-based facilities located in the southern hemisphere can be used to confirm the plane- tary nature of potential candidates. In this context, we are conducting a Chilean based RV follow-up project of K2 candidates which has already discovered a Neptune- sized planet with a period of ≈ 42 d (Espinoza et al., 2 Brahm et al. 2016 2016). The two hot Jupiters presented is this paper were independently discovered by other teams using facilities from the northern hemisphere (Lillo-Box et al., 2016; Johnson et al., 2016; Hirano et al., 2016). The paper is structured as follows. In §2 we present the data, which includes the K2 photometry and the high resolution spectra and radial velocities obtained with the HARPS, FEROS and CORALIE spectrographs. §3 de- scribes the joint analysis that was applied to the data and presents the derived parameters of the planetary systems. Finally, in §4 we discuss our findings. 2. DATA 2.1. K2 Photometry We analysed the photometric data of K2's campaigns 4 and 5. In particular, we obtained all the decorrelated light curves from Vanderburg & Johnson (2014), using the photometry with the optimal aperture. The method that we used to select the transiting planetary candidates is described in detail in Espinoza et al. (2016). After performing a Box Least Squares (BLS, Kov´acs, Zucker & Mazeh, 2002) algorithm, we found that the stars EPIC210957318 (K2-30b) and EPIC212110888 (K2-34b) showed significant periodic signals at 4.1 and 3.0 days, re- spectively. Both of these systems were selected as strong Jovian planetary candidates based on their transit prop- erties (depths, shapes and durations), and due to the lack of evident out of transit variations. Following Espinoza et al. (2016), both decorrelated light curves were normalised by applying a median filter with a 21 point (∼ 10.25 hour) window, which was then smoothed using a Gaus- sian filter with a 5-point standard-deviation. These nor- malised light curves were then used by our transit-fitting pipeline, which results are shown in Section 3.3 (see Fig- ures 1 and 2). 2.2. Spectroscopic follow-up observations Once both targets were identified as strong transit- ing hot Jupiter candidates, we proceeded to acquire high resolution spectra with three different stabilised instru- ments with the goal of measuring the RV variation of the stellar hosts produced by the gravitational pull of the planetary companions. In the case of K2-30 we obtained 4 spectra using the HARPS spectrograph (Mayor et al., 2003) mounted on the ESO 3.6m telescope at La Silla Ob- servatory and 5 spectra using the FEROS spectrograph (Kaufer & Pasquini, 1998) mounted on the MPG 2.2m telescope located in the same observatory. For K2-34 we obtained 7 spectra with FEROS and 3 spectra using the CORALIE spectrograph (Queloz et al., 2001) mounted on the 1.2m Euler Telescope in La Silla Observatory. The FEROS and CORALIE data were obtained us- ing the simultaneous calibration method (Baranne et al., 1996), in which we acquire a spectrum of a ThAr lamp with the comparison fibre while the spectrum of the sci- ence star is acquired with the principal fibre. We use the ThAr spectra to trace the instrumental instrumental velocity drifts produced by environmental changes inside the spectrograph. On the other hand, since the HARPS nightly drift is typically <1 m s−1, the observations with this instrument were performed with the comparison fi- bre pointing to the background sky in order to avoid contamination from saturated ThAr lines. The data from these three instruments were processed through dedicated pipelines built from a modular code that was designed to develop completely robust and au- tomated pipelines for reducing, extracting and analysing echelle spectra of different instruments in a optimal and homogeneous way (Jord´an et al., 2014). Briefly, the pipelines identify the echelle orders using the flat frames, and after correcting by the bias level and the scattered light, the orders of the science and wavelength calibration images are optimally extracted following Marsh (1989). The reference global wavelength calibration solution is computed from the calibration ThAr image acquired in the afternoon by fitting a chebyshev polynomial as a function of the pixel position and echelle order number. If required, the instrumental drifts during the night are computed using the extracted flux of the comparison fi- bre, which is illuminated either by a ThAr lamp or by a Fabry-Perot etalon. The extracted flat frames are used to perform the correction by the blaze function, and then, a low order polynomial is fitted to each order, with an iterative algorithm that avoids the inclusion of absorp- tion lines in the fit, in order to construct a continuum normalised spectrum. The barycentric correction is per- formed using the JPLephem package, and RVs and bisec- tor spans (BSs) are determined by computing the cross- correlation function between the continuum normalised spectrum and a binary mask that resembles the spectrum of a G2 type star. We found that both systems present RV variations in phase with the photometric ephemeris, and with semi- amplitudes consistent with planetary companions. Ta- ble 1 lists the resulting RV variations and BSs computed for both systems, which are plotted against each other in Figure 3. The the lack of correlation between both parameters supports the planetary hypothesis as an ex- planation for the transits observed in K2-30 and K2-34. In Figures 4 and 5 we show the phase-folded RVs for K2-30 and K2-34 obtained in this work, along with the measurements reported by the other groups. 3. ANALYSIS 3.1. Planet scenario validation and transit dilutions We performed a blend analysis for both systems by us- ing the vespa package (Morton, 2012), which allowed us to compute the false-positive probability (FPP) of the transits being produced by different configurations of di- luted eclipsing binaries. By assuming an occurrence rate of 1% for hot Jupiter-like planets (Wang et al., 2015), and using only the K2 light-curves extracted with the optimal aperture according to Vanderburg & Johnson (2014), which corresponds to 3 pixels (12(cid:48)(cid:48)) for K2-30 and 4 pixels (16(cid:48)(cid:48)) for K2-34, we obtain a FFP of 0.18% and 21% for K2-30b and K2-34b, respectively. Given that our RV measurements are in phase with the photometric ephemeris and that their semi-amplitudes are consistent with planetary mass companions, the obtained FFPs cor- respond to upper limits in both cases. Nonetheless, the FPP of K2-30b is smaller than the accepted 1% threshold (e.g., Montet et al., 2015), and therefore it can be vali- dated by the photometry alone. On the other hand, for validating K2-34b we can further use the information ob- tained by our spectroscopic data. Given that we do not observe any evident secondary peaks in the CCF and that Two Hot Jupiters from the K2 mission 3 Figure 1. Phase-folded photometric data of the detrended and normalised K2 light curve for the star K2-30. The red solid line corresponds to the model with the posterior parameters obtained by the exonailer code. Figure 2. Phase-folded photometric data of the detrended and normalised K2 light curve for the star K2-34. The red solid line corresponds to the model with the posterior parameters obtained by the exonailer code. Figure 3. Bisector span measurements as function of the radial velocities for K2-30b (left panel) and K2-34b (right panel) obtained with FEROS (circles), Coralie (squares) and HARPS (triangles). In addition to the velocity variation in phase with the photometric ephemeris for these systems, the lack of correlations between both quantities indicate that the variations are probably produced by the gravitational pull of a giant planets and not originated by blended eclipsing binaries. 0.980.9850.990.9951−3−2−10123−20000−15000−10000−50000NormalizedfluxHoursfromMid-TransitNormalizedflux(ppm)0.990.99250.9950.997511.0025−3−2−10123−10000−7500−5000−250002500NormalizedfluxHoursfromMid-TransitNormalizedflux(ppm)−100−50050100−150−100−50050100150BS[m/s]RV[m/s]−1000100−200−1000100200BS[m/s]RV[m/s]4 Brahm et al. 2016 Radial velocities measured for K2-30 and K2-34. Table 1 σRV BS RV σBS m s−1 m s−1 m s−1 m s−1 35710 35549 35552 35620 35687 35738 35535 35609 35588 46497 46533 46496 46221 46527 46534 46525 46593 46217 46536 −3 −69 −4 −39 24 −49 9 −61 −41 35 61 29 53 06 17 10 −55 43 53 35 35 18 18 17 16 17 16 21 10 10 10 10 10 11 12 19 17 17 27 27 13 13 12 12 12 11 15 10 10 10 10 10 12 13 41 37 38 IF K2 − 30 K2 − 30 K2 − 30 K2 − 30 K2 − 30 K2 − 30 K2 − 30 K2 − 30 K2 − 30 K2 − 34 K2 − 34 K2 − 34 K2 − 34 K2 − 34 K2 − 34 K2 − 34 K2 − 34 K2 − 34 K2 − 34 BJD (UTC) 2457329.620606 2457330.778113 2457331.667931 2457332.689271 2457385.570390 2457386.570970 2457388.603838 2457389.588020 2457401.609029 2457383.765779 2457385.763132 2457386.755100 2457387.764030 2457388.772605 2457389.797512 2457389.698375 2457410.646391 2457408.654077 2457409.673645 Instrument HARPS HARPS HARPS HARPS FEROS FEROS FEROS FEROS FEROS FEROS FEROS FEROS FEROS FEROS FEROS FEROS CORALIE CORALIE CORALIE Two Hot Jupiters from the K2 mission 5 the radial velocity amplitudes are too small to come from stellar objects, we can set the likelihood of all eclipsing binary scenarios to 0, excluding line-of-sight blends and hierarchical triples. With these assumptions, the FPP of K2-34b drops to 0.052%, which validates its planetary nature. While we reject that the observed transits are produced by eclipsing binaries, we cannot rule out that the observed planetary transits are being diluted by the presence of another foreground or background star, or a bound star in a wide orbit. For K2-30b, changes in the inferred planetary radius of the order of the errors can be produced by sources located inside the aperture which are at least ≈ 3.7 magnitudes fainter than the host star. On the other hand, for K2-34b, changes in the inferred planetary radius on the order of the errors can be pro- duced by sources which are at least ≈ 2.0 magnitudes fainter than the host star. However, most of these con- taminant sources can be rejected from the lack of evident additional stars in the POSS images centred in K2-30 and K2-34; and also due to the lack of secondary peaks in the CCF plots. 3.2. Stellar properties In order to obtain the properties of the host stars, we made use of the available photometric and spectroscopic observables for both targets. We retrieved B,V ,g,r and i photometric magnitudes from the AAVSO Photomet- ric All-Sky Survey (APASS, Henden & Munari, 2014) and J, H and K photometric magnitudes from 2MASS for our analysis. For the spectroscopic data, we used the Zonal Atmospherical Stellar Parameter Estimator (ZASPE, Brahm et al., 2015) algorithm with our FEROS spectra as input. ZASPE estimates the atmospheric stel- lar parameters and v sin i from our high resolution echelle spectra via a least squares method against a grid of syn- thetic spectra in the most sensitive zones of the spectra to changes in the atmospheric parameters. ZASPE ob- tains reliable errors in the parameters, as well as the cor- relations between them by assuming that the principal source of error is the systematic mismatch between the data and the optimal synthetic spectra, which arises from the imperfect modelling of the stellar atmosphere or from poorly determined parameters of the atomic transitions. We used a synthetic grid generated using the spectrum code (Gray, 1999) and the ATLAS9 stellar atmospheres (Kurucz, 1993). The spectral region that was considered for the analysis was from 5000 to 6000 A, which includes a large number of atomic transitions and the pressure sensitive Mg Ib lines. The resulting atmospheric parameters obtained by ZASPE were Teff = 5575 ± 50 K, log(g) = 4.60 ± 0.05, [Fe/H] = +0.06 and v sin(i) = 0.5 ± 0.5 km s−1 for K2-30, and Teff = 6149 ± 55 K, log(g) = 4.2 ± 0.09, [Fe/H] = 0.0 ± 0.04 and v sin(i) = 6.31 ± 0.2 km s−1 for K2-34. We used the isochrones package (Morton, 2012) and the Dartmouth Stellar Evolution Database (Dotter et al., 2008) to obtain the physical properties of both stars (mass, radius and age) from the derived atmospheric parameters and the available photometric magnitudes. We took into account the uncertainties in the photomet- ric and spectroscopic properties to estimate the phys- ical properties of the stars, using the MultiNest algo- rithm (Feroz & Hobson, 2008), which allow us to effi- ciently explore the posterior parameter space. For K2-30 we obtained a radius of R∗ = 0.839+0.017 −0.014R(cid:12), a mass M∗ = 0.917+0.014 −0.014M(cid:12), an age of 2.2+1.8−0.8 Gyr, and a distance to the host star of 297.2+6.7−5.6 pc. For K2-34 we obtained a radius of R∗ = 1.58+0.16−0.15R(cid:12), a mass M∗ = 1.226+0.060 −0.045M(cid:12), an age of 4.24+0.39−0.44 Gyr and a distance to the host star of 390+39−37 pc. The stellar pa- rameters of the two host stars are sumarized in Table 2. 3.3. Joint analysis We performed a joint analysis of the detrended and normalised K2 photometry and the radial velocities using the EXOplanet traNsits and rAdIal veLocity fittER, exonailer, which is made publicly available at Github6 and its structure and funcionalities are described in Es- pinoza et al. (2016). Given that both systems were re- cently discovered independently by other groups, we in- cluded also the RV measurements of these projects in the analysis in order to further refine the physical parame- ters of these planets. The joint model fits for the instru- mental velocity offsets between different echelle spectro- graphs and also fits for the jitter of each instrument. For the radial velocities, gaussian priors were set on the semi-amplitude, K, and the RV zero point, µ. The for- mer was centred on zero, while the latter was centred on the observed mean of the RV dataset. Initially we consid- ered the eccentricity of both systems as a free parameter, however we obtained that in both cases the data was con- sistent with circular orbits, and therefore we performed a second joint analysis again by fixing the eccentricity to 0. For the lightcurve modelling, we used the selec- tive resampling technique described in Kipping (2010) in order to account for the 30 min cadence of the K2 photometry, which produces a smearing of the transit shape. In order to minimize the biases in the retrieved transit parameters we fit for the limb darkening coeffi- cients in our analysis (see Espinoza & Jord´an, 2015). We parametrized the limb-darkening effect using the square root law, because for the properties of our two systems, it provides the minimum mean square error, following the method described in Espinoza & Jord´an (2016). We used a white-noise model to treat the photometric residu- als, because we tried first to fit a flicker-noise model, but the parameters obtained with this model were consistent with no 1/f noise component. 500 walkers were used to evolve the MCMC, and each one explored the parameter space in 2000 links, 1500 of which were used as burn-in samples. This gave a total of 500 links sampled from the posterior per walker, giving a total of 250000 sam- ples from the posterior distribution. These samples were tested to converge both visually and using the Geweke (1992) convergence test. The median values of the posterior distributions for each parameter are tabulated in Table 3 along with their errors, which are given by the 16th and 84th percentiles of the posterior distributions for the lower and upper errors, respectively. The priors used for the analysis are shown in Table 4. The modelled light curves with the ob- tained posterior parameters are plotted in Figures 1 and 6 http://www.github.com/nespinoza/exonailer 6 Brahm et al. 2016 Stellar parameters of K2-30 and K2-34. Table 2 K2-30 Value K2-34 Value EPIC210957318 EPIC212110888 03292204+2217577 08301891+2214092 03h29m22.07s 22o17(cid:48)57.86(cid:48)(cid:48) 25.9 ± 2.3 −13.6 ± 2.4 5575 ± 50 0.06 ± 0.04 4.6 ± 0.05 0.5 ± 0.50 G 13.171 14.506 ± 0.030 13.530 ± 0.039 13.346 ± 0.008 12.763 ± 0.042 12.443 ± 0.06 11.63 ± 0.007 11.194 ± 0.008 11.088 ± 0.007 0.917+0.014 −0.014 0.839+0.017 −0.014 2.202+0.09−0.14 0.537+0.035 −0.031 297.2+6.7−5.6 2.2+1.8−0.8 08h30m18.91s 22o14(cid:48)09.27(cid:48)(cid:48) −14.1 ± 0.8 −0.3 ± 0.5 6149 ± 55 0.00 ± 0.04 4.2 ± 0.09 6.31 ± 0.20 F 11.441 12.429 ± 0.033 11.548 ± 0.057 11.892 ± 0.119 11.892 ± 0.119 11.389 ± 0.026 10.264 ± 0.038 10.519 ± 0.004 10.187 ± 0.010 1.226+0.060 −0.045 1.58+0.16−0.15 0.43+0.14−0.09 3.05+0.67−0.57 390+39.0−37.0 4.24+0.39−0.44 Parameter Identifying Information EPIC ID 2MASS ID R.A. (J2000, h:m:s) DEC (J2000, d:m:s) R.A. p.m. (mas/yr) DEC p.m. (mas/yr) Spectroscopic properties Teff (K) Spectral Type [Fe/H] (dex) log(g) (cgs) v sin(i) (km s−1) Photometric properties Kp (mag) B (mag) V (mag) g(cid:48) (mag) r(cid:48) (mag) i(cid:48) (mag) J (mag) H (mag) Ks (mag) Derived properties M∗ (M(cid:12)) R∗ (R(cid:12)) ρ∗ (g/cm3) L∗ (L(cid:12)) Distance (pc) Age (Gyr) Note. Logarithms given in base 10. Source 2MASS EPIC EPIC UCAC4 UCAC4 ZASPE ZASPE ZASPE ZASPE ZASPE EPIC APASS APASS APASS APASS APASS 2MASS 2MASS 2MASS isochrones+ZASPE isochrones+ZASPE isochrones+ZASPE isochrones+ZASPE isochrones+ZASPE isochrones+ZASPE Figure 4. The top panel shows the phase-folded RVs obtained in this work along with the RVs from Johnson et al. (2016) and Lillo-Box et al. (2016) for K2-30b (blue:FEROS, red: HARPS, black:FIES, green: SOPHIE, yellow: HARPS-N). The continuous line corresponds to modelled RV signal with the posterior param- eters. The bottom panel shows the corresponding residuals. 2 for K2-30b and K2-34b, respectively; while the corre- sponding models for the RV curves are shown in Figures 4 and 5. The derived physical and orbital parameters for Figure 5. Same as in Figure 4 but for K2-34b (red: Coralie, black: CAFE, orange: HDS). both systems are consistent with being hot Jupiters. For K2-30b we obtain a mass of Mp = 0.589±0.023MJ , a ra- dius of RP = 1.069±0.021RJ and an equilibrium temper- ature of Teq = 1203±19K, assuming zero albedo. On the other hand, for K2-34b we obtain Mp = 1.698± 0.06MJ , RP = 1.377 ± 0.14RJ and Teq = 1715 ± 17K, where the −100010000.20.40.60.81O-C[m/s]Orbitalphase−100−50050100RV[m/s]−100010000.20.40.60.81O-C[m/s]Orbitalphase−200−1000100200RV[m/s]Two Hot Jupiters from the K2 mission 7 large uncertainty in the radius is dominated by the large uncertainty in the radius of the host star. Priors for the joint analysis of K2-30b and K2-34b. Table 4 Orbital and planetary parameters for K2-30b and K2-34b. Table 3 Parameter Lightcurve parameters K2-30b Prior K2-34b Prior Parameter Lightcurve parameters P (days) . . . . . . . . . . . . . T0 − 2450000 (BJD) . a/R(cid:63) . . . . . . . . . . . . . . . Rp/R(cid:63) . . . . . . . . . . . . . . i (deg) . . . . . . . . . . . . . . . q1 . . . . . . . . . . . . . . . . . . q2 . . . . . . . . . . . . . . . . . . σw (ppm) . . . . . . . . . . . RV parameters K (m s−1) . . . . . . . . . . µFEROS (km s−1) . . µHARPS (km s−1) . . µFIES (km s−1) . . . . . µSOPHIE (km s−1) . µHARPS-N (km s−1) µCoralie (km s−1) . . . µCAFE (km s−1) . . . σFEROS (km s−1) . . σHARPS (km s−1) . . σFIES (km s−1) . . . . . σSOPHIE (km s−1) . σHARPS-N (km s−1) σCoralie (km s−1) . . . σCAFE (km s−1). . . . e . . . . . . . . . . . . . . . . . . . K2-30b K2-34b Posterior Value Posterior Value 4.09849+0.00002 −0.00002 7067.90559+0.00018 −0.00018 2.995629+0.000006 −0.000006 7144.34703+0.00008 −0.00008 10.70+0.26−0.28 0.13097+0.0009 −0.0009 85.86+0.21−0.23 0.53+0.17−0.22 0.42+0.21−0.23 281.2+2.9−2.7 79.0+2.7−3.0 35.634+0.009 −0.008 35.601+0.008 −0.008 35.431+0.010 −0.011 35.506+0.006 −0.005 35.629+0.002 −0.002 - - 0.027+0.015 −0.009 0.017+0.019 −0.013 0.005+0.012 −0.003 0.005+0.009 −0.003 0.002+0.003 −0.001 - - 6.30+0.10−0.10 0.0895+0.0007 −0.0006 82.23+0.19−0.19 0.58+0.10−0.11 0.55+0.14−0.14 78.2+0.7−0.6 207.0+3.1−3.0 46.417+0.008 −0.008 - - 46.311+0.011 −0.010 46.394+0.003 −0.002 46.449+0.009 −0.009 46.075+0.037 −0.035 0.015+0.011 −0.008 - - 0.006+0.015 −0.004 0.008+0.003 −0.002 0.008+0.025 −0.006 0.015+0.042 −0.012 < 0.08 (at 96%) < 0.054 (at 96%) Derived Parameters Mp (MJ ) . . . . . . . . . . . Rp (RJ ) . . . . . . . . . . . . ρp (g/cm3) . . . . . . . . . log gp (cgs) . . . . . . . . . a (AU) . . . . . . . . . . . . . Vesc (km s−1) . . . . . . Teq (K) . . . . . . . . . . . . . Bond albedo of 0.0 Bond albedo of 0.75 0.589+0.023 −0.022 1.069+0.023 −0.019 0.598+0.039 −0.043 3.106+0.022 −0.025 0.0419+0.0016 −0.0012 44.19+0.90−0.95 1203+18−19 851+12−13 Note. Logarithms given in base 10. 1.698+0.061 −0.050 1.377+0.14−0.13 0.80+0.26−0.18 3.35+0.08−0.07 0.0465+0.0046 −0.0047 66.1+2.7−2.3 1715+16−18 1212+11−13 3.4. Searching for additional signals in the K2 photometry We searched for additional signals in the photometry of both targets stars in order to search for additional transiting companions, secondary eclipses and/or opti- cal phase variations due to either reflected light of the detected transiting planets, ellipsoidal variations and/or doppler beaming (e.g. Estevez, De Mooij & Jayaward- hana, 2013). The transit search was performed using the Box Least N (4.098, 0.10) P (days) . . . . . . . . . . . . . N (2.996, 0.10) T0 − 2450000 (BJD) . N (7067.90, 0.10) N (7144.35, 0.10) a/R(cid:63) . . . . . . . . . . . . . . . Rp/R(cid:63) . . . . . . . . . . . . . . i (deg) . . . . . . . . . . . . . . . U (80.0, 90.0) U (80.0, 90.0) U (1, 30) U (0.01, 0.5) U (0.01, 0.5) U (1, 30) q1 . . . . . . . . . . . . . . . . . . q2 . . . . . . . . . . . . . . . . . . σw (ppm) . . . . . . . . . . . U (0.0, 1.0) U (0.0, 1.0) U (0.0, 1.0) U (0.0, 1.0) J(1.0, 2000.0) J(1.0, 2000.0) RV parameters K (m s−1) . . . . . . . . . . µFEROS (km s−1) . . µHARPS (km s−1) . . µFIES (km s−1) . . . . . µSOPHIE (km s−1) . µHARPS-N (km s−1) µCoralie (km s−1) . . . µCAFE (km s−1) . . . σRV (km s−1) . . . . . . . e . . . . . . . . . . . . . . . . . . . N (0.0, 0.1) N (0.0, 0.1) N (35.63, 0.01) N (46.47, 0.1) N (35.60, 0.01) N (35.45, 0.1) N (35.50, 0.01) N (35.63, 0.01) - - J(0.001, 0.1) U (0.0, 1.0) - - N (46.26, 0.1) N (46.40, 0.01) N (46.45, 0.01) N (46.01, 0.1) J(0.001, 0.1) U (0.0, 1.0) Note. N (µ, σ) corresponds to a normal distribution with mean µ and variance σ2. U (a, b) corresponds to an uniform distribution between the values a and b. J(a, b) corresponds to a Jeffrey's prior between the values a and b. Squares (BLS) algorithm on the data with the transit of the detected planetary companion masked. For each significant peak on the BLS periodogram, we visually inspected the phased lightcurve in order to search for additional transits. In addition, the lightcurve was also inspected at periods 2, 3/2, 1/2 and 2/3 times the period of the detected transiting planet presented in this work in order to search for additional companions in 2:1 and 3:2 mean-motion resonances. For both K2-30 and K2-34, no additional transiting companions were found, which limit the possible com- panions to transit depths smaller than 200 ppm and 90 ppm, respectively, at 3-sigma. Also, no secondary eclipses and optical phase variations were detected on ei- ther lightcurve. Given the transit parameters that we ob- tain of K2-30b, the non-detection of a secondary eclipse was expected, as they would have to be smaller than (Rp/a)2 ∼ 150 ppm. Optical phase variations were ex- pected to be below this limit as well. In the case of K2- 34b, (Rp/a)2 ∼ 200 ppm, and our analysis rules out any secondary eclipse larger than 90 ppm at 3-sigma; this im- plies that the geometric albedo of K2-34b is constrained by our data to be less than 0.45, which is in agreement again with the typical geometric albedo of hot Jupiters (Heng & Demory, 2013). No optical phase variations were detected either. 4. DISCUSSION In this paper we present an independent discovery of two transiting hot Jupiters orbiting main sequence stars, that were first selected as candidates from K2 photome- try of campaigns 4 and 5. The planetary nature of these two objects was then confirmed by precision RV measure- 8 Brahm et al. 2016 ments using three high resolution echelle spectrographs located in the southern hemisphere. Both systems were recently announced by other teams that performed independent follow-up campaigns using spectroscopic facilities located in the northern hemi- sphere (Lillo-Box et al., 2016; Johnson et al., 2016; Hi- rano et al., 2016). Even though the data presented in this work was sufficient for confirming the planetary nature of both candidates and to obtain reliable estimations for their physical parameters, we decided to include the ra- dial velocity measurements obtained in these three other works in order to refine the estimation of the planetary parameters. This procedure allowed us to have a better phase coverage for both orbits, which was particularly useful in the case of K2-34b, for which the use of ve- locity measurements obtained from facilities at different geographical longitudes allow us to partially counteract the effect produced by the peculiar value of the orbital period of this planet, which is almost exactly a multi- ple of one day. By combining the radial velocity data we obtained smaller uncertainties in the masses for both planets with respect the the errors reported by the other three groups. However, the planetary mass estimations obtained by all the different groups agree with each other at the level of the reported uncertainties We found that the mass of K2-30b is in-between the Saturn and Jupiter mass (Mp = 0.589± 0.023MJ ), while its radius (Rp = 1.069 ± 0.021RJ ) is slightly larger than the one of Jupiter. In contrast, we found that K2-34 is significantly more massive (Mp = 1.698 ± 0.056MJ ) and larger Rp = 1.377±0.014RJ ) than Jupiter. However, the physical and orbital properties of both of these systems resemble quite well the ones of the typical population of known hot Jupiters, which can be visualised with Fig- ure 6. In the left panel of Figure 6, the mass-radius diagram for the complete population of discovered tran- siting hot Jupiters (P < 10 days, RP > 0.5RJ ), shows that according to our analysis, the physical parameters of K2-30b and K2-34b lie in densely populated regions of the parameters space, and that both planets share a sim- ilar density, close to half the one of Jupiter (ρJ = 0.67 g cm−3). Another particularity to notice from Figure 6 is that, while the inferred radius of K2-30b can be explained with the models of planetary structure of Fortney et al. (2007) by requiring a core mass of ∼ 15 M⊕, the ra- dius of K2-34b is significantly larger than that predicted by these models, and suffers from the radiative and/or tidal inflation mechanisms that typical hot Jupiters are victims of (see Spiegel & Burrrows, 2013). However, it is important to note that, while the radius for K2-30b computed by Johnson et al. (2016) is consistent with the one presented in this work requiring the presence of a solid core, Lillo-Box et al. (2016) found a larger radius for this planet which requires no core and a certain level of inflation. The origin of this discrepancy relies in the estimation of the physical parameters of the host star. While the stellar radius for K2-30 reported in this work (0.839 ± 0.015 R(cid:12)) is consistent with the estimation of Johnson et al. (2016, 0.844 ± 0.032 R(cid:12)) , Lillo-Box et al. (2016) obtains a significantly larger value (0.941 ± 0.041 R(cid:12)). This issue shows the importance of performing ho- mogeneous analysis when global trends and correlations between planetary and stellar parameters are searched. For K2-34b, the different estimations of the planetary radius presented by the other groups are consistent with the value reported in this work. The presented dichotomy in the structure of K2-30b and K2-34b can be explained by the different insolation levels to which they are subjected. The right panel of Figure 6, shows the radii of the discovered transiting hot Jupiters as function of their equilibrium temperatures assuming zero albedo, which shows the correlation first noted by Guilliot (2005). While the equilibrium temper- ature of K2-30b (1203 K) is relatively close to the thresh- old limit proposed by Kov´acs et al. (2010) of Teq = 1000 K, below which the inflation of the radius of Jovian plan- ets is not significant (see also Demory & Seager, 2011), the equilibrium temperature of K2-34b is relatively high (1715 K) and its inferred radius is totally compatible with the correlation. Finally, the two systems are interesting candidates for follow-up studies. The low density of K2-30b combined with the relatively small radius of its host star implies a scale height of 340 km and a transmission spectroscopic signal of 744 ppm (assuming an H2 dominated atmo- sphere and a signal of 5 scale-heights), which means that this system is a good target to be observed via trans- mission spectroscopy to characterize its atmosphere. In addition to the large expected signal, this system pos- sesses a nearby (≈ 1(cid:48)) stellar companion with a similar brightness, which can be used as a comparison source for long slit transmission spectroscopy. On the other hand, the v sin(i) value of K2-34 and its moderately bright na- ture, make of this system a good target for measuring the Rossiter-McLaughlin effect in order to determine the sky- projected obliquity angle. This effect has indeed been al- ready measured by Hirano et al. (2016), who found that K2-34b probably lies in a prograde orbit with low obliq- uity. R.B. are and N.E. supported by CONICYT- PCHA/Doctorado Nacional. A.J. acknowledges support from FONDECYT project 1130857 and from BASAL CATA PFB-06. R.B., N.E., A.J. and C.C. acknowledge support from the Ministry for the Economy, Develop- ment, and Tourism Programa Iniciativa Cient´ıfica Mile- nio through grant IC 120009, awarded to the Millennium Institute of Astrophysics (MAS). J.S.J. acknowledges support from BASAL CATA PFB-06. C.C. acknowl- edges support from FONDECYT project 11150768. This paper includes data collected by the Kepler mission. Funding for the Kepler mission is provided by the NASA Science Mission directorate. It also made use of the SIM- BAD database (operated at CDS, Strasbourg, France), NASA's Astrophysics Data System Bibliographic Ser- vices, and data products from the Two Micron All Sky Survey (2MASS) and the APASS database and the Digi- tized Sky Survey. Based on observations collected at the European Organisation for Astronomical Research in the Southern Hemisphere under the ESO programm 096.C- 0417(A). REFERENCES Anderson, D. R., Hellier, C., Gillon, M., et al., 2010, ApJ, 709, 159. Bakos, G., Noyes, R. W., Kov´acs, G., et al., 2004, PASP, 116, 266. Two Hot Jupiters from the K2 mission 9 Figure 6. Left: the mass-redius diagram for the population of known transiting hot Jupiters. We show as coloured circles and triangles the values obtained for K2-30b and K2-34b, respectively (red: this work, green: Lillo-Box et al. (2016), blue: Hirano et al. (2016), and light blue: Johnson et al. (2016)). The light dashed lines correspond to isodendity curves of 0.67, 1.33 and 3.66 g cm−3, from left to right. The dark dashed lines correspond to the Fortney et al. (2007) models for typical properties of hot Jupiters (a = 0.045 AU, 3 Gyr,), and central core masses of 100 M⊕ (bottom curve) and 0 M⊕ (top curve). According to this figure, the two new systems can be classified as common hot Jupiters having densities close to half the one of Jupiter. However, for K2-30b, the radius calculated by Lillo-Box et al. (2016) seems to be significantly larger than the estimations obtained in this work and in Johnson et al. (2016). Right: radius as function of the theoretical equilibrium temperature for the complete population of discovered transiting hot Jupiters. Symbols and colours are the same than in the left panel. Both systems follow the known correlation between the level of insolation and the degree of inflation in radii. The radius of K2-30b can be explained with the Fortney et al. (2007) models because its equilibrium temperature lies close to the lower limit of 1000 K (dashed line) below which planets are not expected to be inflated. Bakos, G. ´A., Csubry, Z., Penev, K., et al., 2013, PASP, 125, 154. Baranne, A., Queloz, D., Mayor, M., et al., 1996, AAPS, 119, 373. Becker, J. C., Vanderburg, A., Adams, F. C., et al., 2015, ApJL, 812, L18. Brahm, R., Jord´an, A., Hartman, J. D., 2015, AJ, 150, 33. Burrows, A., Hubeny, I., Budaj, J., et al., 2007, ApJ, 661, 502. Crossfield, I. J. M., 2015, PASP, 127, 941. Demory, B. O., & Seager, S., 2011, ApJS, 197, 12 Dotter, A., Chaboyer, B., Jevremovi´c, D., et al., 2008, ApJS, 178, 89. Espinoza, N. & Jord´an, A., 2015, MNRAS, 450, 1879. Espinoza, N. & Jord´an, A., 2016, MNRAS, in press. Espinoza, N., Brahm, R., Jord´an, A., et al., 2016, ArXiv e-prints, 1601.07608. Esteves, L. J., De Mooij, E. J. W. & Jayawardhana, R., A.,2013, ApJ, 772, 51. Feroz, F., Hobson, M. P., 2008, MNRAS, 384, 449. Foreman-Mackey, D., Hogg, D., Lang, D. & Goodman, J., 2013, PASP, 125, 306. Fortney, J. J., Marley, M. S., Barnes, J. W., et al., 2007, ApJ, 659, 1661. Geweke, J., 1992, Bayesian Statistics, 169. Goodman, J. & Weare, J., 2010, Comm. App. Math. Comp. Sci., 5, 65. Gray, R. O., 1999, Astrophysics Source Code Library. Guillot, T., 2005, Annual Review of Earth and Planetary Sciences, 33, 493. Hartman, J. D. and Bakos, G. ´A. and Torres, G., 2011, ApJ, 742, 59. Henden, A. & Munari, U., 2014, Contributions of the Astronomical Observatory Skalnate Pleso, 518. Heng, K. & Demory, B. O., 2013, ApJ, 777, 100. Hirano, T., Nowak, G., Kuzuhara, M., et al., 2016, ArXiv e-prints, 1602.00638. Johnson, M. C., Gandolfi, D., Fridlund, M., et al., 2016, AJ, 151, 171. Jord´an, A., Brahm, R., Bakos, G. ´A., et al. 2014, AJ, 148, 29 Kaufer, A., & Pasquini, L., 1998, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 3355, Optical Astronomical Instrumentation, ed. S. D'Odorico, 844-854. Kipping, D. M., 2010, MNRAS, 408, 1758. Kov´acs, G., Zucker, S. & Mazeh, T., 2002, A&A, 391, 369. Kov´acs, G. , Bakos, G. ´A., Hartman, J. D., 2002, ApJ, 724, 866. Kurucz, R. L., 1993, VizieR Online Data Catalog, 6039, 0. Lillo-Box, J., Demangeon, O., Santerne, A., et al., 2016, ArXiv e-prints, 1601.07635. Lin, D. N. C., Bodenheimer, P., Richardson, D. C., 1996, Nature, 380, 606. Marsh, T. R., 1989, PASP, 101, 1032. Mayor, M. and Pepe, F. and Queloz, D., 2003, The Messenger, 114, 20. McLaughlin, D. B., 1924, ApJ, 60, 22. Montet, B., Morton, T., Foreman-Mackey, D., et al., 2015, ApJ, 809, 25. Morton, T. D., 2012, ApJ, 761, 6. Neveu-VanMalle, M., Queloz, D., Anderson, D. R., 2016, A&A, 586, A93. Queloz, D., Mayor, M., Udry, S., et al., 2001, The Messenger, 105, 1. Pollacco, D. L., Skillen, I., Collier Cameron, A., et al., 2006, PASP, 118, 1407. Rafikov, R. R., 2006, ApJ, 648, 666. Rasio, F. A & Ford, E. B., 1996, Science, 274, 954. Rossiter, R. A., 1924, ApJ, 60, 15. Seager, S. & Deming, D., 2010, ApJ, 772, 76. Spiegel, D. S. & Burrows, A., 2010, ARAA, 48, 631. Steffen, J. H., Ford, E. B., Rowe, J. F., et al., 2012, ApJ, 756, 186. Vanderburg, A. & Jonson, J., 2014, PASP, 126, 948. Wang, J., Fischer, D. A. and Horch, E. P., et al., 2015, ApJ, 799, 229. Winn, J. N., Fabrycky, D., Albrecht, S., et al., 2010, ApJ, 718, L145. 0.50.7511.251.51.7520.010.1110100Radius[RJ]Mass[MJ]0.50.7511.251.51.75250010001500200025003000Radius[RJ]Teq[K]
1007.0029	1	1007	2010-06-30T21:19:20	Modeling the Radial Velocities of HD 240210 with the Genetic Algorithms	[ "astro-ph.EP" ]	More than 450 extrasolar planets are known to date. To detect these intriguing objects, many photometric and radial velocity (RV) surveys are in progress. We developed the Keplerian FITting (KFIT) code, to model published and available RV data. KFIT is based on a hybrid, quasi-global optimization technique relying on the Genetic Algorithms and simplex algorithm. Here, we re-analyse the RV data of evolved K3III star HD 240210. We confirm three equally good solutions which might be interpreted as signals of 2-planet systems. Remarkably, one of these best-fits describes the long-term stable two-planet system, involved in the 2:1~mean motion resonance (MMR). It may be the first instance of this strong MMR in a multi-planet system hosted by evolved star, as the 2:1 MMR configurations are already found around a few solar dwarfs.	astro-ph.EP	astro-ph	Modeling the Radial Velocities of HD 240210 with the Genetic Algorithms A. Rozenkiewicz1, K. Goździewski1 & C. Migaszewski1 1Toruń Centre for Astronomy, Nicolaus Copernicus University, Gagarin Str. 11, 87-100 Toruń, Poland {a.rozenkiewicz,k.gozdziewski,c.migaszewski}@astri.umk.pl More than 450 extrasolar planets are known to date. To detect these intriguing objects, many photometric and radial velocity (RV) surveys are in progress. We developed the Keplerian FITting (KFIT) code, to model published and available RV data. KFIT is based on a hybrid, quasi-global optimization technique relying on the Genetic Algorithms and simplex algorithm. Here, we re-analyse the RV data of evolved K3III star HD 240210. We confirm three equally good solutions which might be interpreted as signals of 2-planet systems. Remarkably, one of these best-fits describes long-term stable two-planet system, involved in the 2:1 mean motion resonance (MMR). It may be the first instance of this strong MMR in a multi-planet system hosted by evolved star, as the 2:1 MMR configurations are already found around a few solar dwarfs. Introduction Most of the discovered exoplanets are found around main sequence stars, perhaps because the determination of stellar parameters is much easier than, for instance for giant or active stars. The formation theory of planets hosted by such stars is still developed and not understood well. Recently, there are only known about of 10 exoplanets orbiting stars with masses greater than 2 M(cid:12) (see, e.g., Extrasolar Planets Encyclopaedia, http://exoplanet.eu). Some recent RV surveys focus on detecting planets around the red giants, which are evolved main sequence stars. Unfortunately, the giants and sub-giants are difficult targets for the RV technique. Usually, they are chromospherically active, pulsating, surface-polluted by large spots, and rotating slowly. The giants produce small number of sharp spectral lines and their chromospheric activity and spots may change the profiles of spectral lines. This intrinsic RV variability (also known as stellar "jitter") is significantly larger than instrumental errors, and may be ∼ 20 − 30 m/s and larger. Actually, the stellar activity may even mimic planetary signals (see, e.g. [1]). Hence, when interpreting the RV variability, we may expect that different uncertainties and errors (usually, of unknown origin) may significantly shift the best fits from the "true" solutions in the parameters space. In such a case, possibly global exploration of the parameter space and the dynamical stability of multiple systems as an additional, implicit observable may help us to correct and verify the derived best-fit solutions for these factors, and to conclude on the architecture of interacting systems, even when limited data are available [4]. Keplerian model of the RV and optimization method We recall the kinematic RV model for the N-planet system, as the first order approximation of the N-body model: N(cid:88) Vr(t) = Ki [cos(ωi + ν(t)) + ei cos(ωi)] + V0, (1) i=1 where, for each planet in the system, Ki is the semi-amplitude of the signal, ei is the orbital eccentricity, ωi is the argument of pericenter, νi(t) is the true anomaly, which depends on the orbital period Pi, the time of periastron passage τi and eccentricity, and V0 is a constant instrumental offset. The N-planet system is then characterized by Np = 5N + 1 free parameters, to be determined from observations. Let us note, that the inclinations and longitudes of nodes are not explicitly present in Eq. (1) and cannot be determined directly from the RV data alone, at least in terms of the kinematic model in Eq. (1). the best-fit model parameters, we seek for the minimum of the reduced χ2 function, min(cid:112)χ2 To fit model Eq. (1) to the RV data, the Gaussian Least-Squares method is commonly used. To estimate ν(p), which 1 YSC'17 Proceedings of Contributed Papers A. Rozenkiewicz, K. Goździewski & C. Migaszewski Clearly, even in the kinematic formulation,(cid:112)χ2 is computed on the basis of synthetic signal Vr(ti, p), where ti are moments of observations V obs r,i with uncertainties σi (i = 1, . . . , NRV is the number of data), and p ≡ (Kp, Pp, ep, ωp, τp, V0), and p = 1 ≡ b, 2 ≡ c, . . . , N are for the model parameters of the N-planet system. ν(p) is a non-linear function. It is well known that it may exhibit numerous local extrema. Hence, the exploration of multi-dimensional parameter space p requires efficient, possibly global numerical optimization. In the past, we found that good results in this problem may be achieved by an application of the Genetic Algorithms (GAs) [2], which mimic the biological evolution. GAs search for the best fit solutions (best adapted population members) to the model (to the environment) through "breading" an initial population (parameters set), under particular genetic operators (e.g., cross- over, mutation), and through a selection of gradually better adapted members. Although the GAs start with purely random initial "population", the search converges to the best fit solutions [2] by deterministic way. Of course, the best fit solutions may be not unique. This is very common in the case of modeling RV data. Overall, the GAs are robust and quasi-global optimization technique, although they not provide efficiency and accuracy of fast local methods, like the Levenberg-Marquardt algorithm. The GAs are used in our hybrid KFIT code, developed for a few years now [3], which also makes use of the local, and accurate simplex algorithm [6]. It helps to quickly refine the final population of the best fits, "grown" by the GAs. local minima of(cid:112)χ2 The results for the RV of HD 240210 In a recent work, Niedzielski et al. [5] detected a planet hosted by evolved dwarf HD 240210. They found an excess in the rms of Keplerian 1-planet model, and interpret this as possible signal of an additional planet. We reproduced their 1-planet solution, see Fig. 1 (the top-left panel). The rms has significant scatter of 39 m/s, compared to the mean instrumental accuracy of ∼ 8 m/s. This rms excess is difficult to explain by the internal variability of the star, hence we tried to find a better 2-planet model. At first, we reproduced (see Fig. 1, the top-right panel) a tentative 2-planet solution illustrated in the discovery paper. However, we notice that the orbital periods are very similar ∼ 440 and ∼ 530 days, respectively, indicating strong 4:3 MMR or 1:1 MMR. In such a case, the kinematic model is not proper anymore, due to significant mutual planetary interactions. Moreover, such configurations could be hardly explained by the planetary formation theory. Still, having in mind that this solution is not unique, we performed an extensive search for the ν with the KFIT code. The statistics of gathered fits is shown in Fig. 2, through their projection onto orbital periods plane. In the range of orbital periods ∈ [60, 3600] days, we found three, equally good best-fit models, which correspond to different orbital configurations, and may be resolved at the 2σ confidence level. In the parameter maps, the mentioned 2-planet model is labeled as Fit I. Two additional fits with orbital periods ratio close to 3:2 and 2:1 are labeled as Fit II and Fit III, respectively (see Table 1). Their synthetic RV curves, with the RV of 1-planet model (thin curve) and observations overplotted, are shown in subsequent panels of Fig. 1. Note, that the alternative 2-planet fits reduce the rms significantly, to ∼ 25 m/s (i.e., by 1/3), consistent with a conclusion in [5]. As may be seen in Table 1, Fits II and III have large eccentricities. A question remains, whether inferred orbital configurations are dynamically stable. In fact, all these Fits I -- III, transformed to osculating elements at the epoch of the first observation, lead to self-disruption of 2-planet systems. Nevertheless, remembering that stable solutions may still be found in their neighborhood, we did dynamical analysis with the N-body, self-consistent GAMP code [4] (which also relies on the GAs), trying to refine Fits I -- III with the requirement of the long-term stability (the edge-on, coplanar models are tested). Certainly, at most one of these models might correspond to the real system. We did not found any stable orbits in the vicinity of Fit I. In a tiny neighborhood of Fit II, there is a stable ν ∼ 1.50 and an rms ∼ 27 m/s which, as the direct numerical integrations show, is stable over 1 Gyr. The best result is found for Fit III as a stable solution, corresponding to the 2:1 MMR ν ∼ 1.36 and an rms of ∼ 25 m/s, i.e., the same as in the kinematic Fit III. The dynamical map [4] around this solution (Fig. 3, the right-hand panel) reveals extended island of stability (∼ 0.15 au). This fit has moderate semi-amplitude librations (∼ 15 − 30◦) of the critical angles θ1 = 2λb − λc − b (around 0◦), θ2 = 2λb − λc − c (around 180◦), and θ3 = b − c (around 180◦). The numerical integrations confirmed that its stability is preserved at least over 1 Gyr. The 2:1 MMR Fit III seems the most promising planetary model explaining the RV variability of the HD 240210. The 2:1 MMR is quite frequent in the sample of ∼ 40 known extrasolar systems with jovian planets, because 5 -- 6 configurations were reported (see, http://exoplanet.eu). Hence, this new system, which could be the first one around evolved star, is likely. We stress that solution III is found in relatively extended stability zone, unlike Fit II, which lies in a tiny, isolated area (∼ 0.01 au, Fig.3, the left-hand panel). These two maps almost overlap in the ac-range, hence other, relatively extended stable islands are rather excluded in this region. configuration with (cid:112)χ2 with(cid:112)χ2 2 YSC'17 Proceedings of Contributed Papers A. Rozenkiewicz, K. Goździewski & C. Migaszewski Figure 1: The RV observations of HD 240210 [5] and synthetic curves for the best fit solutions found in this paper. The left-upper panel: the best 1-planet fit. Subsequent figures are for the best fits labeled with I, II, III in Fig. 2. For a reference, all these plots are accompanied by the model curve of the 1-planet fit. See Table 1 for orbital elements. Figure 2: Statistics of the Keplerian best fits gathered with the KFIT code, projected onto the (Pb, Pc) -- plane of orbital periods. Solutions marked with white circles and labeled by I, II, III are for the best fits in different "islands" of the parameters space. Black filled circles are for all fits within formal 3σ-level of Fit I; for a reference contours with 3,2,1 σ-level of this solution, found through extensive, systematic scanning [4] of the (Pb, Pc)-plane with the Levenberg-Marquardt algorithm, are plotted as curves of increasing thickness. See Table 1 for model parameters. 3 -300-200-100 0 100 200 300Radial Velocity [m/s]HD 240210HD 240210√χ2=1.91σj=20 m/sRMS=39 m/sFit 0-100 010030003250350037504000time [MJD-50,000]5000O-C [m/s]1-planet-300-200-100 0 100 200 300Radial Velocity [m/s]HD 240210√χ2=1.35σj=20 m/sRMS=25 m/sFit I-100 010030003250350037504000time [MJD-50,000]5000O-C [m/s]2-planets-300-200-100 0 100 200 300Radial Velocity [m/s]HD 240210√χ2=1.36σj=20 m/sRMS= 26 m/sFit II-100 010030003250350037504000time [MJD-50,000]5000O-C [m/s]2-planets-300-200-100 0 100 200 300Radial Velocity [m/s]HD 240210√χ2=1.36σj=20 m/sRMS=26 m/sFit III2:1 MMR ?-100 010030003250350037504000time [MJD-50,000]5000O-C [m/s]2-planetsYSC'17 Proceedings of Contributed Papers A. Rozenkiewicz, K. Goździewski & C. Migaszewski circle): Fit II (3:2 MMR,(cid:112)χ2 ν = 1.50, rms ∼ 27 m/s, the left-hand panel), and Fit III (2:1 MMR,(cid:112)χ2 Figure 3: Dynamical maps in terms of the MEGNO indicator [4] around coplanar, edge-on, GAMP N-body fits (crossed ν = 1.36, rms ∼ 25 m/s, the right-hand panel). White color is for unstable configurations, black is for stable solutions. Orbits at the epoch of the first RV of Fit III in terms of (m [mJ], a [au], e, ω [deg],M [deg]) are the following: (4.11, 1.14, 0.284, 304.8, 101.8)b, (2.27, 1.83, 0.592, 162.3, 304.2)c for planets b, and c, respectively, V0 = 21.57 m/s, stellar mass is 0.82 M(cid:12) [5]; Fit II is: (4.50, 1.143, 0.217, 249.24, 163.87)b, (1.94, 1.506, 0.562, 53.09, 243.39)c, V0 = 8.98 m/s. Table 1: Keplerian model parameters of 2-planet best fit solutions to the RV of HD 240210. Formal measurement errors are rescaled by adding the stellar jitter of σj = 20 m/s in quadrature. T0 ≡ 53, 000 days, Nv = NRV − Np = 27. the best fit parameters P [day] K [m/s] e ω [deg] τ − T0 [days] (cid:112)χ2 ν V0[m/s] rms [m/s] I II III b 540 ± 29 131 ± 29 0.05 ± 0.09 204 ± 51 352 ± 65 c 441 ± 27 85 ± 30 0.20 ± 0.14 356 ± 73 610 ± 109 b 484 ± 14 147 ± 16 0.18 ± 0.14 261 ± 41 484 ± 45 c 667 ± 54 82 ± 36 0.74 ± 0.38 36 ± 76 502 ± 72 b 485 ± 18 129 ± 30 0.30 ± 0.13 302 ± 35 536 ± 31 c 994 ± 97 63 ± 23 0.58 ± 0.35 163 ± 87 326 ± 156 -3±7 1.35 25.2 12 ± 9 1.36 25.2 22± 12 1.36 25.2 three Keplerian solutions, which have the same (cid:112)χ2 Conclusions Extrasolar planets hosted by giant or evolved stars bring important border conditions for the planet formation theory. In this work, we re-analysed the literature RV data for evolved dwarf HD 240210, with our KFIT code relying on quasi-global GAs. In the reasonable range of orbital periods less than 3600 days, we found ν and an rms. By further dynamical analysis of these best-fit models, we selected the most likely, stable solution, which corresponds to 2:1 MMR, and is located in relatively extend zone of dynamical stability. Overall, if the 2-planet configuration is assumed, the dynamical constraints seem rule out other two models, but only new observations may confirm the 2:1 MMR hypothesis. Acknowledgements. This work is supported by Polish Ministry of Science, Grant 92/N-ASTROSIM/2008/0. References [1] Berdyugina, S. V., LRSP, 2(8) (1995) [2] Charbonneau, P., ApJS, 101, 309 (1995) [3] Goździewski K., Migaszewski, C., A&A, 449, pp.1219-1232 (2006) [4] Goździewski K., Migaszewski & C., Musieliński, A., Proc. IAU Symp. 249, pp.1219-1232 (2008) [5] Niedzielski A., Nowak G., Adamów M. & Wolszczan A., ApJ, 707, pp. 768-777 (2009) [6] Nelder J. A., Mead R., Comp. J., 7, pp. 308-313 (1965) 4
1506.01608	1	1506	2015-06-04T14:34:21	Cloud structure and composition of Jupiter's troposphere from 5-{\mu}m Cassini VIMS spectroscopy	[ "astro-ph.EP" ]	Jupiter's tropospheric composition and cloud structure are studied using Cassini VIMS 4.5-5.1 {\mu}m thermal emission spectra from the 2000-2001 flyby. We make use of both nadir and limb darkening observations on the planet's nightside, and compare these with dayside observations. Although there is significant spatial variability in the 5-{\mu}m brightness temperatures, the shape of the spectra remain very similar across the planet, suggesting the presence of a spectrally-flat, spatially inhomogeneous cloud deck. We find that a simple cloud model consisting of a single, compact cloud is able to reproduce both nightside and dayside spectra, subject to the following constraints: (i) the cloud base is located at pressures of 1.2 bar or lower; (ii) the cloud particles are highly scattering; (iii) the cloud is sufficiently spectrally flat. Using this cloud model, we search for global variability in the cloud opacity and the phosphine deep volume mixing ratio. We find that the vast majority of the 5-{\mu}m inhomogeneity can be accounted for by variations in the thickness of the cloud decks, with huge differences between the cloudy zones and the relatively cloud-free belts. The relatively low spectral resolution of VIMS limits reliable retrievals of gaseous species, but some evidence is found for an enhancement in the abundance of phosphine at high latitudes.	astro-ph.EP	astro-ph	Cloud structure and composition of Jupiter's troposphere from 5-μm Cassini VIMS spectroscopy R.S. Gilesa,∗, L.N. Fletchera, P.G.J. Irwina aAtmospheric, Oceanic & Planetary Physics, Department of Physics, University of Oxford, Clarendon Laboratory, Parks Road, Oxford, OX1 3PU, UK Abstract Jupiter's tropospheric composition and cloud structure are studied using Cassini VIMS 4.5-5.1 µm thermal emission spectra from the 2000-2001 flyby. We make use of both nadir and limb darkening observations on the planet's nightside, and compare these with dayside observations. Although there is significant spatial variability in the 5-μm brightness temperatures, the shape of the spectra remain very similar across the planet, suggesting the presence of a spectrally-flat, spatially inhomogeneous cloud deck. We find that a simple cloud model consisting of a single, compact cloud is able to reproduce both nightside and dayside spectra, subject to the following constraints: (i) the cloud base is located at pressures of 1.2 bar or lower; (ii) the cloud particles are highly scattering; (iii) the cloud is sufficiently spectrally flat. Using this cloud model, we search for global variability in the cloud opacity and the phosphine deep volume mixing ratio. We find that the vast majority of the 5-μm inhomogeneity can be accounted for by variations in the thickness of the cloud decks, with huge differences between the cloudy zones and the relatively cloud-free belts. The relatively low spectral resolution of VIMS limits reliable retrievals of gaseous species, but some evidence is found for an enhancement in the abundance of phosphine at high latitudes. Keywords: Jupiter, Atmospheres, composition, Atmospheres, structure 1. Introduction The 5-μm atmospheric window is a unique region of Jupiter's spectrum, where a dearth of opacity from the planet's principal infrared absorbers gives us access to parts of the atmosphere that are otherwise hidden from view. The sensitivity of the 4.5-5.2 µm spectrum peaks in the 4-8 bar region of Jupiter's troposphere, beneath the planet's topmost cloud decks (Figure 1). The observed brightness temperatures are therefore highly dependent on the properties of these clouds: the observed radiance varies significantly from the warm, cloud-free belts to the cooler, cloudier zones, a phenomenon first described by Westphal (1969). This sensitivity makes the 5-μm spectral region extremely useful in analysing both the composition and cloud structure of Jupiter's middle troposphere. The 5-μm data used in this study were obtained using the Visual and Infrared Mapping Spectrometer (VIMS, Brown et al., 2004) on the Cassini spacecraft. In late 2000 and early 2001, Cassini performed a fly- by of Jupiter, reaching a closest approach of 9.7 million km (137 Jovian radii) on December 30 2000. During this period, VIMS made spatially-resolved spectroscopic measurements covering the 0.35-5.1 µm wavelength range. ∗Corresponding author Email address: [email protected] (R.S. Giles) Sromovsky and Fry (2010) used this dataset to study fea- tures in reflected sunlight at 3 µm, but the 5-μm segment of this dataset has not yet been analysed. In this paper, we use 5-μm VIMS spectra with a range of phase and emis- sion angles to investigate Jupiter's tropospheric composi- tion and cloud structure; we will draw conclusions about the cloud locations and scattering parameters, and we will assess the degree of constraint offered by Cassini/VIMS on gaseous variability. This builds on previous analyses us- ing 5-μm data from Galileo NIMS and Voyager IRIS. The VIMS data gives us full global coverage of Jupiter, allow- ing us to compare and contrast the 5-μm spectra from the belts and the zones. Additionally, VIMS made observa- tions on Jupiter's nightside and dayside within the space of a few weeks, giving us the opportunity to study the effect of reflected sunlight. theoretical cloud condensation model from Atreya et al. (1997) produces four cloud levels: NH3- ice, NH4SH, H2O-ice and aqueous-ammonia solution. If a solar chemical composition is assumed, the bases of these cloud levels are located at 0.7 bar, 2.2 bar, 5.4 bar and 5.7 bar respectively. Although models such as this pro- vide valuable insights into the approximate condensation levels for each species, we do not necessarily expect to see the clouds form at these precise locations. The models make assumptions about both the bulk composition of the atmosphere and the production of photochemical prod- ucts. They also do not account for the complex dynamics A typical Preprint submitted to Icarus June 19, 2018 1 2 5 10 ) r a b ( e r u s s e r P 4.5 4.6 4.7 4.8 4.9 5.0 5.1 5.2 Wavelength (µm) 0.0 0.2 1.0 Temperature Functional Derivative (Normalised) 0.4 0.6 0.8 (a) The temperature functional derivative for a cloud- free atmosphere, normalised at each wavelength. 0.01 0.10 1.00 10.00 ) r a b ( e r u s s e r P Stratosphere Troposphere Peak Sensitivity 100 150 200 Temperature (K) 250 300 350 400 (b) Jupiter's temperature-pressure profile. Figure 1: The sensitivity of the 5-μm window as a function of wavelength. The peak sensitivity is in the 4-8 bar region of the troposphere, beneath the planet's topmost cloud decks. of the troposphere, neglecting the vertical mixing which brings up material from below and rains down material from above. All of these factors will act to significantly alter the vertical cloud structure. Observations are there- fore fundamental to understanding Jupiter's tropospheric cloud decks. In the following paragraphs, we will briefly summarise the conclusions that have been drawn from pre- vious observational studies. There are many studies supporting the existence of the uppermost NH3-ice cloud deck. 5- and 45-μm observa- tions from Voyager IRIS were analysed by Gierasch et al. (1986), who concluded that there was a highly variable cloud deck at ∼700 mbar, near the expected ammonia con- densation level. Imaging data from the Galileo space- craft showed evidence for thick, variable clouds in the 750±200 mbar region (Banfield et al., 1998), again consis- tent with NH3. Further support came in the form of spec- troscopic identification of ammonia-ice features, using the 1-3 µm region of the Galileo NIMS spectra (Baines et al., 2002) and 9 µm data from Cassini CIRS (Wong et al., 2004). Baines et al. (2002) found that their ammonia- ice spectroscopic signatures were present in less than 1% of Jupiter's clouds, but it has been suggested that this may be due to a hydrocarbon haze coating the 2 ammonia -- ice particles and masking the absorption fea- tures (Kalogerakis et al., 2008). There are additional studies which support the exis- tence of both an upper NH3 cloud and a lower cloud formed of NH4SH. Matcheva et al. (2005) found clouds in the 900-1100 mbar region from Cassini CIRS obser- vations at 7.18 µm, which they concluded were probably composed of an upper layer of NH3 and a lower layer of NH4SH. Sromovsky and Fry (2010) analysed 3-μm data from Cassini VIMS, and found that a 500 mbar cloud com- posed of both NH3 and NH4SH provided the best fit; they suggest that rapid upwelling could lead to the presence of NH4SH particles well above the expected condensa- tion point. Further evidence for a NH4SH cloud comes from analyses of the 5-μm Galileo NIMS data. Irwin et al. (2001) performed retrievals on the NIMS data and deter- mined that the cloud that provides the majority of the 5 µm opacity is located at around 1.5 bar. This was sup- ported by Irwin and Dyudina (2002), who performed a principal component analysis on the same data and once again found that the dominant opacity variations occur in the 1-2 bar region. Based on the location, both studies concluded that the dominant cloud was likely to be com- posed of ammonium hydrosulfide. Because of its predicted location deep in the lower tro- posphere, the water cloud has been difficult to observe. Banfield et al. (1998) used Galileo SSI observations to identify a deep cloud (located at a pressure greater than 4 bar) in one region close to the Great Red Spot, which they concluded was likely to be composed of water. Sim- ilarly, Simon-Miller et al. (2000) found evidence for the cloud in very small regions of the planet by identifying a water ice feature near 44 µm in Voyager IRIS spectra. This limited detection may be due to the fact that the water cloud is only visible in small regions where the rest of the atmosphere is particularly cloud-free. In addition to studying the cloud structure, the 5-μm region allows us to retrieve abundances for tropospheric species and determine if there is any global variability. Disequlibrium species, such as PH3, are expected to have higher mixing ratios in regions of upwelling, as the gas is brought up from deeper pressures where it is more abun- dant. Irwin et al. (2004) and Fletcher et al. (2009) find that Cassini CIRS data shows a PH3 enhancement at the equator compared to the belts on either side, supporting this view. Similarly, zones are expected to have a higher water abundance than belts, as moist air is brought up from below; Carlson et al. (1992), using Voyager IRIS data, found that low relative humidities were correlated with bright 5-μm regions, so the bright belts were found to be depleted in water. In this paper, we investigate the constraints placed on Jupiter's tropospheric clouds by 5-μm VIMS data, and search for any spatial variability in Jupiter's tropospheric gaseous species. We find that the VIMS observations can be modelled using a single, highly reflecting cloud layer lo- cated at a pressure of 1.2 bar or lower, and that this cloud is responsible for the vast majority of the global spectral variability. The relatively low spectral resolution of VIMS limits the gaseous retrievals, but we find some evidence for an enhancement in the abundance phosphine at high latitudes. 2. Observations 2.1. VIMS data and calibration Between October 2000 and March 2001, tens of thou- sands of images of the Jovian system were taken by the Cassini spacecraft. These included 5μm measurements taken by the Visual and Infrared Mapping Spectrome- ter (VIMS, Brown et al., 2004). VIMS is made up of two imaging spectrometers, one covering the visible spec- tral range (VIMS-V) and the other covering the infrared (VIMS-IR). The present study uses data from the 4.5- 5.1 µm region from VIMS-IR. VIMS-IR covers the wavelength range 0.85-5.1 µm with a spectral sampling of 16.6 nm, giving 256 wavelength channels. The spectral resolution varies slightly across the wavelength range, and has an average value of 18.7 nm in the 5-μm window (K. Baines, personal communica- tion). The instantaneous field of view of VIMS-IR is 0.25×0.50 mrad. By using a two-axis scan mechanism, the instantaneous field of view is stepped across the scene to scan a full 64×64 pixel image. The total field of view is 32×32 mrad, which is several times larger than the size of Jupiter's disk at the distance of closest approach (Jupiter has a diameter of 15 mrad at distance of 9.7 million km). The VIMS cubes were processed using ISIS3 (Integrated Software for Imagers and Spectrometers), a software pack- age provided by the USGS (Anderson et al., 2004). The VIMS Science Team calibration (McCord et al., 2004) is included in this software and geometries are assigned to the cubes using NASA/JPL SPICE kernels, so both radio- metric and geometric calibration are achieved. The final result is a data cube of dimensions 64×64×256 (one spec- tral and two spatial dimensions) and a backplane cube of dimensions 64×64×13, including information such as the latitude, longitude, phase angle and emission angle of each spatial point. 2.2. Data selection VIMS observations were made both on the dayside of the planet as the spacecraft approached in late 2000 and on the planet's nightside as the spacecraft moved away from Jupiter in early 2001. Nightside observations simply show the thermal emission from the planet, whereas the dayside observations also include a component from reflected sun- light which further complicates the analysis; Sections 5, 6 and 7 focus on the nightside data, and Section 8 compares these results to the dayside observations. For the nightside, a set of 30 data cubes from a 7 hour period on 11 January 2001 were chosen, when Cassini was e d u t i t a L 40 20 0 -20 -40 360 315 270 225 180 135 90 System III West Longitude 45 0 165 180 195 210 T(K) 225 240 (a) Nightside data from 11 January 2001. e d u t i t a L 40 20 0 -20 -40 360 315 270 225 180 135 90 System III West Longitude 45 0 NEB SEB . NEB SEB 165 180 195 210 T(K) 225 240 (b) Dayside data from 17 December 2000. Figure 2: Brightness temperature maps from 5.0 µm VIMS data on two dates, giving both nightside and dayside ob- servations of Jupiter. Smoothing has been applied to these maps. Black regions correspond to segments where no data was obtained. The GRS is located at ∼ 50◦W and did not move significantly between the two datasets. a distance of 15 million km from Jupiter. The use of mul- tiple cubes vastly increases the number of available data points and allows us to cover more longitudes, while the fairly limited time period means that atmospheric condi- tions are unlikely to have changed. The date was chosen so as to avoid the day-night boundary (to obtain purely nightside emission), while maximising the spatial resolu- tion. For the subsequent dayside analysis, a set of data cubes from a 10 hour period on 17 December 2000 were used. These observations were made at a similar distance from the planet as the 11 January ones (16 million km), giving a comparable spatial resolution of ∼8000 km (∼ half the size of the GRS). Cylindrical projections of both the nightside and the dayside datasets are shown in Figure 2. In both panels, the difference between the warm belts and the cooler zones can be seen. In the nightside observations, there is a ∼70 K brightness temperature difference between the warmest and the coolest regions of the planet. For the dayside ob- servations, this contrast is reduced to ∼50 K; the cloudy zones reflect more light than the relatively cloud-free belts, meaning that the presence of sunlight has a greater impact on the cool zones than on the warm belts. Zonal mean ra- diances were then extracted from both the nightside and the dayside datasets. The spectra were divided into forty- one latitudinal bins covering the range −50◦ to +50◦ at 3 2.5◦ intervals, each with a 5◦ width. For each latitudinal bin, all spectra with an emission angle within 10◦ of the minimum value were averaged to produce one final nadir spectrum for each latitude segment. Following Fletcher et al. (2011), the final errors as- signed to these averaged VIMS spectra were 12% of the mean radiance in the 4.5-5.2 µm range. This is a conserva- tive value, including both quadrature-estimated errors due to pre-flight calibration and forward-model uncertainties on spectral line data. The constant error margin through- out the window prevents retrievals being weighted towards low-radiance regions. The size of the VIMS errors is fur- ther discussed in Section 7. The results of this process for the nightside data can be seen in Figure 3. Figure 3b shows the forty-one spectra from the different latitudes, all normalised to 1.0 to allow a comparison of spectral shapes, while Figure 3a shows the absolute radiances at 5.0 µm as a function of latitude. While Figure 3a shows that there is a factor of 50 differ- ence between the hottest and coldest spectra, Figure 3b shows that the shape of the spectra remains almost identi- cal throughout. This cannot be explained by variations in the chemical composition or temperature profile of the at- mosphere. The only parameter capable of reproducing the huge variability in the radiance while retaining the same spectral shape is the cloud optical thickness, as we show quantitatively in the following sections. 3. Spectral modelling 3.1. Radiative transfer model and retrieval algorithm The Jupiter VIMS spectra were analysed using the NEMESIS software developed by Irwin et al. (2008). This software has previously been used to analyse the 5- μm VIMS spectra of Saturn (Fletcher et al., 2011) and an earlier version was used to analyse Jupiter NIMS data (Irwin et al., 1998, 2001; Nixon et al., 2001). NEME- SIS is made up of a radiative transfer code that computes the emergent radiation for a given atmospheric profile and an optimal estimation retrieval algorithm which iteratively determines the best-fit atmospheric parameters for an ob- served spectrum. The radiative transfer code solves the equation of ra- diative transfer to calculate the top-of-atmosphere radi- ation as a function of wavelength. The input informa- tion required includes a priori estimates of the atmo- spheric composition and the temperature profile, as well as absorption line information for each molecule. Rather than computing each individual absorption line, NEME- SIS uses the correlated-k approximation (Lacis and Oinas, 1991) to reduce the computational time. This ap- proximation allows us to reshuffle the absorption coef- ficients within a small spectral into ascending order, producing a smoothly varying function that re- quires fewer steps in numerical integration (Irwin et al., 2008). NEMESIS has the capability to model a multiple- scattering atmosphere. The full Mie-scattering phase interval 4 ) 1 − m µ 1 − r s 2 − m c W µ ( e c n a d a R i 20 15 10 5 0 −60 STrZ SEB EZ NEB NTrZ −40 −20 0 Latitude 20 40 60 . (a) 5.0 µm radiance as a function of latitude. ) 1 − m µ 1 − r s 2 − m c W µ ( e c n a d a R i 2.0 1.5 1.0 0.5 0.0 4.5 4.6 4.7 4.8 Wavelength (µm) 4.9 5.0 5.1 (b) The normalised spectrum for each latitudinal bin. Figure 3: VIMS nightside spectra, divided into 5◦ wide latitudinal bins and zonally averaged. The upper panel shows how the absolute radiance at 5.0 µm varies with lat- itude. The lower panel shows the full spectrum for each latitudinal bin, normalised to allow a comparison of their shapes. function is approximated using a combined Henyey- Greenstein function (Henyey and Greenstein, 1941) and the calculations are performed using a matrix-operator ap- proach (Plass et al., 1973). The NEMESIS retrieval algorithm determines the at- mospheric parameters from an observed input spectrum through an iterative process. Starting with an initial a priori atmospheric profile, synthetic "forward-models" are generated using the radiative transfer code. These syn- thetic spectra are compared to the observed spectrum and the quality of the fit is assessed by using a cost function comprised of two terms, the residual fit to the data and the deviation from the a priori atmospheric state (each weighted by their respective uncertainties). This cost func- tion is minimised using a Levenburg-Marquardt iterative scheme (Irwin et al., 2008). 3.2. Reference atmosphere For this study, the Jovian atmosphere was divided into 39 levels between 15 bar and 50 mbar, equally spaced in log(p). The temperature and volume mixing ratio of each gas species is defined at each pressure level. The temperature profile is taken from Cassini CIRS observations (Fletcher et al., 2009). The CIRS temper- ature profile is sensitive down to a pressure of 800 mbar, and has been extrapolated below this using a dry adia- bat, consistent with the temperature profile found by the Galileo probe (Seiff et al., 1998) (see Figure 1b). The NH3 and PH3 profiles are also taken from Fletcher et al. (2009) and have deep volume mixing ratios of 1.862 × 10−4 and 1.86 × 10−6 respectively. The CH4 profile was obtained from Nixon et al. (2007) and has a deep volume mixing ratio of 1.81 × 10−3. The CH3D profile assumes a constant ratio of CH3D/CH4 of 8 × 10−5 (Lellouch et al., 2001). The profile of H2O is poorly constrained. For this study, we assume a deep vol- ume mixing ratio, fixed at 1 × 10−3 (∼ the solar abun- dance) with a constant relative humidity at higher alti- tudes. For the reference profile, this relative humidity is set to 10%. The remaining minor atmospheric species, CO, GeH4 and AsH3 were assumed to be well mixed with volume mixing ratios of 1.0 ppb, 0.45 ppb and 0.24 ppb respectively (B´ezard et al., 2002). 3.3. Line data The line data sources that are used to produce the k- tables used in this study are described in Fletcher et al. (2011), along with the assumed broadening parameters and temperature dependences. This was updated to in- clude information about additional GeH4 isotopes ob- tained using STDS (Wenger and Champion, 1998). As with the original GeH4 isotope included in Fletcher et al. (2011), these isotopes were assumed to have a half-width of 0.1 cm−1atm-1 and a temperature dependence of T0.75. As with Fletcher et al. (2011), lines were assumed to have a Voigt profile. The exception to this was NH3, which was updated to have a significantly sub-Lorentzian profile in the 5-μm window, as found by Bailly et al. (2004). 4. Sensitivity analysis The radiative transfer model described in Section 3.1 was used to perform a sensitivity analysis, where each pa- rameter was individually altered in order to observe its effect on the spectrum. Initially, a reference spectrum was produced using the atmospheric profile described in Sec- tion 3.2. For this sensitivity analysis, a single, spectrally- flat cloud layer was inserted at 0.8 bar with an optical thickness of 10. The atmospheric paramaters were then individually altered, and the new spectra were compared to the reference spectrum. For the molecular species, the entire a priori profile was scaled by factors of 0.5 and 2.0. For water, the deep volume mixing ratio was held constant at the solar abundance and the relative humidity was var- ied by factors of 0.5 and 2.0. The cloud opacity was varied by factors of 0.8 and 1.2, and the temperature profile was shifted by ±20 K. The results of this process can be seen in Figure 4, where the initial reference spetrum is shown, alongside nine frames showing the difference between the altered profiles and the reference. The gaseous species with the most significant impact on the shape of the VIMS-resolution 5-μm spectrum are PH3 and H2O. Although NH3, CH3D, GeH4, AsH3 and CO all have spectral lines in this region, the relatively low spectral resolution means that they have a minor impact on the VIMS spectra. The single parameter that has the largest impact on the spectrum is the opacity of the cloud. Because the model cloud is spectrally flat and located well above the weight- ing function peak, the entire wavelength range is affected equally; changing the optical thickness scales the entire spectrum up or down, and a relatively minor change in the opacity has a significant impact on the average radi- ance. Shifting the temperature profile similarly scales the entire spectrum up or down, but to a lesser extent; to re- produce an effect which is comparable to a 20% change of in cloud opacity, the temperature shift would have to be as large as 20 K. A more realistic tropospheric temperature difference 2 K between a moist and a dry adiabat (Lewis, 1995) produces a negligible effect on the spectrum when compared to the variability in cloud opacity. We therefore fix the temperature profile to the dry adiabat used in the reference profile, and note that this may produce a small error in the retrieved cloud opacities. In addition to the degeneracy between cloud and tem- perature, there are further degeneracies between gaseous species which complicate 5-μm retrievals at this spectral resolution. The regions of sensitivity for each parameter are not unique, but instead overlap, leading to difficulties in disentangling the effects of each individual parameter to produce a reliable retrieval. This is further illustrated by Figure 5, which shows the results of a retrievability test. 5 ) 1 − m µ 1 − r s 2 − m c W µ ( e c n a d a R i Reference Spectrum 8 6 4 2 0 4.5 4.6 4.7 4.8 4.9 5.0 5.1 Wavelength (µm) ) 1 − m µ 1 − r s 2 − m c W µ ( e c n a d a R i ) 1 − m µ 1 − r s 2 − m c W µ ( e c n a d a R i Cloud 6 4 2 0 −2 −4 −6 4.5 4.6 4.7 4.8 4.9 5.0 5.1 Wavelength (µm) H2O 2 1 0 −1 −2 4.5 4.6 4.7 4.8 4.9 5.0 5.1 Wavelength (µm) ) 1 − m µ 1 − r s 2 − m c W µ ( e c n a d a R i ) 1 − m µ 1 − r s 2 − m c W µ ( e c n a d a R i CH3D 0.6 0.4 0.2 −0.0 −0.2 −0.4 −0.6 4.5 4.6 4.7 4.8 4.9 5.0 5.1 Wavelength (µm) AsH3 0.6 0.4 0.2 −0.0 −0.2 −0.4 −0.6 4.5 4.6 4.7 4.8 4.9 5.0 5.1 Wavelength (µm) ) 1 − m µ 1 − r s 2 − m c W µ ( e c n a d a R i ) 1 − m µ 1 − r s 2 − m c W µ ( e c n a d a R i ) 1 − m µ 1 − r s 2 − m c W µ ( e c n a d a R i 0.6 0.4 0.2 −0.0 −0.2 −0.4 −0.6 ) 1 − m µ 1 − r s 2 − m c W µ ( e c n a d a R i ) 1 − m µ 1 − r s 2 − m c W µ ( e c n a d a R i Temperature 5 0 −5 4.5 4.6 4.7 4.8 4.9 5.0 5.1 Wavelength (µm) PH3 4 2 0 −2 −4 4.5 4.6 4.7 4.8 4.9 5.0 5.1 Wavelength (µm) NH3 0.6 0.4 0.2 −0.0 −0.2 −0.4 −0.6 4.5 4.6 4.7 4.8 4.9 5.0 5.1 Wavelength (µm) 4.5 4.6 4.7 4.8 4.9 5.0 5.1 Wavelength (µm) CO 0.6 0.4 0.2 −0.0 −0.2 −0.4 −0.6 4.5 4.6 4.7 4.8 4.9 5.0 5.1 Wavelength (µm) Figure 4: The sensitivity of VIMS-resolution spectra to changes in different atmospheric parameters. The refer- ence spectrum shows a synthetic spectrum computed for the atmospheric profile described in Section 3.2, and the remaining panels show the (relative) changes to this spec- trum that are caused by varying each parameter. The red (blue) lines correspond to a decrease (increase) of 20 K for the temperature, a factor of 20% decrease (increase) for the cloud opacity and a factor of 2 decrease (increase) for the gaseous species. GeH4 5. Basic retrievals 75 radiative transfer models were run with random pertur- bations applied to the reference profile. Random noise was then added to these synthetic spectra, in accordance with the 12% noise estimate for VIMS. The noisy spectra were then used as the input for NEMESIS retrievals. Figure 5 compares output was compared with the 'true' input val- ues, and provides the Pearson product-moment correlation coefficient for each parameter. A correlation coeffiect of 1 indicates a perfect positive correlation, while a coefficient of 0 indicates that there is no linear relationship between the two variables. The results tally with Figure 4; changes in cloud cover and the PH3 volume mixing ratio have the most significant impact on the spectrum and can therefore be accurately retrieved, with Pearson product-moment correlation coef- ficients that are close to 1. In contrast, the retrievals for the other gaseous species are poor, and the retrieved val- ues are only weakly correlated with the 'true' input values. Based on these results, the abundances of NH3, CH3D, GeH4, AsH3 and CO will be held fixed at their a priori values during the remainder of this study. Although the retrievals of H2O are fairly poor, we have no reliable a priori profile for water vapour, so the relative humidity of water will be allowed to vary, along with the volume mixing ratio of PH3 and the cloud optical thickness. Having explored the sensitivity of the 5-μm region to different parameters and the potential limitations of any retrievals, we now turn to modelling the VIMS data itself using the NEMESIS retrieval algorithm. As with the sen- sitivity analysis, PH3 and H2O were each allowed to vary via a single parameter. For PH3 this was a scaling param- eter that effectively altered the deep volume mixing ratio, and for H2O this was the relative humidity above a fixed deep volume mixing ratio. For each variable, large errors were assigned to a priori values. Testing was carried out to determine whether allowing additional degrees of freedom in these gases would improve the fit, but this did not make a difference in either case. As an initial model, we use a single, compact, 5-km thick cloud, located at 0.8 bar. The cloud base pres- sure is well above the weighting function peak (4-8 bar) and is the approximate location of the predicted NH3- ice cloud. Figure 3 shows that the shape of the spec- tra are almost identical at all latitudes, despite large variations in the absolute radiance. The cloud param- eters therefore cannot vary strongly with wavelength, as otherwise any spectral features would be more ev- ident in the cool, optically thick spectra than in the warm cloud-free spectra. In this initial model, we there- fore use a spectrally flat cloud, as has been previ- ously suggested by both ground-based and space-based studies, including Drossart et al. (1982), B´ezard et al. (1983), Bjoraker et al. (1986) and Roos-Serote and Irwin (2006). The scattering parameters for this cloud were fixed 6 t u p t u O Cloud: 0.98 16 16 14 14 12 12 10 10 8 8 6 6 4 4 4 6 8 10 12 14 16 4 6 8 10 12 14 16 Input CH3D: 0.56 t t u p u O 2.0 1.5 1.0 0.5 0.0 H2O: 0.22 PH3: 0.96 2.0 1.5 1.0 0.5 0.0 t u p t u O 0.20 0.15 0.10 0.05 0.00 t u p t u O 2.0 1.5 1.0 0.5 0.0 NH3: 0.50 0.0 0.5 1.0 1.5 2.0 Input 0.00 0.05 0.10 0.15 0.20 Input 0.0 0.5 1.0 1.5 2.0 Input AsH3: 0.32 GeH4: 0.39 2.0 1.5 1.0 0.5 0.0 t t u p u O 2.0 1.5 1.0 0.5 0.0 CO: 0.18 t t u p u O 2.0 1.5 1.0 0.5 0.0 t u p t u O t t u p u O 0.0 0.5 1.0 1.5 2.0 Input 0.0 0.5 1.0 1.5 2.0 Input 0.0 0.5 1.0 1.5 2.0 Input 0.0 0.5 1.0 1.5 2.0 Input Figure 5: The retrievability of different atmospheric paramaters for VIMS-resolution spectra. In each case, the true 'input' value is plotted against the retrieved 'output' value. For H2O, the input and output values correspond to the relative humidity, for the cloud, they correspond to cloud opacity and for the remaining gaseous species, they correspond to the scaling factors of the a priori profile. A perfect retrieval would result in the plotted points lying along the diagonal line. The Pearson product-moment correlation coefficients are given in the titles. ) 1 − m µ 1 − r s 2 − m c W µ ( e c n a d a R i 0.4 0.3 0.2 0.1 0.0 4.5 4.6 4.7 Data Retrieval 4.8 Wavelength (µm) 4.9 5.0 5.1 Figure 6: Fit obtained using nightside data from the equa- torial zone. The VIMS data is shown in black (with the error shown in grey) alongside the best-fit retrieved spec- trum in red. at values that are broadly representative of moderately- sized NH3 particles: single-scattering albedo, ω = 0.9 and asymmetry parameter, g = 0.8. An example of the results of these retrievals can be seen in Figure 6; we are able to produce a good fit to the VIMS 5-μm data despite using a very simple cloud model and relatively few free parameters. The one excep- tion to the otherwise good fit is the 4.65-4.75 µm region of the spectrum, where there is an apparent offset be- tween the VIMS data and the retrieved spectrum. Many attempts were made to solve this issue, including allow- ing additional gaseous species to be retrieved, altering the parametrisations for the gases (including PH3), inserting multiple cloud decks and varying the vertical cloud pro- 7 file. Similar issues can be seen in previous 5-μm studies of Jupiter using Galileo/NIMS, including Roos-Serote et al. (1998) and Nixon et al. (2001). Roos-Serote et al. (1998) also had some difficulties in fitting a similar region of the high-resolution ISO/SWS spectra. The consistency of the problem suggests that the problem is with the models rather than the data. The mismatch is similar for both the cool and the warm spectra, suggesting that it is not due to a cloud spectral feature, and is more likely to be due to missing spectral lines from a gaseous species. As with Roos-Serote et al. (1998), we exclude this part of the spectrum to ensure that the retrievals are not driven by an attempt to fit this one region. 6. Cloud constraints Instead of making assumptions about the composition of Jupiter's clouds, we sought to determine the range of cloud parameters that were consistent with the VIMS data. In Section 5, we showed that a simple cloud model, consisting of a single, compact, spectrally flat cloud deck, is able to achieve a good fit to data. In this section, we further explore the cloud parameter space in order to constrain the cloud properties. 6.1. Vertical structure Preliminary conclusions about the vertical location of the tropospheric cloud decks can be drawn by consider- ing the VIMS spectra alone. As described in Section 2, Figure 3 shows that the shape of the spectra are almost identical at all latitudes, despite large variations in the ab- solute radiance. This immediately tells us that the main ) 1 − m µ 1 − r s 2 − m c W µ ( e c n a d a R i 0.8 bar 8 6 4 2 0 4.5 4.6 4.7 4.8 4.9 5.0 5.1 Wavelength (µm) ) 1 − m µ 1 − r s 2 − m c W µ ( e c n a d a R i 2.0 bar 8 6 4 2 0 4.5 4.6 4.7 4.8 4.9 5.0 5.1 Wavelength (µm) Optical Thickness 10 12 14 16 18 20 Figure 7: Sensitivity of spectra to cloud locations. Each panel shows the spectrum for a variety of different cloud optical thicknesses. The upper panel corresponds to a compact cloud located at 0.8 bar and the lower panel corresponds to a compact cloud located at 2.0 bar. cloud decks must be located well above the 4-8 bar region where the weighting functions peak (see Figure 1); oth- erwise, thick clouds would block out the radiation from some parts of spectrum more than others, giving a different shape to to the 5-μm spectrum depending on the bright- ness temperature. Additionally, having an optically thick cloud at a deeper, hotter part of the atmosphere will also start to introduce a blackbody slope into the spectrum, which is not observed in any of the cool spectra. These effects can be seen in Figure 7, which shows the results of inserting a cloud at two different locations in Jupiter's tro- posphere and gradually increasing the optical thickness. In the upper panel, the cloud is located relatively high up in Jupiter's atmosphere, and increasing the optical thickness decreases the observed radiance, but maintains the same spectral shape, just as is observed in the VIMS spectra. In comparison, the lower panel shows the results for a cloud that has been placed deeper in the atmosphere, at 2.0 bar. Although the spectral shape remains similar for the lower opacities, the spectral shape starts to change for high op- tical thicknesses. In addition to changing the shape of the spectrum, this second case simply cannot reproduce the low average radiances observed in the cool zones in Fig- ure 3; at these deeper pressures, the thermal emission from the cloud itself is higher and even an optically thick cloud therefore cannot reproduce the observed low radiances. The constraints on the vertical location of the clouds were further explored by running retrievals with a wide range of base pressures for the single, compact cloud. This was done for both a cool spectrum from the EZ and a warm spectrum from the NEB, and the results for two of the al- titudes tested are shown in Figure 8. When the cloud is located at 0.8 bar, good fits are obtained for both the EZ and the NEB. As the cloud is moved to progressively deeper pressures, the NEB fits initially remains good, but the EZ fits worsens significantly. The cut-off for a reason- able EZ fit is approximately 1.2 bar, while the NEB fits start to deteriorate for cloud base pressures of 3.0 bar or higher. The poor fit at deeper pressure levels is a result of the phenomena shown in Figure 7; as the cloud is moved deeper into the atmosphere, it starts to introduce a slope into the spectrum that is not seen in the data. In order to achieve a good fit for both warm and cool spectra us- ing a single cloud deck, the clouds must be located at an altitude of 1.2 bar or higher. In addition to exploring the location of the primary cloud opacity, testing was also carried out to investigate the vertical profile of the cloud. Additional cloud decks, with independently variable opacities, were included in the retrievals, but this made very little difference to the qual- ity of the fits obtained. Similarly, extended cloud decks were used instead of vertically compact clouds, but this did not improve the fit either. In summary, the VIMS 5-μm nightside data can be fit using a simple cloud model, consisting of a single, compact, spectrally flat cloud deck whose opacity varies as a func- tion of latitude, provided that this cloud deck is located at pressures less than 1.2 bar. It is important to note that this does not rule out more complicated cloud structures, including multiple cloud layers above 1.2 bar. The exis- tence of deeper cloud decks is also possible, provided that the bulk of the 5-μm opacity still originates from the upper clouds. Similar analyses have previously been performed using observations from Galileo NIMS which cover the same wavelength range. Roos-Serote and Irwin (2006) used a grey cloud model to fit the NIMS data, and found it had to be located above 2.0 bar. Our study provides a tighter constraint (p<2.0 bar) because the global VIMS data includes spectra from very cold regions, while the NIMS study was restricted to warmer regions. It is the cooler spectra that provide the stronger constraint on the cloud location. Other NIMS studies have this same dis- crepency; Irwin et al. (2001), Nixon et al. (2001) and Irwin and Dyudina (2002) all place their principal cloud decks at pressures of 1-2 bar, but since they only use warmer spectra, they are able to place the the clouds deeper in the atmosphere and still achieve good fits. The 5-μm VIMS results are consistent with Cassini ob- servations made at other wavelengths. Sromovsky and Fry (2010) analysed the 3-μm segment of the VIMS data. Us- ing their four-layer multiple-scattering model, they found 8 ) 1 − m µ 1 − r s 2 − m c W µ ( e c n a d a R i ) 1 − m µ 1 − r s 2 − m c W µ ( e c n a d a R i EZ: 0.8 bar 0.4 0.3 0.2 0.1 0.0 4.5 4.6 4.7 4.8 4.9 5.0 5.1 Wavelength (µm) NEB: 0.8 bar 20 15 10 5 0 4.5 4.6 4.7 4.8 4.9 5.0 5.1 Wavelength (µm) ) 1 − m µ 1 − r s 2 − m c W µ ( e c n a d a R i ) 1 − m µ 1 − r s 2 − m c W µ ( e c n a d a R i EZ: 1.1 bar 0.4 0.3 0.2 0.1 0.0 4.5 4.6 4.7 4.8 4.9 5.0 5.1 Wavelength (µm) NEB: 1.1 bar 20 15 10 5 0 4.5 4.6 4.7 4.8 4.9 5.0 5.1 Wavelength (µm) Data Retrieval ) 1 − m µ 1 − r s 2 − m c W µ ( e c n a d a R i ) 1 − m µ 1 − r s 2 − m c W µ ( e c n a d a R i EZ: 1.3 bar 0.4 0.3 0.2 0.1 0.0 4.5 4.6 4.7 4.8 4.9 5.0 5.1 Wavelength (µm) NEB: 1.3 bar 20 15 10 5 0 4.5 4.6 4.7 4.8 4.9 5.0 5.1 Wavelength (µm) Figure 8: Retrievals of VIMS spectra, with different assumptions about cloud location. Retrievals from the cool Equa- torial Zone are shown at the top and retrievals from the warm North Equatorial Belt are on the bottom. Each column corresponds to a different cloud base location. that the deepest, highest opacity cloud had a spatially vari- able cloud base that was located between 0.79 and 1.27 bar. Using a narrow spectral window from the CIRS instrument centered on 7.18 µm, Matcheva et al. (2005) found that the cloud absorption coefficient peaks at 0.9-1.1 bar. These lo- cations are broadly consistent with the VIMS data, but we can neither confirm nor rule out the latitudinal variabil- ity in the cloud base pressures; we can achieve a good fit with a single cloud base pressure of 1.2 bar or lower, but a range of pressures is also possible, provided the cloud is located above 3.0 bar for the warm spectra from the belts and above 1.2 bar for the cool spectra from the zones. 6.2. Spectral features As described in Section 5, the similar spectral shapes in Figure 3 suggest that the clouds providing the majority of the 5-μm opacity must be relatively spectrally flat, at least at the VIMS spectral resolution. Having established that a completely grey cloud is able to produce a good fit to the VIMS data, we now explore the extent to which spectral features are compatible which the observations. The two expected cloud materials in the middle tropo- sphere are NH3-ice and NH4SH-ice. Figure 9 shows the extinction cross-section and single-scattering albedo as a function of wavelength for a range of particle sizes. These functions have been calculated through Mie theory, as- suming perfectly spherical particles. From this figure, it is clear that the scattering and absorption parameters of pure particles of NH3 and NH4SH vary significantly with wavelength within the 5-μm window. Larger particles have a relatively flat extinction cross-section, but a strongly 9 ) d e s i l a m r o n ( n o i t c e s − s s o r C ω , l o d e b a g n i r e t t a c s − e g n S i l Cross−section 1.4 1.3 1.2 1.1 1.0 0.9 0.8 4.5 4.6 4.7 4.8 4.9 5.0 Wavelength (µm) Single−scattering albedo 1.0 0.9 0.8 0.7 0.6 0.5 4.5 4.6 4.7 4.8 4.9 5.0 Wavelength (µm) 5.1 5.1 1µm NH3 10µm NH3 40µm NH3 1µm NH4SH 10µm NH4SH 40µm NH4SH Figure 9: Spectral parameters for NH4SH and NH3 parti- cles of different sizes, calculated via Mie theory. Equatorial Zone ) 1 − m µ 1 − r s 2 − m c W µ ( e c n a d a R i 0.6 0.5 0.4 0.3 0.2 0.1 0.0 4.5 4.6 4.7 4.8 4.9 5.0 Wavelength (µm) ) 1 − m µ 1 − r s 2 − m c W µ ( e c n a d a R i 30 25 20 15 10 5 0 4.5 North Equatorial Belt 4.6 4.7 4.8 4.9 5.0 Wavelength (µm) 5.1 5.1 1µm NH3 10µm NH3 40µm NH3 1µm NH4SH 10µm NH4SH 40µm NH4SH Grey Cloud Figure 10: The effect of different cloud particles on the shape of the spectra. The black line in each case gives the fit obtained using a spectrally flat cloud. The spectral fea- tures are more pronounced in the Equatorial Zone, where the clouds are optically thick. wavelength-dependent single scattering albedo, while this is reversed for smaller particles. The effect of the wavelength-dependent cloud parame- ters on the spectral shape can be seen in Figure 10. In each panel, the black line shows the model fit to the VIMS data that is obtained using a spectrally flat cloud. The remaining lines show the effect of using a "real" cloud ma- terial, rather than a grey cloud. To make the impact of the clouds clear, the atmospheric composition (H2O and PH3) has been held constant for each cloud material, and the cloud opacity alone has been varied in order to produce an average radiance as close to the true value as possible. Performing a full retrieval does not however improve the fit sufficiently, and in some cases the algorithm attempts to retrieve an unphysically high abundance of the gaseous species in order to compensate for the poor cloud fit. Figure 10 clearly shows that these "real" clouds are un- able to reproduce the shape of the VIMS spectra. This is particularly the case for Equatorial Zone, where the clouds are optically thick and any spectral features therefore become more prominent. The varying single-scattering albedo is the primary cause of these poor fits; the sharp gradient at ∼4.65 µm seen in Figure 9 for the NH4SH par- ticles translates into the sharp gradient seen in Figure 10, while the gradual slope for the NH3 produces a slope in the spectra. 10 Despite being the most plausible materials for the thick tropospheric clouds, pure NH3-ice and pure NH4SH-ice are not consistent with the VIMS 5-μm observations. How- ever, there are several possible explanations for this. The presence of several cloud decks made up of particles of different materials and/or sizes can act together to "blur out" the individual spectral features, giving a net effect of a roughly grey cloud. This effect, along with the fact that the study focussed on warmer regions where the clouds are thinner, allowed Irwin et al. (1998) to fit the Galileo NIMS 5-μm spectra using a 4-level cloud structure, made up of 0.5 µm tholins, 0.75 µm NH3 particles, and 0.45 µm and 50 µm NH4SH particles. Alternatively, the clouds could be made of a different material altogether, or be masked by deposits from an- other material which causes the cloud to be spectrally flat. Coating of ammonia particles by other substances has been suggested by Baines et al. (2002) as an explanation for the absence of spectroscopically identifiable ammonia clouds in across the majority of the planet and Kalogerakis et al. (2008) found that thin layers of hydrocarbons are able to alter the spectral features at 3 and 9 µm. The 5-μm data alone cannot distinguish between these various possible scenarios, but the net effect in each case is a roughly grey cloud. We choose the simplest option for reproducing this, and continue the analysis using a single spectrally-flat cloud. 6.3. Scattering properties The choice of scattering parameters affects the thermal scattering on the planet's nightside. We now explore the range of scattering parameters that are consistent with the VIMS nightside observations. Using the nightside nadir data described in Section 2.2 and used in Sections 6.1 and 6.2, retrievals were performed using many values of ω and g, covering the full range of physically realistic par- ticles. Although the numerical results of the retrievals var- ied with these different values (a higher single-scattering albedo requires a higher cloud optical thickness for the same spectrum), the fits produced were very uniform. Many different cloud parameters were capable of repro- ducing the nadir VIMS spectra, so the values of ω and g cannot be constrained from the nadir data alone. Additional insights can be gained by considering obser- vations made at a range of emission angles. In the Jovian troposphere, temperature decreases with altitude, leading to limb darkening: the radiance observed decreases to- wards the edge of the disc. The extent of this limb dark- ening is heavily dependent on the cloud decks that are present in the troposphere; by comparing the spectra at different emission angles, we can constrain the scattering parameters of these clouds. The relatively low spatial resolution of the VIMS data means that it is not possible to isolate a single atmospheric feature and compare its appearance at different emission angles. Instead, the scatter plots in Figure 11 show the limb darkening observed in the equatorial region of the ) 1 − m µ 1 − r s 2 − m c W µ ( e c n a d a R i 0.5 0.4 0.3 0.2 0.1 0.0 0 20 ω=0.9, g=0.7 ω=0.8, g=0.8 ω=0.7, g=0.9 40 60 Emission Angle g t , r e e m a r a p y r t e m m y s A 0.9 0.9 0.9 0.8 0.8 0.8 0.7 0.7 0.7 0.6 0.6 0.6 0.5 0.5 0.5 0.5 0.5 0.5 80 100 0.6 0.6 0.6 0.7 0.7 0.7 Single−scattering albedo, ω 0.8 0.8 0.8 0.9 0.9 0.9 Figure 11: Limb darkening for different cloud parameters. The scatter plots show the VIMS data as a function of emission angle, averaged over the 5-μm spectral window. The large spread is due to the spatial inhomogeneity of the clouds, even within a single latitude band. Overplotted are the synthetic limb darkening curves for three example sets of cloud parameters (see Figure 12 for the full range of parameters tested). planet (−2.5◦ to 2.5◦). The equatorial region was chosen as it provides the largest range of emission angles, and exhibits less spatial inhomogeneity than other parts of the planet. Each point corresponds to the mean 4.5-5.2 µm radiance from a single pixel. The large spread of points is due to the inhomogeneity of the cloud thickness, even within a single latitudinal band (as seen in Figure 2). Nevertheless, the general limb dark- ening trend can be seen - as the emission angles increases towards the limb, the radiance decreases. The broad scatter of points in Figure 11 means that a simultaneous retrieval of data at different emission angles is unlikely to be productive. Two points at different posi- tions along the disc are likely to have very different cloud opacities, in addition to having different emission angles. However forward modelling does allow us to place some constraints on the cloud parameters that are consistent with the observed data. For a range of cloud parameter combinations (single- scattering albedo, ω and asymmetry parameter, g), re- trievals were performed using nadir (low-emission angle) data. The retrieved parameters from the nadir data were then used to forward-model spectra corresponding to higher emission angles. The results of this process for three example combinations are shown by the solid lines in Figure 11. Each limb darkening curve is anchored to the same point because of the initial retrieval, but they all vary differently with emission angle. Despite the broad scatter of points, it is immediately apparent from Figure 11a that the case ω = 0.9, g = 0.7 is consistent with the data, but ω = 0.7, g = 0.9 is not. This retrieval and forward-modelling process was re- peated for more parameter combinations. For each case, 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.10 0.11 Radiance at 60° (µWcm−2sr−1µm−1) Figure 12: The synthetic curve radiances at an angle of 60◦, for a range of scattering parameters. The red crosses mark the three example parameters used in Figure 11. The red line marks the minimum observed radiance from the VIMS data; scattering parameters above this line are not consistent with the data. Cases where ω is low and g is high are ruled out. the radiance at an emission angle of 60◦ was recorded and is plotted in Figure 12. The red line corresponds to a ra- diance of 0.062 µW cm−1 sr−1 µm−1, the lowest observed radiance at 60◦ in the VIMS equatorial data. Scattering parameters which produce radiances larger than this value are consistent with the data; scattering parameters which produce smaller radiances are not. Looking at Figure 11, the blue curve falls within the acceptable fit region, the green curve is on the borderline, and the red curve does not give an acceptable fit. While a fairly large range of ω and g values give an acceptable fit to the VIMS limb dark- ening data (the bottom-right segment of the plot), this figure rules out the case where ω is low and g is high. For this analysis, the location of the cloud deck was held constant at 0.8 bar. The effect of varying the altitude of the cloud deck was investigated, and it was found that no additional constraint was provided on top of the results from Section 6.1, i.e. the clouds can be located anywhere above 1.2 bar. The conclusions drawn from Figure 12 are independent of the cloud location, provided that it is above the 1.2 bar level. in results from Roos-Serote and Irwin (2006), who performed a limb darkening analysis using pairs of 5 µm data cubes from Galileo NIMS. With g fixed at 0.8, they found that an optimum fit was obtained with ω = 0.9 ± 0.5. This is good agreement with the 7. Global retrievals Using the best-fit cloud model from the previous sec- tion (a single, compact grey cloud located above 1.2 bar), retrievals were run across the entire planet in order to in- vestigate variability in the atmospheric parameters. As before, the free parameters in the retrievals were the cloud 11 s s e n k c h T i l a c i t p O 25 20 15 10 5 Cloud ) % ( y t i i d m u H e v i t l a e R 20 40 −40 −20 0 Latitude 4 3 2 1 0 H2O ) m p p ( . . R M V . PH3 1.4 1.2 1.0 0.8 0.6 0.4 0.2 −40 −20 0 Latitude 20 40 −40 −20 0 Latitude n / 2 χ 3.0 2.5 2.0 1.5 1.0 0.5 0.0 20 40 Goodness−of−fit −40 −20 0 Latitude 20 40 Figure 13: North-south latitudinal retrievals of the cloud opacity, the H2O relative humidity and the PH3 abundance from Jupiter's nightside. The different colours correspond to the four different conditions placed on the H2O relative humidity: allowed to vary (black), fixed at 0.2% (blue), fixed at 1% (red), fixed at 3% (green). Along with the retrieved quantities is the goodness-of-fit as a function of latitude for each condition. Cloud Opacity e d u t i t a L 40 20 0 -20 -40 360 315 270 225 180 135 90 System III West Longitude 45 0 5 10 15 Optical Thickness Phosphine 20 25 e d u t i t a L 40 20 0 -20 -40 360 315 270 225 180 135 90 System III West Longitude 45 0 NEB SEB . NEB SEB 0.40 0.55 0.70 V.M.R (ppm) 0.85 1.00 Figure 14: Global maps of the retrieved cloud opacity and PH3 abundance from the VIMS nightside observations. The H2O relative humidity has been held fixed at 0.2%. Smoothing has been applied to these maps. opacity, the water vapour relative humidity, and the scal- ing factor of phosphine (effectively the deep volume mixing ratio). Retrievals were run for each of the the forty-one 5◦ latitudinal bins described in Section 2.2 and the results are shown in Figure 13. The initial error estimate of 12% lead to underfitting the data, so the errors were reduced to 6% in order to give a reduced chi-squared value of ∼1 - the errors shown in grey are the formal errors on the retrieval, determined using the 6% error value. 7.1. Water vapour For the initial retrieval, the H2O abundance was in- cluded as a free parameter. The results for this process are shown in black in Figure 13. If allowed to vary, the zonally-averaged relative humidity ranges from ∼0.2% to ∼3%, with a maximum around the equator and a some- what asymmetric appearance. However, Figure 5 previ- ously showed that H2O cannot be accurately retrieved due to extreme degeneracy and correlation with the other pa- rameters, so these results should be taken with caution. To further test these results, we then fixed the H2O relative humidity at various different levels and re-ran the latitudinal retrievals. Figure 13 shows the results for three of these relative humidities: 0.2% (blue), 1% (red) and 3% (green). The fourth panel of this figure shows the impact that each fixed value has on the goodness-of-fit. Although the higher fixed relative humidities (1% and 3%) lead to significantly worse fits at certain latitudes, the lowest value (0.2%) produces fits that are comparable to the free re- trieval. We note that removing water vapour entirely from the retrievals significantly worsened the fit. For certain fixed H2O relative humidities, the cloud opacity and PH3 abundance are able to adjust and pro- vide an equally good fit to the data. By considering the changes in the chi-square parameter, we estimate that the maximum fixed relative humidity allowed is approximately 0.5%. Since we are able to achieve a good fit using a sin- gle H2O profile at all latitudes (provided that the relative humidity is less than 0.5%) we conclude that the VIMS 5-μm spectra do not provide any evidence for latitudinal variability in water vapour. Previous studies using NIMS 12 data (Roos-Serote et al., 2000) have, however, suggested that there is considerable local variability in the H2O hu- midity; since these NIMS analyses made use of high spatial resolution observations of small regions of the planet, this discrepancy may be due to sub-pixel and zonal inhomo- geneity in the relatively low resolution VIMS observations. Any small regions of elevated water may be rendered invis- ible by averaging over the large areas covered by the VIMS pixels, although a thorough exploration of gas and cloud degeneracies in the NIMS dataset is required to confirm this hypothesis. Assessments of spatial variability of tropospheric H2O must await higher spectral resolution space-based mea- surement to better distinguish between the competing ef- fects of water, cloud opacity and phosphine. 7.2. Phosphine The latitudinal distributions of PH3 for the different H2O conditions are shown in Figure 13, and Figure 14 shows the global distribution for the case where the H2O relative humidity is fixed at 0.2%. These global maps were produced by binning the spectra into 10◦ × 10◦ size bins and running a retrieval at each latitude and longitude point. The different water profiles in Figure 13 lead to the fol- lowing global averages for the PH3 deep volume mixing ratio: 0.90±0.09 ppm (allowing H2O to vary), 0.76±0.05 ppm (0.2% relative humidity), 0.92±0.6 ppm (1% relative humidity), 1.09±0.06 ppm (3% relative humidity). The best fits therefore suggest a deep volume mixing ratio of 0.76-0.90 ppm. These retrieved global averages are consis- tent with previous 5-μm studies, including Bjoraker et al. (1986) who gives a value of 0.7 ppm from airborne spec- troscopic observations, and Irwin et al. (1998) who gives a value of 0.77 ppm from Galileo NIMS. However, analyses of Cassini CIRS 7.7-16.6 µm observations have reported higher values: Irwin et al. (2004) fit the data using a deep volume mixing ratio that varies between 1.0 and 1.5 ppm, while Fletcher et al. (2009) found that values of 1.8-1.9 ppm produced the best fit. Since CIRS is sensitive to higher altitudes than NIMS and VIMS, and we expect the phosphine abundance to decrease with altitude rather than increase, it is likely that this discrepancy is due to an in- consistency in the database line strengths between the two spectral regions. Although the different water vapour profiles shown in Figure 13 lead to slightly different PH3 retrievals, they each produce a similar latitudinal pattern: an en- hanced abundance at high latitudes compared to low lat- itudes. This is a phenomenon previously noted in the northern hemisphere by Drossart et al. (1990), who used high-resolution 5-μm spectra to detect a 60% enhance- ment of PH3 at high northern latitudes compared to the NEB; our retrievals give an enhancement of ∼60-75%. Drossart et al. (1990) suggest that this high-latitude en- hancement could be either due to an increase in photo- ) 1 − m µ 1 − r s 2 − m c W µ ( e c n a d a R i Dayside Nightside 20 15 10 5 0 −40 −20 0 Latitude 20 40 Figure 15: Zonally averaged 5.0 µm radiance as a function of latitude for both the nightside and the dayside. chemical dissociation or a decrease in the eddy diffusion coefficient at high latitudes. 7.3. Clouds Figures 2 and 3 both showed that there was significant variation in 5-μm radiance with latitude; as expected, Fig- ures 13 and 14 shows that this is primarily due to variable cloud thickness. The cloud opacity retrieval shown in Fig- ure 14 is the inverse of Figure 2a; where the brightness temperature is high, the opacity is low, and vice versa. The absolute retrieved values of the cloud optical thick- ness are highly dependent on the chosen cloud parameters (and slightly dependent on the H2O relative humidity), but the relative changes are always reproduced, with high optical thicknesses in the zones (0◦,±30◦) and low optical thicknesses in the belts (±15◦). Figure 2 also showed that there is considerable variabil- ity within latitude bands. One region that has significant longitudinal variation is the SEB, where the brightness temperature varies with distance from the Great Red Spot. Figure 14 shows that this is primarily due to cloud opacity. The thickest clouds in the SEB occur to the west of the GRS (50◦W, 20◦S), in its turbulent wake. It is not un- til the opposite side of the planet (∼ 225◦W), that these clouds thin out and the brightness temperature increases. This phenomenon may be explained by turbulence dredg- ing up material from deeper pressures, causing clouds to form. As the distance from the GRS increases, the at- mosphere becomes more quiescent and we return to the ordinary, relatively cloud-free appearance of a belt. 8. Reflected sunlight analysis In addition to the nightside data, VIMS also made mea- surements on Jupiter's dayside. The majority of this study has focussed on the nightside data, as the absence of re- flected sunlight considerably simplifies the analysis. In this final section, we seek to determine whether the results ob- tained from the nightside are consistent with the dayside observations. Figure 2 compared the global maps from the nightside and the dayside, and we noted that the contrast between 13 ) 1 − m µ 1 − r s 2 − m c W µ ( e c n a d a R i 3 2 1 0 4.5 4.6 4.7 Data Retrieval 4.8 Wavelength (µm) 4.9 5.0 5.1 Figure 16: Example of the fit obtained using dayside data from the equatorial zone. the belts and the zones is smaller on the dayside than on the nightside. This can also be seen in Figure 15, which gives the zonal averages as a function of latitude. The ad- ditional component of reflected sunlight makes little differ- ence in the warm belts and there are even points where the nightside is brighter than the dayside. This is because the belts are relatively cloud-free, so there is less reflection of sunlight. If we were looking at precisely the same cloud at different solar zenith angles, then we would expect a small increase due to reflected sunlight; the fact that we some- times see a small decrease is due to the fact that the gap of several weeks between the measurements and the low spa- tial resolution means that we are not observing identical atmospheric conditions. In the cooler, cloudier regions of the planet, such as the equator, the reflected sunlight component becomes more significant. The thicker clouds lead to more reflec- tion, and ensure that the dayside is consistently brighter than the nightside. At the equator, the zonal average radi- ance at 5.0 µm increases from 0.25 µW cm−1 sr−1 µm−1 to 1.33 µW cm−1 sr−1 µm−1, an increase of more than 500%. Again, part of this difference may be due to the fact that we are not necessarily observing identical regions of the planet, but reflected sunlight clearly accounts for a signif- icant part of the dayside radiance from the zones. This flux difference between the nightside and dayside obser- vations of the equatorial zone is consistent with the anal- ysis of Drossart et al. (1998), who studied the solar re- flected component of Jupiter's 5-μm spectra from Galileo NIMS and found that the minimum flux level was six times greater on the dayside than on the nightside. We ran retrievals across the range of latitudes on the dayside, using the set of parameters from the previous sections: the volume mixing ratio of PH3, the relative humidity of water and the opacity of a single, compact, grey cloud located at 0.8 bar. For these dayside retrievals, the reflected solar component is included in the radiative transfer model. We found that we were able to obtain a good fit to the data using this simple cloud model; an ex- ample of the fit obtained in the equatorial region is shown in Figure 16. 14 The full results of the zonally-averaged retrievals are shown in Figure 17. As with Figure 13, the different colours refer to the different conditions placed on the H2O profile. The overall shapes of the retrievals are similar to the equivalent nightside plot (Figure 13); the cloud opac- ities peak at the same latitudes, and the PH3 abundance still exhibits an enhancement at high latitudes. There are, however, a few differences. Firstly, the retrieved values of PH3 exhibit additional minima which correspond to the peaks in cloud opacity. We believe the addition of re- flected sunlight accentuates the degeneracy between the cloud parameters and the PH3 abundance, as it is in the PH3 absorption wing at the short-wavelength edge of the window where the reflected sunlight component is most significant. A small change in the cloud scattering pa- rameters can lead to dramatic changes in the retrieved PH3 values (a phenomenon not seen on the nightside), so retrievals are unreliable. Higher resolution dayside spec- troscopy would reduce this degeneracy, allowing more re- liable PH3 retrievals. Secondly, although the cloud opacity has a very similar shape, the absolute values are slightly different. This is likely to be due to a combination of two factors: the opac- ities may be genuinely different, due to averaging over spa- tially inhomogeneous latitudinal band, and the retrievals may be mismatched due to slightly incorrect cloud scat- tering properties leading to an imbalance between thermal radiation and reflected sunlight. A more comprehensive study, taking into account shorter wavelengths, would be required to jointly constrain the cloud properties. 9. Conclusions This paper uses the 2000/2001 observations of Jupiter made by the Cassini VIMS instrument in the 4.5-5.1 µm range to study the planet's tropospheric composition and cloud structure. This builds on previous work using the Galileo NIMS and Voyager IRIS datasets by making use of (i) the full global coverage afforded by VIMS, covering both warm and cool regions of the planet; (ii) the combi- nation of nightside and dayside observations. We conclude that: 1. VIMS 5-μm data (both nightside and dayside) can be modelled using a very simple cloud model, consisting of a single, compact, spectrally flat cloud. 2. The bulk of the 5-μm opacity must be located suffi- ciently high in the troposphere that is does not impact the shape of the spectrum. For the coolest regions on the planet, with the thickest clouds, this requirement constrains the clouds to altitudes of 1.2 bar or higher. In warmer regions, the clouds can be placed deeper in the atmosphere and still achieve a good fit to the data. 3. The spectra from both cloudy and relatively cloud- free regions have very similar spectral shapes, so the s s e n k c h T i l a c i t p O 25 20 15 10 5 Cloud ) % ( y t i i d m u H e v i t l a e R 20 40 −40 −20 0 Latitude 4 3 2 1 0 H2O ) m p p ( . . R M V . PH3 1.4 1.2 1.0 0.8 0.6 0.4 0.2 −40 −20 0 Latitude 20 40 −40 −20 0 Latitude n / 2 χ 3.0 2.5 2.0 1.5 1.0 0.5 0.0 20 40 Goodness−of−fit −40 −20 0 Latitude 20 40 Figure 17: North-south latitudinal retrievals of the cloud opacity, the H2O relative humidity and the PH3 abundance from Jupiter's dayside. The different colours correspond to the four different conditions placed on the H2O relative humidity: allowed to vary (black), fixed at 0.2% (blue), fixed at 1% (red), fixed at 3% (green). Along with the retrieved quantities is the goodness-of-fit as a function of latitude for each condition. clouds must be relatively spectrally-flat. Pure NH3- ice and pure NH4SH do not have sufficiently grey spec- tra, and are therefore inconsistent with VIMS data. It may be that spectral features in the clouds are masked by coating layers, or that multiple cloud decks act to- gether to blur out any features. 4. The relative lack of limb darkening means that the cloud particles must be highly scattering. There is a degeneracy between the single-scattering albedo and the asymmetry parameter, but cases with a low single- scattering albedo and high asymmetry parameter are ruled out. 5. The majority of the 5-μm global inhomogeneity can be accounted for by variations in the cloud opacity, with thick clouds in the zones and relatively cloud- free belts. 6. The retrieved globally-averaged deep volume mixing ratio for PH3 was 0.76±0.05 ppm (with the H2O rela- tive humidity fixed at 0.2%), consistent with previous 5-μm studies. The latitudinal retrieval of PH3 shows an enhancement in the abundance at high latitudes. 7. The VIMS 5-μm spectra do not provide any evidence for latitudinal variability in the H2O relative humid- ity; if the H2O abundance is held fixed, the cloud opacity and PH3 abundance are able to adjust in or- der to produce an equally good fit to the data. The low spectral resolution and high degeneracy between gases mean that H2O cannot be accurately retrieved, but a fixed relative humidity of less than 0.5% is able to provide a good fit to the data at all latitudes. 8. The low spectral resolution of VIMS also prevents the accurate retrieval of other gaseous species with weaker signatures in the 5-μm window: NH3, CH3D, GeH4, AsH3 & CO. High resolution spectroscopy is required to retrieve abundances of these species and to search for any spatial variability. Acknowledgements Giles was supported via a Royal Society studentship, and Fletcher was supported via a Royal Society Re- search Fellowship at the University of Oxford. Irwin 15 acknowledges the support of the United Kingdom Sci- ence and Technology Facilities Council. We thank the Cassini/VIMS team for planning and implementing these observations. References Anderson, J., Sides, S., Soltesz, D., Sucharski, T., Becker, K., 2004. Modernization of the integrated software for imagers and spec- trometers, in: Lunar and Planetary Science Conference, p. 2039. Atreya, S., Wong, M., Owen, T., Niemann, H., Mahaffy, P., 1997. Three Galileos: The Man, The Spacecraft, The Telescope. Kluwer Academic Publishers, Dordrecht. chapter Chemistry and Clouds of Jupiter's Atmosphere: A Galileo Perspective. pp. 249 -- 260. Bailly, D., Birnbaum, G., Buechele, A., Flaud, P.M., Hartmann, J.M., 2004. Absorption by pure NH3 and NH3 -- H2 mixtures in the 5 µm window region. Journal of Quantitative Spectroscopy and Radiative Transfer 83, 1 -- 13. Baines, K.H., Carlson, R.W., Kamp, L.W., 2002. Fresh ammonia ice clouds in Jupiter: I. Spectroscopic identification, spatial distribu- tion, and dynamical implications. Icarus 159, 74 -- 94. Banfield, D., Gierasch, P., Bell, M., Ustinov, E., Ingersoll, A., Vasavada, A., West, R.A., Belton, M., 1998. Jupiter's cloud struc- ture from Galileo imaging data. Icarus 135, 230 -- 250. B´ezard, B., Baluteau, J.P., Marten, A., 1983. Study of the deep cloud structure in the equatorial region of Jupiter from Voyager infrared and visible data. Icarus 54, 434 -- 455. B´ezard, B., Lellouch, E., Strobel, D., Maillard, J.P., Drossart, P., 2002. Carbon monoxide on Jupiter: Evidence for both internal and external sources. Icarus 159, 95 -- 111. Bjoraker, G.L., Larson, H.P., Kunde, V.G., 1986. The gas composi- tion of Jupiter derived from 5-µm airborne spectroscopic observa- tions. Icarus 66, 579 -- 609. Brown, R., Baines, K., Bellucci, G., Bibring, J.P., Buratti, B., Ca- paccioni, F., Cerroni, P., Clark, R., Coradini, A., Cruikshank, D., et al., 2004. The Cassini visual and infrared mapping spectrometer (VIMS) investigation, in: The Cassini-Huygens Mission. Springer, pp. 111 -- 168. Carlson, B.E., Lacis, A.A., Rossow, W.B., 1992. The abundance and distribution of water vapor in the Jovian troposphere as inferred from Voyager IRIS observations. The Astrophysical Journal 388, 648 -- 668. Drossart, P., Encrenaz, T., Kunde, V., Hanel, R., Combes, M., 1982. An estimate of the PH3, CH3D, and GeH4 abundances on Jupiter from the Voyager IRIS data at 4.5 µm. Icarus 49, 416 -- 426. Drossart, P., Lellouch, E., B´ezard, B., Maillard, J.P., Tarrago, G., 1990. Jupiter: Evidence for a phosphine enhancement at high northern latitudes. Icarus 83, 248 -- 253. Drossart, P., Roos-Serote, M., Encrenaz, T., Lellouch, E., Baines, K., Carlson, R., Kamp, L., Orton, G., Calcutt, S., Irwin, P., in the 4 -- 5 µm range from Galileo/near-infrared mapping spec- trometer observations: Measurements of cloud opacity, water, and ammonia. Journal of Geophysical Research: Planets (1991 -- 2012) 103, 23023 -- 23041. Roos-Serote, M., Irwin, P., 2006. Scattering properties and loca- tion of the Jovian 5-micron absorber from Galileo/NIMS limb- darkening observations. Journal of Quantitative Spectroscopy and Radiative Transfer 101, 448 -- 461. Roos-Serote, M., Vasavada, A., Kamp, L., Drossart, P., Irwin, P., Nixon, C., Carlson, R., 2000. Proximate humid and dry regions in jupiter's atmosphere indicate complex local meteorology. Nature 405, 158 -- 160. Seiff, A., Kirk, D.B., Knight, T.C., Young, R.E., Mihalov, J.D., Young, L.A., Milos, F.S., Schubert, G., Blanchard, R.C., Atkin- son, D., 1998. Thermal structure of Jupiter's atmosphere near the edge of a 5-µm hot spot in the north equatorial belt. Journal of Geophysical Research: Planets (1991 -- 2012) 103, 22857 -- 22889. Simon-Miller, A.A., Conrath, B., Gierasch, P.J., Beebe, R.F., 2000. A detection of water ice on Jupiter with Voyager IRIS. Icarus 145, 454 -- 461. Sromovsky, L., Fry, P., 2010. The source of widespread 3-µm ab- sorption in Jupiter's clouds: Constraints from 2000 Cassini VIMS observations. Icarus 210, 230 -- 257. Wenger, C., Champion, J., 1998. Spherical top data system (STDS) software for the simulation of spherical top spectra. Journal of Quantitative Spectroscopy and Radiative Transfer 59, 471 -- 480. Westphal, J., 1969. Observations of localised 5-micron radiation from Jupiter. The Astrophysical Journal 157, L63 -- L64. Wong, M.H., Bjoraker, G.L., Smith, M.D., Flasar, F.M., Nixon, C.A., 2004. Identification of the 10-µm ammonia ice feature on Jupiter. Planetary and Space Science 52, 385 -- 395. et al., 1998. The solar reflected component in Jupiter's 5-µm spectra from NIMS/Galileo observations. Journal of Geophysical Research: Planets (1991 -- 2012) 103, 23043 -- 23049. Fletcher, L., Orton, G., Teanby, N., Irwin, P., 2009. Phosphine on Jupiter and Saturn from Cassini/CIRS. Icarus 202, 543 -- 564. Fletcher, L.N., Baines, K.H., Momary, T.W., Showman, A.P., Irwin, P.G., Orton, G.S., Roos-Serote, M., Merlet, C., 2011. Saturn's tropospheric composition and clouds from Cassini/VIMS 4.6 -- 5.1 µm nightside spectroscopy. Icarus 214, 510 -- 533. Gierasch, P.J., Conrath, B.J., Magalhaes, J.A., 1986. Zonal mean properties of Jupiter's upper troposphere from Voyager infrared observations. Icarus 67, 456 -- 483. Henyey, L.G., Greenstein, J.L., 1941. Diffuse radiation in the galaxy. The Astrophysical Journal 93, 70 -- 83. Irwin, P., Dyudina, U., 2002. The retrieval of cloud structure maps in the equatorial region of Jupiter using a principal component analysis of Galileo/NIMS data. Icarus 156, 52 -- 63. Irwin, P., Parrish, P., Fouchet, T., Calcutt, S., Taylor, F., Simon- Miller, A., Nixon, C., 2004. Retrievals of jovian tropospheric phos- phine from Cassini/CIRS. Icarus 172, 37 -- 49. Irwin, P., Teanby, N., de Kok, R., Fletcher, L., Howett, C., Tsang, C., Wilson, C., Calcutt, S., Nixon, C., Parrish, P., 2008. The NEMESIS planetary atmosphere radiative transfer and retrieval tool. Journal of Quantitative Spectroscopy and Radiative Transfer 109, 1136 -- 1150. Irwin, P., Weir, A., Smith, S., Taylor, F., Lambert, A., Calcutt, S., Cameron-Smith, P., Carlson, R., Baines, K., Orton, G., et al., 1998. Cloud structure and atmospheric composition of Jupiter retrieved from Galileo near-infrared mapping spectrometer real- time spectra. Journal of Geophysical Research: Planets (1991 -- 2012) 103, 23001 -- 23021. Irwin, P., Weir, A., Taylor, F., Calcutt, S., Carlson, R., 2001. The origin of belt/zone contrasts in the atmosphere of Jupiter and their correlation with 5-µm opacity. Icarus 149, 397 -- 415. Kalogerakis, K.S., Marschall, J., Oza, A.U., Engel, P.A., Meharc- hand, R.T., Wong, M.H., 2008. The coating hypothesis for am- monia ice particles in Jupiter: Laboratory experiments and optical modeling. Icarus 196, 202 -- 215. Lacis, A.A., Oinas, V., 1991. A description of the correlated k distri- bution method for modeling nongray gaseous absorption, thermal emission, and multiple scattering in vertically inhomogeneous at- mospheres. Journal of Geophysical Research: Atmospheres (1984 -- 2012) 96, 9027 -- 9063. Lellouch, E., B´ezard, B., Fouchet, T., Feuchtgruber, H., Encrenaz, T., de Graauw, T., 2001. The deuterium abundance in Jupiter and Saturn from ISO-SWS observations. Astronomy and Astrophysics 370, 610 -- 622. Lewis, J.S., 1995. Physics and Chemistry of the Solar System. Aca- demic Press. chapter Cloud Physics on the Jovian Planets. pp. 154 -- 158. Matcheva, K.I., Conrath, B.J., Gierasch, P.J., Flasar, F.M., 2005. The cloud structure of the jovian atmosphere as seen by the Cassini/CIRS experiment. Icarus 179, 432 -- 448. McCord, T., Coradini, A., Hibbitts, C., Capaccioni, F., Hansen, G., Filacchione, G., Clark, R., Cerroni, P., Brown, R., Baines, K., et al., 2004. Cassini VIMS observations of the Galilean satellites including the VIMS calibration procedure. Icarus 172, 104 -- 126. Nixon, C., Achterberg, R., Conrath, B., Irwin, P., Teanby, N., Fouchet, T., Parrish, P., Romani, P., Abbas, M., LeClair, A., et al., 2007. Meridional variations of C2H2 and C2H6 in Jupiter's atmosphere from Cassini CIRS infrared spectra. Icarus 188, 47 -- 71. Nixon, C., Irwin, P., Calcutt, S., Taylor, F., Carlson, R., 2001. Atmo- spheric composition and cloud structure in jovian 5-µm hotspots from analysis of Galileo NIMS measurements. Icarus 150, 48 -- 68. Plass, G.N., Kattawar, G.W., Catchings, F.E., 1973. Matrix operator theory of radiative transfer. 1: Rayleigh scattering. Applied Optics 12, 314 -- 329. Roos-Serote, M., Drossart, P., Encrenaz, T., Lellouch, E., Carlson, R., Baines, K., Kamp, L., Mehlman, R., Orton, G., Calcutt, S., et al., 1998. Analysis of Jupiter north equatorial belt hot spots 16
1910.11293	1	1910	2019-10-24T17:11:50	Early Dynamics of the Lunar Core	[ "astro-ph.EP" ]	The Moon is known to have a small liquid core, and it is thought that in the distant past the core may have produced strong magnetic fields recorded in lunar samples. Here we implement a numerical model of lunar orbital and rotational dynamics that includes the effects of a liquid core. In agreement with previous work, we find that the lunar core is dynamically decoupled from the lunar mantle, and that this decoupling happened very early in lunar history. Our model predicts that the lunar core rotates sub-synchronously, and the difference between the core and the mantle rotational rates was significant when the Moon had a high forced obliquity during and after the Cassini State transition. We find that the presence of the lunar liquid core further destabilizes synchronous rotation of the mantle for a wide range of semimajor axes centered around the Cassini State transition. CMB torques make it even more likely that the Moon experienced large-scale inclination damping during the Cassini State transition. We present estimates for the mutual core-mantle obliquity as a function of Earth-Moon distance, and we discuss plausible absolute time-lines for this evolution. We conclude that our results are consistent with the hypothesis of a precession-driven early lunar dynamo and may explain the variability of the inferred orientation of the past lunar dynamo.	astro-ph.EP	astro-ph	manuscript submitted to JGR: Planets Early Dynamics of the Lunar Core Matija ´Cuk1, Douglas P. Hamilton2, Sarah T. Stewart3 1SETI Institute, 189 North Bernardo Avenue, Mountain View, CA 94043 2University of Maryland, College Park 3University of California, Davis Key Points: • The liquid lunar core is dynamically decoupled from the mantle. • In the distant past the mutual obliquity between the lunar core and the mantle was large. • The friction at the core-mantle boundary probaly kept the Moon out of synchronous rotation during periods of high obliquity. 9 1 0 2 t c O 4 2 . ] P E h p - o r t s a [ 1 v 3 9 2 1 1 . 0 1 9 1 : v i X r a Corresponding author: Matija ´Cuk, [email protected] -- 1 -- manuscript submitted to JGR: Planets Abstract The Moon is known to have a small liquid core, and it is thought that in the distant past the core may have produced strong magnetic fields recorded in lunar samples. Here we implement a numerical model of lunar orbital and rotational dynamics that includes the effects of a liquid core. In agreement with previous work, we find that the lunar core is dynamically decoupled from the lunar mantle, and that this decoupling happened very early in lunar history. Our model predicts that the lunar core rotates sub-synchronously, and the difference between the core and the mantle rotational rates was significant when the Moon had a high forced obliquity during and after the Cassini State transition. We find that the presence of the lunar liquid core further destabilizes synchronous rotation of the mantle for a wide range of semimajor axes centered around the Cassini State tran- sition. CMB torques make it even more likely that the Moon experienced large-scale in- clination damping during the Cassini State transition. We present estimates for the mu- tual core-mantle obliquity as a function of Earth-Moon distance, and we discuss plau- sible absolute time-lines for this evolution. We conclude that our results are consistent with the hypothesis of a precession-driven early lunar dynamo and may explain the vari- ability of the inferred orientation of the past lunar dynamo. 1 Introduction The Moon has a much smaller bulk density than Earth, indicating that the lunar iron core is much smaller than that of Earth. Lunar laser ranging (LLR) and GRAIL mission results indicate that the Moon's core has a radius of about 340 km (with the 200- 380 km diameter range allowed by data) and is (at least partially) liquid (Williams et al., 2014; Williams, Turyshev, Boggs, & Ratcliff, 2006). The presence of the core is fur- ther suggested by magnetic induction (Hood, Mitchell, Lin, Acuna, & Binder, 1999; Shimizu, Matsushima, Takahashi, Shibuya, & Tsunakawa, 2013) and seismic modeling (Garcia, Gagnepain-Beyneix, Chevrot, & Lognonn´e, 2011; Weber, Lin, Garnero, Williams, & Lognonn´e, 2011). While the Moon does not have a global magnetic field, remanent magnetization observed from orbit and in lunar samples indicates that there was a lunar magnetic field in the distant past. Since the magnetic fields of terrestrial planets are thought to be gen- erated in their liquid iron cores, the rotational dynamics of the Moon's core throughout lunar history is directly relevant to studies of ancient lunar magnetism. -- 2 -- manuscript submitted to JGR: Planets Lunar rotational dynamics is very different from that of Earth. The Moon is in syn- chronous rotation due to Earth's tidal forces and the resulting tidal dissipation, and this state was likely established very soon after the Moon's formation. Tidal dissipation on Earth over billions of years made the Moon's semimajor axis expand from several Earth radii (RE) at formation to the present value of 60.3 RE. As the lunar orbit and orbital period grew, so did the Moon's rotation period. While the Moon's rotational period is equal to its orbital period, the Moon's obliquity is determined by the so-called Cassini States. Cassini States are minimum-energy solutions for the forced obliquity of a rotat- ing body that experiences tidal dissipation, and most synchronous satellites have evolved into stable Cassini States (Colombo, 1966; Peale, 1969). The spin axis of the satellite in a Cassini State is in the same plane as the normals to the orbital plane and the Laplace plane (the plane around which the orbit precesses). When the period of the axial pre- cession is longer than that of orbital precession (as is the case for the present-day Moon), the satellite is in Cassini State 2, in which the spin axis is closer to the Laplace plane normal than it is to the orbit normal (in the present-day Earth-Moon system, the Laplace plane is the ecliptic). Ward (1975) first established that the Moon underwent a transi- tion between the Cassini States 1 and 2 at a semimajor axis of about 34RE, assuming that the Moon's current shape was already established at this time. The necessity of the Cassini State transition is dictated by the dependence of the lunar spin and orbital pre- cession on Earth-Moon separation. The lunar spin precession is driven by Earth's tidal torque and therefore slowed down with the Moon's tidal recession. However, beyond the Earth-Moon distance of about 15RE (within which Earth's oblateness was an important perturbation) the tidal recession led to speeding up of lunar apsidal precession, as it is driven by solar perturbations. Therefore, to reach the present state where the natural periods of nodal and spin precession are about 19 and 80 yr, the Moon had to experi- ence equality of these two periods at some point in its evolution, which Ward (1975) placed at the Earth-Moon distance of about 34 RE. Ward (1975) also found that the Moon would experience very large forced obliquities during the Cassini State transition. Meyer and Wisdom (2011) examined how the Moon's liquid core would behave dur- ing the obliquity evolution established by Ward (1975). Meyer and Wisdom (2011) show that the Moon's free core nutation (FCN) period should be several centuries, much longer than the current nodal precession period of the lunar orbit (18.6 yr). Since the Moon's spin axis is in a Cassini State and its precession follows that of the orbit, the natural fre- -- 3 -- manuscript submitted to JGR: Planets Figure 1. The modern-day relative orientations of the lunar mantle and core spin axes in the plane defined by the ecliptic and lunar orbit normals. The mantle is in Cassini State 2, while the core is in "quasi-Cassini State 2" relative to the mantle. The angles are not to scale: the lunar orbital inclination to the ecliptic is about 5.2◦, the mantle obliquity to the ecliptic is about 1.6◦ and the core obliquity to the ecliptic is only about 0.1◦. quency of nutation of the core around the mantle is too slow to allow the core to follow the mantle's motion. Instead, the lunar core spin axis stays close to the ecliptic normal, in an arrangement somewhat equivalent to Cassini State 2 (the core spin axis is in the plane defined by the ecliptic normal and the mantle spin axis, and the two spin axes are on the opposite sides of the ecliptic normal; Fig. 1). This is in contrast to the case of Earth, where FCN takes only about 400 days, coupling the core to the mantle, which itself is precessing much more slowly, with a 26 kyr period. Meyer and Wisdom (2011) find that this decoupling of the lunar core's rotation axis from that of the mantle should have happened relatively early in lunar orbital evolution, when the Moon had semima- jor axis a = 26 − 29RE, before the mantle's Cassini State transition. Note that the core being "de-coupled" from the mantle means that the two are spinning around dif- ferent axes with different precessional motion; this terminology does not imply that there is no contact or no interaction between the core and the mantle. Dwyer, Stevenson, and Nimmo (2011) recognized that the results of Meyer and Wisdom (2011) imply a large past mutual obliquities between the core and the mantle, and Dwyer et al. (2011) argued -- 4 -- manuscript submitted to JGR: Planets this differential precession may have been the ultimate driving force behind the lunar dynamo. More recently, Chen and Nimmo (2016) have shown that lunar obliquity tides should have significantly affected the past lunar inclination. ´Cuk, Hamilton, Lock, and Stew- art (2016) proposed a model in which the Earth has a very fast spin (<2.5 hour period) and a high obliquity ( 70◦) after lunar formation. Independent of the constraints from lunar inclination, this very fast initial spin of Earth is a requirement by some of the cur- rent lunar formation models in order to explain Earth-Moon isotopic similarity (Lock et al., 2018). As the lunar orbit grows in the ´Cuk et al. (2016) model, it encounters an unstable transition between equatorial and ecliptic Laplace planes (Tamayo, Burns, Hamil- ton, & Nicholson, 2013; Tremaine, Touma, & Namouni, 2009). The lunar orbit is tem- porarily trapped in this zone, where solar perturbations induce large eccentricities, loss of the system's angular momentum (which is transferred to the heliocentric orbit), a re- duction of Earth's obliquity, and a large lunar inclination (≃ 30◦). In this picture, the Moon has a high inclination as it enters the Cassini State transition, with strong obliq- uity tides during the transition reducing the inclination to close to the observed values. In this paper we will consider the dynamics of the lunar core in the context of the ´Cuk et al. (2016) model, and determine the consequences for the core's orientation and the Moon's tidal evolution. 2 Numerical Methods In this work we use a custom made numerical integrator cr-sistem, which com- bines mixed-variable symplectic integrator for orbital motion and a Lie-Poisson mapping for the lunar rotation. In all aspects except the treatment of lunar rotation (including the motions of the core), cr-sistem is identical to r-sistem used by ´Cuk et al. (2016). Therefore we direct the reader to the Methods section of ´Cuk et al. (2016) for all details regarding the treatment of mutual perturbation between the bodies, tides within Earth and the Moon, and the shape and precession of Earth. In this section we will describe our approach for integrating the rotations of the lunar mantle and core, and their mu- tual interaction. ´Cuk et al. (2016) used the approach of Touma and Wisdom (1994), who advance the rotation of the Moon in the body-fixed reference frame implicitly, while the changes -- 5 -- manuscript submitted to JGR: Planets to this motion arising from oblateness and triaxiality are integrated explicitly (this al- gorithm is ultimately based on the Poincar´e-Hough method Hough, 1895; Poincar´e, 1910). The fact that these perturbations to the uniform rotation are relatively small enabled ´Cuk et al. (2016) to efficiently integrate the rotational dynamics with only dozens of time steps required for each orbital/rotational period. However, inclusion of a triaxial core within a triaxial mantle makes the Hamiltonian much more complex. Touma and Wisdom (2001) discuss in some detail the optimal algorithms for nu- merically integrating the rotational dynamics of a planet with a solid mantle and a liq- uid core. While Touma and Wisdom (2001) are able to derive a reasonably efficient al- gorithm for the dynamics of an oblate (i.e. rotationally symmetric) planet with a core, there is no such option for a triaxial core-mantle system. As suggested by Touma and Wisdom (2001), we divide our Hamiltonian into three parts, which correspond to rota- tions around the three Cartesian axes: 1 Hx = Hy = Hz = C 1 2α(cid:16)AC P 2 + AP 2 2β(cid:16)BC Q2 + BQ2 1 2γ(cid:0)CC R2 + CR2 C C − 2FCP PC(cid:17) − 2GC QQC(cid:17) − 2HC RRC(cid:17) (1) (2) (3) where A, B, C are the principal moments of inertia and P, Q, R are the components of angular momentum measured with respect to the mantle's principal axes (values with- out subscript refer to the whole body, and those with the subscript "C" to the core. As Touma and Wisdom (2001) show, PC , QC and RC are not really the angular momen- tum components of the core, but function equivalently. Additionally, the core moments of inertia are FC = 2 5 mC bc, GC = 2 5 mC ac, HC = 2 5 mC ab, and we define α = AAC − F 2 C , β = BBC − G2 C and γ = CCC − H 2 C , where mC is the core's mass and a, b and c are the core's principal axes (our approach and notation follows that of Touma & Wis- dom, 2001). These three Hamiltonians each generate separate rotations of total and core angular momentum vectors M and MC around the corresponding mantle axes. For ex- ample, the angular velocities generated by Hz for M and MC are wz = (CC R−HCRC )/γ and wCz = (CRC − HC R)/γ (with the sense of rotations reversed for w and wC due to our mantle-fixed reference frame). Using these angular velocities, we can define ro- tation matrices that convert vectors between the mantle-fixed and inertial frames, and calculate periodic "kicks" on the rotational motion due to gravitational torques and tidal dissipation. -- 6 -- manuscript submitted to JGR: Planets It is clear from our definition of w that the rotation frequencies of the mantle around all three axis include terms independent of the core's properties (at least to the lowest order) and therefore w does not vanish even in the limit of the core's size and mass be- ing zero. The core-mantle cross-terms do become negligible, but the main components of M remain, and correspond to the rotation of M around the mantle (or, in an iner- tial frame, rotation of the mantle around the total angular momentum vector). This means that the integrator needs to resolve this basic rotational motion which is handled explic- itly, in contrast to the treatment in Touma and Wisdom (1994) (and subsequently ´Cuk et al., 2016) where only the slower nutational motion (separate from principal axis ro- tation) had to be resolved. Our numerical model which includes the lunar core there- fore requires significantly larger number of time-steps per rotation period. There is a di- rect parallel here to orbital integrations, with our integrator being analogous to a ba- sic T +U symplectic integrator (Forest & Ruth, 1990), which requires enough time-steps to resolve basic Keplerian motion, while the Touma and Wisdom (1994) scheme is anal- ogous to mixed-variable symplectic integrators (Wisdom & Holman, 1991) which assume Keplerian orbits, so fewer time-steps are needed per orbit. Therefore, we are not in a good position to replicate the long-term integrations of ´Cuk et al. (2016), as inclusion of the lunar core would increase integration time by at least an order of magnitude. Therefore we will restrict ourselves to examining how the presence of the lunar core affects the Cassini States of the Moon at different Earth-Moon distances, and determining the mutual mo- tions of the core and the mantle for the purposes of understanding the past lunar dy- namo. One remaining choice we had to make was the introduction of core-mantle friction. We calculate the vector difference between the rotational velocities of the mantle and the core ∆~ω, and then apply a torque to MC every time-step: dM = −CC ∆~ωKCM dt (4) where dt is the time-step and KCM = 8.43 × 10−3 year−1 is the magnitude of spin- orbit friction. We based this form of core-mantle torque that is linear in the core-mantle relative motion on Eq. 19 in Pavlov, Williams, and Suvorkin (2016). Our value for KCM corresponds to a damping timescale of (1000/2π) sidereal months, or 120 years, and is equivalent to Pavlov et al. (2016) kv/CT = 16 × 10−9 day−1. Our use of linear core- mantle boundary (CMB) friction follows the usage in LLR community, but the flow at the CMB boundary is likely to be turbulent and non-linear (Williams, Boggs, Yoder, Rat- -- 7 -- manuscript submitted to JGR: Planets cliff, & Dickey, 2001), so our model should be taken as a first approximation. As we ig- nored any tidal forces from Earth or the Sun on the core, the core-mantle friction term (Eq. 4) is the only dissipative influence on the core's rotation, and it acts to match it to the rotation of the mantle. However, non-dissipative core-mantle torques inherent in our integrator make the core precess at a different angle and phase from the mantle, which can lead to equilibrium points very different from core-mantle co-rotation, as we show in the following sections. 3 Core-Mantle (De)coupling After completing our integrator, the first situation we wanted to explore is the present- day dynamics of the lunar core. We used the same lunar shape parameters as ´Cuk et al. (2016), and we decided to set the mass and radius of the lunar core exactly to 0.01 MM and 0.2RM , and we will use these values throughout the paper. Note that the elliptic- ity of the core is by far the most important parameter for its dynamics, and the mass and density have only second-order effects on the core's nutation. Here we used shape parameters (CC −AC )/CC = 2.5×10−4 and (CC −BC)/CC = 2×10−4 for the core (el- lipticity was based on Pavlov et al., 2016). We used long-term average lunar tides (quan- tified by tidal quality factor QM = 60 and Love number k2M = 0.024). Figure 2 shows a 106 yr simulation in which the mantle was put in synchronous ro- tation and Cassini State 2 with obliquity to ecliptic ǫM = 1.5◦, while the core is made to approximately co-rotate with the mantle. For comparison, we also conducted a sep- arate experiment using an entirely solid Moon with no fluid core. In about 300 years the core decouples from the mantle's spin direction and settles in an apparently stable state with obliquity to ecliptic of ǫC = 0.1◦. The spin rate of the core on the other hand set- tles close to that of the mantle, firmly within the range of mantle's rotational librations (bottom). The mantle is meanwhile damping librations around its Cassini state close to 1.58◦, with the damping timescale being much faster than for a solid moon, indicating that the damping is dominated by CMB friction. Figure 3 shows the situation at the end of the same simulation as shown in Fig. 2. Here we plot the y-component the core and mantle spin vectors in the polar repre- sentation ǫi cos(Ωi), where ǫ is the obliquity with respect to ecliptic and Ω the longitude of the node (i.e. intersection between the plane normal to the spin axis and the eclip- -- 8 -- manuscript submitted to JGR: Planets Solid Moon Mantle Core 0 200000 400000 600000 800000 1x106 Time (years) 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 27.45 27.4 27.35 27.3 27.25 27.2 27.15 27.1 27.05 27 ) ° ( y t i u q i l b O ) s y a d ( d o i r e p n p S i Figure 2. Top: Evolution of the ecliptic obliquities of the lunar core (red) and the mantle (blue), as well as a solid moon (gray, mostly covered by blue) in the present Earth-Moon system. Bottom: Evolution of the spin periods of the solid Moon (gray), lunar mantle (blue) and the core (red) in the same simulations (pattern of waves in blue and black curves are an artifact of the output frequency, the real libration period is much shorter). Initial conditions had the core slightly sub-synchronous and with a spin axes aligned with that of the mantle. tic) for either the mantle and the core. This plot enables us to see how the obliquity vec- tors are oriented in space. From Fig. 3, it is clear that the mantle and the core spins are both precessing around the ecliptic normal with the 18.6 yr period dictated by the (or- bital) nodal precession. While the core's obliquity is much smaller, it is precessing ex- actly out-of-phase (i.e. 180◦ away) from the mantle, as expected from a "quasi-Cassini State 2" (see Fig. 1) in which the natural precession frequency of the core (around the mantle) is much slower than the mantle's precession around the reference direction (in this case, the ecliptic normal). As explained in Meyer and Wisdom (2011), the core should -- 9 -- manuscript submitted to JGR: Planets Mantle Core 2 1.5 1 0.5 0 -0.5 -1 -1.5 ) ° ( y t i u q i l b o d e j t c e o r P -2 0 20 40 60 Time (years) 80 100 Figure 3. The very end of the simulation shown in Fig. 2. The lines plot a projection of the obliquity of the core (red) and mantle (blue) on the y-axis in a non-rotating 2-D Cartesian frame centered on the ecliptic pole. It is clear that the mantle and the core spin axes precess around the ecliptic normal with opposite phases, which is analogous to core being in a quasi-Cassini State 2. precess around the mantle with the frequency: ωF CN = ω(cid:16) CC − AC CC (cid:17)(cid:16) C CM (cid:17) (5) where ω is the lunar spin rate, and CM = C − CC the mantle moment of inertia (for our purposes, C/CM ≃ 1). Since we use (CC − AC )/(CC ) = 2.5 × 10−4, and the lunar spin period is 27.3 days, then the period for the core nutation should be ≃ 300 yr. Since ǫC << ǫM , we can estimate ǫC/ǫM ≃ ωF CN / Ω, where Ω is the nodal precession rate for the Moon (based with analogy with Cassini State 2 Ward, 1975). Therefore, the ex- pected forced obliquity of the lunar core should be ǫC = 1.6◦(18.6yr/300yr) ≃ 0.1◦, fully consistent with the results of our simulation. These results are also very similar to the findings of Dumberry and Wieczorek (2016) and Stys and Dumberry (2018) who ad- ditionally took into account the existence of the inner solid core. So we can conclude that for the present Earth-Moon system, our numerical model returns an answer consistent -- 10 -- manuscript submitted to JGR: Planets with the analytical expectations, with the lunar core precessing separately from the man- tle with a much smaller obliquity to the ecliptic. We now turn our attention to the transition between the core being coupled and uncoupled from the mantle, which is in many ways analogous to the Cassini State tran- sition that the mantle experienced somewhat later, with the difference that the former involved the core and the mantle, and the latter the mantle and the lunar orbit. Meyer and Wisdom (2011) calculated that the core should have uncoupled at around 26RE, as- suming that the mantle was not in hydrostatic equilibrium but had a figure identical to the present one. As explained in ´Cuk et al. (2016), this is probably the smallest Earth- Moon distance at which a non-hydrostatic mantle would be a reasonable assumption. However, in our simulations of the Moon at this distance from Earth, we did not exactly reproduce Meyer and Wisdom (2011) result, and we kept getting an uncoupled core in quasi-Cassini State 2 (note that the lunar mantle is at this point in Cassini State 1 rel- ative to the orbit). We tried simulations at even smaller Earth-Moon distances, but even for a = 18RE we were getting a decoupled core (Figs. 4 and 5), using (CC −AC )/CC = 2.5×10−4 (as elsewhere in the paper). However, when we reduced the semimajor axis to a = 15RE, the core in the simulation is now coupled to the mantle (Figs. 6 and 7). Note that these simulations only intended to demonstrate that the core-mantle coupling is possible for small Earth-Moon distances, and are not in any way an accurate repre- sentation of the Moon's dynamics at the time, as the mantle was likely not rigid at this time, and we did not try to accurately match Earth's rotation and oblateness which af- fect the Moon's orbital precession. We conclude that the criterion used by Meyer and Wisdom (2011) for the core-mantle coupling, ωC > Ω, is approximate and cannot be used to accurately determine the dis- tance at which the lunar core dynamically decoupled from the mantle. Numerical inte- grations suggest that the decoupling probably happened at lunar semimajor axis some- what smaller than 26RE estimated by Meyer and Wisdom (2011). This is a relatively minor discrepancy, as the nutation period of the core is about 50% and 33% of the man- tle precession period in our Figs. 4 and 6, respectively. Therefore, our simulations sug- gest that dynamical decoupling happens when the core nutation rate is about twice the nodal precession rate. A more accurate determination would require many more simu- lations and is of limited interest to this paper. Dumberry and Wieczorek (2016) have con- ducted an analysis of core-mantle coupling (but without dissipation) and found that the -- 11 -- manuscript submitted to JGR: Planets 45 40 35 30 25 20 15 10 5 0 6.5 6 5.5 5 4.5 ) ° ( y t i u q i l b O ) s y a d ( d o i r e p n p S i Core-mantle offset Core w.r.t. ecliptic Mantle w.r.t. ecliptic Mantle w.r.t. orbit Core Mantle 4 0 5000 10000 15000 20000 Time (years) Figure 4. Simulation of relaxation to equilibrium state using lunar semimajor axis a = 18RE , eccentricity e = 0, and inclination i = 8◦, with core flattening f = 2.5 × 10−4 . The core is largely decoupled from the mantle in this simulation. Top: Evolution of the ecliptic obliquities of the lunar core (red) and the mantle (blue and green, relative to ecliptic and the lunar orbit, respectively), as well as their relative offset (magenta). Bottom: Evolution of the spin periods of the solid Moon (black), lunar mantle (blue) and the core (red) in the same simulation. exact identity of the mantle precession rate and the FCN is in fact a resonance which forces a large core-mantle tilt; they find that FCN must be significantly faster than man- tle precession to enable dynamical coupling. Therefore, the main conclusion of this sec- tion is that, assuming a rigid lunar mantle with the present shape, the core should be dynamically decoupled from the mantle for lunar semimajor axes a > 25RE that are relevant for the Cassini State Transition and lunar paleomagnetism. -- 12 -- ) ° ( y t i u q i l b o d e j t c e o r P 50 40 30 20 10 0 -10 -20 -30 -40 manuscript submitted to JGR: Planets Mantle Core 50 100 150 200 250 300 350 400 450 500 Time (years) Figure 5. The very end of the simulation shown in Fig. 4. The lines plot a projection of the obliquity vectors of the core (red) and mantle (blue) on one of the axis in the polar representa- tion. It is clear that the mantle and the core spin axes precess around the ecliptic normal with very different phases, indicating dynamical decoupling between the core and the mantle. 4 Effects of the Core on the Lunar Dynamical Evolution In order to have a self-consistent model of past lunar core dynamics, we need to re-examine the dynamical history of the lunar mantle with the core's influence included. Lunar inclination damping due to obliquity tides, as modeled by ´Cuk et al. (2016), di- rectly depends on the mantle's rotational state, which is in principle affected by the pres- ence of the core. It is important to revisit these calculations and determine whether the results obtained by modeling the Moon as a solid body are still valid when core-mantle interaction is included. However, due to the limitations of our current integrator rela- tive to that used by ´Cuk et al. (2016), we are not able to directly integrate the lunar tidal evolution, as complete simulations would take unreasonably long time. In this section we will instead present a series of simulations of relaxation to equi- librium of the lunar core and mantle. These simulations are designed to follow the evo- lutionary track in semimajor axis and inclination found by ´Cuk et al. (2016). We started these simulations with a sub-synchronous Moon, let it damp to a stable state, and recorded -- 13 -- manuscript submitted to JGR: Planets Core w.r.t. ecliptic Mantle w.r.t. ecliptic Core-mantle offset Mantle w.r.t. orbit Core Mantle 0 2000 4000 6000 Time (years) 8000 10000 ) ° ( y t i u q i l b O ) s y a d ( d o i r e p n p S i 50 40 30 20 10 0 25 20 15 10 5 0 Figure 6. Simulation of relaxation to equilibrium state using a = 15RE, e = 0, i = 8◦, with core flattening f = 2.5 × 10−4. Free core nutation is fast enough to couple the core to the mantle. Top: Evolution of the ecliptic obliquities of the lunar core (red) and the mantle (blue and green, relative to ecliptic and the lunar orbit, respectively), as well as their relative offset (magenta). Bottom: Evolution of the spin periods of the lunar mantle (blue) and the core (red) in the same simulation. the final rotational state of the core and the mantle. While this is not a model of lunar orbital history independent of ´Cuk et al. (2016), is an important test of their model. If the rotational state of the mantle, especially the obliquity, is fundamentally changed by the presence of the core, the results of ´Cuk et al. (2016) would not be valid. On the other hand, if the simulations with the core-mantle interactions included produce lunar obliq- uities close to those in the core-less model, the overall nature of the lunar tidal evolu- tion is likely unchanged by core-mantle interactions. -- 14 -- ) ° ( y t i u q i l b o d e j t c e o r P 15 10 5 0 -5 -10 -15 manuscript submitted to JGR: Planets Core Mantle 50 100 150 200 250 300 350 400 450 500 Time (years) Figure 7. The very end of the simulation shown in Fig. 6. The lines plot a projection of the obliquity vectors of the core (red) and mantle (blue) on one of the axis in the polar repre- sentation. It is clear that the mantle and the core spin axes precess around the ecliptic normal in-phase, which indicates that the core is in quasi-Cassini State 1. In Fig. 8, we plot the results of our equilibrium solutions along the track of the lu- nar tidal evolution found by ´Cuk et al. (2016). We generated these solutions by plac- ing the Moon at a sequence of semimajor axes, with the spacing of 1 RE. At each semi- major axis, the Moon was given orbital inclination based on the orbital evolution track taken from Figure 4 in ´Cuk et al. (2016) (the same data are plotted by small red dots in Fig. 8), while the eccentricity was set to e = 0.01. The solid blue circles show the end-states of the simulations after 1 Myr (extended to 2 Myr for the two simulations at 30-31 RE which showed slowest convergence) averaged over the final 1000 years. Due to short-period variations in orbital elements, our initial conditions and averaged final states do not perfectly match in either inclination or semimajor axis, causing the blue circles to be slightly misaligned with the red dots they track. In panel B, we compare our solutions for the mantle obliquity for the specific a and i plotted in panel A (solid blue circles) to the core-less Moon's obliquity found by ´Cuk et al. (2016) (small red dots). The solid black line plots the Cassini State obliquities cor- -- 15 -- manuscript submitted to JGR: Planets A B C 35 30 25 20 15 10 90 80 70 60 50 40 30 20 10 ) ° ( n o i t a n i l c n I ) ° ( y t i u q i l b o r a n u L ) r a e y / d a r ( n p S i r a n u L 300 250 200 150 100 50 24 26 30 34 32 28 36 Semimajor axis (RE) 38 40 Figure 8. Equilibrium solutions for the mantle spin state using our core-mantle code (solid blue circles) along the lunar tidal evolution track found by ´Cuk et al. (2016) (small red dots). Panel A plots the lunar orbital inclination as a function of semimajor axis, while the panels B and C plot the corresponding lunar (i.e. mantle) obliquity (with respect to lunar orbit) and spin rate, respectively. The thin black lines in panels B and C plot the obliquity and spin rate calcu- lated on the assumption of the Moon being in synchronous rotation and Cassini State while on the a-i track shown in panel A. The open blue squares in panel C indicate that the lunar mantle was in synchronous rotation at the end of the simulation. -- 16 -- manuscript submitted to JGR: Planets responding to the inclinations in panel A (unlike the numerical simulations, the Cassini State calculation, by definition, assumes synchronous rotation). In panel C we plot the corresponding rotational rates from our mantle equilibrium solutions (solid blue circles) and the ´Cuk et al. (2016) core-less Moon simulation (small red dots), with the thin black line plotting the synchronous state. Our solutions clearly show similarity to the previously found evolution, including the divergence from a Cassini State. However, while the divergence from the Cassini state for a core-less moon was limited to the immediate aftermath of the Cassini State Tran- sition, our simulations suggest that the Moon was in non-synchronous rotation for all semimajor axes in the a = 25 −40RE range. Particularly interesting is the divergence from ´Cuk et al. (2016) results in semimajor axis range 26-29 RE, where our mantle is in a non-synchronous state, while ´Cuk et al. (2016) find the Moon to be in synchronous rotation. There are two distinct reasons for this divergence. One is from the different ways the two solutions were calculated: ´Cuk et al. (2016) integrated a single continu- ous simulation of lunar tidal evolution, while we performed a number of short simula- tions that started with the Moon in non-synchronous rotation. Therefore, in this seg- ment of ´Cuk et al. (2016) simulations the Moon is likely in a quasi-stable Cassini State 1, which is temporarily stable against Earth's torques, but would not be re-established if it were to be broken by an outside factor (such as a large impact). In the ´Cuk et al. (2016) simulation, the synchronous rotation is broken by an annual libration resonance at 29.7 RE, after which the Moon settles in a sub-synchronous state adjacent to Cassini State 2 (their Extended Data Figs. 8 and 9). In our core-mantle simulations the Moon at a = 27 − 29RE is in a sub-synchronous state adjacent to Cassini State 1 (Fig. 8), which apparently does not exist in the ´Cuk et al. (2016) model (their Extended Data Figs. 7 and 8). Apart from the divergence prior to the Cassini State transition, we also observe that our simulations at semimajor axes 35-40 RE have the Moon in a non-synchronous state, while the ´Cuk et al. (2016) find the Moon in the synchronous Cassini State 2 at this dis- tance (beyond 40 RE we also find that the Moon is mostly in synchronous rotation). The main reason for the prevalence of non-synchronous rotation in our simulations is that the core, which lags behind the mantle in its spin rate, exerts a relatively substantial spin- changing torque on the mantle. This can move the equilibrium spin value outside of the synchronous spin-orbit resonance, producing a stable sub-synchronous state. While for -- 17 -- manuscript submitted to JGR: Planets some semimajor axes a = 30−35RE the synchronous state itself may be unstable (due to too-high Cassini State obliquity; Beletskii, 1972), we find that in the a = 35−40RE range the synchronous rotation is sometimes reachable by spin-down but not by spin- up (i.e. initial conditions may determine the end point rotation state reached by the Moon). In general, the presence of the core drives the sub-synchronous states further away from synchronicity, and as a result the corresponding Cassini States have somewhat higher obliquities. This can be understood by recognizing that the pole of a slower-rotating Moon will precess more quickly, so to match the precession rate of the orbit, axial precession must be slowed down by high obliquity. On the other hand, we find that the Cassini State obliquities for synchronous cases (such as our a = 40RE simulation) are lower when the core is included in the simulation. The reason for this change is that the core, which is significantly misaligned with the mantle, induces additional librations of the mantle in synchronous rotation. Librations make the precession slower (as the Moon presents a dif- ferent average figure to Earth's torques), requiring lower obliquities to reach the same precession rate as in the non-librating case. Despite the modification of synchronous rotation caused by the presence of the fluid core, we find that the lunar orbital evolution history found by ´Cuk et al. (2016) is un- likely to be fundamentally changed. Using the lunar inclination model published by ´Cuk et al. (2016), we find that the lunar mantle has had a large obliquity for the significant part of the Moon's history. Obliquities and associated tides in our model are sometimes lower, sometimes higher than those found by ´Cuk et al. (2016), but there is every rea- son to think that large damping of lunar eccentricity took place. We also note that core- mantle friction also damps obliquity relative to the Laplace plane (Rochester, 1976), which leads to damping of inclination (as the obliquity is forced). Further more extensive nu- merical simulations of the full evolution of lunar spin and orbit in the presence of the core are needed to determine the exact contributions of different inclination-damping mech- anisms. Our results are dependent on our choice for the functional form of CMB friction (Eq. 4), in which the torque is directly proportional to the difference between core and mantle spin rates, with a constant damping timescale of about 120 yr. Other functional forms are possible, including those that are quadratic in relative core-mantle motion (Yo- der, 1981). Given that the current core-mantle tilt is small, these alternative forms would -- 18 -- manuscript submitted to JGR: Planets produce much shorter timescales for damping of core-mantle motion when the lunar obliq- uity to ecliptic was high. We tried such a functional form of CMB friction in some sim- ulations, in order to determine whether such strong friction could couple core to the man- tle. Surprisingly, we find that beyond about 30RE the CMB torque on the mantle is too strong for mantle to stay in the Cassini State, and the Moon evolves into a very slow- rotating, high-obliquity state with the core and the mantle still decoupled. We expect that this state would persist until the lunar inclination is damped, ultimately allowing core-mantle coupling in the absence of mantle's forced obliquity. As a substantial lunar inclination has survived the Cassini State transition, this puts an upper limit on the in- tensity of past CMB friction, and may indicate that past quadratic CMB friction is not consistent with the present lunar orbit. However, our model includes a number of ap- proximations, and we only explored relatively short-term dynamics of the core and the mantle at fixed Earth-Moon distances; full-scale evolutionary integrations are necessary to fully constrain long-term inclination damping due to quadratic CMB friction. 5 Past Rotational State of the Core For the purposes of the generation of the lunar magnetic field, we are chiefly in- terested in the rotational state of the liquid core. In Fig. 9 we plot the final rotational parameters of the core in the simulations shown in Fig. 8. The solid squares and circles plot the results of our simulations featuring the core, while small dots plot the ´Cuk et al. (2016) simulations of a core-less Moon. In the top panel we plot the core and man- tle obliquities relative to the ecliptic, a convenient system of reference because the core's spin axis is close to the ecliptic for the duration of the simulations. Unlike the core, the mantle has a much larger tilt to the ecliptic, both due to large orbital inclination (most important before the Cassini State transition) and the large forced obliquity (dominant after the Cassini State transition). As the core is in a non-synchronous rotation, we also plot the predicted core rotation rate based on that of the mantle using the solid magenta line. This rate is calculated as ωC = ωM cos(ψCM ), where ωC and ωM are the core and mantle spin rates, and ψCM is the core-mantle obliquity. This calculated spin rate as- sumes that the core spin matches the component of the mantle's angular velocity par- allel to the core's spin axis. Comparison between these predicted and numerically sim- ulated spin rates of the core (Fig. 9, bottom panel) shows this to be a good approxima- tion for the core's final spin rate in the simulations. -- 19 -- manuscript submitted to JGR: Planets A B 90 80 70 60 50 40 30 20 10 300 250 200 150 100 50 ) ° ( c i t p i l c E o t y t i u q i l b O ) r a e y / d a r ( e t a r n p S i 0 24 26 28 30 32 34 36 38 40 Lunar semimajor axis (RE) Figure 9. Comparisons between the rotational parameters of the lunar mantle and the core in the simulations shown in Fig. 8. Obliquity to ecliptic is shown in the top panel, while the bottom panel plots the core and mantle rotation rates. Obliquity and spin rate of the mantle (solid blue circles) and the core (solid red squares) are determined by our full core-mantle simulations using cr-sistem. The solid-Moon integration of ´Cuk et al. (2016) is plotted with small black dots. In the bottom panel, thin magenta line plots the rotation rate ωM cos(ψCM ), where ωM is the mantle spin rate, and ψCM is the core-mantle obliquity. Figure 10 re-plots some of our results for the core in terms of quantities measured relative to the mantle, and extends the calculation to the present day using a semi-analytical model. Since the Moon was most likely in Cassini State 2 for a > 40RE, and direct sim- ulations take much longer for large semimajor axes, it is more efficient to calculate core and mantle rotational properties using a simplified model. We find the core obliquity as the solution to the equation: (cid:16) CC − AC CC (cid:17)ωM cos(ψCM ) sin(ψCM ) + Ω sin(ψCM − ǫM ) = 0 (6) -- 20 -- manuscript submitted to JGR: Planets 70 60 50 40 30 20 10 0 w.r.t. mantle w.r.t. ecliptic ) ° ( y t i u q i l b o e r o C i n p S l a i t n e r e f f i D e r o C 0.1 0.01 0.001 25 (ω M-ω C) / ω M 35 30 Lunar semimajor axis (RE) 45 40 50 55 60 Figure 10. Lunar core obliquity (top panel) and relative spin rate (bottom panel) in our model. The points for a < 50RE were determined by a full numerical model also shown in Figs. 8 and 9, while the solid line for a > 40RE was calculated using Eq. 6 and the semi-analytical model described in the text. In the top panel, the upper magenta lines plot the core-mantle obliquity, while the lower red lines plot the core's obliquity relative to the ecliptic. In the bottom panel, we plot the quantity (ωM − ωC)/ωM as a measure of relative difference between the spin rates of the core and the mantle. where Ω is the precession rate of the longitude of the Moon's ascending node, and ǫM is the obliquity of the mantle with respect to the ecliptic. We obtain Eq. 6 from Eq. 5 (in this paper) and Eq. 1 in Ward (1975), and assumed ωC = ωM cos(ψCM ) for the core's rotation rate. Eq. 6 can alternatively be derived from the model of Stys and Dumberry (2018), by setting the size of the inner core to zero. Eq. 6 is solved numerically by sim- ply advancing through obliquities and finding the value for which the sum of terms goes through zero. The mantle obliquity and lunar orbital precession are obtained from the semi analytical model of lunar tidal evolution described by Eqs. M1-M8 in the Meth- -- 21 -- manuscript submitted to JGR: Planets ods section of ´Cuk et al. (2016). We used i = 10◦ and e = 0.01 at a = 40RE as ini- tial conditions, and advanced the model forward, with Earth tidal Q set at lifetime av- erage of QE = 35, and the lunar tidal Q, as well as Earth and Moon tidal Love num- bers, set at present values QM = 38, kE 2 = 0.3 and kM 2 = 0.024 (Williams & Boggs, 2015). This model evolution gives us an inclination just over 5◦ at a = 60RE, consis- tent with the present lunar orbit. Thick lines in Fig. 10 plot the core obliquity and spin rate derived from this model. Fig. 10 shows that, while the numerical and semi-analytical obliquities mostly agree, there is a dramatic divergence for points at a = 43RE and a = 50RE, for which the numerical model does not predict synchronous rotation. This is related to the shifting of the equilibrium spin rate from exact synchronicity due to the CMB friction torque on the mantle. It appears that the equilibrium rotation is just at the outer limit of the syn- chronous spin-orbit resonance, making reestablishment of synchronous rotation sensitive to initial conditions and possible secondary resonances. Here we used a low lunar eccen- tricity (e = 0.01), but a more substantial eccentricity would shift the equilibrium spin value to higher values, counteracting the effects of CMB friction and shifting the equi- librium back into the synchronous rotation. Lunar eccentricity was likely affected by plan- etary orbital resonances at a = 46RE and a = 53RE ( ´Cuk, 2007), with the latter res- onance likely bringing the lunar eccentricity close to the present value. The actual his- tory of the lunar rotation depends on the timing of synchronous-lock-breaking impacts, the relative strength of satellite tides and spin-orbit friction, the time-evolution of lu- nar eccentricity and the (poorly understood but likely) influence of secondary resonances It is evident from Fig. 10 that, in our model, when the Moon had semimajor axis a < 45RE there was a large mutual obliquity between the lunar core and the mantle, and a significant difference between the rotation rates of the two. In order to compare the models of orbital and rotational dynamics to the record of lunar magnetism, we would need to know the absolute timing of the lunar tidal evolution. Unfortunately, there is currently no direct way to constrain the speed of early lunar tidal evolution. While the age of the Earth-Moon system implies Earth's long-term average tidal QE ≃ 34, the current tidal Q of Earth is about 12, meaning that there must have been at least some intervals of higher tidal Q in the past. Bills and Ray (1999) show that often contradic- tory claims of tidal periods captured in the geological record only concern the last bil- lion years or so, and the timeline of the early history of the Earth-Moon system is com- -- 22 -- manuscript submitted to JGR: Planets Table 1. Approximate times when specific lunar semimajor axes were reached for two different values of early Earth's tidal quality factor: Q = 34 (4.5 Gyr average) and Q = 100. For simplic- ity, we assumed that Earth had a Love number kE 2 = 0.3 (like today), and that the Moon formed exactly 4.5 Gyr ago. Lunar a Time for QE = 34 Time for QE = 100 30RE 35RE 40RE 45RE 4.45 Gya 4.35 Gya 4.4 Gya 4.2 Gya 3.8 Gya 4.1 Gya 3.6 Gya 2.5 Gya pletely unconstrained. Some models of ocean dissipation (Webb, 1982) indicate general reduction of Q over geological time, with tidal Q being a factor of several higher than now in the first billion years of the Solar System, but so far there is no widely accepted correspondence between lunar distance and time. In Table 1, we present two possible timelines for lunar tidal evolution as a func- tion of Earth's tidal Q during the system's early history. These numbers do not estab- lish a preferred time-obliquity correspondence, but rather illustrate that a wide range of reasonable parameters would predict a large lunar core-mantle obliquity at the epoch when the lunar magnetic field is thought to have been strongest, 4.25-3.6 Gyr ago (Tikoo et al., 2014; Weiss & Tikoo, 2014). In Table 1 we used a ≃ a−11/2Q−1 E (Murray & Der- mott, 1999), i.e. we assumed a constant Q for early Earth, and ignored lunar eccentric- ity and obliquity tides, as well as temporary halts of tidal evolution due to the Laplace Plane transition ( ´Cuk et al., 2016). The obliquities shown in Figs. 8-10 are model-dependent, and lunar tidal evolu- tion models that produced lower past lunar inclinations (Touma & Wisdom, 1994) would also result in smaller obliquities (although high obliquities close to the Cassini State tran- sition are inevitable). However, lower past lunar inclinations in models predating work by Chen and Nimmo (2013) are primarily an artifact of the obliquity tides being ignored. While Peale and Cassen (1978) considered tidal heating from obliquity tides, they did not investigate associated inclination damping; Williams et al. (2001) proposed that tidal heating during the Cassini State transition powered the lunar dynamo, but did not ad- -- 23 -- manuscript submitted to JGR: Planets dress orbital consequences. Therefore, pre-2013 models of lunar tidal evolution cannot be directly used to derive past lunar mantle obliquities. More recently, the model of Pahle- van and Morbidelli (2015) gets around the inclination damping at the Cassini State tran- sition by exciting lunar inclination through planetesimal flybys after a rapid early lunar orbital expansion. Since Pahlevan and Morbidelli (2015) propose that the Moon under- went a rapid orbital expansion in the first 10 Myr, and had a close-to-present inclina- tion at a ≃ 45RE after the encounters were over (≃ 100 Myr), it appears unlikely the Moon would have had large enough obliquity to power a dynamo at the "magnetic epoch" 4.25-3.6 Gya. While the effects of late planetesimal encounters need more exploration (we ignored them in this work as they happen before the Cassini State transition for tidal parameters we considered), it is likely that the Pahlevan and Morbidelli (2015) model is incompatible with a precession-driven dynamo, and that a different mechanism would be needed to power the ancient lunar dynamo. In our calculations, we ignored the presence of the solid inner core. While the size of the inner core is poorly constrained (Williams et al., 2014), it should have significant effects on the core dynamics by introducing new nutation frequencies (Dumberry & Wiec- zorek, 2016) and it may in some cases significantly change the tilt of the outer core (Stys & Dumberry, 2018). It is thought that the solid inner core grew over time (Scheinberg, Soderlund, & Schubert, 2015), so it would have been much smaller (or not present at all) during the early history of the Earth-Moon system that we study here. However, assum- ing a later formation for the inner core may not be viable, as some models of the mag- netic dynamo generation require the presence of the solid inner core (Stanley, Tian, Weiss, & Tikoo, 2017), so improved dynamical models will have to include a two-component core. Based on our results for the outer core, the inner core is unlikely to be synchronously locked as Stanley et al. (2017) have proposed, as friction with the sub-synchronous outer core would likely be a more important factor for determining the inner core's spin than Earth's tidal torque. Additionally, Stys and Dumberry (2018) have found that the free inner core nutation period is currently in the 10-40 yr range, making a large forced tilt due to resonant interaction with the 18.6 yr mantle precession period likely. Finally, we note that a magnetic field that is locked to the core, or changes its ge- ometry relatively slowly relative to the core, would not be stationary relative to the lu- nar mantle. Earth's core is practically co-rotating with the mantle and its spin axis is not significantly offset from that of the mantle (as the free core nutation is much, much -- 24 -- manuscript submitted to JGR: Planets faster than Earth's axial precession). Therefore, experience from Earth is not directly applicable to the Moon, and a lunar dipole field generated in the core would drift across the lunar surface due to different rotations of the core and the mantle. For example, at a = 40RE, when we find the core-mantle obliquity to be 25◦, a magnetic pole that is aligned with the core spin axis would sweep on the surface a circle at 65◦ latitude in a course of a lunar orbital period (≃ 15 days at that time). The pole of a dipole field that is significantly tilted relative to the core spin axis would trace a much more complex track on the timescales determined by the differential rotation of the core and the mantle. There- fore, directional magnetization of lunar rocks does not record a long-term orientation of the lunar magnetic field, but possibly only a snapshot taken at the critical points of the material's thermal evolution. Some of the recent work on the locations of proposed pa- leopoles does indicate a great variability in the inferred orientation of the lunar dynamo (Nayak, Hemingway, & Garrick-Bethell, 2017; Oliveira & Wieczorek, 2017), and relative motions of the core and the mantle may at least partially explain these findings. 6 Conclusions In this work we numerically modeled the rotational dynamics of the Moon's core and the mantle during the early history of the Earth-Moon system, and we report the following conclusions: 1. The lunar core was precessing independently of the mantle for practically all of lunar history. 2. Our numerical model also suggests that in the distant past the lunar core ro- tated significantly slower than the mantle, assuming that the main dissipative torque on the core comes from friction at the core-mantle boundary. 3. The presence of the small lunar core makes the non-synchronous rotation be- fore and after the Cassini State transition even more likely, and the CMB friction should have caused additional damping of lunar inclination, separate from obliquity tides. 4. For much of the early history of the Earth-Moon system there was a large mu- tual obliquity between the core and the mantle. This has direct implications for the gen- eration of the ancient lunar magnetic field, and we discuss plausible absolute timings of events in lunar tidal evolution. -- 25 -- manuscript submitted to JGR: Planets Finally, we note that this is only the first step in numerically modeling the dynam- ics of the lunar core. Due to numerical limitations, we only explored stationary solutions for the lunar rotation, and in the future we hope to produce a self-consistent lunar tidal evolution model that includes the effects of core-mantle interactions. We also hope that additional factors such as the presence of the solid inner core and the possible non-rigid behavior of the core-mantle boundary will be addressed in future work. Acknowledgments M ´C is supported by NASA Emerging Worlds award NNX15AH65G. We thank Math- ieu Dumberry and an anonymous reviewer for their extremely helpful reviews of the pre- vious version of this paper. The source code for our integrator CR-SISTEM is avail- able through the Astrophysics Source Code Library. References Beletskii, V. V. (1972, Nov). Resonance Rotation of Celestial Bodies and Cassini's Laws. Celestial Mechanics, 6 (3), 356-378. doi: 10.1007/BF01231479 Bills, B. G., & Ray, R. D. (1999). Lunar orbital evolution: A synthesis of re- cent results. Geophysical Research Journal , 26 , 3045-3048. doi: 10.1029/ 1999GL008348 Chen, E. M. A., & Nimmo, F. (2013, December). Tidal dissipation in the early lunar magma ocean and its role in the evolution of the Earth-Moon system. AGU Fall Meeting Abstracts, P51C-1747. Chen, E. M. A., & Nimmo, F. (2016, September). Tidal dissipation in the lunar magma ocean and its effect on the early evolution of the Earth-Moon system. Icarus, 275 , 132-142. doi: 10.1016/j.icarus.2016.04.012 Colombo, G. (1966, Nov). Cassini's second and third laws. Astronomical Journal , 71 , 891. doi: 10.1086/109983 ´Cuk, M. (2007, October). Excitation of Lunar Eccentricity by Planetary Resonances. Science, 318 , 244. doi: 10.1126/science.1146984 ´Cuk, M., Hamilton, D. P., Lock, S. J., & Stewart, S. T. (2016, November). Tidal evolution of the Moon from a high-obliquity, high-angular-momentum Earth. Nature, 539 , 402-406. doi: 10.1038/nature19846 Dumberry, M., & Wieczorek, M. A. (2016, July). The forced precession of the -- 26 -- manuscript submitted to JGR: Planets Moon's inner core. Journal of Geophysical Research (Planets), 121 , 1264-1292. doi: 10.1002/2015JE004986 Dwyer, C. A., Stevenson, D. J., & Nimmo, F. (2011, November). A long-lived lunar dynamo driven by continuous mechanical stirring. Nature, 479 , 212-214. doi: 10.1038/nature10564 Forest, E., & Ruth, R. D. (1990, May). Fourth-order symplectic integration. Physica D Nonlinear Phenomena, 43 , 105-117. doi: 10.1016/0167-2789(90)90019-L Garcia, R. F., Gagnepain-Beyneix, J., Chevrot, S., & Lognonn´e, P. (2011, Sep). Very preliminary reference Moon model. Physics of the Earth and Planetary Interi- ors, 188 (1), 96-113. doi: 10.1016/j.pepi.2011.06.015 Hood, L. L., Mitchell, D. L., Lin, R. P., Acuna, M. H., & Binder, A. B. (1999, Jan). Initial measurements of the lunar induced magnetic dipole moment using Lu- nar Prospector Magnetometer data. Geophysical Research Letters, 26 (15), 2327-2330. doi: 10.1029/1999GL900487 Hough, S. S. (1895, Jan). The Oscillations of a Rotating Ellipsoidal Shell Contain- ing Fluid. Philosophical Transactions of the Royal Society of London Series A, 186 , 469-506. doi: 10.1098/rsta.1895.0012 Lock, S. J., Stewart, S. T., Petaev, M. I., Leinhardt, Z., Mace, M. T., Jacobsen, S. B., & Cuk, M. (2018, April). The Origin of the Moon Within a Terres- trial Synestia. Journal of Geophysical Research (Planets), 123 , 910-951. doi: 10.1002/2017JE005333 Meyer, J., & Wisdom, J. (2011, January). Precession of the lunar core. Icarus, 211 , 921-924. Murray, C. D., & Dermott, S. F. (1999). Solar system dynamics. Nayak, M., Hemingway, D., & Garrick-Bethell, I. (2017, April). Magnetization in the South Pole-Aitken basin: Implications for the lunar dynamo and true polar wander. Icarus, 286 , 153-192. doi: 10.1016/j.icarus.2016.09.038 Oliveira, J. S., & Wieczorek, M. A. (2017, February). Testing the axial dipole hypothesis for the Moon by modeling the direction of crustal magneti- zation. Journal of Geophysical Research (Planets), 122 , 383-399. doi: 10.1002/2016JE005199 Pahlevan, K., & Morbidelli, A. (2015, November). Collisionless encoun- ters and the origin of the lunar inclination. Nature, 527 , 492-494. doi: -- 27 -- manuscript submitted to JGR: Planets 10.1038/nature16137 Pavlov, D. A., Williams, J. G., & Suvorkin, V. V. (2016, November). Determining parameters of Moon's orbital and rotational motion from LLR observations using GRAIL and IERS-recommended models. Celestial Mechanics and Dy- namical Astronomy, 126 , 61-88. doi: 10.1007/s10569-016-9712-1 Peale, S. J. (1969, April). Generalized Cassini's Laws. Astronomical Journal , 74 , 483. Peale, S. J., & Cassen, P. (1978, November). Contribution of tidal dissipation to lu- nar thermal history. Icarus, 36 , 245-269. Poincar´e, H. (1910, Jan). Sur la pr´ecession des corps d´eformables. Bulletin As- tronomique, Serie I , 27 , 321-356. Rochester, M. G. (1976, Jul). The secular decrease of obliquity due to dissipative core-mantle coupling. Geophysical Journal , 46 , 109-126. doi: 10.1111/j.1365 -246X.1976.tb01635.x Scheinberg, A., Soderlund, K. M., & Schubert, G. (2015, July). Magnetic field gener- ation in the lunar core: The role of inner core growth. Icarus, 254 , 62-71. doi: 10.1016/j.icarus.2015.03.013 Shimizu, H., Matsushima, M., Takahashi, F., Shibuya, H., & Tsunakawa, H. (2013, Jan). Constraint on the lunar core size from electromagnetic sounding based on magnetic field observations by an orbiting satellite. Icarus, 222 (1), 32-43. doi: 10.1016/j.icarus.2012.10.029 Stanley, S., Tian, B. Y., Weiss, B. P., & Tikoo, S. M. (2017, March). The Ancient Lunar Dynamo: How to Resolve the Intensity and Duration Conundrums. In Lunar and planetary science conference (Vol. 48, p. 1462). Stys, C., & Dumberry, M. (2018, Nov). The Cassini State of the Moon's Inner Core. Journal of Geophysical Research (Planets), 123 (11), 2868-2892. doi: 10.1029/ 2018JE005607 Tamayo, D., Burns, J. A., Hamilton, D. P., & Nicholson, P. D. (2013, March). Dy- namical Instabilities in High-obliquity Systems. Astronomical Journal , 145 , 54. Tikoo, S. M., Weiss, B. P., Cassata, W. S., Shuster, D. L., Gattacceca, J., Lima, E. A., . . . Fuller, M. D. (2014, October). Decline of the lunar core dy- namo. Earth and Planetary Science Letters, 404 , 89-97. doi: 10.1016/ -- 28 -- manuscript submitted to JGR: Planets j.epsl.2014.07.010 Touma, J., & Wisdom, J. (1994). Evolution of the earth-moon system. Astronomical Journal , 108 (5), 1943-1961. Touma, J., & Wisdom, J. (1994, March). Lie-Poisson integrators for rigid body dy- namics in the solar system. Astronomical Journal , 107 , 1189-1202. doi: 10 .1086/116931 Touma, J., & Wisdom, J. (2001, August). Nonlinear Core-Mantle Coupling. Astro- nomical Journal , 122 , 1030-1050. doi: 10.1086/321146 Tremaine, S., Touma, J., & Namouni, F. (2009, March). Satellite Dynamics on the Laplace Surface. Astronomical Journal , 137 , 3706-3717. Ward, W. R. (1975, August). Past orientation of the lunar spin axis. Science, 189 . Webb, D. J. (1982). Tides and the evolution of the earth-moon system. Geophysical Journal , 70 , 261-271. doi: 10.1111/j.1365-246X.1982.tb06404.x Weber, R. C., Lin, P.-Y., Garnero, E. J., Williams, Q., & Lognonn´e, P. (2011, Jan). Seismic Detection of the Lunar Core. Science, 331 (6015), 309. doi: 10.1126/ science.1199375 Weiss, B. P., & Tikoo, S. M. (2014, December). The lunar dynamo. Science, 346 , 1198. doi: 10.1126/science.1246753 Williams, J. G., & Boggs, D. H. (2015, April). Tides on the Moon: Theory and determination of dissipation. Journal of Geophysical Research (Planets), 120 , 689-724. doi: 10.1002/2014JE004755 Williams, J. G., Boggs, D. H., Yoder, C. F., Ratcliff, J. T., & Dickey, J. O. (2001, November). Lunar rotational dissipation in solid body and molten core. Journal of Geophysical Research (Planets), 106 , 27933-27968. doi: 10.1029/2000JE001396 Williams, J. G., Konopliv, A. S., Boggs, D. H., Park, R. S., Yuan, D.-N., Lemoine, F. G., . . . Zuber, M. T. (2014, July). Lunar interior properties from the GRAIL mission. Journal of Geophysical Research (Planets), 119 , 1546-1578. doi: 10.1002/2013JE004559 Williams, J. G., Turyshev, S. G., Boggs, D. H., & Ratcliff, J. T. (2006). Lunar laser ranging science: Gravitational physics and lunar interior and geodesy. Advances in Space Research, 37 , 67-71. doi: 10.1016/j.asr.2005.05.013 Wisdom, J., & Holman, M. (1991, October). Symplectic maps for the n-body prob- -- 29 -- manuscript submitted to JGR: Planets lem. Astronomical Journal , 102 , 1528-1538. doi: 10.1086/115978 Yoder, C. F. (1981, Dec). The Free Librations of a Dissipative Moon. Philosophical Transactions of the Royal Society of London Series A, 303 (1477), 327-338. doi: 10.1098/rsta.1981.0206 -- 30 --
1509.02707	1	1509	2015-09-09T10:26:44	Gravitational slopes, geomorphology, and material strengths of the nucleus of comet 67P/Churyumov-Gerasimenko from OSIRIS observations	[ "astro-ph.EP" ]	We study the link between gravitational slopes and the surface morphology on the nucleus of comet 67P/Churyumov-Gerasimenko and provide constraints on the mechanical properties of the cometary material. We computed the gravitational slopes for five regions on the nucleus that are representative of the different morphologies observed on the surface, using two shape models computed from OSIRIS images by the stereo-photoclinometry (SPC) and stereo-photogrammetry (SPG) techniques. We estimated the tensile, shear, and compressive strengths using different surface morphologies and mechanical considerations. The different regions show a similar general pattern in terms of the relation between gravitational slopes and terrain morphology: i) low-slope terrains (0-20 deg) are covered by a fine material and contain a few large ($>$10 m) and isolated boulders, ii) intermediate-slope terrains (20-45 deg) are mainly fallen consolidated materials and debris fields, with numerous intermediate-size boulders from $<$1 m to 10 m for the majority of them, and iii) high-slope terrains (45-90 deg) are cliffs that expose a consolidated material and do not show boulders or fine materials. The best range for the tensile strength of overhangs is 3-15 Pa (upper limit of 150 Pa), 4-30 Pa for the shear strength of fine surface materials and boulders, and 30-150 Pa for the compressive strength of overhangs (upper limit of 1500 Pa). The strength-to-gravity ratio is similar for 67P and weak rocks on Earth. As a result of the low compressive strength, the interior of the nucleus may have been compressed sufficiently to initiate diagenesis, which could have contributed to the formation of layers. Our value for the tensile strength is comparable to that of dust aggregates formed by gravitational instability and tends to favor a formation of comets by the accrection of pebbles at low velocities.	astro-ph.EP	astro-ph	Astronomy & Astrophysics manuscript no. 26379_ap November 7, 2018 c(cid:13)ESO 2018 Gravitational slopes, geomorphology, and material strengths of the nucleus of comet 67P/Churyumov-Gerasimenko from OSIRIS observations 5 1 0 2 p e S 9 . ] P E h p - o r t s a [ 1 v 7 0 7 2 0 . 9 0 5 1 : v i X r a O. Groussin1, L. Jorda1, A.-T. Auger1, 2, E. Kührt3, R. Gaskell4, C. Capanna1, F. Scholten3, F. Preusker3, P. Lamy1, S. Hviid3, J. Knollenberg3, U. Keller3, 5, C. Huettig3, H. Sierks6, C. Barbieri7, R. Rodrigo8, 9, D. Koschny10, H. Rickman11, 12, M. F. A'Hearn13, J. Agarwal6, M. A. Barucci14, J.-L. Bertaux15, I. Bertini16, S. Boudreault6, G. Cremonese17, V. Da Deppo18, B. Davidsson11, S. Debei17, M. De Cecco19, M. R. El-Maarry20, S. Fornasier14, M. Fulle21, P. J. Gutiérrez22, C. Güttler6, W.-H Ip23, J.-R. Kramm6, M. Küppers24, M. Lazzarin7, L. M. Lara22, J. J. Lopez Moreno22, S. Marchi25, F. Marzari7, M. Massironi16, 26, H. Michalik27, G. Naletto16, 18, 28, N. Oklay6, A. Pommerol20, M. Pajola16, N. Thomas20, I. Toth29, C. Tubiana6, and J.-B. Vincent6 (Affiliations can be found after the references) Received -- ; accepted -- ABSTRACT Aims. We study the link between gravitational slopes and the surface morphology on the nucleus of comet 67P/Churyumov- Gerasimenko and provide constraints on the mechanical properties of the cometary material (tensile, shear, and compressive strengths). Methods. We computed the gravitational slopes for five regions on the nucleus that are representative of the different morpholo- gies observed on the surface (Imhotep, Ash, Seth, Hathor, and Agilkia), using two shape models computed from OSIRIS images by the stereo-photoclinometry (SPC) and stereo-photogrammetry (SPG) techniques. We estimated the tensile, shear, and compressive strengths using different surface morphologies (overhangs, collapsed structures, boulders, cliffs, and Philae's footprint) and mechani- cal considerations. Results. The different regions show a similar general pattern in terms of the relation between gravitational slopes and terrain mor- phology: i) low-slope terrains (0 -- 20◦) are covered by a fine material and contain a few large (>10 m) and isolated boulders, ii) intermediate-slope terrains (20 -- 45◦) are mainly fallen consolidated materials and debris fields, with numerous intermediate-size boulders from <1 m to 10 m for the majority of them, and iii) high-slope terrains (45 -- 90◦) are cliffs that expose a consolidated mate- rial and do not show boulders or fine materials. The best range for the tensile strength of overhangs is 3 -- 15 Pa (upper limit of 150 Pa), 4 -- 30 Pa for the shear strength of fine surface materials and boulders, and 30 -- 150 Pa for the compressive strength of overhangs (up- per limit of 1500 Pa). The strength-to-gravity ratio is similar for 67P and weak rocks on Earth. As a result of the low compressive strength, the interior of the nucleus may have been compressed sufficiently to initiate diagenesis, which could have contributed to the formation of layers. Our value for the tensile strength is comparable to that of dust aggregates formed by gravitational instability and tends to favor a formation of comets by the accrection of pebbles at low velocities. Key words. Comets: individual: 67P/Churyumov-Gerasimenko -- Comets: general -- Accretion, accretion disks -- Methods: data analysis 1. Introduction Rosetta has been orbiting comet 67P/Churyumov-Gerasimenko (67P) since August 2014. The OSIRIS cameras (Keller et al. 2007) onboard this spacecraft have acquired hundreds of images of the surface with an unprecedented spatial resolution down to the decimeter scale (Sierks et al. 2015). The images reveal a complex nucleus surface made of smooth and hummocky ter- rains that are partially or entirely covered by dust or expose a consolidated material, pits, cliffs, and fractures from the hundred meter scale to the decimeter scale (Thomas et al. 2015). The na- ture and origin of these terrains and morphological features are far from being understood, but remain of paramount importance to better constrain the formation and evolution scenario of the nucleus of 67P and comets in general. This paper focuses on the link between the nucleus gravitational slopes and surface mor- phology to provide constraints on the nature of the cometary ma- terial and its mechanical properties in particular (tensile, shear, and compressive strengths). Gravitational slopes have only been measured on three cometary nuclei so far, 9P/Tempel 1 (Thomas et al. 2007), 81P/Wild 2 (Jorda et al. 2015), and 67P (Jorda et al. 2015). While the slopes of 9P are between 0◦ and 35◦, those of 81P and 67P cover a much wider range from 0◦ to >90◦, slopes exceeding 90◦ indicate overhangs. Beyond the different spatial resolution of the shape models used to compute the gravitational slopes for these three bodies, the differences between 9P on one side and 81P and 67P on the other side are interpreted as an aging effect by Jorda et al. (2015). Following the scenario described by these authors, comets that have spent more time in the inner solar sys- tem like 9P have been smoothed and have a narrower range of gravitational slopes than comets that have spent less time in the inner solar system like 81P and 67P. This planation process of Article number, page 1 of 18 A&A proofs: manuscript no. 26379_ap the nucleus was also proposed by Basilevsky & Keller (2007) for comets 19P/Borrelly, 81P, and 9P. The tensile, shear, and compressive strengths of the cometary material have been estimated by several methods, including the Deep Impact experiment, comet breakup observations, labora- tory experiments, and theoretical modeling (Table 1). Biele et al. (2009) compiled and discussed these different estimates in an excellent review paper. From the Deep Impact experiment, the shear strength was estimated to be <65 Pa by A'Hearn et al. (2005), but might be any value between 0 and 12 kPa according to Holsapple & Housen (2007); this is an uncertainty of three orders of magnitude. From comet breakup, the tensile strength was estimated to be 5 Pa by Asphaug & Benz (1996) from the encounter of comet Shoemaker-Levy 9 with Jupiter, to 100 Pa for a sun-grazing comet with a radius of 1 km (Klinger et al. 1989). Toth & Lisse (2006) and Davidsson (2001) estimated that a tensile strength of <100 Pa and 1-53 Pa, respectively, is sufficient to keep a comet nucleus stable against its rotational breakup. From laboratory experiments, the compressive strength of cometary material analogs (water ice and dust mixture) was estimated to be between 20 kPa and 1 MPa (Jessberger & Kot- thaus 1989; Bar-Nun et al. 2007), while the tensile strength was estimated to be 200 -- 1100 Pa for homogeneous SiO2 dust sam- ples (Blum et al. 2006) and down to 1 Pa for dust-aggregate sam- ples (Blum et al. 2014). From modeling, the tensile strength of fluffy silicate dust/ice material was estimated to be 270 Pa by Greenberg et al. (1995) and to be 5000 Pa by Biele et al. (2009) for the same material made of water ice alone. Again from mod- eling, the compressive strength was estimated to be 6500 Pa for porous icy grains (Sirono & Greenberg 2000). Depending on whether the material is consolidated or un- consolidated, these strength estimates (Table 1) depend, or not, on the scale at which they were measured. While the strength of unconsolidated material might be scale invariant, that of con- solidated material follows a typical d−q power law, where d is the scale and q is the power exponent, with q∼0.6 for water ice (Petrovic 2003). From the kilometer to the millimeter scale, the strength of consolidated material can thus change by three orders of magnitude, still less than the above ranges, which cover up to five orders of magnitude. The different estimates clearly are a priori difficult to reconcile with each other, and large uncertain- ties remain on the tensile, shear, and compressive strengths of the cometary material. Section 2 presents the data, shape models, and methods used in this paper. Section 3 discusses the link between gravitational slopes and surface morphologies for different types of terrains and regions on the nucleus. In Sect. 4 we estimate the tensile, shear, and compressive strengths of the cometary material. Dis- cussions and conclusions are presented in Sect. 5. 2. Data, shape models, and gravitational slopes All the images shown in this paper were acquired with the Nar- row Angle Camera (NAC) of the Optical, Spectroscopic and Infrared Remote Imaging System (OSIRIS) onboard Rosetta (Keller et al. 2007) since August 2014. Their spatial resolution varies between 18 cm pix−1 and 1.8 m pix−1. For this work we used two different shape models of the nu- cleus of 67P, computed from OSIRIS images. The first shape model was computed by Jorda et al. (2015) using the stereo- photoclinometry technique (SPC). The second shape model was computed by Preusker et al. (2015) using the stereo- photogrammetry technique (SPG). The SPC shape model was resampled to match the resolution of the SPG shape model. The Article number, page 2 of 18 Digital Terrain Models (DTMs) extracted from the SPC and SPG shape models have a horizontal sampling of 2 m and a typical vertical accuracy at the decimeter scale. We computed the local gravitational slopes for these two shape models, including the effects of the nucleus rotation and assuming a uniform density inside the nucleus. Details on the method are provided in Jorda et al. (2012). The error on the grav- itational slope is estimated to be 5◦. When not specified, the term slope in this paper always refers to the gravitational slope. 3. Relation between gravitational slopes and surface morphologies The geomorphology of the nucleus surface is diverse and con- strained by several processes related to gravity and cometary ac- tivity (Sierks et al. 2015; Thomas et al. 2015). The link between the different types of terrains (smooth, hummocky, consolidated material, dust covered, with or without boulders, etc.) and their gravitational slope is important to better constrain the processes in play and the nature of the cometary material. In this section we study this relationship for five regions on the nucleus that are representative of the different morphologies observed on the sur- face. We refer to El-Maarry et al. (2015a) for the definition of the regions. 1. The Imhotep region (Fig. 1) -- This region presents a wide variety of terrains and morphologies. The most remarkable ones are the smooth terrains, the largest of which extend over 0.8 km2 , and the roundish features observed near the gravi- tational low of the region. 2. The Ash region (Fig. 2) -- This region is mostly covered by dust that is spatially unresolved at the decimeter scale. It shows several debris fields that are made of boulders, which are located at the feet of steep walls exposing a consolidated material. A large depression of 370 m width is visible on the left side (noted A in Fig. 2). 3. The Seth region (Fig. 3) -- This region is dominated by cir- cular depressions, most of them being accumulation basins with an opening toward a lower basin. Debris accumulates at the feet of the steep walls of basins. A large (∼200 m) and deep pit dominates the bottom right part of the region (noted A in Fig. 3). 4. The Hathor region (Fig. 4) -- This region is dominated by cliffs with a maximum height of ∼900 m, exposing a consol- idated material. Many anisotropies (e.g., fractures) are visi- ble, organized in two main directions roughly perpendicular (Thomas et al. 2015). 5. The Agilkia region (Fig. 5) -- Agilkia is a large "super region" corresponding to the nominal landing site and includes the regions of Hatmehit, Ma'at, Nut, and Maftet. The Agilkia re- gion is dominated by a very large depression of 800 m width on its right side (noted A in Fig. 5). This region shows a vari- ety of exposed consolidated material, dust-covered material, and boulder fields. The histogram of gravitational slope angles computed for each region is shown in Fig. 6. Each region has a unique dis- tribution of gravitational slopes, related to its unique topogra- phy. There are similarities between the different distributions, however. All regions cover a wide range of slopes from 0◦ to almost 100◦. All distributions show a peak (Fig. 6, right panel). The peak of the distribution varies between 2◦ for Imhotep and Groussin, Jorda, Auger et al.: Gravitational slopes, geomorphology and material strengths of the nucleus of comet 67P 64◦ for Hathor. The peak value is a good indication of the overall flatness of the terrain, lower values indicating a flatter terrain on average. Secondary peaks, bumps, or shoulders are also visible; they are related to steep walls (e.g., bump around 55◦ on Seth and Imhotep) or smooth terrains (e.g., shoulder around 20◦ on Imhotep and Agilkia). Slopes in excess of 90◦ are indicative of overhangs and are only visible on the Hathor cliffs; overhangs cover 0.1% of the total Hathor area. Figure 7 allows a direct comparison between the SPC and SPG gravitational slopes. This is important to better estimate the error on the slope. The overall shape of the SPC and SPG slopes distributions is similar. In particular, the peaks are at the same position. The notable exception is Hathor, with a peak at 64◦ for the SPC slopes and 70◦ for the SPG slopes. This leads to a more general comment, which is that the slope distributions of SPC and SPG differ for slopes steeper than 60◦. The SPC shape model tends to systematically underestimate the fraction of the surface with steep slopes compared to the SPG shape model. This exercise shows that slopes are robust up to 60◦, with an error of ±5◦ as given in Sect. 2, but larger uncertainties affect slopes above 60◦, with an error up to 20◦ in some cases. Figures 1 - 5 illustrate that all regions show a similar general pattern in terms of the relation between gravitational slope and terrain morphology. A sketch of this general pattern is shown in Fig. 8: -- Low-slope terrains. Terrains with slopes in the range 0 -- 20◦ are covered by a spatially unresolved material, that is, a ma- terial made of particles smaller than 20 cm that we call the fine material in this paper. A few large isolated boulders are visible on these terrains, with a typical size larger than 10 m. -- Intermediate-slope terrains. Terrains with slopes in the range 20 -- 45◦ are mainly fallen consolidated materials and debris fields, with numerous intermediate-size boulders from <1 m to 10 m for the majority of them (see also Pajola et al. 2015). These terrains are covered by a dust deposit of variable thick- ness, which partially hides some boulders. Most of these ter- rains are located at the feet of high-slope terrains. -- High-slope terrains. Terrains with slopes in the range 45 -- 90◦ are cliffs that expose a consolidated material and do not show boulders or a fine material. These terrains probably show the bare nucleus. This general pattern is very similar to what we find on Earth, particularly in young mountains like the Alps, where boulder fields are frequently observed at the feet of cliffs; theses fields are a result of cliff collapse. The sublimation of ices triggers ero- sion, most likely exacerbated by fractures; gravity controls the collapse like it does on Earth. The presence of large boulders on slopes lower than 20◦ is intriguing (yellow circles in Figs. 1 - 5). These boulders are usu- ally isolated, far from high-slope terrains, and are large, tens of meters. They are too large to be lifted by gas drag resulting from regular cometary activity, this mechanism only applies to boulders smaller than the meter scale (Groussin & Lamy 2003; A'Hearn et al. 2011; Kelley et al. 2013; Gundlach et al. 2015). They could instead be air falls from outburst events, which are limited in time, but much stronger in terms of released energy. Such events are sporadic and spatially localized, however; they may not be frequent enough to explain the presence of large boulders in all the regions we studied. An alternative solution is that large boulders are leftovers from previous basins and de- pression edges, when they were smaller and less eroded. The transition between intermediate- and high-slope terrains is sharp. This indicates that high-slope terrains have a steeper slope than the angle of repose, which can be estimated to be θrepose = 45±5◦, a typical value for gravel on Earth (Julien 1995). The transition between low- and intermediate-slope terrains is softer, with several examples of boulder fields on slopes lower than 20◦ (white circles in Fig. 3) and of smooth terrains on slopes steeper than 20◦ (red circles in Figs. 1 - 5). Boulder fields that end on slopes lower than 20◦ on Seth (white circles in Fig. 3) could result from a progressive erod- ing and degradation process of the nucleus. These boulder fields were probably rock falls, at the feet of the previous emplacement of cliffs, as the above isolated boulders. Following a slow degra- dation process that involves fractal fragmentation into small pieces and progressive covering by dust deposits, they now ap- pear on low-slope terrains and are partially hidden (Pajola et al. 2015). The same degradation process can explain that a fine material of variable thickness partially hides boulder fields on intermediate-slope terrains. Smooth terrains on slopes steeper than 20◦ (red circles in Figs. 1 - 5) could be ancient boulder fields that were able to retain the dust deposit on a slope steeper than that of typical smooth terrains. The origin of the dust could be deposits from regular cometary activity or products of the degradation process of cliffs and boulders. 4. Tensile, shear, and compressive strengths of the cometary material 4.1. Definition of strengths There are three types of strength for a given material: the tensile strength σT , the shear strength σS , and the compressive strength σC. They define the ability of a material to withstand mechanical constraints. Usually, σT < σS < σC. Depending on the nature of the material, consolidated or unconsolidated, the strength de- pends, or not, on the scale at which it is measured (Sect. 1). A review on scaling effects on structural strength can be found in Bažant (1999). As explained in the introduction, large uncertainties remain in the values of σT , σS , and σC for the cometary material, some- times of several orders of magnitude (e.g., Biele et al. 2009). We here intend to provide additional constraints on the strengths of the cometary material; more precisely, the tensile strength of overhangs and collapsed structures, the shear strength of fine ma- terials, boulders and Hathor cliffs, and the compressive strength of fine materials and consolidated materials. 4.2. Tensile strength of overhangs and collapsed structures The tensile strength can be estimated from overhangs. From sim- ple mechanics (e.g., Tokashiki & Aydan 2010), the failure of an overhang of rectangular shape due to bending will occur if the following condition is fulfilled: L2 H σT < 3γ (1) where γ (N m−3) is the unit weight, L (m) the length of the over- hang, and H (m) its height as defined in Fig. 9. Equation (1), which is independent of the width X of the overhang, can be translated into σT < 3ρg L2 H (2) Article number, page 3 of 18 A&A proofs: manuscript no. 26379_ap where ρ = 470 kg m−3 is the density of the material (Sierks et al. 2015) and g is the local gravity. Several examples of overhangs or collapsed structures are shown in Figs. 10 and 11. Figure 10 shows two large structures that collapsed, indicating that the tensile strength was exceeded. Using the DTM of the area, we estimated the length and height of the largest structure to be L = 100 m and H = 30 m. From Eq. (2) with g = 2 × 10−4 m s−2 , this gives an upper limit for the tensile strength σT < 94 Pa. In Fig. 11, several overhangs are visible with an estimated length of 10 m and a height of 5 m. The pres- ence of mass wasting in the form of boulders at the feet of these overhangs indicates that they are close to breaking and thus good estimates of the tensile strength. This argument is also supported by the fact that some of these boulders have a size (∼10 m) sim- ilar to that of the overhangs themselves. From Eq. (2) we derive σT = 5.6 Pa for these overhangs. This determination of the tensile strength is not accurate, mainly because of the uncertainty in the geometry (length and height). Taking these uncertainties into account, the tensile strength of small overhangs (∼10 m) is most likely in the range 3 -- 15 Pa, and the tensile strength of large collapsed structures (∼100 m) is lower than 150 Pa. 4.3. Shear strength of fine materials, boulders, and Hathor cliffs The shear strength can be calculated with Eq. (3), where θ is the slope angle on which the boulder is located, m (kg) is the mass of the considered boulder, and A (m2) is the contact area of the boulder with the terrain underneath. The shear strength results from friction and cohesion, and we cannot separate these two physical quantities in this study. σS = mg sin θ A (3) For a boulder of radius r, Eq. (3) translates into Eq. (4) with A = π(r cos ϕ)2 (Fig. 12). σS = 4 3 πr3ρg sin θ π(r cos ϕ)2 = 4rρg sin θ 3 cos2 ϕ (4) The shear strength is constrained by the largest boulders on the highest gravitational slopes, more precisely. those with the highest r sin θ value. From Fig. 13, adapted from Auger et al. (2015), the highest value for r sin θ is 5.2 m and corresponds to a boulder with radius r = 11.5 m located on a gravitational slope θ = 26.8◦. Assuming an area of contact of 1 % of the total boul- der surface, defined with ϕ = 80◦ in Eq. (4), we obtain a shear strength σS = 22 Pa. Because of the uncertainty on the radius (2 m), slope angle (5◦), and the fact that the area of contact may be larger than 1 % (ϕ < 80◦), the shear strength of the boulders and fine material on which they stand is most likely in the range 4 -- 30 Pa. Figure 13 also illustrates that there are indeed no boul- der on slopes steeper than the angle of repose θrepose = 45 ± 5◦ (Sect. 3). A lower limit of the shear strength of the Hathor cliffs is pro- vided by the lateral pressure at the bottom of these cliffs, given by Eq. (5), where h is the height of the cliff. σS ≥ ρgh(1 − sin θrepose) For a maximum height h = 900 m of the Hathor cliffs, this gives σS ≥ 30 Pa. This value is an approximation since the cliff does not have a slope of 90◦ from top to bottom, meaning that it is not (5) Article number, page 4 of 18 perfectly vertical, and the gravity changes from top to bottom. Nevertheless, it shows that due to the low gravity, a low strength is sufficient to withstand a high cliff. 4.4. Compressive strength of fine and consolidated materials The compressive strength can be estimated from the footprints left by the lander Philae after it bounced on the nucleus surface at the first nominal landing site (Fig. 14). We can derive the com- pressive strength with Eq. (6), where F (N) is the forced applied to a surface of area A (m2), m (kg) is the mass of Philae, v (m s−1) is the lander impact velocity, and d (m) is the depth at which the lander penetrated the surface material. σC = F A = 1 2 mv2 dA (6) The depth of the largest footprint, which has a diameter of ∼5 pix or 150 cm, was estimated to be 20 cm using a shape from shading analysis. This footprint was created by Philae when the first leg touched the ground; we assume here that the three legs did not touch the ground simultaneously. With a mass of 100 kg for the lander, an impact velocity v = 1 m s−1 and a contact area of 0.016 m2 per foot (each foot is made of two disks of 10 cm diameter), we derive a compressive strength of 15.6 kPa for the surface material. This value is an upper limit since the presence of footprints indicates that the impact process exceeded the com- pressive strength of the surface material. Moreover, as we dis- cuss in Sect. 5.4, the penetration of Philae of the surface mate- rials was blocked by a hard layer under the surface, which also leads to overestimating the compressive strength. A more reliable estimate can be given by taking into ac- count mechanical considerations. In practice, the compressive strength of a consolidated material is always larger than its ten- sile strength by typically one order of magnitude (e.g., rocks like sandstone), so that we can reasonably assume σC ∼ 10σT (Sheo- rey 1997). This only applies to consolidated materials, however, such as overhangs or collapsed structures, but not to a uncon- solidated material like the fine material on the surface (Blum & Schräpler 2004; Blum et al. 2006). From the tensile strength cal- culated in Sect. 4.2, we therefore derive a compressive strength of the consolidated material that most likely is in the range 30 -- 150 Pa, with an upper limit of 1.5 kPa; this is well below the upper limit of 15.6 kPa derived for the fine surface materials. 5. Discussions and conclusions 5.1. Summary and comparison with other measurements Values for the tensile, shear, and compressive strengths are summarized in Table 1. Overall, the strengths are low and the cometary material can be considered as weak. Once scaled to the meter scale for comparison, our results agree well with most estimates from other authors who used observations, laboratory experiments, or modeling, in particular for the tensile strength. This scaling should be taken with caution, however, since con- solidated and unconsolidated materials may not follow the same scaling law. Our estimate for the shear strength is at the lower end of the possible range mentioned by Holsapple & Housen (2007). Concerning the compressive strength, the discrepancy with Jessberger & Kotthaus (1989) results from the fact that they measured the compressive strength of a mixture made of liquid water and dust with a high fraction of water (>80%). This mix- ture is cooled from room temperature to low temperature (253 K, Groussin, Jorda, Auger et al.: Gravitational slopes, geomorphology and material strengths of the nucleus of comet 67P 233 K, or 123 K, depending on their test case), and water ice forms as compact hexagonal ice, known to be hard (σC = 5 -- 25 MPa; Petrovic 2003). 5.2. Strength-to-gravity ratio The strengths are low on 67P, but the gravity is low as well. The ratio between the tensile strength at the 1 m scale (7 -- 34 Pa) and gravity (2× 10−4 m s−2) is 35 -- 170× 103 Pa s2 m−1 on 67P. On Earth, weak rocks like siltstone, with a density of ∼2600 kg m−3 and a porosity of 21 -- 41 %, have a typical tensile strength of 0.5 MPa. For this type of weak rocks, the ratio between the tensile strength and Earth's gravity is 51× 103 Pa s2 m−1. This remarkable agreement between the strength-to-gravity ratio on 67P and on Earth explains why the general pattern between grav- itational slopes and surface morphologies looks "familiar" to our eyes, from high-slope terrains (cliffs) to intermediate-slope terrains (mass wasting as boulder fields) and low-slope terrains (fine material). To first order, gravity shapes terrains in a similar way on 67P and on Earth. 5.3. Low strengths and surface morphologies related to activity The low strength of the cometary material may help to under- stand some of the surface morphologies, in particular those im- plying the rising up of a gas bubble, such as the roundish features of 67P (Imhotep region; Auger et al. 2015) and 19P (Brownlee et al. 2004), or the depressions adjacent to smooth terrains on the nucleus of 9P (Belton & Melosh 2009). The saturated vapour pressure of the gas inside the nucleus exceeds 1 kPa when the temperature exceeds 55 K for CO and 155 K for CO2 (Prialnik et al. 2004). This means that the CO and CO2 gas pressure may locally exceed the compressive strength of the cometary material and the gas may push the surrounding materials while it rises to the surface, creating channels inside the nucleus. This process re- mains speculative, however, since it has a severe limitation: how would solar energy penetrate meters below the surface, despite the low thermal inertia of 10 -- 50 J K−1 m−2 s−0.5 (Gulkis et al. 2015)? We currently do not have the answer to this critical ques- tion, but the presence of fractures on the surface (Thomas et al. 2015; El-Maarry et al. 2015b) may help to solve this issue. 5.4. Philae observations A key question is how to reconcile our low strength estimates with the observation made by the Philae MUPUS experiment, which suggests much larger strengths of several MPa since MU- PUS was not able to penetrate the surface material below a few centimeters (Spohn et al. 2014). The solution resides in the hard layer of water ice under the surface dust deposit. As explained in Pommerol et al. (2015), laboratory experiments with analogs such as KOSI (Grun 1991) have shown that a hard layer of water ice can be produced by sublimation/redeposition cycles and/or sintering of water ice close to the surface. This layer has an es- timated thickness of a few centimeters (Pommerol et al. 2015) to several meters (Kossacki et al. 2015) and has been heavily processed, so that it is now hard with strengths up to 1 MPa as measured by Jessberger & Kotthaus (1989), Seiferlin et al. (1995), or Kochan et al. (1989). This layer, which is not repre- sentative of the bulk nucleus material, provides a simple expla- nation for the MUPUS observations. The presence of this layer is compatible with our results on strength since we derived i) the tensile strength from large overhangs and collapsed structures, for which this layer is likely negligible, otherwise the measured tensile strength would be much larger, ii) the shear strength from boulders on the surface, for which this buried layer should not matter, and iii) the compressive strength from the tensile strength using mechanical considerations, for which the same justifica- tion can be used. 5.5. What is representative of the bulk nucleus material? We derived strengths for different materials in different locations on the nucleus, and the question naturally arises whether these materials are representative of the bulk nucleus material. While it is not possible to give a firm positive answer to this question, we provide here some arguments supporting this idea. First, it is very satisfying to see in Table 1 that strength esti- mates derived for different materials at different scales and with different methods are mostly consistent with each other, and this from unconsolidated material with modeling at the microme- ter scale to consolidated material with observations at the meter scale. When translated to the same scale, dust aggregates, peb- bles, boulders, or larger structures have similar strengths. Since they are the building blocks of the nucleus, their mechanical properties should be representative of that of the bulk cometary material. Second, we determined the tensile strength for some large collapsed structures, with a thickness of up to 30 m. These large structures should contain a significant fraction of material that has not been, or has only partially been, affected by activity pro- cesses related to insolation. As a result of the low thermal inertia of 10 -- 50 J K−1 m−2 s−0.5 (Gulkis et al. 2015), the thermal heat wave only penetrates the nucleus by a few meters (at most) (e.g., Groussin et al. 2013). This is confirmed by modeling, which shows that only the first 5 meters are indeed affected by inso- lation (Prialnik et al. 2004). Complex models also show that wa- ter ice is buried under a thin layer (cm) of material depleted in volatiles (Skorov & Blum 2012). The presence of water ice close to the surface (∼1 m) was also observed by the Deep Impact ex- periment (Sunshine et al. 2007). All these arguments support the idea that activity processes only affect the first meters of the nu- cleus. The large 30 m thick structures for which we derived the tensile strength should therefore be a good proxy for the bulk material of the nucleus, even if they are not fully representative of it. We furthermore add that it is currently unclear whether the alteration of the first meters would make the material weaker through forming fractures, or stronger through forming a hard sintered layer. Third, it is important to add that, as demonstrated by Toth & Lisse (2006), Toth & Lisse (2010), and (Davidsson 2001), even a low tensile strength of a few hundred Pa is amply sufficient to prevent the rotational breakup of the nucleus of 67P with its current rotation period of 12.4 h. Finally, although we currently have no evidence for it, we cannot exclude heterogeneities in strengths over the nucleus. For example, the Hathor region, for which we determined a shear strength >1770 Pa at the 1 m scale, may be stronger than some small overhangs. Addressing this problem requires more exten- sive investigations and is beyond the scope of this paper. For now, we can only argue that the general pattern of Fig. 8, which is directly linked to the material strengths, is valid across the en- tire nucleus. Article number, page 5 of 18 A&A proofs: manuscript no. 26379_ap 5.6. Constraining the origin of 67P It is interesting to compare the compressive strength to the pressure resulting from gravity inside the nucleus. Assuming a constant density, the pressure P inside the nucleus is given by Eq. (7), where R is the nucleus radius and r is the distance from the center. 2 3 πGρ2R2 1 − ( P(r) = r R (7) (cid:21) )2 (cid:20) Figure 15 shows the pressure inside the nucleus as a function of depth for different initial radii of the nucleus after its complete accretion and before it entered the inner solar system. Depend- ing on past evolution scenarios, the nucleus of 67P was initially larger by a few hundred meters to a few kilometers (Groussin et al. 2007). The nucleus has shrunk as a result of erosion and now has a radius of ∼2 km (red line in Fig. 15). For an initial radius of 2.2 -- 3.0 km, slightly larger than today, the pressure in- side the nucleus exceeded the compressive strength of the layers we see today (30 -- 150 Pa). This means that if the nucleus was formed by the gentle accretion of pristine materials (see below), their subsequent compaction by gravity inside the nucleus was sufficient to explain the compressive strength we measure today. Continuing with this idea, one may ask whether compaction by gravity inside the nucleus was also sufficient to form the lay- ers we see today on the nucleus (Massironi et al. 2015). On Earth, the diagenesis of rocks starts when they are stressed to more than ten times their compressive strength. If the same mechanism applies to 67P, the primordial material that accreted to form the nucleus had a compressive strength at least ten times lower than the one we measure today to initiate diagenesis, that is, as low as 3 -- 15 Pa. In these conditions, compression could have contributed to the formation of layers. Following the model of Skorov & Blum (2012), the tensile strength of dust aggregates formed by gravitational instability typically is 1 Pa. This value is remarkably consistent with our re- sults (Table 1), particularly if we take into account a compaction scenario where the primordial material was even weaker. Our results therefore tend to favor a formation of comets by pebble accretion in a region of higher concentration of particles such as vortices (Johansen et al. 2014; Barge & Sommeria 1995; Blum et al. 2014), which implies a gentle formation process by accre- tion at low velocity, on the order of 1 m s−1 or lower. In con- trast, the hierarchical accretion model (Weidenschilling 2004) with velocities up to 50 m s−1 for particles larger than 1 m, and the collisional scenario between two large bodies of tens of km or more (Davis & Farinella 1997) with an internal compression by gravity larger than 10 kPa, although not excluded, are less fa- vored. Finally, the low tensile strength indicates that the nucleus of 67P very likely did not experience a global melting/freezing of its water content, neither during the accretion stage by radio- genic heating (Podolak & Prialnik 2000), nor more recently by the exothermic amorphous-to-crystalline water ice reaction (Pri- alnik et al. 2004). Indeed, if a global melting/freezing had oc- curred, the nucleus would be much harder, from 10 kPa (Miles & Faillace 2012) to 1 MPa (e.g., Seiferlin et al. 1995), and denser with a bulk density higher than 1000 kg m−3 (Miles & Faillace 2012). Acknowledgements. OSIRIS was built by a consortium of the Max-Planck- Institut für Sonnensystemforschung, Katlenburg-Lindau, Germany; CISAS Uni- versity of Padova, Italy; the Laboratoire d'Astrophysique de Marseille, France; the Instituto de Astrofísica de Andalucia, CSIC, Granada, Spain; the Research and Scientific Support Department of the ESA, Noordwijk, Netherlands; the Instituto Nacional de Técnica Aeroespacial, Madrid, Spain; the Universidad Article number, page 6 of 18 Politéchnica de Madrid, Spain; the Department of Physics and Astronomy of Uppsala University, Sweden; and the Institut für Datentechnik und Kommunika- tionsnetze der Technischen Universität Braunschweig, Germany. The support of the national funding agencies of Germany (DLR), France (CNES), Italy (ASI), Spain (MEC), Sweden (SNSB), and the ESA Technical Directorate is grate- fully acknowledged. We thank the Rosetta Science Operations Centre and the Rosetta Mission Operations Centre for the successful rendezvous with comet 67P/Churyumov-Gerasimenko. We thank David Romeuf from the University Claude Bernard Lyon 1 (France) for creating the red/blue anaglyph in Fig. 11. We thank the referee, J. Blum, for his helpful and constructive report. References A'Hearn, M. F., Belton, M. J. S., Delamere, W. A., et al. 2011, Science, 332, A'Hearn, M. F., Belton, M. J. S., Delamere, W. A., et al. 2005, Science, 310, 258 Asphaug & Benz. 1996, Icarus, 121, 225 Auger, A.-T., Groussin, O., Jorda, L., et al. 2015, Astronomy & Astrophysics, in Bar-Nun, A., Pat-El, I., & Laufer, D. 2007, Icarus, 187, 321 Barge, P. & Sommeria, J. 1995, Astronomy & Astrophysics, 295, L1 Basilevsky, A. & Keller, H. 2007, Solar System Research, 41, 109 Bažant, A. P. 1999, Archive of Applied Mechanics, 69, 703 Belton, M. J. & Melosh, J. 2009, Icarus, 200, 280 Biele, J., Ulamec, S., Richter, L., et al. 2009, Acta Astronautica, 65, 1168 Blum, J., Gundlach, B., Mühle, S., & Trigo-Rodriguez, J. M. 2014, Icarus, 235, 1396 press 156 Blum, J. & Schräpler, R. 2004, Physical Review Letters, 93, 115503 Blum, J., Schräpler, R., Davidsson, B. J. R., & Trigo-Rodríguez, J. M. 2006, The Astrophysical Journal, 652, 1768 Brownlee, D. E., Horz, F., Newburn, R. L., et al. 2004, Science, 304, 1764 Davidsson, B. J. R. 2001, Icarus, 149, 375 Davis, D. R. & Farinella, P. 1997, Icarus, 125, 50 El-Maarry, M., Thomas, N., Giacomini, L., et al. 2015a, Astronomy & Astro- El-Maarry, M., Thomas, N., Gracia-Berná, A., & et al. 2015b, Geophysical Re- Greenberg, J. M., Mizutani, H., & Yamamoto, T. 1995, Astronomy & Astro- physics, in press search Letters, in press physics, 295, L35 Groussin, O., Hahn, G., Lamy, P. L., Gonczi, R., & Valsecchi, G. B. 2007, Monthly Notices of the Royal Astronomical Society, 376, 1399 Groussin, O. & Lamy, P. 2003, Astronomy & Astrophysics, 412, 879 Groussin, O., Sunshine, J., Feaga, L., et al. 2013, Icarus, 222, 580 , star- dust/EPOXI Grun, E. 1991, Space Science Reviews, 56, 105 Gulkis, S., Allen, M., von Allmen, P., et al. 2015, Science, 347, 709 Gundlach, B., Blum, J., Keller, H. U., & Skorov, Y. V. 2015, submitted to As- Güttler, C., Krause, M., Geretshauser, R. J., Speith, R., & Blum, J. 2009, ApJ, Holsapple, K. A. & Housen, K. R. 2007, Icarus, 187, 345 Jessberger, H. L. & Kotthaus, M. 1989, Physics and Mechanics of Cometary tronomy & Astrophysics 701, 130 Materials, 302, 141 Johansen, A., Blum, J., Tanaka, H., et al. 2014, Protostars and Planets VI, 547 Jorda, L., Gaskell, G., Capanna, C., & et al. 2015, submitted to Icarus Jorda, L., Lamy, P. L., Gaskell, R. W., et al. 2012, Icarus, 221, 1089 Julien, P. 1995, Cambridge University Press: Cambridge, New York, 280 pp Keller, H. U., Barbieri, C., Lamy, P., et al. 2007, Space Sci. Rev., 128, 433 Kelley, M. S., Lindler, D. J., Bodewits, D., et al. 2013, Icarus, 222, 634 , star- Klinger, J., Espinasse, S., & Schmidt, B. 1989, Physics and Mechanics of Kochan, H., Feuerbacher, B., Joo, F., et al. 1989, Advances in Space Research, Kossacki, K. J., Spohn, T., Hagermann, A., Kaufmann, E., & Kührt, E. 2015, dust/EPOXI Cometary Materials, 302, 197 9, 113 submitted to Icarus Massironi, M., Simioni, E., Marzari, F., et al. 2015, submitted to Nature Miles, R. & Faillace, G. A. 2012, Icarus, 219, 567 Pajola, M., Vincent, J.-B., Lee, J.-C., et al. 2015, Astronomy & Astrophysics, in press Petrovic, J. J. 2003, Journal of materials and science, 38, 1 Podolak, M. & Prialnik, D. 2000, in Astronomical Society of the Pacific Con- ference Series, Vol. 213, Bioastronomy 99, ed. G. Lemarchand & K. Meech, 231 Pommerol, A., Thomas, N., Elmaarry, M., et al. 2015, Astronomy & Astro- physics, in press Groussin, Jorda, Auger et al.: Gravitational slopes, geomorphology and material strengths of the nucleus of comet 67P Preusker, F., Scholten, F., Matz, K.-D., et al. 2015, Astronomy & Astrophysics, in press Prialnik, D., Benkhoff, J., & Podolak, M. 2004, Comets II, 359 Seiferlin, K., Spohn, T., & Benkhoff, J. 1995, Advances in Space Research, 15, 35 Sheorey, P. R. 1997, Empirical rock failure criterion, CRC press Sierks, H., Barbieri, C., Lamy, P. L., et al. 2015, Science, 347 [http://www.sciencemag.org/content/347/6220/aaa1044.full.pdf] Sirono, S.-i. & Greenberg, J. M. 2000, Icarus, 145, 230 Skorov, Y. & Blum, J. 2012, Icarus, 221, 1 Spohn, T., Knollenberg, J., & Team, M. 2014, AGU conference San Fransisco, P34B Sunshine, J. M., Groussin, O., Schultz, P. H., et al. 2007, Icarus, 191, 73 Thomas, N., Sierks, H., Barbieri, C., et al. 2015, Science, 347 [http://www.sciencemag.org/content/347/6220/aaa0440.full.pdf] Thomas, P. C., Veverka, J., Belton, M. J., et al. 2007, Icarus, 187, 4 , deep Impact Mission to Comet 9P/Tempel 1, Part 1 Tokashiki, N. & Aydan, O. 2010, Dokobu Gakkai Ronbunshuu, C 66, 297 Toth, I. & Lisse, C. M. 2006, Icarus, 181, 162 Toth, I. & Lisse, C. M. 2010, in IAU Symposium, Vol. 263, IAU Symposium, ed. J. A. Fernandez, D. Lazzaro, D. Prialnik, & R. Schulz, 131 -- 140 Weidenschilling, S. 2004, Comets II, 745, 97 1 Aix Marseille LAM (Laboratoire d'Astrophysique de Marseille) UMR 7326, 13388, Marseille, France Université, 2 Laboratoire GEOPS (Géosciences Paris Sud), Bat. 509, Université CNRS, Paris Sud, 91405 Orsay Cedex, France 3 Institute of Planetary Research, DLR, Rutherfordstrasse 2, 12489, Berlin, Germany 4 Planetary Science Institute, Tucson (AZ), USA 5 Institute for Geophysics and Extraterrestrial Physics, TU Braun- schweig, 38106, Germany 6 Max-Planck-Institut für Sonnensystemforschung, 37077 Göttingen, Germany 7 Department of Physics and Astronomy, Padova University, Vicolo dell'Osservatorio 3, 35122, Padova, Italy 8 Centro de Astrobiologia (INTA-CSIC), 28691 Villanueva de la 9 International Space Science Institute, Hallerstrasse 6, CH-3012 Canada, Madrid, Spain Bern, Switzerland wijk, The Netherlands 10 Scientific Support Office, European Space Agency, 2201, Noord- 11 Department of Physics and Astronomy, Uppsala University, Box 516, 75120, Uppsala, Sweden 12 PAS Space Research Center, Bartycka 18A, PL-00716 Warszawa, 13 Department of Astronomy, University of Maryland, College Park, Poland MD, 20742-2421, USA 14 LESIA, Obs. de Paris, CNRS, Univ Paris 06, Univ. Paris-Diderot, 5 place J. Janssen, 92195 Meudon, France 15 LATMOS, CNRS/UVSQ/IPSL, 11 boulevard d'Alembert, 78280, Guyancourt, France 16 Centro di Ateneo di Studi ed Attività Spaziali, "Giuseppe Colombo" (CISAS), University of Padova, via Venezia 15, 35131 Padova, Italy 17 Department of Mech. Engineering University of Padova, via Venezia 1, 35131 Padova, Italy 18 CNR-IFN UOS Padova LUXOR, via Trasea 7, 35131 Padova, Italy 19 UNITN, Universit di Trento, via Mesiano, 77, 38100 Trento, Italy 20 Physikalisches Institut, Sidlerstr. 5, University of Bern, CH-3012 Bern, Switzerland Italy 21 INAF - Osservatorio Astronomico, Via Tiepolo 11, 34143, Trieste, 22 Instituto de Astrofisica de Andalucía (CSIC), Glorieta de la As- tronomía s/n, 18008 Granada, Spain 23 Institute for Space Science, Nat. Central Univ., 300 Chung Da Rd., 32054, Chung-Li, Taiwan 24 Operations Department, European Space Astronomy Centre/ESA, P.O. Box 78, 28691 Villanueva de la Canada, Madrid, Spain 25 Southwest Research Institute, 1050 Walnut St., Boulder, CO 80302, 26 INAF, Osservatorio Astronomico di Padova, 35122 Padova, Italy. 27 Institut für Datentechnik und Kommunikationsnetze der TU Braun- schweig, Hans-Sommer-Str. 66, 38106 Braunschweig, Germany 28 University of Padova, Department of Information Engineering, Via Gradenigo 6/B, 35131 Padova, Italy 29 Konkoly Observatory, Budapest H-1525, P.O. Box 67, Hungary USA Article number, page 7 of 18 A&A proofs: manuscript no. 26379_ap Table 1. Summary of the tensile strength (σT ), shear strength (σS ), and compressive strength (σC) for cometary materials, including our own estimates. The list is not exhaustive, but representative of values found in the literature. We scaled each value to the meter scale using a power law with an exponent equal to -0.6 (Sect. 1). This scaling should be taken with caution since consolidated and unconsolidated materials may not follow the same scaling law. Reference Method Value (Pa) Scale Value (Pa) at 1 m scale 100 1 -- 53 5 200 -- 1 100 2 000 -- 4 000 1 5 000 270 1 m 1 m 1 m 1 cm 100 µm 1 mm 10 µm 10 µm <150 3 -- 15 30 m 5 m 100 1 -- 53 5 13 -- 69 8 -- 16 0.02 5 0.3 <1 150 8 -- 39 Observations (sungrazing comets) Observations (rotational breakup) Observations (D/1993 F2 SL9) Laboratory experiments Laboratory experiments Laboratory experiments Modelling Modelling Observations (67P, Rosetta) -- Collapsed structures -- Overhangs Tensile strength (σT ) Klinger et al. (1989) Davidsson (2001) Asphaug & Benz (1996) Blum et al. (2006) Bar-Nun et al. (2007) Blum et al. (2014) Biele et al. (2009) Greenberg et al. (1995) This work Shear strength (σS ) Holsapple & Housen (2007) A'Hearn et al. (2005) This work Observations (9P, Deep Impact) Observations (9P, Deep Impact) Observations (67P, Rosetta) -- Hathor cliffs -- Fine surface materials and boulders 0 -- 12 000 65 1 m 1 m 0 -- 12 000 65 >30 4 -- 30 900 m 1 m >1 770 4 -- 30 Compressive strength (σC) Jessberger & Kotthaus (1989) Laboratory experiments Laboratory experiments Bar-Nun et al. (2007) Laboratory experiments Güttler et al. (2009) Laboratory experiments Blum et al. (2014) Sirono & Greenberg (2000) Modelling This work Observations (67P, Rosetta) -- Fine surface materials -- Collapsed structures -- Overhangs 30 000 -- 1 000 000 20 000 <1 000 15 6 500 1 cm 1 890 -- 63 100 80 16 0.2 7 100 µm 1 mm 1 mm 10 µm <15 600 <1 500 30 -- 150 10 cm 30 m 5 m <3 920 <11 600 79 -- 394 Article number, page 8 of 18 Groussin, Jorda, Auger et al.: Gravitational slopes, geomorphology and material strengths of the nucleus of comet 67P Fig. 1. Left panel: view of the Imhotep region. The letter A indicates the region of roundish features, observed near the gravitational low of the region. Image NAC_2014-08-25T23.12.54 (spatial resolution: 0.94 m pix−1). Right panel: gravitational slopes of the Imhotep region, computed from the SPC shape model (Jorda et al. 2015), superimposed on the background image of the left panel. Blue corresponds to terrains with slope angles between 0◦ and 20◦, yellow to terrains with slope angles between 20◦ and 45◦ , and red to terrains with slope angles between 45◦ and 90◦. Yellow circles show examples of large and isolated boulders on slopes lower than 15◦. White circles show examples of boulder fields that end on slopes lower than 15◦. Fig. 2. Left panel: view of the Ash region. The letter A indicates the largest depression of the region. Image NAC_2014-08-07T18.20.34 (spatial resolution: 1.5 m pix−1). Right panel: gravitational slopes of the Ash region, computed from the SPC shape model (Jorda et al. 2015), superimposed on the background image of the left panel. Blue corresponds to terrains with slope angles between 0◦ and 20◦, yellow to terrains with slope angles between 20◦ and 45◦ , and red to terrains with slope angles between 45◦ and 90◦. Yellow circles show examples of large and isolated boulders on slopes lower than 15◦. Red circles show examples of smooth terrains on slopes in the range 25 -- 30◦. Article number, page 9 of 18 A&A proofs: manuscript no. 26379_ap Fig. 3. Left panel: view of the Seth region. The letter A indicates the largest pit of the region. Image NAC_2014-08-16T10.59.16 (spatial resolution: 1.8 m pix−1). Right panel: gravitational slopes of the Seth region, computed from the SPC shape model (Jorda et al. 2015), superimposed on the background image of the left panel. Blue corresponds to terrains with slope angles between 0◦ and 20◦, yellow to terrains with slope angles between 20◦ and 45◦ , and red to terrains with slope angles between 45◦ and 90◦. Yellow circles show examples of large and isolated boulders on slopes lower than 15◦. Red circles show examples of smooth terrains on slopes in the range 25 -- 30◦. White circles show examples of boulder fields that end on slopes lower than 15◦. Article number, page 10 of 18 Groussin, Jorda, Auger et al.: Gravitational slopes, geomorphology and material strengths of the nucleus of comet 67P Fig. 4. Upper panel: View of the Hathor region. Image NAC_2014-08-07T20.20.34 (spatial resolution: 1.5 m pix−1). The red circle shows an example of a smooth terrain on slopes in the range 25 -- 30◦. Green circles show examples of overhangs with slopes >90◦. Lower panel: Extracted Digital Terrain Model (left) of the Hathor region with the corresponding gravitational slopes (right) computed from the SPC shape model (Jorda et al. 2015). For technical reasons related to the DTM extraction, we were not able to produce a figure for Hathor like Figs. 1, 2, 3, or 5. Article number, page 11 of 18 A&A proofs: manuscript no. 26379_ap Fig. 5. Left panel: view of the Agilkia region. The letter A indicates the largest depression of the region. Image NAC_2014-08-25T15.42.54 (spatial resolution: 0.94 m pix−1). Right panel: gravitational slopes of the Agilkia region, computed from the SPC shape model (Jorda et al. 2015), superimposed on the background image of the left panel. Blue corresponds to terrains with slope angles between 0◦ and 20◦, yellow to terrains with slope angles between 20◦ and 45◦ , and red to terrains with slope angles between 45◦ and 90◦. Yellow circles show examples of large and isolated boulders on slopes lower than 15◦. The red circle shows an example of a smooth terrain on slopes in the range 25 -- 30◦. Fig. 6. Histogram of gravitational slope angles for the five regions, with one color per region, in logarithmic scale (left) and linear scale (right) for the Y axis. The values are normalized so that the sum of all Y values equals 1. Each region has a unique distribution of gravitational slopes, related to its unique topography. All regions cover a widee range of slopes from 0◦ to almost 100◦. The gravitational slopes were derived from the SPC shape model (Jorda et al. 2015). Article number, page 12 of 18 Groussin, Jorda, Auger et al.: Gravitational slopes, geomorphology and material strengths of the nucleus of comet 67P Fig. 7. Histograms of gravitational slope angles for the five regions, derived from the SPC shape model (solid line) and from the SPG shape model (dashed line). The black arrow on Hathor indicates a bump on the SPG slopes distribution that must be ignored since it results from the fact that the SPG DTM includes a large part of the Hapi region that is absent from the SPC DTM. The Hapi region, which is smooth and flat, thus introduces a bias in the distribution for low-slope angles. Article number, page 13 of 18 A&A proofs: manuscript no. 26379_ap Fig. 8. Sketch of the general pattern in terms of relation between gravitational slopes and terrain morphologies: i) low-slope terrains (0 -- 20◦) are covered by fine material and contain a few large (>10 m) and isolated boulders, ii) intermediate-slope terrains (20 -- 45◦) are mainly fallen consolidated materials and debris fields, with numerous intermediate-size boulders from <1 m to 10 m for the majority of them, and iii) high- slope terrains (45 -- 90◦) are cliffs that expose a consolidated material and do not show boulders or fine material. The borders between low- and intermediate-slope terrains are not always sharp (see text for details). Fig. 9. Scheme of an overhang with rectangular shape L × X × H. Article number, page 14 of 18 Groussin, Jorda, Auger et al.: Gravitational slopes, geomorphology and material strengths of the nucleus of comet 67P Fig. 10. Examples of large structures that already collapsed, with a length L of 100 m or 50 m and a height H of 30 m (determined from the Digital Terrain Model). These structures are located at the northern border of the Imhotep region. Image NAC_2014-09-16T15.44.07 (spatial resolution: 0.52 m pix−1). Article number, page 15 of 18 A&A proofs: manuscript no. 26379_ap Fig. 11. Red/blue anaglyph of overhangs at a high spatial resolution of 18 cm pix−1, in the Maftet region. Several overhangs are visible in this anaglyph; they are indicated by white arrows, all with a length of ∼10 m and a height of ∼5 m. Image NAC_2014-10-19T13.18.55 (spatial resolu- tion: 0.18 m pix−1). Article number, page 16 of 18 Groussin, Jorda, Auger et al.: Gravitational slopes, geomorphology and material strengths of the nucleus of comet 67P Fig. 12. Scheme of a boulder lying on the surface. The radius of the boulder is r, and its contact area with the surface is defined by the angle ϕ. Fig. 13. Left panel: size (radius r) of 2207 boulders of the Imhotep region as a function of their gravitational slope angle θ (adapted from Auger et al. 2015). The black horizontal lines correspond to the angle of repose θrepose = 45 ± 5◦; there are no boulder on slopes larger than the angle of repose. Right panel: value r sin θ for each boulder, as a function of its radius. Both panels: for robustness, we rejected the upper 1% of the r sin θ distribution (blue points). For the remaining 99%, the boulder with the highest value r sin θ is highlighted in green and corresponds to r = 11.5 m and θ = 26.8◦, as indicated by the green horizontal and vertical dashed lines. Article number, page 17 of 18 A&A proofs: manuscript no. 26379_ap Fig. 14. Images of the first nominal landing site, before and after the Philae touchdown. The footprint of the three landing feet are clearly visible. Image NAC_2014-11-12T15.18.52 and NAC_2014-11-12T15.43.51 (spatial resolution: 0.31 m pix−1). Fig. 15. Pressure versus depth inside the nucleus, for different initial sizes of the nucleus (radius R=2.2, 3.0, 5.0 or 7.0 km). The pressure increases with depth. The blue range corresponds to the best value of 30-150 Pa for the compressive strength of the surface materials we see today (30 -- 150 Pa). The red line at 2 km indicates the current size of the nucleus, which is what we see today. For an initial radius of 2.2 -- 3.0 km, slightly larger than today, the pressure inside the nucleus exceeds the compressive strength of the layers we see today. Article number, page 18 of 18
1805.02385	2	1805	2018-06-18T10:49:12	Estimating the porosity structure of granular bodies using the Lane-Emden equation applied to laboratory measurements of the pressure-density relation of fluffy granular samples	[ "astro-ph.EP" ]	The porosity structure of a granular body is an important characteristic that affects evolutionary changes in the body. We conducted compression experiments using fluffy granular samples with various particle sizes, shapes, and compositions. We approximated the pressure-filling factor relationship of each sample with a power law (a modified polytropic relationship). We also fit our previous data and literature data for fluffy granular samples using a power-law equation. The fitting with a power-law form was as good as that achieved with the equations used for powders in previous studies. The polytropic indices obtained in the current study ranged from ~0.01 to ~0.3 and tended to decrease with increasing particle size for samples of similar porosities. We calculated the radial porosity structure and bulk porosity of granular bodies with various radii using the Lane-Emden equation. The results provide the initial, most porous structures of accreted primordial bodies, or re-accumulated rubble-pile bodies consisting of particles that have compression properties similar to those of the assumed granular materials. A range of porosity structures is allowed for a body of given size and macroporosity, depending on the compression properties of the constituent granular material.	astro-ph.EP	astro-ph	1 Estimating the Porosity Structure of Granular Bodies Using the Lane– Emden Equation Applied to Laboratory Measurements of the Pressure– Density Relation of Fluffy Granular Samples Tomomi Omura and Akiko M. Nakamura Department of Planetology, Kobe University, 1-1 Rokkodai-cho, Nada-ku, Kobe, Hyogo, 657-8501, Japan Accepted for publication in ApJ Send editorial correspondence to: Tomomi Omura Department of planetology Kobe University 1-1 Rokkodai-cho, Nada-ku Kobe, 657-8501 Japan Email address: [email protected] Abstract The porosity structure of a granular body is an important characteristic that affects evolutionary changes in the body. We conducted compression experiments using fluffy granular samples with various particle sizes, shapes, and compositions. We approximated the pressure-filling factor relationship of each sample with a power law (a modified polytropic relationship). We also fit our previous data and literature data for fluffy granular samples using a power-law equation. The fitting with a power-law form was as good as that achieved with the equations used for powders in previous studies. The polytropic indices obtained in the current study ranged from ~0.01 to ~0.3 and tended to decrease with increasing particle size for samples of similar porosities. We calculated the radial porosity structure and bulk porosity of granular bodies with various radii using the Lane–Emden equation. The results provide the initial, most porous structures of accreted primordial bodies, or re-accumulated rubble-pile bodies consisting of particles that have compression properties similar to those of the assumed granular materials. A range of porosity structures is allowed for a body of given size and macroporosity, depending on the compression properties of the constituent granular material. Key words: comets: general ― methods: laboratory: solid state ― minor planets, asteroids: general― planets and satellites: formation 2 1. INTRODUCTION Granular objects, such as planetesimals consisting of dust grains and rubble-pile bodies consisting of impact debris, have been ubiquitous throughout the evolution of the solar system (Davis et al. 1979; Weidenschilling 1980; Fujiwara et al. 2006; Kataoka et al., 2013b). The porosity of these objects is an important characteristic that affects their response to impacts as well as their thermal properties. Therefore, the porosity structure determines the evolutionary path of an object. For instance, laboratory and numerical simulations of impacts with small bodies with different internal structures have shown that the damaged or fractured region remains localized when the body has a porous internal structure (Love et al. 1993; Asphaug et al. 1998). The thermal evolution of meteorite parent bodies, coated with a thick, insulating, porous regolith accumulated through repeated impacts, and the re-accumulation of ejecta on the surface of bodies with a primordial porous granular structure, has been shown to differ greatly from that of simple rocky bodies (Akiridge et al. 1998; Henke et al. 2012). Recently, there has been an increase in data on the bulk density of small bodies (Consolmagno et al. 2008; A'Hearn 2011; Baer et al. 2011), providing some indication of internal density structure. A range of density structures should arise from soil pressure due to the equilibrium of self-gravity, centrifugal force, tidal force, the presence of rocks, compression by impacts, and compaction due to impact-induced vibrations. In particular, the density structure due to soil pressure provides the initial, most porous structure when the centrifugal and tidal forces can be neglected. Previous studies estimated bulk density corresponding to the pressure due to self-gravity based on compression curves obtained in laboratory experiments (Yasui & Arakawa 2009) or numerical simulations (Kataoka et al. 2013b), where uniform pressure and density were assumed and a density gradient was not taken into account. In this study, we conducted compression experiments with fluffy samples having various initial porosities, using particles with various sizes, shapes, and compositions. Power-law equations were fit to the compression curves and the Lane– Emden equation was then applied, even though this equation was originally intended for fluids, to calculate the density structure of granular bodies with different radii, assuming that the density structure is determined by soil pressure due only to self-gravity in the vertical direction. 2. COMPRESSION EXPERIMENTS 2.1 Experimental Procedure We used irregular alumina particles, polydisperse spherical silica particles, and 3 polydisperse irregular silica sand with three different sized grains as sample powders. Table 1 lists the physical properties of the sample powders. The silica sand grains were used in a previous study (Omura et al. 2016), but we updated the density of the grain according to a measurement of true density (2.645 g cm-3) made using a pycnometer. We mimicked the porosity of fluffy granular bodies using these materials. The compression properties of a granular body may vary depending on the structure of the layer (i.e., porosity) and the physical properties of constituent particles. The relationship between the characteristics of the powder and the compression properties of samples is discussed in Section 2.3. The cumulative volume fractions of the sample were determined using a laser diffractometer (SHIMAZU SALD-3000S) installed at Kobe University and are shown in Fig. 1. The cumulative volume fraction is defined as Fvol (< x) = dx. Table 1 lists the median diameter (d50) and ratio d85/d15 of each sample. The ratio d85/d15 is an indicator of the broadness of the size distribution: the larger the ratio, the broader the size distribution. Note that the measured sample powders may have contained lumps; therefore, the particle size distribution shown in Fig. 1 does not represent the individual particle size distribution, especially in the case of alumina particles. These lumps may be similar in structure to the dust aggregate used in a previous study (Weidling et al. 2012) and cometary dust particles (Bentley et al. 2016). For this sample, we show both median diameter obtained by measurement and manufacturer's information about particle diameter. We sieved the samples using a 500 µm mesh screen into a cylindrical container and then used a spatula to level off the part of the bed that exceeded the height of the container. The initial filling factor of the samples was adjusted by tapping and piston loading. In our previous study, granular samples were compressed by a centrifuge to simulate the body force of soil pressure due to self-gravity inside small bodies (Omura et al. 2016). The results of centrifuge compression are consistent with those of piston compression when the density gradient inside the centrifuge sample is taken into consideration (Suzuki et al. 2004). Therefore, in the current study, we performed piston compression to obtain the compression curves of the samples. The samples were loaded by a piston fixed to a compressive testing machine installed at Kobe University (EZ- Graph). The uniaxial pressure on the surface of a sample ranged from 1,000 to 4 × 105 Pa for the alumina (0.1 μm) and silica beads (1.7 μm), and from 1.5 × 104 to 4 × 105 Pa for the silica sand. The pressure acting on the sample was calculated by dividing the compressive force by the cross-sectional area of the piston. In fact, friction between the particles and the container wall produces vertical pressure acting on the bottom of the sample that is smaller than that acting on the surface (Janssen effect; e.g. Duran 2000). 4 We calculated the pressure on the bottom of the sample in the same way as described in Omura & Nakamura (2017), and confirmed that the ratio of the vertical pressures acting on the bottom and surface of a sample was between 0.75 and 0.93. The bulk density of the samples was calculated by the volume and mass of the sample, and the volume of the sample was estimated from its height, which was calculated using the height of the piston. The details of the experimental procedure are the same as described in Omura & Nakamura (2017) for alumina (0.1 μm) and silica beads (1.7 μm), and in Omura et al. (2016) for silica sand. 2.2 Experimental Results Figure 2 shows the results of the compression experiments. The vertical axis shows the applied pressure (P) normalized by the tensile strength of the sample particles (Y). We adopted Y = 6.0 × 107 Pa for SiO2 (Kaye & Laby 2005) and Y = 2.8 × 108 Pa for Al2O3 (manufacturer's information). We took tensile strength to be the strength of the sample grains, because tensile strength and crushing strength have been shown to be almost the same in laboratory measurements (S. Shigaki 2016, private communication). Here, we neglected the effects of size on particle strength, which increases with decreasing particle size (Yashima & Saito 1979). The horizontal axis is the filling factor of the sample: 𝑓 = 𝜌𝑏𝑢𝑙𝑘 𝛿⁄ (1) where 𝜌𝑏𝑢𝑙𝑘 and δ are the bulk density and grain density of the sample, respectively (listed in Table 1). For alumina and silica beads, the average values of three experiments are shown. The vertical and horizontal error bars correspond to the standard deviation of the averaged pressure and the uncertainty of the filling factor, respectively. For silica sand, the result of a single measurement with an error bar corresponding to the uncertainty in sample volume is shown (see Omura et al., 2016; Omura & Nakamura 2017). The filling factor for the samples remained almost constant at its initial value until the threshold pressure, defined as the yield strength, was reached, at which point compaction started (Omura & Nakamura. 2017). Here, we show only compression curves with a pressure range larger than the yield strength. For silica sand, no such region existed and all data are shown in Fig. 2. The density change of the largest silica sand particles (73 µm) due to the applied pressure was smaller than that of the 13 µm and 19 µm samples; in other words, that particular sample was the least compressible, probably because it had the narrowest size distribution. The silica sand (19 µm) with the broadest size distribution had the highest density, because void spaces between large particles can be filled effectively by small grains. 5 2.3 Fitting the Results The polytropic relationship used in the Lane–Emden equation (e.g. Chandrasekhar, 1957) is as follows: 𝑃 = 𝐾𝜌 𝑛+1 𝑛 , (2) where P and  denote the pressure and density, respectively. Equation (2) can be modified as 𝑃 𝑌 = 𝐾′ 𝑌 ( 𝜌𝑏𝑢𝑙𝑘 𝛿 𝑛+1 ) 𝑛 = 𝛼𝑓 𝑛+1 𝑛 , (3) where K' and α are constants. We approximated the compression curves in Fig. 2 with Equation (3) and then obtained the factor K', α and polytropic index n for each sample. The results are listed in Table 2. We calculated K' and n from the data in our previous work (Omura et al., 2016; Omura and Nakamura 2017) and Güttler et al. (2009) using Equation (3). The pressure range of the dataset used for fitting was chosen because the data fitted the power law well. We also calculated n from the data of Castellanos et al. (2005). In this case, the pressure of the original data was normalized; therefore, we show only the value of n. These values are listed in Table 2. Previous studies also used a power law to describe the compression curve of granular material. For example, Kataoka et al. (2013a) gave the following relationship between pressure and the filling factor of highly porous granular aggregate: 𝑃 = 𝐸𝑟𝑜𝑙𝑙 𝑟0 3 𝑓3, (4) where P is the pressure, r0 is the radius of the constituent particles, f is the filling factor of the granular aggregate, and Eroll is the rolling energy of the constituent particles (see Wada et al. 2007). In this equation, pressure is proportional to the cube of the filling factor. Sirono (2004) fitted compressive strength data obtained experimentally using mixtures of toner particles and different amounts of fumed silica particles with the power law 𝛴(𝑓) = 𝛴0𝑓𝛽, where 𝛴(𝑓) is the compressive strength of a sample with a given filling factor f. 𝛴0 and β are fitting parameters. The power exponent β corresponds to (n + 1)/n in Eq. (1). We calculated n from the power exponent value shown in Kataoka et al. (2013a) and Sirono (2004) and these are listed in Table 2. The values of n obtained from the compression curves measured in our study ranged from ~0.02 to ~0.2. The value obtained from the compression curves of silica spheres (Güttler et al. 2009) was 0.34, which was slightly larger than our results. The values obtained from a compression curves of mixture of toner particles and fumed silica (Sirono 2004; Castellanos et al., 2005) or TiO2 particles (Castellanos et al. 2005), and the samples used in our previous studies, were also within 6 this range, except for the values of the 77 μm alumina particles. The value of n obtained from the simulation (Kataoka et al. 2013a), which started with a more porous aggregate (f ≅ 10-3) than the samples used in this study, was 0.5, which was larger than the experimental results. Figures 3a and 3b show the relationships between n and the ranges of the filling factor and particle diameter, respectively. For the toner particles with additives, we adopted the diameter of the toner particles as the particle diameter. For alumina (0.1 μm), we adopted both the manufacturer's diameter and the median diameter obtained by measurement. In Figure 3b, both are shown in different symbols. Although the range of filling factors for silica beads and sand grains is similar, the value of n varies by an order of magnitude and decreases with increasing particle diameter. This occurs because when the sample consists of smaller particles, there are more contact points, i.e., movable points, in the same volume. Particles can slide or rotate at contact points. This is probably why compression of a sample consisting of small particles is easier than compression of a sample consisting of large particles. For alumina (0.1 μm), the value of n is similar to that of silica beads (1.7 μm) and the median diameter obtained by measurement is close to the median diameter of silica beads (1.7 μm). For the alumina sample, compression seems to be caused by rearrangement of aggregates prior to rearrangement of individual particles. A popular formula for compression curves used in pharmacology is the Kawakita equation (Kawakita & Lüdde 1971) 𝑃𝑎 𝐶 = 1 𝑎𝑏 + 𝑃𝑎 𝑎 , (5) where Pa is the applied axial pressure and C is the relative volume decrease, or 𝐶 = ( 𝑉0−𝑉 𝑉0 ) = ( 𝑓−𝑓𝑖 𝑓 ) (6) where V0 is the initial sample volume, V is the sample volume, 𝑓𝑖 is the initial filling factor of the sample, and a and b are constants. Note that C corresponds to a strain in the geometry of piston compression. The Nutting equation is another equation that involves the relative volume decrease as a parameter (Nutting 1921; Scott Blair & Caffyn 1949) 𝛾 = 𝜓−1𝜎𝜇𝑡𝜈 (7), where γ is the strain, σ is the stress, t is the duration of the stress, μ and ν are physical exponents that change with the particle diameter and water content (Taneya & Sone 1962), and ψ is a physical property called "firmness" (Scott Blair & Valda Coppen 1940). For μ = 1 and ν = 1, ψ corresponds to the viscosity, while in the case of μ = 1 and ν = 0, ψ corresponds to the elastic modulus (Nutting 1921). Here, γ and σ in Equation (7) are 7 rewritten using C and Pa, respectively, and Equation (7) is modified as 𝑃𝑎 𝐶 = 𝜓𝑃𝑎 1−𝜇𝑡−𝜈 (8). In our experiments, the loading velocity was fixed. Therefore, we regarded 𝑡−𝜈 as a constant. We fitted our experimental data with the Kawakita and Nutting equations. The results are shown in Fig. 4. Our experimental data were fitted well by both equations; however, the fit with the power-law form shown above was as good as the fittings with these equations. The correlation coefficients for all of the fittings exceeded 0.99. 3. INTERNAL POROSITY STRUCTURE OF GRANULAR OBJECTS We calculated the internal structure of a granular body with a certain radius using the power-law relationships between pressure and the filling factor (i.e., density) determined by the measurements given in the previous section and the Lane–Emden equation: 1 𝜉2 𝑑 𝑑𝜉 (𝜉2 𝑑𝛷 𝑑𝜉 ) = −Φ𝑛. (9) denotes a dimensionless parameter introduced for rewriting the density using the density of the center c as follows: 𝜌 = 𝜌𝑐Φ𝑛, (10) and denotes a dimensionless form of the radius of the object and is given by where ξ = 𝑟 𝑎 , (11) 𝑎 = √(𝑛+1)𝐾𝜌𝑐 4𝜋𝐺 1−𝑛 𝑛 . (12) Here, G is the gravitational constant. The boundary conditions of the Lane–Emden equation are Φ = 1 and 𝑑Φ 𝑑𝜉 = 0 at the center of the body. When we choose a set of constants K and n obtained from measurements of the compression properties of a granular sample, we can obtain the relationship between the radius R and mass M of a spherical body. Equation (9) was solved using the fourth-order Runge–Kutta method. Figure 5 shows the results for bodies with radii of 1, 10, 60, and 100 km when using the compression properties of the silica beads (1.7 μm) and silica sand (13 µm). Density was converted into porosity using = 1 8 − (bulk/), where  is the porosity. The porosity decreases with increasing distance from the surface and the radius of the object. The porosity structure depends on the compression properties of the constituent granular particles. In case of the silica beads, the porosity dependence on the depth from the surface of the body is stronger than that of the silica sand. The bulk porosity depends on the size of body more strongly for silica beads than the silica sand. The bulk porosity is almost equal for bodies 60 km in radius but consisting of different particles, i.e., silica beads and silica sand, although the radial porosity structure is different. The results calculated in this study provide the porosity structure arising from soil pressure due only to self-gravity. That is, these are the initial, most porous structures of bodies consisting of the assumed granular materials when the centrifugal and tidal forces are negligibly small. Therefore, we can calculate the porosity structure of a planetesimal and a re-accumulated rubble-pile body that consists of granular material with certain compression properties. In comparison, asteroids lose their void spaces during their lifetimes, e.g., by impact compression or by compaction due to vibration, or because the bodies may contain rocks inside them. In these cases, we can constrain the internal structure of the body when we assume the compression properties of the constituent granular material. For example, 702 Alauda, which has a radius of ~100 km (Tedesco et al. 2002), has a bulk porosity calculated from its bulk density (Rojo & Margot 2011) and the density of CM chondrites (Consolmagno et al. 2008) of ~0.46. This value is smaller than the result calculated using the compression properties of silica sand shown in Fig. 5. If we assume that 702 Alauda consists of granular material that has the same compression properties as silica sand, this body contains rocks or has lost its void spaces during its lifetime. In this study, we used compression curves measured in the laboratory. The compression properties of a granular material should depend on the interparticle force of its constituent particles, which varies with the measurement environment due to adsorbed molecules under ambient conditions (Perko et al. 2001; Kimura et al. 2015). The interparticle force increases when adsorbed molecules are removed; the compression of granular material may become more difficult in space compared to in a laboratory setting. In addition, the internal structure near the surface of a body was not constrained in this study because under a pressure smaller than Yield strength, the layer is not compressed to large degree and so the porosity of the layer near the surface may have remained essentially constant (Omura & Nakamura 2017). 4. SUMMARY 9 In this study, we conducted compression experiments using fluffy granular samples with various particle sizes, shapes, and compositions. We obtained a relationship between the applied pressure and filling factor for each sample and approximated it with a power-law form, i.e., a modified polytropic relationship. The approximation with a power-law form was as good as the approximation with the Kawakita equation, which is used in pharmacology. The polytropic index n for our samples ranged from ~0.02 to ~0.2 and decreased with the size of constituent particles. The Lane–Emden equation was applied and we presented a model of the radial porosity structures of granular bodies with different radii based on the measured compression curves. Different porosity structures were estimated based on different compression curves for bodies with the same size and bulk porosity. The results provide the initial, most porous structures of accreted primordial bodies or re-accumulated rubble-pile bodies when the centrifugal and tidal forces are negligibly small. The internal structure of small bodies can be constrained by comparing the results with the values based on observations. We thank M. Hyodo for allowing us access to the laser diffractometer. We are grateful to an anonymous reviewer for constructive comments on this paper. This research was supported by the Hosokawa Powder Technology Foundation and a grant-in-aid for scientific research from the Japanese Society for the Promotion of Science of the Japanese Ministry of Education, Culture, Sports, Science, and Technology (MEXT; No. 25400453). References A'Hearn, M. F. 2011, ARA&A, 49, 281 Akiridge, G., Benoit, P. H., & Sears, D. W. G. 1998. Icar, 132, 185 Asphaug, E., Ostro, S. J., Hudson, R. S., Scheeres, D. J., & Benz W., 1998, Nature, 393, 437 Baer, J., Chesley, S. R., & Matson, R. D. 2011, AJ, 141, 143 Bentley, M. S., Schmied, R., Mannel, T., et al. 2016, Natur, 537, 73 Castellanos, A., Valverde, J. M., & Quintanilla, M. A. S. 2005, PhRvL, 94, 075501 Chandrasekhar, S. 1957, An Introduction to the Study of Stellar Structure, (Dover, New York) Consolmagno, G. J., Britt, D. T., & Macke, R. J. 2008, ChEG, 68, 1 Davis, D. R., Chapman, C. R., Greenberg, R., Weidenschilling, S. J., & Harris, A. W. 1979, in Asteroids, ed. T. Gehrels (Tucson, Univ. Arizona Press), 528 Duran, J. 2000, Sands, Powders, and Grains: An Introduction of Physics of Granular Materials. (New York: Springer) Trans. A. Reisinger 10 Fujiwara, A., Kawaguchi, J., Yeomans, D. K., et al. 2006, Sci, 312, 1330 Güttler, C., Krause, M., Geretshauser, R. J., Speith, R., & Blum, J. 2009, ApJ, 701, 130 Henke, S., Gail, H. -P., Trieloff, M., Schwarz, W.H., & Kleine, T. 2012, A&A, 537, A45 Kataoka, A., Tanaka, H., Okuzumi, S., & Wada, K. 2013a, A&A, 554, A4 Kataoka, A., Tanaka, H., Okuzumi, S., & Wada, K. 2013b, A&A, 557, L4 Kawakita, K., & Lüdde, K-H. 1971, Powder Technol., 4, 61 Kaye, G. W. C., & Laby, T. H., 2005. Tables of Physical & Chemical Constants (16th edition 1995). 2.2.2 Elasticities and strengths. Kaye & Laby Online. Version 1.0, www.kayelaby.npl.co.uk Kimura, H., Wada, K., Senshu, H., & Kobayashi, H. 2015, ApJ, 812, 67 Love, S.G., Hörz, F., & Brownlee, D. E. 1993, Icar, 105, 216 Nutting, P. G. 1921, FrInJ 191, 679 Omura, T., Kiuchi, M., Güttler, C., & Nakamura, A. M. 2016, Transaction of JSASS, Aerospace Technology Japan, 14, Pk_17 Omura, T., & Nakamura, A. M. 2017, P&SS, 149, 14 Perko, H. A., Nelson, J. D., & Sadeh W. Z. 2001, J. Geotech. Geoenviron. Eng., 127, 371 Rojo, P. & Margot, J. L. 2011, ApJ, 727, 69 Scott Blair, G. W., & Valda Coppen, F. M. V. 1940, Natur, 146, 840 Scott Blair, G. W., & Caffyn, J. E. 1949, The London, Edinburgh, and Dublin Philosophical Magazine and Journal of Science, 40, 80 Sirono, S. 2004, Icar, 167, 431 Suzuki, M., Ojima, K., Iimura, K., & Hirota, M. 2004, J. Soc. Powder Technol., Japan, 41, 663 (in Japanese) Taneya, S., & Sone, T. 1962. Journal of the Japan Society for Testing Materials, 11, 289 (in Japanese) Tedesco, E. F., Noah, P. V., Noah, M., & Price, S. D. 2002. AJ, 123, 1056 Wada, K., Tanaka, H., Suyama, T., Kimura, H., & Yamamoto, T. 2007. ApJ, 661, 320 Weidenschilling, S. J. 1980, Icar, 44, 172 Weidling, R., Güttler, C., & Blum, J. 2012. Icar, 218, 688 Yashima, S. & Saito, F. 1979, J. Soc. Powder Technol., Japan, 16, 714 (in Japanese) Yasui, M., & Arakawa, M. 2009, JGR, 114, E09004 11 Table 1 Properties of the sample powders. Name Grain Shape Material Median d85/d15 Density δ (gcm-3) Diameter (μm) Alumina (0.1 μm) Silica beads (1.7 μm) 3.9 2.2 spherical irregular Al2O3 ＜0.1a, 2.6b SiO2 SiO2 SiO2 SiO2 1.7b 13 b 19 b 73 b 2.9 2.8 12 22 2.0 Silica sand (13 μm) 2.645 irregular Silica sand (19 μm) 2.645 irregular Silica sand (73 μm) 2.645 irregular a Manufacturer's information, Goodfellow. b Determined from the measured particle size distribution. Table 2 Fitting parameters for the compression curves. K' (Pa) α n Ref. 12 Alumina (0.1 μm) Silica beads (1.7 μm) Silica sand (13 μm) Silica sand (19 μm) Silica sand (73 μm) Alumina (4.5 μm) Alumina (23 μm) Alumina (77 μm) Fly ash (4.8 μm) Alumina (6.5 μm) Alumina (15 μm) Alumina (23 μm) Fly ash (4.8 μm) Glass beads (18 μm) Silica sphere (1.5 μm) Toner (12.7 μm) + Silica (0.01 %) Toner (12.7 μm) + Silica (0.05 %) Toner (12.7 μm) + Silica (0.1 %) Toner (12.7 μm) + Silica (0.2 %) Toner (12.7 μm) + Silica (0.4 %) (1.213±0.035)×1011 (4.33±0.13)×102 (2.3164±0.0056)×10-1 (8.35±0.23)×106 (4.96±0.49)×1012 (6.32±0.45)×1011 (4.22±0.36)×1021 (1.59±0.12)×1013 (4.50±0.80)×1020 (1.24±0.49)×1041 (1.371±0.042)×109 (7.36±0.54)×1011 (3.13±0.22)×1012 (6.9±1.3)×1019 (1.649±0.025)×108 (2.951±0.093)×1016 (8.6±3.9)×105 1.9×106 1.5×105 3.3×105 2.6×105 3.3×106 (1.391±0.039)×10-1 (8.27±0.81)×104 (1.056±0.075)×104 (2.201±0.013)×10-1 (4.437±0.025)×10-2 (4.244±0.019)×10-2 (7.05±0.60)×1013 (2.0111±0.0045)×10-2 … … … … … … … … … … … … … … … (5.647±0.023)×10-2 (2.521±0.013)×10-2 (8.430±0.041)×10-3 (6.844±0.023)×10-2 (5.463±0.023)×10-2 (4.761±0.018)×10-2 (2.559±0.014)×10-2 (8.134±0.014)×10-2 (2.3496±0.0027)×10-2 (3.420±0.043)×10-1 0.152 0.192 0.159 0.152 0.105 … … … … … 1 1 1 1 2 2 2 2 2 3 4 4 4 4 4 Toner (7.8 μm) + TiO2 (100 % SAC) Toner (7.8 μm) + Silica (32 % SAC) Toner (11.8 μm) + Silica (32 % SAC) Toner (15.4 μm) + Silica (32 % SAC) Toner (19.1 μm) + Silica (32 % SAC) Numerical simulation … … … … … 4.7×105 (ice 0.2 μm) 6.6×103 (silica 1.2 μm) … … … … … … 13 5 5 5 5 5 6 (6.47±0.12)×10-2 (5.93±0.21)×10-2 (5.51±0.19)×10-2 (5.32±0.15)×10-2 (4.96±0.16)×10-2 0.50 References. (1) Omura et al. (2016), (2) Omura & Nakamura (2017), (3) Güttler et al. (2009), (4) Sirono (2004), (5) Castellanos et al. (2005), (6) Kataoka et al. (2013a). 14 Fig. 1. Particle size distribution of the sample powders. Fig.2. Relationships between the filling factor (f) and the applied pressure normalized by the tensile strength of the sample particles (P/Y). The thin lines are fitted according to Equation (3). 0204060801000.1110100Alumina (0.1 μm)Silica beads (1.7 μm)Silica sand (13 μm)Silica sand (19 μm)Silica sand (73 μm)Cumulative volume fraction, %Diameter, μm 15 Fig. 3. Fig. 3. Relationship between the polytropic index (n) and (a) filling factor (f) and (b) particle diameter. Purple squares indicate the values of alumina (0.1 μm) and an open purple square is the median diameter obtained by measurement (see Table 1). SB indicates silica beads. Gray and orange symbols indicate the values obtained from Omura et al. (2016) and Omura & Nakamura (2017), respectively. GB indicates glass beads. The value indicated by a brown square was obtained from Güttler et al. (2009). Open black symbols show the values obtained by Sirono (2004). T+S indicates toner + fumed silica particles (7 nm) and values in parentheses are the concentrations of the additives in mass (%). Values calculated with the data obtained by Castellanos et al. (2005) are denoted by open pink symbols. T+TiO2/S indicates toner + TiO2 or fumed silica particles (8 or 50 nm) and the values in parentheses are the surface area coverage (SAC) of toner particles by additives (%). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) 16 Fig. 4. Experimental results (marks) fitted with the Kawakita equation (thin dotted lines) and the Nutting equation (thin solid lines). To make the lines clearer, only 33 % (a) or 10 % (b) of the data points are plotted. Fig. 5. Calculation results of the internal porosity structure of four different sized bodies. The horizontal axis shows the distance from the center normalized by the radius of the body. Curves show the results of the calculation. Left: calculated results using the compression properties of silica beads (K = 2.48 × 10-12, n = 0.2201). The bulk porosity is 0.95, 0.82, 0.51, and 0.34 for bodies of radius 1, 10, 60, and 100 km, respectively. Right: calculated results using the compression properties of 13 μm silica sand (K = 1.4 × 10-68, n = 0.04437). The bulk porosities are 0.67, 0.59, 0.51, and 0.49 for bodies of radius 1, 10, 60, and 100 km, respectively.
1110.0280	1	1110	2011-10-03T07:24:06	WISE/NEOWISE Observations of the Jovian Trojans: Preliminary Results	[ "astro-ph.EP" ]	We present the preliminary analysis of over 1739 known and 349 candidate Jovian Trojans observed by the NEOWISE component of the Wide-field Infrared Survey Explorer (WISE). With this survey the available diameters, albedos and beaming parameters for the Jovian Trojans have been increased by more than an order of magnitude compared to previous surveys. We find that the Jovian Trojan population is very homogenous for sizes larger than $\sim10$km (close to the detection limit of WISE for these objects). The observed sample consists almost exclusively of low albedo objects, having a mean albedo value of $0.07\pm0.03$. The beaming parameter was also derived for a large fraction of the observed sample, and it is also very homogenous with an observed mean value of $0.88\pm0.13$. Preliminary debiasing of the survey shows our observed sample is consistent with the leading cloud containing more objects than the trailing cloud. We estimate the fraction to be N(leading)/N(trailing) $\sim 1.4 \pm 0.2$, lower than the $1.6 \pm 0.1$ value derived by others.	astro-ph.EP	astro-ph	WISE/NEOWISE Observations of the Jovian Trojans: Preliminary Results T. Grav Department of Physics and Astronomy, Johns Hopkins University, Baltimore, MD21218, USA; [email protected] A. K. Mainzer, J. Bauer1, J. Masiero Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA 91109, USA Minor Planet Center, Harvard-Smithsonian Center for Astrophysics, Cambridge MA 02138, T. Spahr USA R. S. McMillan Lunar and Planetary Laboratory, University of Arizona, Tucson, AZ 85721, USA R. Walker Monterey Institute for Research in Astronomy, Marina, CA 93933 Infrared Processing and Analysis Center, California Institute of Technology, Pasadena, CA R. Cutri 91125, USA E. Wright UCLA Astronomy, PO Box 91547, Los Angles, CA 90095, USA E. Blauvelt, E. DeBaun, D. Elsbury, T. Gautier IV, S. Gomillion, E. Hand Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA 91109, USA A. Wilkins2 -- 2 -- Department of Astronomy, University of Maryland, College Park, MD 20742 Received ; accepted 1Infrared Processing and Analysis Center, California Institute of Technology, Pasadena, CA 91125, USA 2Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA 91109, USA -- 3 -- ABSTRACT We present the preliminary analysis of over 1739 known and 349 candidate Jovian Trojans observed by the NEOWISE component of the Wide-field Infrared Survey Explorer (WISE). With this survey the available diameters, albedos and beaming parameters for the Jovian Trojans have been increased by more than an order of magnitude compared to previous surveys (Tedesco et al. 1992, 2002; Fern´andez et al. 2003, 2009; Ryan & Woodward 2010). We find that the Jovian Trojan population is very homogenous for sizes larger than ∼ 10km (close to the detection limit of WISE for these objects). The observed sample consists almost exclusively of low albedo objects, having a mean albedo value of 0.07± 0.03. The beaming parameter was also derived for a large fraction of the observed sample, and it is also very homogenous with an observed mean value of 0.88 ± 0.13. Preliminary debiasing of the survey shows our observed sample is consistent with the leading cloud containing more objects than the trailing cloud. We estimate the fraction to be N (leading)/N (trailing) ∼ 1.4 ± 0.2, lower than the 1.6 ± 0.1 value derived by Szab´o et al. (2007). Subject headings: Minor planets, asteroids, general - Infrared: planetary systems - surveys - -- 4 -- 1. Introduction The Jovian Trojan asteroids comprise two clouds around the L4 and L5 Lagrangian points in Jupiter's orbit. There are currently around 4800 known Jovian Trojans, with about half having multi opposition orbits. The population is believed to have a total number similar to that of the main-belt asteroids (MBAs; Tedesco et al. 2005; Yoshida & Nakamura 2005; Jewitt et al. 2000). The two clouds of Trojans librate around the L4 (leading cloud) and L5 (trailing cloud) Lagrange points with periods of the order of a few hundred years, and their orbital eccentricities (< 0.3) and inclinations (< 40◦) are similar to that of the MBAs. About three quarters of the Trojans that have been studied spectroscopically to date have feature-less (D-type) spectra (Bus & Binzel 2002; Roig et al. 2008; Emery et al. 2011) that are found to have low optical albedo (Tedesco et al. 1989; Sheppard & Jewitt 2003; Fern´andez et al. 2003, 2009). The remaining fraction of the Trojans that have been spectroscopically characterized have P- or C-type spectra, mostly in the trailing swarm (Fitzsimmons et al. 1994; Emery et al. 2011). These spectral properties are similar to that of the cometary nuclei and are consistent with an origin in the outer Solar System. Studies of the size distributions support collisional grinding (Jewitt et al. 2000). The Jovian Trojans lie at the core of several of the most important aspects of planetary science, and there are several hypotheses have put forth to explain their origin: 1) mutual collisions of planetesimals populating the region around Jupiter's orbit could have injected fragments into stable Trojan orbits (Shoemaker et al. 1989); 2) nebular gas drag could have produced drift of smaller planetesimals into the resonance gap, where they grew to present size through mutual collisions (Yoder 1979; Peale 1993; Kary & Lissauer 1995); 3) they were formed simultaneously with Jupiter in the early phase of the solar nebula, where a growing Jupiter captured and stabilized the planetesimals near the L4 and L5 points (Marzari & Scholl 1998a,b; Fleming & Hamilton 2000). Marzari et al. (2002) gives an excellent overview of the details of these different early proposed scenarios. -- 5 -- It has also been suggested that, depending on the importance of gas drag during formation, the two clouds could have different dynamics, with the significant gas drag helping to stabilize orbits around the trailing cloud. Planetary migration, on the other hand, would destabilize the trailing cloud, causing it to evolve differently than the leading cloud (Gomes 1998). More recently the so-called Nice-model suggested a more complex scenario: the current Trojan populations are objects that formed together with the Kuiper belt objects in a primordial disk ranging from roughly ∼ 15 − 30AU (Morbidelli et al. 2005; Gomes et al. 2005; Tsiganis et al. 2005). The Jovian Trojans were captured after the mutual 1:2 mean motion resonance crossing of Jupiter and Saturn during migration. This suggests the possibility that the physical and orbital properties of the leading and trailing clouds could be quite different. Such differences have yet to be found. It is, however, clear that the dynamical and physical distributions of the Trojan asteroids offer a critical window into differentiating between several models of Solar System dynamical evolution. It is important to note that there are severe observational biases in the sample of known Jovian Trojans. Jupiter and the Trojans take 12 years to complete a circuit around the ecliptic, and in the last decade the leading cloud has spent significantly more time in the northern hemisphere than the trailing cloud. During this time the number of known Trojans has increased ten-fold. With most of the optical large sky surveys located in the northern hemisphere, the leading cloud has seen significantly better coverage during this time. In addition, the trailing cloud has spent the last few years around the galactic center, an area that most surveys avoid due to the significant increase in star density that makes moving object identification correspondingly difficulty. While the number of MBAs and Trojans to a given size is similar, only about 1% of the known asteroids are in the latter population. This ratio is a consequence of the larger distance that makes a trojan four -- 6 -- magnitudes fainter than an MBA of similar size in the middle of the Main-Belt (this is not even accounting for differences in albedo that generally make the apparent magnitudes of the Trojans even fainter). In this paper we present the analysis of thermal measurements of more than 2000 known and candidate Jovian Trojans performed by the NEOWISE component of the Wide-field Infrared Survey Explorer (WISE; Wright et al. 2010; Mainzer et al. 2011a). WISE, although a mission funded by the NASA's Astrophysical Division, is contributing significantly to the study of the Solar System, and has observed more than 157,000 minor planets during its 1 year long survey. The large sample of Trojans with accurate thermal measurements will allow us to address the following questions: 1) What is the size distribution of the Jovian Trojans with diameters larger than WISE's detection limit of ∼ 5km?; 2) Do the leading and trailing swarms have the same size and absolute number distribution above this detection limit?; 3) What is the albedo distribution of the WISE sample?; 4) Do the leading and trailing clouds have the same albedo distribution? The WISE/NEOWISE observations are described in Section 2, and in Section 3 we describe the Trojan sample and how we select candidate Trojans from the sample of WISE/NEOWISE observations that do not have any optical follow-up. Section 4 describes the thermal modeling in details. The analysis of the results of the thermal modeling is given in Section 5. 2. Observations WISE is a NASA Medium-class Explorer class mission designed to survey the entire sky in four infrared wavelengths, 3.4, 4.6, 12 and 22 µm (denoted W1, W2, W3, and W4 respectively; Wright et al. 2010; Liu et al. 2008; Mainzer et al. 2005). The survey -- 7 -- collected observations of over 157,000 asteroids, including Near-Earth Objects, Main Belt Asteroids, comets, Hildas, Jupiter Trojans, Centaurs and scattered disk objects (Mainzer et al. 2011a). WISE has collected infrared measurements of nearly two orders of magnitude more asteroids than its predecessor, the Infrared Astronomical Satellite (IRAS; Tedesco et al. 2002; Matson et al. 1989). The survey started on 2010 January 14 and the mission exhausted its secondary tank cryogen on 2010 August 5. The ecliptic x- and y-positions of the objects observed during the cryogenic part of the survey is shown in Figure 1. Exhaustion of the primary cryogen tank occurred on 2010 September 29, but the survey was continued until 2011 February 1 as the NEOWISE Post-Cryogenic Mission using only bands W1 and W2, yielding a survey that observed the entire main-belt once. The WISE survey cadence resulted in most minor planets receiving on average of 10-12 observations over ∼ 36 hours (Wright et al. 2010; Mainzer et al. 2011a). The WISE observations of the Trojans were retrieved by querying the Minor Planet Center (MPC) observation files to look for all instances of individual WISE detections of the desired objects that were reported using the WISE Moving Object Processing System (Mainzer et al. 2011a, WMOPS;). The resulting set of position/time pairs were used as the basis for a query of WISE source detections in individual exposures (known as "Level 1b" images) using the Infrared Science Archive (IRSA). To ensure that only observations of the moving objects were returned from the query, a search radius of 0.3" from the position listed in the MPC observation file was used. Furthermore, since WISE collected a single exposure every 11 seconds, the time of our observation was required to be within 4 seconds of the time specified by the MPC. Only observations with 0 or p in the artifact identification flag cc flag were used. A cc flag value of 0 indicates that no evidence of known artifacts was found at the position, while a cc flag of p indicates that an artifact may be present. We have found that observations with cc flag of p produce fluxes that are similar to non-flagged fluxes, resulting in recovery of 20% more observations. Some -- 8 -- of the Trojans observed have W3 magnitudes smaller than 4, at which point the detector approached experimentally-derived saturation limits. In order to account for the inaccuracy of the point-spread-function fitting of these slightly saturated observations, the WISE W3 and W4 magnitude error bars were set to 0.2 magnitudes (Mainzer et al. 2011c). In order to avoid having low-level noise detections and/or cosmic rays contaminating our thermal model fits we required each object to have at least three uncontaminated observations in a band. Any band that did not have at least 40% of the observations of the band with the most numerous detections (W3 or W4 for the Trojans), even if it has 3 observations, was discarded. WMOPS was designed to reject inertially fixed objects such as stars and galaxies in bands W3 and W4. However, with stars having ∼ 100 times higher density in bands W1 and W2, it is more likely that asteroid detections in these bands are confused with inertial sources. We removed such confused asteroid detections by cross-correlating the asteroid detections with sources in the WISE atlas and daily co-added catalogs from IRSA. Objects within 6.5" (equivalent to the WISE beam size at bands W1, W2 and W3) of the asteroid position appearing in the co-added sources at least twice and in more than 30% of the total number of coverages of a given area of sky were considered to be inertially fixed sources contaminating the asteroid positions, and these positions were removed from the thermal fitting. 3. Object Selection As of 2011 April 5, there are currently 4846 known objects that the MPC has identified as Jovian Trojans. Some of these have observational arc lengths that makes their identification tenuous at best. During the cryogenic part of its survey NEOWISE observed 1751 objects that have orbits with semi-major axis between 5.0. and 5.4 AU with arcs long enough to securely identify them as Trojans (here set somewhat arbitrarily to 18 days, with -- 9 -- 981 of these in the leading and 770 in the trailing cloud. In the rest of the paper we will call this sample the long-arc Trojans (LAT). In addition, NEOWISE detected 20,685 objects with length-of-arcs of less than 18 days (here donated the SLA sample). Most of these short-arc objects are NEOWISE discoveries and their orbits remain highly uncertain with clear non-natural feature seen in their orbital parameter distributions caused by the discrete steps the MPC uses to assign their orbital elements (see Figure 2). However, there exist a few ways to use the observed quantities to remove objects that cannot be associated with the Jovian Trojan clouds. Figure 3 shows the observed right ascension and declination of this short length-of-arcs sample and the LAT. Imposing a set of right ascension criteria reduces the number of significantly. Furthermore, due to their heliocentric distance any Jovian Trojan will be moving at very low sky-plane velocities. Using the long-arc members of the cloud we set the velocity limit to < 0.18 degrees per day. This criterion reduces the short-arc sample to 1722 objects (see Figure 4). We then use the thermal color measurements to further remove objects that are inconsistent with the LAT sample. We use W 3 − W 4 − 5v > 2.25, where v is the on-sky velocity in degrees per day. It is important to note that these criteria could remove potential trojans that have physical properties very different than that of the LAT sample. However, it seems unlikely that there would be a large number of such objects that would affect our ability to debias the survey. These criteria yield a sample of 349 objects, 208 in the leading cloud and 141 in the trailing cloud, and in the rest of the paper we call this sample the short-arc Trojans (SAT). It is expected that this SAT sample will receive additional follow-up as the major surveys observe the Jovian Trojan clouds in the future, allowing the MPC to link our observations to those of the optical surveys. It should also be noted that the SAT are only candidate objects. Many of the longer length-of-arc Hildas and even some MBAs have velocities and colors consistent with the LAT sample and so a fair fraction of the SAT sample is expected to be from these populations rather than the Jovian Trojans. -- 10 -- NEOWISE has reported 16,556 observations of the 1751 Trojans with well defined orbits. We extracted the observations from the archive using the method described in Mainzer et al. (2011c) and received 16,551 data lines. There are 2949 observations in the MPC observation catalog of the 349 possible Jovian Trojans. We extracted the observations from the archive and found 2946 corresponding data lines. 4. Preliminary Thermal Modeling Preliminary thermal models for each of the Trojans observed by WMOPS during the cryogenic portion of the survey and using the First-Pass Data Processing Pipeline (version 3.5) described above has been computed (these thermal models will be recomputed when the final data processing is completed some time during the fall of 2011). As described in Mainzer et al. (2011c) the spherical near-Earth asteroid thermal model (NEATM; Harris 1998) was used. The NEATM uses the so-called beaming parameter η to account for cases intermediate between zero thermal inertia (the Standard Thermal Model, or STM; Lebofsky et al. 1978) and high thermal inertia (the Fast Rotating Model, or FRM; Veeder et al. 1989; Lebofsky & Spencer 1989). In the STM, η is set to 0.756 to match the occultation diameters of Ceres and Pallas, while in the FRM, η is equal to π. In the NEATM η is a free parameter than can be fitted if two or more thermal bands are available, or using a single thermal band if a priori information of diameter and albedo is available from space craft or occultation observations. The effects of rotational variability are discussed in more detail below. For each object a spherical surface was approximated using a set of triangular facets (c.f. Kaasalainen 2004). While Trojans may be significantly non-spherical, the WISE observations generally consist of 8 − 10 observations uniformly distributed over ∼ 36h for each object, such that rotational variation in generally is averaged out. Caution needs to -- 11 -- be exercised when interpreting the meaning of an effective diameter in cases of objects with higher rotational amplitudes (Wright 2007). Thermal models were computed for each individual WISE measurement, ensuring that the correct Sun-observer-object geometry were used. The temperature of each facet was computed and color corrections were applied to each facet based on Mainzer et al. (2011c). Nightside facets were assumed to contribute no thermal flux. Adjustments of the W3 effective wavelength blue-ward by 4% from 11.5608µm to 11.0984µm, and the W4 effective wavelength red-ward by 2.5% from 22.0883µm to 22.6405µm, were used. In addition the −8% and +4% offsets to the W3 and W4 magnitude zeropoints (respectively) due to the red-blue calibrator discrepancy were also used (Wright et al. 2010; Mainzer et al. 2011c). In general, orbital elements and absolute magnitudes were taken from the MPC catalogs, and we assumed the absolute magnitude H to have an error equal to 0.3 magnitudes. Emissivity, , was assumed to be 0.9 for all wavelengths (c.f. Harris et al. 2009), and the slope parameter, G, in the magnitude-phase relationship (Bowell et al. 1989) was set to 0.15 unless an improved value exists in the MPC catalogs. For Jovian Trojans with measurements in both W3 and W4, the beaming parameter η was determined using a least square minimization (but was constrained to be less than the upper bound set by the FRM case, π). Figure 5 shows the fitted η value histogram for the objects with both thermal bands, along with the best fitting double Gaussian distribution. The median value of the 1534 objects in the LAT with fitted η is 0.88 ± 0.13, while the weighted mean value is 0.84 ± 0.11. The best-fit double Gaussian shown in Figure 5 has a mean value of 0.84 ± 0.10 and 0.97 ± 0.18 with the lower mean Gaussian having a peak ∼ 3 times higher than the higher mean. For the 216 objects in the LAT with only one thermal measurement, the beaming value cannot be fitted, and we assumed a value of 0.87 ± 0.13. For the Jovian Trojans, bands W1 and W2 are generally dominated by reflected light. -- 12 -- The flux from reflected sunlight was computed for each WISE band as described in Mainzer et al. (2011c) using the IAU phase curve correction (Bowell et al. 1989). Those facets that were illuminated by reflected light and observable by WISE were corrected using color corrections appropriate for a G2 V star (Wright et al. 2010). In order to compute the fraction of total luminosity due to reflected light, the albedo in W1 and W2, dubbed pIR was introduced (we assume that pIR is the same for both bands). The geometric albedo pV is defined as the ratio of brightness of an object observed at zero phase angle to that of a perfectly diffusing Lambertian disk of the same radius located at the same distance. Related to the visible geometric albedo, is the Bond albedo, A, given by A ≈ AV = qpV , where the phase integral q is given by 0.290 + 0.684G (Bowell et al. 1989). The albedo in W1 and W2, pIR, is assumed to obey the same relationship, although it is possible that it varies with wavelength, so what we denote here as pIR for convenience may not be exactly analogous to pV . The resulting distribution is shown in Figure 6 and the mean of the pIR/pV value for the 100 objects for which we were able to fit both pIR and pV is 2.0 ± 0.5. The error in the fits of diameter (D), albedo (pV ), W1/W2 albedo (pIR) and beaming (η) were determined for each object by running 50 Monte Carlo (MC) trials that varied the object's absolute magnitude (H) values by the errors described above and the WISE magnitudes by their error bars using Gaussian probability distributions. The minimum magnitude error for all WISE measurements was set to 0.03 magnitudes, in line with the repeatability given in Wright et al. (2010). If the source was brighter than the saturation limits of 6, 6, 4 or 3 in bands 1 to 4, respectively, the magnitude error of that band was set to 0.2 magnitude to reflect the tests performed in Mainzer et al. (2011c) using calibration objects with known diameters. For objects with fixed η, errors on derived parameters were computing by varying η by a gaussian centered on 0.875 with a full width half max (fwhm) of 0.13. For objects which the W1/W2 albedo, pIR could not be fit, the MC trials varied pIR/pV using a gaussian with center at 2.0 and a fwhm of 0.5. -- 13 -- As described in Mainzer et al. (2011c) the minimum diameter error that can be achieved using WISE observations is ∼ 10%, while the minimum albedo error is ∼ 20% of the stated value, where two thermal bands are available and η can be fitted. Since a significant fraction of the Trojan population has been found to have non-spherical shapes (Hartmann et al. 1988; Binzel & Sauter 1992; Mottola et al. 2011) some care has to be taken in interpreting the derived values presented herein. All diameters given are considered effective diameters, where the assumed sphere has a volume close to that of the actual body observed. Tests using a variety of synthetic triaxial ellipsoidal bodies with different sizes, elongations and pole orientations show that even for objects with significant rotational lightcurves, ∼ 1 magnitude, the effective diameter derived is generally found to have a 1 sigma error bar of ∼ 20% when compared to the spherical-equivalent diameter of the highly elongated ellipsoidal test bodies. This results holds as long as the rotational period is not significantly lower than our sample rate of ∼ 3 hours or significantly longer than the average coverage of an object of ∼ 36 hours. For the shorter rotational periods, the error quoted above is still generally valid, but for the longer rotational periods the derived effective diameters can be anywhere from the highest to the lowest extent of the axes depending on what part of the rotational lightcurve our sample covered. Identifying objects with short or long rotational periods will have to be done in conjunction with existing or new optical lightcurve data, as WISE/NEOWISE in general does not sample often enough to determine periods shorter than the spacecraft's orbital period and does not cover time spans long enough to sample enough of the lightcurve of the objects with periods longer than ∼ 36 hours. A more comprehensive study of the rotational lightcurves by combining existing and new optical data with the thermal data collected from WISE/NEOWISE and the influence of these lightcurves on the individual objects' fits will be covered in a future paper. However, the results presented herein are a statistically valid sample of effective diameters and albedos for the Jovian Trojan population. -- 14 -- Table 1 shows some examples of the results of the thermal model fits and a full electronic version of the table for the 1739 Jovian Trojans detected by WMOPS using the First-Pass processing pipeline during the cryogenic WISE/NEOWISE mission and that have the necessary filtered observations is available at the journal website. 5. Results Table 1: Example of Electronic Table of the Thermal Model Fits Object H G D 00588 8.67 0.15 00617 8.19 0.15 00624 7.20 0.15 00659 8.99 0.15 00884 8.81 0.15 00911 7.89 0.15 01172 8.33 0.15 01173 8.89 0.15 01208 8.99 0.15 01404 9.30 0.15 01437 8.30 0.15 160.6 ± 11.9 185.1 ± 13.1 163.9 ± 7.2 122.4 ± 10.7 116.9 ± 7.6 143.8 ± 4.5 138.7 ± 3.9 114.4 ± 8.0 134.1 ± 6.1 88.4 ± 3.0 128.1 ± 9.8 pV 0.023 ± 0.006 0.027 ± 0.006 0.087 ± 0.016 0.030 ± 0.007 0.039 ± 0.012 0.060 ± 0.012 0.043 ± 0.009 0.038 ± 0.009 0.025 ± 0.006 0.043 ± 0.008 0.052 ± 0.012 η 1.02 ± 0.09 0.90 ± 0.08 0.96 ± 0.05 0.93 ± 0.10 1.01 ± 0.08 1.07 ± 0.04 1.03 ± 0.03 0.88 ± 0.12 0.86 ± 0.05 0.93 ± 0.04 1.00 ± 0.10 pIR # obs mJD 0.076 ± 0.010 0.057 ± 0.007 0.107 ± 0.012 0.059 ± 0.008 0.092 ± 0.012 0.094 ± 0.008 0.088 ± 0.009 0.075 ± 0.019 0.059 ± 0.007 0.065 ± 0.008 0.103 ± 0.024 9 9 9 9 55205.5 10 10 11 11 55281.5 11 11 11 11 55228.2 8 11 11 11 55215.4 9 10 10 10 55317.7 21 21 21 21 55221.2 9 10 10 10 55315.7 0 0 0 7 55277.6 11 12 12 12 55303.5 12 12 12 12 55209.7 0 0 10 10 55244.1 Thermal models were derived for 1739 Jovian Trojans from the LAT sample (for 9 objects there were not enough uncontaminated detections to derive thermal fits), with 985 objects in the leading cloud and 754 objects in the trailing cloud. The diameter versus albedo distribution is shown in Figure 7 . The trojans are compared to the IRAS result -- 15 -- (Ryan & Woodward 2010) and ground based and Spitzer results from Fern´andez et al. (2003, 2009), and are in very good agreement. Our sample does not have the higher albedo seen in some of the smaller objects in the Fern´andez et al. (2009) sample. The best-fit albedo distribution of the 1739 objects in the LAT sample is given in Figure 8. The mean value for the entire sample 0.07 ± 0.03, with both the leading and trailing cloud having the same value. We attempt to fit the sample to a double Gaussian sample. First we used 50 Monte-Carlo trials to derive different sample albedo distributions by varying the individual albedos by its error. For each trial distribution the histogram was generated, and the set of 50 histograms was used to derive the mean value and standard deviation in each bin. The results are shown in Figure 8 as the points with error bars. We then attempted to fit a double Gaussian distribution to these mean values. We find that the LAT sample is best fit with a low albedo Gaussian with mean value of 0.06 and a fwhm of 0.02 and a higher albedo Gaussian with mean value of 0.10 and fwhm of 0.3. The peak amplitude of the two Gaussians has a ratio of ∼ 3. in favor of the low albedo distribution. The mean value is slightly higher than the historically canonical value of 0.040 ± 0.005 (Tedesco et al. 1989; Jewitt et al. 2000). The albedos of the larger objects are, however, consistent with the albedo derived by other authors (Ryan & Woodward 2010; Fern´andez et al. 2009) The albedos of the small objects are also consistent with that of Fern´andez et al. (2003), although that project found a few small objects with albedos that are on the high end of our sample. The homogeneously low albedos strengthen the hypothesis that the Jovian Trojans consist of the low albedo C-, D- and P-type asteroids (Tedesco et al. 1989; Sheppard & Jewitt 2003; Fitzsimmons et al. 1994; Emery et al. 2011). No difference is, however, found among the visible albedo distributions of the leading and trailing clouds that indicate any differences in the taxonomic distribution between the two clouds as suggested by -- 16 -- Fitzsimmons et al. (1994) and Emery et al. (2011). It could simply be that there is no good way to distinguish between these low albedo taxanomic types based on visible albedo alone (Mainzer et al. 2011d). We will, however, study the correlation of albedo with taxonomic class, spectral slope and broadband colors in a future paper. Two thermal bands were available for 1523 objects (831 in the leading and 692 in the trailing cloud), allowing us to derive beaming values. The results are shown in Figure 9 and comparing the beaming value for these objects with those derived by Fern´andez et al. (2003) and Ryan & Woodward (2010) show that our values are generally consistent with those found by these authors. The distribution of beaming values is shown in Figure 5 and is very similar for the two clouds. The mean beaming value is 0.89 for the leading cloud, 0.87 for the trailing cloud and 0.88 for the full sample. The beaming values for the population is very well defined with standard deviation of ∼ 0.13 in both clouds and the full sample. This low dispersion points to the fact that the thermal properties of the Jovian Trojan LAT sample on a whole are very homogeneous compared to the Near-Earth Objects (Mainzer et al. 2011b) and Main-Belt Asteroids (Masiero et al. 2011). In the LAT sample there are 100 objects (60 in the leading and 40 in the trailing) for which measurements are available in bands W1 and/or W2. As mentioned above this allows for the determination of the albedo at these wavelengths as these bands are almost exclusively reflected light. The results are shown versus diameter in Figure 10. While there appears to be a trend with higher pIR for smaller objects, this is highly likely to be due to observational biases as smaller objects with low infrared albedo most likely have fluxes below the sensitivity of WISE in W1 and/or W2. Figure 11 shows the diameter versus ratio of pIR over pV for the 100 objects for which infrared albedo were fit. We see that larger objects are generally darker in the near-infrared than the small objects, but their slopes, i.e. ratio between pIR and pV , are steeper (i.e. redder). -- 17 -- From the LAT sample we can also look at the observed cumulative size distribution (see Figure 12). The observed distributions are remarkably similar, but we caution again that these are the raw distributions. As can be seen from Figure 1 the two clouds were not uniformly covered. The survey started partially into the leading cloud and the part closest to the planet was not observed until the Post-Cryogenic Mission, at which point only W1 and W2 were functioning, leading to a significant loss of objects in this part of the cloud. While the trailing cloud was completely covered during the cryogenic survey, the cloud's tail was in the galactic plane, close to the galactic center, at the time of observation. Careful debiasing is thus needed to properly compare the size distributions of the two clouds. In general the two clouds as observed are remarkably similar. Both the albedo and beaming distributions are the same to within the model errors, and even the infrared albedo distributions have similar means and widths. 6. Preliminary Debiasing of the Trojan Population In order to derive the true size and albedo distributions of the two Trojan clouds, we have to remove the inherent biases that exist in the WISE/NEOWISE sample. The strength of the NEOWISE survey is that it carried out a "blind" search for moving objects, meaning that all moving objects were detected in the same fashion regardless of whether they were known beforehand. This uniformity allows the NEOWISE survey to be debiased independently of the biases of other surveys. Presented below is a preliminary attempt at debiasing, with focus on arriving an estimate on the relative abundance of objects within the two clouds. The full debiasing of this population will be presented in a future paper. To model the NEOWISE survey bias, a high fidelity simulation of it was created. The time of observation, coordinates, orientation and footprint (47 x 47 arcminutes) of all -- 18 -- 1.2 million pointings used by WMOPS during the cryogenic survey was used to recreate the survey history. The WISE Known Solar System Object Possible Association List (KSSOPAL) was used to assess the survey detection efficiency at various locations across the sky. KSSOPAL used a list of known minor planet ephemerides to predict where the object should have been in each WISE frame and generated a list of probable matches. However, unlike WMOPS, it made no attempt at eliminating matches to inertially fixed sources such as stars and galaxies, nor did it check for spurious associations with artifacts or cosmic rays. We limited ourselves to the numbered asteroids in KSSOPAL as these in general have well-determined orbits. In order to reduce the possibility of spurious associations with stars or galaxies, we checked each source location from KSSOPAL against the WISE level 3 Atlas source table and used the n out of m statistics provided to search for sources that repeated; these sources were flagged. For each magnitude bin, the total number of available detections predicted by KSSOPAL was counted and compared to the total number of matches found. The estimate of single image completeness as function of flux for a particular region of the sky for bands W3 and W4 is shown in Figure 13. This completeness curve was computed for a number of different locations throughout the sky to sample the surveys sensitivity as a function of ecliptic and galactic latitude and longitude. The result is the probability that a moving object of a particular flux was detected by the WISE pipeline, and the detection probability curves P were fitted for both W3 and W4 using the following function: P = a0 2 (1 − tanh(a2M − a1)) + a3 (1) where M is the W3 or W4 magnitude and ai are the fitted coefficients. The orbital parameters for our synthetic population were created using the methodology described in Grav et al. (2011), while the physical parameters were constructed using the observed distributions presented above. We assume here that the size, albedo and beaming distributions are the same for the two clouds. Looking at the results in the previous section -- 19 -- this seems to be a reasonable assumption, but we will investigate this in deeper detail in future work. The synthetic populations were given a Gaussian albedo distribution with mean of 0.07 and a fwhm of 0.03. For the beaming a Gaussian with mean 0.88 and a fwhm of 0.13 were used. For the synthetic cumulative diameter distribution we use a power-law of the form: N (> D) = aoD−α (2) where we found α = 2 to work very well for our preliminary debiasing (c.f. Jewitt et al. 2000). In the following preliminary foray into debiasing of our survey we limited ourselves to objects with sizes larger than 10km, which yielded a sample of 1660 Jovian Trojans detected by NEOWISE/WISE. Figure 14 shows the comparison of the simulated and observed size distribution, which are in fair agreement. The difference between the synthetic population and the simulated survey is a measure of the bias introduced by the survey. Figure 15 shows the relative number of objects in the two clouds using the preliminary debiasing simulations. We again caution that this is a preliminary result, but in this early work we were unable to derive any synthetic population with the equal number of objects in the two clouds that yielded simulated physical and dynamical distributions that were similar to that of the observed sample. Even adding in the SAT sample is unable to account for this relative number difference. If we assume that all the 141 objects in the trailing SAT sample and none of the 208 objects in the leading SAT sample are indeed Jovian Trojans (something that is highly unlikely), the number of trailing objects increases to 911. This highly unlikely scenario would only reduce the relative fraction from ∼ 1.4 for the LAT sample alone to ∼ 1.2. The lack of inclusion of the SAT sample in the preliminary debiasing is most likely the dominant error in determining the fraction of objects in the two clouds at this point. The two clouds thus have a fractional number of N (leading)/N (trailing) ∼ 1.4 ± 0.2, which is lower than the fraction of 1.6 ± 0.1 derived by Szab´o et al. (2007) based on optical observations from the SDSS. This example shows, -- 20 -- however, that full debiasing is the key to fully understand the similarities and differences between the two populations. This work is underway and will be presented in a future paper. 6.1. Conclusions We have derived thermal models of 1739 Jovian Trojans together with a sample of 349 objects with observational characteristics that make them possible Trojans. This sample represents an increase by more than one order of magnitude in the number of Jovian Trojans with thermal measurements compared to previous surveys (Tedesco et al. 1992, 2002; Fern´andez et al. 2003, 2009; Ryan & Woodward 2010). We find that the Jovian Trojan population is very homogenous for sizes larger than ∼ 10km (close to the lower size limit for which WISE is sensitive to these objects). The observed sample consists almost exclusively of low albedo objects, with the observed sample having a mean albedo value of 0.07 ± 0.0.3. The uniformly low albedos strengthens the notion that the population consists almost exclusively of C-, P- and D-type asteroids (Gradie et al. 1989). The beaming parameter was also derived for a large fraction of the observed sample, and is also very homogenous with an observed mean value of 0.88 ± 0.13. Preliminary debiasing of the survey shows our observed sample is consistent with the leading cloud containing more objects than the trailing cloud. We estimate the fraction to be N (leading)/N (trailing) ∼ 1.4 ± 0.2, somewhat lower than with the 1.6 ± 0.1 value derived by Szab´o et al. (2007). The size distribution is also found to broadly be consistent with the power-law slope found in Jewitt et al. (2000), and work is underway to fully debias this interesting population of objects. -- 21 -- 7. Acknowledgments This publication makes use of data products from the Wide-field Infrared Survey Explorer, which is a joint project of the University of California, Los Angeles, and the Jet Propulsion Laboratory/California Institute of Technology, funded by the National Aeronautics and Space Administration. This publication also makes use of data products from NEOWISE, which is a project of the Jet Propulsion Laboratory/California Institute of Technology, funded by the Planetary Science Division of the National Aeronautics and Space Administration. We gratefully acknowledge the extraordinary services specific to NEOWISE contributed by the International Astronomical Union's Minor Planet Center, operated by the Harvard-Smithsonian Center for Astrophysics , and the Central Bureau for Astronomical Telegrams, operated by Harvard University. We also thank the worldwide community of dedicated amateur and professional astronomers devoted to minor planet follow-up observations. This research has made use the NASA/IPAC Infrared Science Archive, which is operated by the Jet Propulsion Laboratory/California Institute of Technology, under contract with the National Aeronautics and Space Administration. -- 22 -- Fig. 1. -- The ecliptic x and y positions at 2010 August 5, of the 142,716 objects detected during the cryogenic part of the WISE survey that was assigned an orbit from the MPC (from 2010 January 14 to 2010 August 5). The two Jovian Trojans clouds are clearly seen ∼ 60 degrees leading and trailing the planet Jupiter. The orbits and positions of Mercury, Venus, Earth, Mars and Jupiter are also shown. -- 23 -- Fig. 2. -- The semi-major axes and eccentricities of the objects detected with NEOWISE during the cryogenic part of the survey. The objects with length-of-arc longer than 18 days are given in grey. Objects with shorter length-of-arcs (SLA) are given in black and are mostly NEOWISE discoveries with no optical follow-up. The systematic pattern seen in the objects with short length-of-arcs (SLA) is an artifact of the MPC's discrete steps in assigning the elements of the highly uncertain orbits for these objects. -- 24 -- Fig. 3. -- The observed right ascension and declination of the short arc population (arc- lengths shorter than 18days) compared to the objects with long arc-lengths in the leading and trailing clouds. The effect of the galactic center can be seen clearly, with a dearth of detections near RA=269, DEC=-26 degrees. -- 25 -- Fig. 4. -- The W3 and W4 magnitude color is plotted versus the sky-plane velocity of the SLA (grey) and LAT (black) samples. This distribution shows that we can do a cut at sky-plane velocity V < 0.18 and W 3 − W 4 − 5 ∗ V > 2.25, W3 and W4 are the magnitudes in the two bands, to derive the SAT sample. -- 26 -- Fig. 5. -- The beaming distribution of the long-arc Trojans (LAT) objects for which beaming was derived. The leading and trailing clouds are shown in grey dashed and grey solid, respectively. The solid points are derived by generating 100 Monte-Carlo (MC) trials to yield different sample beaming distributions by varying the individual beaming values by their errors. For each trial distribution the histogram was generated, and the set of 100 histograms was used to derive the mean value and standard deviation in each bin. A double Gaussian distribution was then fitted to these MC mean values and associated standard deviations, plotted as a dashed black line. -- 27 -- Fig. 6. -- The IR albedo/Visible albedo distribution of the 100 long-arc Trojans (LAT) objects for which pIR was derived. The leading and trailing clouds are shown in dashed and solid grey, respectively. The solid points are derived by generating 100 MC trials to yield different sample beaming distributions by varying the individual beaming values by their errors. For each trial distribution the histogram was generated, and the set of 100 histograms was used to derive the mean value and standard deviation in each bin. A single Gaussian distribution was then fitted to these MC mean values and associated standard deviations, plotted as a dashed black line. -- 28 -- Fig. 7. -- The 1739 long-arc Trojans (LAT) objects for which diameter and albedo was de- rived. Diameter and albedos derived by Fern´andez et al. (2003, 2009) and Ryan & Woodward (2010) are shown for comparison. Fig. 8. -- The albedo distribution of the long-arc Trojans (LAT) objects for which albedo was derived. The leading and trailing clouds are shown in dashed and solid grey, respectively. -- 29 -- Fig. 9. -- The long-arc Trojans (LAT) objects for which diameter and beaming value were derived. Diameter and beaming value derived by Fern´andez et al. (2003) and Ryan & Woodward (2010) are shown for comparison. Fig. 10. -- The 100 long length-of-arc Trojans (LAT) objects for which diameter and infrared albedo were derived. -- 30 -- Fig. 11. -- The 100 long length-of-arc Trojans (LAT) objects for which diameter and infrared albedo were derived. Fig. 12. -- The observed (biased) cumulative size distribution of full sample (black), together with the leading (dashed grey) and trailing (solid grey) clouds. -- 31 -- Fig. 13. -- The detection efficiency of the W3 and W4 bands as a function of magnitude as found using the KSSOPAL. -- 32 -- Fig. 14. -- Shown here is an example results from our debiasing simulations. The dashed grey line shows the synthetic population, consisting of 2750 objects with D > 10km. The solid grey line gives the resulting simulated observed population. The black line gives the population with D > 10km observed by WISE/NEOWISE. The dot-dashed grey line shows the simulated population divided by the synthetic population, i.e. the bias introduced by the survey as a function of diameter, normalized by 104. The plot only shows objects larger than 10km and observed arcs longer than 18 days. -- 33 -- Fig. 15. -- Shown here is an example mean longitude results from our debiasing simulations. The dashed grey line gives the synthetic population used in the simulation. The grey solid line gives the resulting simulated population. The black solid line gives the population observed by WISE/NEOWISE. The plots shows objects larger than 10km and observed arcs longer than 18 days. -- 34 -- REFERENCES Bowell, E., Hapke, B., Domingue, D., Lumme, K., Peltoniemi, J., & Harris, A. W. 1989, in Asteroids II, ed. R. P. Binzel, T. Gehrels, & M. S. Matthews, 524 -- 556 Binzel, R. P. & Sauter, L. M. 1992, Icarus, 95, 222 Bus, S. J. & Binzel, R. P. 2002, Icarus, 158, 146 Emery, J. P., Burr, D. M., & Cruikshank, D. P. 2011, AJ, 141, 25 Fern´andez, Y. R., Jewitt, D., & Ziffer, J. E. 2009, AJ, 138, 240 Fern´andez, Y. R., Sheppard, S. S., & Jewitt, D. C. 2003, AJ, 126, 1563 Fitzsimmons, A., Dahlgren, M., Lagerkvist, C., Magnusson, P., & Williams, I. P. 1994, A&A, 282, 634 Fleming, H. J. & Hamilton, D. P. 2000, Icarus, 148, 479 Gomes, R., Levison, H. F., Tsiganis, K., & Morbidelli, A. 2005, Nature, 435, 466 Gomes, R. S. 1998, AJ, 116, 2590 Gradie, J. C., Chapman, C. R., & Tedesco, E. F. 1989, in Asteroids II, ed. R. P. Binzel, T. Gehrels, & M. S. Matthews, 316 -- 335 Grav, T., Jedicke, R., Denneau, L., Chesley, S., Holman, M. J., & Spahr, T. B. 2011, PASP, 123, 423 Harris, A. W. 1998, Icarus, 131, 291 Harris, A. W., Mueller, M., Lisse, C. M., & Cheng, A. F. 2009, Icarus, 199, 86 -- 35 -- Hartmann, W. K., Binzel, R. P., Tholen, D. J., Cruikshank, D. P., & Goguen, J. 1988, Icarus, 73, 487 Jewitt, D. C., Trujillo, C. A., & Luu, J. X. 2000, AJ, 120, 1140 Kaasalainen, M. 2004, A&A, 422, L39 Kary, D. M. & Lissauer, J. J. 1995, Icarus, 117, 1 Lebofsky, L. A. & Spencer, J. R. 1989, in Asteroids II, ed. R. P. Binzel, T. Gehrels, & M. S. Matthews, 128 -- 147 Lebofsky, L. A., Veeder, G. J., Lebofsky, M. J., & Matson, D. L. 1978, Icarus, 35, 336 Liu, F., Cutri, R., Greanias, G., Duval, V., Eisenhardt, P., Elwell, J., Heinrichsen, I., Howard, J., Irace, W., Mainzer, A., Razzaghi, A., Royer, D., & Wright, E. L. 2008, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 7017, Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series Mainzer, A., Bauer, J., Grav, T., Masiero, J., Cutri, R. M., Dailey, J., Eisenhardt, P., McMillan, R. S., Wright, E., Walker, R., Jedicke, R., Spahr, T., Tholen, D., Alles, R., Beck, R., Brandenburg, H., Conrow, T., Evans, T., Fowler, J., Jarrett, T., Marsh, K., Masci, F., McCallon, H., Wheelock, S., Wittman, M., Wyatt, P., DeBaun, E., Elliott, G., Elsbury, D., Gautier, T., Gomillion, S., Leisawitz, D., Maleszewski, C., Micheli, M., & Wilkins, A. 2011a, ApJ, 731, 53 Mainzer, A., Grav, T., Masiero, J., Bauer, J., Wright, E., Cutri, R., McMillan, R. S., Cohen, M., Ressler, M., & Eisenhardt, P. 2011b, ApJ, 736, 100. Mainzer, A., Grav, T., Bauer, J., Masiero, J., McMillan, R. S., Cutri, R. M., Walker, R., Wright, E., Eisenhardt, P., & Spahr, T. 2011c, accepted for publication in ApJ. -- 36 -- Mainzer, A., Grav, T., Masiero, J., Hand, E., Bauer, J., Tholen, D., McMillan, R. S., Spahr, T., Cutri, R. M., & Maleszewski, C. 2011d, accepted for publication in ApJ. Mainzer, A. K., Eisenhardt, P., Wright, E. L., Liu, F.-C., Irace, W., Heinrichsen, I., Cutri, R., & Duval, V. 2005, in Presented at the Society of Photo-Optical Instrumentation Engineers (SPIE) Conference, Vol. 5899, UV/Optical/IR Space Telescopes: Innovative Technologies and Concepts II. Edited by MacEwen, Howard A. Proceedings of the SPIE, Volume 5899, pp. 262-273 (2005)., ed. H. A. MacEwen, 262 -- 273 Marzari, F. & Scholl, H. 1998a, Icarus, 131, 41 -- . 1998b, A&A, 339, 278 Marzari, F., Scholl, H., Murray, C., & Lagerkvist, C. 2002, Asteroids III, 725 Masiero, J., Mainzer, A., Grav, T., Bauer, J., Cutri, R. M., Dailey, J., McMillan, R. S., Walker, R., & Wright, E. 2011, accepted for publication in ApJ. Matson, D. L., Veeder, G. J., Tedesco, E. F., & Lebofsky, L. A. 1989, in Asteroids II, ed. R. P. Binzel, T. Gehrels, & M. S. Matthews, 269 -- 281 Morbidelli, A., Levison, H. F., Tsiganis, K., & Gomes, R. 2005, Nature, 435, 462 Mottola, S., Di Martino, M., Erikson, A., Gonano-Beurer, M., Carbognani, A., Carsenty, U., Hahn, G., Schober, H.-J., Lahulla, F., Delb`o, M., & Lagerkvist, C.-I. 2011, AJ, 141, 170 Peale, S. J. 1993, Icarus, 106, 308 Roig, F., Ribeiro, A. O., & Gil-Hutton, R. 2008, A&A, 483, 911 Ryan, E. L. & Woodward, C. E. 2010, AJ, 140, 933 -- 37 -- Sheppard, S. S. & Jewitt, D. C. 2003, Nature, 423, 261 Shoemaker, E. M., Shoemaker, C. S., & Wolfe, R. F. 1989, in Asteroids II, ed. R. P. Binzel, T. Gehrels, & M. S. Matthews, 487 -- 523 Szab´o, G. M., Ivezi´c, Z., Juri´c, M., & Lupton, R. 2007, MNRAS, 377, 1393 Tedesco, E. F., Cellino, A., & Zappal´a, V. 2005, AJ, 129, 2869 Tedesco, E. F., Noah, P. V., Noah, M., & Price, S. D. 2002, AJ, 123, 1056 Tedesco, E. F., Veeder, G. J., Fowler, J. W., & Chillemi, J. R. 1992, The IRAS Minor Planet Survey, Tech. rep. Tedesco, E. F., Williams, J. G., Matson, D. L., Weeder, G. J., Gradie, J. C., & Lebofsky, L. A. 1989, AJ, 97, 580 Tsiganis, K., Gomes, R., Morbidelli, A., & Levison, H. F. 2005, Nature, 435, 459 Veeder, G. J., Hanner, M. S., Matson, D. L., Tedesco, E. F., Lebofsky, L. A., & Tokunaga, A. T. 1989, AJ, 97, 1211 Wright, E. L. 2007, arXiv:astro-ph/0703085 Wright, E. L., Eisenhardt, P. R. M., Mainzer, A. K., Ressler, M. E., Cutri, R. M., Jarrett, T., Kirkpatrick, J. D., Padgett, D., McMillan, R. S., Skrutskie, M., Stanford, S. A., Cohen, M., Walker, R. G., Mather, J. C., Leisawitz, D., Gautier, T. N., McLean, I., Benford, D., Lonsdale, C. J., Blain, A., Mendez, B., Irace, W. R., Duval, V., Liu, F., Royer, D., Heinrichsen, I., Howard, J., Shannon, M., Kendall, M., Walsh, A. L., Larsen, M., Cardon, J. G., Schick, S., Schwalm, M., Abid, M., Fabinsky, B., Naes, L., & Tsai, C. 2010, AJ, 140, 1868 Yoder, C. F. 1979, Icarus, 40, 341 -- 38 -- Yoshida, F. & Nakamura, T. 2005, AJ, 130, 2900 This manuscript was prepared with the AAS LATEX macros v5.2.
1506.02019	1	1506	2015-06-05T19:26:47	The first radial velocity measurements of a microlensing event: no evidence for the predicted binary	[ "astro-ph.EP", "astro-ph.IM", "astro-ph.SR" ]	The gravitational microlensing technique allows the discovery of exoplanets around stars distributed in the disk of the galaxy towards the bulge. However, the alignment of two stars that led to the discovery is unique over the timescale of a human life and cannot be re-observed. Moreover, the target host is often very faint and located in a crowded region. These difficulties hamper and often make impossible the follow-up of the target and study of its possible companions. Gould et al. (2013) predicted the radial-velocity curve of a binary system, OGLE-2011-BLG-0417, discovered and characterised from a microlensing event by Shin et al. (2012). We used the UVES spectrograph mounted at the VLT, ESO to derive precise radial-velocity measurements of OGLE-2011-BLG-0417. To gather high-precision on faint targets of microlensing events, we proposed to use the source star as a reference to measure the lens radial velocities. We obtained ten radial velocities on the putative V=18 lens with a dispersion of ~100 m/s, spread over one year. Our measurements do not confirm the microlensing prediction for this binary system. The most likely scenario is that the assumed V=18 mag lens is actually a blend and not the primary lens that is 2 magnitude fainter. Further observations and analyses are needed to understand the microlensing observation and infer on the nature and characteristics of the lens itself.	astro-ph.EP	astro-ph	Astronomy & Astrophysics manuscript no. letter_UVES_arXiv July 14, 2021 c(cid:13)ESO 2021 Letter to the Editor The first radial velocity measurements of a microlensing event: no evidence for the predicted binary(cid:63) I. Boisse1, A. Santerne2, J.-P. Beaulieu3, W. Fakhardji1, N.C. Santos2, 4, P. Figueira2, S. G. Sousa2, and C. Ranc3 1 Aix Marseille Université, CNRS, LAM (Laboratoire d'Astrophysique de Marseille) UMR 7326, 13388, Marseille, France e-mail: [email protected] 2 Instituto de Astrofísica e Ciências do Espaço, Universidade do Porto, CAUP, Rua das Estrelas, 4150-762 Porto, Portugal 3 CNRS, Université Pierre et Marie Curie, UMR 7095, Institut d'Astrophysique de Paris, 98 bis boulevard Arago, F-75014 Paris, 4 Departamento de Física e Astronomia, Faculdade de Ciências, Universidade do Porto, Rua do Campo Alegre, 4169-007 Porto, France Portugal Received ; accepted ABSTRACT The gravitational microlensing technique allows the discovery of exoplanets around stars distributed in the disk of the galaxy towards the bulge. However, the alignment of two stars that led to the discovery is unique over the timescale of a human life and cannot be re-observed. Moreover, the target host is often very faint and located in a crowded region. These difficulties hamper and often make impossible the follow-up of the target and study of its possible companions. Gould et al. (2013) predicted the radial-velocity curve of a binary system, OGLE-2011-BLG-0417, discovered and characterised from a microlensing event by Shin et al. (2012). We used the UVES spectrograph mounted at the VLT, ESO to derive precise radial-velocity measurements of OGLE-2011-BLG-0417. To gather high-precision on faint targets of microlensing events, we proposed to use the source star as a reference to measure the lens radial velocities. We obtained ten radial velocities on the putative V=18 lens with a dispersion of ∼100 m s−1, spread over one year. Our measurements do not confirm the microlensing prediction for this binary system. The most likely scenario is that the assumed V=18 mag lens is actually a blend and not the primary lens that is 2 mag fainter. Further observations and analyses are needed to understand the microlensing observation and infer on the nature and characteristics of the lens itself. Key words. planetary systems -- radial-velocity -- microlensing -- OGLE-2011-BLG-0417 1. Introduction Different exoplanet detection methods (radial velocity, hereafter RV, stellar transits, direct imaging, pulsar timing, astrometry, and microlensing) are currently used to probe different populations of planets over a wide range of orbital radii, masses and host types. To date, around 25 exoplanets have been discovered via microlensing, and roughly as many await publication (Beaulieu, priv com). These numbers are relatively modest compared with that discovered by the RV method or by the Kepler satellite. However, microlensing probes a domain of the parameter space (host separation vs. planet mass) which is often not accessible currently to other methods. The detection of cold planets down to a few Earth masses (Beaulieu et al. 2006) or the observation of free floating planets put the planetary formation scenario to the test (Cassan et al. 2012). On the other hand, microlensing events have the weaknesses of not being repeatable, and to focus on faint stars on crowded fields that are difficult to characterise precisely. A reobservation of the system to get further parameter characterization is often very difficult. From the observed microlensing light curve, the first deter- mined parameters are the mass ratio and the sky-projected angu- lar separation of the system. Additionnal effects, parallax, xal- larap, terrestrial parallax, finite source effects, detection of the light coming from the lens (thanks to high angular resolution), or Bayesian analysis are used in order to derive the physical param- eters of the different planetary systems and to proceed further in the analysis. In this context, Skowron et al. (2011) showed that the deformation of the microlensing light curve can be used to constrain all the orbital parameters of a binary lens system. Shin et al. (2012) presented the microlensing event OGLE- 2011-BLG-0417 and modelled it as due to a binary lens sys- tem. The source star is a K3 red giant located in the galactic bulge at 8 kpc with Isource=16.74 (Vsource=19.42). They identified the blended light as the primary lens, (Ilens=16.30, Vlens=18.23), which would make it one of a few case where the lens pri- mary is significantly brighter than the source star. Gould et al. (2013) adopted the new calibration of Nataf et al. (2013) for the Bulge giant and revised the initial error budget. Located at 0.95± 0.06 kpc, the lens binary is composed of a primary star of 0.524 ± 0.036 M(cid:12), orbited by a M dwarf of 0.153 ± 0.011 M(cid:12). They showed that this event could be tested by RV measurements and published revised Keplerian parameters (reported here in Ta- ble 1). With a large expected RV amplitude, of 6.4 km s−1, and with an eccentricity of 0.069, this system can be detected us- ing only relatively low-precision instruments, benchmarking the microlensing detection. (cid:63) Based on observations made with ESO Telescope at the Paranal Observatory under program ID 092.C-0763(A) and 093.C-0532(A). In this letter, we present the first radial velocity observations of this microlensing target, using the UVES spectrograph. Article number, page 1 of 6 5 1 0 2 n u J 5 . ] P E h p - o r t s a [ 1 v 9 1 0 2 0 . 6 0 5 1 : v i X r a Table 1. RV parameters derived by Gould et al. (2013) from the analysis of the microlens event. A&A proofs: manuscript no. letter_UVES_arXiv K km s−1 6.352 0.340 P yr 1.423 0.113 e 0.688 0.027 ω deg 341.824 2.655 Value Error Tperi HJD 5686.344 6.960 2. Observations We obtained a total of 9 hours of observations with the UVES cross-dispersed echelle spectrograph (Dekker et al. 2000) mounted on the VLT in P92 and P93 to measure the RV vari- ations of the OGLE-2011-BLG-0417 binary. Ten spectra were acquired between October 2013 and September 2014. We used the two arms of the spectrograph in parallel with a dichroic beam splinter, the standard mode DIC-1 (390+580), with a wavelength coverage of 326-445 and 476-684 nm. It probes a domain where the late-K dwarf lens emits sufficient flux and where the spectra is not strongly polluted by telluric lines. The exposure time was set to one hour in order to reach a signal-to-noise ratio (SNR) of ∼ 20 at 550 nm. We used a slit of 1 arcsec that gives a spectral resolution of 40000, sufficient to re- solve the lines and calculate RV without loosing too much light due to slit losses. Due to the faintness of the target, the guiding was done with the red camer, the slow readout mode of the CCD was used and we requested a seeing no larger than 1". Still, a slightly larger seeing than the width of the slit leads to some flux loss, but this is compensate in RV precision by a gain in the sta- bility of the illumination of the slits of the spectrographs (Boisse et al. 2010). The log of the observations is given in the online Table 3. All the measurements were kept for the analysis. UVES is not stabilised in pressure and temperature. An im- portant RV drift of the zero point is expected to be present and depend on the temporal evolution of the two parameters. To min- imize this effect, a thorium-argon calibration was done before and after the scientific observation. The target star being in the Galactic bulge, the field is densely crowded. We then fixed the angle of the slit with the sky as shown in the top of Fig. 1 in order to minimise the pollution of contaminant star inside the slit. 3. Data reduction We used the Reflex ESO pipeline to reduced the spectra (Freudling et al. 2013). We first corrected them from cosmic rays with a 5σ-clipping process. We then computed the weighted cross-correlation functions (CCF) of the spectra with a template K5 mask (Pepe et al. 2002). The CCF of the co-added spectrum is plotted in Fig. 2. The source and the lens are superimposed. Their RV values are different by several tens of km.s−1, as ex- pected since the lens is in the Galactic disk whereas the source is in the bulge. The observation of two significant components shows that the observed target was indeed OGLE-2011-BLG- 0417. We compared the correlations obtained for the three detec- tors (one in the blue and two in the red arm) to disentangle the source star from the lens. The variation as a function of wavelength of the relative contrast of the CCF is equivalent to flux ratio. So, the bluer source would show a deeper contrast in the blue wavelength domain. With a (V − I)lens=1.93 and a (V − I)source=2.68, the source is indeed expected to be bluer than the lens. On the contrary, the redder lens shows a higher contrast Article number, page 2 of 6 Fig. 1. Images of the slit-viewer camera of UVES red-arm showing the crowding of the field. The black horizontal shadow is the slit opened at 1". On the top, position that we chose when the slit is on the target, OGLE-2011-BLG-0417. On the bottom, the target is decentered (top left of the slit). North is up and East is on the right. in the red. The source also presents a deepest CCF as presumed for a lower log g (giant star). We checked with the slit-viewer camera that no other source entered the 1 arcsec slit (see bottom of Fig. 2). On three obser- vations, however, the seeing increased or the slit drifted from the pointed target (see the online Table 3). As a consequence, two of the close-by stars entered the slit. In the individual spectra, two components with low contrast appeared in the CCF profiles. However, they are well separated from the lines of the source and the lens. We tested that the CCF width of the source and the lens are constant within the error bars for our 10 measurements, and hence that the measured RV are not polluted by contami- nant stars. We note, however, that the presence of contaminants would in principle increase the RV dispersion. To be put in the same RV heliocentric reference, each CCF profiles were corrected in wavelength from the barycentric earth radial velocity (BERV) values given by the UVES pipeline. The CCF profile were then fitted with a two-Gaussian model to de- rive the RV of the lens and the source. A third and fourth Gaus- sians were used when contaminant stars were visible. For each observation, we calculated two RV values, one for each of the wavelength calibration secured before or after the science expo- sure: RVbe f ore and RVa f ter (respectively). The derived RV is the average between the two, which assumes that the instrumental drift is nearly linear between the two calibrations. We measured the spectrograph RV drift to be from 15 up to 400 m s−1, within one hour. Using the median exposure as calculated from the pho- tometer count of the detectors and weighting by the spectral in- I. Boisse et al.: The first radial velocity measurements of a microlensing event: no evidence for the predicted binary Fig. 2. CCF compute on three bandpasses of the sum of all the spectra, allowing to identify the thinner and bluer line to be the giant source and the other, the late-K dwarf primary of the binary lens. Fig. 3. UVES RV of the source and the lens. The velocities of the two stars share the same instrumental systematics. formation between the blue and red arms does not change the result above a few m.s−1. This correction of the drift allows us to reach root mean square (RMS) for each of the measured stars of ∼800 m.s−1. A zero-point drift of the spectrograph may have not been corrected by the thorium-argon calibrations. The RV is more- over dispersed by another source of noise due to changes in the illumination of the slit for the different spectrum (and so differ- ent point spread functions). The long exposure time (one hour) was expected to average the small movement of the target in the slit. But, the seeing was often significantly smaller than the slit (see the online Table 3). At this stage, the main contribution to the dispersion should come from this effect, however, difficult to measure precisely. We then computed a telluric mask from O2 lines (Figueira et al. 2010) and cross-correlate it with each spectra. The obtained CCF was fitted by a Gaussian to derive a zero-point RV value. Only 20 O2 lines could be fitted in the spectral domain, and the derived RV present an error of ∼200 m.s−1. When we further corrected the stellar RV values from the zero-point drift as cal- culated from the telluric lines, the RMS decreases to ∼250 m.s−1. We plotted in Fig. 3, the RV of the source and the lens. The two stars are not gravitationally bound and their RV variations are dominated by the same instrumental systematics. The main con- tribution comes from the error on the telluric correction. We then decided to use the RV from the source as a reference to measure the lens RV. It leads to the ∆RV value. This could be done because the flux of the two stars are blended in the spectra and followed the same path in the spectrograph. By doing that, the systematics observed in both stars are canceled out. The photon-noise uncertainty was estimated from the em- pirical calibration of Bouchy et al. (2005) on UVES/FLAMES (see also Loeillet et al. 2008). That leads to a mean error bar of σpn ∼50 m.s−1 on the individual targets, and we took their quadratic sum for the ∆RV photon noise error. Another source of noise comes from the drift of the instrument during the expo- sure. Considering this drift is linear between the calibrations and assuming that the value of the RV has a Gaussian distribution that is within the two calibrations with a probability of 99.9%, we calculated an error of σcalib = (RVbe f ore − RVa f ter)/(2 × 3). The RV of the source, the lens and the ∆RV, as well as their error bars are reported in the online Table 2. The reported er- rors are the quadratic sum of σpn and σcalib. The mean error is ∼110 m.s−1 for the individuals RV and 65 m.s−1 on the ∆RV. 4. Data analysis and results The ten secured RV present an RMS of 94 m s−1, with no signif- icant variations. We plotted them in Fig. 4 with the best model from Gould et al. (2013). There is a clear disagreement between the observations and the predicted model. Using the PASTIS val- idation software (Diaz et al. 2014), we estimated the probability of this predicted model. We modelled a Keplerian orbit with nor- mal priors that matched the values and uncertainties reported by Gould et al. (2013). We used a uniform prior for the systemic RV as well as for an extra RV jitter. We ran 20 Markov Chain Monte Carlo analysis of 3.105 iterations each. We repeated the same analyses with an opposite sign for the RV amplitude (the sign of the RV curve is not determined by the microlensing prediction) and for a no-variation scenario. After thinning and merging the chains, we ended with about 10 000 independent samples of the posterior distribution. The best-fit model is superimposed in grey in Fig. 4. The residuals exhibit a RMS of 340 m s−1, hence three times larger than the dispersion of the data. Moreover, this best-fit model de- parts from the prediction joint confidence interval by 3.7 sigma (a priori probability of ∼ 2.10−4). We then estimated the statis- tical evidence of the Gould et al. (2013) prediction, using the method described in Tuomi & Jones (2012). We found a prob- ability of 7.10−8 or 2.10−7 depending if we consider a RV am- plitude that is positive or negative (respectively). Therefore, our spectroscopic observations unambiguously reject the microlens- ing prediction of Gould et al. (2013) for this binary system. We decided to double check if the bright blend (Iblend = 16.29, Vblend =18.23) is the primary component of the lensing system as claimed by Gould et al. (2013). We adopt a set of isochrones (Girardi et al. 2012) with ages in the range 1-10 Gyr for a solar metallicity and the distance modulus of 950 pc. We find a G8 star with a mass of ∼0.82 M(cid:12), almost 2 magnitudes brighter than the primary lens. The bright blend cannot be the primary lens of the microlensing event OGLE 2011-BLG-417. We will revisit the system with high angular resolution to see if the bright blend could be a distant companion to the lensing system or a chance alignement on the line of sight of the source. 5. Conclusions We used the UVES spectrograph to derive Doppler measure- ments of the reflex motion of the primary component of the pre- dicted binary lens OGLE-2011-BLG-0417. The lens, composed of a late-K dwarf orbited by a M dwarf, is brighter (V=18) than the microlensing source (V=19.3). The huge semi-amplitude of Article number, page 3 of 6 −80−60−40−20020406080Radialvelocity[km.s−1]0.750.800.850.900.951.001.05CCFcontrastsourcelens3300 -- 4500A4800 -- 5800A5800 -- 6800A655066006650670067506800685069006950Time[BJD-2450000]−0.6−0.4−0.20.00.20.40.60.8RelativeRV[km.s−1]lenssourceA&A proofs: manuscript no. letter_UVES_arXiv Fig. 4. UVES RV of OGLE-2011-BLG-0417 (dots) with the predicted model from Shin et al. (2012) and Gould et al. (2013) (black line). The systemic was fixed to the median value of the observations. The best-fit model that departs from the prediction by 3.7 sigma, is superimposed in grey. The observations reject the modelled Keplerian at a level greater than 2.10−7 (see text). The insert is an enlargement around the observations. ∼ 6.35 kms−1 of the predicted eccentric orbit should have been easily detected with the RV precision reported her. Our ten measurements with a dispersion of 94 ms−1 and a mean error bar of 65 ms−1 do not confirm the microlensing anal- ysis. These are the first published RV measurements on a mi- crolensing target. We are led to believe that an error on the anal- ysis of the microlensing event is the reason of the discrepancy. A quick look at the photometric data of the microlensing event seems to indicate that the brightness of the lens has been overes- timated. The most likely scenario is that the bright blend of the microlensing light curve is not the light from the primary lens in contradiction with the prediction of Gould et al. (2013). As a consequence, the RV modulation of the lensing system could not be detected, because the primary is 2 magnitude fainter. It is not clear yet if the bright blend is a distant companion to the OGLE 2011-BLG-417 system or a chance alignement. A more complete reanalysis of the system taking into account spectro- scopic and photometric data will be performed in due time, but this is outside the scope of this letter. UVES allowed us to reach a RV precision of 100 m s−1 on a target of 18th magnitude in V in 1 hour-time integration. This precision was made possible thanks to the fact that the source star is observed simultaneously with the binary microlens and can serve as calibration. The technique is similar to differential photometry, but in the spectral domain. With a modest allocation of telescope time (9 h), it would have been possible to characterised with RV, a binary system de- tected by microlensing. This has strong implication for the mod- elling of the microlensing observations. This shows that spectro- scopic follow-up observation of microlensing events is possible with large telescopes. Understanding the reasons of the discrep- ancy on this event will help to improve the characterisation of microlensing systems, already detected or that will be detected with the K2, WFIRST and Euclid space missions. In the coming years, using future ESO facilities such as ESPRESSO @ VLT, or HIRES @ E-ELT it will be possible to perform such mea- surements on planetary systems detected by microlensing such Article number, page 4 of 6 as OGLE-2007-BLG-109 (Gaudi et al. 2008) and OGLE-2012- BLG-0026 (Han et al. 2013), once it has been confirmed via high angular resolution observation that there is no strong contamina- tion by a blend, for example thanks to high angular resolution observations. Acknowledgements. We thanks the UVES operation astronomers that performed these observations and Christian Hummel for his help in the preparation of the p2pp process. I.B. thanks A.S. for his enthusiastic work. Between the first and last measurements, both of us became parent. What a great change ! A.S. warmly thanks Rodrigo F. Díaz and José-Manuel Almenara for their sub- stantial contribution in the development of the PASTIS software. A.S. is sup- ported by the European Union under a Marie Curie Intra-European Fellow- ship for Career Development with reference FP7-PEOPLE-2013-IEF, number 627202. This work was supported by Fundação para a Ciência e a Tecnologia (FCT) through the research grant UID/FIS/04434/2013. P.F., N.C.S., and S.G.S. also acknowledge the support from FCT through Investigador FCT contracts of reference IF/01037/2013, IF/00169/2012, and IF/00028/2014, respectively, and POPH/FSE (EC) by FEDER funding through the program Programa Op- eracional de Factores de Competitividade -- COMPETE. P.F. further acknowl- edges support from FCT in the form of an exploratory project of reference IF/01037/2013CP1191/CT0001. References [2006] Beaulieu, J.-P., Bennett, D., Fouqué, P. et al. 2006, Nature, 439, 437 [2011] Boisse, I., Bouchy, F., Chazelas, B. et al. 2011, EPJWC, 1602003B [2005] Bouchy, F., Pont, F., Melo, C. et al. 2005, A&A, 431, 1105B [2012] Cassan, A., Jubas, D., Beaulieu, J.-P. et al. 2012, Nature, 481, 167 [2000] Dekker, H., D'Odorico, S., Kaufer, A. et al. 2000, SPIE, 4008, 534 [2014] Díaz, R.F, Almenara, J.M, Santerne, A. et al. 2014, MNRAS, 441, 983 [2013] Freudling, W., Romaniello, M., Ballester, P. et al. 2013, A&A, 559, 96 [2008] Gaudi, B., Patterson, J., Spiegel, D. et al. 2008, ApJ, 677, 1268 [2012] Girardi, L., Barbieri, M., Groenewegen, M. et al. 2012, Springer, 165 [2013] Gould, A., Shin, I.-G., Han, C. et al. 2013, ApJ, 768, 126 [2013] Han, C., Udalski, A., Choi, J.-Y. et al. 2013, ApJ, 762, L28 [2008] Loeillet, B., Bouchy, F., Deleuil, M. et al. 2008, A&A, 479, 865 [2013] Nataf, D., Gould, A., Fouqué, P. et al. 2013, ApJ, 769, 88 [2002] Pepe, F., Mayor, M., Galland, F. et al. 2002, A&A, 388, 632 [2012] Shin, I.-G., Han, C. Choi, J.-Y. et al. 2012, ApJ, 755, 91 [2011] Skowron J., Udalski, A., Gould, A. et al. 2011, ApJ, 738, 87 [2012] Tuomi, M. & Jones, H. R. A. 2012, A&A, 544, 116 650066006700680069007000Time[BJD-2450000]4042444648505254DifferentialRV(RVlens-RVsource)[km.s−1]OGLE-2011-BLG-0417predictedbestmodelobserved65006600670068006900700042.242.442.642.843.0A&A -- letter_UVES_arXiv, Online Material p 5 Table 2. RV measurements and their associated 1-σ error bars. RVsource ±1 σ(2) RVlens ±1 σ(2) ∆RV km s−1 -13.166 -13.606 -13.011 -13.489 -13.543 -13.196 -12.779 -13.439 -13.140 -13.074 ±1 σ(2) km s−1 km s−1 km s−1 km s−1 km s−1 42.315 0.076 0.199 29.149 0.164 42.369 0.052 0.047 28.762 0.063 0.025 29.426 0.058 42.437 0.053 42.384 0.064 0.195 28.895 0.161 42.650 0.109 0.051 29.107 0.103 42.351 0.082 0.247 29.155 0.197 42.458 0.060 0.162 29.678 0.150 0.024 29.030 0.043 42.469 0.045 42.363 0.049 0.132 29.223 0.112 0.070 29.449 0.064 42.522 0.059 BJD(1) -2 456 000 573.52363 721.88561 722.85366 729.83683 746.84698 769.80834 771.82660 830.73591 864.62468 925.52640 Notes. (1) The BJD are UTC. (2) The 1-σ error take into account the estimated photon-noise and the error due to the drift of the instrument (in the text σpn and σcalib, respectively). A&A -- letter_UVES_arXiv, Online Material p 6 e r a s e u l a v s s a m r i a d n a g n i e e s , D J B . ) t x e t e e s ( s e g n a h c V R t n a c fi i n g i s e c u d n i t o n d i d s n o i t a d a r g e d r e h t a e w e s u a c e b s i s y l a n a e h t n i t p e k e r a s t n e m e r u s a e m V R l l A . s n o i t a v r e s b o e h t f o g o L . e r u s o p x e - d i m . 3 t a e l b a T n e v i g s t n e m m o C g n i e e S s s a m r i A ) 3 ( e r u s o p x e - d i M ) 2 ( R N S 6 4 . 1 6 1 . 1 0 3 . 1 8 2 . 1 7 0 . 1 p x e T c e s 0 8 4 3 0 8 4 3 0 8 4 3 0 8 4 3 0 8 0 3 4 4 . 0 4 5 . 0 1 5 . 0 4 4 . 0 1 5 . 0 o o t h g i h s a w d n i W . d e d n e s a w B O e h t t a h t t n i o p e h t o t d e t a r o i r e t e d g n i e e S t i l s e h t f o t u o d e t f i r d t e g r a T % 0 1 + g n i e e S . " 0 1 - " 8 . 0 d n u o r a e b o r p e d i u G 9 0 1 . 1 0 1 . 5 7 0 . 9 6 0 . 6 7 1 . 0 6 0 . 1 0 1 . 0 0 1 . 1 7 0 . 4 7 0 . 3 0 . 1 1 0 . 1 7 0 . 1 4 0 . 1 4 2 . 1 0 8 4 3 0 0 6 3 0 0 6 3 0 0 6 3 0 0 6 3 0 5 . 0 9 4 . 0 0 5 . 0 2 5 . 0 8 4 . 0 6 . 8 1 6 . 0 2 0 . 2 2 6 . 1 2 5 . 3 1 5 . 9 1 0 . 1 2 0 . 4 2 0 . 4 2 6 . 1 2 ) 1 ( s 1 − 2 O V R m k 5 7 2 1 . 1 5 2 1 2 . 0 8 1 3 9 . 0 4 6 9 6 . 0 - 3 3 9 0 . 0 - 1 − V R E B s m k 6 4 2 4 9 . 7 2 - 1 8 7 0 7 . 9 2 8 7 4 3 8 . 9 2 5 5 0 4 1 . 0 3 7 8 4 2 0 . 9 2 D J B 0 0 0 0 0 4 2 - 3 6 3 2 5 1 6 5 8 8 6 6 3 5 8 3 8 6 3 8 8 9 6 4 8 . . . . . 3 7 5 6 5 1 2 7 6 5 2 2 7 6 5 9 2 7 6 5 6 4 7 6 5 4 2 3 1 . 0 - 7 2 2 4 . 0 8 2 9 9 . 0 6 9 8 0 . 0 - 6 5 4 4 . 0 - 4 6 1 9 7 . 3 2 2 7 4 0 1 . 3 2 5 3 2 4 2 . 3 - 8 1 7 9 4 . 8 1 - 0 3 3 2 5 . 9 2 - 4 3 8 0 8 0 6 6 2 8 1 9 5 3 7 8 6 4 2 6 0 4 6 2 5 . . . . . 9 6 7 6 5 1 7 7 6 5 0 3 8 6 5 4 6 8 6 5 5 2 9 6 5 e t a D 7 0 - 0 1 - 3 1 0 2 4 0 - 3 0 - 4 1 0 2 5 0 - 3 0 - 4 1 0 2 2 1 - 3 0 - 4 1 0 2 9 2 - 3 0 - 4 1 0 2 1 2 - 4 0 - 4 1 0 2 3 2 - 4 0 - 4 1 0 2 1 2 - 6 0 - 4 1 0 2 5 2 - 7 0 - 4 1 0 2 4 2 - 9 0 - 4 1 0 2 r e t e m o t o h p e h t m o r f d e t a l u c l a c s i e r u s o p x e - d i m e h T ) 3 ( . m n 0 5 5 ∼ t a l e x i p r e p s i R N S e h T ) 2 ( . k s a m . s m r a 2 O c i r u l l e t a h t i w a r t c e p s e h t f o n o i t a l e r r o c - s s o r c d e r e h t d n a e u l b e h t n e e w t e b n o i t a m r o f n i l a r t c e p s e h t e h t m o r f d e v i r e d V R e h t s i 2 O V R ) 1 ( y b g n i t h g i e w d n a s r o t c e t e d e h t f o . s e t o N t n u o c
1507.04367	1	1507	2015-07-15T20:04:34	Double-ringed debris discs could be the work of eccentric planets: explaining the strange morphology of HD 107146	[ "astro-ph.EP" ]	We investigate the general interaction between an eccentric planet and a coplanar debris disc of the same mass, using analytical theory and n-body simulations. Such an interaction could result from a planet-planet scattering or merging event. We show that when the planet mass is comparable to that of the disc, the former is often circularised with little change to its semimajor axis. The secular effect of such a planet can cause debris to apsidally anti-align with the planet's orbit (the opposite of what may be naively expected), leading to the counter-intuitive result that a low-mass planet may clear a larger region of debris than a higher-mass body would. The interaction generally results in a double-ringed debris disc, which is comparable to those observed in HD 107146 and HD 92945. As an example we apply our results to HD 107146, and show that the disc's morphology and surface brightness profile can be well-reproduced if the disc is interacting with an eccentric planet of comparable mass (~10-100 Earth masses). This hypothetical planet had a pre-interaction semimajor axis of 30 or 40 au (similar to its present-day value) and an eccentricity of 0.4 or 0.5 (which would since have reduced to ~0.1). Thus the planet (if it exists) presently resides near the inner edge of the disc, rather than between the two debris peaks as may otherwise be expected. Finally we show that disc self-gravity can be important in this mass regime and, whilst it would not affect these results significantly, it should be considered when probing the interaction between a debris disc and a planet.	astro-ph.EP	astro-ph	Mon. Not. R. Astron. Soc. 000, 1 -- 14 (2002) Printed 9 October 2018 (MN LaTEX style file v2.2) Double-ringed debris discs could be the work of eccentric planets: explaining the strange morphology of HD 107146 Tim D. Pearce(cid:63) and Mark C. Wyatt Institute of Astronomy, University of Cambridge, Madingley Road, Cambridge, CB3 0HA, UK Released 2002 Xxxxx XX ABSTRACT We investigate the general interaction between an eccentric planet and a coplanar debris disc of the same mass, using analytical theory and n-body simulations. Such an interaction could result from a planet-planet scattering or merging event. We show that when the planet mass is comparable to that of the disc, the former is often circularised with little change to its semimajor axis. The secular effect of such a planet can cause debris to apsidally anti-align with the planet's orbit (the opposite of what may be naıvely expected), leading to the counter-intuitive result that a low-mass planet may clear a larger region of debris than a higher- mass body would. The interaction generally results in a double-ringed debris disc, which is comparable to those observed in HD 107146 and HD 92945. As an example we apply our results to HD 107146, and show that the disc's morphology and surface brightness profile can be well-reproduced if the disc is interacting with an eccentric planet of comparable mass (∼ 10 − 100 Earth masses). This hypothetical planet had a pre-interaction semimajor axis of 30 or 40 au (similar to its present-day value) and an eccentricity of 0.4 or 0.5 (which would since have reduced to ∼ 0.1). Thus the planet (if it exists) presently resides near the inner edge of the disc, rather than between the two debris peaks as may otherwise be expected. Finally we show that disc self-gravity can be important in this mass regime and, whilst it would not affect these results significantly, it should be considered when probing the interaction between a debris disc and a planet. Key words: planets and satellites: dynamical evolution and stability - planet -- disc interactions - circumstellar matter - stars: individual: HD 107146 1 INTRODUCTION Given the calm order of the Solar System today, where most planets and minor bodies occupy near circular and coplanar orbits, one could be forgiven for forgetting that planetary systems can be violent places. Indeed, our own system probably had a tumultuous youth; planets may have scattered off each other and collided (Hartmann & Davis 1975), switched places with one-another (Tsiganis et al. 2005) and ploughed into regions of debris (Walsh et al. 2011). This turbulent picture is also inferred from extrasolar planets; many of these objects are eccentric (Schneider et al. 2011), a possible hallmark of previous scattering events (Juri´c & Tremaine 2008). Some, such as Fomalhaut b, may also pass through regions of de- bris (Kalas et al. 2013; Pearce, Wyatt & Kennedy 2015). (cid:63) [email protected] c(cid:13) 2002 RAS Furthermore, major orbital evolution is required to ex- plain some classes of extrasolar planets, such as the Hot Jupiters (Rasio & Ford 1996; Wu & Lithwick 2011). So if planet-planet scattering, mergers and dynamical insta- bilities could be the norm then it is pertinent to ask how planets affected by these processes interact with other bodies in the system, and whether we can use this infor- mation to probe their past or ongoing dynamical evolu- tion. In this paper we examine the general evolution of a system hosting a debris disc interacting with an equal- mass, coplanar, eccentric planet, assuming the planet's eccentricity was rapidly driven up by one of the above processes. Our Solar System hosts two debris discs, the Asteroid and Kuiper Belts, and many extrasolar discs have been detected as infrared excesses in the spectra of stars (e.g. Rhee et al. 2007; Eiroa et al. 2013). So given that debris discs are reasonably common, it is likely that dynamically evolving planets interact with these structures in at least some systems. Instruments such as Spitzer, the HST and ALMA have resolved some extra- solar debris discs (e.g. Backman et al. 2009; Schneider et al. 2009; Dent et al. 2014), and the eccentricity or clumpiness of these discs can be used to infer the pres- ence of planets which would be otherwise undetectable. Hence one aim of this work is to characterise this inter- action in general, to ascertain the signatures an eccentric planet leaves on a disc which can then be compared to observations. This part of the paper is similar in its goals and methodology to our previous investigation (Pearce & Wyatt 2014), in which we considered only planets much more massive than the disc; the main difference now that the planet and disc are of equal mass is the disc's ability to significantly alter the planet's orbit, which can have major effects on the system evolution. Our second aim is to apply the general results to the debris disc of HD 107146, which has been resolved in both infrared emission and scattered light (Ardila et al. 2004; Williams et al. 2004; Carpenter et al. 2005; Corder et al. 2009; Hughes et al. 2011; Ricci et al. 2015). This disc is seen nearly face on, and is broad (∼ 100 au) and axisym- metric with a ∼ 30 au inner hole. The curiosity of this system is that the 1.25mm debris surface density profile appears to either first decrease and then increase with radius, or generally increase but with a gap at around 80 au (Ricci et al. 2015). This differs from the surface density profiles of protoplanetary discs, which decrease with radius (e.g Andrews & Williams 2007); it has been suggested that the unusual profile could be the result of planetary perturbations. Given the large disc mass (pos- sibly ∼ 100M⊕ in total; Ricci et al. 2015), this system is a prime application of our results. We will show that the strange disc morphology can be explained as the after- math of the interaction between an eccentric planet and a coplanar debris disc of the same mass. The layout of this paper is as follows. We examine the general outcomes of this interaction in Section 2, using a combination of theory and n-body simulations. We then apply these results to HD 107146 in Section 3, where we run simulations over a broad region of parameter space to replicate the disc structure and surface brightness profile observed with ALMA. We discuss the implications of the results in Section 4, and conclude in Section 5. 2 GENERAL INTERACTION OUTCOMES In this section we describe the general outcomes of the in- teraction between an eccentric planet and a comparable- mass, coplanar debris disc. We first summarise the dy- namical processes involved in the interaction in sections 2.1 to 2.3, and describe their effects on the orbits of both the debris particles and the planet. Wherever possible we give quantitative predictions from dynamical theory. In Section 2.4 we present an n-body simulation as an ex- ample of the general outcomes of this interaction, and explain these in the context of the theoretical results. c(cid:13) 2002 RAS, MNRAS 000, 1 -- 14 Equal-mass planet disc interactions 2 2.1 Secular effects on debris Secular interactions are long-term angular momentum exchanges between bodies, which can cause a particle's eccentricity and orbital plane (but not semimajor axis) to evolve. As secular timescales are much longer than or- bital timescales then, providing the objects are not in a mean-motion resonance, each interacting body may be thought of as an extended "wire" of material in the shape of the object's orbit, with a density at each point inversely proportional to the body's velocity there (this approach to secular perturbations is known as Gauss averaging). The influence of other masses causes this wire to change shape and orientation. There exists no general analytic evaluation of this interaction (although semi-analytic so- lutions can be constructed in some cases; see Beust et al. 2014), so a common approach is to derive an analyti- cal form complete up to second order in eccentricities and inclinations, and disregard all higher terms. This is valid for small eccentricities and inclinations, but intro- duces errors if larger values are considered; however, we showed in Pearce & Wyatt (2014) that this still produces qualitatively correct results for very large eccentricities so long as the inclinations are small. We now summarise the behaviour of a particle undergoing a secular inter- action with a planet according to second-order secular theory, and apply the results to the interaction between an eccentric planet and a debris disc. Second-order secular theory predicts that a body's eccentricity e and longitude of pericentre are coupled (for details see Murray & Dermott 1999). Specifically, for a test particle undergoing secular perturbations from one or more massive bodies, these quantities satisfy the equations e cos = ep cos p + ef cos f , e sin = ep sin p + ef sin f . (1) (2) Here ep and p denote the "free" parameters of the par- ticle; ep is constant, and p increases linearly with time t such that p = At + β, (3) where A and β are constants. The quantities ef and f are the "forced" values, which depend on the parame- ters of the massive bodies in the system and may evolve in time. Hence the test particle moves around a circle on the e cos , e sin plane, of radius ep and at a rate A. Meanwhile the centre of this circle also moves as the perturbing bodies evolve over time. We now consider a system comprised of a star of mass M∗, a test particle at semimajor axis a and a planet of mass Mplt at semimajor axis aplt (where a > aplt). We also assume that some mechanism causes the planet's pericentre to precess ( plt (cid:54)= 0). We set both the test particle's eccentricity and the planet's longitude of peri- centre to zero at time zero for simplicity, i.e. β = π, f = plt, ep = ef0 (where ef0 is ef evaluated at t = 0). The forcing eccentricity is now Equal-mass planet disc interactions 3 Figure 1. Evolution of a test particle's orbit under the influence of a single, precessing planet, according to second-order secular theory. Shown are the cases where its orbit precesses more quickly than that of the planet (A > plt) and more slowly (A < plt), for a planet of a given eccentricity. Left plot: the coupled evolution of eccentricity e and longitude of pericentre , where the large black circle denotes the planet. A particle of a given semimajor axis will move around one of the two paths in the direction indicated, depending on the sign of A − plt. Right plots: the corresponding physical areas swept out by the particles, in a frame instantaneously aligned with the planet's orbit. The asterisk and thick red line denote the star and the planet's orbit respectively, and the dashed lines show the extreme orbits of the particle. They grey region is the area swept out by the particle, resulting from the superposition of all orbits between the two extremes. Both plots show systems with identical parameters, but with differing signs of A − plt. Hence a particle which would never cross the planet's orbit if A > plt may do so if A < plt. Also note that the planet is precessing, so the structures on the right hand plots will rotate whilst remaining aligned with the planet's orbit. ef = Aplt A − plt eplt, where (cid:114) (cid:114) Aplt = − 1 4 A = 1 4 GM∗ a3 GM∗ a3 Mplt M∗ Mplt M∗ aplt a aplt a b(2) 3/2(aplt/a), b(1) 3/2(aplt/a), (4) (5) (6) G is the gravitational constant and b(j) coefficient, such that s (α) is a Laplace (cid:90) 2π 0 s (α) ≡ 1 b(j) π cos(jψ)dψ (1 − 2α cos ψ + α2)s . (7) Equation 4 shows that the orbit of the test parti- cle will evolve differently in the regimes A > plt and A < plt. Firstly, if A > plt then the precession rate of the particle is faster than that of the planet, and ef > 0. The particle will therefore move anticlockwise about a circle on the e cos(− plt), e sin(− plt) plane, which crosses the origin and is offset from the origin in the di- rection of the planet. This is shown by the solid line on the left hand plot of Figure 1. Hence the particle's eccen- tricity will be maximised when its orbit is aligned with that of the planet, and small when antialigned. Super- imposing all intermediate orbits shows that the particle will sweep out a broad, eccentric disc aligned with the planet's orbit (central plot of Figure 1). This allows the particle to attain high eccentricities and yet be shielded from scattering by the planet. For a complete summary of particle evolution in this regime, see Pearce & Wyatt (2014) and Faramaz et al. (2014). c(cid:13) 2002 RAS, MNRAS 000, 1 -- 14 If A < plt, the planet precesses more quickly than the particle and the forcing eccentricity becomes nega- tive. The particle will now move clockwise about a cir- cle on the e cos( − plt), e sin( − plt) plane, again crossing the origin but offset in the direction opposing the planet. This is the dashed line on the left hand plot of Fig- ure 1. The particle's eccentricity will now be maximised when its orbit is antialigned with that of the planet, and superimposing all intermediate orbits results in a broad, eccentric disc antialigned with the planet's orbit (right hand plot of Figure 1). Hence a particle which would never cross the planet's orbit if A > plt may now do so, and could therefore be ejected by the latter. Equation 6 shows that the particle's precession rate A is set by the perturber's mass and the semimajor axes of the particle and perturber, all of which are constants in the secular problem. On Figure 2 we plot A as a function of test particle semimajor axis for an example system, which contains a precessing planet. Generally, A → 0 if the test particle's semimajor axis is very small or very large, and A → ∞ if the semimajor axes of the particle plt (cid:54)= 0 then there will and planet are similar. Hence if always be a region in particle semimajor axis space where A > plt, and two regions where A < plt. In this case a sufficiently broad disc of test particles will cover both the A > plt and A < plt regimes. Figure 3 shows a schematic of the resulting debris structure, for particles external to the planet. The result is a superposition of the debris structures on Figure 1, with the innermost particles in the A > plt regime and the outermost in the A < plt regime. There exists a crescent-shaped region devoid of debris, and also a location where A > plt and A < plt particles overlap. However this overlap does not necessarily correspond to debris overdensity, and it is unlikely to be a site of increased dust production; the A < ϖpltA > ϖplte sin(ϖ - ϖplt)-0.50.00.5e cos(ϖ - ϖplt)-0.50.00.5A > ϖpltx / aplt−202A < ϖplty / aplt−202x / aplt−202Equal-mass planet disc interactions 4 star. Hence for all but the broadest discs, if the planet is much more massive than the disc then the result will be an eccentric debris structure apsidally aligned with the planet's orbit. This was the case investigated in Pearce & Wyatt (2014). Alternatively if the planet mass is com- parable to that of the disc (ignoring planet evolution for now), the planet will likely precess more rapidly than the outermost debris. Thus the farthest debris will as- sume an eccentric structure antialigned with the planet's pericentre. Whether the innermost debris also forms this structure, or forms a structure apsidally aligned with the planet (as on Figure 3), depends on the parameters of the system; a sufficiently eccentric planet will eject all particles with similar semimajor axes, leaving only dis- tant debris which may be slowly precessing. In this case only the antialigned debris structure would be formed. This also leads to a counter-intuitive result; a planet of comparable mass to the disc will clear a larger region of debris than a much more massive planet. This is be- cause particle orbits antialign with that of a low mass (i.e. rapidly precessing) planet, and are therefore more likely to be ejected than if under the influence of a more massive planet (which their orbits would align with). Thus far we have only considered the secular evolu- tion of debris which does not intersect the planet's or- bit. Particles with orbits crossing that of the planet will eventually be scattered unless in a mean-motion reso- nance, however secular evolution may still occur before then. Beust et al. (2014) simulated the evolution of de- bris under the influence of an eccentric planet, when the planet's orbit crosses that of the debris. They showed that the secular interaction still drives up particle eccen- tricities as before. However their orbits do not preferen- tially align or antialign with that of the planet, but rather initially orientate themselves such that for much of the time their pericentres are misaligned with the planet's by ∼ 70◦. We observed similar evolution of particles on planet-crossing orbits in our simulations using high mass planets (Pearce & Wyatt 2014), suggesting that this be- haviour arises because the orbits intersect. That we are now concerned with comparable planet and disc masses does not make a difference; the particles affected by this mechanism have semimajor axes similar to the planet's, and therefore still precess faster than the latter even if the planet is rapidly precessing. Thus in addition to the long-term secular structures described above, particles crossing the planet's orbit will have their eccentricities driven up and their orbits misaligned with the planet's by ∼ 70◦, before eventually being scattered. 2.2 The effect of scattering on debris Material which regularly crosses the planet's orbit (again, if not in a mean-motion resonance) may be scattered by the planet. A particle's post-scattering orbit may differ significantly from its pre-scattering orbit, but both must pass through the scattering point. Hence a planet which scatters material at all points around its orbit will form an overdensity of debris tracing its orbit, caused by the overlapping orbits of scattered particles. This overdensity is often strongest at planet apocentre, where scattering Figure 2. Precession rate of a test particle as a function of its semimajor axis, derived using second-order secular theory. Here the planet (black circle) has a semimajor axis of 40 au and precesses at a rate plt = 4 × 10−5 ◦yr−1. Particles pre- cessing more rapidly than the planet form an eccentric disc apsidally aligned with the planet's orbit (central plot on Fig- ure 1), whilst those with slower precession rates will antialign with the planet's orbit (right hand plot on Figure 1). Note that this plot does not take account of mean motion resonances, the effect of which could dominate over secular behaviour. Figure 3. General shape of a debris disc undergoing a secular interaction with a interior, precessing planet. The grey regions are disc material, the red ellipse denotes the planet's orbit and the asterisk shows the star. There are two distinct debris popu- lations; an inner population apsidally aligned with the planet, and an outer one which is antialigned. The inner population may or may not be present depending on the parameters of the system, and likewise for the outer population. collision velocities between particles are actually greater within the inner (aligned) disc than those in this inner- outer disc overlap region. One mechanism which may cause planet precession is the secular effect of the debris. If the planet is much more massive than the total disc mass, the precession rate of the former will be slower than that of the de- bris, unless the disc extends very close to or far from the c(cid:13) 2002 RAS, MNRAS 000, 1 -- 14 A / ○yr-110−810−610−410−2100a / au050100150200y / aplt−6−4−20246x / aplt−6−4−20246is most efficient; this is because the planet spends more time around apocentre than at other points in its orbit, and the relative velocity between the planet and (non- scattered) disc particles is smallest here. Objects repeatedly scattered by the planet will even- tually leave the system or collide with other bodies. In the meantime, scattered objects may attain very large semimajor axes and eccentricities, and hence their orbits could extend far beyond the initial outer edge of the disc. Inclinations are typically excited less than eccentricities in repeated scattering encounters (Ida & Makino 1992), so scattered material will form a broad disc superimposed on the secular debris structure discussed in Section 2.1. The surface density Σ of such a scattered disc will follow an r−3.5 profile. This is an empirical result which is observed in all of our simulations regardless of parameters (as well as those of Duncan, Quinn & Tremaine 1987), but it can also be obtained using the following semi-analytic method. According to the model of Yabushita (1980), particles repeatedly scattered by a planet will diffuse in semimajor axis space, such that the number of particles n with x in the range x → x + dx (where x ≡ 1/a) at time t is given by (cid:20) (cid:16) 1 +(cid:112)x/x0 (cid:17)(cid:21) I2 (cid:20) 16 τ (cid:21) 4(cid:112)x/x0 . n(x, τ ) = 4 xτ exp − 8 τ (8) Here x0 is the initial value of x, I2 is the modified Bessel function, and τ ≡ t/tD(x0) where tD(x) is the diffusion timescale: tD(x) ≡ 0.01Tplt √ apltx , (9) (cid:18) Mplt (cid:19)−2 M∗ where Tplt is the orbital period of the perturbing planet. A reasonable approximation is that scattered particles diffuse in x whilst their pericentre distance and incli- nation remain constant (Duncan, Quinn & Tremaine 1987). Hence we may build a simple model of the sys- tem, whereby debris particles initially on circular orbits with semimajor axes equal to aplt diffuse in x, whilst their pericentres remain at aplt. We calculate the distribution of x values at time t, where t (cid:29) tD(x0), using Equation 8. We then create a virtual scattered disc consisting of a large number of particles, with semimajor axes drawn from the above distribution and pericentres equal to aplt. For each orbit we calculate the instantaneous radial dis- tance r of the particle at a randomised mean anomaly, and calculate the surface density profile resulting from the summation of these r values for all particles. Regard- less of the parameters used (planet mass, semimajor axis and stellar mass) this profile always goes as r−3.5. Hence an r−3.5 profile appears to be a natural consequence of scattering, and a population of scattered material could potentially be identified from such a slope. 2.3 Planet evolution Unlike when the planet is much more massive than the disc, if the two are of comparable mass then the planet's orbit may undergo significant evolution. It was noted in c(cid:13) 2002 RAS, MNRAS 000, 1 -- 14 Equal-mass planet disc interactions 5 Section 2.1 that secular perturbations from the disc cause the planet's orbit to precess, and its orbital plane will also evolve if the planet and disc planes are initially mis- aligned (although no significant plane evolution will occur if the two are roughly coplanar at the start of the inter- action). However the planet's eccentricity may evolve sig- nificantly, through planet-particle scattering and secular interactions with the disc. These mechanisms individu- ally affect eccentricity in different ways, so overall the planet's eccentricity behaviour combines two effects. Sec- ular perturbations from the disc will cause the planet's eccentricity to increase and decrease periodically, whilst scattering damps the eccentricity and will circularise the orbit (given enough scattering events). Hence the planet's eccentricity will undergo a long-term decline, with ad- ditional oscillatory behaviour in the meantime. If the planet scatters sufficient material before circularisation then there may be too little debris remaining to continue the damping process; in this case, the planet's eccentric- ity may not tend to zero but to some higher value. Scattering will also change the planet's semimajor axis. For a single planet scattering debris, the lack of inte- rior planets to remove material scattered inwards means than particles may only leave the system through colli- sions or ejection. The former mechanism will be rare as the star and planet pose small targets, hence the even- tual location of scattered material is likely exterior to its initial orbit. The planet will lose energy to counter this increase in particle energy, hence its semimajor axis will tend to decrease. In section 5.3 of Pearce & Wy- att (2014) we derived a theoretical upper limit on this semimajor axis change, and showed that a planet cannot undergo significant migration if much more massive than the disc except for a contrived set of circumstances. The same arguments still apply even when the planet and disc are of comparable mass; in the context of the pa- rameters in equations 14-16 of Pearce & Wyatt (2014), we require Γ ∼ 1 for significant migration, which is un- likely for broad discs. Hence scattering is unlikely to cause any significant change in planet semimajor axis. Recall- ing that secular interactions also have no effect on this quantity, we conclude that the semimajor axis of an ec- centric planet interacting with a comparable-mass debris disc is unlikely to evolve significantly. 2.4 General numerical simulations We now present n-body simulations of an eccentric planet interacting with a comparable-mass coplanar debris disc, to demonstrate the physical effects described in Sections 2.1 to 2.3. We ran almost 100 n-body simulations using the Mercury 6.2 integrator (Chambers 1999), covering a broad region of parameter space. The general simulation setup is as follows. A planet of mass Mplt orbits a star of mass M∗, with an initial semimajor axis aplt and eccen- tricity eplt. The planet's pericentre is typically of order 1-10 au; this is roughly the location of the water snow line for solar-type stars, and hence the region where giant planets may be expected to form. The planets have initial eccentricities ranging from 0.1 to 0.9. We fix M∗ = 1M(cid:12); Equal-mass planet disc interactions 6 Figure 4. Example n-body simulation of an interaction be- tween an eccentric planet and an equal mass, coplanar debris disc, with the simulation parameters described in the text. The left panels show the debris surface density and the planet's orbit (white ellipse), at time zero and then at subsequent evo- lutionary stages. The right panels show the radially averaged surface density at these times; the thick black lines are the sur- face density profiles, the dashed lines are the analytic surface density at t = 0, and the points show the planet's semimajor axis and pericentre/apocentre distances. The red line on the Stage 3 and 4 surface density plots shows an r−3.5 profile, typical of scattered debris, and material beyond 150 au fol- lows this profile. The evolutionary stages are common to all our simulations, and are described in Section 2.4. changing this parameter affects the timescales in the in- teraction, but not the nature of the evolution. The star also hosts a debris disc exterior to the planet's pericentre, of mass Mdisc (where Mplt = Mdisc), composed of N equal-mass particles. The disc midplane lies in the planet's orbital plane. We consider discs with initial inner and outer radii (r1 and r2 respectively) of the order of 10 − 100 au. The semimajor axes a of disc particles have initial values between r1 and r2, and are distributed such that n(a) ∝ a1−γ, (10) where γ is the surface density index. These particles are initially on circular orbits, which are randomised in lon- gitude of ascending node and have inclinations up to an opening angle I with respect to the disc midplane. We use a pre-interaction opening angle of I = 5◦, that of the classical Kuiper Belt (Bernstein et al. 2004), and γ = 1.5, that of the Minimum Mass Solar Nebula (Hayashi 1981), in our simulations. We probe disc and planet masses from 0.1 Earth masses (0.1 M⊕) to 3 Jupiter masses. The discs contain N = 103 − 104 equal-mass debris particles, representing the more massive bodies (i.e. those unaffected by radi- ation pressure and PR-drag); hence only gravitational forces are included. Each particle exerts a force on the planet and vice-versa, but does not perturb other debris. Thus we ignore the self-gravity of the disc; this is dis- cussed in Section 4.1. Each simulation lasts of the order of 10 − 100 Myr. Despite the broad range of parameters tested, the interaction always produced the same qualitative results. We found four distinct evolutionary stages occurring on logarithmic timescales, which we describe below. We also present an example simulation; we show the disc surface density at each evolutionary stage on Figure 4, and the planet's eccentricity evolution on Figure 5. The example shows a 10 M⊕ planet interacting with a comparable- mass disc, with aplt = 40 au, eplt = 0.6, r1 = 50 au, r2 = 150 au and N = 103. Stage 1: The planet begins to scatter material from the inner regions of the disc, depleting the debris surface density inwards of planet apocentre. Non-scattered par- ticles with orbits crossing that of the planet have their eccentricity increased via the planet's secular influence; these orbits are preferentially misaligned by ± ∼ 70◦ to the planet's orbit, due to the effect described in Beust et al. (2014). Particles beyond the planet's apocentre still have roughly circular orbits, and the similar secular phases of neighbouring particles cause the formation of a spiral-shaped overdensity beyond the planet's orbit (Wyatt 2005). The planet's eccentricity undergoes its most rapid decline due to debris scattering, and secular effects may also cause this eccentricity to oscillate. Stage 2: All debris initially crossing the planet's orbit has been scattered at least once; an overdensity of scattered material forms along the planet's orbit, and this overdensity is strongest around planet apocentre. A population of scattered material with surface density c(cid:13) 2002 RAS, MNRAS 000, 1 -- 14 Initialy / au−1000100x / au−10001000 Myrr / au050100150200Normalised Σ0.00.20.40.60.81.0Stage 1y / au−10001000.5 MyrNormalised Σ0.00.20.40.60.8Stage 2y / au−10001005 MyrNormalised Σ0.00.20.40.60.8Stage 3y / au−100010050 MyrNormalised Σ0.00.20.40.60.8Stage 4y / au−1000100x / au−1000100500 MyrNormalised Σ0.00.20.40.60.8r / au050100150200Equal-mass planet disc interactions 7 Hence the innermost peak of the surface density pro- file has been reduced or even removed. An overdensity of scattered material may still exist just exterior to the planet's orbit; this material no longer comes close to the planet since the latter's eccentricity decreased, so this de- bris is now stable. Particles driven to high eccentricities by secular effects early on may now have their eccen- tricities frozen, as the forcing eccentricity becomes small owing to the decrease in planet eccentricity. If the planet circularisation timescale is much longer than the secu- lar timescale of the outermost particles then surviving non-scattered debris forms a smooth disc apsidally an- tialigned with the planet; otherwise, the spiral overden- sity may still be present in the outer debris and remain there indefinitely. Generally, whilst the planet's eccentricity and longi- tude of pericentre evolve significantly throughout the in- teraction, its other orbital elements remain roughly con- stant. In the example simulation aplt changes by less than 5 per cent, and iplt never exceeds 0.6◦ (from an initial value of 0◦). The qualitative results presented in this section are general. The quantitative results will differ for specific systems, but rough scaling rules can be applied. For ex- ample, increasing the mass of the planet and disc simul- taneously will decrease the interaction timescales, whilst increasing the planet semi-major axis and disc radii will increase timescales. Increasing the planet eccentricity will decrease the circularisation timescale, and moving the disc mass inwards (either through reducing r1 or increas- ing γ) makes the planet circularise faster and to a greater degree. Whilst this paper primarily considers the case where Mplt = Mdisc, the results are applicable to the Mplt > Mdisc regime too. Even if the planet were orders of mag- nitude more massive than the disc, the general secular behaviour is the same as in the equal mass case. A differ- ence between the two mass regimes is that the transition between aligned and antialigned particles occurs farther from the star if the planet is more massive than the disc; this is because the planet would precess more slowly rel- ative to debris than the equal mass case, so the location where particles precess more slowly than the planet is far- ther from the star (see Figure 2). This is why we did not observe this secular behaviour in Pearce & Wyatt (2014); our discs simply did not extend far enough outwards to probe this regime. The main qualitative evolutionary dif- ference between the equal mass case and that when the planet is much more massive is that the planet will not undergo the same degree of orbital evolution in the latter regime. An interesting result may occur if 1 (cid:46) Mplt/Mdisc (cid:46) 10, whereby particles can change between the aligned and antialigned secular regimes. This occurs because the planet initially precesses rapidly (leading to antialign- ment of some particle orbits), yet the planet is massive enough to eject a significant fraction of the disc particles. The declining disc mass causes the planet precession to slow, meaning that the precession rate of some particles can "overtake" that of the planet. The final result of such an interaction is the formation of a coherently eccentric Figure 5. Eccentricity evolution of the planet in the simula- tion shown on Figure 4, as an example of the general behaviour observed in all our simulations. At early times (up to ∼ 10 Myr) secular eccentricity oscillations are noticeable, on top of the long-term decline from debris scattering. The dotted lines and numbers in boxes refer to the stages of system evolution, for comparison with Figure 4. going as ∼ r−3.5 begins to form beyond the planet's orbit, extending beyond the initial outer edge of the disc. This population is initially small, and hence is not visible on Figure 4 until the final panel. However a loga- rithmic surface density plot demonstrates that a ∼ r−3.5 population has begun forming by the second stage. Material originating just exterior to the planet's orbit may form a coherently eccentric disc apsidally aligned with the planet, depending on the planet's precession rate and eccentricity (see Section 2.1). The spiral-shaped overdensity continues to develop beyond the planet's orbit. The debris surface density profile hence has two peaks: a broad peak of scattered material stretching between the planet's initial pericentre and apocentre distances, and sharp peak farther out corresponding to the spiral overdensity. Stage 3: At least one secular period has elapsed for particles initially orbiting just exterior to the planet's apocentre; some initially stable material has been driven onto eccentric orbits apsidally antialigned with the planet, crossed its orbit and been scattered. Hence the surface density of scattered material tracing the planet's orbit is increased, and a large crescent shaped gap forms in the disc in the direction of planet pericentre. The spiral overdensity exterior to the planet continues to move outwards, as more distant particles are still in secular phase with their neighbours. The planet's eccentricity may by now have reduced significantly. Note that the planet's location corresponds to a region of overdensity in the disc, rather than the region of underdensity (as might naıvely be expected). Stage 4: The planet has scattered all material crossing its orbit, and its eccentricity evolution essentially ceases. c(cid:13) 2002 RAS, MNRAS 000, 1 -- 14 21t / Myreplt0.00.10.20.30.40.50.60246810431001000disc (as in Pearce & Wyatt 2014), but with a more messy structure due to additional particles which have changed their secular behaviour. We do not wish to comment on the case where Mplt < Mdisc, as disc self gravity would be very important in this regime and thus the results of this paper probably do not apply there (see Section 4.1). Our results may be used to predict the outcome of an eccentric planet interacting with a coplanar debris disc of the same or greater mass. They may also be used to infer the presence of an unseen perturber from the structure of an imaged debris disc, as we will now demonstrate for HD 107146. 3 APPLICATION TO HD 107146 HD 107146 is a 80-200 Myr old G2V star, located 27.5 pc from the Sun (van Leeuwen 2007; Williams et al. 2004). In 2000, IRAS imaging revealed excess infrared emission in the stellar spectrum, indicative of a debris disc (Silver- stone 2000). As noted in Section 1, further observations resolved the disc in both infrared emission and scattered light (Ardila et al. 2004; Williams et al. 2004; Carpenter et al. 2005; Corder et al. 2009; Hughes et al. 2011). For an excellent summary of work on HD 107146 up until 2011, see Ertel et al. (2011). The recent 1.25mm ALMA image reveals the disc at millimetre wavelengths in unprecedented detail (Ricci et al. 2015). These data show that the disc spans 30−150 au from the star and, assuming it is circular, inclined by 21◦ to the sky plane at a position angle (E of N) of 140◦. These observations detected 0.2M⊕ of dust at 1.25mm, and by extrapolating this up to bodies of diameter D = 1000 km (with the number of bodies of a given diameter n(D) ∝ D−3.6) the authors inferred a total disc mass of 100M⊕. The ALMA image and corresponding surface brightness profile are shown on the top two plots of Figure 6. It is clear from these plots that the disc has an un- usual morphology. The outer regions are brighter than the inner, and the brightness profile decreases with ra- dius before increasing again farther out. Lower-resolution 880µm SMA observations also show that the surface brightness does not decrease with radius as expected (Hughes et al. 2011). These data have been interpreted as the disc's surface density profile either being double- peaked or increasing with radius with a gap at 80 au (Ricci et al. 2015), and these models are indistinguishable at the observation resolution. Possible causes of these pro- file include embedded Pluto-sized objects inducing colli- sions between large debris bodies, or perturbations from an unseen planetary companion. The ALMA observations failed to detect any CO gas, suggesting that a dust-gas interaction is not responsible for the disc morphology. We wish to ascertain whether a past (or ongoing) interaction between the disc and a hypothetical eccen- tric planet can explain the disc features, and if so, esti- mate the pre-interaction orbit of the planet as well as its present day location. We assume the planet originated interior to the disc, where some event placed it onto an eccentric orbit; this could have been a planet-planet scat- tering or merger event, for example (Lin & Ida 1997; Ford c(cid:13) 2002 RAS, MNRAS 000, 1 -- 14 Equal-mass planet disc interactions 8 & Rasio 2008). We aim to reproduce the 1.25mm ALMA observations with an n-body simulation of such an in- teraction. Millimetre grains are unaffected by radiation pressure and PR-drag, so should act as tracers of the par- ent debris bodies (those most important for the dynamics of the system). Hence the ALMA data is well-suited to modelling with purely gravitational n-body simulations. 3.1 Simulation setup In addition to the simulations described in Section 2.4, we ran a further ∼ 150 simulations specifically aimed at reproducing the HD 107146 debris disc. We describe their setup now. HD 107146 is a G2V star, so we fix its mass at 1M(cid:12). We also fix the disc mass to the 100M⊕ value of Ricci et al. (2015). However this still leaves nine physical variables: the planet's mass, its initial semimajor axis, eccentricity, inclination and argument of pericentre, and the disc's initial inner and outer radii, opening an- gle and surface density profile. Computational limitations prevent us from exploring this whole parameter space, so we make several assumptions about the pre-interaction system to reduce the number of variables. We again fix the pre-interaction disc opening angle at I = 5◦ and γ = 1.5, and again assume the planet initially orbits in the disc midplane. These assumptions leave five physical parameters: Mplt, aplt, eplt, r1 and r2. However we can use physical reasoning to fix a further two of these. Firstly, the disc of HD 107146 appears to be roughly axisymmetric. In Pearce & Wyatt (2014) we showed that if Mplt (cid:29) Mdisc, the planet's eccentricity will not be significantly damped by the disc, and external debris will form a coherently eccentric disc aligned with the planet's orbit. Conversely, in Sections 2.3 and 2.4 we showed that if Mplt ∼ Mdisc then the planet's eccentricity will be significantly damped, and hence the outer edge of the disc will remain roughly circular. Thus if HD 107146's disc is interacting with a reasonably eccentric planet (or did so in the past) then Mplt ∼ Mdisc, so we fix the planet mass to be 100M⊕ in our simulations. This means that the outer edge of the disc will be largely unchanged by the interaction, so we fix r2 = 151 au (which best fits the data in the outermost regions). Again, we use N = 103 − 104 equal-mass debris particles to simulate the disc, and omit disc self-gravity (see Section 4.1). We are thus left with three free parameters: the ini- tial values of aplt and eplt, and the initial inner disc radius r1. These three are somewhat degenerate, so we cannot fix any at a single value. Instead, we run simulations with aplt = 20, 30, 40 and 50 au (noting that the inner peak of the observed surface density profile is at 50 au, and that the planet's initial semimajor axis is typically interior to this peak in our general simulations), and for each aplt we run simulations with various values of eplt and r1. We disfavour simulations where the planet's initial pericentre is within 3 Hill radii of the disc inner edge, or external to this location; in these cases the disc would be unstable before the interaction started. Once our simulations are complete we compare them to the observations, using the method described below. Equal-mass planet disc interactions 9 Figure 6. ALMA observations of HD 107146, along with our best-fitting simulation. Top left: ALMA 1.25mm continuum image (Ricci et al. 2015). Note that we use a different colour scale to that in the aforementioned paper. The white ellipse represents the beam size and orientation. Top right: points show the normalised, radially-averaged surface brightness profile of the disc as observed by ALMA, measured using elliptical apertures. The solid line is the profile from our best-fitting n-body simulation, at the time (19 Myr after the start of the interaction) of the best fit; the two agree with a reduced χ2 value of 0.4. The simulation parameters and the method used to compare the data and simulation are described in Section 3. Bottom left: positions of debris particles in the best-fitting n-body simulation at 19 Myr. The x − y plane is the initial disc midplane, with planet pericentre initially pointing along the x axis. The orbit of each particle has been populated with 100 points with randomised mean anomalies, to increase the effective number of particles plotted. The white point is the star, and the white ellipse the planet's orbit. Bottom right: simulated ALMA image of the n-body disc. The particles have been scaled for emission, the image rotated, and smoothed with a 2D Gaussian representing the ALMA beam (white oval). Compare this to the ALMA observation in the top left, noting that we have not added noise and hence our image is smoother. 3.2 Constructing simulated observations Throughout each simulation we compare the instanta- neous distribution of debris to that observed by Ricci et al. (2015). This requires the simulated debris to be converted into an image and surface brightness profile as would be observed by ALMA, for which we use the fol- lowing method. Firstly, we populate each particle's orbit with 100 points at randomised mean anomalies, to in- crease the effective number of particles simulated. We then scale for emission by weighting each point by a black body; a point at radial distance r from the star is weighted to have a luminosity L, where Here Bν (λ, T ) is the spectral radiance of a body of tem- perature T at a wavelength λ, given by Planck's law: (cid:20) (cid:18) hc (cid:19) λkBT (cid:21)−1 − 1 Bν (λ, T ) ∝ exp , (12) where h is the Planck constant, c is the speed of light and kB is the Boltzmann constant. The temperature of the body is determined by the flux it receives from the star (again assuming black body behaviour), hence (cid:18) L∗ (cid:19)1/4 T = 4πσ −1/2 r (13) L(r) ∝ Bν (λ, T ). (11) where L∗ is the star's luminosity and σ is the Stefan- Boltzmann constant. For HD 107146, we use L∗ = L(cid:12) and λ = 1.25mm, that of the ALMA observations. For c(cid:13) 2002 RAS, MNRAS 000, 1 -- 14 1.00.50.0Normalised surface brightnessDec offset / arcsec−6−3036RA offset / arcsec−6−3036Normalised surface brightness-0.20.00.20.40.60.81.01.2r / au050100150200y / au−200−1000100200RA offset / arcsec−200−10001002001.00.50.0Normalised surface brightnessDec offset / arcsec−6−3036RA offset / arcsec−6−3036these parameters, Bν (λ, T ) roughly scales as r−1/2. For this analysis we have assumed the disc is optically thin; this is valid here since, whilst the disc is massive, it covers a broad region. Whilst the optical depth could be high if the disc were extremely thin, even a moderate 5◦ open- ing angle would be large enough that we do not have to consider flux attenuation here. To produce images for comparison with the ALMA observations, we rotate the simulated (emission scaled) disc to an inclination of 21◦ and a position angle of 143◦. We then convolve our image with a two-dimensional Gaussian to simulate the ALMA point spread function (PSF); this Gaussian has a standard deviation along its major axis of 13.4 au, along its minor axis of 9.8 au, and its major axis has a position angle of 19.8◦. We also calcu- late the radially averaged surface brightness profile of the simulated disc, using elliptical apertures on the simulated image as in Ricci et al. (2015). We may then compare our simulations to the ALMA observations of HD 107146. 3.3 Fitting the HD 107146 disc We identified the simulations which best replicate the HD 107146 disc using a χ2 analysis. At 100 time intervals throughout each simulation we calculated the χ2 value comparing the observed radial surface brightness profile with the simulated profile at this time (found using the method in Section 3.2). On Figure 7 we plot the aplt, eplt, r1 parameter space tested in our simulations, and colour each point by min(χ2 red) (the minimum value of reduced χ2, that is the minimum value of χ2 attained during that simulation, divided by the number of degrees of freedom). If the data are independent, a reduced χ2 of order 1 means that the obs1ervations are consistent with the model, and the smaller the value, the better the fit (although values much smaller than 1 imply the data is overfitted). Here the observed surface brightness profile points are correlated with each other, so little should be inferred from the exact value of min(χ2 red) itself; however this value does allow a comparison between simulations, to identify that which best reproduces the HD 107146 disc. Figure 7 shows that there are several regions of tested parameter space which produce discs consistent with ALMA observations. The best fit is attained using a planet with initial semimajor axis aplt = 40 au and eccen- tricity eplt = 0.4, interacting with a disc with initial in- ner radius r1 = 50 au. For this case the simulated surface brightness profile is most similar to the ALMA observa- tions 19 Myr after the start of the interaction, and we plot the simulated ALMA image and surface brightness profile at this time on the lower two plots of Figure 6. These well reproduce the observations; the surface brightness profile yields a min(χ2 red) value of 0.4, and the simulated image resembles the ALMA observation by eye. Note that glob- ular structures in the observed image are probably noise, which has not been accounted for in the simulated image and hence the latter appears smoother than the observa- tion. This best fit occurs when the simulated system is at stage 3 or 4 in its evolution (as described in Section 2.4); the planet has removed most of the material crossing its c(cid:13) 2002 RAS, MNRAS 000, 1 -- 14 Equal-mass planet disc interactions 10 orbit, and its orbital evolution has essentially stalled. By this point the planet's eccentricity has decreased from 0.4 to 0.05, whilst its semimajor axis (initially 40 au) has only red ∼ 1 at reduced by 4 au. The simulation first reaches χ2 10 Myr, and this parameter remains less than 1 until the end of the simulation (at 30 Myr); hence the simulation also provides a good fit to the observations over a long time interval. However the best-fitting simulation is not unique in reproducing the observations. A well-defined χ2 mini- mum also exists for a planet semimajor axis of 30 au, cen- tred on eplt = 0.55 and r1 = 50 au, and this is almost as good as the best-fitting 40 au solution (min(χ2 red) = 0.5). Again, the planet in the aplt = 30 au simulation un- dergoes minimal semimajor axis evolution whilst its ec- centricity is significantly reduced, and the simulation fits best once it has evolved to stage 3 or 4 and resembles the observations for a long time. Conversely, we find that planets with initial semimajor axes of 20 and 50 au do not reproduce the observed disc well. Also note that our well-fitting simulations have the disc's initial inner edge exterior to its present day value, and material has since been scattered inwards by the planet. In conclusion, a planet with an initial semimajor axis of 30 or 40 au and an eccentricity of 0.4-0.5, interacting with a comparable- mass debris disc with initial inner edge at 50 au, can well reproduce the disc of HD 107146. At present, the planet is most likely on a roughly circular orbit at 30-40 au. 4 DISCUSSION We have examined the general interaction between an eccentric planet and a coplanar, comparable-mass debris disc, and applied our results to HD 107146 in an attempt to explain its unusual disc. In this section we discuss a possible limitation of our work: the omission of disc self- gravity. We also discuss the timescale of the HD 107146 interaction. Finally, we examine the implications of this paper for planet searches, both for general systems with debris discs and also for HD 107146. 4.1 Disc self-gravity Our simulated debris particles exert a force on the planet (and vice-versa), but do not interact with each other; hence we do not include disc self-gravity in our simu- lations. This omission dramatically increases computa- tional efficiency, allowing us to run several hundred sim- ulations for this paper. However whilst self-gravity does not affect the interaction outcome if the planet is much more massive than the disc (as in Pearce & Wyatt 2014), if the two are of comparable mass then this effect could become important. Debris in a self-gravitating disc would undergo ad- ditional secular and scattering evolution from the influ- ence of other disc particles. Secular interactions work over large distances on timescales scaling inversely with the object masses, whilst scattering works over short dis- tances on timescales going as the inverse-square of the masses (see equations 17 and 18 in Pearce & Wyatt 2014). Equal-mass planet disc interactions 11 Figure 7. Reduced χ2 of our simulated radially-averaged surface density profiles compared to that of HD 107146, at the time in each simulation when this parameter is minimised. We varied the initial inner disc radius r1 and the initial planet eccentricity eplt for four initial planet semimajor axes aplt, fixing all other parameters as described in the text. The points show our simulations, with the colourmap and contours interpolated between them. Contours show log10[min(χ2 red)] = 0, 0.5 and 1 respectively. The hatched regions show an unphysical area of parameter space, where the planet's initial pericentre is closer than three Hill radii to the disc inner edge (see Section 3.1). (cid:90) r2 (cid:20) Therefore given the small debris particle masses, the ma- jor effect of self-gravity is likely to be on the secular evo- lution of the disc. We investigate the possible secular effect of self- gravity by analytically calculating the precession rate of a test particle embedded in a disc. This gives us a feel for how the initial disc in our best-fitting HD 107146 simulation would evolve due to self-gravity alone (in the absence of any planetary perturbations), and allows us to compare the magnitude of the self-gravity effect to that of the planet. To calculate the precession rate, we consider a test particle at position (R, φ, z) in cylindrical coordinates, which experiences a force from a 2 dimensional, axisym- metric disc in the z = 0 plane. Equation 2-146 in Binney & Tremaine (1987) gives the radial acceleration of the particle due to the disc as Fr(R) = − G R3/2 − R (cid:18) R(cid:48) × R r1 R(cid:48) + z2 R(cid:48)R K(k) − 1 4 (cid:19) E(k) (cid:21) k2 1 − k2 (cid:48) kΣ(R √ R(cid:48)dR (cid:48) , (14) ) where R(cid:48) is the radial location of a point in the disc, k2 ≡ 4RR(cid:48) (R + R(cid:48))2 + z2 , (15) and K(k) and E(k) are the complete elliptical integrals of the first and second kind respectively. We wish to con- sider a particle in the disc midplane, i.e. z = 0. However Binney & Tremaine (1987) note that Equation 14 has an unphysical singularity at R = R(cid:48) if z = 0, because k = 1 here and so the K(k) and (1− k2)−1 terms become unde- fined. This issue can be resolved by setting 0 < z (cid:28) R, so we use z = 10−4 au in our evaluation. A particle in the midplane experiences no vertical acceleration, and its tangential acceleration is also zero because the disc is axisymmetric. Hence the self-gravity of an axisymmet- c(cid:13) 2002 RAS, MNRAS 000, 1 -- 14 ric disc exerts only a radial force on a disc particle. The precession rate of a particle at true anomaly f moving in a Keplerian potential and perturbed by an additional radial force Fr is ω = − 1 e a(1 − e2) µ Fr cos f (16) (section 2.9 of Murray & Dermott 1999), and we average this over the orbital period T : (cid:115) (cid:90) T 0 (cid:90) 2π 0 (cid:104) ω(cid:105) ≡ 1 T ωdt ≈ √ 1 1 − e2 2πa2 ωr2df, (17) where we have assumed that a and e are constant over one orbital period. Finally, if e (cid:28) 1 then Fr will not vary significantly over the particle's orbit. In this case (cid:114) a µ (cid:104) ω(cid:105) ≈ Fr(a), (18) and similar analyses for semimajor axis and eccentricity yield a ≈ e ≈ 0. We evaluated Equation 14 numerically for a disc with the initial parameters of that in our best- fitting HD 107146 simulation. The force, and the resulting precession rate, are shown as functions of radius by the black lines on Figure 8. The plot shows that the precession rate of particles inwards of 75 au is still dominated by the planet, even when disc self-gravity is considered. In the n-body sim- ulation (without self-gravity), the planet drives up the eccentricities of these particles and scatters the majority of them, with the remainder forming an eccentric disc aligned with the planet's orbit. Hence this would still oc- cur with the inclusion of self-gravity. However the disc's gravity may initially dominate beyond this region; in the n-body simulation, particles beyond 80 au preferentially antialign with the planet's orbit, so debris out to 100 au is removed. Self-gravity would effectively cause these aplt = 20 aur1 / au20304050607080eplt0.20.40.60.8aplt = 30 aueplt0.20.40.60.8aplt = 40 aueplt0.20.40.60.8log10[min (χ2 red)]-0.50.00.51.01.5aplt = 50 aur1 / au20304050607080eplt0.20.40.60.8Equal-mass planet disc interactions 12 107146 simulation, the inclusion of disc self-gravity would not qualitatively affect the resultant disc structure inte- rior to 75 au and exterior to 95 au. In the region between these radii, the potential effect of self-gravity is unclear. This region might not undergo the same level of depletion as in the simulations, and the spiral structure visible in Figure 6 might not be present. Hence our observational fit might not be as good as that on Figure 6. More gener- ally, depending on the simulation parameters, self-gravity may affect our predicted outcomes for the interaction in- vestigated in this paper. The main effect of self-gravity would probably be the reduction of debris depletion in the region immediately interior to the outermost peak of the disc. However we stress that a more sophisticated self- gravity analysis is required to fully explore its potential effect, which is beyond the scope of this paper. 4.2 HD 107146 interaction timescale Our best-fitting HD 107146 simulations all reproduce the observed disc ∼ 10 Myr after the start of the interac- tion, compared to the 100 Myr age of the star. These two timescales are compatible, but scenarios in which they are comparable would be preferable. The interac- tion timescales are set by the disc (and hence planet) mass and, since the disc mass derived from observations is uncertain (Ricci et al. 2015), there is scope to change this in our simulations. The secular interactions between the planet and disc are the dominant effects in the sim- ulations, and Equations 4 - 7 show the secular preces- sion rate to scale linearly with mass whilst the forcing eccentricity is independent of mass. Equation 14 shows that the effect of disc self-gravity also scales linearly with disc mass, so scaling both Mplt and Mdisc simul- taneously will not change the importance of self-gravity relative to planetary perturbations. Hence changing the disc and planet masses should affect the secular interac- tion timescales, but not the nature of this interaction. Reducing the masses will make the planet less efficient at ejecting debris, but seeing as the main effects of the in- teraction are secular in nature, this should not affect the outcome too much. Hence if we reduce the disc and planet masses in our simulations by an order of magnitude (so Mdisc ∼ 10M⊕), then roughly the same interaction will occur over a timescale comparable with the stellar life- time. Hence whilst our interaction timescales are by no means incompatible with the system age, if we assume that this interaction is responsible for the observed disc structure then our results might suggest the disc mass is closer to 10M⊕ than 100M⊕. Alternatively the planet may have only recently been placed on an eccentric orbit, and we happen to have observed the system at this stage in its evolution. 4.3 Implications for planet searches Our findings have interesting implications for the infer- ence of unseen planets from debris disc features, both generally and for HD 107146. An important result is that the planets in this interaction generally circularise with little change in semimajor axis, having ejected much of Figure 8. The evolution of a test particle under the influence of a massive axisymmetric disc. Top plot: the radial force im- parted by the disc on a coplanar particle at radius R, from Equation 14. The black line shows a 100 M⊕ disc with an r−1.5 surface density profile, with inner and outer radii of 50 and 150 au respectively. These are the initial disc parameters in the best-fitting HD 107146 simulation. The red line shows the same disc but with all mass inwards of 75 au removed, representing the truncation of the disc by the planet. Bottom plot: the magnitudes of the resulting particle precession rates. The green dotted line shows the precession rate due to the sec- ular influence of a 100 M⊕ planet with aplt = 40au (Equation 6). The plot shows that the planet dominates particle evolu- tion in the inner regions of the disc, whilst disc self-gravity may be more important in the outer regions. particles' precession rates to be uncorrelated with the planet's evolution, preventing both preferential antialign- ment and also significant eccentricity excitation. So were disc self-gravity included, the depletion of the disc in the best-fitting HD 107146 simulation would initially extend out to 75 au rather than 100 au. This would not fit the observational data. However we have not yet considered the evolution of the disc self-gravity. Particles inwards of 75 au would be depleted even with self-gravity, and this would change the disc potential. The red lines on Figure 8 show a disc with the same parameters as discussed above, but with all mass inwards of 75 au removed. Now the planet is still influential out to about 95 au, so much of this debris may still eventually undergo scattering. Beyond this re- gion the disc self-gravity will always dominate, although in the simulation these particles were not significantly perturbed by the planet anyway. Hence the inclusion of self-gravity will not affect the overall simulation results in the outermost regions. Our analysis suggests that, for our best-fitting HD c(cid:13) 2002 RAS, MNRAS 000, 1 -- 14 Fr / 10-6 au yr-2−202468ϖ / rad yr-110−810−710−610−5R / au050100150200the debris interior to their final orbital distance. Hence the eventual location of the planet is typically near the inner edge of the disc, on an orbit traced by a debris over- density, beyond which lies a gap followed by another over- dense ring. This configuration would not otherwise be ex- pected; having observed a double-peaked debris disc, the naıve assumption would be that any perturbing planet lies in the underdense region between the two peaks. Hence future planet finding missions should not be dis- couraged if no planets are found in a debris disc gap; indeed, the absence of planets in this region may hint at a violent dynamical history, and could motivate the search for planets near the inner edge of the disc instead. If our hypothesis on the evolution of the HD 107146 system is correct, then a ∼ 100M⊕ planet currently orbits near the inner edge of the disc, with a semimajor axis of 30− 40 au and an eccentricity of ∼ 0.1. This planet orig- inated interior to the disc; assuming it was scattered out of its original location by another body then, based on its initial pericentre, a second companion with mass at least equal to that of the scattered planet exists at (cid:38) 10−25 au from the star. Companions of less than 10 Jupiter masses (3000M⊕) have not been ruled out anywhere in the sys- tem by imaging (Apai et al. 2008), so this scenario is possible and could be tested with deeper planet searches. Furthermore, our simulations suggest that the outermost debris peak actually forms a thin spiral, rather than a continuous ring. This structure would be detectable in observations with ∼ 3 times the resolution of the ALMA image, and such a detection would favour our hypothesis on the history of the system (although with the caveat that the disc self-gravity would also have an effect, and may partially or completely wash-out this spiral). Such a resolution may well be possible with current instrumen- tation. Another potential application of this work is to HD 92945; this system may also harbour a double-peaked de- bris disc (Golimowski et al. 2011), so our results could be used to invoke a perturbing planet in that system too. However the HD 92945 disc was imaged in scattered light, so the emitting dust would be affected by radiation forces. Hence it is unclear without more detailed analysis whether the more massive debris also follows this double- peaked profile (as in HD 107146), or whether the observed morphology is a consequence of non-gravitational forces on small dust. 5 CONCLUSIONS Broad, double-ringed debris discs could potentially have evolved to their present state under the influence of an eccentric, comparable-mass planet. We investigate this interaction in general, and show that it follows four dis- tinct stages on logarithmic timescales. A key result is that planet precession may cause distant debris orbits to anti-align with that of the planet, whilst the inner- most debris orbits align with the planet's. This results in distinct inner and outer debris regions with a gap or de- pletion between them, akin to the double-peaked debris structures potentially observed in HD 107146 and HD 92945. It also produces the counter-intuitive result that a c(cid:13) 2002 RAS, MNRAS 000, 1 -- 14 Equal-mass planet disc interactions 13 low-mass planet may clear a larger region of debris than a higher-mass body. In general the planet undergoes a rapid eccentricity decrease whilst its semimajor axis re- mains constant; thus if the planet initially scattered off another body then the two would quickly decouple, so our results still hold in the presence of additional mas- sive planets (providing the eccentricity damping is fast enough). We then modelled the HD 107146 system in detail, confirming that the debris disc's unusual morphology can be well explained by this interaction. If an unseen eccen- tric planet did sculpt debris into the structure seen today, then this hypothetical planet initially had pericentre in the inner regions of the system and apocentre within the disc itself; based on our best-fitting model, the planet is currently on a low-eccentricity orbit 30-40 au from the star. This is below the companion detection thresholds of current observations of the system, but could potentially be found by future imaging projects. 6 ACKNOWLEDGEMENTS We thank Luca Ricci for allowing us the use of his ALMA image, and Mher Kazandjian for discussions concerning the modelling of disc self-gravity. We also thank Herv´e Beust for his very constructive and helpful review. TDP acknowledges the support of an STFC studentship, and MCW is grateful for support from the European Union through ERC grant number 279973. REFERENCES Andrews S. M., Williams J. P., 2007, ApJ, 659, 705 Apai D. et al., 2008, ApJ, 672, 1196 Ardila D. R. et al., 2004, ApJ, 617, L147 Backman D. et al., 2009, ApJ, 690, 1522 Bernstein G. M., Trilling D. E., Allen R. L., Brown M. E., Holman M., Malhotra R., 2004, AJ, 128, 1364 Beust H. et al., 2014, A&A, 561, A43 Binney J., Tremaine S., 1987, Galactic Dynamics Carpenter J. M., Wolf S., Schreyer K., Launhardt R., Henning T., 2005, AJ, 129, 1049 Chambers J. E., 1999, MNRAS, 304, 793 Corder S. et al., 2009, ApJ, 690, L65 Dent W. R. F. et al., 2014, Science, 343, 1490 Duncan M., Quinn T., Tremaine S., 1987, AJ, 94, 1330 Eiroa C. et al., 2013, A&A, 555, A11 Ertel S., Wolf S., Metchev S., Schneider G., Carpen- ter J. M., Meyer M. R., Hillenbrand L. A., Silverstone M. D., 2011, A&A, 533, A132 Faramaz V. et al., 2014, A&A, 563, A72 Ford E. B., Rasio F. A., 2008, ApJ, 686, 621 Golimowski D. A. et al., 2011, AJ, 142, 30 Hartmann W. K., Davis D. R., 1975, Icarus, 24, 504 Hayashi C., 1981, Progress of Theoretical Physics Sup- plement, 70, 35 Hughes A. M., Wilner D. J., Andrews S. M., Williams J. P., Su K. Y. L., Murray-Clay R. A., Qi C., 2011, ApJ, 740, 38 Ida S., Makino J., 1992, Icarus, 96, 107 Equal-mass planet disc interactions 14 Juri´c M., Tremaine S., 2008, ApJ, 686, 603 Kalas P., Graham J. R., Fitzgerald M. P., Clampin M., 2013, ApJ, 775, 56 Lin D. N. C., Ida S., 1997, ApJ, 477, 781 Murray C. D., Dermott S. F., 1999, Solar system dy- namics Pearce T. D., Wyatt M. C., 2014, MNRAS, 443, 2541 Pearce T. D., Wyatt M. C., Kennedy G. M., 2015, MN- RAS, 448, 3679 Rasio F. A., Ford E. B., 1996, Science, 274, 954 Rhee J. H., Song I., Zuckerman B., McElwain M., 2007, ApJ, 660, 1556 Ricci L., Carpenter J. M., Fu B., Hughes A. M., Corder S., Isella A., 2015, ApJ, 798, 124 Schneider G., Weinberger A. J., Becklin E. E., Debes J. H., Smith B. A., 2009, AJ, 137, 53 Schneider J., Dedieu C., Le Sidaner P., Savalle R., Zolo- tukhin I., 2011, A&A, 532, A79 Silverstone M. D., 2000, PhD thesis, UNIVERSITY OF CALIFORNIA, LOS ANGELES Tsiganis K., Gomes R., Morbidelli A., Levison H. F., 2005, Nature, 435, 459 van Leeuwen F., 2007, A&A, 474, 653 Walsh K. J., Morbidelli A., Raymond S. N., O'Brien D. P., Mandell A. M., 2011, Nature, 475, 206 Williams J. P., Najita J., Liu M. C., Bottinelli S., Carpenter J. M., Hillenbrand L. A., Meyer M. R., Soderblom D. R., 2004, ApJ, 604, 414 Wu Y., Lithwick Y., 2011, ApJ, 735, 109 Wyatt M. C., 2005, A&A, 440, 937 Yabushita S., 1980, A&A, 85, 77 c(cid:13) 2002 RAS, MNRAS 000, 1 -- 14
1910.05156	1	1910	2019-10-11T13:07:28	ROME/REA: A gravitational microlensing search for exo-planets beyond the snow-line on a global network of robotic telescopes	[ "astro-ph.EP", "astro-ph.IM" ]	Planet population synthesis models predict an abundance of planets with semi-major axes between 1-10 au, yet they lie at the edge of the detection limits of most planet finding techniques. Discovering these planets and studying their distribution is critical to understanding the physical processes that drive planet formation. ROME/REA is a gravitational microlensing project whose main science driver is to discover exoplanets in the cold outer regions of planetary systems. To achieve this, it uses a novel approach combining a multi-band survey with reactive follow-up observations, exploiting the unique capabilities of the Las Cumbres Observatory (LCO) global network of robotic telescopes combined with a Target and Observation Manager (TOM) system. We present the main science objectives and a technical overview of the project, including initial results.	astro-ph.EP	astro-ph	Draft version October 14, 2019 Typeset using LATEX default style in AASTeX62 9 1 0 2 t c O 1 1 . ] P E h p - o r t s a [ 1 v 6 5 1 5 0 . 0 1 9 1 : v i X r a ROME/REA: A gravitational microlensing search for exo-planets beyond the snow-line on a global network of robotic telescopes Yiannis Tsapras,1 R.A. Street,2 M. Hundertmark,1 E. Bachelet,2 M. Dominik,3 V. Bozza,4, 5 A. Cassan,6 J. Wambsganss,1, 7 K. Horne,3 S. Mao,8, 9 W. Zang,8 D.M. Bramich,10 and A. Saha11 1Zentrum fur Astronomie der Universitat Heidelberg, Astronomisches Rechen-Institut, Monchhofstr. 12-14, 69120 Heidelberg, Germany 2Las Cumbres Observatory Global Telescope Network, 6740 Cortona Drive, suite 102, Goleta, CA 93117, USA 3SUPA, School of Physics & Astronomy, University of St Andrews, North Haugh, St Andrews KY16 9SS, UK 4Dipartimento di Fisica "E.R. Caianiello", Universit`a di Salerno, Via Giovanni Paolo II 132, 84084 Fisciano, Italy. 5Istituto Nazionale di Fisica Nucleare, Sezione di Napoli, Via Cintia, 80126, Napoli, Italy. 6Institut d'Astrophysique de Paris, Sorbonne Universit´e, CNRS, UMR 7095, 98 bis boulevard Arago, 75014 Paris, France 7International Space Science Institute (ISSI), Hallerstrasse 6, 3012 Bern, Switzerland 8Department of Astronomy and Tsinghua Centre for Astrophysics, Tsinghua University, Beijing 100084, China 9National Astronomical Observatories, Chinese Academy of Sciences, 100012 Beijing, China 10New York University Abu Dhabi, PO Box 129188, Saadiyat Island, Abu Dhabi, UAE 11National Optical Astronomy Observatory, 950 North Cherry Ave., Tucson, AZ 85719, USA (Received -- ; Revised -- ; Accepted -- ) Submitted to PASP ABSTRACT Planet population synthesis models predict an abundance of planets with semi-major axes between 1-10 au, yet they lie at the edge of the detection limits of most planet finding techniques. Discovering these planets and studying their distribution is critical to understanding the physical processes that drive planet formation. ROME/REA is a gravitational microlensing project whose main science driver is to discover ex- oplanets in the cold outer regions of planetary systems. To achieve this, it uses a novel approach combining a multi-band survey with reactive follow-up observations, exploiting the unique capabilities of the Las Cumbres Observatory (LCO) global network of robotic telescopes combined with a Target and Observation Manager (TOM) system. We present the main science objectives and a technical overview of the project, including initial results. Keywords: stars: planetary systems -- gravitational lensing -- Galaxy: bulge -- methods: observa- tional 1. INTRODUCTION The prolific discoveries of planets orbiting distant stars over the past two decades have radically transformed our understanding of the properties of planetary systems (Howard 2013). Despite these discoveries, fundamental questions about the formation, physical properties and distribution of exoplanets still remain open due to the dearth of low-mass planets (Mp < 100M⊕) detected at large orbital distances from their host stars (Udry et al. 2003; Foreman-Mackey et al. 2014). Current formation theories postulate that proto-planetary cores form in metal-rich accretion disks surrounding the host star. These proto-planets co-evolve with the disk and can undergo orbital decay due to torque asymmetries in the surrounding disk material (Armitage 2011; Mordasini et al. 2012; Dullemond 2013). Corresponding author: Yiannis Tsapras [email protected] 2 The so-called snow-line is defined as the distance from a star beyond which the disk temperature drops below ∼160K and water turns to ice (Min et al. 2011). Theory predicts that beyond the snow-line, the formation of ice grains allows planetary embryos to develop sufficiently massive solid cores and gradually grow by accreting material from the surrounding gaseous disk, transforming them into gas giants. Population synthesis simulations predict an abundance of low and intermediate mass planets (Mp < 100M⊕) beyond the snow-line (Ida et al. 2013; Mordasini et al. 2009), but they remain exceedingly hard to detect and little is known about their properties. Indeed, even all of the transiting planets discovered by the highly accomplished Kepler space mission are far too close to their host stars to begin investigating these predictions (Batalha et al. 2013; Borucki et al. 2010). A recent statistical analysis of microlensing planets by Suzuki et al. (2018) found a discrepancy between the microlens- ing results and the number of intermediate-mass giant planets predicted by planet population synthesis simulations. Since the latter often rely on runaway gas accretion to produce gas giants, a standard assumption in core accretion theory, the microlensing results imply that there may be physical processes involved in giant planet formation that have been overlooked or underestimated in existing models. The significance of this apparent discrepancy between theory and observations has wide-ranging implications and can only be assessed by concentrated efforts to increase the sample of exoplanets discovered beyond the snow-line. Besides the tantalising possibility of discovering cold Earths, finding these planets is therefore crucial in understanding the physical processes that drive planet formation (Gaudi 2012). Each planet detection method is sensitive to a different domain of the planet distribution in mass and distance and the emerging pattern provides a basis for testing and developing our understanding of how planets form and how their orbits evolve. Gravitational microlensing detects planets by measuring how light rays from a background source star bend as they pass through the gravitational field of an intervening planetary system (lens) on their way to our telescopes (Mao & Paczynski 1991; Gould & Loeb 1992). What is actually observed during a microlensing event is a gradual increase in the brightness of the source star as the lens appears to move closer to it on the plane of the sky, followed by a gradual dimming back to its normal brightness as the lens moves away.The gravitational influence of planets in orbits of a few au around the lens star, typically an M or K-dwarf, can further bend the light rays coming from the source star. The presence of these unseen planets is then revealed through the detection of brief but intense changes in the measured brightness. This method opens up a unique window to the population of low-mass exoplanets at or beyond the snow-line, which is unavailable to other detection methods (Tsapras et al. 2016; Cassan et al. 2012; Gould et al. 2010). Microlensing is thus the fastest, cheapest1 and most time-efficient way to probe the population of exoplanets at moderate to large separations (1-10 au), exploring the region where rock/ice cores are predicted to grow and undergo runaway gas accretion2 and providing essential data to back-calibrate planet population synthesis models. Furthermore, the method is uniquely sensitive to planetary systems at distances of several kilo-parsecs, affording us a view of the true Galactic population of planets (Tsapras 2018; Street et al. 2018b). In Section 2, we discuss the motivation for the project and how it compares with and can benefit existing efforts within the microlensing community. A technical overview of the facilities and instruments used for the project is given in Section 3. The observing strategy is described in Section 4. Expected yields based on a simulation of a full observing season are given in Section 5. We conclude with the presentation of some initial results in Section 6. 2. PROJECT MOTIVATION ROME/REA3 is an observational science project running on the global robotic telescope network of the Las Cumbres Observatory (LCO) with the aim of discovering exoplanets beyond the snow-line of their host stars using the technique of gravitational microlensing. Microlensing event rates are highest in a ∼4 square degree area close to the Galactic centre due to the sheer number of available source and lens stars (Sumi et al. 2013). Although this area is in the observing footprint of existing surveys, their observations are typically obtained in a single band and only occasionally (weekly or monthly) in a second band, making it difficult to characterise the source star 4 (Udalski et al. 2015; Kim et al. 2016; Bond et al. 2001). 1 Ground-based observations every 15 minutes using modest 1m-class telescope facilities are sufficient to detect the signals of planets as small as the Earth (Dominik et al. 2006). 2 Provided the core mass has grown enough to be comparable to the mass of the surrounding gaseous envelope. 3 Robotic Observations of Microlensing Events/Reactive Event Assessment 4 Source stars are generally too faint for spectroscopy. (cid:16) 1 c2θ2 S (cid:17) , ML = 4Gρ2 − 1 DS DL 3 (1) Source star characterisation plays an important part in interpreting a microlensing event since the angular size of the source can be used to estimate the angular Einstein radius of the lens and thus its mass. The relevant equation is θS = ρθE, where θS is the angular source size, θE the angular Einstein radius and ρ the angular size of the source normalised to the angular Einstein radius of the lens (Witt & Mao 1994; Yoo et al. 2004). The latter is obtained during the model fitting process when finite source effects are detected in the event light curve. The mass of the lens can then be derived from where DL is the distance to the lens and DS the distance to the source. Therefore, knowledge of the distance and spectral type of the source (and hence its radius) is essential in order to constrain the mass of the lens, and if the distance to the lens can be inferred from the relative lens-source parallax, πrel (Gould 1992), then the lens mass can be uniquely determined. For the faint (I∼15-19mag) stars that populate our target fields, multi-band photometry provides this crucial information about the source. Furthermore, since the microlensing effect is achromatic, regular observations in multiple bands can also be used to distinguish between light coming from the source star, which contains the microlensing signal, and blended light from faint stars at roughly the same position on the sky as the source, which dilutes that signal by different amounts in different bands. We now turn to how this is done in practice. The ROME/REA project is conducted exclusively on the 1m telescopes that belong to LCO's network in the southern hemisphere. We use a novel observing strategy that relies on LCO's robotic telescope observing framework and which complements the scientific goals of other microlensing surveys by ensuring that the source stars within our survey footprint are well characterised and therefore that the physical parameters of the lenses can be well determined. We achieve this by extracting the source type through the use of time series data in three observing bands, which also allows us to constrain how blended the event is. Three bands is the minimum required to enable blend analysis using a colour-colour diagram. On this diagram, an unblended source will appear as a single point for the full duration of the microlensing event, whereas a blended source will trace a curved path as the event evolves5 because the flux ratio between pure-source and pure-blend fluxes is going to change as it is only the source that experiences magnification due to microlensing. Once the microlensing event has run its course, a line may be fit to the path. If the line is short, then it is an event with low blending. Conversely, if the line is long, the event is heavily blended. The length of the line is therefore a direct indicator of the degree of blending. Furthermore, the colour of the blend itself will change the slope of the line in the colour-colour diagram. Information about the stellar type of the source can be obtained by over-plotting the colour differences for (unblended) stars of known spectral types on this diagram and identifying their (fixed) locations. The location of the source, after accounting for extinction, can then be directly compared with these known positions in the diagram and the source star can be associated with a particular spectral type. The angular radius of the source θS can then be estimated using the relationships derived in Boyajian et al. (2014). A detailed first demonstration of the strategy and associated results can be found in Street et al. (2019). In this analysis, the microlensing event investigated (OGLE-2018-BLG-0022) was found to be due to a binary star with individual components of 0.375 ± 0.020 M(cid:12) and 0.098 ± 0.005 M(cid:12). For comparison, an independent analysis of the same event by Han et al. (2019), relying on the more common approach using only two bands and a more densely sampled light curve, determined the masses of the components to be 0.40 ± 0.05 M(cid:12) and 0.13 ±0.01 M(cid:12), respectively. This suggests that the three-band approach can potentially improve the accuracy of the mass estimates by at least a factor of two. Microlensing observations in the past have generally not been associated with extensive log-keeping and metadata structures6, so it is usually not possible to appreciate the reasoning behind human-driven observing decisions or to trace how different instrumental and environmental factors influence the quality of the data (Bachelet et al. 2015). In addition, surveys have redefined their observing fields and cadences several times over the years, which affects their relative sensitivity to planets at different regions of the sky and can lead to biased estimates of the planet mass function if not carefully accounted for. In our ROME survey, the observing strategy is software-driven and follows a predetermined pattern, while the REA target selection process is also entirely automated (see Section 4). All observing 5 If a flux ratio diagram (FRD) is used instead, this path will be straight. 6 These can contain useful information about the regions of sensitivity of the instruments used, information about local observing conditions at the time of the observation, etc. 4 Table 1. LCO sites and telescopes used by the ROME/REA project Observing site Siding Spring Observatory South African Astronomical Observatory Cerro Tololo Interamerican Observatory (cid:48) 31◦16 (cid:48) 32◦22 (cid:48) 30◦10 Coordinates (cid:48) S 149◦4 (cid:48) 20◦48 (cid:48) 70◦48 (cid:48)(cid:48) 23.88 (cid:48)(cid:48) S 48 (cid:48)(cid:48) 2.64 S (cid:48)(cid:48) 15.6 (cid:48)(cid:48) 36 E 17.28 E (cid:48)(cid:48) W Elevation (m) Time zone 1116 1460 2198 UTC+10 UTC+2 UTC-3 decisions are logged and tracked, allowing us to reconstruct the decision tree that led to any particular observation. Any additional observing requests by members of our team are tagged as such and analysed separately when estimating the planet detection sensitivity of the project as a whole. Thus, knowledge of the conditions under which each observation was performed helps us to better quantify our biases. The observing sites used to conduct this project are listed in Table 1. Apart from the TAC7 allocated observing time, the ROME/REA project is enabled by contributions to the total time budget by the University of St Andrews, Heidelberg University and the Chinese National Academy of Sciences. 3. FACILITIES OVERVIEW The telescope network description that follows is an update on the information presented in Brown et al. (2013). We present here sufficient information to place the following sections in context but refer the interested reader to the original paper for a more detailed description. LCO is an organisation dedicated to time-domain astronomy. To facilitate this, LCO operates a homogeneous network of 2m, 1m and 0.4m telescopes on multiple sites around the world8, covering both hemispheres. Each observing site hosts between one to five telescopes, which are outfitted for imaging and spectroscopy. The instruments and filters used are the same for all telescopes, allowing 'network redundancy', so that observations can be shifted to alternate sites at any time in the case of technical problems or poor weather. The 1m telescopes are currently equipped with custom made 4K×4K 15-micron Fairchild CCDs (Sinistro). The pixel scale is 0.389 arcseconds in 1×1 binning mode, giving a field of view of 26.5×26.5 arcminutes. The standard filter loadout is a complete Johnson-Cousins/Bessell set (UBVRI) and a SDSS/PanSTARRS set (u(cid:48)g(cid:48)r(cid:48)i(cid:48)z(cid:48) sY w). We note that ROME observations are performed only in three bands: SDSS-g(cid:48), SDSS-r(cid:48) and SDSS-i(cid:48), while REA observations are only done in SDSS-i(cid:48). The specified horizon limit of the telescopes is 15 degrees (3.7 airmasses) and their slewing speed is ∼6 degrees per second. The robotic system is responsible for all telescope functions, including slewing, tracking, auto-guiding, but also controls the functions of the instruments and filter wheels. It features built-in recovery mechanisms to address problems in an automated fashion and has a weather station providing it with continuous information about local conditions so it can shut the enclosure should the humidity or cloud cover exceed the limiting parameter values. The telescopes are controlled by a single robotic scheduler program, capable of orchestrating complex and highly responsive9 observing programs using the entire network to provide round-the-clock observations of any astronomical target of interest, weather permitting. Sequences of complex observing requests can be submitted programmatically to the network through an application program interface (API). This is flexible enough to allow combinations of different observing strategies within the same science program. For example, a user can write software to specify regular survey observations with a fixed instrumental setup and cadence on the network but also specify conditions for a rapid response function with dynamically adjustable observing times based on the current brightness of the target that will trigger when specific conditions are met. All observation requests to the network are transferred to a central database, classified by the dynamic scheduler software (Saunders et al. 2014; Lampoudi et al. 2015) and handed out to the telescopes in order of scientific importance (assigned by the time-allocation committee to all science programs), observability and internal relative project priority. 4. OBSERVING STRATEGY 7 TAC: Time Allocation Committee. 8 https://lco.global/ 9 Can respond to and observe any newly alerted target within minutes. 4.1. Project setup 5 Figure 1. [Left]:ROME/REA targets twenty crowded stellar fields rich in microlensing events close to the Galactic centre. The exact position of each field on the sky was optimised using an algorithm that avoided regions with very bright stars or very high extinction (Drew et al. 2014; Nataf et al. 2013), while maximising the microlensing event rate and avoiding field overlap. The colours represent the expected number of microlensing events per year. [Right]: Project system architecture. Both the ROME survey and REA responsive mode automatically submit observing requests to the LCO telescopes through the Telescope Scheduler. New images obtained at the telescopes are promptly identified and processed by our software, and the resulting products are stored in a database. Photometric analysis can be performed either automatically or refined manually. The possibility of having autonomous software agents in control of a science program make the LCO network an indispensable tool for our project. The ROME/REA project combines survey and follow-up observations to maximise the planet detection rate. Microlensing observations take place when the Galactic bulge is observable for extended periods of time from the LCO sites, roughly from 1st April to 15th of October each year. Our observing strategy for the ROME survey involves monitoring an event-rich 3.76 square degree area of the sky close to the Galactic centre every seven hours, seamlessly switching the observations between the three southern LCO sites. Twenty target fields with the highest star counts were selected to maximise the microlensing event rate (see Figure 1), based on the coordinates of OGLE microlensing alerts announced between 2013 and 2015. The field centres are given in Table 2. Regions with very bright stars (Vmag<7) that would cause the detector chip to saturate were avoided, as were regions of very high extinction. A sequence of single exposures in each of three bands (SDSS-g(cid:48), SDSS-r(cid:48) and SDSS-i(cid:48)) is obtained during every field visit. Although LCO offers a range of filters, these bands were selected because they give the optimal balance between throughput/CCD sensitivity and spectral type classification efficiency. The exposure times are fixed to 300 seconds to reach ∼19th magnitude in the SDSS-i(cid:48) band at an estimated S/N∼50. Follow-up observations form part of our reactive REA strategy, which complements our ROME survey and aims to increase the sensitivity to planets below Neptune-mass. At any given time during the observing season, there are about 40-50 concurrent microlensing events. Our algorithm selects those where the probability of detection of planets is highest per time spent observing (Hundertmark et al. 2018; Horne et al. 2009), typically between two to four events. Extra observing requests for each of these events are then automatically submitted to the robotic observing queue of a single telescope at each southern LCO site with a requested cadence of 1 hour. The software will attempt to avoid queuing reactive REA observations to the telescope performing the ROME survey observations, but that is not always possible or practical. In the event that any of these high-interest targets show anomalous features, the observing cadence can be further reduced to 15 minutes. REA observations are performed in a single band (SDSS-i(cid:48)), with the exposure time evaluated in real-time based on the current brightness of the microlensing event (Dominik et al. 2008), and the pointing is matched to the coordinates of the ROME field the target event appears in. We note in passing 6 Table 2. ROME/REA field centers Field Identifier RA Dec ROME-FIELD-01 ROME-FIELD-02 ROME-FIELD-03 ROME-FIELD-04 ROME-FIELD-05 ROME-FIELD-06 ROME-FIELD-07 ROME-FIELD-08 ROME-FIELD-09 ROME-FIELD-10 ROME-FIELD-11 ROME-FIELD-12 ROME-FIELD-13 ROME-FIELD-14 ROME-FIELD-15 ROME-FIELD-16 ROME-FIELD-17 ROME-FIELD-18 ROME-FIELD-19 ROME-FIELD-20 17:51:20.62 17:58:32.82 17:52:00.01 17:52:43.24 17:53:25.04 17:53:25.47 17:54:07.10 17:54:50.34 17:55:31.47 17:56:11.64 17:56:57.32 17:57:34.75 17:58:15.29 17:59:02.12 17:59:08.06 18:00:18.00 18:03:14.40 18:01:09.81 18:01:15.06 18:03:20.82 -30:03:38.94 -27:58:41.76 -28:49:10.41 -29:16:42.65 -30:15:28.21 -29:46:22.74 -28:41:37.35 -29:11:12.21 -29:46:13.68 -28:38:38.64 -29:16:18.01 -30:05:57.25 -28:26:32.04 -29:10:46.57 -29:38:21.86 -28:32:15.21 -28:05:52.20 -27:59:54.97 -29:00:30.33 -28:50:35.37 Note -- Each pointing covers a field of view of 26.5×26.5 arcminutes. that in case the system is unable to perform REA observations, the reasons are usually technical and are independent of knowledge about the events themselves, therefore statistical results would not be biased. This automated procedure reassesses ongoing events every quarter of an hour and ensures that high-cadence comple- mentary reactive observations are obtained for those microlensing events that show the highest sensitivity to planets, but only if these events happen to lie within the survey areas regularly monitored by ROME. Microlensing events outside our survey footprint do not enter the observing queue, although exceptions can be made in case of particularly remarkable events (Nucita et al. 2018; Dong et al. 2019). For any ROME or REA observations to be scheduled, the angular distance between the moon and the target field must be greater than 15 degrees. Furthermore, no ROME observations are scheduled during full moon (2 nights/month). A simple overview of the channels of communication between the different parts of the project is shown in the right panel of Figure 1. Although LCO offers a Target of Opportunity (ToO) rapid response observing option, we do not use it for our project since the scheduler picks up our observing requests within minutes so that a ToO response has been deemed unnecessary. Since REA observations are submitted with higher priority, they are preferentially selected by the scheduler. We note that the ROME/REA system comprehensively logs and time-stamps all steps that lead to software-driven observing decisions. We are thus able to reconstruct the configuration of the system at any given time and understand why particular observing decisions were made. 4.2. Target and Observation Manager The aforementioned features of the project are embedded in a specialised software framework that is generically referred as a "Target and Observation Manager" (TOM) system (Street et al. 2018a). TOM systems function as proxy astronomers, making observing decisions based on available information. They are typically used to harvest alert streams coming from different surveys, assess the relative importance of these alerts for a given science program, 7 Figure 2. [Left]A colour-magnitude diagram for a single ROME field, simulated with the Besan¸con model (Robin et al. 2003). Red polygons mark the brightness changes of an unblended source star as it gets magnified by a passing foreground lens. Black polygons mark the brightness and colour changes of a blended source star during the course of its magnification. The skewness of this line reveals the degree of blended light contributing to the event. [Right] An unblended microlensed source star would not 'move' in this colour-colour diagram since all colours are equally magnified. Source stars with high degrees of blending can be traced through the diagram and reveal the actual source type. Noise has been removed from the microlensing event for illustrative purposes. schedule or recommend observations, trigger reduction pipelines, update databases, as well as serve the information back to users in easily accessible formats, such as web feeds. They offer a powerful way to visualise and interact with data through a web browser or a graphical user interface (GUI). TOM systems have been rapidly gaining popularity in the advent of next-generation wide-field surveys, such as ZTF10 and LSST11, which are expected to produce more than a million alerts of astronomical transients per night. It is humanly impossible to parse such a high volume of alerts on any reasonable timescale and decide the best way to allocate limited observing resources in order to maximise scientific returns. However, a well-designed TOM system can handle all practical aspects. The ROME/REA TOM system operates in two modes. The regular monitoring of fixed ROME survey fields is configured and scheduled independently from the REA mode. ROME observations are scheduled daily with a fixed order, cadence and exposure time. For REA, the TOM automatically harvests alerts from a number of surveys, including OGLE, MOA and ZTF. It uses the ARTEMiS12 microlensing alert broker (Dominik et al. 2007, 2008), designed to identify and track ongoing anomalous microlensing events, passing all relevant information to an automated target selection and prioritisation algorithm (Hundertmark et al. 2018). For the targets selected, the algorithm evaluates the exposure times to be used based on the predicted current brightness of the event and submits groups of observing requests to the LCO telescopes without requiring manual intervention. Our TOM system continuously runs another piece of software in the background which identifies recently acquired images for the ROME/REA project in the LCO image archives and automatically transfers them back for processing. Incoming images automatically trigger a customised photometric pipeline to produce or update light curves in the ROME/REA database. A web front-end is also provided as part of the TOM for team members to track and assess the performance of the algorithms at any given time, to identify potential problems at a glance, and to visualise the light curve of any given target of interest. 5. SIMULATING A SEASON 10 Zwicky Transient Facility (Bellm et al. 2019) 11 Large Synoptic Survey Telescope (LSST Science Collaboration et al. 2009) 12 http://www.artemis-uk.org/ 8 In order to estimate the expected yield of our project, we simulated a full microlensing observing season as it would be observed with the strategy outlined in section 4. To generate the sample of microlensing events, we used the published parameters of microlensing events detected during the 2015 microlensing season by the OGLE-IV survey (Udalski et al. 2015), provided their coordinates matched our ROME survey footprint. The light curves of these events were then sampled based on our observing strategy. The simulation was done using the open-source pyLIMA microlensing modeling software (Bachelet et al. 2017). It included losses due to weather, based on historical weather data at the sites of the LCO telescopes, and observing limits set by the proximity of the target field to the moon. The noise model was calibrated using data from previous microlensing observing seasons at LCO with a similar technical setup, specifically the 2016 season of the RoboNet microlensing project (Tsapras et al. 2009). For each event, the planet detection probability was evaluated assuming each star has one planet between 0.5-10 au (uniform in log(a)). The signal of an artificial planet with a mass derived from the Cassan et al. (2012) planet mass function was injected into the light curve and we evaluated whether it would be detected or missed given our sampling. To consider a planet "detected", we required at least seven consecutive observations during the planetary anomaly (Dominik et al. 2010), with each corresponding point deviating by more than 3σ from the unperturbed single-lens microlensing light curve. The general procedure is described in Hundertmark et al. (2018). In effect, we do this in order to avoid mis-estimating crucial statistical properties due to small samples. For the simulation, all high magnification events were already selected and being "observed" in REA 1-hour mode. ROME "observations", one in each band every seven hours, were also included in this assessment. We did not include a REA 15-minute observing cadence in the simulations because the automatic alert assessment system produced many false positives, which, since these observations are very time-costly, would deplete our allocated time very quickly. Our simulation results can be summarised as follows: Assuming that the technical performance of the telescope network remains stable, we expect our project to yield at least ∼10 new cool planet discoveries in its three-year lifetime, if such planets are as abundant as theory predicts. We estimate that ∼79% of the planets we find will have masses between 30 M⊕ and 10 MJup, and ∼21% between 5 and 30 M⊕. The exact number and type of planets will depend on the true underlying planet population statistics and our detection efficiency, but any findings in this planetary parameter space are of great scientific interest since recent statistical results from microlensing (Suzuki et al. 2016, 2018) identify a break in the mass-ratio function with few planets detected in the low-mass range (Mp ≤ 20M⊕). If confirmed, this result would imply that planet formation processes are not as efficient in producing smaller mass planets as current planet population synthesis models suggest (Ida et al. 2013). Although most of these planets are expected to be identified separately by other surveys, the ability to characterise the source star using observations in three bands is particular to this project. Figure 2 was generated from our simulation using the Besan¸con model of stellar population synthesis of the Galaxy and demonstrates the usefulness of multi-band photometry, as previously described in Section 2. We used the under- lying distributions to draw the lens and source distances for our simulations. The left panel shows a colour-magnitude diagram (CMD) for a single ROME survey field and how our program can distinguish between different degrees of blending. On the right, the advantage of colour-colour information is illustrated: highly blended source stars can be identified and traced through this diagram as the microlensing event evolves, thereby revealing their stellar type. Our simulations also showed that it is possible to obtain useful data in all three bands for about a third of our total star sample. It is difficult in practice to extract accurate measurements for the fainter events close to our detection limits, especially in SDSS-g(cid:48), as well as for very short events. 6. INITIAL RESULTS 6.1. A new photometric pipeline Image subtraction involves the subtraction of all constant-in-time features from time-series of astronomical images of the same field, using them to self-calibrate the photometric scale (Tomaney & Crotts 1996; Alard & Lupton 1998). This is done by constructing a reference image, which can be a single image, or a combination of a number of images, obtained under good observing conditions, and adjusting it to match the observing conditions of every other image taken at different epochs. The adjustment for a single pair of images involves a convolution that registers the images, blurs them to match the atmospheric seeing, and scales them to match the atmospheric transmission (Alard 2000). The adjusted reference image is then subtracted from the target image, removing all constant-in-time features and leaving behind only residual noise and the signals of sources that have varied between the two images. The brightness 9 Figure 3. Image subtraction results. A reference image, shown on the left, obtained under good observing conditions is adjusted to match the orientation and seeing distortions of a target image, shown in the centre, obtained at a different epoch. The two are subtracted and the resulting difference image is presented in the right panel. The image thumbnail is centred on event OGLE-2019-BLG-0171. fluctuations of a variable object, such as a microlensing target, can then be measured on the full set of difference images. Bramich (2008) presented an alternative method to standard image subtraction for determining the convolution kernel that matches pairs of images of the same field. The technique involves defining the kernel as a discrete pixel array which, at the cost of some computational speed, can deal sufficiently well with asymmetric point-spread functions and small image misalignments. Image subtraction software, developed in IDL13 and used for the RoboNet microlensing follow-up project (Tsapras et al. 2009), employed this technique to produce consistently reliable precision photometry optimised for the extraction of single microlensing event light curves. However, the expected computational speed drop for a project of the scale of ROME/REA necessitated the development of new algorithms, more suited to the particular needs of the project. Specifically, we want each incoming ∼ 4k×4k pixel image to be automatically picked up and processed by our photometric pipeline and new photometric points extracted for all stars in a timely fashion. Each updated light-curve can then be used to search for and assess ongoing microlensing events in real-time using our machine-learning event classifier. Our new photometric pipeline (pyDANDIA14) is written in the Python programming language and is optimised for the cameras of the LCO network. We achieved improved computational speed by trading some photometric accuracy at the fainter end of the magnitude distribution. The results of a detailed evaluation of the performance of the pipeline will be published in a separate paper. A typical example of image subtraction is presented in Figure 3. 6.2. Producing a star catalogue A triplet of high-quality template images that have been obtained as part of the same exposure sequence, one in each observing band, are used to generate a star catalogue for each of the twenty ROME survey fields. This is done to ensure that the measured colours are derived from data taken approximately contemporaneously and under as similar observing conditions as possible. Individual sources are identified and measured on each of the three template images using a custom implementation of the DAOFIND algorithm (Stetson 1987), followed by a refinement of the World Coordinate System (WCS) solution using the Gaia DR2 catalogue and photometric calibration by cross-matching with the VPHAS+ Public Survey catalogue15. Sources are then cross-matched between all three images and a final catalogue of stars is produced for the particular ROME field16. The calibrated star positions and magnitudes in these catalogues remain fixed as long as the template images used for the photometric reduction do not change. We expect that the catalogues will need to be revised two or three times 13 Interactive Data Language. 14 https://github.com/pyDANDIA/pyDANDIA 15 http://www.vphas.eu/data.shtml 16 We note that inclusion in the final catalogue requires that a source is detected in at least one of the three template images. 10 (a) (b) Figure 4. (a) Color-Magnitude diagram of ROME-FIELD-16. Measurements for 125593 stars were cross-matched between the SDSS-g(cid:48), SDSS-r(cid:48) and SDSS-i(cid:48) bands. (b) Colour-Colour diagram of ROME-FIELD-16. The different spectral types marked by the grey dots are obtained from the atlas of synthetic spectra of Pickles (1985). (Values not corrected for extinction.) during the course of the project once stacked template images have been constructed, offering improved sensitivity to fainter magnitudes. Figure 4 shows a colour-magnitude diagram from our initial reduction of a single ROME field. The mean number of stars detected per field, averaged over all ROME fields, is ∼150 thousand down to magnitude I ∼19. We note that this value is based on our initial analysis performed using a single template image in each band. By generating stacked template images at the end of the project, we expect the magnitude sensitivity to improve by between 0.5 to 1 magnitudes, and the final catalogue release will contain multi-band photometry for about 3 million stars. When analysing individual microlensing events, our colour analysis only uses stars within 2 arcmin from the event coordinates to generate these diagrams, so as to minimise the effect of differential extinction across the field of view. Correcting for extinction then involves identifying the location of the Red Clump (RC) stars on the colour-magnitude diagram, measuring its offset from known values (Nataf et al. 2013; Ruiz-Dern et al. 2018) and using it to estimate of photometric properties of the source17. Furthermore, since the magnification of an event can be assumed to be approximately the same in each set of SDSS-g(cid:48),-r(cid:48),-i(cid:48) ROME images, taken within 15 minutes of each other, it is possible to measure the colour of the source independent of the model by calculating the colour-colour slope for different sets of pass-bands and for different magnifications. An end-to-end colour analysis of microlensing event OGLE-2018-BLG- 0022 using ROME/REA observations exclusively was recently presented in Street et al. (2019), where this procedure is described in detail. Figure 5(a) presents the photometric accuracy in SDSS-i(cid:48) for 68293 stars in a 3k×3k sub-region of ROME field 5. Such diagrams are part of a set of diagrams automatically constructed for each ROME field and are used to evaluate the quality of the photometry. Our data set contains observations in three bands for thousands of variable stars. For example, Figure 5(b) shows the phase-folded light curve of known RR Lyrae variable star OGLE-BLG-RRLYR-07770 (RA/Dec(J2000): 17:56:01.90, -29:51:45.3, I=15.461, V=17.591). For clarity, only SDSS-i(cid:48) observations from the three southern LCO sites are displayed. 17 Under the assumption that the source is at the same distance and suffers the same extinction as the RC. 11 (a) (b) Figure 5. (a) Accuracy of ROME/REA SDSS-i(cid:48) photometry from an initial analysis of SAAO observations. (b) Sample SDSS-i(cid:48) phase-folded light curve of RR Lyrae variable OGLE-BLG-RRLYR-07770, as observed from LCO telescopes in Chile (CTIO), South Africa (SAAO) and Australia (SAO). The solid black line shows our best RR Lyrae model fit to the data, following the method described in Tsapras et al. (2017). 6.3. Machine-learning event identification Our team has developed an efficient machine-learning classifier that uses the Random Forest algorithm to identify microlensing signals in time-series data18. The classifier is designed for flexibility of use and has been successfully applied on OGLE-II survey data, recovering ∼95% of known microlensing events. In addition, it has been tested on archival data from the Palomar Transient Factory (PTF) optical wide-field survey and alert-stream data from the ZTF survey. The top part of Figure 6 presents the light curve of one of the microlensing events recovered by running the classifier on the initial photometry of a single ROME/REA field. For the purposes of this exercise, light curves containing fewer than 5 photometric measurements were removed from the sample. The bottom panel of the figure demonstrates how the classifier performs in real-time. Four different classes of variability are considered: CONS (constant star), CV (cataclysmic variable), VAR (RR Lyrae or Cepheid) and ML (microlensing). Every incoming data point changes the relative probability of the light curve belonging to each of the classes. The microlensing 'detection' happens at the epoch when the green line crosses above the others. The methodology, features considered and technical details of the classifier can be found in Godines et al. (2019). 7. SUMMARY the brief signals cold exoplanets can produce in the light curves of microlensing events. The ROME/REA project uses the Las Cumbres Observatory network of robotic telescopes to detect and characterise The ROME survey is the first extensive campaign to map an area of ∼4 degrees close to the Galactic centre in multiple bands with a cadence of 7 hours. Observations in three bands are useful for characterising the spectral type of the source stars and identifying contaminating blended light. With this information at hand, it is possible to place stronger constraints on the mass of the lens, thereby improving the precision of the mass estimate of any planetary companions. 18 https://github.com/dgodinez77/LIA 12 Figure 6. [Top]: The light curve of a microlensing event identified in a test run of our Random Forest classifier on a subset of ROME data. [Bottom]: Drip-feeding analysis reveals at which epoch the classifier would have identified this as a microlensing event (t=7925). Four different classes were considered: CONS for constant star, CV for cataclysmic variable, VAR for RR Lyrae or Cepheid and ML for microlensing. The reactive observing REA campaign, with its default 1 hour observing cadence, complements the ROME survey by targeting specific microlensing events where the probability of detecting the signals of planetary companions to the main lens star are highest. This higher cadence offers enhanced sensitivity to planets with smaller masses (less than 10 M⊕) and, when anomalous features are detected, can be further adjusted to sample the light curve every 15 minutes. For the great majority of high-magnification events in the ROME/REA sample, there will be sufficiently dense coverage of the light curve to characterise the physical nature of the event, thereby leading to either entirely new discoveries or independent validations of newly discovered events by other microlensing surveys. In addition, even in the cases where ROME/REA observations are too sparse for characterisation, multi-band data will still be of great use to other surveys in constraining the stellar type of the source and a combined analysis of all available data can lead to a better understanding of the properties of the lens (Bachelet et al. 2012; Bennett et al. 2016; Street et al. 2019; Tsapras et al. 2019). A public release of the ROME/REA photometric catalogue in three bands for about 3 million stars brighter than I ∼ 20mag in our observing fields is planned after the end of the project. The final star catalogue will be a valuable side- product of our survey and will be freely available to the astronomical community. All data products and codes developed during the course of this project are to be released under a public license, including our new image subtraction pipeline, and user-friendly software to model microlensing event light curves (pyLIMA19(Bachelet et al. 2017), muLAn20(Ranc & Cassan 2018)). Studies of variable stars, Galactic extinction, transiting exoplanets and stellar remnants are some potential secondary science by-products of our final data release. The dynamic observing techniques we have developed within the framework of the LCO network rely on parsing incoming alert streams and modifying observing proposals in real-time. They introduce new targets, adjust relative priorities between targets, alter the observing cadence and exposure times, always using the latest available information. Such intelligent robotic observing agents are of great interest to the wider transient astronomical community as new 19 https://github.com/ebachelet/pyLIMA 20 https://github.com/muLAn-project/ 13 methods are urgently sought to handle the high-volume alert streams from upcoming wide-field surveys like the LSST (Bloom et al. 2012; Narayan et al. 2018). ACKNOWLEDGMENTS YT and JW acknowledge the support of DFG priority program SPP 1992 "Exploring the Diversity of Extrasolar Planets" (WA 1047/11-1). RAS and EB gratefully acknowledge support from NASA grant NNX15AC97G. DMB acknowledges the support of the NYU Abu Dhabi Research Enhancement Fund under grant RE124. KH acknowledges support from STFC grant ST/R000824/1. This work was partly supported by the National Science Foundation of China (Grant No. 11333003, 11390372 and 11761131004 to SM). This work makes use of observations from the LCOGT network. This research uses data obtained through the Telescope Access Program (TAP), which has been funded by the National Astronomical Observatories of China, the Chinese Academy of Sciences, and the Special Fund for Astronomy from the Ministry of Finance. This work was partly supported by the National Science Foundation of China (Grant No. 11390372 and 11761131004 to SM). REFERENCES Alard, C. 2000, A&AS, 144, 363. Boyajian, T. S., van Belle, G., & von Braun, K. 2014, AJ, https://ui.adsabs.harvard.edu/abs/2000A&AS..144..363A 147, 47. Alard, C., & Lupton, R. H. 1998, ApJ, 503, 325. https://ui.adsabs.harvard.edu/abs/2014AJ....147...47B https://ui.adsabs.harvard.edu/abs/1998ApJ...503..325A Bramich, D. M. 2008, MNRAS, 386, L77. https: Armitage, P. J. 2011, ARA&A, 49, 195. //ui.adsabs.harvard.edu/abs/2008MNRAS.386L..77B http://adsabs.harvard.edu/abs/2011ARA%26A..49..195A Brown, T. M., Baliber, N., Bianco, F. B., et al. 2013, Bachelet, E., Norbury, M., Bozza, V., & Street, R. 2017, PASP, 125, 1031. AJ, 154, 203. https://ui.adsabs.harvard.edu/abs/2017AJ....154..203B Bachelet, E., Shin, I. G., Han, C., et al. 2012, ApJ, 754, 73. https://ui.adsabs.harvard.edu/abs/2012ApJ...754...73B Bachelet, E., Bramich, D. M., Han, C., et al. 2015, ApJ, 812, 136. https://ui.adsabs.harvard.edu/abs/2015ApJ...812..136B Batalha, N. M., Rowe, J. F., Bryson, S. T., et al. 2013, ApJS, 204, 24. http://adsabs.harvard.edu/abs/2013ApJS..204...24B Bellm, E. C., Kulkarni, S. R., Graham, M. J., et al. 2019, PASP, 131, 018002. https: //ui.adsabs.harvard.edu/abs/2019PASP..131a8002B Bennett, D. P., Rhie, S. H., Udalski, A., et al. 2016, AJ, 152, 125. https://ui.adsabs.harvard.edu/abs/2016AJ....152..125B Bloom, J. S., Richards, J. W., Nugent, P. E., et al. 2012, Publications of the Astronomical Society of the Pacific, 124, 1175. https://ui.adsabs.harvard.edu/abs/2012PASP..124.1175B Bond, I. A., Abe, F., Dodd, R. J., et al. 2001, MNRAS, 327, 868. https: //ui.adsabs.harvard.edu/abs/2001MNRAS.327..868B http://adsabs.harvard.edu/abs/2013PASP..125.1031B Cassan, A., Kubas, D., Beaulieu, J.-P., et al. 2012, Nature, 481, 167. http://adsabs.harvard.edu/abs/2012Natur.481..167C Dominik, M., Horne, K., & Bode, M. F. 2006, Astronomy and Geophysics, 47, 3.25. https://ui.adsabs.harvard.edu/abs/2006A&G....47c..25D Dominik, M., Rattenbury, N. J., Allan, A., et al. 2007, MNRAS, 380, 792. http://esoads.eso.org/abs/2007MNRAS.380..792D Dominik, M., Horne, K., Allan, A., et al. 2008, Astronomische Nachrichten, 329, 248. https://ui.adsabs.harvard.edu/abs/2008AN....329..248D Dominik, M., Jørgensen, U. G., Rattenbury, N. J., et al. 2010, Astronomische Nachrichten, 331, 671. https://ui.adsabs.harvard.edu/abs/2010AN....331..671D Dong, S., M´erand, A., Delplancke-Strobele, F., et al. 2019, ApJ, 871, 70. https://ui.adsabs.harvard.edu/abs/2019ApJ...871...70D Drew, J. E., Gonzalez-Solares, E., Greimel, R., et al. 2014, MNRAS, 440, 2036. https: //ui.adsabs.harvard.edu/abs/2014MNRAS.440.2036D Borucki, W. J., Koch, D., Basri, G., et al. 2010, Science, Dullemond, C. P. 2013, Astronomische Nachrichten, 334, 327, 977. http://adsabs.harvard.edu/abs/2010Sci...327..977B 589. http://adsabs.harvard.edu/abs/2013AN....334..589D 14 Foreman-Mackey, D., Hogg, D. W., & Morton, T. D. 2014, Nataf, D. M., Gould, A., Fouqu´e, P., et al. 2013, ApJ, 769, ApJ, 795, 64. https://ui.adsabs.harvard.edu/abs/2014ApJ...795...64F 88. https://ui.adsabs.harvard.edu/abs/2013ApJ...769...88N Gaudi, B. S. 2012, ARA&A, 50, 411. http://adsabs.harvard.edu/abs/2012ARA%26A..50..411G Godines, D., Bachelet, E., Narayan, G., & Street, R. A. Nucita, A. A., Licchelli, D., De Paolis, F., et al. 2018, MNRAS, 476, 2962. https: //ui.adsabs.harvard.edu/abs/2018MNRAS.476.2962N 2019, Astronomy and Computing, 28, 100298. https://doi.org/10.1016/j.ascom.2019.100298 Gould, A. 1992, ApJ, 392, 442. https://ui.adsabs.harvard.edu/abs/1992ApJ...392..442G Gould, A., & Loeb, A. 1992, ApJ, 396, 104. https://ui.adsabs.harvard.edu/abs/1992ApJ...396..104G Gould, A., Dong, S., Gaudi, B. S., et al. 2010, ApJ, 720, 1073. http://adsabs.harvard.edu/abs/2010ApJ...720.1073G Han, C., Bond, I. A., Udalski, A., et al. 2019, ApJ, 876, 81. https://ui.adsabs.harvard.edu/abs/2019ApJ...876...81H Horne, K., Snodgrass, C., & Tsapras, Y. 2009, MNRAS, 396, 2087. https: //ui.adsabs.harvard.edu/abs/2009MNRAS.396.2087H Howard, A. W. 2013, Science, 340, 572. http://adsabs.harvard.edu/abs/2013Sci...340..572H Hundertmark, M., Street, R. A., Tsapras, Y., et al. 2018, A&A, 609, A55. https://ui.adsabs.harvard.edu/abs/2018A&A...609A..55H Ida, S., Lin, D. N. C., & Nagasawa, M. 2013, ApJ, 775, 42. http://adsabs.harvard.edu/abs/2013ApJ...775...42I Kim, S.-L., Lee, C.-U., Park, B.-G., et al. 2016, Journal of Korean Astronomical Society, 49, 37. https://ui.adsabs.harvard.edu/abs/2016JKAS...49...37K Lampoudi, S., Saunders, E., & Eastman, J. 2015, arXiv e-prints, arXiv:1503.07170. https://ui.adsabs.harvard.edu/abs/2015arXiv150307170L LSST Science Collaboration, Abell, P. A., Allison, J., et al. 2009, arXiv e-prints, arXiv:0912.0201. https://ui.adsabs.harvard.edu/abs/2009arXiv0912.0201L Mao, S., & Paczynski, B. 1991, ApJ, 374, L37. https://ui.adsabs.harvard.edu/abs/1991ApJ...374L..37M Min, M., Dullemond, C. P., Kama, M., & Dominik, C. 2011, Icarus, 212, 416. http://adsabs.harvard.edu/abs/2011Icar..212..416M Mordasini, C., Alibert, Y., & Benz, W. 2009, A&A, 501, 1139. http://adsabs.harvard.edu/abs/2009A%26A...501.1139M Mordasini, C., Alibert, Y., Klahr, H., & Henning, T. 2012, A&A, 547, A111. http://adsabs.harvard.edu/abs/2012A%26A...547A.111M Pickles, A. J. 1985, ApJS, 59, 33. https://ui.adsabs.harvard.edu/abs/1985ApJS...59...33P Ranc, C., & Cassan, A. 2018, muLAn: gravitational MICROlensing Analysis Software, Astrophysics Source Code Library, , , ascl:1811.012. http://adsabs.harvard.edu/abs/2018ascl.soft11012R Robin, A. C., Reyl´e, C., Derri`ere, S., & Picaud, S. 2003, A&A, 409, 523. https://ui.adsabs.harvard.edu/abs/2003A&A...409..523R Ruiz-Dern, L., Babusiaux, C., Arenou, F., Turon, C., & Lallement, R. 2018, A&A, 609, A116. https: //ui.adsabs.harvard.edu/abs/2018A&A...609A.116R Saunders, E. S., Lampoudi, S., Lister, T. A., Norbury, M., & Walker, Z. 2014, in Proc. SPIE, Vol. 9149, Observatory Operations: Strategies, Processes, and Systems V, 91490E. http://adsabs.harvard.edu/abs/2014SPIE.9149E..0ES Stetson, P. B. 1987, PASP, 99, 191. http://adsabs.harvard.edu/abs/1987PASP...99..191S Street, R. A., Bowman, M., Saunders, E. S., & Boroson, T. 2018a, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 10707, Software and Cyberinfrastructure for Astronomy V, 1070711. http://adsabs.harvard.edu/abs/2018SPIE10707E..11S Street, R. A., Lund, M. B., Khakpash, S., et al. 2018b, arXiv e-prints, arXiv:1812.03137. http://adsabs.harvard.edu/abs/2018arXiv181203137S Street, R. A., Bachelet, E., Tsapras, Y., et al. 2019, AJ, 157, 215. https://ui.adsabs.harvard.edu/abs/2019AJ....157..215S Sumi, T., Bennett, D. P., Bond, I. A., et al. 2013, ApJ, 778, 150. https://ui.adsabs.harvard.edu/abs/2013ApJ...778..150S Suzuki, D., Bennett, D. P., Sumi, T., et al. 2016, ApJ, 833, 145. https://ui.adsabs.harvard.edu/abs/2016ApJ...833..145S Suzuki, D., Bennett, D. P., Ida, S., et al. 2018, ApJ, 869, L34. https://ui.adsabs.harvard.edu/abs/2018ApJ...869L..34S Tomaney, A. B., & Crotts, A. P. S. 1996, AJ, 112, 2872. Narayan, G., Zaidi, T., Soraisam, M. D., et al. 2018, The https://ui.adsabs.harvard.edu/abs/1996AJ....112.2872T Astrophysical Journal Supplement Series, 236, 9. https://ui.adsabs.harvard.edu/abs/2018ApJS..236....9N Tsapras, Y. 2018, Geosciences, 8, 365. http://adsabs.harvard.edu/abs/2018Geosc...8..365T 15 Tsapras, Y., Street, R., Horne, K., et al. 2009, Tsapras, Y., Cassan, A., Ranc, C., et al. 2019, MNRAS, Astronomische Nachrichten, 330, 4. https://ui.adsabs.harvard.edu/abs/2009AN....330....4T Tsapras, Y., Hundertmark, M., Wyrzykowski, (cid:32)L., et al. 2016, MNRAS, 457, 1320. http://adsabs.harvard.edu/abs/2016MNRAS.457.1320T Tsapras, Y., Arellano Ferro, A., Bramich, D. M., et al. 2017, MNRAS, 465, 2489. https: //ui.adsabs.harvard.edu/abs/2017MNRAS.465.2489T 1368. https: //ui.adsabs.harvard.edu/abs/2019MNRAS.tmp.1368T Udalski, A., Szyma´nski, M. K., & Szyma´nski, G. 2015, AcA, 65, 1. https://ui.adsabs.harvard.edu/abs/2015AcA....65....1U Udry, S., Mayor, M., & Santos, N. C. 2003, A&A, 407, 369. https://ui.adsabs.harvard.edu/abs/2003A&A...407..369U Witt, H. J., & Mao, S. 1994, ApJ, 430, 505. http://adsabs.harvard.edu/abs/1994ApJ...430..505W Yoo, J., DePoy, D. L., Gal-Yam, A., et al. 2004, ApJ, 603, 139. http://adsabs.harvard.edu/abs/2004ApJ...603..139Y
1806.07354	1	1806	2018-06-19T17:31:16	From Planetesimal to Planet in Turbulent Disks. II. Formation of Gas Giant Planets	[ "astro-ph.EP", "astro-ph.SR" ]	In the core accretion scenario, gas giant planets are formed form solid cores with several Earth masses via gas accretion. We investigate the formation of such cores via collisional growth from kilometer-sized planetesimals in turbulent disks. The stirring by forming cores induces collisional fragmentation and surrounding planetesimals are ground down until radial drift. The core growth is therefore stalled by the depletion of surrounding planetesiamls due to collisional fragmentation and radial drift. The collisional strength of planetesimals determines the planetesimal-depletion timescale, which is prolonged for large planetesiamls. The size of planetesiamls around growing cores is determined by the planetesimal size distribution at the onset of runaway growth. Strong turbulence delays the onset of runaway growth, resulting in large planetesimals. Therefore, the core mass evolution depends on turbulent parameter $\alpha$; the formation of cores massive enough without significant depletion of surrounding planetesimals needs strong turbulence of $\alpha \gtrsim 10^{-3}$. However, the strong turbulence with $\alpha \gtrsim 10^{-3}$ leads to a significant delay of the onset of runaway growth and prevents the formation of massive cores within the disk lifetime. The formation of cores massive enough within several millions years therefore requires the several times enhancement of the solid surface densities, which is achieved in the inner disk $\lesssim 10$AU due to pile-up of drifting dust aggregates. In addition, the collisional strength $Q_{\rm D}^$ even for kilometer-sized or smaller bodies affects the growth of cores; $Q_{\rm D}^ \gtrsim 10^7 \,{\rm erg/g}$ for bodies $\lesssim 1$ km is likely for this gas giant formation.	astro-ph.EP	astro-ph	Draft version June 20, 2018 Preprint typeset using LATEX style emulateapj v. 01/23/15 FROM PLANETESIMAL TO PLANET IN TURBULENT DISKS. II. FORMATION OF GAS GIANT PLANETS Department of Physics, Nagoya University, Nagoya, Aichi 464-8602, Japan Hiroshi Kobayashi Astronomical Institute, Tohoku University, Aramaki, Aoba-ku, Sendai 980-8578, Japan Hidekazu Tanaka Draft version June 20, 2018 ABSTRACT In the core accretion scenario, gas giant planets are formed form solid cores with several Earth masses via gas accretion. We investigate the formation of such cores via collisional growth from kilometer- sized planetesimals in turbulent disks. The stirring by forming cores induces collisional fragmentation and surrounding planetesimals are ground down until radial drift. The core growth is therefore stalled by the depletion of surrounding planetesiamls due to collisional fragmentation and radial drift. The collisional strength of planetesimals determines the planetesimal-depletion timescale, which is prolonged for large planetesiamls. The size of planetesiamls around growing cores is determined by the planetesimal size distribution at the onset of runaway growth. Strong turbulence delays the onset of runaway growth, resulting in large planetesimals. Therefore, the core mass evolution depends on turbulent parameter α; the formation of cores massive enough without significant depletion of surrounding planetesimals needs strong turbulence of α & 10−3. However, the strong turbulence with α & 10−3 leads to a significant delay of the onset of runaway growth and prevents the formation of massive cores within the disk lifetime. The formation of cores massive enough within several millions years therefore requires the several times enhancement of the solid surface densities, which is achieved In addition, the collisional in the inner disk . 10AU due to pile-up of drifting dust aggregates. D & 107 erg/g for strength Q∗ bodies . 1 km is likely for this gas giant formation. Subject headings: planets and satellites: formation - planets and satellites: gaseous planets D even for kilometer-sized or smaller bodies affects the growth of cores; Q∗ 1. INTRODUCTION Gas giant planets such as Jupiter, Saturn, and mas- sive exoplanets are formed in protoplanetary disks con- taining solid and gas. In the core accretion scenario, once planetary embryos grow to ∼ 5M⊕, planetary at- mospheres that Mars sized or larger embryos acquire become too massive to keep hydrostatic states, result- ing in the rapid gas accretion to form gas giants (e.g., Mizuno 1980). Therefore, gas giant formation requires such massive solid cores to be formed within the disk lifetime (∼ 4 Myr). In protoplanetary disks, sub-micron sized dust grains accumulate to be planetesimals. Collisional growth of grains produces centimeter sized or larger particles, which drift onto the host star on significantly short timescales due to moderate coupling with gas. If par- ticles accumulate up to the Roche density, gravita- tional instability creates kilometer sized or larger plan- etesimals (Goldreich & Ward 1973; Youdin & Goodman 2005; Michikoshi et al. 2012; Takeuchi & Ida 2012). On the other hand, collisional growth of dust grains naturally form fluffy aggregates (Suyama et al. 2008, 2012). The bulk density of dust aggregates becomes ∼ 10−4 g/cm3 and the collisional growth timescale of such bodies is much shorter than their drift timescale, so that kilome- ter sized planetesimals are formed via direct collisional growth prior to radial drift (Okuzumi et al. 2012). [email protected] [email protected] Planetesimals grow through mutual collisions. In a turbulent disk, the stirring by turbulence increases the random velocity of planetesimals vr. For large vr, the gravitational focusing of planetesimals is negligi- ble and the orderly growth occurs until vr & 1.5vesc (Kobayashi et al. 2016), where vesc are the surface escape velocities of planetesimals. In the orderly growth, the mass-weighted average radius of planetesimals is compa- rable to the radius of largest planetesimals. As planetes- imals grow, vesc increases to ∼ vr, runaway growth oc- curs (Wetherill and Stewart 1989). The mass-weighted average radius of planetesimals at the onset of run- away growth, which is determined by turbulent strength (Kobayashi et al. 2016), almost remains the same after runaway growth. Runaway growth forms a planetary embryo in each an- nulus of the disk. Embryos further grow mainly through collisions with surrounding planetesimals. As embryos become massive, their viscous stirring increases vr, re- sulting in destructive collisions between planetesimals. Collisional fragments of planetesimals get further smaller via collisions between themselves. This collisional cas- cade decreases the size of bodies until radial drift. This process reduces surrounding planetesimals and embryo growth is then stalled (Kobayashi et al. 2010). Collisional strength depends on the radius of plan- etesimals, r (e.g., Benz & Asphaug 1999). For r & 1 km, the collisional outcome of a single collision is con- trolled by the self-gravity of colliders so that larger bodies are effectively stronger for collisions. Larger 2 planetesimals, which have a longer collisional deple- tion timescales, contribute more to the growth of massive embryos. Embryo growth strongly depends on the mass-weighted average radius of planetesi- mals surrounding embryos (Kobayashi et al. 2010, 2011; Kobayashi & Dauphas 2013), which is mainly deter- mined by the strength of turbulence (Kobayashi et al. 2016). Therefore, the growth and formation of solid cores of gas giants depends on the strength of disk turbulence. In this paper, we investigate gas giant planet formation via core accretion in turbulent disks. In § 2, we introduce the critical core mass from simple analysis. In § 3, we explain the model of simulations. In § 4, we perform simulations of collisional evolution of bodies from r = 1 km. We show the dependence of embryo growth on turbulent strength and collisional property and find the conditions for the formation of cores massive enough for the onset of rapid gas accretion to form gas giants. In § 5, we discuss the growth timescale of embryos in turbulent disks, and the type I migration of growing embryos and the gas dispersal timescale of disks. In § 6, we summarize our findings. 2. CRITICAL CORE MASS The runaway growth of planetesimals produces a plan- etary embryo with mass ME and radius RE in each annu- lus of a protoplanetary disk. Once RE is larger than its Bondi radius, RB defined by RB = GME/c2 s , the embryo has an atmosphere, where G is the gravitational constant and cs is the sound velocity of gas. The density of a hy- drostatic atmosphere, ρa, at the distance R from the cen- ter of the planetary embryo is given by (derivation in Ap- pendix A; Mizuno 1980; Stevenson 1982; Inaba & Ikoma 2003) ρa = πσSB 12κLeR3 (cid:18) GMEµmH k (cid:19)4 , (1) where σSB is the Stefan-Boltzmann constant, κ is the opacity of the atmosphere, k is the Boltzmann constant, µ is the mean molecular weight, mH is the mass of a hy- drogen atom, and Le is the planetary luminosity. The accretion of bodies onto planetary embryos mainly de- termines Le so that Le = GME RE dME dt . The total atmospheric mass, MA, is given by (2) (3) (4) MA =Z RB 4πR2ρadR, RE π2σSBG3M 3 ERE = 3κ ME k (cid:17)4 (cid:16) µmH ln(cid:18) RB RE(cid:19) . Once MA is comparable to ME, the hydrostatic atmo- sphere is not maintained and then rapid gas accretion occurs to form gas giants. The embryo mass at the on- set of rapid gas accretion is called the critical core mass, which is approximately estimated from MA/ME ≈ 1/3 (Mizuno 1980; Stevenson 1982; Ikoma et al. 2000). Us- ing Eq. (4) with MA = ME/3, the critical core mass Mcrit is given by Mcrit = 5M⊕ ME/ ME 10 Myr !−3/4 (cid:18) κ 0.01 cm2 g−1(cid:19)3/4 , (5) where the dependence of ln(RB/RE) on ME is ignored for this derivation. Therefore, forming embryos with ME & 5M⊕ are re- quired for giant planet formation via core accretion. Note that the critical core mass may be smaller than 5M⊕ because the depletion of planetesimals due to ME (e.g., collisional fragmentation results in smaller Kobayashi et al. 2011). We below perform simulations for the formation and growth of embryos and find conditions for the formation of embryos with the critical core mass. Although we here simply estimate MA using the radiative tempera- ture gradient with constant κ, we calculate MA from ME and ME obtained in simulations, taking into account the convective temperature gradient as well as the radiative one with κ dependent on temperature (Inaba & Ikoma 2003). 3. MODEL FOR PLANETARY ACCRETION The surface number density of planetesimals with masses m to m + dm orbiting around the host star with mass M∗ at the distance a, ns(m, a)dm, evolves through collisions and radial drift. The governing equation is given as (e.g., Kobayashi et al. 2016), ∂mns(m, a) ∂t = m 0 0 dm2ns(m1, a)ns(m2, a) 2 Z ∞ dm1Z ∞ ×K(m1, m2)δ(m − m1 − m2 + me) −mns(m)Z ∞ dm2ns(m2, a)K(m, m2) ∂mZ ∞ ×K(m1, m2)Ψ(m, m1, m2) 1 a dm1Z m1 [amns(m, a)vdrift(m, a)], ∂ ∂a − ∂ + 0 0 0 dm2ns(m1, a)ns(m2, a) (6) where the collisional kernel K(m1, m2) between bodies with masses m1 and m2 is given by K(m1, m2) = (hm1,m2a)2hPcol(m1, m2)iΩ, (7) with the reduced mutual Hill radius hm1,m2 = [(m1 + m2)/3M∗]1/3, the dimensionless mean collisional rate hPcol(m1, m2)i, and the Kepler orbital frequency Ω, me and Ψ(m, m1, m2) are, respectively, the total and cu- mulative masses of bodies produced by a single colli- sion between bodies with masses m1 and m2, and vdrift is the drift velocity of a body due to gas drag. The collisional rate is determined according to the orbits of colliding bodies and the amount of their atmospheres: hPcol(m1, m2)i is given by a function of relative eccentric- ities and inclinations between bodies and the accretion rates of the bodies, which are summarized in Inaba et al. (2001). The enhancement of collisional cross-section due to embryo's atmosphere (Inaba & Ikoma 2003) and due to gas drag for small bodies (Ormel & Klahr 2010) is taken into account. The atmospheric opacity is given by the sum of those of gas and grains, given by Formation of Gas Giant Planets 3 (Inaba & Ikoma 2003) κ =  0.01 + 4f cm2 g−1 for T ≤ 170 K, 0.01 + 2f cm2 g−1 for 170 K < T ≤ 1700 K, 0.01 cm2 g−1 for T > 1700 K, (8) where f is the grain depletion factor comparing to the interstellar one. The effective growth of dust aggregates deplete the opacity significantly (Okuzumi et al. 2012), so that we put f = 10−4. The collisional kernel depends on orbital eccentricities e and inclinations i of colliders, which depend on mass m and distance a. The time variation of e(m, a) and i(m, a) via gravitational interactions between bodies depends on the mass distribution of bodies. Therefore, the evolution of e(m, a) and i(m, a) have to be treated simultaneously with collisional evolution. The square of dispersions for eccentricities and inclinations of bodies evolve according to de2(m, a) dt di2(m, a) dt = = + + de2 di2 dt (cid:12)(cid:12)(cid:12)(cid:12)g dt (cid:12)(cid:12)(cid:12)(cid:12)g + + di2 de2 dt (cid:12)(cid:12)(cid:12)(cid:12)d dt (cid:12)(cid:12)(cid:12)(cid:12)d + + di2 de2 dt (cid:12)(cid:12)(cid:12)(cid:12)c dt (cid:12)(cid:12)(cid:12)(cid:12)c di2 de2 dt (cid:12)(cid:12)(cid:12)(cid:12)t dt (cid:12)(cid:12)(cid:12)(cid:12)t , , (9) (10) where the subscripts "g", "d", "c", and "t" indicate gravitational interaction between bodies (Ohtsuki et al. 2002), damping by gas drag (Adachi et al. 1976), colli- sional damping (Ohtsuki 1992), and turbulent stirring due to density fluctuation (Okuzumi & Ormel 2013) and aerodynamical friction (Volk et al. 1980), respectively. We use the damping rates by gas drag linear functions of e and i given by Inaba et al. (2001) because e and i ≪ 1, although the e and i damping rates by gas drag are greater for higher e and i (Kobayashi 2015). We use the damping rate and radial drift due to gas drag, tak- ing into account three gas drag regimes (see details in Kobayashi et al. 2010). For 10 meter-sized or larger bodies, the stirring by tur- bulence is mainly caused by turbulent density fluctua- tion, given by (Okuzumi & Ormel 2013) de2 di2 M∗ (cid:19)2 dt (cid:12)(cid:12)(cid:12)(cid:12)t ≈ fd(cid:18) Σga2 dt (cid:12)(cid:12)(cid:12)(cid:12)t ≈ ǫ2 de2 dt (cid:12)(cid:12)(cid:12)(cid:12)t , Ω, (11) (12) where Σg is the gas surface density, fd is the dimension- less factor, and ǫ = 0.01 (see Kobayashi et al. 2016). According to the magnetohydrodynamical simulations, fd is given by (Okuzumi & Ormel 2013), fd = 0.94α (1 + 4.5Hres,0/H)2 , (13) where H is the scale height of the disk, Hres,0 is the half vertical width of the dead zone, α is the dimensionless turbulent viscosity at the midplane. For simplicity, we set Hres,0 = H. Dynamical stirring by large bodies increases collisional velocities of bodies. The collisional outcome is deter- mined by me and Ψ(m, m1, m2) in Eq. (6), which are mainly controlled by the fragmentation energy Q∗ D (see 9 10 8 10 primordial,100m/s 50m/s m o n olith, 1 0 0 m/s 50m/s ] g / g r e [ * D Q 107 106 0.001 0.01 0.1 Radius 1 [km] 10 100 Fig. 1.- The fragmentation energy Q∗ D for primordial and mono- lith bodies with impact velocities of 50 m/s and 100 m/s, given in Eqs. (14)–(19). detailed setting Appendix B). We use the following for- mula for Q∗ D; Q∗ D = Q0s r 1 cm!βs + Q0gρs r 1 cm!βg + Cgg 2Gm r , (14) where ρs is the density of a body, Q0s, Q0g, βs, βg, and Cgg characterize the collisional strength. On the right hand side of Eq. (14), the first term is dominant for kilo- meter sized or smaller bodies. The second term is im- portant for larger bodies. The third term is determined by pure gravity, which is dominant for r & 107 cm: we set Cgg = 10 (Stewart & Leinhardt 2009). The values of Q0s, Q0g, βs, and βg depend on the impact velocity (e.g., Benz & Asphaug 1999) and the structure of a body (e.g., Wada et al. 2013). For monolith bodies, these values are estimated from the interpolation of simulation data for the impact velocities vimp of 3 and 5 km/s obtained by Benz & Asphaug (1999), given by erg/g, (15) 3 km/s(cid:19)γsv Q0s = 1.6 × 107(cid:18) vimp 3 km/s(cid:19)γbsv βs =−0.39(cid:18) vimp 3 km/s(cid:19)γgv Q0g = 1.2(cid:18) vimp 3 km/s(cid:19)γbgv βg = 1.26(cid:18) vimp , , erg cm3/g2, (16) (17) (18) where γsv = −0.82, γgv = −0.31, γbsv = −0.080, and γbgv = 0.032. However, growing bodies via collisions are porous (e.g., Wada et al. 2013), which might not have monolith-like structures until melting. Therefore, bod- ies smaller than about 10 km have Q∗ D of porous bodies, which is given by (Wada et al. 2013), Qs0 = 4 × 107(cid:18) vimp 100 m/s(cid:19) erg/g, βs = 0. (19) (20) We investigate the growth of melted bodies (monolith) and primordial bodies. For melted bodies, Q∗ D is given by Eqs. (15)–(18). For primordial bodies, Q∗ D is calculated from Eqs. (17)–(20). For melted and primordial bodies, Q∗ D are shown in Fig. 1. 4 4. MASS EVOLUTION OF BODIES We perform simulations for planet formation in disks from 4.8 to 26 AU via the time integration of Eqs. (6), (9), and (10) for M∗ = M⊙, where M⊙ is the solar mass. The disk is divided into 10 annuli and the mass distribution is described using mass bins with radios between adjacent mass bins 1.05. We fix the bulk density of bodies at ρs = 1 g/cm3; the mass-radius relation is given by m = 4πρsr3/3. The mass corresponding to the smallest mass bin is set to 4.2×106 g (r = 1 m). We set the initial bodies at radius of 1 km. The bodies initially have e = i/√ǫ = 3vesc/aΩ ≈ 9.9 × 10−5. The collisional growth of bodies is almost independent of the initial radius and orbits of bodies if the new born planetesimals are smaller than those at the onset of runaway growth (Kobayashi et al. 2016). (a) Δ Δ Δ Δ 2 2 2 Σ Σ Σ Σ [g/cm] (r,a) (r,a) (r,a) (r,a) [g/cm [g/cm [g/cm ] ] ] s s s s 10 10-3 10- 10 100 101 10-3 10 10 100 101 10 10 "# 10 10 ! -1 -1 -3 -3 -2 -2 0 0 1 1 (b) 8 6 4 ( $%& 5 10 15 5 Distance, a 6 ' [AU] 10 )* 20 +, Fig. 2.- Size distribution for mass surface density ∆Σs(r, a) of bodies at 2.2 × 105 yr (a), 7.2 × 105 yr (b), 1.4 × 106 yr (c), and 3.3 × 106 yr (d) in a disk with xs = xg = 3 for α = 3 × 10−3 with collisional fragmentation for primordial collisional strength, as a function of the distance from the host star, a, and the radius of bodies corresponding to their mass m. 108 10 104 ] m c [ 102 r , s u i d a R (a) (c) e(r,a) e(r,a) e(r,a) e(r,a) (b) -4 -4 -4 -4 10 10 DE WX -3 -3 -3 -3 10 10 FG YZ -2 -2 -2 -2 10 10 HI [\ -1 -1 -1 -1 10 10 JK ]^ LMN 8 10 6 10 4 10 ] m c [ 10 r , s u i d a R 2 8 6 4 2 5 6 . /3 89 BC @A >? <= 20 O :; Distance, a 6 P [AU] QR ST 20 UV Fig. 3.- Orbital eccentricity distribution obtained from the same simulation as Fig. 2. The initial surface densities of solid and gas, Σg,0 and Σs,0, are, respectively, set to have power-law radial dis- tributions; Σg,0 = 1700xg(cid:16) a Σs,0 = 30xs(cid:16) a 1 AU(cid:17)−1.5 1 AU(cid:17)−1.5 g cm−2, g cm−2, (21) (22) where xg and xs are the scaling factors and the disk with xg = xs = 1 corresponds to the minimum-mass solar Fig. 4.- Orbital inclination distribution obtained from the same simulation as Fig. 2. nebula (MMSN) model (Hayashi 1981). The solid sur- face density Σs evolves due to the radial drift of bodies, while the gas surface density artificially decreases from the formula Σg = Σg,0 exp(−t/τgas) with τgas = 107 yr. We discuss τgas and the time dependence of depletion in § 5. In addition, we set the temperature at the disk midplane as (23) T = 280(cid:16) a 1 AU(cid:17) K, which affects RB and the aerodynamical turbulent stir- ring. However, they are unimportant for the evolution of ME and MA. We show the size and velocity evolution of bodies in the whole disk in § 4.1 and focus on the size and veloc- ity distributions around 5 AU in § 4.2. The evolution of planetary embryos dependent on α is shown in § 4.3. We discuss the formation of cores with the critical core masses from ME and MA obtained within disk lifetimes by simulations in § 4.4. 4.1. Size and Velocity Evolution of Bodies in the Whole Disk We perform a simulation for planet formation with pri- mordial strength in the disk with xs = xg = 3 (3MMSN) for α = 3×10−3, which results in the size and orbit evolu- tion of bodies as shown in Figs. 2-4. The surface density of solid bodies is given by Σs(a) = R mns(m, a)dm = R m2ns(m, a)d ln m, so that we define the surface density of bodies in the logarithmic mass interval as ∆Σs(r, a) ≡ m2ns(m, a), (24) with r corresponding to m. Fig. 2 shows ∆Σs(r, a). The collisional growth of bodies occurs inside-out. Or- derly growth initially occurs so that ∆Σs have a maxi- mum around radii of largest bodies in each annulus for t . 1 Myr (Fig. 2a,b). Larger bodies tend to have greater e and i for r & 10 m (Fig.3a,b and 4a,b). Density- fluctuation turbulent stirring and gas damping mainly control e and i for r & 10 m, while the fluid-dynamical turbulent stirring is dominant instead of that by den- sity fluctuation for r . 10 m. The runaway growth of bodies produces planetary embryos inside 8 AU by 4 Myr (Fig. 2c,d). The stirring by embryos increases e and i (Figs. 3c,d and 4c,d), which induces collisional fragmen- tation due to high speed collisions. Collisional fragments become smaller due to further collisions between them- selves until radial drift, which reduces the surface density of bodies with r . 10 – 100 km (compare Figs. 2c and Formation of Gas Giant Planets 5 2d). Planetary embryo formation in the outer disk in- duces collisional fragmentation and radial drift, which supplies small bodies in the inner disk. However, the net flux of small bodies reduces the surface density of bodies around planetary embryos in the inner disk. This stalls the growth of embryos (e.g., Kobayashi et al. 2010, 2011). In order to see detailed mass and orbit evolu- tion, we focus on the evolution at 5.2 AU in the following paragraphs. ] 2 m c / g [ Σ Δ 1 10 0 10 -1 10 s -2 10 -1 10 -2 10 i , e -3 10 10 -4 10 2 3 10 4 10 5 10 Radius 106 [cm] 7 10 8 10 9 10 Fig. 5.- Size distribution of ∆Σs of bodies (top) and their eccen- tricities e (bottom; solid curves) and inclinations i (bottom; dotted curves) at 5.2 AU in a disk with xs = xg = 3 for α = 3 × 10−3 without collisional fragmentation. The curves indicate those at 1.7 × 105 yr (blue), 8.7 × 105 yr (purple), 1.9 × 106 yr (magenta), and 3.2 × 106 yr (red). The time evolution results in the existence of large bodies, so that the curves move from left to right. ] 2 m c / g [ Σ Δ 1 10 0 10 -1 10 s -2 10 -1 10 -2 10 i , e -3 10 10 -4 10 2 3 10 4 10 5 10 Radius 6 10 [cm] 7 10 8 10 9 10 Fig. 6.- Same as Fig.5 but with collisional fragmentation for primordial strength. The curves indicate those at 2.2 × 105 yr (blue), 7.2 × 105 yr (purple), 1.4 × 106 yr (magenta), and 3.3 × 106 yr (red). ] 2 m c / g [ Σ Δ 1 10 0 10 -1 10 s -2 10 -1 10 -2 10 i , e -3 10 10 -4 10 2 3 10 4 10 5 10 Radius 6 10 [cm] 7 10 8 10 9 10 Fig. 7.- Same as Fig.5 but with collisional fragmentation for monolith strength. The curves indicate those at 2.1×105 yr (blue), 4.9×105 yr (purple), 1.1×106 yr (magenta), and 3.1×106 yr (red). 4.2. Size and Velocity Distributions around 5 AU As shown in Figs. 2–4, collisional fragmentation natu- rally occurs during planet formation and plays an impor- tant role for planet formation (e.g., Wetherill & Stewart 1993; Inaba et al. 2003; Kobayashi et al. 2010, 2011, 2012; Kobayashi & Dauphas 2013). Therefore collisional fragmentation is necessary to be taken into account for planet formation. For comparison, however, we first show the size and orbit evolution of bodies in the case with- out collisional fragmentation (me = Ψ = 0 in Eq. 6). Fig. 5 shows the evolution of ∆Σs, e, and i of bodies at 5.2 AU in the 3MMSN disk with α = 3 × 10−3, as a function of r corresponding to m. For t . 1 Myr, the surface densities ∆Σs(r) have single peaks, which move to large r. This is caused by orderly growth of bodies. The peak radius rpk is approximated to be the mass-weighted average radius of bodies. For t & 1 Myr, the size distribution becomes wider and then collisional evolution produces another peak at the high-mass end, which indicates planetary embryos. This is caused by the runaway growth of bodies. After the onset of run- away growth, rpk does not change significantly so that the peak radius at the onset of runaway growth, rrg, is approximated to be rpk, which is estimated to ∼ 100 km from the size distribution of bodies at ∼ 1 Myr. The on- set of runaway growth happens if vr . 1.5vesc for bodies of r ∼ rpk (Kobayashi et al. 2016). Since vr/aΩ ≈ e and vesc/aΩ ≈ 6× 10−3(r/100 km), we estimate rrg ≈ 100 km from the data of e at 1.9 Myr in Fig. 5, which is consistent with the mass distribution in Fig. 5. The size distributions of bodies have single peaks at r = rpk prior to runaway growth, t . 1 Myr. Bod- ies with r . rpk have similar slopes of the size distri- butions of bodies. Even after runaway growth, bod- ies with r . rrg have similar slopes. The slope of the mass distribution of bodies is estimated by eye to d ln ∆Σs(r)/d ln r ≈ 1.8. For collisional cascade with Q∗ D and vr independent of m, d ln ns(m)/d ln m = 1/2 (Dohnanyi 1969; Tanaka et al. 1996). The slopes ob- tained from simulations without collisional fragmenta- tion are steeper than that of the simple collisional cas- cade. On the other hand, the slopes of bodies with r ranging from ≈ rrg to 105 km are d ln ∆Σs(r)/d ln r ≈ −2.0 in t & 1 Myr. The collisional cross-section is pro- portional to m5/3/v2 r due to gravitational focusing and vr ∝ m−1/2 due to dynamical friction, which analyti- cally gives d ln ∆Σs(r)/d ln r = −2 (Makino et al. 1998). The slopes formed via runaway growth is roughly ex- plained by gravitational focusing and dynamical friction. In more detailed analysis, the runaway-growth slopes bend slightly (see Morishima 2017). Fig. 6 shows the evolution of bodies at 5.2 AU in the same condition as Figs. 2-4 (i.e., the case with collisional fragmentation for primordial strength). For t . 1 Myr, ∆Σs(r) have a single peak, although ∆Σs(r) have a long tail at low-mass side, which is produced by col- lisional fragmentation. These small bodies make col- lisional damping effective, which reduces vr. Runaway growth occurs slightly earlier than the case without col- lisional fragmentation, resulting in rrg ≈ 60 km smaller than rrg without fragmentation. After the onset of run- away growth (t & 1 Myr), rpk moves insignificantly. The stirring by planetary embryos created by runway 6 growth increases e and i of surrounding planetesimals, which induces collisional fragmentation of planetesimals of r ∼ rrg. The solid surface density is decreased via col- lisional fragmentation and radial drift of resultant frag- ments (Kobayashi et al. 2010, 2011). Planetary embryos grow through the accretion of planetesimals and frag- ments until their depletion. The size distributions of bodies have single peaks at r = rpk prior to runaway growth, which is similar to the case without collisional fragmentation. However, bod- ies smaller than the peaks have multiple slopes: The slopes are estimated by eye to d ln ∆Σs(r)/d ln r ≈ 2.5 for r ≪ 30 m, d ln ∆Σs(r)/d ln r ≈ 0.6 from r ≈ 30 m to ∼ 0.2rpk, and d ln ∆Σs(r)/d ln r ≈ 2.0 for r ≈ 0.2rpk – rpk (see Fig. 6). The slope for large bodies of r = 0.2rpk – rpk is similar to that for no fragmentation, which is simply determined by collisional growth of bodies in or- derly growth. The intermediate sized bodies for r = 30 m – 0.2rpk, which are mainly produced by erosive colli- sions of bodies with r ∼ rpk, have so small e and i that collisional fragmentation is negligible. The colli- sional growth among them results in the slope deter- mined by collisional cascade to the positive mass di- rection, d ln ∆Σs(r)/d ln r = 1/2 (Tanaka et al. 1996). For bodies with r ≪ 30 m, radial drift as well as the collisional cascade affects the mass distribution so that the slope is given by d ln ∆Σs(r)/d ln r = 5/21. On the other hand, the runaway growth produces a dif- ferent power-law size distribution of bodies with r & rrg; d ln ∆Σs(r)/d ln r = −2.0, which is caused by run- away growth similar to the case without fragmenta- tion. For r . rrg, the slope is controlled by collisional fragmentation due to large e and i. Collisional cas- cade therefore occurs, resulting in d ln ∆Σs(r)/d ln r = (1 + 3p)/(2 + p) with 3p = d ln Q∗ D/d ln r − 2d ln vr/d ln r (Kobayashi & Tanaka 2010). Although d ln Q∗ D/d ln r de- pends on collisional velocities, p ≈ 0 is roughly estimated and then d ln ∆Σs(r)/d ln r ≈ 0.5, which is similar to the slope for r ∼ 30 m – 50 km. For r . 30 m, bodies have so small e and i that collisional fragmentation no longer occurs. The slope is determined by collisional growth and radial drift, which is almost the same as the slope of small bodies prior to runaway growth. Fig. 7 shows the result of another simulation with monolith strength for the same disk condition. The evo- lution of size distribution is similar to the case for pri- mordial strength; but the onset of runaway growth oc- curs slightly earlier, resulting in rrg ≈ 30 km. This is because the small bodies produced by collisional frag- mentation prior to runaway growth results in slightly effective collisional damping. After the onset of run- away growth (t & 1 Myr), forming planetary embryos induce collisional fragmentation of planetesimals, which mainly produces bodies with r ∼ 10 m–10 km through collisional cascade. The surface density ∆Σs(r) has a peak at the radius of 10–100 m. This is because the colli- sional cascade starting from collisional fragmentation of bodies with r ∼ rrg is stalled by e and i damping by gas drag in the Stokes regime for bodies with r . 10 – 100 m 1 Such bodies feel gas drag in the Stokes regime so that vr ∝ r−2. The collisional cascade gives (d ln ∆Σs(r)/d ln r)cas = 1/2 and the modulation due to radial drift results in d ln ∆Σs(r)/d ln r = (d ln ∆Σs(r)/d ln r)cas − d ln vr/d ln r. (Kobayashi et al. 2010). The magnitudes of such peaks depend on rrg, Q∗ D, Σs, and so on. Planetary embryos mainly grow through collisions with bodies of ∼ rrg or 10 – 100 m (Kobayashi et al. 2010, 2011). On the other hand, ∼ 300 m-sized bodies are quickly destroyed via col- lisions with small bodies because of low Q∗ D. The mass distribution prior to runaway growth is sim- ilar to the case with primordial strength. The slopes of small bodies are almost the same, while the popu- lation of small bodies is larger because of a high pro- duction caused by collisional fragmentation due to low Q∗ D. On the other hand, although runaway growth re- sults in a slope for r & 30 km ≈ rrg similar to the case with primordial strength, bodies with r . rrg have a "wavy" structure, where the slopes d ln ∆Σs(r)/d ln r varies in small size ranges; d ln ∆Σs(r)/d ln r ≈ 0.1 for r ≈ 3–60 km, d ln ∆Σs(r)/d ln r ≈ 0.2 for r ≈ 0.3– 3 km, d ln ∆Σs(r)/d ln r ≈ −0.2 for r ≈ 30–300 m, and d ln ∆Σs(r)/d ln r ≈ 0.2 for r . 30 m (t = 1.9 Myr in Fig. 7). The slope controlled by collisional cascade is analytically estimated to d ln ∆Σs(r)/d ln r ≈ 0.1 for r & 500 m and d ln ∆Σs(r)/d ln r ≈ 0.0 for r ≈ 30–500 m. Collisional cascade may explain the slope only around r ≈ 3–60 km. The "wavy" pattern for r . 3 km is formed due to large v2 D (e.g., Campo Bagatin et al. 1994; Durda & Dermott 1997; Th´ebault et al. 2003; Krivov 2007; Lohne et al. 2008). The value of v2 D becomes significant around r = 500 m because of minimum Q∗ D (see Fig. 1), which produces a bump at r ≈ 500 m in the size distribution. r /Q∗ r /Q∗ 1 10 0 10 ] ⊕ M [ -1 10 s s a -2 10 M o y r b m E -3 10 -4 10 -5 10 0.3 0.6 1 Time [Myr] 3 6 10 Fig. 8.- Time evolution of planetary-embryo masses for no frag- mentation (magenta), monolith strength (purple), and primordial strength (blue). Fig. 8 shows growth of planetary embryos obtained from simulations shown in Figs. 5–7. For t . 0.4 – 1 Myr, embryo masses ME grow as ME ∝ t3 because of orderly growth (Kobayashi et al. 2016). The onset of runaway growth occurs around 0.4 – 1 Myr, resulting in strong time dependence of ME. The rapid growth occurs until ME ∼ 10−2–10−1 M⊕. The growth timescale in runaway growth depends on the collisional model, which is caused by different rrg. After the rapid growth, slow growth occurs again, which is called oligarchic growth. Planetary embryos grow through surrounding planetes- imals whose typical radii are rrg (see Figs. 5–7). The viscous stirring by embryos increases e and i of plan- etesimals, which induces collisional cascade. Resultant bodies of ∼ 10 – 100 m drift inward rapidly. Therefore, the oligarchic growth is stalled by the depletion of plan- etesimals due to collisional fragmentation and radial drift Formation of Gas Giant Planets 7 of fragments. The growth of embryos at the oligarchic stage depends on Q∗ D for r . rrg that controls the deple- tion of the solid surface density (see Figs. 6 and 7). The oligarchic growth for primordial strength stalls later than that for melted materials. For primordial strength, the characteristic radii of planetesimals, rrg, are large due to the late onset of runaway growth, which have great Q∗ D. The insignificant collisional depletion of planetesimals re- sults in the long accretion of planetesimals onto embryos so that embryos grow to be massive. In addition, colli- sional cascade grinds planetesimals down to 10 – 100 m. Strong gas drag damps vr for r . 100 m so that colli- sional cascade stalls without destructive collisions. The radius of bodies at the low mass end of collisional cas- cade, rcc, depends on Q∗ D; primordial strength has large rcc. Bodies with r ∼ rcc have so small vr that they effec- tive accrete onto embryos, while their drift timescales are short because of strong gas drag. The accretion efficiency of bodies with r ∼ rcc depends on rcc (Kobayashi et al. 2010). For primordial material, embryos effectively grow via accretion of bodies with r ∼ rcc because of slow radial drift for large rcc due to great Q∗ D. 4.3. Turbulent Strength Dependence of Embryo Growth We carry out the collisional-evolution simulations in 3MMSN disks with α = 3 × 10−5 – 3 × 10−3 for pri- mordial Q∗ D, comparing with the results without colli- sional fragmentation (see Fig. 9). For weak turbulence (α . 10−4), the growth of embryos is similar to the case without fragmentation until embryos are larger than the Moon mass (∼ 10−2M⊕). For α = 3 × 10−3, the early collision fragmentation due to strong turbulent stirring leads to effective collisional damping, inducing the onset of runaway growth earlier comparing to the case with- out collisional fragmentation. After runaway growth, large embryos controls vr instead of turbulence, which induces significant collisional fragmentation of planetesi- mals. Collisional cascade results in bodies of r ∼ rcc ∼ 10 – 100 m. The effective accretion of such small bodies leads to the rapid growth of embryos (compare the results with or without collisional fragmentation). However, the growth is stalled by the depletion of small bodies due to collisional fragmentation and radial drift. For small α the runaway growth occurs early, resulting in small rrg. If we ignore collisional fragmentation, the early formation and growth of embryos due to small α results in large embryos (see Fig. 9). However, collisional fragmentation affects the growth of embryos. Small plan- etesimals tend to be destroyed via collision due to small Q∗ D for weak self-gravity (see Fig. 1). The stirring by small embryos increases vr of planetesimals moderately, which can induces collisional fragmentation of planetes- imals if small rrg. Resultant fragments accelerate the growth of embryos initially, while the depletion of sur- rounding planetesimals stalls embryo growth. Weak tur- bulence (small α) results in small masses of embryos in the late stage (t & 4 Myr in Fig. 9), while large α en- hances rrg and then ME at the late stage. Fig. 10 shows embryo growth for monolith Q∗ D. The α dependence is similar to the result for primordial Q∗ D; higher α produces more massive embryos. However, fi- nal embryos are smaller. The accretion of small bodies with r ∼ rcc ≈ 10 – 100 m contributes to embryo growth. However, the depletion of bodies with r ∼ rcc due to 1 10 0 10 -1 10 ] ⊕ M [ s s a -2 10 -3 10 M o y r b m E -4 10 -5 10 0.01 0.1 Time [Myr] 1 10 Fig. 9.- Time evolution of planetary-embryo masses in 3MMSN disk with α = 3 × 10−5 (blue), 3 × 10−4 (magenta), and 3 × 10−3 D (solid) and without (red) for collisional model with primordial Q∗ fragmentation (dotted). radial drift stalls embryo growth. Low Q∗ D for r ∼ 10 – 100 m makes rcc small. Therefore, the growth for mono- lith strength stalls earlier than that for the primordial strength (see Figs. 9 and 10). 1 10 0 10 -1 10 ] ⊕ M [ s s a -2 10 -3 10 M o y r b m E -4 10 -5 10 0.01 0.1 Time [Myr] 1 10 Fig. 10.- Same as Fig. 9, but for monolith Q∗ D (solid). 4.4. Forming Cores with Critical Core Masses As discussed in § 2, gas giant formation via gas accre- tion requires embryos to be larger than ∼ 5M⊕ prior to significant gas depletion. Fig.11 shows embryo masses ME at 5.2 AU at 4 Myr with primordial strength. Mas- sive embryos tend to be formed in the disks with strong turbulence (large α), while too strong turbulence can- not produce massive embryos because the onset of run- away growth is too late. In 3MMSN disks, embryos grow up to 1 – 2M⊕, which are too small to start gas accretion. However, the critical core mass depends on ME. To confirm the possibility of gas giant for- mation, we calculate the masses of static atmospheres around embryos, MA, based on the analytical model for atmospheric radial density profile ignoring the gravity of atmospheres (Inaba & Ikoma 2003). The mass ratio MA/ME is smaller than 0.2 so that runaway gas accre- tion does not occurs within 4 Myr. In the later stage, embryos grow more massive (see Fig. 9); MA/ME & 1/3 at t ≈ 5 Myr and 6–7Myr for α = 3× 10−4 and 3× 10−3, respectively. In 2MMSN, the α dependence of ME and MA is similar to that in 3MMSN, while ME and MA are smaller. Therefore, embryos cannot reach the critical core mass within the disk lifetime . 4 Myr in the disks less massive than 3MMSN. The collisional evolution of fluffy dust aggregates over- comes the radial drift barrier if the dust aggregates that 8 ] ⊕ M [ E M , s s a 10 5 2 1 xg=3,xs=12 xg=2,xs=8 3MMSN 2MMSN 2 1 0.5 E M 0.2 A M 0.1 / 0.5 0.2 0.05 0.02 M o y r b m E xg=3,xs=12 xg=2,xs=8 2MMSN 3MMSN -5 10 -4 10 -3 10 -2 10 -1 10 α Strength, Turbulent -5 10 -4 10 -3 10 -2 10 -1 10 Turbulent Strength, α Fig. 11.- The masses of largest bodies, embryo mass ME (left panel) and the mass ratio of atmosphere to core MA/ME (right panel) for the primordial strength bodies at 4 Myr at 5.2 AU in the disks with xg = xs = 2 (2MMSN; green), xg = xs = 3 (3MMSN; blue), xg = 2 and xs = 8 (red), and xg = 3 and xs = 12 (magenta). Once MA/ME exceeds 1/3, runaway gas accretion to form gas giants occurs. drift most effectively are controlled in the Stokes gas drag regime, resulting in planetesimals within ∼ 10 AU (Okuzumi et al. 2012). During collisional evolution, the radial drift of dust aggregates induces a pile-up in the planetesimal forming region so that the solid surface den- sity increases by a factor 3 – 4 (Okuzumi et al. 2012). According to the result, we set xs/xg = 4 2. In the solid-enhanced disks with xg = 3 and xs = 12, embryo masses at 4 Myr exceed 5M⊕ for α & 3 × 10−4 so that their atmospheric masses MA are much larger than ME/3 (Fig. 11). For xg = 2 and xs = 8, ME for α . 10−3 is similar to that in the case of 3MMSN, while large ME for α & 10−3 results in MA/ME & 1/3. Fig. 12 shows the mass evolution of embryos in the disk with xg = 2 and xs = 8. The growth is stalled at t & 0.1 Myr for α = 3×10−4. The embryo growth occurs within ∼ 1 Myr even for α & 10−3, which forms massive cores. There- fore, in the solid-enhanced disks, runaway gas accretion for gas giant formation occurs within the disk lifetime ∼ 4 Myr. 1 10 0 10 -1 10 ] ⊕ M [ s s a -2 10 -3 10 M o y r b m E -4 10 -5 10 0.01 0.1 Time [Myr] 1 10 Fig. 12.- Same as Fig. 9, but in the disk with xg = 2 and xs for α = 3 × 10−4 (magenta), 3 × 10−3 (red), and 3 × 10−2 (blown). For monolith strength, embryo masses are smaller than that for primordial strength. In the 3MMSN disks, em- bryos are too small to start gas accretion. Even in the 5MMSN disk, MA/ME is smaller than 0.2. Therefore, gas giant formation via core accretion is difficult in the disk 2 Although the solid enhancement due to the aggregate growth occurs within about 10 AU, the solid surface densities are set to be enhanced from 4.8 to 26 AU in the simulations. We also conduct the simulation with the outer disk edge at 9.4 AU for α = 3 × 10−4 and 3 × 10−3 in the disks with xg = 3 and xs = 12 and then the masses of embryos at 4 Myr are the same within 20% for different outer edges. The supply from & 10 AU is insignificant because the late formation of planetary embryos beyond 10 AU. less massive than 5MMSN. In the solid enhanced disks with xg = 3 and xs = 12, embryo masses reach ∼ 1 – 2M⊕ for α & 3×10−3. For α = 3×10−2, MA/ME > 1/3. In the disk with xg = 5 and xs = 20, MA/ME > 1/3 for α ∼ 3 × 10−3. Therefore, melted bodies may form gas giant planets via core accretion in more massive disks, comparing to primordial material. 5. DISCUSSION 5.1. Growth Timescale Prior to the runaway growth, the mass distribution of bodies is approximated to be a single mass population (Kobayashi et al. 2016), so that the collisional timescale at the onset of runaway growth is estimated to be τcol ≈ 4iρsrrg 3eΣsΩ . (25) i/e = 0.5. The balance between turbulent stirring and gas drag gives i/e as ǫ prior to the onset of runaway growth (Kobayashi et al. 2016), while the gravitational inter- action during runaway growth results in the energy equipartition, The embryo formation timescale via runaway growth is proportional to τcol with i/e = 0.5 (Ormel et al. 2010; Kobayashi & Tanaka 2010). The subsequent embryo growth timescale τg, which depends on the accretion of planetesimals or small bodies (Kobayashi et al. 2010, 2011), is simply approxi- mated to be τg ≈ 3τcol; Myr. (26) τg = 4.5(cid:16) xs 3 (cid:17)−1(cid:16) rrg 30 km(cid:17)(cid:16) a 5.2 AU(cid:17)3 The random velocity is determined by the balance be- tween turbulent stirring and collisional damping prior to runaway growth. Once vr ≈ 1.5vesc, runaway growth occurs. The radius of bodies at the onset of runaway growth is given by (Kobayashi et al. 2016) rrg = 32(cid:18) α 3 × 10−3(cid:19)(cid:16) xg 3 (cid:17)2(cid:16) xs 3 (cid:17)−1(cid:16) a 5.2 AU(cid:17)1.5 (27) From Eqs. (26) and (27), the timescale of embryo growth depending on turbulence strength is given by km α xs(cid:19)2(cid:18) τg = 4.8(cid:18) xg As seen in Fig. 11, growing embryos at t = 4 Myr for α = 3×10−3 are less massive than those for α = 3×10−4 for xg = xs, while embryos at t = 4 Myr increases with 3 × 10−3(cid:19)(cid:16) 5.2 AU(cid:17)4.5 Myr. (28) a Formation of Gas Giant Planets 9 5 2 1 ] ⊕ M [ E M , s s a 0.5 M o y r b m E xgcdefs=20 xg_`abs=12 5MMSN 3MMSN 0.2 -5 10 -4 10 -3 10 -2 10 -1 10 α Strength, Turbulent xgklmns=20 0.5 0.2 E M / A M 0.1 0.05 5MMSN 0.02 0.01 0.005 10 -5 xgghijs=12 3MMSN -4 10 -3 10 -2 10 -1 10 Turbulent Strength, α Fig. 13.- Same as Fig. 11, but for monolith strength in the disks with xg = xs = 3 (3MMSN; blue), xg = xs = 5 (5MMSN; grey), xg = 3 and xs = 12 (magenta), and xg = 5 and xs = 20 (purple), α for xs/xg = 4. That is explained by the dependence of τg on xs/xg; For xs/xg = 4, τg < 4 Myr even for α . 3 × 10−2, resulting in embryo formation within the disk lifetime. Collisional fragmentation of planetesimals stalls em- bryo growth so that large rrg tends to form embryos mas- sive enough. The formation of cores with the critical core masses requires rrg & 10 km around 5 AU for primordial strength in the disks with xg ≈ 3. On the other hand, large rrg results in a long embryo-growth timescale (see Eq. 26); τg ≪ 4 Myr is needed. These conditions are not satisfied for 3MMSN, while the solid-enhanced disks create massive cores for α & 10−3 because of these con- ditions (see Fig. 12). 5.2. Planetary Migration and Gas Dispersal Planetary embryos migrate due to density waves caused by interaction with disks; the migration timescale is estimated to be (e.g., Tanaka et al. 2002) Ω−1, τmig = γ−1(cid:18) ME ≈ 1.2γ−1(cid:18) ME ×(cid:16) M∗(cid:19)−1(cid:18) Σga2 M⊕(cid:19)−1(cid:18) 5.2 AU(cid:17)−1/2(cid:18) M∗ M∗ (cid:19)−1(cid:18) cs vK(cid:19)2 0.05 (cid:19)2 450 g cm−2(cid:19)−1(cid:18) cs/vK M⊙(cid:19)3/2 Myr, Σg a (29) where γ is the migration coefficient and the value of Σg is given from that of at 5.2 AU in the 3 MMSN disk. In the isothermal disk, γ ≈ 4 (Tanaka et al. 2002) and the co- rotation torque may reduce γ ∼ 1 (Paardekooper et al. 2011). If the migration timescale is shorter than the time reaching the critical core mass, embryos are lost due to migration prior to gas giant formation. We here estimate the migration timescale for embryos obtained in our simulations using γ = 1. For primordial strength, embryos as large as the critical core mass are formed within several Myrs in the disk with xs/xg ≈ 4 for α ≈ 10−3 – 10−1 (Fig. 11). Fig. 14 shows 3MA/ME at 5.2 AU for α = 3×10−4 and 3×10−3; MA be- comes ME/3 at t ≈ 2 Myr and 3 – 4 Myr for α = 3× 10−4 and 3 × 10−3, respectively and embryos then reach the critical core masses. Using ME obtained from the simu- lations and Eq. (29), we calculate τmig/t. Embryos may grow rather than migration unless τmig/t . 1. There- fore, embryos migrate inward prior to the formation of embryos with the critical core masses; gas giant forma- tion is inhibited by migration. The disk dispersion timescale is required to be com- parable to or shorter than τmig. For τgas = 1 Myr, τmig/t > 1 is satisfied in the simulations. However the gas surface density significantly decreases prior to the onset of the gas accretion of the core (MA = ME/3). Therefore, a more realistic gas dispersal should be taken into consideration. For accretion disks with constant α for simplicity, the gas surface density is proportional to Σg ∝ (1 + t/τdep)−1.5 (Lynden-Bell & Pringle 1974), where τdep = α−1(vK/cs)2Ω−1/3 at the disk radius rcut of the expo- nential cutoff for the surface density. We estimate α τdep = 0.63(cid:18) ×(cid:18) M∗ 3 × 10−3(cid:19)−1(cid:18) cs M⊙(cid:19)−1/2 Myr. 0.1vK(cid:19)−2 (cid:16) rcut 50 AU(cid:17)3/2 (30) Therefore, Σg may decrease on the timescale of 1 Myr. 10 5 2 1 0.5 0.2 τmig / t 3 MA / ME opq rst uvw xyz 1 2 3 4 [Myr] {}~ Fig. 14.- Dimensionless migration timescales τmig/t (solid curves) calculated from the results of simulations with primordial strength in disks with xg = 3 and xs = 12 for α = 3 × 10−4 (blue) and 3 × 10−3 (red). Dotted curves indicate 3MA/ME; The onset of runaway gas accretion is estimated from 3MA/ME = 1. We perform the simulations assuming Σg ∝ Σg,0/(1 + t/τdep)3/2 with τdep = 0.5 Myr (Fig. 15), which mimics to the accretion disk. Embryos reach the critical core mass at 3 – 4 Myr. The early gas depletion weakens the damping of eccentricity and inclination due to gas drag, and then the accretion of planetesimals on embryos is suppressed due to large e and i. However, small bod- ies produced by collisional fragmentation feel the Stokes gas drag, which is independent of gas density. The early gas depletion does not affect the accretion of such small bodies significantly. Therefore, embryos exceed the crit- 10 ical core mass within 4 Myr. On the other hand, the early gas depletion affects migration significantly. The gas depletion with τdep . 1 Myr prolongs τmig, resulting in τmig/t > 1. Therefore, embryos may start rapid gas accretion prior to migration. τmig / t 10 5 2 1 0.5 0.2 1 [Myr] 3 MA / ME 2 3 4 Fig. 15.- Same as Fig. 14, but for Σg = Σg,0/(1 + t/τdep)3/2 with τdep = 106 yr. The disk with xg = 3 initially has ∼ 17MJ, where we set the disk edge at 50 AU and MJ is the mass of Jupiter. The disk mass decreases with ∝ (1 + t/τdep)−1/2 (Lynden-Bell & Pringle 1974) so that the disk mass be- comes 6MJ at t = 4 Myr for τdep = 0.5 Myr. The gas accretion of embryos with the critical core masses oc- curs not only from around embryos with the critical core masses but also from the whole disk during disk evo- lution (Tanigawa & Ikoma 2007), so that embryos may acquire an atmosphere comparable to Jupiter. There- fore, gas accretion occurs in such a depleting disk, which saves forming gas giants from type II migration (e.g., Ida & Lin 2008). It should be noted that τdep does not correspond to the disk lifetime inferred from the infrared observation. In the self-similar solution for accretion disks (Lynden-Bell & Pringle 1974), the disk evolution timescale prolongs for t ≫ τdep. Even for τdep . 1 Myr, the large amount of gas remaining in several Myrs may be compatible with the disk lifetime from observations. 6. SUMMARY We investigate planet formation in a turbulent disk. Turbulence suppresses the runaway growth of planetesi- mals. Once the random velocity of planetesimals is com- parable to their escape velocity, runaway growth occurs (Kobayashi et al. 2016). The mass-weighted average ra- dius during and after runaway growth is approximated to be that at the onset of runaway growth, rrg. Em- bryos formed through runaway growth become massive through the accretion of planetesimals with r ∼ rrg. However, the stirring by massive embryos induces de- structive collisions of planetesimals. Collisional cascade grinds bodies down to 10 – 100 m. Embryos grow through effective accretion of small bodies, while radial drift re- duces small bodies. Eventually, the embryo growth stalls due to the depletion of bodies surrounding embryos via destructive collisions and radial drift. Therefore, the for- mation and growth of embryos strongly depends on the collisional properties for r ∼ rrg, which is controlled by turbulent strength. We have carried out simulations of collisional evolution of bodies with collisional strengths (Fig. 1) for the forma- tion and growth of embryos in various disks, especially taking into account the stirring by density fluctuation caused by turbulence (Eq. 13). We find the followings. • Strong turbulence delays the onset of runaway growth and increases rrg. Embryos forming within disk lifetimes, t ≈ 4 Myr, tend to be large for high α, while the onset of runaway growth is too late to form massive embryos within disk lifetimes for α & 10−3. Cores massive enough are formed from rg & 10 km corresponding to α & 10−3. However, the formation timescales of such cores are longer than the disk depletion time for α & 10−3. • Solid-enhanced disks are preferable for the forma- tion of massive embryos. Such a local enhancement of solid can occur through radial drift of dust ag- gregates. For weak turbulence α . 10−3, embryo masses at t = 4Myr are similar even in the en- hancement of solid. However, embryos may grow within the disk lifetime even for strong turbulence α & 10−3. Therefore, embryos are as large as the critical core mass within disk lifetimes in the gas disk more massive than 2MMSN. • The embryo growth depends on collisional strength of bodies. We investigate collisional evolution for primordial and melted bodies using the model de- scribed in Eqs. (14)–(20) as shown in Fig. 1. For weak turbulence disks, embryo growth is indepen- dent of collisional strength until ME ≈ 0.01M⊕. However, once collisional fragmentation of bod- ies with r ∼ rrg is effective, the embryo growth strongly depends on collisional strength. The low- mass end of collisional cascade is important for the efficiency of the accretion of bodies with r = 10– D & 107 erg/g for 100 m. The collisional strength Q∗ r . 1 km, which is satisfied for primordial bodies rather than monolith material, is likely to form a massive core to be a gas giant (see Figs. 11 and 13). • We have estimated the timescale of planetary mi- gration during embryo growth. We have taken into account the gas density evolution similar to the accretion disk model. For example, embryos grow to the critical core mass for gas giant for- mation prior to migration in the disk initially con- taining 3MMSN gases for the dispersal timescale of 0.5 Myr (see Fig. 15). In spite of the short dispersal timescale, the disk mass at the onset of rapid gas accretion remains much larger than Jupiter masses. Therefore, moderately massive disks with short dis- persal timescales are likely to form gas giants with- out significant migration. The work is supported by Grants-in-Aid for Scien- tific Research (26287101,17K05632,17H01105,17H01103) from MEXT of Japan and by JSPS Core-to-Core Pro- gram "International Network of Planetary Sciences". Formation of Gas Giant Planets APPENDIX PLANETARY ATMOSPHERE The atmosphere structure is governed by (e.g., Inaba & Ikoma 2003) dP dR dT dR GMEρa , R2 =− =( − 3 4πσSB κLeP T 4 −(cid:16)1 − 1 Γ2(cid:17) GMρa P R2T if 3κLeP 4πσSBT 4 < 1 − 1 Γ2 otherwise, 11 (A1) (A2) , where P is the pressure, T is the temperature, and Γ2 is the second adiabatic exponent. In this study, we assume the atmospheric mass MA is much smaller than ME . However, if MA/ME & 1/3, Eq. (A1) is invalid and then the rapid gas accretion forms gas giants (e.g. Mizuno 1980). In Eq. (A2), the first term on the right hand side is given by the radiative energy transfer, while the second term is determined by convective transfer. Here, we derive a solution using the radiative term, although we take into account the both terms for simulations below § 3. Dividing Eq. (A2) by Eq. (A1) and integrating over P , we have 16πσSBGMET 3µmH . ρa = 3kκLe Inserting Eq. (A3) into Eq. (A2) and integrating it, we then obtain T = GMEµmH 4kR . Substitution of Eq. (A4) into Eq. (A3) gives Eq. (1). COLLISIONAL OUTCOME MODELING (A3) (A4) Collisional outcome from the collision between bodies with m1 and m2 is expressed by me and Ψ(m, m1, m2), which are determined as follows. me = φ 1 + φ (m1 + m2), (B1) where φ is the dimensionless impact energy. Using the impact velocity vimp and Q∗ energy needed for ejection of half bodies, φ = m1m2vimp/2(m1 + m2)2Q∗ D. D, which is the specific impact (m1 + m2)Ψ(m, m1, m2) =  me me(cid:18) m mL(cid:19)2−b (m > mL), (m ≤ mL), (B2) where φ is the dimensionless impact energy and mL is the largest mass of fragments produced by a single collision between bodies with m1 and m2, given by mL = ǫL 1 + φ me = ǫLφ (1 + φ)2 (m1 + m2), (B3) b and ǫL are constants. We set b = 5/3 and ǫL = 0.2. The timescale of collisional cascades is insensitive to values of b and ǫL (Kobayashi & Tanaka 2010). REFERENCES Adachi, I., Hayashi, C., & Nakazawa, K. 1976, Progress of Theoretical Physics, 56, 1756 Benz, W., & Asphaug, E. 1999, Icarus, 142, 5 Campo Bagatin, A., Cellino, A., Davis, D. R., Farinella, P., & Paolicchi, P. 1994, Planet. Space Sci., 42, 1079 Dohnanyi, J. S. 1969, J. Geophys. Res., 74, 2531 Durda, D. D., & Dermott, S. F. 1997, Icarus, 130, 140 Goldreich, P., & Ward, W. R. 1973, ApJ, 183, 1051 Hayashi, C. 1981, Progress of Theoretical Physics Supplement, 70, 35 Ida, S., & Lin, D. N. C. 2008, ApJ, 673, 487 Ikoma, M., Nakazawa, K., & Emori, H. 2000, ApJ, 537, 1013 Inaba, S., & Ikoma, M. 2003, A&Ap, 410, 711 Inaba, S., Wetherill, G. W., & Ikoma, M. 2003, Icarus, 166, 46 Inaba, S., Tanaka, H., Nakazawa, K., Wetherill, G. W., & Kokubo, E. 2001, Icarus, 149, 235 Kobayashi, H. 2015, Earth, Planets and Space 67, 60 Kobayashi, H., & Dauphas, N. 2013, Icarus, 225, 122 Kobayashi, H., Ormel, C. W., & Ida, S. 2012, ApJ, 756, 70 Kobayashi, H., & Tanaka, H. 2010, Icarus, 206, 735 Kobayashi, H., Tanaka, H., Krivov, A. V., & Inaba, S. 2010, Icarus, 209, 836 Kobayashi, H., Tanaka, H., & Krivov, A. V. 2011, ApJ, 738, 35 Kobayashi, H., Tanaka, H., & Okuzumi, S. 2016, ApJ, 817, 105 Krivov, A. V. 2007, Dust in Planetary Systems, 643, 123 Lohne, T., Krivov, A. V., & Rodmann, J. 2008, ApJ, 673, 1123-1137 Lynden-Bell, D., & Pringle, J. E. 1974, MNRAS, 168, 603 Makino, J., Fukushige, T., Funato, Y., & Kokubo, E. 1998, New A, 3, 411 Michikoshi, S., Kokubo, E., & Inutsuka, S.-i. 2012, ApJ, 746, 35 Mizuno, H. 1980, Progress of Theoretical Physics, 64, 544 Morishima, R. 2017, Icarus, 281, 459 Ohtsuki, K. 1992, Icarus, 98, 20 Ohtsuki, K., Stewart, G. R., & Ida, S. 2002, Icarus, 155, 436 Okuzumi, S., & Ormel, C. W. 2013, ApJ, 771, 43 12 Okuzumi, S., Tanaka, H., Kobayashi, H., & Wada, K. 2012 ApJ, 752, 106 Ormel, C. W., Dullemond, C. P., & Spaans, M. 2010b, Icarus, 210, 507 Ormel, C. W., & Klahr, H. H. 2010, A&A, 520, A43 Ormel, C. W., & Okuzumi, S. 2013, ApJ, 771, 44 Paardekooper, S.-J., Baruteau, C., & Kley, W. 2011, MNRAS, 410, 293 Stevenson, D. J. 1982, Planet. Space Sci., 30, 755 Stewart S. T., Leinhardt Z. M., 2009, ApJ, 691, L133 Suyama, T., Wada, K., & Tanaka, H. 2008, ApJ, 684, 1310-1322 Suyama, T., Wada, K., Tanaka, H., & Okuzumi, S. 2012, ApJ, 753, 115 Takeuchi, T., & Ida, S. 2012, ApJ, 749, 89 Tanaka, H., Inaba, S., & Nakazawa, K. 1996, Icarus, 123, 450 Tanaka, H., Takeuchi, T., & Ward, W. R. 2002, ApJ, 565, 1257 Tanigawa, T., & Ikoma, M. 2007, ApJ, 667, 557 Th´ebault, P., Augereau, J. C., & Beust, H. 2003, A&A, 408, 775 Volk, H. J., Jones, F. C., Morfill, G. E., & Roser, S. 1980, A&A, 85, 316 Wada K., Tanaka H., Okuzumi S., Kobayashi H., Suyama T., Kimura H., Yamamoto T., 2013, A&A, 559, A62 Wetherill, G. W., Stewart, G. R. 1989, Icarus77, 330 Wetherill, G. W., & Stewart, G. R. 1993, Icarus, 106, 190 Youdin, A. N., & Goodman, J. 2005, ApJ, 620, 459
1503.06955	1	1503	2015-03-24T09:14:07	Stellar wind induced soft X-ray emission from close-in exoplanets	[ "astro-ph.EP" ]	In this paper, we estimate the X-ray emission from close-in exoplanets. We show that the Solar/Stellar Wind Charge Exchange Mechanism (SWCX) which produces soft X-ray emission is very effective for hot Jupiters. In this mechanism, X-ray photons are emitted as a result of the charge exchange between heavy ions in the solar wind and the atmospheric neutral particles. In the Solar System, comets produce X-rays mostly through the SWCX mechanism, but it has also been shown to operate in the heliosphere, in the terrestrial magnetosheath, and on Mars, Venus and Moon. Since the number of emitted photons is proportional to the solar wind mass flux, this mechanism is not very effective for the Solar system giants. Here we present a simple estimate of the X-ray emission intensity that can be produced by close-in extrasolar giant planets due to charge exchange with the heavy ions of the stellar wind. Using the example of HD~209458b, we show that this mechanism alone can be responsible for an X-ray emission of $\approx 10^{22}$~erg~s$^{-1}$, which is $10^6$ times stronger than the emission from the Jovian aurora. We discuss also the possibility to observe the predicted soft X-ray flux of hot Jupiters and show that despite high emission intensities they are unobservable with current facilities.	astro-ph.EP	astro-ph	Draft version July 8, 2021 Preprint typeset using LATEX style emulateapj v. 05/12/14 STELLAR WIND INDUCED SOFT X-RAY EMISSION FROM CLOSE-IN EXOPLANETS K.G. Kislyakova1, L. Fossati2, C.P. Johnstone3, M. Holmstrom4, V.V. Zaitsev5, & H. Lammer1 1Space Research Institute, Austrian Academy of Sciences, Graz, Austria; [email protected] 2Argelander-Institut fur Astronomie der Universitat Bonn, Bonn, Germany 3University of Vienna, Department of Astrophysics, Vienna, Austria 4Swedish Institute of Space Physics, Kiruna, Sweden 5Institute of Applied Physics, Russian Academy of Sciences, Nizhny Novgorod, Russia (Dated: July 8, 2021) Draft version July 8, 2021 ABSTRACT In this paper, we estimate the X-ray emission from close-in exoplanets. We show that the So- lar/Stellar Wind Charge Exchange Mechanism (SWCX) which produces soft X-ray emission is very effective for hot Jupiters. In this mechanism, X-ray photons are emitted as a result of the charge exchange between heavy ions in the solar wind and the atmospheric neutral particles. In the Solar System, comets produce X-rays mostly through the SWCX mechanism, but it has also been shown to operate in the heliosphere, in the terrestrial magnetosheath, and on Mars, Venus and Moon. Since the number of emitted photons is proportional to the solar wind mass flux, this mechanism is not very effective for the Solar system giants. Here we present a simple estimate of the X-ray emission intensity that can be produced by close-in extrasolar giant planets due to charge exchange with the heavy ions of the stellar wind. Using the example of HD 209458b, we show that this mechanism alone can be responsible for an X-ray emission of ≈ 1022 erg s−1, which is 106 times stronger than the emission from the Jovian aurora. We discuss also the possibility to observe the predicted soft X-ray flux of hot Jupiters and show that despite high emission intensities they are unobservable with current facilities. Subject headings: planets and satellites: aurorae -- planet-star interactions -- planets and satellites: individual (HD 209458b) -- X-rays: general -- radiation mechanisms: non-thermal 1. INTRODUCTION e.g. X-ray the Solar system objects, emission has been observed for many of for Mars (Holmstrom et al. 2001; Gunell et al. 2004; Dennerl 2002), Venus (Bhardwaj et al. 2007; Dennerl et al. 2002), Earth and the Moon (Collier et al. 2014; Bhardwaj et al. and the Galilean satellites (Cravens et al. 2003; Bhardwaj et al. 2007), Saturn comets (Cravens 2002; Lisse et al. 2004), and in the heliosphere (Cravens et al. 2001). (Branduardi-Raymont et al. 2007), Jupiter 2010), For the Solar system planets, X-rays are known to be generated via different mechanisms. The main mecha- nisms are: -- continuum Bremsstrahlung emission due to collisions with electrons (produces mostly hard X-rays); -- excitation of neutral species and ions due to colli- sions, e.g., with electrons (charged particle impact), fol- lowed by line emission; -- stellar X-ray photon scattering from neutrals in plan- etary atmospheres (elastic scattering and K-shell fluores- cent scattering, requires a significant column density); -- charge exchange between the solar wind ions with neutrals (SWCX), followed by X-ray emission; -- X-ray production from the charge exchange of en- ergetic (energies of about a MeV/amu) heavy ions of planetary magnetospheric origin with neutrals or by di- rect excitation of ions in collisions with neutrals (this is known to be effective on Jupiter, e.g. Cravens et al. 2003; Bhardwaj et al. 2007). The cross sections at solar wind energies for charge ex- change with the solar wind heavy ions are several magni- tudes larger than the cross sections for the excitation of the neutral species by electrons (Bhardwaj et al. 2007), which makes the SWCX mechanism more effective. In the present article, we discuss the SWCX mecha- nism and X-ray scattering as applied to close-in giant exoplanets, in particular to HD 209458b. We discuss the observability of the X-ray emission from HD 209458b, X- ray emission from other giant planets, and the influence of the host star age. Other X-ray production mechanisms are beyond the scope of the present article, but will be the goal of a future study. 2. SOLAR WIND CHARGE EXCHANGE MECHANISM In the SWCX mechanism, an electron is transferred from a neutral atom or molecule to a highly charged heavy ion of the solar wind. This mechanism is known to produce soft X-rays in cometary comas (Cravens 2002). In the case of a magnetized planet, these ions can enter the neutral atmosphere following the open field lines near the polar cusp. It is known from experimental and theoretical studies that solar wind heavy ions can undergo charge exchange reactions when they are within approximately 1 nm of a neutral atomic species (e.g., Lisse et al. 2004; Cravens 2002; Bhardwaj et al. 2007 and references therein): Aq+ + B → A(q−1)+∗ + B+, (1) where A is a charged heavy ion in the solar wind (the projectile), q is the projectile charge and B is a neutral component (target). The product ion A(q−1)+∗ is still highly charged and is almost always left in an excited state (marked by an "*"). Then, the excited ion emits 2 Planetary and stellar wind parameters, HD 209458b TABLE 1 Name Planetary mass, Mp Planetary radius, Rp Semi-major axis Stellar wind velocity, vsw Stellar wind density, nsw Fraction of heavy ionsb, f Planetary magnetic moment, Mp Magnetospheric stand-off distance, Rs Valuea ≈ 0.71MX ≈ 1.38RX ≈ 0.047 AU 4 × 107 cm s−1 5 × 103 cm−3 10−3 ≈ 0.1MX 2.9Rp aadopted from Kislyakova et al. (2014) bsolar wind value assumed one or several X-ray photons in the following reaction: A(q−1)+∗ → A(q−1)++hν. (2) Although the de-excitation usually represents a number of cascading processes through intermediate states, if q is high then an X-ray photon (at least one, though usually several) is emitted (Cravens 2002). The composition of the solar wind by volume is 0.92 hydrogen, 0.08 helium, and ≈ 10−3 heavier elements. Since the solar wind quickly becomes collisionless as it expands, the charge states that the heavy ions have in the hot solar corona are frozen-in, and therefore the heavy elements are usually highly charged. In the solar wind, the most common heavy ions are C4+, C5+, N6+, O6+, O7+, O8+, Ne8+, Si9+, Fe12+. The cross sections for such charge transfer collisions are very high at solar wind energies, exceeding 1015 cm2 (e.g., Greenwood et al. 2001). The types of the species that undergo charge exchange define the energy of the emitted X-ray photons, which is usually in the range of 0.3 -- 0.5 keV. 3. SWCX MECHANISM ON HOT JUPITERS: THE CASE STUDY OF HD 209458B In this section, we discuss the soft X-ray emis- sion which can be produced by the SWCX mechanism on close-in exoplanets. As an example, we consider HD 209458b, which is a well-studied close-in gas giant or- biting a 4±2 Gyr old G-type star. Planetary and stellar wind parameters are summarized in Table 1. In our fur- ther estimations, we rely on results of Kislyakova et al. (2014), who investigated the magnetosphere and stellar wind parameters in the vicinity of HD 209458b by means of modelling. Their result support a magnetic moment of HD 209458b of approximately 10% of the Jupiter's and a stellar wind with a velocity of 4 × 107 cm s−1 at the time of observation. For a very simple estimate of the X-ray intensity, I, emitted in the region of the atmosphere exposed to heavy ion precipitation one can use the following expression (Cravens et al. 2003): 4πI ≈ 2nswvswf N, (3) where N is a factor of 2 or 3 and represents the number of photons emitted per ion (below we assume N = 3). The additional factor of 2 on the right hand side is a flank magnetosheath enhancement factor (Cravens et al. 2003). For Jupiter, Equation 3 yields 4πI ≈ 105 cm−2 s−1 while values of 2 × 106 − 2 × 107 cm−2 s−1 are neces- sary to explain the observed auroral soft X-ray emission, which means that SWCX is not the main mechanism that produces soft X-ray emission for Jupiter. For HD 209458b, we assume the stellar wind param- eters estimated by Kislyakova et al. (2014), nsw = 5 × 103 cm−3, vsw = 4 × 107 cm s−1 and f ≈ 10−3 which gives 4πI ≈ 1.2 × 109 cm−2 s−1 or 60 -- 600 times the observed Jovian values. Given its proximity to its host star, it is unclear whether HD 209458b is located in the sub-Alfv´enic or super-Alfv´enic region of the wind. The exact regime depends on the magnetic moment of the host star and stellar wind parameters. Although the results of Kislyakova et al. (2014) support that HD 209458b is rather in the super-Alfv´enic regime and thus outside the stellar plasma corotation region, we make an es- timate also for the corotation case. HD 209458 has been observed to have a rotational velocity of 4.4 km/s which corresponds to a rotational period of ≈11.5 days (Mazeh et al. 2000). This gives the corotational velocity of plasma at 0.047 AU of vcor ≈ 2.7 × 106 cm s−1. Taking into account also the Keplerian orbital speed vorb ≈ 1.4× 107 cm s−1, this corresponds to the plasma flow velocity in the vicinity of HD 209458b of vflow ≈ 1.2 ×107 cm s−1. Substituting it into the equation 3 instead of vsw, one ob- tains an estimate of 4πI ≈ 3.6 × 108 cm−2 s−1, which is still 18 -- 180 times the observed Jovian value. To estimate the aurora size of HD 209458b we follow the approach of Vidotto et al. (2011). The fractional area of the planetary surface that has open magnetic field lines is (1 − cos α0) for both the north and south auroral caps, where α0 = arcsin[(Rp/Rs)1/2], (4) Rp is the radius of a planet, and Rs is the magnetosphere stand off distance at the substellar point. Assuming Rs = 2.9Rp estimated by Kislyakova et al. (2014), we obtain α0 ≈ 0.63 and 1 − cos α0 ≈ 0.19. This gives the size of the aurora of A ≈ 2.17 × 1020 cm2, or ≈ 217 times the Jovian aurora AX ≈ 1018 cm2 (Cravens et al. 2003). Now we can estimate the power of the soft X-ray emis- sion from HD 209458b in both the corotation and non- corotation regimes. For simplicity, we assume the energy of each emitted X-ray photon is 0.3 keV (Cravens et al. 2003). Using the solar value of f = 10−3, we esti- mate the total X-ray power of HD 209458b to be ≈ 1.3 × 1020 erg s−1 in the non-corotation regime (point C on Fig.1) and ≈ 2.3 × 1019 erg s−1 in the corotational regime (Fig.1, point D). Note that these values can still present a lower limit. If charge exchange occurs not only in the auroral regions of HD 209458b, but in the whole hemisphere with the radius of Rs, the values should be multiplied by a factor of ≈ 88, which is the ratio of the interaction area sizes. This gives ≈ 1.1 × 1022 erg s−1 in the non-corotation regime (point A in Fig.1) and ≈ 2.0 × 1021 erg s−1 in the corotation regime (Fig.1, point B). Since the atmo- spheres of hot Jupiters in general and HD 209458b in particular are highly inflated and are believed to extend beyond the magnetosphere (Kislyakova et al. 2014), this is a more realistic case than interaction only in the au- rora region (Jupiter type). The X-ray production can be even larger if the region outside the magnetosphere (the volume between the magnetopause and the bow shock) is included (Robertson & Cravens 2003). 3.1. Contribution of stellar X-ray photon scattering Although only a small fraction of the incident stellar X- rays are reflected by the planetary atmosphere, in the So- lar system this mechanism is known to contribute to the total soft X-ray luminosity of planets and dominates, for example, the X-rays from Venus (Dennerl et al. 2002). Cravens et al. (2006) showed that the scattering albedo for the outer planets is quite small and equals 10−3 at 3 nm. We assume this albedo for HD 209458b as a crude estimate. The total X-ray luminosity of HD 209458 was first ob- served to be log LX ≈ 27.02 ± 0.2 erg s−1 (Kashyap et al. 2008). However, later it was reported that this result may present a luminosity of a nearby star and a new upper limit of log LX . 26.12 erg s−1 was reported (Sanz-Forcada et al. 2010). Both values are close to the Solar X-ray luminosity of log LX⊙ ≈ 26 − 27 erg s−1. The value of log LX = 26.12 erg s−1 yields an X-ray flux of ≈ 39 erg cm−2 s−1 in the vicinity of HD 209458b and an X-ray luminosity of reflected soft X-rays from the planet of ≈ 2.3 × 1019 erg s−1, which is comparable only to our lowest estimate for the X-ray flux produced by the SWCX mechanism (point D in Fig1). 4. OBSERVABILITY In order to see whether the soft X-ray exoplane- tary emission described above could be observable with currently available facilities, we have considered the maximum X-ray luminosity estimated for HD 209458b (i.e., 1.1 × 1022 erg s−1) and that of its host star (log LX = 27.02 erg s−1; Kashyap et al. 2008). Note that Sanz-Forcada et al. (2010) reported only an upper limit of log LX < 26.12 erg s−1 on the X-ray luminosity of HD 209458 and concluded that Kashyap et al. (2008) might have confused the star with a nearby object. For our purposes we are interested in the best possible sce- nario and therefore assume the X-ray luminosity given by Kashyap et al. (2008). A lower stellar X-ray lumi- nosity would make the detection of the planetary X-ray emission harder to detect compared to what described here. By rescaling both luminosities to the distance of HD 209458 (d = 49.6 pc; van Leeuwen 2007), we ob- tained that the maximum X-ray flux of HD 209458b would be 3.9 × 10−20 erg cm−2 s−1, while from the star we would get an X-ray flux of 3.6 × 10−15 erg cm−2 s−11, resulting in a difference of about 5 orders of magnitude. The planetary X-ray emission could be observed us- ing the secondary transit, as commonly done in the in- frared, for example. The ratio between the in-transit and out-of-transit fluxes is expected to be of the order of 10−5 erg cm−2 s−1, which would require a signal-to-noise ratio (S/N) of the measurements of the order of 105 to be detected. Such a high precision is currently reached in the optical and infrared bands (mostly with space ob- servations), but it is prohibitive at X-ray wavelengths. 1 The slight difference in the stellar X-ray flux with that given by Kashyap et al. (2008) is due to the use of a different distance. 3 To highlight this, we used the available count rate sim- ulator2 for the XMM-Newton telescope that, among the facilities currently available, has the largest efficiency in the soft X-rays. Taking into account the 0.1 -- 1.0 keV band, a plasma with a temperature of 106 K, and the count rate given for the pn detector and the "thin" filter we obtained a count rate of 3.9 × 10−3 counts s−1 for the X-ray emission of HD 209458. As a result, the S/N ob- tained exposing for 103 seconds is about 2 and it would require several thousand years of exposure time to reach the S/N required to detect the planetary X-ray emission. The situation might slightly improve if the planetary X- ray emission has a different spectral behaviour compared to that of the star, but the detectability would probably remain unfeasible. Here we have not considered the in- trinsic stellar X-ray variability which will hamper the detection of the planetary X-ray emission. 5. SWCX MECHANISM ON OTHER GIANT EXOPLANETS In this section, we briefly consider the influence of the orbital distance on the soft X-ray emission generated by the SWCX mechanism (the radius of HD 209458b is as- sumed). For the stellar wind parameters, we consider two cases: for one case, we assume the wind has the proper- ties of the slow component of the current solar wind, and for the other case, we scale the wind properties to match what we might expect for the young solar analogue EK Dra. We calculate the solar slow wind parameters as a func- tion of distance from the star using a 1D hydrodynamic wind model that is constrained by in situ spacecraft mea- surements of the real solar wind. The model was devel- oped by Johnstone et al. (2015b) and provides a very good description of the real solar wind outside of the solar corona. Although little is known about the properties of winds from other stars, it is suspected that more active stars have mass fluxes that are significantly higher than the mass flux of the current solar wind (Wood et al. 2005; Holzwarth & Jardine 2007; Suzuki et al. 2013). To scale the slow solar wind model to EK Dra, we use the scaling relation for mass loss rate derived by Johnstone et al. (2015a) and the parameters for EK Dra given by Gudel (2007). We find a mass loss rate, and therefore corresponding values for nswvsw, for EK Dra that are approximately a factor of 15 higher than in the current solar wind. Based on the example of the solar wind, we might ex- pect that the abundances of heavy ions are similar in the corona and in the wind. While it is known that the coronal abundances are correlated with coronal activity for Sun-like stars, the coronal abundances of the most active stars are only approximately a factor of two dif- ferent from the solar values (Telleschi et al. 2005). Since this is an insignificant difference compared to all other uncertainties, for simplicity we assume a solar value of f ≈ 10−3. Magnetic moments of tidally locked gas giants are be- lieved to be smaller than MX because of their slower rotation (Griessmeier et al. 2004; Khodachenko et al. 2012). For HD 209458b, this hypothesis was lately also 2 http://heasarc.gsfc.nasa.gov/cgi-bin/Tools/w3pimms/w3pimms.pl 4 Fig. 1. -- Dependence of the soft X-ray emission produced by the SWCX mechanism on the orbital semi-major axis of a HD 209458b-like giant planet. Left panel: the calculation using the stellar wind parameters f , nsw, vsw of the current solar wind (4.5 Gyr old G2V dwarf). The bright shaded area shows the locations where planets could potentially be tidally locked and the dark shaded area shows where planets will be tidally locked. The letters A, B, C, and D mark the estimates for HD 209458b. The upper and lower green lines are estimates for emission from auroral region with aurora size calculated according to Eq. 4 and for a fixed Jovian aurora of AX = 1018 cm2 respectively. The upper and lower red lines are estimates for a tidally locked planet for the charge exchange in the whole hemisphere restricted by Rs and only auroral regions respectively. The upper and lower blue lines illustrate the estimates for a tidally locked planet in a corotation regime. The vertical dashed blue lines show the semi-major axes of some known hot Jupiters orbiting G dwarf stars. The point labelled "Jupiter" marks the observed soft X-ray emission from the Jovian aurora, and "Jupiter (SWCX)" stands for the intensity produced via the SWCX mechanism only. Right panel: the same as the left panel, but assuming the stellar wind parameters of the young stellar analogue EK Dra (100 Myr old G1.5V dwarf). supported by a modelling result based on the Lyα tran- sit observations (Kislyakova et al. 2014) which predicted a planetary magnetic moment of Mp ≈ 0.1MX. Al- though a prediction of magnetic moments of hot Jupiters ≥ MX also exists (Christensen et al. 2009), in the present study we assume a moment value in the range of ≈ 0.05 − 0.5MX (see Fig. 2 in Khodachenko et al. 2012). The size of the aurora is calculated according to Eq. 4 and the magnetopause stand off distance following the relation (Baumjohann & Treumann 1996) Rs = µ0f 2 0 M2 p 8π2ρswv2 sw!1/6 , (5) where µ0 is the diamagnetic permeability of free space, f0 ≈ 1.22 is magnetosphere form-factor. Fig. 1 presents the dependence of the emitted soft X- ray power on the orbital distance of the planet. The let- ters mark the emission levels for HD 209458b estimated above. We should note that we did not take into account the unknown rotation rate of the host star, which leads to an overestimate of the plasma flow speed and, respec- tively, the emission level in the corotation regime. The letter marks for HD 209458b don't lie exactly on the lines because of the difference between simple estimates used for Rs and Mp to those estimated via comprehensive modelling by Kislyakova et al. (2014). As a consequence these plots only qualitatively describe the behaviour of the soft X-ray emission due to the SWCX mechanism. For every particular planet, an individual consideration should be made similar to the one above for HD 209458b. The main conclusion of our results is that the soft X- ray emission is the highest for closest hot Jupiters and strongly depends on the interaction area size (see the difference between the aurora and the whole hemisphere case -- lower and upper red and blue lines, respectively). The simple Equation 4 (Vidotto et al. 2011) can be used only for hot Jupiters and yields a significant overestimate for Jupiter (see the two green lines). For non tidally locked gas giants on wide orbits, the lower green line presents the most plausible estimate. We should also note that the corotation regime proba- bly breaks closer to the star than shown on Fig. 1. How- ever, this is not so easy to restrict because of many un- known parameters. Soft X-ray emission from exoplanets orbiting a younger star with a denser stellar wind is always stronger than the emission from an planet embedded in the current solar wind (eq. 3), which is confirmed by Fig. 1b, simply because the number of emitted photons is proportional to the wind mass flux assuming the same f . 6. CONCLUSIONS In this work, we presented a simple estimate of the pos- sible X-ray emission from the close-in gas giants emitted due to the SWCX mechanism. We have shown for the example of HD 209458b that this mechanism alone can be responsible for the X-ray emission intensity of the or- der of ≈ 1022 erg s−1, which is ≈ 106 times higher than the X-ray emission from Jupiter. We have discussed the possibility to observe the soft X-ray flux from close-in extrasolar giant planets and have shown that although this emission exceeds the intensity of the Jovian soft X-ray emission by several orders of magnitude, it is unobservable with present-day facilities because of the large distances to the systems. The main conclusion of the study is that hot Jupiters should be bright X-ray sources in comparison to the So- lar system giant planets. The spectrum of this emission as well as the influence of other X-ray producing mecha- 5 nisms should be the subject of future study. This study was carried out with the support by the FWF NFN project S116601-N16 "Pathways to Habit- ability: From Disk to Active Stars, Planets and Life" and the related subprojects S116 604-N16 and S116 607-N16. L.F. acknowledges financial support from the Alexander vun Humboldt foundation. L.F. thanks Lorenzo Lovisari for useful discussions. V.Z. acknowledges support from MES RF project 14.Z50.31.0007 "Lab.Astrophysics". REFERENCES Baumjohann, W., & Treumann, R. A. 1996, Basic space plasma physics, London: Imperial College Press Holzwarth, V., & Jardine, M. 2007, A&A, 463, 11 Johnstone, C. P., Gudel, M., Brott, I., & Luftinger, T. 2015a, Bhardwaj, A., Elsner, R. F., Randall Gladstone, G., et al. 2007, A&A, submitted Planet. Space Sci., 55, 1135 Johnstone, C. P., Gudel, M., Luftinger, T., Toth, G., & Brott, I. Branduardi-Raymont, G., Bhardwaj, A., Elsner, R. F., & 2015b, A&A, submitted Rodriguez, P. 2010, A&A, 510, A73 Christensen, U. R., Holzwarth, V., & Reiners, A. 2009, Nature, Kashyap, V. L., Drake, J. J., & Saar, S. H. 2008, ApJ, 687, 1339 Khodachenko, M. L., Alexeev, I., Belenkaya, E., et al. 2012, ApJ, 457, 167 744, 70 Collier, M. R., Snowden, S. L., Sarantos, M., et al. 2014, Journal Kislyakova, K. G., Holmstrom, M., Lammer, H., Odert, P., & of Geophysical Research (Planets), 119, 1459 Cravens, T. E. 2002, Science, 296, 1042 Cravens, T. E., Clark, J., Bhardwaj, A., et al. 2006, Journal of Geophysical Research (Space Physics), 111, 7308 Cravens, T. E., Robertson, I. P., & Snowden, S. L. 2001, J. Geophys. Res., 106, 24883 Khodachenko, M. L. 2014, Science, 346, 981 Lisse, C. M., Cravens, T. E., & Dennerl, K. 2004, X-ray and extreme ultraviolet emission from comets, ed. G. W. Kronk, 631 -- 643 Mazeh, T., Naef, D., Torres, G., et al. 2000, ApJ, 532, L55 Robertson, I. P., & Cravens, T. E. 2003, Geophys. Res. Lett., 30, Cravens, T. E., Waite, J. H., Gombosi, T. I., et al. 2003, Journal 1439 of Geophysical Research (Space Physics), 108, 1465 Dennerl, K. 2002, A&A, 394, 1119 Dennerl, K., Burwitz, V., Englhauser, J., Lisse, C., & Wolk, S. 2002, A&A, 386, 319 Greenwood, J. B., Williams, I. D., Smith, S. J., & Chutjian, A. 2001, Physica Scripta Volume T, 92, 150 Sanz-Forcada, J., Ribas, I., Micela, G., et al. 2010, A&A, 511, L8 Suzuki, T. K., Imada, S., Kataoka, R., et al. 2013, PASJ, 65, 98 Telleschi, A., Gudel, M., Briggs, K., et al. 2005, ApJ, 622, 653 van Leeuwen, F. 2007, A&A, 474, 653 Vidotto, A. A., Jardine, M., & Helling, C. 2011, MNRAS, 414, 1573 Griessmeier, J.-M., Stadelmann, A., Penz, T., et al. 2004, A&A, Wood, B. E., Muller, H.-R., Zank, G. P., Linsky, J. L., & 425, 753 Gudel, M. 2007, Living Reviews in Solar Physics, 4, 3 Gunell, H., Holmstrom, M., Kallio, E., Janhunen, P., & Dennerl, K. 2004, Geophys. Res. Lett., 31, 22801 Holmstrom, M., Barabash, S., & Kallio, E. 2001, Geophys. Res. Lett., 28, 1287 Redfield, S. 2005, ApJ, 628, L143
1710.03250	1	1710	2017-10-09T18:06:59	Coupling SPH and thermochemical models of planets: Methodology and example of a Mars-sized body	[ "astro-ph.EP" ]	Giant impacts have been suggested to explain various characteristics of terrestrial planets and their moons. However, so far in most models only the immediate effects of the collisions have been considered, while the long-term interior evolution of the impacted planets was not studied. Here we present a new approach, combining 3-D shock physics collision calculations with 3-D thermochemical interior evolution models. We apply the combined methods to a demonstration example of a giant impact on a Mars-sized body, using typical collisional parameters from previous studies. While the material parameters (equation of state, rheology model) used in the impact simulations can have some effect on the long-term evolution, we find that the impact angle is the most crucial parameter for the resulting spatial distribution of the newly formed crust. The results indicate that a dichotomous crustal pattern can form after a head-on collision, while this is not the case when considering a more likely grazing collision. Our results underline that end-to-end 3-D calculations of the entire process are required to study in the future the effects of large-scale impacts on the evolution of planetary interiors.	astro-ph.EP	astro-ph	Coupling SPH and thermochemical models of planets: Methodology and example of a Mars-sized body G.J. Golabek1,2,, A. Emsenhuber3, M. Jutzi3, E.I. Asphaug4,5 & T.V. Gerya2 1 Bayerisches Geoinstitut, University of Bayreuth, Universitätsstrasse 30, 95440 Bayreuth, Germany. 2 ETH Zurich, Institute of Geophysics, Sonneggstrasse 5, 8092 Zürich, Switzerland. 3 University of Bern, Physics Institute, Space Research and Planetary Sciences, Center for Space and Habitability, Gesellschaftsstrasse 6, 3012 Bern, Switzerland. 4 Arizona State University, School of Earth and Space Exploration, PO Box 876004, Tempe, AZ 85287, USA. 5 University of Arizona, Lunar and Planetary Laboratory, 1269 E. University Blvd, Tucson, AZ 85721, USA. Corresponding author. G.J. Golabek ([email protected]) Accepted for publication in Icarus. Abstract Giant impacts have been suggested to explain various characteristics of terrestrial planets and their moons. However, so far in most models only the immediate effects of the collisions have been considered, while the long-term interior evolution of the impacted planets was not studied. Here we present a new approach, combining 3-D shock physics collision calculations with 3-D thermochemical interior evolution models. We apply the combined methods to a demonstration example of a giant impact on a Mars-sized body, using typical collisional parameters from previous studies. While the material parameters (equation of state, rheology model) used in the impact simulations can have some effect on the long- term evolution, we find that the impact angle is the most crucial parameter for the resulting spatial distribution of the newly formed crust. The results indicate that a dichotomous crustal pattern can form after a head-on collision, while this is not the case when considering a more likely grazing collision. Our results underline that end-to-end 3-D calculations of the entire process are required to study in the future the effects of large-scale impacts on the evolution of planetary interiors. 1. Introduction Towards the end of terrestrial planetary accretion, giant impacts are inevitable [Wetherill, 1985; Melosh, 1990; Canup & Asphaug, 2001]. It has been suggested that they can explain various characteristics of the terrestrial planets like Mercury's anomalously thin silicate mantle [Benz et al., 1988, 2007; Asphaug and Reufer, 2014], the retrograde spin of Venus [Alemi and Stevenson, 2006], the origin of the Earth's Moon [Hartmann and Davis, 1975; Cameron and Ward, 1976; Canup and Asphaug, 2001; Canup, 2004, 2012; Ćuk and Stewart, 2012; Reufer et al., 2012] and the formation of the martian dichotomy [Wilhelms and Squyres, 1984; Frey and Schultz, 1988; Andrews-Hanna et al., 2008; Nimmo et al., 2008; Marinova et al., 2008, 2011]. For the case of Mars, numerical model results suggest that the northern lowlands of Mars could correspond to a giant impact basin formed after primordial crust formation. However in these models only the impact process and its immediate aftermath are considered, while the longer-term evolution featuring post-collision crust formation is not studied. Thermochemical codes consider the physical evolution of the planetary interior, and are better suited than impact models to study the long-term evolution of planetary interiors following a major collision. For the case of the martian dichotomy these endogenic models [e.g., Weinstein, 1995], suggest a degree-1 mantle upwelling underneath the southern highlands [Zhong and Zuber, 2001; Roberts and Zhong, 2006; Zhong, 2009; Keller and Tackley, 2009; Šramek and Zhong, 2010, 2012]. These models rely on a high Rayleigh number and a particular viscosity profile to form a low degree convective planform within the geological constraints for Mars dichotomy formation. Also for the case of the martian dichotomy formation, a hybrid exogenic-endogenic approach was suggested [Reese and Solomatov, 2006, 2010; Reese et al., 2010; Golabek et al., 2011], proposing an impact- induced regional magma ocean and subsequent superplume in the southern hemisphere. These thermochemical simulations usually rely on highly simplified impact models, for instance a uniform or localized deposition of energy or matter representative of the collision [e.g. Wilhelms and Squyres, 1984]. Although a simplified approach might be suitable for the limited case of a head-on impact, it is well established that impacts at 45 degrees are most likely [Gilbert, 1893; Shoemaker, 1962]. This is true for cratering and for giant impacts, however for cratering, the infinite target means that impact angle does not matter very much. No matter what the angle, the projectile can not find its way beyond the target, and there is no abrupt transition to 'grazing' behavior until impact angles of around 75 degrees from vertical [Melosh, 1989]. However for giant impacts and other similar-sized collisions [Asphaug, 2010], even a close-to-head-on impact (less than 20 degrees) can allow for lopsided evolution of material. Thus, the combined effects of gravity, angular momentum, and linear downrange momentum require 3-D models. Thus to better understand the influence of giant impacts on the early evolution of terrestrial planets, methodological improvements are necessary, ideally using the advantages of both methods to overcome the current model limitations. In a recent study, results of 2-D axisymmetric impact calculations were used as input in thermochemical evolution models [Rolf et al., 2017]. We present here a new approach using the results of 3-D large-scale collision models performed with a shock physics code to be used as initial conditions in a thermochemical code and present demonstration calculations. In section 2 we introduce the methods used by both codes, one for giant impacts and one for the thermochemical evolution. In section 3 we describe the model scenario of our demonstration calculation, which we use to test the coupled method. In section 4 we discuss the effect of parameters like the equation of state, the rheology and the transfer time on the post-impact crust formation. The two final sections are devoted to discussion and outlook. 2. Methodology 2.1 SPH code We use a smoothed particle hydrodynamics (SPH) code [Benz and Asphaug, 1994; Reufer et al. 2012; Jutzi et al., 2013] to model the collision. This code is based on the SPHLATCH code developed by Reufer et al. [2012] and includes self-gravity, a newly implemented strength model and various equations of state (EoS). The details of the strength and EoS models are described in Emsenhuber et al., in press. Here we only give a short summary of the basic method. SPH uses a Lagrangian representation where material is divided into particles. Quantities are interpolated ('smoothed') over a certain length by summing over surrounding particles (called neighbours) using a kernel function 𝐵(𝑥⃗) = ∑ 𝐵𝑖 𝑖 𝑊(𝑥⃗ − 𝑥⃗𝑖, ℎ𝑖) (1) where 𝑥⃗𝑖, ℎ𝑖 and 𝑉𝑖 are the position, smoothing length, and volume of particle i, respectively. 𝐵𝑖 represents the quantity (field variable) to be interpolated and 𝑊(𝑥⃗, ℎ) is the kernel. This interpolation scheme is used in SPH to solve the relevant differential equations [see Emsenhuber et al., in press]. SPH is an adaptive resolution technique, thus large volumes of space can be modelled economically. Indeed 'space' has no particles and costs no resolution. The method is therefore quite popular in three-dimensional (3-D) planetary collisional studies [e.g. Jutzi et al., 2015], and includes good equations of state and treatments of shock and self-gravity. But being a 2nd order method limited to the Courant timescale (that is, the sound-crossing time of a particle smoothing length), it is not suited to model the longer-term post-impact evolution. Body accelerations are the result of the pressure gradient, and for this SPH uses the pressure as computed from an equation of state (EoS; see below) which is a function P(ρ,u) of density ρ and internal energy u. For solid materials the pressure gradient is generalized into a stress tensor [Benz and Asphaug, 1994]. We use a Drucker- Prager-like yield criterion [Collins et al., 2004, Jutzi, 2015] with yield strength 𝜎𝑖 of intact material, which is a function of temperature [Collins et al., 2004]. A more sophisticated model for geological materials, which includes also a tensile fracture and a porosity model, and its application in SPH (using a different code) is described in Jutzi [2015]; for problems at the global-scale of a Mars-sized body these aspects can be ignored. As for the EoS we use either Tillotson [Tillotson, 1962] or ANEOS [Thompson and Lauson, 1972; Melosh, 2007]. The Tillotson equation of state provides both the pressure P and the speed of sound vs as output, but it does not give the temperature T, so as an approximation we use the internal energy u as a proxy for temperature by dividing by the heat capacity cp. However it should be noted that eq. (2) does not take into account that heat capacity is temperature-dependent. 𝑇 = 𝑢/𝑐𝑝 (2) For a more physical computation of temperature, and to allow for phase transitions, which do not occur in the Tillotson EoS, we use ANEOS [Thompson and Lauson, 1972] for iron and M-ANEOS [Melosh, 2007] for silicates. These equations of state provide numerous output variables including the temperature and phase information (e.g. melt and vapor fraction). Depending on the parameters used in the equation of state, this phase information can be given in different ways. For iron, it can be used to infer the melting temperature at a given pressure. However, this is not the case for silicates, for which we apply the same procedure as used in [Senft and Stewart, 2009] to obtain the melting temperature at a given pressure. Numerous tests have been made comparing SPH with other codes in the area of impact contact and compression and the early stages of dynamical evolution. Canup et al. [2013] find that SPH and CTH (a popular grid-based shock physics hydrocode) both give similar answers for the amount of mass and momentum ejected into a protolunar disk for a Moon-forming giant impact scenario, although varying in detailed aspects such as clumping. Pierazzo et al. [2008] make comparisons and find overall agreement for the early impact ejection phase of large-scale impact cratering calculations. 2.2 I3ELVIS code The study of giant impacts over the past 30 years has depended on hydrocode methods such as SPH, whose methods are generally familiar to the planetary science community. Global thermochemical methods are required to study the post-collision evolution of a planetary object after a giant impact, so we describe that methodology in more detail. We consider 3-D creeping flow using the Boussinesq approximation, in which both thermal and chemical buoyancy forces are included. For this purpose we performed 3-D simulations using the code I3ELVIS by applying the 'spherical-Cartesian' methodology [see also Gerya and Yuen, 2007]. The methodology combines finite differences on a fully staggered rectangular Eulerian grid with a Lagrangian marker-in-cell technique for solving the momentum, continuity and temperature equations. In order to model a planetary body on a Cartesian grid, the code calculates self-consistently the gravitational field for a self-gravitating planetary body. In detail, the gravitational potential can be described using the Poisson equation 𝜕2𝜙 𝜕𝑥2 + 𝜕2𝜙 𝜕𝑦2 + 𝜕2𝜙 𝜕𝑧2 = 4𝜋𝐺𝜌 (3) where 𝜙 is the gravitational potential and G is the gravitational constant. Knowing the gravitational potential, the components of the gravity vector can be computed as follows: 𝑔𝑥 = − 𝜕𝜙 𝜕𝑥 ; 𝑔𝑦 = − 𝜕𝜙 𝜕𝑦 ; 𝑔𝑧 = − 𝜕𝜙 𝜕𝑧 (4) We assume the Boussinesq approximation in 3-D geometry: 𝜕𝑣𝑥 𝜕𝑥 + 𝜕𝑣𝑦 𝜕𝑦 + 𝜕𝑣𝑧 𝜕𝑧 = 0 (5) where vx, vy and vz are the x, y and z component of the velocity vector, respectively. The local density ρ depends explicitly on temperature T, pressure P, composition c and melt fraction φ. The density of iron and both solid and molten silicates varies with P-T conditions according to the relation: where ρ0 is the density of the distinct material at T0 = 298.15 K and P0 = 105 Pa, α is the thermal expansion 𝜌 = 𝜌0[1 − 𝛼(𝑇 − 𝑇0)] ∙ [1 + 𝛽(𝑃 − 𝑃0)] (6) coefficient and β is the compressibility coefficient. For partially molten silicates we take additionally into account that the density will vary with the melt fraction φ as follows [see Burg and Gerya, 2005]: 𝜌𝑒𝑓𝑓 = 𝜌𝑆𝑖−𝑠𝑜𝑙 − 𝜑[𝜌𝑆𝑖−𝑠𝑜𝑙 − 𝜌𝑆𝑖−𝑙𝑖𝑞](7) where ρSi-sol and ρSi-liq are the pressure and temperature-dependent densities of the solid silicates and liquid silicates, respectively. For this purpose we use a batch-melting model assuming a peridotite composition. Pressure-dependent parameterizations for both the solidus and liquidus temperatures, Tsol and Tliq of peridotite are applied [Herzberg et al. 2000; Trønnes and Frost, 2002; Wade and Wood 2005]: Tsol [K] = 1416.2 + 58.3 P [GPa] + 52.3 P2 [GPa] - 16.3 P3 [GPa] + 2.3 P4 [GPa] - 0.2 P5 [GPa] + 8·10-3 P6 [GPa] - 2·10-4 P7 [GPa] + 2·10-6 P8 [GPa] (8) Tliq [K] = 1973 + 28.57 P [GPa] (9) For T ≤ Tsol, the silicate melt fraction φ is zero, for T ≥ Tliq, it is equal to 1. In the intermediate temperature range Tsol < T < Tliq, the melt fraction is assumed to increase linearly with temperature according to the following relation [Burg and Gerya, 2005]: 𝜑 = 𝑇−𝑇𝑠𝑜𝑙 𝑇𝑙𝑖𝑞−𝑇𝑠𝑜𝑙 (10) Both consumption and release of latent heat due to melting and freezing of silicates are taken into account. The 3-D Stokes equations for creeping flow take the form: 𝜕𝜎𝑥𝑥 𝜕𝑥 + 𝜕𝜎𝑥𝑦 𝜕𝑦 + 𝜕𝜎𝑥𝑧 𝜕𝑧 − 𝜕𝜎𝑦𝑥 𝜕𝑥 + 𝜕𝜎𝑦𝑦 𝜕𝑦 + 𝜕𝜎𝑦𝑧 𝜕𝑧 − 𝜕𝜎𝑧𝑥 𝜕𝑥 + 𝜕𝜎𝑧𝑦 𝜕𝑦 + 𝜕𝜎𝑧𝑧 𝜕𝑧 − 𝜕𝑃 𝜕𝑥 𝜕𝑃 𝜕𝑦 𝜕𝑃 𝜕𝑧 = −𝑔𝑥𝜌 (11) = −𝑔𝑦𝜌 (12) = −𝑔𝑧𝜌 (13) where σij represents the components of the deviatoric stress tensor and P is the total pressure, which includes both dynamic and lithostatic components. We have adopted a Lagrangian frame in which the energy conservation equation takes the following form [Gerya and Yuen, 2003, 2007]: 𝜌𝑐𝑝 ( 𝐷𝑇 𝐷𝑡 ) = − 𝜕𝑞𝑥 𝜕𝑥 − 𝜕𝑞𝑦 𝜕𝑦 − 𝜕𝑞𝑧 𝜕𝑧 + 𝐻𝑟 + 𝐻𝑠 + 𝐻𝐿 (14) where D/Dt is the substantive time derivative, Hr is the radiogenic heating term, Hs is the shear heating term, HL is the latent heating term and qi is a heat flux component. The shear heating term is given by 𝐻𝑠 = 𝜎𝑥𝑥𝜀𝑥𝑥 + 𝜎𝑦𝑦𝜀𝑦𝑦 + 𝜎𝑧𝑧𝜀𝑧𝑧 + 2𝜎𝑥𝑦𝜀𝑥𝑦 + 2𝜎𝑥𝑧𝜀𝑥𝑧 + 2𝜎𝑦𝑧𝜀𝑦𝑧 (15) where 𝜀𝑖𝑗 are the components of the strain-rate tensor defined as The heat flux qi is defined as 𝜀𝑖𝑗 = 1 2 𝜕𝑣𝑖 ( 𝜕𝑥𝑗 + 𝜕𝑣𝑗 𝜕𝑥𝑖 ) (16) 𝑞𝑥 = −𝑘 𝜕𝑇 𝜕𝑥 , 𝑞𝑦 = −𝑘 𝜕𝑇 𝜕𝑦 , 𝑞𝑧 = −𝑘 𝜕𝑇 𝜕𝑧 (17) where k is the thermal conductivity. We employ a viscous constitutive relationship between stress and strain-rate with η representing the effective viscosity: 𝜎𝑖𝑗 = 2𝜂𝜀𝑖𝑗 (18) We use for solid silicates a viscosity η, which depends on temperature T, pressure P and strain rate 𝜀 defined in terms of deformation invariants [Ranalli, 1995] as: 𝜂𝑒𝑓𝑓 = 𝐴1/𝑛𝑒𝑥𝑝 ( 𝐸𝑎+𝑃𝑉𝑎 𝑛𝑅𝑇 (1−𝑛)/𝑛 ) 𝜀𝐼𝐼 (19) where 𝜀𝐼𝐼 = √ 1 2 𝜀𝑖𝑗𝜀𝑖𝑗 is the second invariant of the strain-rate tensor and A, Ea and n are the pre-exponential parameter, the activation energy and the power law coefficient, respectively. R is the gas constant and Va is the activation volume. For solid silicates, this ductile rheology is combined with a brittle rheology to yield an effective viscoplastic rheology. For this purpose the Mohr-Coulomb [e.g., Ranalli, 1995] and Peierls [e.g., Kameyama et al., 1999] yield criteria are simultaneously implemented by limiting the creep viscosity as follows: and for 𝐶 + 𝑃𝑓 ≥ 𝜎𝑃 𝜂𝑒𝑓𝑓 ≤ 𝐶+𝑃𝑓 2𝜀 𝐼𝐼 (20) 𝜂𝑒𝑓𝑓 ≤ 𝜎𝑃 2𝜀 𝐼𝐼 (21) where C is the cohesion, f is the internal friction coefficient and σP is the Peierls stress limit. The presence of a melt fraction φ > 0.1 has for silicate material an additional influence on the effective viscosity as given here [Pinkerton and Stevenson, 1992] 𝜂𝑒𝑓𝑓 = 𝜂𝑆𝑖−𝑙𝑖𝑞𝑒𝑥𝑝 {[2.5 + ( 0.48 1−𝜑 𝜑 ) ] (1 − 𝜑)} (22) where ηSi-liq is the viscosity of molten silicates, which we assume to be a constant. 2.2.1 Effective thermal conductivity of largely molten silicates: Within a narrow silicate melt fraction range, the silicate behaviour undergoes a transition from a solid-like material to a low viscosity crystal suspension [see Costa et al., 2009; Solomatov, 2015]. Thus, the viscosity of largely molten silicates is ηSi-liq ~ 10-4-102 Pa s [Rubie et al., 2003; Liebske et al., 2005; Bottinga and Weill, 1972]. Hence, due to the low viscosity, both the Rayleigh Ra and the Nusselt number Nu will be high and cooling will be a very efficient process. Due to numerical restrictions, the lower cut-off viscosity in the numerical model is ηnum = 1018 Pa s, many orders of magnitude higher than realistic viscosities for largely molten silicates. It was suggested that the heat flux from a magma ocean [e.g. Solomatov, 2015] can be described using the so-called soft turbulence model [Kraichnan, 1962; Siggia, 1994]. In this model the expected convective heat flux q is given as: where the Rayleigh number Ra is defined as 𝑞 = 0.089 𝑘(𝑇−𝑇𝑠𝑢𝑟𝑓) 𝐿 𝑅𝑎1/3 (23) 𝑅𝑎 = 𝛼𝑔(𝑇−𝑇𝑠𝑢𝑟𝑓)𝜌𝑒𝑓𝑓 2 𝑐𝑃𝐷3 𝑘𝜂𝑆𝑖−𝑙𝑖𝑞 (24) Tsurf is the surface temperature and D is the depth of the magma ocean. Depending on the actual silicate melt viscosity in the numerical model ηnum, one can estimate an increased effective thermal conductivity keff by using the theoretically expected heat flux from a low viscosity magma ocean q from eq. (23). This effective thermal conductivity can simulate the heat flux of a medium with a realistic magma ocean viscosity [Zahnle et al., 1988; Tackley et al., 2001; Hevey and Sanders, 2006; Golabek et al., 2011, 2014], despite our numerical limitations. Combining eq. (23) and (24) this can be done using the following expression for the effective thermal conductivity keff: 𝑘𝑒𝑓𝑓 = ( 3 2 ) 𝑞 0.089 1 2 (𝑇−𝑇𝑠𝑢𝑟𝑓) 𝜌𝑒𝑓𝑓 1 − 2 (25) ) ( 𝛼𝑔𝑐𝑃 𝜂𝑛𝑢𝑚 2.2.2 Crust formation Partial melting of the mantle, melt extraction and percolation toward the bottom of the forming basaltic crust is implemented in a simplified manner. According to our model, mafic magma added to the crust is balanced by melt production and extraction in the mantle. However, melt percolation is not modelled directly and is considered to be nearly instantaneous. The standard (i.e. without melt extraction) volumetric degree of mantle melting φ changes with pressure and temperature according to the linear batch melting model (see eq. 10). Lagrangian markers track the amount of melt extracted during the evolution of each numerical experiment. The total amount of melt, φtot for every marker takes into account the amount of previously extracted melt and is calculated as 𝜑𝑡𝑜𝑡 = 𝜑 − ∑ 𝜑𝑒𝑥𝑡 𝑁 (26) where ∑ 𝜑𝑒𝑥𝑡 𝑁 is the total melt fraction extracted during the previous N extraction episodes. The rock is considered to be non-molten (refractory) when the extracted melt fraction is larger than the current one (i.e. when ∑ 𝜑𝑒𝑥𝑡 𝑁 > 𝜑). Since the extracted melt fraction propagates much faster than the rocks deform [Condomines et al., 1988], melts produced at depth are moved towards the surface and added to the bottom of the forming crust. In order to ensure melt volume conservation and account for mantle compaction and subsidence in response to the melt extraction, melt addition to the bottom of the crust is performed at every time step by converting the shallowest markers of mantle into crustal markers. The local volume of these new crustal markers matches the local volume of extracted melt computed for the time step. Basaltic melts are assumed to be only extracted from relatively shallow (≤300 km depth) mantle regions with low degree of melting (φ≤0.2). This corresponds roughly to the pyroxene fraction of a fertile mantle, being the main component in the basaltic to andesitic composition of crustal material [McKenzie and Bickle, 1988]. For simplicity, we do not take partitioning of heat producing elements into the crustal material into account. 3. Model setup We now apply the combined methods to the example of a giant impact on a Mars-sized body, using typical collisional parameters from previous studies. We perform a series of 8 SPH collision models, using a Mars- mass target body (R = 3436 km) and an impactor with 1000 km radius made of silicates and iron, with no initial crust on both bodies. The collision occurs at 5 Myr after CAI formation, within the constraints for the last giant impact onto a smaller planetary embryo [Nimmo & Kleine, 2007; Mezger et al. 2013; Morishima et al. 2013]. For each combination of physical properties (solid and fluid rheologies with either Tillotson or ANEOS EoS) we test different impact angles of 0° (head-on collision) and 45° (grazing collision), as defined in [Asphaug, 2010]. For details see Table 1. We use in all our models a collision at mutual escape velocity [Asphaug, 2010]. With a value ~3·10-3 for the ratio of kinetic impactor energy over gravitational binding energy, this model setup is expected to cause significant post-impact melting [Reese and Solomatov, 2006], resulting in significant crust formation following the giant impact. The simulations used here are a subset of the suite of calculations performed in [Emsenhuber et al., in press]. The target and impactor both start with a Mars-like internal structure with an iron core radius half of the body radius. For simulations using Tillotson EoS, we start with an isothermal mantle with an initial temperature close to the surface solidus temperature of peridotite (TSi = 1500 K) [Reese and Solomatov, 2010]. The core temperature is initially set for both objects to a constant value of TFe = 1800 K, which ensures that at the pressure conditions at the center of the target body the assumed eutectic Fe-FeS is molten [Chudinkovich and Boehler, 2007]. For the simulations with ANEOS, an isentropic profile is used for which the temperature at the center of the core is 1800 K and 1500 K for the silicate at the core-mantle boundary. We begin setting up self-gravitating, hydrostatically-equilibrated planetary objects, starting with one dimensional (1-D) spherically symmetric bodies modelled using a Lagrangian hydrocode [Benz, 1991], with the same EoS model as for SPH. This profile is evolved by computing the force balance between self-gravity and pressure (including a damping term), until hydrostatic equilibrium is reached so that radial velocities are small (less than 1% of the escape velocity). Afterwards we transfer the 1-D radial profile onto SPH particles that are placed onto a 3-D lattice. Parameters of each particle are copied from the 1-D profile according to the radius. As particles are equally spaced on the lattice, variation of density is taken into account by adjusting the particle mass. The spherical SPH bodies are then also evolved in a last initializing step to reach hydrostatic equilibrium and negligible radial velocities. Thus the SPH simulations start with two relaxed, differentiated spherical planetary objects. To model the collision they start at an initial distance of several radii, so that during the approach they begin to deform tidally prior to the collision, which has an effect on the collision outcome. The SPH simulations are performed with a resolution of about one million particles for the target. The number of particles for the impactor is scaled according to the mass ratio between the two planetary objects, so that particle spacing h is approximately constant, for the greatest numerical accuracy during the collision. The corresponding smoothing length is then approximately 60 km for both bodies. For more details of the SPH calculations see Emsenhuber et al., in press. The results from SPH are transferred to I3ELVIS at earliest after the collision when (i) the seismic shock waves in the target body have decayed and (ii) after the bulk of ejecta material has either fallen back to the surface of the target body or has escaped (see section 4.1.2). For most cases the transfer time is 18 hours after the collision, several times the self-gravity timescale. To study the sensitivity of the post-collision model evolution to the transfer time, we performed for the reference case (ANEOS EoS, grazing collision) additional thermochemical calculations starting at 8 and 12 hours after the collision (see also Table 1). The temperature and composition results are interpolated (using the SPH Kernel function as in eq. (1)) onto a Eulerian equally spaced grid of (10,000 km)3 size centered on the middle of the post-impact target body. The resulting grid spacing is 20 km, which corresponds to a grid resolution of 5013. The computational domain is filled with 5.24·108 randomly distributed Lagrangian markers. The informations are then transferred from these nodal points to the markers using a bilinear interpolation scheme described in detail elsewhere [Gerya and Yuen, 2003, 2007]. To avoid unrealistic silicate melt densities, we introduce in the I3ELVIS models a lower density cut-off at 2500 kg/m3. The I3ELVIS model requires some cleanup of the SPH input. Thus material at a distance of more than 3700 km from the planetary center is considered to be remnant ejecta material and is therefore deleted. The rest of the box is filled with sticky air material of nearly zero density, constant viscosity (sa = 1018 Pa s) and constant temperature of Tsa = 220 K. This layer represents an infinite reservoir to absorb heat released from the planetary body [Golabek et al., 2014; Tkalcec et al., 2013] and ensures a free surface of the planetary body [Schmeling et al., 2008; Crameri et al., 2012]. For numerical reasons, the viscosity range is limited to six orders of magnitude. The silicate viscosity is cut off at lower and upper limits of 1018 and 1024 Pa s, whereas the iron viscosity is kept constant at 1018 Pa s. A similar value for the maximum silicate viscosity (1024 - 1028 Pa s) was suggested before [Karato and Murthy, 1997]. It should be noted that the iron viscosity in the model (Fe = 1018 Pa s) is, for numerical reasons, many orders of magnitude higher than its physical value, which is expected to range from 10 -2 Pa s (in the liquid state) [Rubie et al., 2015] up to 1012 Pa s in the solid state under high temperature and pressure conditions [Yunker and Van Orman, 2007]. For all models, we consider time-dependent radioactive heating by both short- (26Al, 60Fe) and long-lived (40K, 235U, 238U, 232Th) radiogenic isotopes. In the early solar system, 26Al is by far the dominant radioactive energy source and the initial 26Al/27Al is taken as 5.85·10-5 [Thrane et al., 2006], this value representing an upper limit for the abundance of 26Al [see Jacobsen et al., 2008]. We use a wet olivine rheology [Ranalli, 1995]. This is reasonable as olivine represents the majority of the martian mantle composition and is weaker relative to pyroxenes, thus controlling mantle deformation [Mackwell, 1991]. The Peierls stress limit P employed for wet olivine rheology [Katayama and Karato, 2008] is 2.9 GPa. As suggested for a Mars-like mantle composition, we apply a density crossover between solid and molten silicates at 600 km depth [Suzuki et al., 1998]. For all other physical parameters employed see Table 2. 4. Results 4.1 SPH results In this section, we briefly summarize the results of the SPH impact simulations. A detailed description of the full set of simulations, as well as a discussion of other physical and numerical effects is given in Emsenhuber et al. (in press). 4.1.1 Material and temperature distribution resulting from the SPH simulations A snapshot of the grazing impact simulations at the time of the first contact of the impactor with the target is shown in figure 1 for the two different material rheologies. We note that with fluid rheology, the impactor is tidally deformed prior to the collision in a more pronounced way than with solid rheology. Qualitative differences in the collision outcome due to the equation of state and the rheology can be seen in figure 2, which shows for the two cases the material and temperature distributions (at t = 18 h after the impact) resulting from the impact. There are two main effects of using solid rheology observable in the final results: the impactor's material location and the heat distribution around the impact zone. Considering material location, about 25 wt% of impactor's core remains in the upper part of the target's mantle under solid rheology whereas for fluid rheology this is insignificant. Also the distribution of the mantle of the impactor within the target is quite different. For solid rheology, about half of the impactor's mantle ends up in the interior while the remaining impactor mantle material is close to the surface. In contrast, with fluid rheology, about 10 wt% of the impactor's mantle lies close to the core-mantle boundary, very little across the rest of the target's mantle, and the majority is located close to the surface. Assuming solid rheology the heat distribution is focussed around the impact zone; in contrast, mostly the impactor material gets heated in models assuming a fluid rheology and the temperature distribution is less localised. This results in initially less silicate melt being present in all fluid rheology cases than in cases assuming a solid rheology. The differences due to the different equations of state are more subtle. Material with high internal energy has a lower temperature with ANEOS compared to Tillotson. While ANEOS includes the latent heat of vaporisation, we compute the temperature from the specific energies in the runs with the Tillotson EoS using eq. (2) that does not take this into account. However, this affects only a small fraction of material, which is located close to the surface. This issue is discussed in more detail in Emsenhuber et al., in press. In future simulations with much higher surface resolution the effect may be significant. 4.1.2 Re-impacting ejecta At the standard transfer time (t = 18 h) at which we switch from the SPH impact simulations to the interior evolution code I3ELVIS, there is still some ejecta material in orbit, which will potentially get accreted onto the target body later on. We use the following procedure to compute the amount of ejecta remaining to re-impact the Mars-sized target body as a function of time: For each SPH particle it is checked whether it is (1) bound to the target body, (2) on a crossing orbit and (3) outside a sphere with a radius of 3700 km around the center of mass of the target body. Figure 3 shows that within ~18 hours after the collision the remaining ejecta mass in orbit drops to values below <10-3 MMars. To estimate the potential thermal effect of this remaining ejecta on the global evolution, we can calculate its thickness when distributed into a global surface layer of average thickness d by 3 𝑑 = √ 3(𝑀+𝑚) 4𝜋𝜌 − 𝑅 (27) where M is the mass of the target planet, m is the mass of the ejecta layer and 𝜌 is the average density of the planet and R its average radius without the ejecta blanket. For simplicity we assume that the remaining orbiting ejecta and the planet have the same average density. However it should be kept in mind that local effects of falling ejecta not considered here may be important. Using the characteristic diffusion time scale t ~ d2/, where  is the thermal diffusivity, we can estimate the minimum cooling timescale of this remaining ejecta material (see Figure 4). For remaining ejecta masses <10-3MMars the characteristic diffusion timescale becomes very short compared to the crust formation timescale (see section 4.2), therefore we expect that the disregarded ejecta mass will not affect the evolution of the post- impact planet significantly. 4.2 I3ELVIS results As described in section 3, for each impact simulation (see table 1) the SPH output is transferred into the I3ELVIS code used to study the post-impact interior evolution. In these longer-term calculations, the crust formation is computed by the procedure outlined in section 2.2.2. The resulting crustal distributions at t ~ 5·105 yr after the collision are displayed in Figures 5-7. Since the early post-collision evolution involves the sinking of iron diapirs through the target's mantle, additional mantle melting occurs due to the release of potential energy. For the head-on collisions the bulk of the sinking iron is concentrated in one iron diapir, as a result most of the heat release occurs in the impact region. For the grazing collisions the post-impact iron distribution is more complex and various-sized iron diapirs are initially distributed throughout a larger portion of the target mantle (see Figure 2). Due to the power law rheology the sinking velocity of the iron diapirs is faster than in a Newtonian medium [e.g. Weinberg & Podladchikov, 1994]. Thus merging of the iron diapirs with the target's core occurs in all calculations on a timescale of ~104 yrs. Due to the shear heating related to the sinking of the iron diapirs hot wakes form [Ziethe and Spohn, 2007], evolving on the longer-term into thermal plumes. Due to the cooling of the silicate melt a basaltic crust forms, however delamination from the bottom occurs repeatedly. This delaminating crust sinks towards the core-mantle boundary and limits the crustal thickness. To determine the time-dependent total volume and distribution of crust, we consider only stable crustal material, meaning that disconnected crustal drips are disregarded. While we observe differences in the time evolution and the final volume of stable crust, all models show that the amount of stable crustal volume stabilizes after several hundred thousand years of evolution (see Figure 8). However, we observe a factor of two difference in the volume of stable crustal material for the different calculations, especially when considering the grazing collisions using Tillotson EoS with different rheologies. Figure 5 shows the cumulative distribution of the crustal thickness for the head-on and grazing impacts at ~5·105 yr after the collision, respectively. A comparison of the results for the two different EoS is shown as well. In Figures 6 and 7, the corresponding crustal thickness maps are displayed, showing the spatial distribution for each case. Overall, the main differences in the spatial crustal pattern result from the impact angle. This effect can be clearly seen in the crustal thickness maps, which for head-on impacts show very different patters than for grazing impacts. In all cases considered, head-on impacts result in a hemispherical crustal dichotomy (see Figure 6). On the other hand the crustal pattern resulting from the grazing impacts is less consistent (see Figure 7). As above, the differences between the runs with fluid and with solid rheology are generally more pronounced in the cases using Tillotson EoS (see Figure 5). These variations appear to mostly emerge from the different distributions of hot material just after the collision. As mentioned above a larger fraction of the silicates is molten in simulations with solid SPH rheology (coll02; see table 1), preferentially resulting in the formation of more crust in the long-term. However, the results for fluid rheology (coll08; see table 1) show at 300-1000 km depth the presence of hotter (~300 K) and more abundant (>5 times larger volume) impactor-derived melt in the mantle than in cases assuming a solid SPH rheology. For simulation coll02 this allows for fast cooling of the near surface layers and the crust tends to remain stable since no extended regions of still largely molten, thus low viscosity material, are present beneath the crust that would favour its destabilization. On the other hand, in simulation coll08 these regions of very hot silicate material are present for an extended time. Due to the presence of this low viscosity material beneath the newly formed crust the growth of crustal Rayleigh- Taylor instabilities is favoured. This is because the instability growth rate depends on the viscosity of the underlying material [Turcotte & Schubert, 2014]. This causes more crustal dripping events in simulation coll08 and more crustal material sinks into the mantle (see Figure 9). Comparing the two simulation outcomes it can be observed that only ~34% of the crustal material is stable at the end of simulation coll08 compared to ~56 % for simulation coll02. For our nominal case (coll04; see table 1), we performed additional longer-term evolution simulations using different SPH-I3ELVIS transfer times. Figure 10 shows the corresponding crustal thickness distributions. These look relatively similar for the three cases (transfer time of 8 h, 12 h, 18 h) investigated, indicating that as expected the timing of the data transfer does not have a major effect on the crustal distribution. Although this paper is only about the demonstration of the methodology, it is interesting to compare our results to the observed crustal volumes on a planet like Mars. At ~5·105 yr after the collision the volume of stable crust lies between 1.23·1018 m3 and 2.39·1018 m3, this corresponds to 17-34 % of the present-day volume of the martian crust [Taylor & McLennan, 2009]. Data indicate that at the end of the simulations 10-15% of the mantle material is partially molten and buoyant. This is related to the presence of hot plumes formed from the wakes left by the sinking iron diapirs. However most of the material in the shallow mantle is already highly depleted, therefore future crust formation is limited. Despite this some additional crust formation related to plumes can be expected to occur on the longer-term. 5. Discussion The presented numerical models employing SPH and thermochemical simulations performed in series show that while the total volume of crust formed after a collision is comparable for both head-on and grazing collisions, the spatial crustal distribution is strongly affected by the impact angle. While we observe variations in the distribution and total crustal volume, the present results show that for both EoS and both SPH rheologies employed the resulting post-collision crust displays a crustal dichotomy, when considering a head-on collision, while this is not the case for all models assuming the more likely grazing collision. Focussing on the outcome of the calculations based on SPH models with the more realistic solid rheology, the influence of the EoS on both the total volume and the spatial distribution of crustal material is more limited. The results also show that the rheology employed in the SPH models does have a significant influence on the immediate post-impact internal heat distribution and amount of silicate melt, but for the long-term crust formation process it is only a second-order effect. We note that its influence is more pronounced in the Tillotson cases. This might be related to the simplified treatment of silicate melting employed in models employing the Tillotson EoS. As discussed this can have an effect on the volume of stable crust since the presence of very hot material during the crust formation stage favours crustal dripping events. 5.1 Possible future improvements As mentioned above, some partial melt is still present at the end of the presented simulations. To study how much additional crust can be formed afterwards and whether all of the post-collision crust will be stable on the long-term additional calculations will be necessary in the future. A possible approach would be the transfer of the I3ELVIS results into the code StagYY [Tackley, 2008] suited to model the long-term evolution of terrestrial planets as done previously in 2-D geometry in Golabek et al. [2011]. Also it would be useful to study the effect of the impact angle on crust formation in more detail and determine where the transition between a dichotomous crust as observed for all head-on models and the more complex crustal pattern found for grazing collisions occurs in the parameter space. Also due to numerical limitations we have to use a lower cut-off for viscosity of all materials at 1018 Pa s. This limits the possibility of the post-collision magma ocean to spread due to isostatic relaxation [Reese and Solomatov, 2006]. A possible approach would require a reduced effective thermal conductivity, so magma ocean cooling would be extended, allowing for a longer spreading time. Additionally, a more realistic initial temperature profile with an increasing temperature with depth would allow for more efficient spreading since the viscosity of the deeper mantle would be reduced, thus isostatic relaxation could occur on a shorter timescale. Clearly these features have to be investigated in the future in more detail. 6. Conclusions The present demonstration calculations suggest that the coupled SPH-thermochemical approach in 3-D geometry, although computationally expensive, can help in the future to obtain a more realistic post-collision evolution of planetary interiors as these models are able to capture both the collision process, the following magma ocean cooling, crust formation and solid-state deformation processes in detail. Our results indicate that significant amounts of crustal material taking a dichotomous crustal pattern can form after a head-on collision, while this is not the case when considering a more likely grazing collision. This coupled approach can be applied to various planetary bodies to improve our understanding of their early evolution. Acknowledgements We thank two anonymous reviewers for detailed comments that helped to improve the manuscript considerably. Helpful discussions with Francis Nimmo and Don Korycansky on giant impacts are appreciated. A.E. acknowledges the financial support of the Swiss National Science Foundation under grant 200020_17246. This work has been carried out within the frame of the National Centre for Competence in Research PlanetS supported by the Swiss National Science Foundation. References Alemi, A., Stevenson, D.J., 2006. Why Venus has No Moon, American Astronomical Society, DPS meeting #38, #07.03; Bulletin of the American Astronomical Society, 38, 491. Andrews-Hanna, J.C., Zuber, M.T., Banerdt, W.B., 2008. The Borealis basin and the origin of the martian crustal dichotomy. Nature 453, 1212-1215. Asphaug, E., 2010. Similar-sized collisions and the diversity of planets. Chem. Erde - Geochemistry 70, 199- 219. Asphaug, E. & Reufer, A., 2014. Mercury and other iron-rich planetary bodies as relics of inefficient accretion. Nature Geosci. 7, 564-568. Benz, W., 1991. An Introduction to Computation Methods in Hydrodynamics. In: Loore, C.B. de, (Ed.), Lecture Notes in Physics 373, Late stages of stellar evolution, computational methods in astrophysical hydrodynamics. Springer, pp. 258-312. Benz, W., Cameron, A.G.W., Slattery, W.L., 1988. Collisional Stripping of Mercury's Mantle. Icarus 74, 516- 528. Benz, W., Asphaug, E., 1994. Impact simulations with fracture. I - Method and tests. Icarus 107, 98-116. Benz, W., Anic, A., Horner, J., Whitby, J.A., 2007. The Origin of Mercury. Space Sci. Rev. 132, 189-202. Bottinga, Y., Weill, D.F., 1972. The viscosity of magmatic silicate liquids: A model for calculation. Am. J. Sci. 272, 438-475. Burg, J.-P., Gerya, T.V., 2005. The role of viscous heating in Barrovian metamorphism of collisional orogens: Thermomechanical models and application to the Lepontine Dome in the Central Alps. J. Metamorp. Geol. 23, 75-95. Cameron, A.G.W., Ward, W.R., 1976. The origin of the Moon. Lunar Sci. 7, 120-122. Canup, R.M., Asphaug, E., 2001. Origin of the Moon in a giant impact near the end of the Earth's formation. Nature, 412, 708-712. Canup, R.M., 2004. Simulations of a late lunar-forming impact. Icarus 168, 433-456. Canup, R.M., 2012. Forming a Moon with an Earth-like Composition via a Giant Impact. Science 338, 1052- 1055. Canup, R.M., Barr, A.C., Crawford, D.A. 2013. Lunar-forming impacts: High-resolution SPH and AMR-CTH simulations. Icarus 222, 200-219. Collins, G.S., Melosh, H.J., Ivanov, B.A., 2004. Modeling damage and deformation in impact simulations. Meteor. Planet. Sci. 39, 217-231. Condomines, M., Hemond, C., Allègre, C.J., 1988. U-Th-Ra Radioactive Disequilibria and Magmatic Processes. Earth Planet. Sci. Lett. 90, 243-262. Ćuk, M., Stewart, S.T., 2012. Making the Moon from a Fast-Spinning Earth: A Giant Impact Followed by Resonant Despinning. Science 338, 1047-1052, 2012. Chudinovskikh, L., Boehler, R., 2007. Eutectic melting in the system Fe-S to 44 GPa. Earth Planet. Sci. Lett. 257, 97-103. Costa, A., Caricchi, L., Bagdassarov, N., 2009. A model for the rheology of particle-bearing suspensions and partially molten rocks. Geochem. Geophys. Geosyst. 10, Q03010. Crameri, F., Schmeling, H., Golabek, G.J., Duretz, T., Orendt, R., Buiter, S.J.H., May, D.A., Kaus, B.J.P., Gerya T.V., Tackley, P.J., 2012. A comparison of numerical surface topography calculations in geodynamic modelling: an evaluation of the 'sticky air' method. Geophys. J. Int. 189, 38-54. Emsenhuber, E., Jutzi, M., Benz, W., SPH calculations of Mars-scale collisions: The role of the equation of state, material rheologies, and numerical effects. Icarus, in press. Escartín, J., Hirth, G., Evans, B., 2001. Strength of slightly serpentinized peridotites: Implications for the tectonics of oceanic lithosphere. Geology 29, 1023-1026. Frey, H., Schultz, R.A., 1988. Large impact basins and the mega-impact origin for the crustal dichotomy on Mars. Geophys. Res. Lett. 15, 229-232. Gerya, T.V., Yuen, D.A., 2003. Characteristics-based marker-in-cell method with conservative finite- differences schemes for modeling geological flows with strongly variable transport properties. Phys. Earth Planet. Int. 140, 293-318. Gerya, T.V., Yuen, D.A., 2007. Robust characteristics method for modelling multiphase visco-elasto-plastic thermo-mechanical problems. Phys. Earth Planet. Int. 163, 83-105. Gilbert, G.K., 1893. The Moon's face, a study of the origin of its features. Bull. Philos. Soc. Wash. (DC) 12, 241-292. Golabek, G.J., Keller, T., Gerya, T.V., Zhu, G., Tackley, P. J., Connolly, J.A.D., 2011. Origin of the martian dichotomy and Tharsis from a giant impact causing massive magmatism. Icarus 215, 346-357. Golabek, G.J., Bourdon, B., Gerya, T.V., 2014. Numerical models of the thermomechanical evolution of planetesimals: Application to the acapulcoite-lodranite parent body. Meteorit. Planet. Sci. 49, 1083-1099. Hartmann, W. K., Davis, D. R., 1975. Satellite-sized planetesimals and lunar origin. Icarus 24, 504-515. Hevey, P.J., Sanders, I.S., 2006. A model for planetesimal meltdown by 26Al and its implications for meteorite parent bodies, Meteorit. Planet. Sci. 41, 95-106. Herzberg, C., Raterron, P., Zhang, J., 2000. New experimental observations on the anhydrous solidus for peridotite KLB-1. Geochem. Geophys. Geosyst. 1, 2000GC000089. Jacobsen, B., Yin, Q.-Z., Moynier, F., Amelin, Y., Krot, A.N., Nagashima, K., Hutcheon, I.D., Palme, H., 2008. 26Al-26Mg and 207Pb-206Pb systematics of Allende CAIs: Canonical solar initial 26Al/27Al ratio reinstated. Earth Planet. Sci. Lett. 272, 353-364. Jutzi, M., Asphaug, E., Gillet, P., Barrat, J.-A., Benz, W., 2013. The structure of the asteroid 4 Vesta as revealed by models of planet-scale collisions. Nature 494, 207-210. Jutzi, M., 2015. SPH calculations of asteroid disruptions: The role of pressure dependent failure models. Planet. Space Sci. 107, 3-9. Jutzi, M., Holsapple, K.A., Wünnemann, K., Michel, P., 2015. Modeling asteroid collisions and impact processes. In Asteroids IV (P. Michel, F. DeMeo, and W. F. Bottke, eds.), Univ. of Arizona, Tucson, pp. 679- 700. Kameyama, M., Yuen, D.A., Karato, S.-i., 1999. Thermal-mechanical effects of low-temperature plasticity (the Peierls mechanism) on the deformation of a viscoelastic shear zone, Earth Planet. Sci. Lett. 168, 159-172. Karato, S.-i., Murthy, V.R., 1997. Core formation and chemical equilibrium in the Earth I. Physical considerations. Phys. Earth Planet. Int. 100, 61-79. Katayama, I., Karato, S.-i., 2008. Low-temperature, high-stress deformation of olivine under water-saturated conditions. Phys. Earth Planet. Int. 168, 125-133. Keller, T., Tackley, P.J., 2009. Towards self-consistent modelling of the martian dichotomy: The influence of one-ridge convection on crustal thickness distribution. Icarus 202, 429-443. Kraichnan R.H., 1962. Turbulent thermal convection at arbitrary Prandtl number. Phys. Fluids 5, 1374-1389. Liebske, C., Schmickler, B., Terasaki, H., Poe, B.T., Suzuki, A., Funakoshi, K.-I., Ando, R., Rubie, D.C., 2005. Viscosity of peridotite liquid up to 13 GPa: Implications for magma ocean viscosities. Earth Planet. Sci. Lett. 240, 589-604. Lodders, K., Fegley, B., 1998. The Planetary Scientist's Companion. Oxford Univ. Press, New York, 371pp. Mackwell, S.J., 1991. High-temperature rheology of enstatite: Implications for creep in the mantle. Geophys. Res. Lett. 18, 2027-2030. Marinova, M.M., Aharonson, O., Asphaug, E., 2008. Mega-impact formation of the Mars hemispheric dichotomy. Nature 453, 1216-1219. Marinova, M.M., Aharonson, O., Asphaug, E., 2011. Geophysical consequences of planetary scale impacts into a Mars-like planet. Icarus 211, 960-985. Melosh, H.J., 1989. Impact Cratering: A Geologic Process. Monographs on Geology and Geophysics. 11. Melosh, H.J., 1990. Giant impacts and the thermal state of the early Earth. In: Origin of the Earth, edited by H. E. Newsom & J. H. Jones, Oxford Univ. Press, 69-83. Melosh, H.J., 2007. A hydrocode equation of state for SiO2. Meteor. Planet. Sci. 42, 2079-2098. McKenzie, D., Bickle, M.J., 1988. The volume and composition of melt generated by extension of the lithosphere. J. Petr. 29, 625-679. Mezger, K., Debaille, V., Kleine, T., 2013. Core Formation and Mantle Differentiation on Mars. Space Science Rev. 174, 27-48. Morishima, R., Golabek, G.J. and Samuel, H., 2013. N-body simulations of oligarchic growth of Mars: Implications for Hf-W chronology. Earth Planet. Sci. Lett. 366, 6-16. Nimmo, F., Kleine, T., 2007. How rapidly did Mars accrete? Uncertainties in the Hf-W timing of core formation. Icarus 191, 497-504. Nimmo, F., Hart, S.D., Korycansky, D.G., Agnor, C.B., 2008. Implications of an impact origin for the martian hemispheric dichotomy. Nature 453, 1220-1223. Pinkerton, H., Stevenson, R.J., 1992. Methods of determining the rheological properties of magmas at subliquidus temperatures. J. Volcan. Geotherm. Res. 53, 47-66. Ranalli, G., 1995. Rheology of the Earth, second ed., Chapman & Hall, London, UK, 436pp. Reese, C.C., Solomatov, V.S., 2006. Fluid Dynamics of local martian magma oceans. Icarus 184, 102-120. Reese, C.C., Solomatov, V.S., 2010. Early martian dynamo generation due to giant impacts. Icarus 207, 82-97. Reese, C.C., Orth, C.P., Solomatov, V.S., 2010. Impact origin for the martian crustal dichotomy: Half-emptied or half-filled? J. Geophys. Res. 115, E05004. Reufer, A., Meier, M.M.M., Benz, W. & Wieler, R., 2012. A hit-and-run Giant Impact scenario. Icarus 221, 296-299. Roberts, J.H., Zhong, S., 2006. Degree-1 convection in the Martian mantle and the origin of the hemispheric dichotomy. J. Geophys. Res. 111, E06013. Rolf, T., Zhu, M.-H., Wünnemann, K., Werner, S.C., 2017. The role of impact bombardment history in lunar evolution. Icarus 286, 138-152. Rubie, D., Melosh, H.J., Reid, J.E., Liebske, C., Righter, K., 2003. Mechanisms of metal-silicate equilibration in the terrestrial magma ocean. Earth Planet. Sci. Lett. 205, 239-255. Rubie, D.C., Nimmo, F., Melosh, H.J., 2015. Formation of the Earth's core. In Treatise on Geophysics, 2nd ed., Volume 9: Evolution of the Earth. Stevenson D. J. (ed.), Amsterdam: Elsevier B.V. pp. 43-79. Schmeling, H., Babeyko, A.Y., Enns, A., Faccenna, C., Funiciello, F., Gerya, T.V., Golabek, G.J., Grigull, S., Kaus, B.J.P., Morra, G., Schmalholz, S.M., van Hunen, J., 2008. A benchmark comparison of spontaneous subduction models - Towards a free surface. Phys. Earth Planet. Int. 171, 198-223. Shoemaker, E.M., 1962. Interpretation of lunar craters. In: Kopal, Z. (Ed.), Physics and Astronomy of the Moon. Academic Press, New York, pp. 283-359. Siggia, E.D., 1994. High Rayleigh number convection. Annu. Rev. Fluid Mech. 26, 137-168. Solomatov, V.S., 2015. Magma oceans and primordial mantle differentiation. In: Stevenson, D.J. (Ed.), Treatise on Geophysics, 2nd ed., Evolution of the Earth, vol. 9., Elsevier B.V., Amsterdam, Netherlands, pp. 81-104. Senft, L.E., Stewart, S.T., 2009. Dynamic fault weakening and the formation of large impact craters. Earth Planet. Sci. Lett. 287, 471-482. Šra mek, O., Zhong, S., 2010. Long-wavelength stagnant-lid convection with hemispheric variation in lithospheric thickness: link between Martian crustal dichotomy and Tharsis? J. Geophys. Res. 115, E09010. Šra mek, O., Zhong, S., 2012. Martian crustal dichotomy and Tharsis formation by partial melting coupled to early plume migration. J. Geophys. Res. 117, E01005. Stolper, E., Hager, B.H., Walker, D., Hays, J.F., 1981. Melt segregation from partially molten source regions - The importance of melt density and source region size. J. Geophys. Res. 86, 6261-6271. Suzuki, A., Ohtani, E., Kato, T., 1998. Density and thermal expansion of a peridotite melt at high pressure. Phys. Earth Planet. Int. 107, 53-61. Tackley, P.J., Schubert, G., Glatzmaier, G.A., Schenk, P., Ratcliff, J.T., 2001. Three-dimensional simulations of mantle convection in Io. Icarus 149, 79-93. Tackley, P.J., 2008. Modelling compressible mantle convection with large viscosity contrasts in a three- dimensional spherical shell using the yin-yang grid. Phys. Earth Planet. Int. 171, 7-18. Taylor, S.R., McLennan, S.M., 2009. Planetary Crusts: Their Composition, Origin and Evolution. Cambridge Univ. Press, Cambridge, UK, 404pp. Thompson, S.L., Lauson, H.S., 1972. Improvements in the CHART-D Radiation-hydrodynamic code III: Revised analytic equations of state. Technical report SC-RR-71 0714. Thrane, K., Bizzarro, M., Baker, J.A., 2006. Extremely brief formation interval for refractory inclusions and uniform distribution of 26Al in the early Solar System. Astrophys. J. 646, L159–L162. Tillotson, J.H., 1962. Metallic Equations of State for Hypervelocity Impact. GA-3216, General Atomic, San Diego, CA. Tkalcec, B.J., Golabek, G.J., Brenker, F.E., 2013. Solid-state plastic deformation in the dynamic interior of a differentiated asteroid. Nat. Geosci. 6, 93-97. Trønnes, R.G., Frost, D.J., 2002. Peridotite melting and mineral-melt partitioning of major and minor elements at 22–24.5 GPa. Earth Planet. Sci. Lett. 197, 117-131. Turcotte D. L., Schubert G., 2014. Geodynamics, 3rd ed. New York: Cambridge University Press. 636 pp. Wade, J., Wood, B.J., 2005. Core formation and the oxidation state of the Earth. Earth Planet. Sci. Lett. 236, 78-95. Weinberg, R.F., Podladchikov, Y., 1994. Diapiric ascent of magmas through power law crust and mantle. J. Geophys. Res. 99, 9543-9559. Weinstein, S.A., 1995. The effects of a deep mantle endothermic phase change on the structure of thermal convection in silicate planets. J. Geophys. Res. 100, 11719-11728. Wetherill, G.W., 1985. Occurrence of giant impacts during the growth of the terrestrial planets. Science 228, 877-879. Wilhelms, D.E., Squyres, S.W., 1984. The martian dichotomy may be due to a giant impact. Nature 309, 138- 140. Yunker, M.L., Van Orman, J.A., 2007. Interdiffusion of solid iron and nickel at high pressure, Earth Planet. Sci. Lett. 254, 203-213. Zahnle, K.J., Kasting, J.F., Pollack, J.B., 1988. Evolution of a steam atmosphere during Earth's accretion. Icarus 74, 62-97. Ziethe, R., Spohn, T., 2007. Two-dimensional stokes flow around a heated cylinder: A possible application for diapirs in the mantle. J. Geophys. Res. 112, B09403. Zhong, S., 2009. Migration of Tharsis volcanism on Mars caused by differential rotation of the lithosphere. Nat. Geosci. 2, 19-23. Zhong, S., Zuber, M.T., 2001. Degree-1 mantle convection and the crustal dichotomy on Mars. Earth Planet. Sci. Lett. 189, 75-84. Figures Figure 1: Grazing impact simulations at t = 0 s. The impactor's trajectory is clockwise. Shown is a slice at 1000 km depth inside the impact plane. Plots are centered on the center of mass of the target body. Only simulations using ANEOS are shown here; models using Tillotson show almost no difference at this stage. Temperature distributions for solid and fluid rheologies according to captions. Colors for material distribution are as follows: blue for target mantle, purple for impactor mantle, red for target core and orange for impactor core. Figure 2: Material and temperature distributions resulting from the (a) grazing collision simulations and (b) head-on collisions, at t = 18 hours after the impact. Colors for material distribution as given in Figure 1. Note that due to the grazing impacts, the targets have rotated clockwise by ~270 degrees. Figure 3: Remaining ejecta in orbit as a function of time for the grazing cases shown in Figure 2a. Figure 4: Characteristic cooling timescale for the disregarded ejecta material when spread out into a global surface layer. Figure 5: Cumulative crustal thickness distributions at ~5·105 yr after the collision for different impact angle (upper row) and different rheology used in the SPH simulation (lower row). The curves show for each case the fraction of the surface area that has an underlying crust of a thickness larger than the value given on the x- axis. Figure 6: Maps of crustal thickness distribution at ~5·105 yr after the collision resulting from head-on collisions using the Mercator equal-area projection. Figure 7: Maps of crustal thickness distribution at ~5·105 yr after the collision resulting from grazing collisions using the Mercator equal-area projection. Figure 8: Time evolution of the volume of stable crust (disconnected crustal drips are disregarded). Figure 9: Time evolution of the volume of stable crust (cyan coloured lines) and total volume of crust (auburn coloured lines) for the two extreme cases (coll02, coll08) from Figure 8. Figure 10: Cumulative crustal thickness distributions at ~5·105 yr after the collision for the nominal case using different SPH-I3ELVIS transfer times. Tables Table 1: List of numerical models Model name Impact angle [degrees] EoS Transfer time SPH-I3ELVIS [h] SPH rheology coll01 coll02 coll03 coll04 coll05 coll06 coll07 coll08 coll09 coll10 Tillotson Tillotson M-ANEOS M-ANEOS M-ANEOS M-ANEOS Tillotson Tillotson M-ANEOS M-ANEOS 0 45 0 45 45 45 0 45 0 45 18 18 18 18 8 12 18 18 18 18 solid solid solid solid solid solid fluid fluid fluid fluid Table 2: Physical parameters used in I3ELVIS Parameter Symbol Value Units Reference Density of uncompressed silicate melt Density of uncompressed solid silicates Temperature of space (sticky air) Activation energy Activation volume Dislocation creep onset stress Power law exponent Cohesion Internal friction coefficient of solid silicates Latent heat of silicate melting Silicate melt fraction at rheological transition Heat capacity of silicates Thermal expansivity of solid silicates Thermal expansivity of molten silicates Thermal conductivity of solid silicates Thermal conductivity of molten silicates ρSi-liq ρSi-sol Tsa Ea Va σ0 n C f LSi φcrit cP αSi-sol αSi-liq k keff 2900 3500 220 470 kg/m3 kg/m3 K kJ/mol 8.10-6 m3/mol (1,2) (1) (3) (4) (5) (6) (4) (5) (7) (6) Pa Pa kJ/kg 3.107 4 108 0.3 400 0.4 (8,9) 1000 J/(kg K) 3.10-5 6.10-5 1/K 1/K 3 W/(m K) (6) (2) (2) (6) ≤106 W/(m K) (10) References: (1) Stolper et al. (1981), (2) Suzuki et al. (1998), (3) Lodders & Fegley (1998), (4) Ranalli (1995), (5) Golabek et al. (2011), (6) Turcotte & Schubert (2014), (7) Escartín et al. (2001), (8) Solomatov (2015), (9) Costa et al. (2009), (10) Golabek et al. (2014)
1603.00042	1	1603	2016-02-29T21:13:40	A Transiting Jupiter Analog	[ "astro-ph.EP" ]	Decadal-long radial velocity surveys have recently started to discover analogs to the most influential planet of our solar system, Jupiter. Detecting and characterizing these worlds is expected to shape our understanding of our uniqueness in the cosmos. Despite the great successes of recent transit surveys, Jupiter analogs represent a terra incognita, owing to the strong intrinsic bias of this method against long orbital periods. We here report on the first validated transiting Jupiter analog, Kepler-167e (KOI-490.02), discovered using Kepler archival photometry orbiting the K4-dwarf KIC-3239945. With a radius of $(0.91\pm0.02)$ $R_{\mathrm{Jup}}$, a low orbital eccentricity ($0.06_{-0.04}^{+0.10}$) and an equilibrium temperature of $(131\pm3)$ K, Kepler-167e bears many of the basic hallmarks of Jupiter. Kepler-167e is accompanied by three Super-Earths on compact orbits, which we also validate, leaving a large cavity of transiting worlds around the habitable-zone. With two transits and continuous photometric coverage, we are able to uniquely and precisely measure the orbital period of this post snow-line planet ($1071.2323\pm0.0006$ d), paving the way for follow-up of this $K=11.8$ mag target.	astro-ph.EP	astro-ph	Draft version March 2, 2016 Preprint typeset using LATEX style emulateapj v. 5/2/11 6 1 0 2 b e F 9 2 . ] P E h p - o r t s a [ 1 v 2 4 0 0 0 . 3 0 6 1 : v i X r a A TRANSITING JUPITER ANALOG D. M. Kipping1, G. Torres2, C. Henze3, A. Teachey1, H. Isaacson4, E. Petigura4, G. W. Marcy4, L. A. Buchhave5, J. Chen1, S. T. Bryson3, E. Sandford1 Draft version March 2, 2016 ABSTRACT Decadal-long radial velocity surveys have recently started to discover analogs to the most influential planet of our solar system, Jupiter. Detecting and characterizing these worlds is expected to shape our understanding of our uniqueness in the cosmos. Despite the great successes of recent transit surveys, Jupiter analogs represent a terra incognita, owing to the strong intrinsic bias of this method against long orbital periods. We here report on the first validated transiting Jupiter analog, Kepler-167e (KOI- 490.02), discovered using Kepler archival photometry orbiting the K4-dwarf KIC-3239945. With a radius of (0.91 ± 0.02) RJ, a low orbital eccentricity (0.06+0.10−0.04) and an equilibrium temperature of (131± 3) K, Kepler-167e bears many of the basic hallmarks of Jupiter. Kepler-167e is accompanied by three Super-Earths on compact orbits, which we also validate, leaving a large cavity of transiting worlds around the habitable-zone. With two transits and continuous photometric coverage, we are able to uniquely and precisely measure the orbital period of this post snow-line planet (1071.2323± 0.0006 d), paving the way for follow-up of this K = 11.8 mag target. Subject headings: techniques: photometric - planetary systems - planets and satellites: detection - stars: individual (KIC-3239945, KOI-490, Kepler-167) 1. INTRODUCTION Jupiter is the dominant member of our planetary sys- tem with a mass exceeding twice that of all the other planets combined. Theories of the formation and evolu- tion of our neighboring planets are usually conditioned upon the properties and location of our system's gargan- tuan world (see, e.g. Walsh et al. 2011), yet exoplanetary surveys have only recently begun to assess the prevalence of such objects (see, e.g. Gould et al. 2010). Jupiter's presiding mass led to it playing a critical role in the dynamical evolution of the early Solar System (Morbidelli, et al. 2007). The final architecture of our solar system, including the Earth, is thus intimately con- nected to the existence and dynamical history of Jupiter (Batygin & Laughlin 2015). The mass, location and exis- tence of Jupiter also likely affect the impact rate of minor bodies onto the Earth (Horner et al. 2010), thereby in- fluencing the evolution of terrestrial life. The search for Jupiter analogs has therefore emerged as a scientific pri- ority, linked to the fundamental goal of understanding our uniqueness in the cosmos. Around 20 extrasolar Jupiter analogs have been dis- covered with the radial velocity method (see Table 4 of Rowan et al. 2015), indicating that these cool worlds are not unique to the Solar System. Occurrence rate esti- mates, including constraints from microlensing surveys, (cid:39) 3% (Cumming et al. 2008; Gould et al. 2010; Wittenmyer et al. 2011; Rowan et al. 2015) (depending upon the definition of an "analog"), typically converge at η(cid:88) 1 Dept. of Astronomy, Columbia University, 550 W 120th St., New York, NY 10027, USA; email: [email protected] 2 Harvard-Smithsonian Center for Astrophysics, Cambridge, MA 02138, USA 3 NASA Ames Research Center, Moffett Field, CA 94035, USA 4 University of California, Berkeley, CA 94720, USA 5 Centre for Star and Planet Formation, Natural History Mu- seum of Denmark, University of Copenhagen, DK-1350 Copen- hagen, Denmark It is interesting to note that η(cid:88) is approxi- although recently Wittenmyer et al. (2016) argued for 6.1+2.8−1.6%. mately equal to the prevalance of Earth analogs orbit- ing FGK stars as measured using the Kepler transit sur- vey, specifically η⊕ = 1.7+2.6−1.0% (Petigura et al. 2013; Foreman-Mackey et al. 2014) (periods of 200–400 days; 0.5–1.5 R⊕), although again with the caveat depending upon how one defines "analogous". The transit method has dominated the exoplanet de- tection game over the last decade. This technique has demonstrated a sensitivity to planets ranging from sub- Earths (Barclay et al. 2013) to super-Jupiters (Fortney et al. 2011) orbiting a diverse array of stars, such as M- dwarfs (Dressing & Charbonneau 2015), giants (Quinn et al. 2015) and even binaries (Doyle et al. 2011). From 2010-2015, the number of confirmed/validated exoplan- ets discovered via the transit method is seven-fold that of all other exoplanet hunting methods combined. One of the last regions of parameter-space which has been stub- bornly resistant to the reign of transits are those planets found beyond the snow-line, owing to their long orbital periods. Indeed, whilst ∼ 20 Jupiter analogs have been found with radial velocities (Rowan et al. 2015), no tran- siting examples have been previously announced. Jupiter analogs have both a reduced geometric tran- sit probability and a lower chance of being observed to transit within a fixed observing window less than twice the planet's orbital period (which is usually the case). In general, one expects the planey yield of a transit survey to scale as P −5/3 (Beatty & Gaudi 2008), implying that a 3 day period Jupiter is ∼16,000 times easier to find than the same planet at 1000 days. Nevertheless, in a large transit survey spanning multiple years, such as Ke- pler (∼200,000 stars over ∼4 years), these obstacles are expected to yield and Kepler should expect detections if the occurrence rate is (cid:38) O[10−2]. Pursuing this possibility, we here report the dis- 2 Kipping et al. covery of a 1071 day period transiting planet, Kepler- 167e (formeley KOI-490.02), orbiting the K-dwarf KIC- 3239945 (see Table 1) with a size and insolation com- parable to Jupiter. This planet, which lies comfortably beyond the snow-line, is found to transit twice over the duration of the Kepler mission allowing for a precise de- termination of the orbital period and making it schedu- lable for future follow-up work. The data processing and follow-up observations required for this discovery are dis- cussed in § 2 &§ 3 respectively. In § 4, we discuss how we are able to validate Kepler-167e and the other three tran- siting candidates (KOI-490.01, .03 & .04; P ∼ 4.4, 7.4 & 21.8 d) within the system using BLENDER. Light curve fits, leveraging asterodensity profiling, are discussed in § 5, allowing us to infer the radius and even eccentricity of Kepler-167e. Finally, we place this discovery in context in § 6, discussing the system architecture and prospects for follow-up. 2. KEPLER PHOTOMETRY 2.1. Data Acquisition We downloaded the publicly available Kepler data for KOI-490 from the Mikulski Archive for Space Telescopes (MAST). The downloaded data were released as part of Data Release 24 and were processed using Science Opera- tions Center (SOC) Pipeline version 9.2.24. All quarters from 1-17 were available in long-cadence (LC) and from 9-17 there was also short-cadence (SC), which was used preferentially over LC. 2.2. Data Selection To fit light curve models to the Kepler data, it is neces- sary to first remove instrumental and stellar photomet- ric variability which can distort the transit light curve shape. We break this process up into two stages: (i) pre- detrending cleaning (ii) long-term detrending. In what follows, each quarter is detrended independently. 2.3. Pre-detrending Cleaning The first step is to visually inspect each quarter and remove any exponential ramps, flare-like behaviors and instrumental discontinuities in the data. We make no at- tempt to correct these artifacts and simply exclude them from the photometry manually. We inspect all points occurring outside of a transit for In-transit points are defined as those occur- outliers. ring within ±0.6 transit durations of the nominal lin- ear ephemeris for each KOI. For these durations and ephemerides, we adopt the NASA Exoplanet Archive Akeson et al. (2013) parameters. We then clean the out-of-transit Simple Aperture Photometry (SAP) light curve of 3 σ outliers, identified using a moving median smoothing curve with a 20-point window. 2.4. Detrending with CoFiAM For the data used in the transit light curve fits in §5, it is also necessary to remove the remaining long-term trends in the time series. These trends can be due to instrumental effects, such as focus drift, or stellar ef- fects, such as rotational modulations. For this task, data are detrended using the Cosine Filtering with Autocor- relation Minimization (CoFiAM) algorithm. CoFiAM was specifically developed to protect the shape of a transit light curve and we direct the reader to our previous work (Kipping et al. 2013) for a detailed description. Each transit of each KOI is detrended independently using CoFiAM, setting the protected timescale to twice the associated transit duration. After detrending, the light curves were fitted with the same light curve model and algorithm described later in §5. The maximum likelihood duration and ephemeris were saved from these fits. We then used these values to go back to §2.3 and repeat the entire detrending process, to ensure we used accurate estimates of these terms. 3. FOLLOW-UP OBSERVATIONS 3.1. Spectroscopy KOI-490 was observed on 2011 October 16 at the Keck I telescope on Mauna Kea (HI) with the HIRES spectrometer (Vogt et al. 1994), in order to help charac- terize the star as described below in § 3.4. The exposure time was 30 minutes and the spectrograph slit was set using the C2 decker (0.(cid:48)(cid:48)86 × 14(cid:48)(cid:48)). Reductions were per- formed with the standard procedures employed by the California Planet Search (Howard et al. 2010; Johnson et al. 2010). This resulted in an extracted spectrum with R ∼ 60, 000 covering the approximate wavelength range 360–800 nm, with a signal-to-noise ratio of 90 per resolu- tion element in the region of the Mg I b triplet (519 nm). We examined the spectrum for signs of absorption lines from another star that might be causing the transit sig- nal, if located within the slit. This was done by first subtracting a spectrum closely matching that of the tar- get star (after proper wavelength shifting and continuum normalization), and then inspecting the residuals (see Kolbl et al. 2015). We saw no evidence of secondary spectral lines. In order to quantify our sensitivity to such companions we performed numerical simulations in which we subjected the residuals to a similar fitting pro- cess by injecting mock companions over a range of tem- peratures from 3500 to 6000 K, and with a broad range in relative velocities. We then attempted to recover them, and this allowed us to estimate that we are sensitive to companions down to about 1% of the flux of the primary star, with velocity separations greater than 10 km s−1. For smaller relative velocities the secondary lines would be blended with those of the primary and would not be detected. This spectroscopic constraint is used below for the validation of the candidates in § 4. 3.2. High-resolution Imaging Images from the J-band UK Infrared Telescope survey (UKIRT; Lawrence et al. 2007) available on the Kepler Community Follow-up Observing Program (CFOP) Web site6 show a nearby companion about five magnitudes fainter than KOI-490 at an angular separation of 2.(cid:48)(cid:48)1 in position angle 62.◦7, which falls within the photometric aperture of Kepler. Ancillary information for this source based on automatic image classification indicates a prob- ability of 99.4% that it is a galaxy, 0.3% that it is a star, and 0.3% that it is noise, though it is unclear how robust these assessments are. The UKIRT images have a typical seeing-limited resolution of about 0.(cid:48)(cid:48)8 or 0.(cid:48)(cid:48)9. Additional imaging efforts reported on CFOP include speckle inter- ferometry observations on the WIYN 3.5m telescope at 6 https://cfop.ipac.caltech.edu/home/ . A Transiting Jupiter Analog 3 692 nm and 880 nm (see Horch et al. 2014; Everett et al. 2015), lucky imaging observations on the Calar Alto 2m telescope at 766 nm (Lillo-Box et al. 2012), and imaging with the Robo-AO system on the Palomar 1.5m telescope approximately in the R band (Law et al. 2014), none of which detected this companion likely due to its faintness at the optical wavelengths probed by these observations. To investigate this detection further and to explore the inner regions around the target, we observed KOI- 490 with near-infrared adaptive optics (AO) using the 1024 × 1024 NIRC2 imager (Wizinowich et al. 2004; Jo- hansson et al. 2008) on the Keck II, 10m telescope on the night of 2014 June 12. We used the natural guide star system, as the target star was bright enough to be used as the guide star. The data were acquired using the narrow- band Br-γ filter and the narrow camera field of view with a pixel scale of 9.942 mas pixel−1. The Br-γ filter has a narrower bandwidth (2.13–2.18 µm), but a similar cen- tral wavelength (2.15 µm) compared the 2MASS Ks filter (1.95–2.34 µm; 2.15 µm) and allows for longer integra- tion times before saturation. A 3-point dither pattern was utilized to avoid the noisier lower left quadrant of the NIRC2 array. The 3-point dither pattern was ob- served three times with two coadds per dither position for a total of 18 frames; each frame had an exposure time of 30 s, yielding a total on-source exposure time of 3 × 3 × 2 × 30 s = 540 s. The target star was measured with a resolution of 0.(cid:48)(cid:48)051 (FWHM). The object 2(cid:48)(cid:48) to the NE of KOI-490 seen in the UKIRT images was clearly detected in the NIRC2 data (see Fig- ure 1), and no other stars were detected within 10(cid:48)(cid:48). The image of the companion appears stellar, so we consider it a star in the following. In the Br-γ passband the data are sensitive to companions that have a K-band contrast of ∆K = 3.8 mag at a separation of 0.(cid:48)(cid:48)1 and ∆K = 8.8 mag at 0.(cid:48)(cid:48)5 from the central star. We estimated the sensi- tivities by injecting fake sources with a signal-to-noise ratio of 5 into the final combined images at distances of N × FWHM from the central source, where N is an integer. We also observed KOI-490 in the J-band (1.248 µm) with NIRC2 in order to obtain the J − K color of the companion star. The J-band observations employed the same 3-point dither pattern with an integration time of 10 s per coadd for a total on-source integration time of 180 s. These data had slightly better resolution (0.(cid:48)(cid:48)046) and a sensitivity of ∆J = 4.4 mag at 0.(cid:48)(cid:48)1 and ∆J = 7.4 mag at 0.(cid:48)(cid:48)5. Full sensitivity curves in both J and Br-γ are shown in Figure 1. The companion star was found to be fainter than the primary star by ∆J = 3.84 ± 0.03 mag and ∆K = 3.51 ± 0.01 mag, and separated from the primary by ∆α = 1.(cid:48)(cid:48)97 ± 0.(cid:48)(cid:48)01 and ∆δ = 1.(cid:48)(cid:48)00 ± 0.(cid:48)(cid:48)01 in right ascension and dec- lination (corresponding to an angular separation of 2.(cid:48)(cid:48)21 in position angle 63.◦1). After deblending the 2MASS photometry we find that the primary has J- and K-band magnitudes of J1 = 12.47 ± 0.02 mag and K1 = 11.87 ± 0.02 mag, and the secondary has J2 = 16.32 ± 0.04 mag and K2 = 15.38 ± 0.02 mag. The companion is a redder star than the primary: the individual colors are (J − K)1 = 0.60 ± 0.03 and (J − K)2 = 0.94 ± 0.04. Fig. 1.- Br-γ Keck/NIRC2 AO image of KOI-490 shown along with the sensitivity curves in the J (1.248 µm) and Br-γ (2.15 µm) bands. 3.3. Centroid motion analysis The very precise astrometry that can be obtained from the Kepler images enables a search for false positives that may be causing one or more of the signals in KOI-490, such as a background eclipsing binary. This can be done by measuring the location of the transit signals relative to the target by means of difference images, formed by subtracting an average of in-transit pixel values from out- of-transit pixel values. If a transit signal is caused by a stellar source, then the difference image will show that stellar source, and its location can be determined by pixel response function centroiding (Bryson et al. 2013). The centroid of an average out-of-transit image provides the location of KOI-490 because the object is well isolated. The centroid of the difference image is then compared to that of the out-of-transit image, which provides the location of the transit source relative to KOI-490. The automatic pipeline processing of Kepler provides these offsets for each quarter in the Data Validation Re- ports, which are available through the CFOP Web site. For KOI-490.01 and KOI-490.03 the multi-quarter av- erages of the offsets indicate a position for the source of the transits that is consistent with the location of target. Based on the 1σ uncertainties associated with those multi-quarter average offsets we adopted 3σ radii of confusion for these two candidates of 0.(cid:48)(cid:48)396 and 0.(cid:48)(cid:48)603, respectively, within which the centroid motion analy- sis is insensitive to the presence of contaminating stars. As these limits are smaller than the separation of the 2.(cid:48)(cid:48)2 companion reported earlier, that star cannot be the source of these transits. For KOI-490.02 and KOI-490.04 the automatic fits per- formed by the Kepler pipeline failed, as indicated in the Data Validation Reports, so no centroid information is available for these candidates. 3.4. Stellar properties The spectroscopic properties of KOI-490 were deter- mined from an analysis of our Keck/HIRES spectrum. Our analysis was performed using the Stellar Parame- ter Classification (SPC) pipeline (Buchhave et al. 2012), which cross-correlates the observed spectrum against a large library of calculated spectra based on model atmo- spheres by R. L. Kurucz, and assigns stellar properties 4 Kipping et al. TABLE 1 Stellar properties of KOI-490. Property Value 4890 ± 50 Teff (K)a . . . . . . . . . . . . . . . 4.61 ± 0.10 log g (dex)a . . . . . . . . . . . . . [Fe/H] (dex)a . . . . . . . . . . . −0.03 ± 0.08 v sin i (km s−1)a . . . . . . . . < 2 log10[ρ(cid:63)/(kg m−3)]b . . . . 3.460+0.031 −0.065 0.770+0.024 M(cid:63) (M(cid:12)) . . . . . . . . . . . . . . −0.028 0.726+0.018 R(cid:63) (R(cid:12)) . . . . . . . . . . . . . . . −0.015 log10[L(cid:63)/L(cid:12)] . . . . . . . . . . . −0.570+0.036 −0.034 6.53 ± 0.12 MV (mag) . . . . . . . . . . . . . . 4.21 ± 0.06 MKs (mag). . . . . . . . . . . . . 330 ± 10 Distance (pc) . . . . . . . . . . . 3.3+5.8−0.8 Age (Gyr) . . . . . . . . . . . . . . a Value from SPC. b Mean stellar density constraint from tran- sit light curve fits to KOI-490.01, KOI- 490.03, and KOI-490.04 (see text). interpolating amongst those of the synthetic spectra pro- viding the best match. This analysis gave Teff = (4890± 50) K, log g = (4.61 ± 0.10), [Fe/H] = (−0.03 ± 0.08), and v sin i < 2 km s−1. The measured radial velocity is (−29.3 ± 1.0) km s−1, and the effective temperature corresponds to a spectral type of K3 or K4. The mass and radius of the star, along with other prop- erties, were estimated by comparing the SPC parameters against a grid of Dartmouth isochrones (Dotter et al. 2008) with a χ2 procedure similar to that described by Torres et al. (2008). Because the stellar radius and age are largely determined by the surface gravity, and our log g determination provides a relatively weak constraint for KOI-490 given its uncertainty, we supplemented it with an estimate of the mean stellar density obtained by fitting the Kepler light curves of KOI-490.01, KOI- 490.03, and KOI-490.04, on the assumption that they are true planets (justified below) and that they orbit the same star. The stellar parameters derived in this way are listed in Table 1, along with the inputs from SPC and the photometric mean density (posteriors shown in Figure 13). The inferred distance is based on the appar- ent Ks-band magnitude from 2MASS and a reddening estimate of E(B − V ) = (0.075 ± 0.030) from the Kepler Input Catalog (KIC; Brown et al. 2011). Given these properties for KOI-490, we investigated whether the measured brightness and color of the 2.(cid:48)(cid:48)2 neighbor reported earlier are consistent with those ex- pected for a physically associated main-sequence star of later spectral type, i.e., one falling on the same Dart- mouth isochrone as the primary. We find that an M4 or M5 dwarf with a mass around 0.20–0.21 M(cid:12) would have approximately the right brightness compared to the pri- mary, though its J − K color would be about 0.16 mag bluer than we measure. However, given the uncertainties that may be expected in the theoretical flux predictions for cool stars (based here on PHOENIX model atmo- spheres implemented in the Dartmouth models), as well as variations in color that may occur in real stars due, e.g., to chromospheric activity, we consider the measured properties to be still consistent with a bound companion, although a chance alignment cannot be ruled out. 4. STATISTICAL VALIDATION Transiting planet candidates require extra care to show that the periodic dips in stellar brightness are not as- trophysical false positives, caused by other phenomena such as an eclipsing binary blended with the target (a "blend"). Because KOI-490 is a faint star (V ≈ 14.3), it is challenging to confirm the planetary nature any of the candidates in this system dynamically, by measuring the Doppler shifts they induce on the host star. The al- ternative is to validate them statistically, showing that the likelihood of a true planet is far greater than that of a false positive. Rowe et al. (2014) followed this ap- proach and reported the validation of two of the can- didates, KOI-490.01 and KOI-490.03, based on the ar- gument that most candidates in multiple systems can be shown statistically to have a very high chance of be- ing true planets (Lissauer et al. 2012, 2014). These two planets received the official designations Kepler-167b and Kepler-167c. The validations relied in part on an exami- nation of existing follow-up observations including spec- troscopy and high-resolution imaging, and on an anal- ysis of the flux centroids. The other two candidates in the system, KOI-490.02 and KOI-490.04, were not con- sidered validated by Rowe et al. (2014) because the cen- troid information available was insufficient to determine whether the source of the photometric signals coincided with the location of the target, within errors. Addition- ally, the period of KOI-490.02 was not precisely known, since only one transit had occurred in the data at their disposal (Q1–Q10). After the publication of the Rowe et al. (2014) work, AO imaging of KOI-490 was obtained that showed the presence of a 2.(cid:48)(cid:48)2 companion that was unknown at the time (see § 3.2), and could possibly be the source of one of the signals. However, as pointed out earlier, the refined centroid information now available that includes Kepler observations from Q1–Q17 firmly rules out that the com- panion is causing the transits in KOI-490.01 and KOI- 490.03, as it is well beyond the 3 σ exclusion regions for these candidates (0.(cid:48)(cid:48)396 and 0.(cid:48)(cid:48)603, respectively; § 3.3). Thus, the validations of Rowe et al. (2014) stand. We describe below our efforts to validate the other two candidates, KOI-490.02 (the snow-line candidate) and KOI-490.04, using the BLENDER technique (Torres et al. 2004, 2011; Fressin et al. 2012). This procedure has been applied successfully to the validation of many other tran- sit candidates from Kepler (for recent examples see, e.g., Meibom et al. 2013; Ballard et al. 2013; Kipping et al. 2014; Torres et al. 2015; Jenkins et al. 2015). BLENDER addresses the possibility that the signals originate in an unseen background/foreground eclipsing binary (BEB) along the line of sight, a background or foreground star transited by a larger planet (BP scenario), or a stellar companion physically associated with the target that is in turn transited by another star or by a planet. These types of blends are usually the most difficult to rule out. The companions in the last two cases are usually close enough to the target as to be spatially unresolved. We refer to those hierarchical triple configurations as HTS or HTP, respectively, depending on the nature of the eclipsing object (star or planet). Other types of false positives that do not involve contamination by another object along the line of sight include grazing eclipsing bi- A Transiting Jupiter Analog 5 naries, and transits of a small star in front of a giant star. However, these cases can be easily ruled out as their sig- nals would be inconsistent with the observed durations of transit ingress and egress for the two candidates. Our validations with BLENDER follow closely the pro- cedure described by Kipping et al. (2014) or Torres et al. (2015); the reader is referred to these sources for de- tails of the methodology. In essence, BLENDER uses the shape of a transit light curve to rule out blend scenar- ios that would lead to the wrong shape for a transit. Large numbers of false positives of different kinds are simulated, and the synthetic light curves are then com- pared with the Kepler observations in a χ2 sense. Blends giving poor fits to the real data are considered to be excluded, and the ensemble of results places tight con- straints on the detailed properties of viable blends in- cluding the sizes or masses of the objects involved, their brightness and colors, the linear distance between the background/foreground eclipsing pair and the KOI, and even the eccentricities of the orbits. 4.1. KOI-490.02, a Snow-line Candidate Our simulations with BLENDER rule out background eclipsing binaries as the source of the signal. This is illustrated in Figure 2 (left panel), where we show the χ2 landscape in a representative cross-section of parameter space. The diagram shows the linear separation between the BEB and the target as a function of the mass of the primary star in the BEB. The only scenarios of this kind that provide acceptable fits to the Kepler light curve are those in which the main star of the binary is about twice as massive as KOI-490 (i.e., M ∼ 1.4 M(cid:12)). These false positives occupy a narrow vertical strip on the lower right corner of the first panel in the figure (darker region con- tained within the white, 3σ contour). However, all of these configurations result in a combined r − Ks color index for the blend that is much bluer than the mea- sured value for KOI-490 (r − Ks = 2.095 ± 0.027)7, as indicated by the hatched blue region in the figure within which all blends have the wrong color. Furthermore, in these false positive configurations with F-type primaries the BEB is brighter than the target itself (see dashed green line). This conflicts with our spectroscopic classifi- cation of KOI-490 as an early K dwarf. We conclude that BEBs cannot mimic the transits of KOI-490.02 and si- multaneously satisfy all observational constraints. This also rules out the 2.(cid:48)(cid:48)2 companion as the source of the transits. Blends involving background or foreground stars tran- sited by a larger planet (BP scenario) are more easily able to match the transit shape and depth. This is shown in the middle panel of Figure 2, in which the per- mitted region is larger and accommodates chance align- ments with stars between about 0.25 M(cid:12) and 1.0 M(cid:12). The spectroscopic constraint represented by the hatched green area excludes all such blends if the intruding stars have ∆Kp < 5 mag and fall within the spectrometer slit (unless their spectral lines are blended with those of the target), as we would have detected them in our Keck/HIRES observation. Most other scenarios are also ruled out by the color constraint, but there is a narrow 7 This accounts for zero-point corrections to the Sloan magni- tudes in the KIC, as prescribed by Pinsonneault et al. (2012). strip of viable blends in which the contaminating star has the same color (mass) as the target (see figure) so that it does not alter the combined r − Ks index. These would then be near twin stars of our target, and they would have to be brighter than our nominal target be- cause they are in the foreground. However, a star so similar to our target that is transited by a planet and is along the same line of sight but is brighter would effec- tively be our target, so we do not consider this as a false positive. BLENDER indicates that physical companions eclipsed by another star (HTS scenario) invariably have the wrong shape for the transit, or produce secondary eclipses that are not seen in the Kepler data for KOI-490.02. Even in cases that show only a single eclipse due to a high eccen- tricity and a special orientation (Santerne et al. 2013) the properties of the primary of the eclipsing binary would be such that the overall brightness would make the binary detectable and/or make its color inconsistent with the measured color index of the target. These configurations are therefore easily ruled out. Physically associated stars transited by a larger planet (HTP scenario) can mimic the light curve only for a very narrow range of parame- ters, as illustrated in Figure 2, but those blends are all too blue because the companion needs to be even more massive than the target in order to produce the right shape for the transit, after accounting for dilution. We can thus exclude this category of blends completely. In summary, our BLENDER simulations for KOI-490.02 combined with the observational constraints allow us to easily validate the candidate as a bona fide planet, rul- ing out as the source of the transits not only unseen background stars but also the known 2.(cid:48)(cid:48)2 companion. A significant factor aiding in this process is the very high signal-to-noise ratio of the deep transits, thanks to which the shape is so well defined (particularly the ingress/egress phases) that very few configurations in- volving another star along the line of sight can match the Kepler photometry as well as a true transiting planet fit. 4.2. KOI-490.04 The transits of this candidate are much shallower than those of KOI-490.02, and as a result the constraint on the detailed shape of the light curve provided by the Kepler data is considerably weaker. Our BLENDER simulations indicate that false positives involving a bound companion eclipsed by a smaller star (HTS scenario) do not provide acceptable fits to the light curve, as in the previous case. However, not all blends corresponding to the BEB, BP, and HTP configurations can be ruled out. We illustrate this in Figure 3. For example, background eclipsing bi- naries that are more than five magnitudes fainter than the target in the Kepler band can match the shape of the light curve just as well as a model of a planet transiting the target, for a wide range of masses of the primary star of the binary between 0.6 M(cid:12) and 1.4 M(cid:12) (left panel). An even wider range of masses is permitted for back- ground stars transited by a larger planet (BP, middle panel). Similarly, small stars physically bound to the target can mimic the light curve closely if transited by a planet of suitable size (right panel, HTP). Our follow-up observations may rule out some fraction of these blends, but not all of them. To compute the expected rates of occurrence of each of 6 Kipping et al. Fig. 2.- Map of the χ2 surface (goodness of fit) for KOI-490.02 for three different blend scenarios, as labeled. Only blends within the solid white contours (darker shading) provide fits to the Kepler light curves that are within acceptable limits (3σ, where σ is the significance level of the χ2 difference compared to a transiting planet model fit; see Fressin et al. 2012). Other concentric colored areas (lighter colors) represent fits that are increasingly worse (4 σ, 5 σ, etc.), which we consider to be ruled out. The hatched green areas indicate regions of parameter space where blended stars can be excluded if they are within 0.(cid:48)(cid:48)43 of the target (half-width of the spectrometer slit), within five magnitudes in brightness (1% relative flux), and have a radial velocity differing from the target by 10 km s−1 or more. In all of the above cases they would have been detected spectroscopically. Blends in the hatched blue areas can also be ruled out because they would be either too red (left) or too blue (right) compared to the measured r− Ks color of KOI-490, by more than three times the measurement uncertainty. Left: BEB scenario. The vertical axis represents the linear distance between the eclipsing binary and the target (DBEB − Dtarg), cast for convenience in terms of the distance modulus difference ∆δ = 5 log(DBEB/Dtarg). The dashed green line shown for reference is the locus of blends of equal apparent Kp brightness as the target. Middle: BP scenarios. As before, only blends that are brighter than the target (below the dashed green line) are able to mimic the light curve. The r − Ks color constraint rules out most of those. Right: HTP scenario. The vertical axis now shows the size of the planet transiting the companion star, in units of Jupiter's radius. All blends of this kind that provide acceptable fits to the light curve are too blue, and are therefore ruled out. Fig. 3.- Similar to Figure 2 for KOI-490.04, using the same color scheme. For this candidate all three scenarios feature blends that cannot be ruled out from the shape of the transits or constraints from our follow-up observations (darker areas not overlapping the hatched regions). The expected frequencies of each of these types of blends are estimated in the text. these scenarios we followed the Monte Carlo procedure described by Torres et al. (2015), in which we simulated large numbers of blends and counted those that satisfy the constraints from BLENDER and the follow-up obser- vations. The technique relies on the number densities of stars in the vicinity of KOI-490 (at Galactic latitude +9.◦4), the estimated rates of occurrence of eclipsing bi- naries and transiting planets of various sizes and orbital periods, and other known properties of binary stars such as the distributions of their periods, eccentricities, mass ratios, etc. Details of these calculations may be found in the work cited above. We obtained estimated frequencies of 4.17 × 10−8 for the BEB scenario, 1.09 × 10−7 for BP, and 4.10× 10−6 for HTP configurations. The total blend frequency is the sum of these, or 4.25× 10−6. While this value may seem small in absolute terms, the a priori rate of occurrence of transiting planets of a given period and size ("planet prior", PL) is also expected to be small. For a secure validation we require here that the "odds ratio" PL/(BEB + BP + HTP) be large enough so that the planet hypothesis is clearly favored over a false pos- itive. Our estimate of the planet prior is 2.02 × 10−3, based on the number of known KOIs of similar size and period as the candidate. The resulting odds ratio is then 475, which corresponds to a confidence level of 99.79% that the signal is due to a bona fide planet. As this ex- ceeds the 3σ significance threshold typically adopted in BLENDER applications, it formally validates KOI-490.04 as a planet. There is, however, an important caveat to make: an implicit assumption for BLENDER is that there is no visible sign of a blend, whereas we know of the pres- ence of the 2.(cid:48)(cid:48)2 companion. Our BLENDER simulations in fact show (Figure 3) that a faint star such as this could well be causing the signal if transited by a larger planet, both as a physical companion to KOI-490 or as a background/foreground interloper. Under these circum- stances the validation with BLENDER is not sufficient as it applies only to unseen sources, and we must seek alter- nate ways of ruling out the 2.(cid:48)(cid:48)2 companion as the cause of the transit signals. This is successfully achieved through the use of asterodensity profiling, as described in §5.2. 5. LIGHT CURVE FITS 5.1. Joint fit to Kepler-167b and Kepler-167c Planets Kepler-167b and Kepler-167c were both vali- dated by Rowe et al. (2014) and new centroid information since that time excludes the possibility of these two ob- jects orbiting the 2.(cid:48)(cid:48)2 companion. This therefore estab- A Transiting Jupiter Analog 7 Priors for the joint fit to KOI-490.01 & KOI-490.03. J is a Jeffreys prior, N is a normal and U is uniform. TABLE 2 Description Global parameters Mean stellar density [kg m−3] Limb darkening coefficient 1 Limb darkening coefficient 2 Log of the blending flux ratio Local parameters Ratio-of-radii Impact parameter Orbital period [d] Time of transit minimum [d] Symbol Prior θ(cid:63) ρ(cid:63) q1 q2 log10 β θP (RP /R(cid:63)) b P τ J [10, 106] U [0, 1] U [0, 1] N (−1.976, 0.036) U [0, 1] U [0, 2] U [ ¯P − 0.1, ¯P + 0.1] U [¯τ − 0.1, ¯τ + 0.1] lishes that Kepler-167b and Kepler-167c orbit the same star, namely the target star Kepler-167, to high confi- dence. Fitting the transit light curve of a planet includes mul- tiple parameters pertaining to the star itself, which may be described by the vector θ(cid:63). The vector θ(cid:63) contains the limb darkening coefficients describing the stellar in- tensity profile and the mean stellar density, ρ(cid:63). Since Kepler-167b and Kepler-167c orbit the same star, we can fit the transit lightcurves of both simultaneously, adopting a global θ(cid:63). By conditioning θ(cid:63) on the data describing both planets, we obtain a higher signal-to- noise measurement of these terms, which in turn leads to somewhat better precision on the local parameters, θP, describing each planet (due to the inter-parameter covariances Carter et al. 2008). In this work, we use the quadratic limb darkening law with the optimal parameterization (q1 & q2) de- scribed by Kipping (2013b). The light curves are gen- erated with the Mandel & Agol (2002) algorithm using 30-point resampling to account for those point which are long-cadence, as described in Kipping (2010). Quarter- to-quarter contamination factors are accounted for, us- ing the "CROWDSAP" header information in the raw fits files via the method described by Kipping & Tinetti (2009). A blending factor affecting all quarters due to the 2.(cid:48)(cid:48)2 companion is also accounted for via this method. For the global blending factor, we convert the J and Ks colors observed in the NIRC2 AO images to a Kepler bandpass magnitude using the fifth-order polynomial re- lation in Appendix A of Howell et al. (2012). Assum- ing either a dwarf or a giant leads to the same result (within the estimated uncertainty) of Kp = 19.0 ± 0.1. Assuming Gaussian errors on the J and Ks colors from AO and adding in quadrature an extra Gaussian uncer- tainty reflecting the 0.05 magnitude error in the Howell et al. (2012) relation, we estimate a blending factor of log β = (−1.976 ± 0.036), where β is the flux ratio be- tween the target and the companion in the Kepler band- pass. This is treated as a normal prior in our fits and added to the θ(cid:63) vector, since it is a term affecting all of the planets. The parameters and priors used are listed in Table 2, with the θ(cid:63) plus two sets of θP parameters giving a total of 12 free parameters in our model. Fits were achieved using the multimodal nested sam- pling algorithm MultiNest (Feroz & Hobson 2008; Feroz et al. 2009) with 4000 live points and a target efficiency set to 0.1. The eccentricities of Kepler-167b Fig. 4.- Posterior distribution of the mean stellar density of Kepler-167 conditioned on the transits of Kepler-167b & c (green) and Kepler-167b, c & d (blue). We only assume that the plan- ets orbit the same star and have circular orbits. For comparison, the posterior derived using isochrone matching of the SPC stellar atmosphere constraints is shown in black. and Kepler-167c are assumed to be zero in the fits. Multi-planet Kepler systems are known to have low ec- centricities, with Van Eylen & Albrecht (2015) finding a Rayleigh distribution with σe = 0.049 ± 0.013 de- scribes the overall population. Moreover, Kepler-167b and Kepler-167c orbit the same star with orbital peri- ods of 4.4 d and 7.4 d, placing them in close proximity to both the star and each other. We therefore expect these planets to be particularly likely to have near-zero eccentricity, from a dynamical perspective. Rather than describe the posteriors found for each pa- rameter, we focus here on the term ρ(cid:63), since the oth- ers will be superseded by the global fits performed later. The posterior distribution for log10(ρ(cid:63)) (it is more ap- propriate to discuss the log since we invoked a Jeffreys prior) yields log10[ρ(cid:63)(kg m−3)] = 3.446+0.034 −0.098 and is plot- ted in Figure 4. This may be compared to the den- sity expected by iscohrone-matching using the effective temperature, metallicity and surface gravity found using SPC. Drawing random samples from three normal dis- tributions describing each and finding the nearest Dart- mouth isochrone each time, we derive a wholly indepen- dent stellar density for Kepler-167 of log10[ρ(cid:63)(kg m−3)] = 3.455+0.015 −0.016. The close agreement between the two is fur- ther evidence that Kepler-167b and Kepler-167c orbit the target star, although their validation (Rowe et al. 2014) was performed independent of this fact. 5.2. Two Fits for KOI-490.04 From the validation discussion in § 4, it was estab- lished that KOI-490.04 does not orbit an unseen com- panion to 3 σ confidence, but may still orbit the seen 2.(cid:48)(cid:48)2 companion. Here, we perform two fits to explore these two hypotheses. ��-��+��-��+��-��-��-��+��-��+��-��-��-��+��-��+��-��-��(ρ★[��-�])��8 Kipping et al. In hypothesis A (HA), we assume that KOI-490.04 or- bits the target star, which we have established also hosts Kepler-167b and Kepler-167c. The posterior distribution of the light curve derived stellar density from the fit in § 5.1 becomes the prior for the same term in this hy- pothesis. Note that this distribution does not invoke any information from the spectroscopic analysis. All of the other parameters retain the same priors listed in Table 2. Specifically, the limb darkening parameters are treated as free again, since these were poorly constrained from the previous fit. For HA, we relax the assumption of a circular or- bit. Since KOI-490.04 is further from both the star and the other two planets (P ∼ 21.8 d), higher eccentricities are possible. In this hypothesis though, KOI-490.04 be- longs to a typical Kepler multi-planet system and thus we adopt the eccentricity distribution derived by Van Eylen & Albrecht (2015); a Rayleigh distribution with σe = 0.049. The prior for the argument of periapsis be- comes increasingly non-uniform as eccentricity diverges from zero, due to a geometric effect (see Kipping 2014a). Nevertheless, a uniform prior is reasonable in this case given the low-eccentricity nature of the Van Eylen & Al- brecht (2015) distribution. In hypothesis B (HB), we assume that KOI-490.04 or- bits the 2.(cid:48)(cid:48)2 companion. Here, the object can no longer be considered to belong to a multi-planet system, since Kepler-167b and Kepler-167c orbit a different star now. Therefore, the potential for high eccentricities becomes even greater and we consider the object to follow a Beta distribution calibrated to the population of radial veloc- ity planets with orbital periods less than one year, as described by Kipping (2013a) (∼ Beta[0.694, 3.252]). In this case, the geometric bias due to the transit probabil- ity is more significant and requires accounting for. We therefore use the ECCSAMPLES code described by Kip- ping (2014a) to modify the Beta prior to a prior con- ditioned on the fact we know this object is transiting. However, rather than directly sampling from the prior, we penalized the likelihood function appropriately to im- prove computational efficiency. Additionally, since the star is different to Kepler-167, the mean stellar density follows an uninformative Jeffreys prior, J [1, 109] kg m−3. Finally, the blending prior is flipped to consider the blend source being Kepler-167. We fitted both models using MultiNest, which re- turns the Bayesian evidence, Z, enabling Bayesian model selection between the two. Note that this is essentially a more advanced treatment of using the photo-blend effect (Kipping 2014b) employed in validating several candi- dates by Torres et al. (2015). Bayesian model selection favors hypothesis A with ∆(log Z) = (5.57±0.10). Given that only two hypothesis exist, the statistical significance of hypothesis A being the preferred model is 2.9 σ. We therefore use the photo-blend effect to show that KOI- 490.04 orbits the target star to 99.6% confidence. However, in hypothesis B, the mean stellar density required to explain the data is log10[ρ(cid:63)(kg m−3)] = 3.49+0.16−0.28. This would make the companion star of sim- ilar spectral type to that of Kepler-167. This essen- tially excludes the possibility of a bound binary, since the AO companion could not be 5 magnitudes fainter in this case. If this were a chance alignment of a back- Fig. 5.- Posterior distribution (solid) of the orbital eccentricity of Kepler-167d, with comparison to the prior (dashed) describing Kepler multis (Van Eylen & Albrecht 2015). The eccentricity is constrained by the comparison of the transit light curve shape of Kepler-167d to the ρ(cid:63) constraint from the earlier joint fit of Kepler- 167b & c. ground star, we would still expect the host star to have a similar color to the primary. However, J − Ks of Kepler-167 is (0.60 ± 0.03) whereas the AO companion has J − Ks = (0.94 ± 0.05). Reddening can not explain such a large difference either and thus we conclude that hypothesis B is even less likely than indicated by the formal Bayesian evidence comparison. We therefore con- clude that KOI-490.04 orbits Kepler-167 to a confidence exceeding 3 σ and thus may be considered a "validated" planet, designated Kepler-167d. Since hypothesis A invokes the posterior of ρ(cid:63) from the earlier joint fit of Kepler-167b & c as a prior, the ec- centricity of Kepler-167d is constrained without any use of stellar evolution or atmosphere models. This demon- strates perhaps the first applied example of Multibody Asterodensity Profiling (MAP), proposed by Kipping et al. (2012). However, we find that the light curve of Kepler-167d is of insufficient data quality to provide a meaningful improvement on the eccentricity constraint over the prior. Specifically, in hypothesis A, the eccen- tricity posterior closely resembles the prior (see Figure 5) and the credible intervals on e change from 0.058+0.036 −0.029 (the prior) to 0.055+0.035 −0.027 (the posterior). 5.3. Parameters and Eccentricity of Kepler-167e Having established that Kepler-167b, Kepler-167c and Kepler-167d orbit the same star, we perform a new global fit adopting a common θ(cid:63) for the parent star. Since the eccentricity of Kepler-167d is inferred to be consis- tent with a low eccentricity prior, we assume the or- bit is nearly circular in this global fit. Including the third planet provides a modest improvement in the con- straint on the mean stellar density, as expected. This may be seen in Figure 4 where the constraint tightens ��-��+��-��+��-��-��A Transiting Jupiter Analog 9 TABLE 3 Comparison of the basic transit parameters of Kepler-167e when epochs 1 & 2 are fitted independently. The last three differ by 0.9, 1.2 & 1.2 σ respectively. Parameter Epoch 1 Epoch 2 τ [BKJDUTC-2,455,000] (RP /R(cid:63)) . . . . . . . . . . . . . . T14 [hours] . . . . . . . . . . . . . T23 [hours] . . . . . . . . . . . . . 253.28698+0.00042 −0.00043 0.1265+0.0014 −0.0011 16.241+0.090 −0.079 12.48+0.14−0.18 1324.51928+0.00044 −0.00045 0.1284+0.0015 −0.0016 16.114+0.070 −0.071 12.15+0.20−0.20 up to log10[ρ(cid:63)(kg m−3)] = 3.460+0.031 −0.065. We use this pos- terior on the mean density as an additional constraint for the fundamental stellar parameters through isochrone matching, as described earlier in § 3.4. The full set of posteriors from this fit are shown in Figure 11 and are available for download at this URL. Using the new revised isochrone modeling and the ratio-of-radii posteriors from this three-planet joint fit, we infer our best-estimate of the radii of planets Kepler- 167b, c & d to be Rb = 1.615+0.047 −0.043 R⊕, Rc = 1.548+0.050 −0.048 R⊕. The maxi- mum likelihood light curve models for the planets in the Kepler-167 system are shown in Figure 6. −0.048 R⊕ and Rd = 1.194+0.049 Note that despite using three transiting planets, the limb darkening coefficients are poorly constrained re- turning only marginally different posteriors from the Specifically, we measure q1 = 0.63+0.26−0.30 and priors. q2 = 0.17+0.18−0.10. We now turn our attention to the outer planet, Kepler- 167e, which was validated earlier to orbit the target star (see § 4.1). We first verified that the period of 1071 d is the correct one, by inspecting the raw light curve around integer ratios of the candidate period. Given the depth of 1.6%, even a simple visual inspection excludes this possibility. As was done by Kipping et al. (2014), we also verified that the shape of the two transit events observed are consistent which is also visually evident in Figure 6. Basic parameters derived from two independent fits of each event reveal that the events are consistent, as shown in Table 3 (note that the first transit is long-cadence and the second, short). Given that Kepler-167e is a much longer orbital period planet than the other three, the potential for an eccen- tric orbit is much higher both a-priori as a member of the long-period planet sample (Kipping 2013a) and dy- namically since it is essentially decoupled from the other three. We treat the posterior distribution for the mean stellar density of the Kepler-167b, c & d joint fit as a prior in the fits of planet e. Although there is a weak constraint on the limb darkening coefficients, we consider the infor- mation too weak to be worth including and thus treat the limb darkening coefficients as independent and free. Therefore the priors largely follow those listed in Table 2, except for ρ(cid:63). It is also crucial to include the orbital eccentricity and argument of periapsis. Given the potential for a large ec- centricity, the geometric bias effect described by Kipping (2014a) becomes pronounced and must be accounted for. As was done in hypothesis B of the KOI-490.04 fits, we use the geometry-corrected Beta prior from Kip- ping (2014a) via a likelihood penalization implementa- tion, adopting Beta shape parameters calibrated to the long-period (> 1 year) radial velocity exoplanet catalog (Kipping 2013a) (specifically ∼ Beta[1.12, 3.09]). Unlike the case of Kepler-167d, the light curve plus stellar density prior does provide a more constraining posterior on the derived eccentricity than the prior. As shown in Figure 7, the orbit is measured to be close to circular, with e = 0.062+0.104 −0.043. Using the posteriors for the fundamental stellar parameters derived using SPC along with the Kepler-167b, c & d ρ(cid:63) constraint plus isochrone matching, we estimate that Kepler-167e is ∼ 10% smaller than Jupiter at 10.15+0.24−0.23 R⊕. The posteriors from the fit of Kepler-167e are shown in Figure 12 and are available at this URL. The median and 68.3% credible intervals for the basic parameters of Kepler-167b, c, d & e are shown in Table 4. 6. DISCUSSION 6.1. A Transiting Jupiter Analog Kepler-167e appears to be the first example of a tran- siting Jupiter analog, as defined by its size (0.91 RJ), low eccentricity (e = 0.06+0.10−0.04) and location beyond the snow-line (see Figure 8). Although Jupiter analogs have been found via other methods (e.g. see Rowan et al. 2015), the geometric biases affecting the transit method make it highly unfavorable for discovering such worlds. Kepler-167e has an a priori transit probability of (cid:39) (cid:39) 3% occurrence rate 0.18%. Combined with the η(cid:88) of Jupiter analogs (Rowan et al. 2015), one should ex- pect O[10] transiting examples to exist amongst the ∼ 200, 000 stars observed by Kepler. However, only those objects just beyond the snow-line and orbiting later than Solar-type stars will have a chance to produce the two transits needed to resolve the orbital period. Con- sequently, it is possible that Kepler-167e could be the only transiting Jupiter analog for which we can precisely measure the period until the next generation of surveys. The fact that Kepler-167e is transiting offers the op- portunity to probe the atmosphere of a genuine Jupiter analog (Dalba et al. 2015), which has thus far been im- possible. Whilst Kepler-167 is relatively faint in the V band at 14.3, the fact that this is a late-type star means the planet may be characterizable toward the near- and mid-infrared bandpasses, where the K-band magnitude is 11.8. This fact, combined with the very deep tran- sit depth of 1.6%, makes atmospheric characterization a challenging, but not impossible, task. As was done by Kipping et al. (2014), we used the Kennedy & Kenyon (2008) predictions for the time- evolving snow-line of a ∼ 0.8 M(cid:12) star to estimate that Kepler-167e's present location corresponds to the snow- line after ∼ 800, 000 years. This time is less than the median lifetimes of protoplanetary disks of Solar-type stars (e.g., Strom et al. 1993; Haisch et al. 2001) and (Yasui et al. 2012), since disk lifetimes scale as M Kepler-167e could have formed at its present location. Unlike Kepler-421b (Kipping et al. 2014), which was "near" the snow-line, Kepler-167e is comfortably beyond it for the majority of the disk lifetime. −1/2 (cid:63) 6.2. Could Kepler-167e be a brown dwarf ? 10 Kipping et al. Fig. 6.- Folded transit light curves of Kepler-167b, Kepler-167c, Kepler-167d and Kepler-167e. For the upper three, data (gray points) are binned to a 10 minute cadence. Light curve of Kepler-167e uses 30 minute binning and uses circles to denote the first transit (Q4) and squares to denote the second transit (Q16). Note that all of the transits were fitted using the original unbinned data. With a radius close to that of Jupiter, Kepler-167e is a member of the so-called "degenerate worlds" class8, where the mass-radius relation is nearly flat (Chen & Kipping 2016). This class encompasses a diverse range of masses, from that of Saturn to the most massive brown dwarfs. On this basis, Kepler-167e's radius is consistent with being either a brown dwarf or a Jovian-like planet. 8 Formally, using the Chen & Kipping (2016) model, we estimate an 80% probability that Kepler-167e is a degenerate world. As argued by Chen & Kipping (2016), the division be- tween brown dwarfs and gas giants is somewhat con- trived, with both belonging to a continuum. Neverthe- less, we evaluate the possibility that Kepler-167e's mass is greater than the 13 MJ canonical threshold (Spiegel et al. 2011) here. Using the radius-to-mass probabilistic forecasting code of Chen & Kipping (2016), our radius samples can be converted to predicted masses. Being a degenerate world, ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○��-��-��-�-�-��-��-��-��-��-��Δ(��)[��]●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○��-��-��-�-�-�-��-��-��-��-��-��Δ(��)[��]●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○��-��-��-��-�-�-�-��-��-��-��-��Δ(��)[��]●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○○■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□□��-��-��-��-��-��-��-��-��-��-��Δ(��)[��]A Transiting Jupiter Analog 11 Final parameter estimates for the planets orbiting Kepler-167. ‡ = fixed; † = assuming a Bond albedo similar to Jupiter of 0.34 (whereas we simply adopt 0 for the other cases due to the uncertainty in what kind of planet they are, as shown in Figure 10). ∗ = equivalent semi-major axis of the planet if it orbited the Sun with e = 0 and insolation level Seff . TABLE 4 Parameter Kepler-167b Kepler-167c Kepler-167d Kepler-167e Fitted parameters log10[ρ(cid:63) (kg m−3)] . . . . . . q1 . . . . . . . . . . . . . . . . . . . . . . q2 . . . . . . . . . . . . . . . . . . . . . . log10 β . . . . . . . . . . . . . . . . . (RP /R(cid:63)) . . . . . . . . . . . . . . . b . . . . . . . . . . . . . . . . . . . . . . . P [days] . . . . . . . . . . . . . . . τ [BKJDUTC-2,455,000] e . . . . . . . . . . . . . . . . . . . . . . . ω [rads] . . . . . . . . . . . . . . . . Other transit parameters (a/R(cid:63)) . . . . . . . . . . . . . . . . . u1 . . . . . . . . . . . . . . . . . . . . . u2 . . . . . . . . . . . . . . . . . . . . . i [◦] . . . . . . . . . . . . . . . . . . . . T14 [hours] . . . . . . . . . . . . . . . . . . . . . . . . . . T23 [hours] T12 (cid:39) T34 [mins] . . . . . . . Physical parameters RP [R⊕] . . . . . . . . . . . . . . . a [AU] . . . . . . . . . . . . . . . . . Teq [K] . . . . . . . . . . . . . . . . . Seff [S⊕] . . . . . . . . . . . . . . . a∗ eff [AU] . . . . . . . . . . . . . . . Mforecast (1 σ interval) . 3.460+0.031 −0.065 0.63+0.26−0.30 0.17+0.18−0.10 −1.976 ± 0.036 0.02036+0.00034 −0.00032 0.17+0.18−0.12 4.3931632+0.0000046 −0.0000045 831.78317+0.00034 −0.00036 0‡ - 3.460+0.031 −0.065 0.63+0.26−0.30 0.17+0.18−0.10 −1.976 ± 0.036 0.01952+0.00042 −0.00044 0.25+0.13−0.13 7.406114+0.000012 −0.000011 552.15774+0.00129 −0.00084 0‡ - 14.33+0.35−0.69 0.63+0.15−0.20 0.14+0.28−0.25 89.33+0.47−0.80 2.350+0.035 −0.035 2.249+0.035 −0.033 2.89+0.342 −0.086 20.30+0.49−0.98 0.63+0.15−0.20 0.14+0.28−0.25 89.30+0.36−0.43 2.746+0.096 −0.061 2.630+0.098 −0.064 3.36+0.37−0.14 1.615+0.047 −0.043 0.0483+0.0017 −0.0025 914+26−16 115.8+13.0−8.0 0.0929+0.0034 −0.0048 2.4M⊕–6.3M⊕ 1.548+0.050 −0.048 0.0684+0.024 −0.0035 768+21−14 57.7+6.5−4.0 0.1316+0.0048 −0.0068 2.2M⊕–5.8M⊕ 3.460+0.031 −0.065 0.63+0.26−0.30 0.17+0.18−0.10 −1.976 ± 0.036 0.01507+0.00050 −0.00052 0.474+0.076 −0.063 21.803855+0.000078 −0.000119 669.7934+0.0015 −0.0018 < 0.12 - 41.7+1.0−2.0 0.63+0.15−0.20 0.14+0.28−0.25 89.352+0.090 −0.140 3.582+0.131 −0.073 3.440+0.137 −0.077 4.09+0.49−0.27 1.194+0.049 −0.048 0.1405+0.0050 −0.0071 536.0+14.4−9.6 13.68+1.54−0.95 0.270+0.010 −0.014 1.2M⊕–2.9M⊕ 3.460+0.031 −0.065 0.452+0.072 −0.063 0.463+0.062 −0.053 −1.976 ± 0.036 0.12810+0.00091 −0.00093 0.233+0.049 −0.068 1071.23228+0.00056 −0.00056 253.28699+0.00039 −0.00040 0.062+0.104 −0.043 3.5+2.6−2.9 560+11−15 0.915+0.020 −0.019 −0.243+0.040 −0.038 89.9760+0.0070 −0.0052 16.13+0.44−0.34 12.29+0.38−0.33 115.9+3.7−3.7 10.15+0.24−0.23 1.890+0.058 −0.067 † 130.9+2.0−3.0 0.0739+0.0047 −0.0091 3.64+0.12−0.13 0.3MJ–50MJ Fig. 7.- Posterior distribution (solid) of the orbital eccentricity of Kepler-167e, with comparison to the prior (dashed) describing long-period transiting planets (Kipping 2014a). The eccentricity is constrained by the comparison of the transit light curve shape of Kepler-167e to the ρ(cid:63) constraint from the earlier joint fit of Kepler-167b, c & d. Fig. 8.- Catalog of known transiting exoplanets with the color depicting the peak wavelength color of the parent star. Solar Sys- tem worlds are shown with black symbols and the Kepler-167 plan- ets with squares. The blue-box depicts Jovian-sized planets beyond the snow-line (∼ 0.25 S⊕), with Kepler-167e being the first tran- siting exoplanet to fill this space. Data comes from the Exoplanet Orbit Database (Wright et al. 2011). this unsurprisingly returns a very broad distribution, with the 68.3% credible interval spanning 0.3 MJ to 50 MJ. We find that the probability of the mass be- ing less than the hydrogen-burning limit to be 97.4%, indicating that Kepler-167e is very unlikely to be a star. Moreover, we find 1.33 times more samples below the 13 MJ threshold than above it, implying a slight prefer- ence for a Jupiter-like object. ��-��+��-��+��-��-��♀♀⊕⊕♂♂♃♃♄♄■■■■□□□□��-��[�⊕]��[�⊕]Jupiter-analogs12 Kipping et al. This estimate can be improved by including a suitable occurrence rate prior, for which we here use Cumming et al. (2008) power-law of occurrence rate ∼ M−0.31. This adds further weight to the Jupiter-like scenario with an odds ratio of 4.28. We therefore estimate that Kepler- 167e is four times more likely to be a Jupiter-like planet than a brown dwarf. It may be possible for observers exclude the brown dwarf hypothesis using radial velocities. If Kepler-167e is a brown dwarf, then the radial velocity amplitude would be K ≥ 31619−11 m s−1. In contrast, the (broad) forecasted radial velocity amplitude is log10[K (m s−1)] = 2.32+0.75−1.47 (i.e. ∼ 211 m s−1). 6.3. Multibody Asterodensity Profiling The eccentricity of Kepler-167e is measured purely us- ing the transit shapes of the orbiting planets, represent- ing a first for the field. In all previous cases, independent information constraining the mean stellar density was used, such as spectroscopic + isochrone analysis (Daw- son & Johnson 2012), asteroseismology (Sliski & Kipping 2014) or flicker (Kipping et al. 2014). The eccentricity of a transiting planet can be measured using asterodensity profiling (Kipping 2014b), specifi- cally via the photo-eccentric effect (Dawson & Johnson 2012). This essentially compares the light curve derived stellar density (related to the T14 and T23 transit du- rations) to that derived via some independent method. Although eccentricity is the dominant effect, for Kepler- 167 the photo-blend effect is in play too, due to the AO detected companion. Whilst the most common and accessible method to get an independent mean stellar density is spectroscopy com- bined with isochrone modeling, this approach essentially makes the unrealistic assumption of zero-model error. Multibody Asterodensity Profiling (Kipping et al. 2012) was conceived with the idea of comparing the light curves of planets orbiting the same star against one another, to obviate the need to ever go through evolutionary mod- els. In the case of multiple eccentric planets, the inverse problem is quite challenging but the compact, inner three planets of Kepler-167 are likely on near-circular orbits, providing a so-called "stellar anchor" we can use to char- acterize the star. This inference is then used to measure the eccentricity of the outer planet, which a priori could be much more eccentric. We are able to measure the eccentricity to be e = 0.06+0.10−0.04, which further supports the case that Kepler-167e is a Jupiter analog. Critically, we empha- size that this measurement used nothing more than the Kepler photometric time series of a 14th magnitude star and a single night of AO imaging on a 10 m telescope. As a comparison, HD 32963b is a recently discovered Jupiter analog found using radial velocities (Rowan et al. 2015) for which 199 nights of precise radial velocities on a 10 m class telescope led to the comparable constraint of e = (0.07 ± 0.04), in spite of the fact that HD 32963 is six and a half magnitudes brighter than Kepler-167. However, the transit method does have the major drawback that transiting Jupiter analogs are far less numerous than their non-transiting counterparts. Fig. 9.- Schematic illustrating the scale of the Kepler-167 sys- tem. Planet sizes are scaled relative to the key, rather than the or- bital distances in order to make them visible. The four known plan- ets display remarkable coplanarity and near-circular orbits with the habitable-zone (Kopparapu et al. 2013) notably devoid of transit- ing planets. 6.4. System Architecture All three inner planets orbit interior to the inner edge of the habitable-zone (Kopparapu et al. 2013) and are unlikely to be interesting from an astrobiological per- spective. The architecture of the Kepler-167 system is curious, with a compact multi followed by a large cavity of transiting planets and then an outer Jupiter analog (see Figure 9). Thus, Kepler-167 resembles a fusion of the compact Kepler multis and the classic Solar System. Since transit surveys have a very poor sensitivity to long- period planets like Kepler-167e (planet yield ∼ P −5/3; Beatty & Gaudi 2008), it is plausible that Kepler-167e- like planets are found frequently in the Kepler compact multis. A radial velocity survey targeting the bright Ke- pler multis would be able to resolve this question. Additional non-transiting planets could reside in the Kepler-167 cavity, although the known four planets dis- play remarkable coplanarity and low eccentricities, sug- gestive of a dynamically cold system. Amongst the inner planets, the planet sizes increase as one approaches the parent star. Using the Chen & Kipping (2016) mass- radius model, we estimate that the inner two planets are most likely gaseous worlds whilst the outer planet is most likely rocky (see Figure 10). Whilst this pattern ostensi- bly jars our anthropocentric prior, as well as the expected outcome of photo-evaporation (Lopez & Fortney 2013), Ciardi et al. (2013) find that there is no preferential or- dering of compact Kepler multis for planets R (cid:46) 3R⊕. The Kepler-167 system teases the possibility that com- pact multis may plausibly harbor distant Jupiter analogs, inviting the community to pursue this question with cur- rent and future facilities. Moreover, whilst Kepler-167 is not a bright star, transiting Jupiter analogs represent an important new class of targets in the on-going cam- 1.89AU1.39 AU0.91RJ= 1 R⊕ed=0.06-0.04+0.10snowline@ 2 MyrKepler-167eA Transiting Jupiter Analog 13 Discovering members of this population around brighter stars is a challenging but likely rewarding task for future missions. ACKNOWLEDGEMENTS This work made use of the Michael Dodds Comput- ing Facility and the Pleiades supercomputer at NASA Ames. GT acknowledges partial support for this work from NASA grant NNX14AB83G (Kepler Participating Scientist Program). DMK acknowledges partial support from NASA grant NNX15AF09G (NASA ADAP Pro- gram). This research has made use of the Exoplanet Orbit Database and the Exoplanet Data Explorer at ex- oplanets.org, and the corner.py code by Dan Foreman- Mackey at github.com/dfm/corner.py. We offer our thanks and praise to the extraordinary scientists, engi- neers and individuals who have made the Kepler Mis- sion possible. Finally, the authors wish to extend spe- cial thanks to those of Hawaiian ancestry on whose sacred mountain of Mauna Kea we are privileged to be guests. Without their generous hospitality, the Keck observa- tions presented herein would not have been possible. Fig. 10.- Forecasted masses for the planets Kepler-167b (green), Kepler-167c (turquoise) and Kepler-167d (blue) using our radii pos- terior samples and the radius-to-mass forecasting model of Chen & Kipping (2016). Gray region denotes the 1 σ confidence interval of the transition from rocky to gaseous worlds found by Chen & Kipping (2016). paigns to characterize the atmospheres of alien worlds. REFERENCES Akeson, R. L., Chen, X., Ciardi, D. et al. 2013, PASP, 125, 989 Ballard, S., Charbonneau, D., Fressin, F. et al. 2013, ApJ, 773, 98 Barclay, T., Rowe, J. F., Lissauer, J. J. et al. 2014, Nature, 494, 452 Batygin, K. & Laughlin, G. 2015, PNAS, 112, 4214 Beatty, T. G. & Gaudi, S. B., 2008, ApJ, 686, 1302 Brown, T. M., Latham, D. W., Everett, M. E., & Esquerdo, G. A., AJ, 142, 112 Bryson, S. T., Jenkins, J. M., Gilliland, R. L., et al. 2013, PASP, 125, 889 Buchhave, L. A., Latham, D. W., Anders, J. et al. 2012, Nature, 486, 375 Carter, J. A., Yee, J. C., Eastman, J. et al., 2008, ApJ, 689, 499 Chen, J. & Kipping, D. M., 2016, submitted Ciardi, D. R., Fabrycky, D. C., Ford, E. B., Gautier, T. N., Howell, S. B., Lissauer, J. J., Ragozzine, D. & Rowe, J. F. 2013, ApJ, 763, 41 Cumming, A., Butler, R. P., Marcy, G. W., Vogt, S. S., Wright, J. T. & Fischer, D. A. 2008, PASP, 120, 531 Dalba, P. A., Muirhead, P. S., Fortney, J. J., Hedman, M. M., Nicholson, P. D., Veyette, M. J. 2015, ApJ, 814, 152 Dawson, R. I. & Johnson J. A. 2012, ApJ, 756, 122 Dotter, A., Chaboyer, B., Darko, J., Veselin, K., Baron, E. & Ferguson, J. W. 2008, ApJS, 178, 89 Doyle, L. R., Carter, J. A., Fabrycky, D. C. et al. 2011, Science, 333, 1602 Dressing, C. D. & Charbonneau, D. 2015, ApJ, 807, 45 Everett, M. E., Barclay, T., Ciardi, D. R., et al. 2015, AJ, 149, 55 Feroz, F. & Hobson, M. P., 2008, MNRAS, 384, 449 Feroz, F., Hobson, M. P. & Bridges, M., 2009, MNRAS, 398, 1601 Foreman-Mackey, D., Hogg, D. W.; Morton, T. D. 2014, ApJ, 795, 64 Fortney, J. J., Demory, B.-O., D´esert, J.-M. et al. 2011, ApJS, 197, 9 Fressin, F., Torres, G., Rowe, J. F. et al. 2012, Nature, 482, 195 Gould, A., Dong, S., Gaudi, S. B. et al. 2010, ApJ, 720, 1073 Haisch, Jr., K. E., Lada, E. A. & Lada, C. J., 2001, ApJ, 553, L153 Horch, E. P., Howell, S. B., Everett, M. E. & Ciardi, D. R. 2014, ApJ, 795, 60 Horner, J., Jones, B. W. & Chambers, J. 2010, International Journal of Astrobiology, 9, 1 Howard, A. W., Johnson, J. A., Marcy, G. W. et al. 2010, ApJ, 721, 1467 Howell, S. B., Rowe, J. F., Bryson, S. T. et al., 2012, ApJ, 746, 123 Jenkins, J. M., Twicken, J. D., Batalha, N. M., et al. 2015, AJ, 150, 56 Johansson, E. M., van Dam, M. A., Stomski, P. J. et al. 2008, Proc. SPIE, 7015, 91 Johnson, J. A., Howard, A. W., Bowler, B. P. et al. 2010, PASP, 122, 701 Kennedy, G. M. & Kenyon, S. J., 2008, ApJ, 673, 502 Kipping, D. M., 2010, MNRAS, 408, 1758 Kipping, D. M., 2013a, MNRAS, 434, L51 Kipping, D. M. 2013b, MNRAS, 435, 2152 Kipping, D. M. 2014a, MNRAS, 444, 2263 Kipping, D. M., 2014b, MNRAS, 440, 2164 Kipping, D. M. & Tinetti, G. 2013, MNRAS, 407, 2589 Kipping, D. M., Dunn, W. R., Jasinski, J. M. & Manthri, V. P. 2012, MNRAS, 421, 1166 Kipping, D. M., Hartman, J., Buchhave, L. A., Schmitt, A., Nesvorn´y, D. & Bakos, G. ´A., 2013, ApJ, 770, 101 Kipping, D. M., Torres, G., Buchhave, L. A., et al. 2014, ApJ, 795, 25 Kolbl, R., Marcy, G. W., Isaacson, H. & Howard, A. W. 2015, AJ, 149, 18 Kopparapu, R. K., Ramirez, R., Kasting, J. F. et al., 2013, ApJ, 765, 131 Law, N. M., Morton, T., Baranec, C., et al. 2014, ApJ, 791, 35 Lawrence, A., Warren, S. J., Almaini, O. et al. 2007, MNRAS, 379, 1599 Lillo-Box, J., Barrado, D. & Bouy, H. 2012, A&A, 546, A10 Lissauer, J. J., Marcy, G. W., Rowe, J. F., et al. 2012, ApJ, 750, 112 Lissauer, J. J., Marcy, G. W., Bryson, S. T., et al. 2014, ApJ, 784, 44 Lopez, E. D. & Fortney, J. J. 2013, ApJ, 776, 2 Mandel, K. & Agol, E., 2002, ApJ, 580, 171 Meibom, S., Torres, G., Fressin, F. et al. 2013, Nature, 499, 55 Morbidelli, A., Tsiganis, K., Crida, A., Levison, H. F. & Gomes, R. 2007, AJ, 134, 1790 Petigura, E. A., Howard, A. W. & Marcy, G. W. 2013, PNAS, 110, 19273 Pinsonneault, M. H., An, D., Molenda- Zakowicz, J., et al. 2012, ApJS, 199, 30 Quinn, S. N., White, T. R., Latham, D. W. et al. 2015, ApJ, 803, 49 Rowan, D., Meschiari, S., Laughlin, G. et al., 2015, ApJ, 817, 104 Rowe, J. F., Bryson, S. T., Marcy, G. W., et al. 2014, ApJ, 784, 45 Santerne, A., Fressin, F., D´ıaz, R. F., et al. 2013, A&A, 557, A139 ��(��)=��%��(��)=��%��(��)=��%��(��)=��%��(��)=��%��(��)=��%��[�⊕]��14 Kipping et al. Sliski, D. H. & Kipping, D. M. 2014, ApJ, 788, 148 Spiegel, D. S., Burrows, A. & Milsom, J. A. 2011, ApJ, 727, 57 Strom, S. E., Edwards, S. & Skrutskie, M. F., 1993, in Protostars and Planets III, ed. E. H. Levy & J. I. Lunine, University of Arizona Press, Tucson, 837866 Torres, G., Konacki, M., Sasselov, D. D. & Jha, S. 2004, ApJ, 614, 979 Torres, G., Winn, J. N. & Holman, M. J. 2008, ApJ, 677, 1324 Torres, G., Fressin, F., Batalha, N. M. et al. 2011, ApJ, 727, 24 Torres, G., Kipping, D. M., Fressin, F., et al. 2015, ApJ, 800, 99 Van Eylen, V. & Albrecht, S. 2015, ApJ, 808, 126 Vogt, S. S., Allen, S. L., Bigelow, B. C. et al. 1994, Proc. SPIE, 2198, 362 Walsh, K. J., Morbidelli, A., Raymond, S. N., O'Brien, D. P. & Mandel, A. M. 2011, Nature, 475, 206 Wittenmyer, R. A., Tinney, C. G., O'Toole, S. J. et al. 2011, ApJ, 727, 102 Wittenmyer, R. A., Butler, R. P., Tinney, C. G. et al., ApJ, accepted (arXiv e-print:1601.054657) Wizinowich, P. L., Le Mignant, D., Bouchez, A. et al. 2004, Proc. SPIE, 5490, 1 Wright, J. T., Fakhouri, O., Marcy, G. W. et al., 2011, PASP, 123, 412 Yasui, C., Kobayashi, N., Tokunaga, A. T. & Saito, M., 2012, AAS Meeting #219, #439.06 APPENDIX:POSTERIOR DISTRIBUTIONS A Transiting Jupiter Analog 15 Fig. 11.- Triangle plot of the posterior distributions of the 16 parameters explored in the joint fit of Kepler-167b, c & d. Contours mark the 0.5, 1.0, 1.5 & 2.0 σ confidence intervals and dashed lines on the histograms mark the median and surrounding 1 σ confidence interval. Posteriors may be downloaded at this URL. logρq1q2logβpbbbPbτbpcbcPcτcpdbdPdτd16 Kipping et al. Fig. 12.- Triangle plot of the posterior distributions of the 10 parameters explored in the fit of Kepler-167e. Contours mark the 0.5, 1.0, 1.5 & 2.0 σ confidence intervals and dashed lines on the histograms mark the median and surrounding 1 σ confidence interval. Posteriors may be downloaded at this URL. logρq1q2logβpebePeτeeeωeA Transiting Jupiter Analog 17 Fig. 13.- Triangle plot of the posterior distributions of fundamental stellar parameters using SPC plus Dartmouth isochrones. Contours mark the 0.5, 1.0, 1.5 & 2.0 σ confidence intervals and dashed lines on the histograms mark the median and surrounding 1 σ confidence interval. Posteriors may be downloaded at this URL. Teff[Fe/H]MRlogglogLAge
1605.04066	1	1605	2016-05-13T07:14:05	Blueshifted [OI] lines from protoplanetary discs: the smoking gun of X-ray photoevaporation	[ "astro-ph.EP", "astro-ph.SR" ]	Photoevaporation of protoplanetary discs by high energy radiation from the central young stellar object is currently the favourite model to explain the sudden dispersal of discs from the inside out. While several theoretical works have provided a detailed pictured of this process, the direct observational validation is still lacking. Emission lines produced in these slow moving protoplanetary disc winds may bear the imprint of the wind structure and thus provide a potential diagnostic of the underlying dispersal process. In this paper we primarily focus on the collisionally excited neutral oxygen line at 6300A. We compare our models predictions to observational data and demonstrate a thermal origin for the observed blueshifted low-velocity component of this line from protoplanetary discs. Furthermore our models show that while this line is a clear tell-tale-sign of a warm, quasi-neutral disc wind, typical of X-ray photoevaporation, its strong temperature dependence makes it unsuitable to measure detailed wind quantities like mass-loss-rate.	astro-ph.EP	astro-ph	Mon. Not. R. Astron. Soc. 000, 1 -- 7 (2011) Printed 8 March 2021 (MN LATEX style file v2.2) Blueshifted [OI] lines from protoplanetary discs: the smoking gun of X-ray photoevaporation. Barbara Ercolano1,2(cid:63), James E. Owen3,4 1Universitats-Sternwarte Munchen, Scheinerstr. 1, 81679 Munchen, Germany 2Excellence Cluster Origin and Structure of the Universe, Boltzmannstr.2, 85748 Garching bei Munchen, Germany 3Institute for Advanced Study, Einstein Drive, Princeton, NJ 08540, USA 4Hubble Fellow 8 March 2021 ABSTRACT Photoevaporation of protoplanetary discs by high energy radiation from the cen- tral young stellar object is currently the favourite model to explain the sudden dis- persal of discs from the inside out. While several theoretical works have provided a detailed pictured of this process, the direct observational validation is still lacking. Emission lines produced in these slow moving protoplanetary disc winds may bear the imprint of the wind structure and thus provide a potential diagnostic of the underlying dispersal process. In this paper we primarily focus on the collisionally excited neutral oxygen line at 6300A. We compare our models predictions to observational data and demonstrate a thermal origin for the observed blueshifted low-velocity component of this line from protoplanetary discs. Furthermore our models show that while this line is a clear tell-tale-sign of a warm, quasi-neutral disc wind, typical of X-ray photoevap- oration, its strong temperature dependence makes it unsuitable to measure detailed wind quantities like mass-loss-rate. Key words: protoplanetary discs 1 INTRODUCTION Understanding disc dispersal is a key piece in the puzzle of planet formation as it sets the timescale over which (gas giant) planet formation must occur. Furthermore the sim- ilarity between the observed timescales over which Young Stellar Objects (YSOs) lose their disc and the theoretically estimated timescale for the formation of planets suggests that the two processes are probably coupled and feed back on each other. The final build up of gaseous planets occurs in discs on their last gasp before dispersal, and the dispersal process itself can lead to rapid evolution of young planets (Owen & Wu, 2016), making evolved discs all the more in- teresting to study. A popular model to drive disc dispersal is photoevapo- ration by radiation from the central star (e.g. Clarke et al. 2001). The exact nature of the driving radiation is however still open to debate. (Extreme and Far) Ultraviolet (UV) radiation as well as X-ray radiation has been shown to be able to drive winds from the disc upper layers (Alexander et al. 2006; Gorti, Hollenbach & Dullemond 2009; Ercolano et al. 2008, 2009; Owen et al. 2010) that are efficient enough to (cid:63) E-mail: [email protected] (BE) c(cid:13) 2011 RAS disperse the discs in the observed timescales. However both the location and intensity of the wind depend strongly on the driving radiation, with differences of more than two or- ders of magnitude for mass loss rates predicted by different models. Thus these different models obviously have profound implications for disc evolution and hence for the formation of planets and their subsequent evolution (e.g. Ercolano & Rosotti 2015). The presence of a warm, at least partially, ionised disc wind has been confirmed via the observation of a few km/s blue-shift in the profile of the [NeII] 12.8µm fine structure line (Pascucci et al. 2007). Unfortunately modelling of this line cannot shed light on the radiation source which drives the wind, due to the different routes to formation of the Ne+ ion which have different efficiencies in a fully ionised EUV-driven wind and in a quasi-neutral X-ray driven wind. Ne+ formation occurs via the removal of a valence electron from Ne atoms in the fully-ionised winds driven by EUV radiation. In an X-ray driven wind Ne+ is predominantly produced via charge exchange of Ne2+ with neutral H atoms, which are abundant in the quasi-neutral winds driven by X- rays. Atomic physics thus conspires to the result that both an EUV- and an X-ray driven wind, whose mass-loss rates differ by over two orders of magnitude, can equally well fit 2 Ercolano, Owen the observations (Ercolano & Owen 2010; Alexander 2008; Pascucci et al. 2011). Forbidden lines from low-ionisation and atomic states of common elements may present an alternative way to study the wind dispersal mechanism. In particular a few km/s blueshift has been measured in the profiles of, for exam- ple, the [OI] 6300, [O I] 5577, [S II] 6731, and [N II] 6583 lines (e.g. Hartigan et al. 1995, White & Hillenbrand 2004, Mohanty et al. 2005, Rigliaco et al. 2013, Natta et al. 2014). The profiles are often double-peaked, with one component blue-shifted to a few hundred km/s (high velocity compo- nent, HVC) and a second component typically blue-shifted by only a few km/s (low velocity component, LVC), typical of a photoionised wind. The HVC is generally attributed to emission from a dense outflows closer to the star or jets (Har- tigan et al. 1995). The collisionally excited neutral oxygen line at 6300A, in particular, gained significant attention for its potential to discriminate between an EUV- and an X-ray driven wind. EUV-driven winds are, by construction, fully ionised and cannot match the observed luminosities of the LVCs of the [OI] 6300 line (Font et al. 2004). X-ray winds, on the other hand, are only weakly ionised and warm enough to produce the [OI] 6300 line. Ercolano & Owen (2010, EO10) calculated synthetic spectra from the X-ray photoevaporation models of Owen et al. (2010,2011), providing an atlas of atomic and low-ionising emission lines, including the [OI] 6300 line. The line intensi- ties and profiles predicted by EO10 were in agreement with the observations available at the time, thus suggesting that the blue-shifted LVC of the [OI] 6300 line is produced via collisional excitation of neutral oxygen atoms with electrons and hydrogen atoms in the slow-moving photoevaporative winds driven by X-ray radiation. Later observational studies by Rigliaco et al. (2013) and Natta et al. (2014) showed some significant inconsistencies between the model predictions of the [OI] 6300 line by EO10 and the new data. This led Rigliaco et al. (2013) to suggest a non-thermal origin for the [OI] 6300 line (see also Gorti et al. 2011). Natta et al. (2014), on the other hand, still argue against a non-thermal origin of the OI lines, based on the simultaneous presence of the [SII] 4068 line in their spectra. In this paper we revisit the theoretical models in light of these later observations and show that the [OI] 6300 data can be indeed explained in the context of an X-ray driven photoevaporative wind and confirm a thermal origin for this line. We show that the over-simplistic scaling of the illumi- nating flux in the work by EO10 is to blame for the mislead- ing conclusions on the origin of the [OI] 6300. In Section 2 we present the methods employed in our study. In Section 3 we describe the modelling strategy and present the results. A final discussion and our conclusions are given in Section 4. 2 METHODS 2.1 X-ray photoevaporative wind structure We use the set of wind solutions (density and velocity distri- bution of gas in the wind) for primordial discs (i.e. gas-rich, optically thick discs, which do not have an evacuated inner cavity) calculated by Owen et al. (2010,2011) and EO10 for a 0.7 M(cid:12) star and X-ray luminosities (0.1 keV (cid:54) hν (cid:54) 10 keV) of LX = 2 × 1028, 2 × 1029 and 2 × 1030 erg/sec. These were obtained by means of two-dimensional hydrodynamic calculations using the zeus code (Stone et al. 1992a,b,c; Hayes et al. 2006), modified to include the effects of X- ray irradiation with a parametrisation of the gas temper- ature as a function of the local ionisation parameter. The dust radiative transfer and photosionisation code mocassin (Ercolano et al. 2003, 2005, 2008b), modified according to Ercolano et al (2008a), was used produce the temperature parametrisation. The atomic database of the mocassin code included opacity data from Verner et al. (1993) and Verner & Yakovlev (1995), energy levels, collision strengths and transition probabilities from Version 5.2 of the CHIANTI database (Landi et al. 2006, and references therein) and hy- drogen and helium free-bound continuous emission data of Ercolano & Storey (2006). The ionising spectrum used to calculate the temperature parametrisation was calculated by Ercolano et al (2009a), using the plasma code of Kashyap & Drake (2000) from an emission measure distribution based on that derived for RS CVn type binaries by Sanz-Forcada et al. (2002), which peaks at 104 K and fits to Chandra spectra of T-Tauri stars by Maggio et al. (2007), which peaks at around 107.5 K. This spectrum has a significant EUV component (13.6 eV (cid:54) hν (cid:54) 0.1keV ), with roughly LEU V = LX . Solar abundances (Asplund et al. 2005), de- pleted according to Savage & Sembach (1996) were as- sumed, namely (number density, with respect to hydrogen): He/H = 0.1, C/H = 1.4 × 104, N/H = 8.32 × 105, O/H = 3.2 × 104, N e/H = 1.2 × 104, M g/H = 1.1 × 106, Si/H = 1.7×106, S/H = 2.8×105. More details about the codes and setup of the models can be found in Ercolano et al. (2008a, 2009a) and Owen et al. (2010). The hydrodynamical calculations where performed in spherical co-ordinates with a domain spanning [0, π/2] in the θ direction and [rin, rout] in the radial direction, with rout set to 100 AU. In the calculations taken from Owen et al. (2010,2011) rin was set to 0.33 AU and a resolution of 100 uniformly spaced cells in the angular direction and 250 non-uniformly spaced cells in the radial was used. As we shall discuss later in order to asses the role of [OI] emission from the bound inner disc (R < 1 AU) we perform a new set of hydrodynamical calculations, this time with rin=0.04 AU with a resolution of 256 uniformly spaced cells in the angular direction and 384 non-uniformly spaced cells in the radial direction. Since the wind is launched from approximately 1 AU in the calculations, a smaller inner boundary did not effect the dynamics it just gave the hydrostatic density and temperature structure of the inner disc which we could use to calculate the [OI] emissivities. In all cases the simulations were run for at least 10 dynamical time-scales at the outer boundary until a steady-state was achieved (see discussion in Owen et al. 2010). 2.2 Photoionisation calculations Following the approach in EO10, we perform photoionisa- tion calculations of the wind structures with the aim of pre- dicting the intensity and spectral profile of the collisionally excited neutral hydrogen line at 6300A and compare it with the observational results of Rigliaco et al. (2013). The mo- cassin code is employed for this task, with exactly the same c(cid:13) 2011 RAS, MNRAS 000, 1 -- 7 settings as described in the previous section. The ionising spectrum we use is the same as described above for the X- ray region, and used by EO10, additionally we consider here a softer spectral component due to accretion on the young stellar object (YSO). The latter is approximated as a black- body of temperature 12000K and luminosity expressed as a multiple of the stellar bolometric luminosity, Lacc = a × Lbol, with a ranging from 10 to 10−5. It is important to note that since the wind itself is driven by the X-rays (in the range 0.1-1 keV), modifying the soft UV spectrum will not effect the dynamics of the wind itself (Owen et al. 2012), and it is thus not necessary to run new wind simulations for the purpose of this work. 2.3 Emissivity and line profile calculations We have used our two-dimensional map of emissivities and gas velocities obtained from mocassin to reconstruct a three-dimensional cube of the disc and calculate the line- of-sight emission profiles for several of the [OI] transitions. The emissivities are calculated using 120 logarithmically spaced radial points (NR=120)1 and 1500 logarithmically spaced height points (NZ =1500). By assuming azimuthal and reflection symmetry the resultant 3D grid has dimen- sions NR × 2NZ × Nϕ, where we adopt a values of Nϕ = 400 which was sufficient to resolve the line profiles. For the [OI] lines the disc atmosphere is optically thin and the contri- bution to the line can escape freely, provided that the line of sight does not intercept the disc mid-plane which is com- pletely optically thick due to dust (c.f. EC10). We neglect attenuation due to dust in the disc's atmosphere, this will only effect the results for the largest inclinations. The line luminosity is then computed by including a Doppler broadening term in each cell. Thus, the luminosity at a given velocity u is computed using numerical integration by direct summation, taking the emissivities and velocity to be constant in each cell. Such that the line luminosity at a given velocity L(u) is given by: (cid:19) (cid:90) L(u) = d3r (cid:18) (cid:96)(r)(cid:112)2πvth(r)2 exp − [u − ulos(r)]2 2vth(r)2 Blueshifted OI lines from PPDs 3 3 STRATEGY AND RESULTS On the basis of their observations Rigliaco et al. (2013) high- lighted a number of discrepancies with the models of EO10, which argued against a thermal origin of the [OI] 6300 line in an X-ray driven photoevaporative wind, as suggested by EO10. In particular, in contrast to the models of EO10, the observations showed: (1) no correlation between the [OI] lu- minosity, L[OI], and the X-ray luminosity, LX , (2) a corre- lation of L[OI] with the FUV luminosity LF U V , (3) higher full width half maximum (FWHM) of the [OI]6300 line and (4) a lower [OI]6300/[OI]5577 ratio. We will show here that the above discrepancies can be fully explained by considering the assumptions made by EO10, with regards to the ionising spectrum of the central Young Stellar Object (YSO). As described above, the illu- minating spectrum used by EO10 extends to the EUV re- gion, with LEU V = LX . In order to explore the relation be- tween of L[OI] and LX , EO10 scaled their entire spectrum by the same factor, hence also increasing/decreasing the LEU V reaching the X-ray driven wind. We suggest here that the L[OI]-LX correlation reported by EO10 is actually the L[OI]- LEU V correlation, resulting from the homogeneous scaling of the input spectrum at all wavelengths. We demonstrate this here by performing the numerical experiment of decoupling the UV and the X-ray regions of the illuminating spectrum, where we define X-ray/EUV energies higher/lower than 0.1 keV, respectively. Practically we include an additional in- put spectrum in the form of a blackbody of temperature 12000K, loosely representative of an accretion component, whose luminosity can be scaled independently from the X- ray luminosity. We then set up a grid of models at constant Lacc and vary only LX and a grid of models at constant LX and vary only Lacc. The models are summarised in Table 1 and the results are described in the next section. For the figures presented in this section, we used all the observational data reported in the Rigliaco et al. (2013) study, which was partially based on re-analysis of previous data by Hartigan et al. (1995). 3.1 Models at constant accretion luminosity: no (1) correlation with LX where (cid:96)(r) is the volume averaged power emitted at a point r, ulos is the projected gas velocity along the line of sight and vth is the local rms velocity of the emitting atom. The lines were computed with a velocity resolution of 0.25 km s−1. The lines were then degraded to an instrumental resolution of R = 25, 000 -- i.e. the resolution of the Hartigan et al. (1995) and Rigliaco et al. (2013) study -- and R = 50, 000 a higher resolution representative of a future study. The degra- dation was performed by convolving the line profiles with a Gaussian profile of the appropriate width. The line profiles were calculated for disc inclinations of 0 to 90 degrees at 10 degree intervals. EO10 pubilshed [OI]6300 line intensities obtained from their X-ray photoevaporative wind model of a 0.7 M(cid:12) star with LX = 2 × 1028, 2 × 1029 and 2 × 1030 erg/sec. The resulting L[OI] showed a near linear correlation with LX (filled dia- monds in Figure 1), which is not seen in the observations of Rigliaco et al. (2013) (empty symbols in Figure 1). As men- tioned already in the previous section, we show here that this apparent correlation is driven by the LEU V -LX correlation in the scaling of the illuminating spectra in the EO10 model. Indeed, as shown by the filled circles in Figure 1, our models with constant accretion luminosity where we vary LX only do not show any correlation at all between L[OI]-LX . 3.2 Models at constant X-ray luminosity: correlation with Lacc 1 Note throughout this paper we use {r, θ, φ} and {R, ϕ, z} to distinguish between spherical and cylindrical polar co-ordinates respectfully. The observations of Rigliaco et al. (2013) showed a clear relation between the L[OI] with the FUV luminosity LF U V . We suggest that the observed correlation is explainable in c(cid:13) 2011 RAS, MNRAS 000, 1 -- 7 4 Ercolano, Owen Figure 1. [O I] 6300 luminosity versus X-ray luminosity for the subsample of 21 Sample II objects from Rigliaco et al. (2013) (empty diamonds and triangles, where triangles denote upper limits in X-ray luminosity), compared to the model predictions of EO10 (filled diamonds) and those from this work with con- stant accretion luminosity and suppressed chromospheric emis- sion <100eV (filled circles). See section 3.1 for details. Figure 3. FWHM versus sine of the disk inclination from Rigli- aco et al. (2013) (empty diamonds) compared to the model pre- dictions for the LX = 2 × 1030 erg/sec model of EO10 (filled di- amonds) and those from the LX = 2 × 1030 erg/sec, Lacc = Lbol model from this work (filled circles). All model results shown here are for a spectral resolution of R=25000. See section 3.3 for de- tails. that the observed correlation is reproduced for a constant X-ray luminosity, and by varying only the input accretion luminosity. Both the Rigliaco et al. (2013) observations and the models seem to show that the correlation flattens as the accretion luminosity (i.e. the UV flux reaching the wind) decreases. In the models this is due to the [OI] 6300 line luminosity reaching the floor value set by the X-ray illumi- nation. However, Figure 2 also includes the observational sam- ple in the Lupus star forming region from Natta et al. (2014), represented by the crosses. These data extends to much lower accretion luminosities and includes mostly late M stars with masses typically around 0.2 M(cid:12), which is the median of their sample. The Lupus data does not show a flattening in the correlation. It is important to note at this point that our models are not appropriate for a comparison at such low masses, as they were computed for a stellar mass of 0.7 M(cid:12) (note that only 7 out of 44 stars in the Natta et al., 2014, sample have masses larger than 0.5 M(cid:12)). The X-ray prop- erties of such low mass stars are also not very well known. A recent attempt at characterisation of the X-ray properties in the TW Hya region by Kastner et al. (2016), found that log (LX /Lbol) decreases for stars with spectral type M4 or later, and its distribution broadens. This study, while suffer- ing from poor statistics for stars later than M3, also showed that the later type stars had more long-lived discs, probably due to their less efficient X-ray irradiation. A new modeling campaign to cover this new parameter space is under way and will be the focus of a future work. 3.3 The FWHM of the [OI]6300: a consequence of the emission region A further discrepancy between the model predictions of EO10 and the observational results of Rigliaco et al. (2013) c(cid:13) 2011 RAS, MNRAS 000, 1 -- 7 Figure 2. [O I] 6300 luminosity versus accretion luminosity for the Samples I and II objects from Rigliaco et al. (2013) (empty di- amonds) and the Lupus sample from Natta et al. (2014) (crosses) compared to the model predictions from this work with constant X-ray luminosity (filled circles). Emission from the X-ray spec- trum at <100eV is not suppressed in these models. See section 3.2 for details. terms of the correlation between the emission region of the collisionally excited [OI] line at 6300A and the EUV flux reaching the wind. As the EUV luminosity is dominated by the accretion luminosity of the YSO, we show this point here by running a set of models of constant X-ray luminosity (LX = 2 × 1030erg/sec) and vary only the accretion lumi- nosity, Lacc = a × Lbol, over a range of a from 1 to 10−5. The results are shown in Figure 2, where the filled circles show our models and the empty symbols the observational results of Rigliaco et al (2013). It is clear from the figure Blueshifted OI lines from PPDs 5 Figure 4. Temperature (left) and density (right) maps showing the location of the 85% emission region of the OI 6300 line (black contour). The velocity field is represented by the red arrows. Plotted is the model with LX = 2 × 1030 erg/sec, Lacc = Lbol. are the much lower FWHM of [OI]6300 predicted by the models (filled diamonds in Figure 3) compared to the ob- servations (empty symbols in Fugure 3). As listed in Table 2, the FWHM of the [OI] 6300 line predicted in our new models (e.g. LX = 2 × 1030 erg/sec and Lacc = Lbol, filled circles in Figure 3) have broader FWHM, which are better in agreement with the observations. The reason for this is that the emission region of the [OI] 6300 line in our new models extends to about 35 AU above the disc (Left panel in Figure 4), compared to only up to 15AU in the models of EO10 (see their Figure 3). The [OI] 6300 line thus samples a wider range of wind velocities, which naturally results in a broader spectral profile. The corresponding line profiles are shown in Figure 6 for 10 inclinations between 0 and 90 degrees and for R=25,000 resolution (similar to the data used by Rigliaco et al. 2013) and R=50,000 (which is more representative of future work). We stress that the underlying wind model is the same as that of EO10, the difference in the extension of the [OI]6300 line emission region is purely driven by the presence of the extra accretion luminosity component which is more efficient at heating the wind to the temperatures required to produce the [OI]6300 line in the wind at higher vertical distances above the disc. We note however that the FWHM values of the new models are still somewhat narrower than the observational data. The Natta et al. (2014) sample in Lupus (not shown in this plot) have on average still broader profiles, with a me- dian FWHM of 55.5Km/s. New observational data at higher spectral resolution is needed to determine the exact spectral profile of these lines. 3.4 A thermal origin for the [OI] 6300: the [OI] 6300/ [OI] 5577 ratio Perhaps the strongest doubts that the [OI] 6300 may not have a thermal origin were cast by Rigliaco et al (2013) on the basis of their observational mesurements of the [OI]6300/[OI]5577 line ratio in their sample of YSOs. Fig- ure 5 shows that the measurements (empty symbols) scatter around a ratio of about 5, while the models of EO10 (filled c(cid:13) 2011 RAS, MNRAS 000, 1 -- 7 Figure 5. EW [OI] 6300A/EW [OI] 5577 A versus accretion lumi- nosity for the observational data of Rigliaco et al. (2013) (empty diamonds) and Natta et al. (2014) (crosses), compared to the model predictions of this work with (filled diamonds) and with- out (filled circles) neutral hydrogen collision contributions to the [OI] 5577A line. See section 3.4 for details. diamonds) show significantly higher ratios between 7 and 10, in clear disagreement with the observations. The empty dia- monds and the crosses represent, respectively, the data from Rigliaco et al. (2013) and Natta et al. (2014). This discrep- ancy led Rigliaco et al. (2013) to favour a non-thermal origin for the formation of the [OI] 6300 line. However, as was al- ready discussed by EO10, the luminosity of the [OI]5577 line calculated from the models is only a lower limit. Both [OI]6300 and [OI]5577 can be excited by collisions with elec- trons or neutral hydrogen atoms, however in the literature no neutral hydrogen collisional strengths are available for the [OI]5577 line, which is then necessarily underestimated, while the [OI]6300 predictions include both contributions from electron and neutral hydrogen collision. To illustrate this point better we have perfomed cal- culations where we turn off neutral hydrogen collisions for 0510152025303540Radius [AU]510152025303540Height [AU]11.522.533.544.5log10(Temperature) [K]0510152025303540Radius [AU]510152025303540Height [AU]-22-20-18-16-14-12-10log10(Density) [g cm-3]6 Ercolano, Owen LX <100eV [2E30erg/sec] suppressed Lacc [Lbol] L[OI]63000 [L(cid:12)] L[OI]5577 [L(cid:12)] 1.0 1.0 1e-1 1e-2 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 yes yes yes yes yes yes yes yes no no no no no no yes yes yes yes 10 1 1 1 1e-1 1e-2 1e-3 1e-4 10 1 1e-1 1e-2 1e-3 1e-4 1 1e-1 1e-2 1e-3 2.9e-4 3.7e-5 1.2e-5 1.1e-5 1.2e-5 8.3e-6 7.3e-6 6.2e-6 3.4e-4 4.5e-5 1.5e-5 1.1e-5 1.0e-5 1.0e-5 >2.2e-5 >5.4e-6 >3.7e-6 >3.4e-6 > 3.0e-5 >3.8e-6 >2.3e-6 >2.0e-6 > 7.0e-7 > 3.6e-7 > 2.3e-7 > 2.1e-7 >4.e-5 >4.8e-6 >1.5e-6 >9.7e-7 >8.3e-7 >8.2e-7 >4.7e-6 >1.3e-6 >9.5e-7 > 8.2e-7 Table 1. Summary of models and corresponding line luminosity predictions. See text for detail. inclination [degrees] FWHM25 [km s−1] v25 peak [km s−1] FWHM50 [km s−1] v50 peak [km s−1] 0 10 20 30 40 50 60 70 80 90 14.36 15.76 18.67 21.73 24.63 27.29 29.58 31.36 32.49 32.89 -2.1 -2.6 -3.6 -4.6 -5.1 -5.1 -4.4 -3.4 -1.9 0.0 8.97 11.25 14.82 18.32 21.62 24.63 27.25 29.33 30.72 31.38 -1.6 -2.9 -4.4 -5.9 -6.6 -6.9 -6.3 -6.1 -5.1 0.0 Table 2. Velocity at the peak and full-width-half- maximum (FWHM) of the [OI] 6300 line for the LX = 2 × 1030 erg/sec, Lacc = Lbol model from this work as a function of disc inclina- tion. The quantities were calculated for spectral resolutions of 25000 and 50000. the [OI]6300, so that both [OI]6300 and [OI]5577 only in- clude collisions with electrons. These results are also shown in Figure 5 where the predictions from the electron colli- sion only calculations are shown as filled circles. These new calculations are consistent with the observational measure- ments, implying also that hydrogen collision strengths must contribute in a comparable fashion to the fluxes of both the [OI]6300 and the [OI]5577 lines. 4 CONCLUSIONS We have revisited the question of the origin of the low veloc- ity component (LVC) of the blueshifted [OI] line at 6300A. Our new models suggest that this line is produced by col- lisional excitation of neutral Oxygen by free electrons and neutral hydrogen atoms in the quasi-neutral X-ray photo- evaporative winds of young stellar objects up to approxi- mately 1 M(cid:12). Our current models are not relevant to higher mass Herbig-type stars as the photoevaporative wind struc- ture is likely to change due to the fact that these objects have much lower (or absent) X-ray luminosities compared to their solar-like counterparts. Our models show that while the wind structure is driven by X-ray radiation (0.1 keV< E <1keV) reaching the bound atmosphere of a protoplanetary disc, the size of the emit- ting region of optical forbidden lines is determined by the EUV (13.6 eV< E <100eV) photons reaching the wind. The wind is optically thick to EUV radiation, which thus cannot reach the bound disc atmosphere at all. In fact EUV pho- tons penetrate into and heat only a thin vertically extended (up to about 37 AU) region of the wind above the inner disc (<5AU). The luminosity of the [OI] 6300 line and, in fact, of all forbidden lines that have an exponential temperature dependence (due to the Boltzmann term in their emissivity), are strongly weighted to the hottest regions, where neutral hydrogen is still abundant. This is shown in the left panel of Figure 4. This means that the line luminosities scale with the size of the wind region that can be heated by the EUV to the appropriate temperature. From this it follows, as confirmed by our models, that the luminosity of optical forbidden lines is correlated with the accretion luminosity of the YSO. As a consequence the luminosity of lines like [OI] 6300 cannot be used to measure wind properties, such as mass loss rate, as their production is not due to the same radiation that causes the photoevaporation of the disc atmosphere. Nevertheless the blueshifted low velocity components de- tected by (e.g.) Hartigan et al. (1995), Rigliaco et al. (2013) and Natta et al. (2014) are a clear tell-tale sign of a disc wind. High spectral resolution observations of these lines, particularly in combination, remain an important diagnos- tic tool of mass loss processes from disc atmospheres. 5 ACKNOWLEDGEMENTS We thank an anonymous referee for helpful comments on the first version of this paper. We thank Elisabetta Rigli- aco for making electronic versions of their data available to us and for the useful discussion. JEO acknowledges sup- port by NASA through Hubble Fellowship grant HST-HF2- 51346.001-A awarded by the Space Telescope Science Insti- tute, which is operated by the Association of Universities for Research in Astronomy, Inc., for NASA, under contract NAS 5-26555. REFERENCES Alexander, R. D., Clarke, C. J., & Pringle, J. E. 2006, MNRAS, 369, 216 Alexander, R. D. 2008, MNRAS, 391, L64 Asplund, M., Grevesse, N., & Sauval, A. J. 2005, Cosmic Abundances as Records of Stellar Evolution and Nucle- osynthesis, 336, 25 Ercolano, B., Barlow, M. J., Storey, P. J., & Liu, X.-W. 2003, MNRAS, 340, 1136 Ercolano, B., Barlow, M. J., & Storey, P. J. 2005, MNRAS, 362, 1038 Ercolano, B., & Storey, P. J. 2006, MNRAS, 372, 1875 c(cid:13) 2011 RAS, MNRAS 000, 1 -- 7 Blueshifted OI lines from PPDs 7 Figure 6. [OI] 6300 line profiles from our high resolution hydrodynamical simulations. See text for detailed. The left panel shows line profiles computed with a spectral resolution of 25,000 whereas the right panel is at 50,000. Each line indicates a different viewing inclination, where the inclination is indicated in the legend in degrees. Ercolano, B., Drake, J. J., Raymond, J. C., & Clarke, C. C. ApJ, 570, 799 2008a, ApJ, 688, 398 Ercolano, B., Young, P. R., Drake, J. J., & Raymond, J. C. 2008b, ApJS, 175, 534-542 Stone, J. M., & Norman, M. L. 1992a, ApJS, 80, 753 Stone, J. M., & Norman, M. L. 1992b, ApJS, 80, 791 Stone, J. M., Mihalas, D., & Norman, M. L. 1992c, ApJS, Ercolano, B., Clarke, C. J., & Drake, J. J. 2009a, ApJ, 699, 80, 819 Verner, D. A., Ferland, G. J., Korista, K. T., & Yakovlev, D. G. 1996, ApJ, 465, 487 Verner, D. A., & Yakovlev, D. G. 1995, AAPS, 109 White, R. J., & Hillenbrand, L. A. 2004, ApJ, 616, 998 1639 Ercolano, B., & Owen, J. E. 2010, MNRAS, 406, 1553 Ercolano, B., & Rosotti, G. 2015, MNRAS, 450, 3008 Gorti, U., Dullemond, C. P., & Hollenbach, D. 2009, ApJ, 705, 1237 Hartigan, P., Edwards, S., & Ghandour, L. 1995, ApJ, 452, 736 Hayes, J. C., Norman, M. L., Fiedler, R. A., et al. 2006, ApJS, 165, 188 Kashyap, V., & Drake, J. J. 2000, Bulletin of the Astro- nomical Society of India, 28, 475 Landi, E., Del Zanna, G., Young, P. R., et al. 2006, ApJS, 162, 261 Maggio, A., Flaccomio, E., Favata, F., et al. 2007, ApJ, 660, 1462 Mohanty, S., Jayawardhana, R., & Basri, G. 2005, ApJ, 626, 498 Natta, A., Testi, L., Alcal´a, J. M., et al. 2014, A&A, 569, A5 Owen, J. E., Ercolano, B., Clarke, C. J., & Alexander, R. D. 2010, MNRAS, 401, 1415 Owen, J. E., Ercolano, B., & Clarke, C. J. 2011, MNRAS, 412, 13 Owen, J. E., Clarke, C. J., & Ercolano, B. 2012, MNRAS, 422, 1880 Owen, J. E., & Wu, Y. 2016, ApJ, 817, 107 Pascucci, I., Hollenbach, D., Najita, J., et al. 2007, ApJ, 663, 383 Pascucci, I., Sterzik, M., Alexander, R. D., et al. 2011, ApJ, 736, 13 Rigliaco, E., Pascucci, I., Gorti, U., Edwards, S., & Hollen- bach, D. 2013, ApJ, 772, 60 Sanz-Forcada, J., Brickhouse, N. S., & Dupree, A. K. 2002, c(cid:13) 2011 RAS, MNRAS 000, 1 -- 7 -60-40-200204060Velocity [km s-1]024681012Luminosity [Scaled]R=25,0000102030405060708090-60-40-200204060024681012141618R=50,000
1904.02409	1	1904	2019-04-04T08:47:05	ALMA survey of Class II protoplanetary disks in Corona Australis: a young region with low disk masses	[ "astro-ph.EP", "astro-ph.SR" ]	In recent years, the disk populations in a number of young star-forming regions have been surveyed with ALMA. Understanding the disk properties and their correlation with those of the central star is critical to understand planet formation. In particular, a decrease of the average measured disk dust mass with the age of the region has been observed. We conducted high-sensitivity continuum ALMA observations of 43 Class II young stellar objects in CrA at 1.3 mm (230 GHz). The typical spatial resolution is 0.3". The continuum fluxes are used to estimate the dust masses of the disks, and a survival analysis is performed to estimate the average dust mass. We also obtained new VLT/X-Shooter spectra for 12 of the objects in our sample. 24 disks are detected, and stringent limits have been put on the average dust mass of the non-detections. Accounting for the upper limits, the average disk mass in CrA is $6\pm3\,\rm M_\oplus$, significantly lower than that of disks in other young (1-3 Myr) star forming regions (e.g. Lupus) and appears consistent with the 5-10 Myr old Upper Sco. The position of the stars in our sample on the HR diagram, however, seems to confirm that that CrA has age similar to Lupus. Neither external photoevaporation nor a lower than usual stellar mass distribution can explain the low disk masses. On the other hand, a low-mass disk population could be explained if the disks are small, which could happen if the parent cloud has a low temperature or intrinsic angular momentum, or if the the angular momentum of the cloud is removed by some physical mechanism such as magnetic braking. In order to fully explain and understand the dust mass distribution of protoplanetary disks and their evolution, it may also be necessary to take into consideration the initial conditions of star and disk formation process, which may vary from region to region, and affect planet formation.	astro-ph.EP	astro-ph	Astronomy & Astrophysics manuscript no. Corona_Australis_arXiv April 5, 2019 c(cid:13)ESO 2019 9 1 0 2 r p A 4 . ] P E h p - o r t s a [ 1 v 9 0 4 2 0 . 4 0 9 1 : v i X r a ALMA survey of Class II protoplanetary disks in Corona Australis: a young region with low disk masses(cid:63) P. Cazzoletti1, C. F. Manara2, Hauyu Baobab Liu2, 3, E. F. van Dishoeck1, 4, S. Facchini2, J. M. Alcalà5, M. Ansdell6, 7, L. Testi2, J. P. Williams8, C. Carrasco-González9, R. Dong10, J. Forbrich11, 12, M. Fukagawa13, R. Galván-Madrid9, N. Hirano3, M. Hogerheijde4, 14, Y. Hasegawa15, T. Muto16, P. Pinilla17, M. Takami3, M. Tamura13, 18, 19, M. Tazzari20, and J. P. Wisniewski21 (Affiliations can be found after the references) Received December XX, 2017; accepted September XX, 2017 ABSTRACT Context. In recent years, the disk populations in a number of young star-forming regions have been surveyed with the Atacama Large Millimeter/submillimeter Array (ALMA). Understanding the disk properties and their correlation with the properties of the central star is critical to understand planet formation. In particular, a decrease of the average measured disk dust mass with the age of the region has been observed, consistent with grain growth and disk dissipation. Aims. We want to compare the general properties of disks and their host stars in the nearby (d = 160 pc) Corona Australis (CrA) star forming region to those of the disks and stars in other regions. Methods. We conducted high-sensitivity continuum ALMA observations of 43 Class II young stellar objects in CrA at 1.3 mm (230 GHz). The typical spatial resolution is ∼ 0.3(cid:48)(cid:48). The continuum fluxes are used to estimate the dust masses of the disks, and a survival analysis is performed to estimate the average dust mass. We also obtained new VLT/X-Shooter spectra for 12 of the objects in our sample for which spectral type information was missing. Results. 24 disks are detected, and stringent limits have been put on the average dust mass of the non-detections. Taking into account the upper limits, the average disk mass in CrA is 6 ± 3 M⊕. This value is significantly lower than that of disks in other young (1-3 Myr) star forming regions (Lupus, Taurus, Chamaeleon I, and Ophiuchus) and appears to be consistent with the average disk mass of the 5-10 Myr old Upper Sco. The position of the stars in our sample on the Herzsprung-Russel diagram, however, seems to confirm that that CrA has age similar to Lupus. Neither external photoevaporation nor a lower than usual stellar mass distribution can explain the low disk masses. On the other hand, a low-mass disk population could be explained if the disks are small, which could happen if the parent cloud has a low temperature or intrinsic angular momentum, or if the the angular momentum of the cloud is removed by some physical mechanism such as magnetic braking. Even in detected disks, none show clear substructures or cavities. Conclusions. Our results suggest that in order to fully explain and understand the dust mass distribution of protoplanetary disks and their evolution, it may also be necessary to take into consideration the initial conditions of star and disk formation process. These conditions at the very beginning may potentially vary from region to region, and could play a crucial role in planet formation and evolution. Key words. protoplanetary disks -- submillimeter: ISM -- planets and satellites: formation -- stars: pre-main sequence -- stars: variables: T Tauri, Herbig Ae/Be -- stars: formationt 1. Introduction Planets form in protoplanetary disks around young stars, and the way these disks evolve also impacts what kind of planetary sys- tem will be formed (Morbidelli & Raymond 2016). The evolu- tion of the disk mass with time is one of the key ingredients of planetary synthesis models (Benz et al. 2014). For a long time infrared telescopes (e.g., Spitzer) have shown how the inner re- gions of disks dissipate on a timescale of ∼3-5 Myr (Haisch et al. 2001; Hernández et al. 2007; Fedele et al. 2010; Bell et al. 2013). Only recently, however, we have been able to measure the bulk disk mass for statistically significant samples of disks, thanks to the high sensitivity of the Atacama Large Millime- ter/submillimeter Array (ALMA). Pre-ALMA surveys of disk (cid:63) Based on observations made with ESO Telescopes at the La Silla Paranal Observatory under programme ID 299.C-5048 and 0101.C- 0893 masses were restricted to the northern hemisphere Taurus, Ophi- uchus and Orion Nebula Cluster regions (Andrews & Williams 2005; Andrews et al. 2009, 2013; Eisner et al. 2008; Mann & Williams 2010). In the first years of operations of ALMA this has changed dramatically: hundreds of disks have been surveyed to determine the disk population in the ∼1-3 Myr old Lupus, Chamaeleon I, Orion Nebula Cluster, Ophiuchus, IC348 and Taurus regions (Ansdell et al. 2016; Pascucci et al. 2016; Eisner et al. 2018; Cieza et al. 2019; Ruíz-Rodríguez et al. 2018; Long et al. 2018), in the ∼3-5 Myr old σ-Orionis region (Ansdell et al. 2017), and in the older ∼5-10 Myr Upper Scorpius associa- tion (Barenfeld et al. 2016). These surveys have shown that the typical mass of protoplanetary disks decreases with the age of the region, in line with the observations that the inner regions of disks are dissipated within ∼ 3-5 Myr, similar to the dissipation time scale measured in the infrared. A positive correlation be- tween disk and stellar mass was also found, and a steepening of its slope with time was identified (Ansdell et al. 2016, 2017; Pas- Article number, page 1 of 20 A&A proofs: manuscript no. Corona_Australis_arXiv Fig. 1: Spatial distribution of the CrA sources from the Peterson et al. (2011) catalogue on top of the Herschel 250 µm map of the Corona Australis molecular Cloud. The different colours represent the classification of the YSOs. The blue star indicates the position of R CrA cucci et al. 2016). This is consistent with the result that massive planets form and are found preferentially around more massive stars (e.g. Bonfils et al. 2013; Alibert et al. 2011). Finally, the steepening of the relation with time is explained with more ef- ficient radial drift around low mass stars (Pascucci et al. 2016), and it suggests that a significant portion of the planet forma- tion process, especially around low mass stars, must happen in the first ∼1-2 Myr, when enough material to form planets is still available in disks (Testi et al. 2016; Manara et al. 2018). Study- ing the evolution of the Mdisk − M(cid:63) relation in as many different environments as possible is therefore critical for understanding how the planet formation process is affected by the mass of the central stars. We present here a survey of the Class II disks in the Corona Australis star forming region (CrA). Located at an average dis- tance of about 154 pc (Gaia Collaboration et al. 2018; Dzib et al. 2018), the CrA molecular cloud complex is one of the nearest star-forming regions (see review in Neuhäuser & Forbrich 2008). It has been the target of many infrared surveys, the most recent being the Gould Belt (GB) Spitzer Legacy program presented in Peterson et al. (2011). At the center of the CrA region is located the Coronet cluster, which is a region of young embedded ob- jects in the vicinity of R CrA (Herbig Ae star, Neuhäuser et al. 2000), on which many of the previous studies have focused. All studies agree in assigning to the Coronet an age < 3 Myr (e.g. Article number, page 2 of 20 Meyer & Wilking 2009; Sicilia-Aguilar et al. 2011). However, there are also some indications of a more evolved population (e.g. Neuhäuser et al. 2000; Peterson et al. 2011; Sicilia-Aguilar et al. 2011). A deep, sub-mm wavelength survey of the disk pop- ulation in the region can help to further understand the formation and evolutionary history of CrA. We therefore use ALMA to conduct a high-sensitivity mil- limeter wavelength survey of all the known Class II sources in CrA and compare the results with other regions surveyed to- date. In Sec. 2 the sample is described, while the ALMA ob- servations are detailed in Sec. 3. We also describe there new VLT/X-Shooter observations to determine the stellar charachter- istics. The continuum millimeter measurements, their conversion to dust masses and a comparison with other star-forming regions is presented in Sec. 4. Our findings are interpreted in the context of disk evolution in Sec. 5. Finally, the work is summarized in Sec. 6. 2. Sample selection Peterson et al. (2011) present in their work a comprehensive cat- alogue of known Young Stellar Objects (YSOs) in the CrA star forming region selected based on Spitzer, 2MASS, ROSAT, and Chandra data. In addition to the these, they also added more YSOs from the literature. Their final catalogue includes a total P. Cazzoletti et al.: ALMA band 6 survey for Class II YSOs in CrA Table 1: Stellar properties of the central sources of the disks in the sample. The RA and DEC in J2000 are from the Spitzer data presented in Peterson et al. (2011) RA DEC Ref. 2MASS ID Name 1 J18563974-3707205 CrA-1 2 J18595094-3706313 CrA-4 3 J19002906-3656036 CrA-6 4 J19004530-3711480 CrA-8 3 J19005804-3645048 CrA-9 5 J19005974-3647109 CrA-10 6 J19011629-3656282 CrA-12 7 J19011893-3658282 CrA-13 7 J19013232-3658030 CrA-15 7 J19013385-3657448 CrA-16 7 J19014041-3651422 CrA-18 8 J19015112-3654122 CrA-21 1 J19015180-3710478 CrA-22 7 J19015374-3700339 CrA-23 1 J19020682-3658411 CrA-26 5 J19021201-3703093 CrA-28 ... J19021464-3700328 CrA-29 5 J19022708-3658132 CrA-30 1 J19023308-3658212 CrA-31 3 J19031185-3709020 CrA-35 1 J19032429-3715076 CrA-36 1 J19012576-3659191 CrA-40 9 J19014164-3659528 CrA-41 ... J19015037-3656390 CrA-42 1 J19031609-3714080 CrA-45 E 1 J19031609-3714080 CrA-45 W 1 J18564024-3655203 CrA-47 1 J18570785-3654041 CrA-48 10 J19000157-3637054 CrA-52 1 J19011149-3645337 CrA-53 9 J19013912-3653292 CrA-54 11 J19015523-3723407 CrA-55 4 J19021667-3645493 CrA-56 1 J19032547-3655051 CrA-57 6 J19010860-3657200 SCrA N 6 J19010860-3657200 SCrA S 6 J19015878-3657498 TCrA 6 J19014081-3652337 TYCrA 12 J19041725-3659030 Halpha15 4 J19025464-3646191 ... ... ... J19015173-3655143 Haas17 ... J19020410-3657013 IRS10 References. (1) This work, (2) Bouy et al. (2004), (3) Romero et al. (2012), (4) López Martí et al. (2005), (5) Sicilia-Aguilar et al. (2011), (6) Forbrich & Preibisch (2007), (7) Sicilia- Aguilar et al. (2008), (8) Currie & Sicilia-Aguilar (2011), (9) Meyer & Wilking (2009), (10) Walter et al. (1997), (11) Herczeg & Hillenbrand (2014), (12) Patten (1998) SpT -37:07:20.8 M6 -37:06:31.6 M8 -36:56:03.8 M4 -37:11:48.2 M8.5 M1 -36:45:05.0 M4 -36:47:11.2 -36:56:28.3 M5 -36:58:28.4 M2 -36:58:03.0 M3.5 -36:57:44.9 M2.5 -36:51:42.3 M1.5 -36:54:12.4 M2 -37:10:44.7 M4.5 -37:00:33.9 M7.5 -36:58:41.0 M7 -37:03:09.4 M4.5 -37:00:32.9 -36:58:13.1 M0.5 -36:58:21.2 M3.5 -37:09:02.1 M5 -37:15:07.7 M5 -36:59:19.1 M4.5 M2 -36:59:52.7 -36:56:38.4 ... -37:14:08.2 M3.5 -37:14:08.2 M3.5 M6 -36:55:20.8 M5 -36:54:04.4 M1 -36:37:06.2 -36:45:33.8 M5 K7 -36:53:29.4 K5 -37:23:41.0 -36:45:49.4 M4 -36:55:05.3 M4.5 -36:57:20 K3 M0 -36:57:20 F0 -36:57:49 B9 -36:52:33.88 M4 -36:59:03.0 -36:46:19.1 M4.5 -37:01:06.0 -36:55:14.2 -36:57:01.2 18:56:39.76 18:59:50.95 19:00:29.07 19:00:45.31 19:00:58.05 19:00:59.75 19:01:16.29 19:01:18.95 19:01:32.31 19:01:33.85 19:01:40.41 19:01:51.12 19:01:51.86 19:01:53.75 19:02:06.80 19:02:12.00 19:02:14.63 19:02:27.07 19:02:33.07 19:03:11.84 19:03:24.29 19:01:25.75 19:01:41.62 19:01:50.48 19:03:16.09 19:03:16.09 18:56:40.28 18:57:07.86 19:00:01.58 19:01:11.49 19:01:39.15 19:01:55.23 19:02:16.66 19:03:25.48 19:01:08.62 19:01:08.62 19:01:58.78 19:01:40.83 19:04:17.25 19:02:54.65 19:01:58.34 19:01:51.74 19:02:04.09 ISO-CrA-177 G09-CrA-9 ... ... ... ... of 116 YSOs, 14 of which are classified as Class I, 5 as Flat Spectrum (FS), 43 as Class II and 54 as Class III. The Infrared Class was determined by calculating the spectral slope α over the widest possible range of IR wavelengths as follows: α = ∆ log (λFλ) ∆ log λ , (1) where λ is the wavelength and Fλ the flux at λ. Sources with α ≥ 0.3 are classified as Class I; FS have −0.3 ≤ α < 0.3; Class II have −1.6 ≤ α < −0.3; sources with α < −1.6 are Class III (Evans et al. 2009; Peterson et al. 2011). Fig. 1 shows the spatial distribution of the sources and their classification on top of the Herschel 250 µm map of the molecular Cloud. Our sample includes all the Class II sources from the Peter- son et al. (2011) catalogue. Two of them (CrA-49 and CrA-51) were later identified as background, evolved stars based on par- allax measurements with Gaia (Gaia Collaboration et al. 2016; Lindegren et al. 2018; Luri et al. 2018; Gaia Collaboration et al. 2018) and on our VLT/X-Shooter spectra (see Sec. 3.2). We then checked our sample against the more recently published Article number, page 3 of 20 A&A proofs: manuscript no. Corona_Australis_arXiv two epochs of observations followed the standard procedure of ALMA quality assurance (i.e., QA2). The bootstrapped flux val- ues of the calibrator quasar J1924-2914 were consistent with the SMA Calibrator list1 (Gurwell et al. 2007) to ∼10%. After cal- ibration, we fit the continuum baseline and subtract it from the spectral line data, using the CASA task uvcontsub. The continuum data imaging was performed with multi- frequency synthesis (MFS) imaging of the continuum data us- ing the CASA-clean task, and correcting for the primary beam. By jointly imaging all three epochs of data, for each target source field, the achieved continuum root-mean-square (RMS) noise level is ∼0.15 mJy beam−1, and the synthesized beam is θmaj × θmin=0(cid:48)(cid:48).33×0(cid:48)(cid:48).31 (P.A.=67◦), corresponding to a spatial resolution of ∼ 50 au at d = 154 pc. The imaged detections are presented in Fig. 3. It is important to note that because of an error when setting the observation coordinates, the decimal places of the target RAs have been trimmed: this results in an offset of the sources of up to 15(cid:48)(cid:48) east of the phase center: as a consequence, our images had to be primary beam corrected. The images in Fig. 3 have therefore been re-centered using the best-fit positions in Tab. 3. 3.2. VLT/X-Shooter observations The spectroscopic follow-up observations for the 13 targets with missing spectral type information were carried out in Pr.Id. 299.C-5048 (PI Manara) and Pr.Id. 0101.C-0893 (PI Cazzoletti) with the VLT/X-Shooter spectrograph (Vernet et al. 2011). This instrument covers the wavelength range from ∼300 nm to ∼2500 nm simultaneously, dividing the spectrum in three arms, the UVB (λλ ∼ 300-550 nm), the VIS (λλ ∼ 500-1050 nm), and the NIR (λλ ∼ 1000-2500 nm). All targets were observed both with a narrow slit - 1.0(cid:48)(cid:48) in the UVB, 0.9(cid:48)(cid:48) in the VIS and NIR arms - leading to R∼9000 and ∼10000, respectively, and a wide slit of 5.0(cid:48)(cid:48) used to obtain an accurate flux calibration of the spectra. The log of the observations is reported in Table B.2. The spec- tra of all the observed targets are detected in the NIR arm, while only 5 targets are bright enough and not extincted too much to be detected also in the UVB arm. The reduction of the data was performed using the ESO X- Shooter pipeline 2.9.3 (Modigliani et al. 2010). The pipeline per- forms the typical reduction steps, such as flat fielding, bias sub- traction, order extraction and combination, rectification, wave- length calibration, flux calibration using standard stars observed in the same night. We extracted the 1D spectra from the 2D im- ages produced by the pipeline using IRAF and then removed telluric absorption lines in the VIS and NIR arms using telluric standard stars observed close in time and airmass (see e.g., Al- calá et al. 2014). The S/N of the spectra at different wavelengths is reported in Table B.2. 4. Results and analysis 4.1. Stellar properties The spectral type for the targets were obtained from the litera- ture (see Tab. 1) or from the VLT/X-Shooter spectra. The proce- dure used for the analysis of the X-Shooter spectra was as fol- lows. First, we corrected the spectra for extinction using the val- ues from the literature (Dunham et al. 2015; Sicilia-Aguilar et al. 2008, 2011) and the reddening law by Cardelli et al. (1989) with RV =3.1, as suggested by Sicilia-Aguilar et al. (2008). Then, 1 http://sma1.sma.hawaii.edu/callist/callist.html Fig. 2: Distribution of the spectral types of the stars in CrA (Red) compared to that of Lupus (Orange) and Upper Sco (Blue). survey by Dunham et al. (2015) in which the Spitzer data are re-analysed and the spectral slopes re-calculated. We find broad agreement between the classification in Peterson et al. (2011) and Dunham et al. (2015), except for a few very marginal cases at the boundaries of classes. Our final sample contains 41 targets, two of which are clearly resolved binaries (S CrA and CrA-45). Of the 43 targeted disks, 24 are detected with ALMA. The spectral type (SpT) was known for only 26 of the stars from the literature. We obtained VLT/X- Shooter spectra for 11 of the remaining targets, and derived their properties as explained in Sec. 4.1. The basic stellar properties for the CrA sample are given in Table 1, the distribution of SpTs is shown in Fig. 2, while the mil- limeter observations, flux densities, and calculated disk masses are presented in Table 3. 3. Observations 3.1. ALMA observations We have carried out three executions of observations at 1.3 mm towards 43 Class II YSOs in the Corona Australis molecular cloud, using ALMA (2015.1.01058.S, PI: H. B. Liu).. Each one of the 43 target sources were integrated for approximately 1 minute in each epoch. The spectral setup consists of six spec- tral windows, of which the (central frequency [GHz], total band- width [MHz], and frequency channel width [kHz]) are (216.797, 1875, 488), (219.552, 59, 61), (219.941, 59, 61), (220.390, 117, 61), (230.531, 117, 31), (231.484, 1875, 488), respectively. Ad- ditional observational details are summarized in Table 2. 12CO (2-1), 13CO (2-1) and C18O (2-1) transitions were also targeted with our spectral setup, but no clear detection was found be- cause of strong foreground contamination. SO (6-5) and SiO (5- 4) lines were also covered and not detected. The data were manually calibrated using the CASA v5.1.1 software package (McMullin et al. 2007) . The gain calibrator for the first epoch of observations was faint. To yield reasonably high signal-to-noise (S/N) ratios when deriving the gain phase solutions, the phase offsets among spectral windows were first solved using the passband calibration scan. After applying the phase offsets solution, the gain phase solution was then derived by combining all spectral windows. The calibration of the other Article number, page 4 of 20 M0M5L0K3G8G0F0A0Spectral Type0.000.050.100.150.20U ScoLupusCrAP. Cazzoletti et al.: ALMA band 6 survey for Class II YSOs in CrA Table 2: ALMA observations towards Class II objects in the CrA molecular cloud Epoch Time (UTC; 2016) Project baseline lengths (min-max) [m] Absolute flux calibrator Gain calibrator Bootstrapped gain calibrator flux [Jy] Passband calibration Bootstrapped passband calibrator flux [Jy] 1 (Aug.01) 03:32-04:54 14-1108 Pallas J1937-3958 0.26 J1924-2914 4.1 2 (Aug.01) 05:01-06:23 15-1075 Pallas J1924-2914 3.9 J1924-2914 3.9 3 (Aug.02) 03:18-04:40 15-1110 Pallas J1924-2914 4.1 J1924-2914 4.1 Fig. 3: ALMA Band 6 1.3 mm continuum images of the 24 detections in Corona Australis. The size of the images is 3(cid:48)(cid:48) × 3(cid:48)(cid:48). The size of the beam is indicated at the bottom-left corner of the first panel (0(cid:48)(cid:48).31 × 0(cid:48)(cid:48).33). The north of each image is upwards. The presented images have not been primary beam corrected. Article number, page 5 of 20 50 auA&A proofs: manuscript no. Corona_Australis_arXiv Table 3: 1.3 mm continuum properties of the sources targeted in our sample. Name F1.3 mm [mJy] RMS [mJy beam−1] 0.10 0.14 0.08 0.16 0.33 0.22 0.10 0.21 0.08 1.00 0.35 0.08 0.12 0.12 0.11 0.09 0.11 0.09 0.19 0.11 0.82 0.11 0.21 0.34 1.92 1.92 0.09 0.12 0.16 0.10 0.09 0.10 0.10 0.10 10.4 10.4 0.28 0.13 0.08 0.08 0.11 0.11 0.09 amaj [(cid:48)(cid:48)] ... ... ... 0.365 ± 0.018 0.389 ± 0.008 0.481 ± 0.119 0.661 ± 0.100 0.367 ± 0.021 0.478 ± 0.008 0.380 ± 0.008 ... amin [(cid:48)(cid:48)] ... ... ... 0.31 ± 0.014 0.31 ± 0.005 0.25 ± 0.034 0.32 ± 0.030 0.35 ± 0.019 0.44 ± 0.007 0.32 ± 0.006 ... PA [◦] ... ... ... 81.90 ± 13.3 81.07 ± 3.4 73.11 ± 7.9 90.67 ± 4.8 126.02 ± 60.2 13.42 ± 11.5 93.93 ± 5.3 ... Mdust [M⊕] ... ... ... 1.50 ± 0.12 3.70 ± 0.12 0.48 ± 0.15 1.00 ± 0.17 2.02 ± 0.19 14.84 ± 0.39 3.92 ± 0.14 ... ... ... 0.401 ± 0.120 ... ... 0.28 ± 0.061 ... ... 108.44 ± 24.8 ... ... 0.26 ± 0.11 ... ... ... ... 0.504 ± 0.139 0.416 ± 0.019 0.384 ± 0.004 0.384 ± 0.017 0.377 ± 0.010 0.400 ± 0.003 0.393 ± 0.013 ... ... ... 0.404 ± 0.023 0.577 ± 0.189 0.354 ± 0.037 ... ... ... 0.451 ± 0.005 0.439 ± 0.004 0.568 ± 0.033 0.362 ± 0.044 0.529 ± 0.128 0.535 ± 0.146 ... ... ... ... ... 0.27 ± 0.046 0.33 ± 0.013 0.32 ± 0.002 0.30 ± 0.011 0.31 ± 0.007 0.34 ± 0.002 0.33 ± 0.009 ... ... ... 0.29 ± 0.012 0.26 ± 0.044 0.29 ± 0.025 ... ... ... 0.40 ± 0.004 0.39 ± 0.003 0.37 ± 0.016 0.28 ± 0.026 0.39 ± 0.078 0.25 ± 0.037 ... ... ... ... ... 96.59 ± 10.6 80.58 ± 7.3 74.97 ± 2.1 85.01 ± 6.7 81.88 ± 4.9 46.21 ± 1.7 87.43 ± 7.0 ... ... ... 70.00 ± 5.4 105.67 ± 7.8 96.62 ± 19.4 ... ... ... 75.03 ± 4.3 80.73 ± 3.7 20.25 ± 4.3 90.13 ± 14.6 142.60 ± 27.2 71.91 ± 7.3 ... ... ... ... ... 0.37 ± 0.13 2.05 ± 0.14 9.41 ± 0.15 2.04 ± 0.14 3.55 ± 0.16 21.76 ± 0.26 4.64 ± 0.25 ... ... ... 1.43 ± 0.12 0.35 ± 0.14 0.59 ± 0.10 ... ... ... 94.51 ± 1.52 102.36 ± 1.46 3.64 ± 0.27 0.66 ± 0.13 0.50 ± 0.16 0.38 ± 0.13 ... ... ... ... ... ... ... ... ... band magnitudes and using the bolometric correction from Her- czeg & Hillenbrand (2014), assuming for all the target the av- erage distance of 154 pc calculated by Dzib et al. (2018). With this information, we have been able to plot our data on the HR diagram (Fig. 6) and to estimate the stellar masses (M(cid:63)) for all the targets using the evolutionary tracks by Baraffe et al. (2015) for M(cid:63) < 1.4M(cid:12) and Siess et al. (2000) for higher M(cid:63) and ages younger than 1 Myr . The stellar parameters for the targets are reported in Tab. A.1. ... ... ∆α [(cid:48)(cid:48)] ... ... ... ∆δ [(cid:48)(cid:48)] ... ... ... 0.39 0.35 0.34 0.42 0.27 ... 0.45 0.52 ... ... 0.76 ... ... ... 0.61 0.60 ... 0.38 ... 0.62 0.34 0.63 0.28 ... ... ... ... ... ... ... ... -0.14 -0.26 ... ... ... -0.36 -0.28 -0.23 -0.20 0.35 ± 0.15 -0.14 -0.01 0.07 -0.05 0.06 ... 2.06 ± 0.17 5.07 ± 0.16 0.65 ± 0.11 1.37 ± 0.24 2.77 ± 0.26 20.34 ± 0.53 5.36 ± 0.19 CrA-1 CrA-4 CrA-6 CrA-8 CrA-9 CrA-10 CrA-12 CrA-13 CrA-15 CrA-16 CrA-18 CrA-21 CrA-22 CrA-23 CrA-26 CrA-28 CrA-29 CrA-30 CrA-31 CrA-35 CrA-36 CrA-40 CrA-41 CrA-42 CrA-45 E CrA-45 W CrA-47 CrA-48 CrA-52 CrA-53 CrA-54 CrA-55 CrA-56 CrA-57 S CrA S S CrA N T CrA TY CrA IRS10 Halpha15 ISO-CrA-177 Haas17 G09-CrA-9 Notes. † Offset with respect to coordinates listed in Tab. 1. 0.51 ± 0.18 2.82 ± 0.19 12.9 ± 0.21 2.80 ± 0.19 4.87 ± 0.21 29.82 ± 0.35 6.36 ± 0.34 129.53 ± 2.09 140.30 ± 2.00 4.99 ± 0.37 0.91 ± 0.18 0.69 ± 0.22 0.52 ± 0.17 -0.14 0.24 ... ... 1.34 0.13 1.34 0.46 ... 0.54 0.49 ... ... 1.95 ± 0.17 0.48 ± 0.19 0.81 ± 0.14 -0.34 -1.00 -0.36 0.05 ... ... 0.05 ... 0.61 -0.26 -0.36 0.23 -0.06 0.01 ... -0.07 -0.01 -0.27 ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... we calculated the values of a number of spectral indices both at wavelengths in the VIS and the NIR arms, taken by those calibrated by Riddick et al. (2007), Jeffries et al. (2007), and Herczeg & Hillenbrand (2014), as in Manara et al. (2017), and by Testi et al. (2001), as in Manara et al. (2013a). The spectral types derived from these indices are presented in Tab. B.1 in Ap- pendix B. The spectral indices in the VIS arms are more reliable, and we select the spectral type from these indices when avail- able. The observed spectra along with a template of the relative Spectral Types are presented in Fig. B.1. The spectral types are converted in effective temperatures (Teff) using the relation by Herczeg & Hillenbrand (2014). Stel- lar luminosity (L(cid:63)) is obtained from the reddening-corrected J- Article number, page 6 of 20 P. Cazzoletti et al.: ALMA band 6 survey for Class II YSOs in CrA 4.2. mm continuum emission Among the 41 targets, 20 of them show a clear (≥ 4σ) detec- tion within a 1(cid:48)(cid:48) radius from the nominal Spitzer location from Peterson et al. (2011). In addition, CrA-42 and T CrA show a ∼ 36σ and a ∼ 22σ detection respectively at a slightly larger distance from their nominal Spitzer positions (1(cid:48)(cid:48).05 for CrA-42 and 1(cid:48)(cid:48).34 T CrA), and are also regarded as detections. S CrA is a known binary (Reipurth & Zinnecker 1993; Ghez et al. 1997; Takami et al. 2003), and we detected millimeter emission associ- ated with both binary components. CrA-45 is also identified as a binary. The total number of detections is therefore 24 out of the 43 targeted disks, so the detection rate is ∼ 56% . None of the disks show clear substructures, no transition disk with cavities with radius > 25 au are found and all of them appear to be unresolved or marginally resolved: a Gaussian is therefore fitted to the detected sources (two Gaussians for the bi- naries) in the image plane using the imfit task in CASA. The task returns the total flux-density F1.3 mm of the source along with the statistical uncertainty, the FWHM along the semi-major (amaj) and semi-minor (amin) axis and the position angle (PA). The re- sults of the fit are shown in Table 32The right ascension offset (∆α) and the declination offset (∆δ) with respect to the Spitzer coordinates is also shown. The rms noise for the non-detections was calculated using the imstat task within a 1(cid:48)(cid:48) radius centered at the Spitzer coordinates; for the detection, it was calculated in an annular region centered on the source and with inner and outer radii equal to 2(cid:48)(cid:48) and 4(cid:48)(cid:48), respectively. In order to constrain the average flux density of individually undetected sources, a stacking analysis was also performed. The images were centered at their Spitzer coordinates (Table 1) and then stacked. Even after the stacking, no detection was found and an average rms noise is 0.017 mJy beam−1, corresponding to a 3σ upper limit of 0.051 mJy is found assuming unresolved disks. However, it should be noted that the average offset between the disks and the Spitzer positions, measured on the detections, are < ∆α >= −0.13(cid:48)(cid:48) and < ∆δ >= 0.47(cid:48)(cid:48): it is therefore possible that the undetected sources did not overlap during the stacking, and that the upper limit is actually higher than that quoted. 4.3. Dust masses Assuming that the observed sub-millimeter emission is optically thin and isothermal, the relation between the emitting dust mass (Mdust) and the observed continuum flux at frequency ν (Fν) is as follows (Hildebrand 1983): Mdust = Fν d2 κν Bν(Tdust) ≈ 2.19 × 10−6 160 (cid:32) d (cid:33)2 F1.3 mm [M(cid:12)], (2) Fig. 4: Comparison of the cumulative dust mass distributions of Lupus, CrA, Cham I, σ Ori and Upper Sco, derived using a survival analysis accounting for the upper limits. 2005), rather than the Tdust = 25 K×(L∗/L(cid:12))0.25 relation based on two-dimensional continuum radiative transfer by Andrews et al. (2013) and used in other works (e.g. Law et al. 2017). We adopt this simplified approach with a single grain opacity and temper- ature for all the disks in the sample following the approach of Ansdell et al. (2016) and to facilitate the comparison with other star-forming regions (see Sec. 4.4). Moreover, it should be noted that no dependence of the average dust temperature on the stel- lar parameters was found with the more detailed modelling by Tazzari et al. (2017) for the Lupus disks. The dust masses of the disks in our sample are presented in Tab. 3, along with the relative uncertainty calculated from the flux uncertainty. Only 3 disks out of 24 detections have a dust content ≥ 10 M⊕3 and large enough to form the cores of giant planets in the future. However, it is still possible that a similar amount of dust mass is hidden at the inner few region due to very high optical depth (e.g. Zhu et al. 2010; Liu et al. 2017; Vorobyov et al. 2018). Also note that very recent high angular resolution ALMA and VLA observations of disks are revealing that an important amount of dust is located in dense regions such as rings (e.g. Andrews et al. 2018), which are optically thick at wavelengths around 1 mm (Dullemond et al. 2018). When optically thin emission is detected, higher masses are estimated (Carrasco-González et al. 2016). The stacking of the non-detections gives an average 3σ upper limit corresponding to 0.036 M⊕, about 3 Lunar masses. where d is the distance of the object, Fν is measured the flux- density, Bν(Tdust) is the Planck function for a given dust temper- ature Tdust and κν is the dust opacity at frequency ν. To make the comparison with previous surveys easier, for the dust opacity κν we follow the same approach of Ansdell et al. (2016), assum- ing κν = 10 cm2 g−1 at 1000 GHz (Beckwith et al. 1990) and scaling it to our frequency using β = 1. The adopted value is therefore κν = 2.3 cm2 g−1 at ν = 230 GHz (1.3 mm). In the right-hand side of Eq. 2, the distance d is measured in pc and the flux density F1.3 mm is in mJy. For each object, the average distance of the cluster d = 154 pc was used. For the dust tem- perature, we use a constant Tdust = 20 K (Andrews & Williams 2 Note that the F1.3 mm uncertainty only includes the statistical uncer- tainty from the fit, and not the 10% absolute flux calibration uncertainty. 4.4. Comparison with other regions The surveys of nearby star forming regions over the last years have shown growing evidence of a decrease in the mass of the disks with age, reflecting dust growth and disk dispersal. Ans- dell et al. (2016, 2017) found consistent results, calculating the highest average mass in the youngest regions (1-3 Myrs), and the lowest for the the oldest Upper Sco association (5-10 Myrs). The 2-3 Myrs old IC348 is the only exception, showing an average dust mass of only 4 ± 1 M⊕ (between the average σ Orionis and 3 5 sources in total have a dust content ≥ 10 M⊕ if we also consider CrA-16 and CrA-36, which have dust masses only marginally below 10 M⊕. Article number, page 7 of 20 101100101102Mdust [M]0.00.20.40.60.81.0P MdustCham ILupus OriUScoCrAA&A proofs: manuscript no. Corona_Australis_arXiv Table 4: Global properties of the star forming regions surveyed with ALMA in order of age. Name Distance Average dust mass Age [Myr] 1-3 1-3 1-3 2-3 2-3 3-5 5-10 [pc] 129.51 1601,(cid:63) 154 1 1921 3211 3881 1443 Taurus Lupus CrA Chameleon I IC 348 σ Ori Upper Sco References. (1) Dzib et al. (2018) (2) Comerón (2008) (3) de Zeeuw et al. (1999) (cid:63) The average distance of the 4 Lupus clouds was used. [M⊕] 13 ± 2 14 ± 3 6 ± 3 24 ± 9 4 ± 1 7 ± 1 5 ± 3 that of Upper Sco) despite its young age. This can be explained by the low-mass stellar population in the region (Ruíz-Rodríguez et al. 2018) (also see Tab. 4). The same analysis was done here for CrA. The dust masses are uniformly calculated following the approach used by Ans- dell et al. (2016), namely using Eq. 2 with the continuum fluxes (or the 3 σ upper limits) from our ALMA data or from the litera- ture, assuming a uniform T = 20 K, and inputting the frequency of the observation for each specific dataset. The distances as- sumed for each region are listed in Tab. 4. For the Upper Sco region, only the disks classified as "full", "evolved" and "tran- sitional" from the Barenfeld et al. (2016) sample are included, while the "debris" and Class III YSOs, which likely represent a separate evolutionary stage, are excluded. Finally, in order to facilitate the comparison with the other samples, in this analy- sis we only include the disks around stars with masses above the brown-dwarf limit (M(cid:63) ≥ 0.1 M(cid:12)). The Kaplan-Meier esti- mator from the lifelines4 and ASURV (Lavalley et al. 1992) packages were then used to estimate the cumulative mass distri- bution and to calculate the average dust mass and its uncertainty while properly accounting for the upper limits by using well- established techniques for left-censored data sets. Fig. 4 presents the results accounting for the upper limits given by the non-detections. With an average dust mass of 6 ± 3 M⊕, the distribution of the CrA disks appear closer to that of the old Upper Sco region rather than to those of the younger systems. 4.5. Mdisk − M(cid:63) relation A clear correlation between the dust mass of disks and the mass of the central star has been identified across all protoplanetary disk populations surveyed (Pascucci et al. 2016; Ansdell et al. 2017). This finding highlights how the disk properties are af- fected by the central star, and is consistent with the correlation between frequency of giant planets and mass of the host star, both from the observational and theoretical points of view (Al- ibert et al. 2011; Bonfils et al. 2013). Moreover, the slope of this relation has been observed to steepen with time, with the young Taurus, Lupus and Chemeleon I regions (∼ 1 − 3 Myr) having slopes similar to each other and shallower than that found for the disks in the Upper Sco association (5 − 10 Myr). Studying the Mdust − M(cid:63) relation for the disks in the CrA sample contributes to gain insight on the origin of the overall 4 10.5281/zenodo.1495175 Article number, page 8 of 20 Fig. 5: Correlation between dust disk flux scaled at 330 GHz (assum- ing α = 2.25, as in Ansdell et al. 2018) and at a distance of 150 pc with stellar mass for the objects in CrA. The slopes of Lupus and Upper Sco are also plotted for comparison. We show the results of the Bayesian fitting procedure by Kelly (2007). The solid line represents the best fit model, while the light lines show a subsample of models from the chains, giving an idea of the uncertainties. low mass of the disk population found in Sec. 4.4. We derive the Mdust−M(cid:63) relation using the same linear-regression Bayesian ap- proach followed by Ansdell et al. (2017) and presented by Kelly (2007)5. Unlike other linear regression methods, this approach is capable of simultaneously accounting for the uncertainties in both the measurements of Mdust and M(cid:63), of the intrinsic scatter of the data and of the disk non detections, which result in upper- limits on the disk masses. Note that the SpT, and therefore the stellar mass, is missing for 5 of our targets: for these objects the stellar mass is randomly drawn from the stellar mass distribution of the entire sample. In particular, 4 of the objects with unknown SpT are also not detected with ALMA, while the other one (CrA- 42) shows a clear detection of a disk at mm-wavelengths. For the 4 non detections, the stellar mass is therefore randomly drawn among the masses of the stars with non-detected disks, while the mass of CrA-42 is drawn from those showing a detection with ALMA. This uncertainty is also taken into account in the Bayesian approach we adopt by performing 100 different draws. In our fit, a standard uncertainty of 20% of M(cid:63) on the stellar mass is assumed (Alcalá et al. 2017; Manara et al. 2017), while the uncertainties shown in Tab. 3 were used for the Mdust values. Finally, it should be noted that only 1 out of 89 sources in Lupus was a Herbig Ae/Be star, while Upper Sco did not include any Herbig. We therefore decided not to include T CrA and TY CrA in the fit , for which the Mdust − M(cid:63) relation might not hold. The best fit relation we find is then plotted in Fig. 5 in dark red, along with a subsample of all the models in the chains to show the uncertainty. As in the other surveys, we also find a cor- relation, where the best-fit model has a slope β = 2.32±0.77 and intercept α = 1.29 ± 0.60. This regression intercept is lower that that of other regions, as a consequence of the low disk masses found in the region. The uncertainties of the best-fit parameters reflect the large scatter in the data and the low number statistics. In order to test that no strong bias was introduced by our pro- cedure, we also run the fit described above without any random draw, finding consistent results. 5 https://github.com/jmeyers314/linmix 0.11.0M [M]0.11101001000Flux at 150 pc [mJy]UScoLupusCrAP. Cazzoletti et al.: ALMA band 6 survey for Class II YSOs in CrA In order to further test if the Class II population in our sam- ple indeed includes an older population, we have placed them on the HR diagram, by using the spectral types listed in Tab. 1 and by deriving effective temperatures and bolometric correc- tions using the relationships in Herczeg & Hillenbrand (2014) and tables in Herczeg & Hillenbrand (2015), respectively. The obtained diagram is presented in Fig. 6. For comparison, the Up- per Sco and Lupus objects are also plotted. In contrast with what Fig. 4 suggests, the HR diagram supports the scenario of a young CrA cluster with an age more consistent to that of Lupus than to Upper Sco. In order to make this conclusion evident, the median val- ues of the bolometric luminosities for each temperature are also shown (solid coloured lines in Fig. 6). The indicative age of the cluster is the isochrone closer to those median values: these lines also suggest that CrA is younger than Upper Sco. However, a more extended spectral classification for a larger number of ob- jects in CrA would be needed to fully test this older-population scenario. 5.2. Is CrA young? If the whole CrA is coeval with an age of 1 − 3 Myr, some other mechanism has to be invoked to explain the low observed mm fluxes. For example, these fluxes could be due to low metallicity. However, James et al. (2006) determined metallicities for three T Tauri stars in CrA, finding them to be only slightly sub-solar, and not low enough to explain our obesrvations. External photo-evaporation is also known to play an impor- tant role in the disk mass evolution (Facchini et al. 2016; Win- ter et al. 2018a), and evidence of it occurring has been found in σ Ori (Maucó et al. 2016; Ansdell et al. 2017), where a clear correlation between disk mass and distance from the central Her- big O9V star has been observed and in the Orion Nebula Cluster (Mann & Williams 2010; Eisner et al. 2018). However, in CrA no correlation between the mass of the disks (or the disk detection rate) and the distance from the brightest star (R CrA) is found. Moreover, in σ Ori external photo-evaporation has been shown to affect disks up to 2 pc away from the Herbig star, where the ge- ometrically diluted far-ultraviolet (FUV) flux reached a value of ∼ 2000 G0. The spectral type of R CrA is still uncertain, ranging from F5 (e.g. Garcia Lopez et al. 2006) to B8 (e.g. Hamaguchi et al. 2005). Even in the latter case, assuming a typical FUV lu- minosity for a B8 star of LFUV ∼ 10 L(cid:12) (Antonellini et al. 2015) and accounting for geometric dilution, we find that the FUV flux would drop to ∼ 1 G0 in the first inner pc from R CrA, thus ruling-out external photo-evaporation as an explanation. Also, this calculation neglects dust absorption, which is probably very effective in the Coronet cluster around R CrA. Because of the Mdisk− M(cid:63) relation presented in Sec. 4.5, it is also possible that a system dominated by low-mass stars shows a low-mass disk population, regardless of its age, as in the case of IC348 (Ruíz-Rodríguez et al. 2018). It is therefore important, when comparing disk dust masses from different regions, to ver- ify that they have the same stellar mass distribution. In order to do this, we employ a Monte Carlo (MC) approach similar to that used by Andrews et al. (2013). We first normalize the stellar pop- ulations by defining stellar mass bins and randomly drawing the same number of sources in each bin from the reference sample (CrA) and from a comparison sample (Lupus, Chamaeleon I or Upper Sco). We then perform a two-sample logrank test for cen- sored datasets between the disk dust masses of the two samples, to test the probability (pφ value) that the two samples are ran- domly drawn from the same parent population. A low pφ value Article number, page 9 of 20 Fig. 6: HR diagram showing the sources in our CrA sample (red), in the Lupus sample from Ansdell et al. (2016) (orange) and in the Upper Sco sample (blue) from Barenfeld et al. (2016). The evolutionary tracks for different stellar masses and the relative isochrones from Baraffe et al. (2015) are also plotted for reference. The isochrones refer (from top to bottom) to the 1 Myr, 2Myr, 5 Myr and 10 Myr isochrone. The coloured solid lines show the approximate median value of the luminosity at each temperature. 5. Discussion 5.1. Is CrA old? The observed low disk dust masses suggest that the CrA objects targeted in our survey may have an age comparable to that of the Upper Sco association, rather than to the young Lupus region. Unlike CrA, however, Upper Sco shows no presence of Class 0 or Class I sources, as expected for a 5− 10 Myr region (Dunham et al. 2015). Moreover, most studies agree in assigning Corona Australis an age < 3 Myr (e.g. Meyer & Wilking 2009; Nisini et al. 2005; Sicilia-Aguilar et al. 2008, 2011). On the other hand, most of these studies focused only on the Coronet cluster, a small region extending ∼ 1 pc around the R CrA YSO, and where most of the young embedded Class 0 and Class I sources are located (see Fig. 1). The hypothesis that the large scale YSO population of the whole CrA cloud also includes a population of older objects therefore cannot be entirely ruled out. Some evidence of an additional older population has already been presented in previous studies. Neuhäuser et al. (2000) for example identify two classical T Tauri stars located outside the main cloud with an age of ∼ 10 Myr using ROSAT data. In ad- dition, Peterson et al. (2011) perform a clustering analysis of the 116 YSOs in their sample, identifying a single core (cor- responding to the Coronet) and a more extended population of PMS stars showing an age gradient west of the Coronet. They also observe that in the central core, the ratio Class II/ Class I=1.8, while the same ratio is Class II/ Class I=2.3 when all the objects in the sample are considered, again hinting toward a younger population inside the Coronet. Finally, Sicilia-Aguilar et al. (2011) point out that the relatively low disk fraction ob- served in the Coronet (∼ 50% López Martí et al. 2010, based on near IR photometry) is in strong contrast with the young age of the system: this inconsistency could be solved if an older pop- ulation were also present. The large scatter in the Mdust − M(cid:63) relation could also be a consequence of two stellar populations of different ages. 300040005000Teff[K]32101log (L/L)0.02 M0.05 M0.10 M0.2 M0.4 M0.8 M1.2 MCrACrA-XSLupusUpper ScoA&A proofs: manuscript no. Corona_Australis_arXiv bey et al. 1984). In particular, clouds with higher cs and Ω0 (i.e. warmer or more turbulent) will form more massive disks (also see Appendix A in Visser et al. 2009). Therefore, a cold par- ent cloud or one with low intrinsic angular momentum Ω0, will form disks with a lower mass, and with a lower Mdisk/M(cid:63) as ob- served in CrA. Consistently, observations of dense cloud cores in the CrA cloud show line-widths lower than in other regions (Tachihara et al. 2002). Moreover, because of the smaller circu- larization radius, the formed disks would alse be smaller (e.g. Dullemond et al. 2006) and potentially mostly optically thick, thus hiding an even larger fraction of the mass . Alternatively, small and optically thick disks could result from magnetic brak- ing of the disks by means of the magnetic field threading the disk and the surrounding molecular cloud at the formation stage (e.g. Mellon & Li 2008; Herczeg & Hillenbrand 2014; Krumholz et al. 2013). The same scenario was proposed by Maury et al. (2019) to explain the low occurrence of large (> 60 AU) Class 0 disks in the CALYPSO sample. Such scenarios, although not testable with the present dataset, are consistent with the low disk mass distribution and with the low intercept of the Mdisk − M(cid:63) in CrA and are not in contradiction with the young age of the stellar popultion. If the parent cloud initial conditions are indeed responsible for the low masses observed, this would be an additional critical aspect to be considered when studying planet formation and evolution. Since the conditions at the epoch of disk formation can be different in each star-forming region, proper modelling is required to assert to which extent they can affect the initial disk mass distribution, the subsequent disk evolution, planet formation and planetary populations. Observationally, this could be tested by observing the mass of disks around Class 0 and Class I objects in CrA: if the disks are born with a low-mass, the disk mass distribution even at these younger stages should be significantly lower than in other re- gions. 6. Conclusion We presented the first ALMA survey of 43 Class II protoplane- tary disks in the Corona Australis nearby (d = 160 pc) star form- ing region, in order to measure their dust content and understand how it scales with the stellar properties. The ultimate goal was to test if the relations between disk properties, age of the stellar population found in other surveys also hold for this region. 1. The average mm fluxes from the disks in CrA is low. This in turn converts into a low disk mass distribution. Even though our observations are able to constrain dust masses down to ∼ 0.2 M⊕, the detection rate is only 56%. Moreover, we find that only 3 disks in our sample have a dust mass ≥ 10 M⊕ and thus sufficient mass to form giant planet cores. 2. We obtained VLT/X-Shooter spectra for 8 objects with previ- ously unknown spectral type, and derived their stellar physi- cal properties. 3. Despite the apparent young age of the CrA stellar population, we find that the dust mass distribution of the disks in CrA is much lower than that of the Lupus young star forming region which shares a similar age, while it appears to be consistent with that in the 5-10 Myr old Upper Sco association. The correlation between disk dust mass Mdust and stellar mass M(cid:63) previously identified in all other surveyed star forming regions is confirmed. However, because of the low mass of the disks in our sample we find a much lower intercept. The large scatter of the data points does not allow the slope of the relation to be well constrained for CrA. Fig. 7: Comparison of the mass distributions of Lupus and USco to that of CrA, following the MC analysis proposed by Andrews et al. (2013). pφ is the probability that the synthetic population drawn from the com- parison sample (Lupus and Upper Sco) and the reference sample come from the same parent population. f (< pφ) is the cumulative distribution for pφ resulting from the logrank two-sample test for censored datasets after 104 MC iterations. indicates that the difference in disk masses cannot only be as- cribed to different stellar populations and that some other factor, such as disk evolution and the age of the system, must play a role. This process is repeated 104 times, and the results are used to create the cumulative distributions shown in Fig. 7. When using Upper Sco as a comparison sample, we find a median pφ value of 0.53, while the median pφ for Lupus is only 0.004. The conclu- sion is that even when accounting for the Mdust−M(cid:63) relation, the disk dust mass distribution of CrA appears to be statistically dif- ferent from that of Lupus, while it is significantly more similar to that from that of Upper Sco. Therefore, the comparably low masses of the protoplanetary disks in CrA cannot be explained in terms of the low stellar masses. Another way a disk can lose part of its mass is via tidal in- teraction with other stars (e.g. Clarke & Pringle 1993; Pfalzner et al. 2005). This mechanism is, however, only effective in much denser environments than CrA (e.g. Winter et al. 2018b). In prin- ciple, it is possible to imagine that at very early stages most of the stars were located in a dense region (e.g. the Coronet) where they interacted violently before being ejected. However, the very low velocity dispersion of the stars in the cluster makes this scenario very unlikely (Neuhäuser et al. 2000). Tidal interaction can be effective in removing dust mass from a disk even in later stages when the disk is in a binary system (e.g. Artymowicz & Lubow 1994), as proposed to explain the low mm flux of some objects in Taurus by Long et al. (2018). A higher than usual binary fraction could therefore explain the low disk masses observed in CrA. However, Ghez et al. (1997) show that the binary fraction of CrA is indistinguishable from those of Lupus and Chameleon I. Finally it is possible that the low mass distribution observed today is a consequence of a population of disks that has formed with a low mass from the very beginning. For example, the disk formation efficiency in a cloud with mass M0 depends on the sound speed cs and on the solid body rotation rate Ω0, where we have defined the disk formation efficiency as the fraction of M0 that is in the disk at the end of the collapse stage, or as the ratio between Mdisk/M(cid:63) at that time (Cassen & Moosman 1981; Tere- Article number, page 10 of 20 0.00010.0010.010.11.0p0.00.20.40.60.81.0f(p)23Upper ScoLupusP. Cazzoletti et al.: ALMA band 6 survey for Class II YSOs in CrA 4. Since most of the age estimates of the CrA regions are based on the population of the compact Coronet cluster, a possible explanation for the low disk masses might be in principle that CrA also hosts an old population of disks, consistently with previous observations. The position of the objects of our sample on the HR diagram, however, seems to support the idea of a mostly coeval, young population. 5. Low disk masses in a young star forming region can be ex- plained by external photo-evaporation (as in the case of σ Ori) or by a low stellar mass population (as in IC348). With our analysis, we can rule out both these scenarios for CrA. Tidal interaction between different members of CrA, strip- ping material from the disks, as well as close binaries can also be ruled out. 6. We suggest that initial conditions may play a crucial role in setting the initial disk mass distribution and its subsequent evolution. Small disks with low mass can originate from a cloud with very low turbulence or sound speed, or can alter- natively result from disk magnetic braking. It is therefore im- portant to better study the impact of initial conditions on the disk properties, especially if planet formation occurs even before 1 Myr age, as the recent results from Tychoniec et al. (2018) and Manara et al. (2018) suggest. Future surveys including younger Class 0 and I objects in CrA and other star forming regions will help testing wether or not ini- tial conditions play a critical role in shaping the physical proper- ties of circumstellar disks. Acknowledgements. We thank S. van Terwisga, S. Andrews, G. Lodato and A. Hacar for very useful discussion, and Dr. Mark Gurwell for compiling the SMA Calibrator List (http://sma1.sma.hawaii.edu/callist/callist.html). We also acknowledge the DDT Committee and the Director of the La Silla and Paranal Observatory for granting DDT time. This work was partly supported by the Ital- ian Ministero dell'Istruzione, Università e Ricerca through the grant Progetti Pre- miali 2012 -- iALMA (CUP C52I13000140001), by the Deutsche Forschungs- gemeinschaft (DFG, German Research Foundation) - Ref no. FOR 2634/1 TE 1024/1-1, and by the DFG cluster of excellence Origin and Structure of the Universe (www.universe-cluster.de). This project has received funding from the European Union's Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No 823823. H.B.L. is supported by the Ministry of Science and Technology (MoST) of Taiwan (Grant Nos. 108-2112- M-001-002-MY3 and 108-2923-M-001-006-MY3). J.M.A. acknowledges finan- cial support from the project PRIN-INAF 2016 The Cradle of Life -- GENESIS- SKA (General Conditions in Early Planetary Systems for the rise of life with SKA). C.F.M. and S.F. acknowledge an ESO Fellowship. M.T. has been sup- ported by the DISCSIM project, grant agreement 341137 funded by the Eu- ropean Research Council under ERC-2013-ADG and by the UK Science and Technology research Council (STFC). Y.H. is supported by the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the Na- tional Aeronautics and Space Administration. C.C.G and R.G.M acknowledge financial support from DGAPA UNAM. This paper makes use of the following ALMA data: ADS/JAO.ALMA#2015.1.01058.S. ALMA is a partnership of ESO (representing its member states), NSF (USA) and NINS (Japan), together with NRC (Canada) and NSC and ASIAA (Taiwan), in cooperation with the Repub- lic of Chile. The Joint ALMA Observatory is operated by ESO, AUI/NRAO and NAOJ. All the figures were generated with the python-based package matplotlib (Hunter 2007). References Alcalá, J. M., Natta, A., Manara, C. F., et al. 2014, A&A, 561, A2 Alcalá, J. M., Manara, C. F., Natta, A., et al. 2017, A&A, 600, A20 Alibert, Y., Mordasini, C., & Benz, W. 2011, A&A, 526, A63 Artymowicz, P., & Lubow, S. H. 1994, ApJ, 421, 651 Andrews, S. M., & Williams, J. P. 2005, ApJ, 631, 1134. Andrews, S. M., Wilner, D. J., Hughes, A. M., Qi, C., & Dullemond, C. P. 2009, Andrews, S. M., Rosenfeld, K. A., Kraus, A. L., & Wilner, D. J. 2013, ApJ, 771, ApJ, 700, 1502 129 Andrews, S. M., Huang, J., Pérez, L. M., et al. 2018, ApJ, 869, L41 Ansdell, M., Williams, J. P., van der Marel, N., et al. 2016, ApJ, 828, 46 Ansdell, M., Williams, J. P., Manara, C. F., et al. 2017, AJ, 153, 240. Ansdell, M., Williams, J. P., Trapman, L., et al. 2018, ApJ, 859, 21 Antonellini, S., Kamp, I., Riviere-Marichalar, P., et al. 2015, A&A, 582, A105 Baraffe, I., Homeier, D., Allard, F., & Chabrier, G. 2015, A&A, 577, A42 Barenfeld, S. A., Carpenter, J. M., Ricci, L., et al. 2016, ApJ, 827, 142. Beckwith, S. V. W., Sargent, A. I., Chini, R. S., & Guesten, R. 1990, AJ, 99, 924 Bell, C. P. M., Naylor, T., Mayne, N. J., et al. 2013, MNRAS, 434, 806. Benz, W., Ida, S., Alibert, Y., Lin, D., & Mordasini, C. 2014, Protostars and Planets VI, 691 Bonfils, X., Delfosse, X., Udry, S., et al. 2013, A&A, 549, A109. Bouy, H., Brandner, W., Martín, E. L., et al. 2004, A&A, 424, 213 Cardelli, J. A., Clayton, G. C., & Mathis, J. S. 1989, ApJ, 345, 245 Carmona, A., van den Ancker, M. E., & Henning, T. 2007, A&A, 464, 687 Carrasco-González, C., Henning, T., Chandler, C. J., et al. 2016, ApJ, 821, L16 Cassen, P., & Moosman, A. 1981, Icarus, 48, 353 Cieza, L. A., Ruíz-Rodríguez, D., Hales, A., et al. 2019, MNRAS, 482, 698. Clarke, C. J., & Pringle, J. E. 1993, MNRAS, 261, 190 Comerón, F. 2008, Handbook of Star Forming Regions, Volume II, 5, 295 Currie, T., & Sicilia-Aguilar, A. 2011, ApJ, 732, 24 Cutri, R. M., Skrutskie, M. F., van Dyk, S., et al. 2003, VizieR Online Data Deller, A. T., Forbrich, J., & Loinard, L. 2013, A&A, 552, A51 de Zeeuw, P. T., Hoogerwerf, R., de Bruijne, J. H. J., Brown, A. G. A., & Blaauw, Catalog , II/246. A. 1999, AJ, 117, 354 378 Dullemond, C. P., Natta, A., & Testi, L. 2006, ApJ, 645, L69 Dullemond, C. P., Birnstiel, T., Huang, J., et al. 2018, ApJ, 869, L46 Dunham, M. M., Allen, L. E., Evans, N. J., II, et al. 2015, ApJS, 220, 11 Dzib, S. A., Loinard, L., Ortiz-León, G. N., Rodríguez, L. F., & Galli, P. A. B. 2018, ApJ, 867, 151 Eisner, J. A., Plambeck, R. L., Carpenter, J. M., et al. 2008, ApJ, 683, 304 Eisner, J. A., Arce, H. G., Ballering, N. P., et al. 2018, ApJ, 860, 77 Evans, N. J., II, Dunham, M. M., Jørgensen, J. K., et al. 2009, ApJS, 181, 321 Facchini, S., Clarke, C. J., & Bisbas, T. G. 2016, MNRAS, 457, 3593 Fedele, D., van den Ancker, M. E., Henning, T., et al. 2010, A&A, 510, A72. Forbrich, J., & Preibisch, T. 2007, A&A, 475, 959 Gaia Collaboration, Prusti, T., de Bruijne, J. H. J., et al. 2016, A&A, 595, A1 Gaia Collaboration, Brown, A. G. A., Vallenari, A., et al. 2018, A&A, 616, A1 Galván-Madrid, R., Liu, H. B., Manara, C. F., et al. 2014, A&A, 570, L9 Garcia Lopez, R., Natta, A., Testi, L., & Habart, E. 2006, A&A, 459, 837 Ghez, A. M., McCarthy, D. W., Patience, J. L., & Beck, T. L. 1997, ApJ, 481, Gurwell, M. A., Peck, A. B., Hostler, S. R., Darrah, M. R., & Katz, C. A. 2007, From Z-Machines to ALMA: (Sub)Millimeter Spectroscopy of Galaxies, 375, 234 Haisch, K. E., Jr., Lada, E. A., & Lada, C. J. 2001, ApJ, 553, L153 Hennebelle, P., & Ciardi, A. 2009, A&A, 506, L29 Herczeg, G. J., & Hillenbrand, L. A. 2014, ApJ, 786, 97 Herczeg, G. J., & Hillenbrand, L. A. 2015, ApJ, 808, 23 Hernández, J., Hartmann, L., Megeath, T., et al. 2007, ApJ, 662, 1067 Hildebrand, R. H. 1983, QJRAS, 24, 267 Hamaguchi, K., Yamauchi, S., & Koyama, K. 2005, ApJ, 618, 360 Hunter, J. D. 2007, Computing In Science & Engineering, 9, 3 James, D. J., Melo, C., Santos, N. C., & Bouvier, J. 2006, A&A, 446, 971 Jeffries, R. D., Oliveira, J. M., Naylor, T., Mayne, N. J., & Littlefair, S. P. 2007, MNRAS, 376, 580 Kelly, B. C. 2007, ApJ, 665, 1489 Kenyon, S. J., Gómez, M., & Whitney, B. A. 2008, Handbook of Star Forming Regions, Volume I, 4, 405 Krumholz, M. R., Crutcher, R. M., & Hull, C. L. H. 2013, ApJ, 767, L11 Lavalley, M., Isobe, T., & Feigelson, E. 1992, Astronomical Data Analysis Soft- ware and Systems I, 25, 245 Law, C. J., Ricci, L., Andrews, S. M., Wilner, D. J., & Qi, C. 2017, AJ, 154, 255 Luhman, K. L. 2008, Handbook of Star Forming Regions, Volume II, 5, 169 Liu, H. B., Galván-Madrid, R., Forbrich, J., et al. 2014, ApJ, 780, 155 Liu, H. B., Vorobyov, E. I., Dong, R., et al. 2017, A&A, 602, A19 Lindegren, L., Hernandez, J., Bombrun, A., et al. 2018, arXiv:1804.09366 Long, F., Pinilla, P., Herczeg, G. J., et al. 2018, ApJ, 869, 17 López Martí, B., Eislöffel, J., & Mundt, R. 2005, A&A, 444, 175 López Martí, B., Spezzi, L., Merín, B., et al. 2010, A&A, 515, A31 Luri, X., Brown, A. G. A., Sarro, L. M., et al. 2018, A&A, 616, A9 Manara, C. F., Testi, L., Rigliaco, E., et al. 2013a, A&A, 551, A107 Manara, C. F., Testi, L., Herczeg, G. J., et al. 2017, A&A, 604, A127 Manara, C. F., Morbidelli, A., & Guillot, T. 2018, A&A, 618, L3 Mann, R. K., & Williams, J. P. 2010, ApJ, 725, 430 Maucó, K., Hernández, J., Calvet, N., et al. 2016, ApJ, 829, 38 Maury, A. J., André, P., Testi, L., et al. 2019, A&A, 621, A76 McMullin, J. P., Waters, B., Schiebel, D., Young, W., & Golap, K. 2007, Astro- nomical Data Analysis Software and Systems XVI, 376, 127 Mellon, R. R., & Li, Z.-Y. 2008, ApJ, 681, 1356 Article number, page 11 of 20 A&A proofs: manuscript no. Corona_Australis_arXiv 5 INAF-Osservatorio Astronomico di Capodimonte, via Moiariello Leiden, The Netherlands 16, 80131 Napoli, Italy 1 Max-Planck-Institute for Extraterrestrial Physics (MPE), Giessen- bachstr. 1, 85748, Garching, Germany e-mail: [email protected] 2 European Southern Observatory (ESO), Karl-Schwarzschild-Str. 2, 3 Academia Sinica Institute of Astronomy and Astrophysics, Roo- D-85748 Garching, Germany sevelt Rd, Taipei 10617, Taiwan 4 Leiden Observatory, Leiden University, Niels Bohrweg 2, 2333 CA 6 Center for Integrative Planetary Science, University of California at Berkeley, Berkeley, CA 94720, USA 7 Department of Astronomy, University of California at Berkeley, 8 Institute for Astronomy, University of Hawai'i at M¯anoa, Honolulu, Berkeley, CA 94720, USA HI 96822, USA 9 Instituto de Radioastronomía y Astrofísica (IRyA-UNAM), Univer- sidad Nacional Autónoma de México, Campus Morelia, Apartado Postal 3-72, 58090 Morelia, Michoacán, Mexico 10 Department of Physics & Astronomy, University of Victoria, Victo- ria, BC, V8P 1A1, Canada 11 Centre for Astrophysics Research, University of Hertfordshire, Col- lege Lane, Hatfield AL10 9AB, UK 12 Harvard-Smithsonian Center for Astrophysics, 60 Garden St, Cam- 13 National Astronomical Observatory of Japan, 2-21-1 Osawa, Mi- bridge, MA 02138, USA taka, Tokyo 181-8588 Japan 14 Anton Pannekoek Institute for Astronomy, University of Amster- dam, PO Box 94249, 1090 GE, Amsterdam, the Netherlands 15 Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA 91109, USA 16 Division of Liberal Arts, Kogakuin University, 1-24-2 Nishi- Shinjuku, Shinjuku-ku, Tokyo 163-8677, Japan 17 Department of Astronomy/Steward Observatory, The University of Arizona, 933 North Cherry Avenue, Tucson, AZ 85721, USA 18 Department of Astronomy, The University of Tokyo, 7-3-1, Hongo, Bunkyo-ku, Tokyo 113-0033, Japan 19 Astrobiology Center of NINS, 2-21-1, Osawa, Mitaka, Tokyo 181- 8588, Japan 20 Institute of Astronomy, University of Cambridge, Madingley Road, Cambridge CB3 0HA, UK 21 Homer L. Dodge Department of Physics and Astronomy, University of Oklahoma, 440 W. Brooks Street, Norman, OK 73019, USA Meyer, M. R., & Wilking, B. A. 2009, PASP, 121, 350 Morbidelli, A., & Raymond, S. N. 2016, Journal of Geophysical Research (Plan- ets), 121, 1962 Mordasini, C., Alibert, Y., Benz, W., et al. 2012, A&A, 541, A97. Modigliani, A., Goldoni, P., Royer, F., et al. 2010, Proc. SPIE, 7737 Neuhäuser, R., Walter, F. M., Covino, E., et al. 2000, A&AS, 146, 323 Neuhäuser, R., & Forbrich, J. 2008, Handbook of Star Forming Regions, Volume II, 5, 735 Nisini, B., Antoniucci, S., Giannini, T., & Lorenzetti, D. 2005, A&A, 429, 543 Ortiz-León, G. N., Loinard, L., Kounkel, M. A., et al. 2017, ApJ, 834, 141 Pascucci, I., Testi, L., Herczeg, G. J., et al. 2016, ApJ, 831, 125. Patten, B. M. 1998, Cool Stars, Stellar Systems, and the Sun, 154, 1755 Peterson, D. E., Caratti o Garatti, A., Bourke, T. L., et al. 2011, ApJS, 194, 43 Pfalzner, S., Vogel, P., Scharwächter, J., & Olczak, C. 2005, A&A, 437, 967 Reipurth, B., & Zinnecker, H. 1993, A&A, 278, 81 Riddick, F. C., Roche, P. F., & Lucas, P. W. 2007, MNRAS, 381, 1067 Romero, G. A., Schreiber, M. R., Cieza, L. A., et al. 2012, ApJ, 749, 79 Ruíz-Rodríguez, D., Cieza, L. A., Williams, J. P., et al. 2018, MNRAS, 478, 3674 Schaefer, G. H., Hummel, C. A., Gies, D. R., et al. 2016, AJ, 152, 213 Sicilia-Aguilar, A., Henning, T., Juhász, A., et al. 2008, ApJ, 687, 1145 Sicilia-Aguilar, A., Henning, T., Kainulainen, J., & Roccatagliata, V. 2011, ApJ, 736, 137 Siess, L., Dufour, E., & Forestini, M. 2000, A&A, 358, 593 Tazzari, M., Testi, L., Natta, A., et al. 2017, A&A, 606, A88 Tachihara, K., Onishi, T., Mizuno, A., & Fukui, Y. 2002, A&A, 385, 909 Takami, M., Bailey, J., & Chrysostomou, A. 2003, A&A, 397, 675 Terebey, S., Shu, F. H., & Cassen, P. 1984, ApJ, 286, 529 Testi, L., D'Antona, F., Ghinassi, F., et al. 2001, ApJ, 552, L147. Testi, L., Natta, A., Scholz, A., et al. 2016, A&A, 593, A111 Tychoniec, Ł., Tobin, J. J., Karska, A., et al. 2018, ApJS, 238, 19 Vernet, J., Dekker, H., D'Odorico, S., et al. 2011, A&A, 536, A105 Visser, R., van Dishoeck, E. F., Doty, S. D., & Dullemond, C. P. 2009, A&A, 495, 881 Vorobyov, E. I., Akimkin, V., Stoyanovskaya, O., Pavlyuchenkov, Y., & Liu, H. B. 2018, A&A, 614, A98 Walter, F. M., Vrba, F. J., Wolk, S. J., Mathieu, R. D., & Neuhauser, R. 1997, AJ, 114, 1544 Winter, A. J., Clarke, C. J., Rosotti, G., et al. 2018, MNRAS, 478, 2700 Winter, A. J., Clarke, C. J., Rosotti, G., & Booth, R. A. 2018, MNRAS, 475, 2314 Zhu, Z., Hartmann, L., & Gammie, C. 2010, ApJ, 713, 1143 Article number, page 12 of 20 P. Cazzoletti et al.: ALMA band 6 survey for Class II YSOs in CrA Appendix A: Additional stellar properties Tab. A.1 shows a compilation of the most relevant stellar param- eters used in our analysis. The J magnitude is taken from the 2MASS survey (Cutri et al. 2003). The extinctions are either de- rived from our VLT/X-Shooter spectra or from the references in Column 4. Note that the extinctions from Dunham et al. (2015) were not derived from the stellar spectra but from extinction maps and might therefore systematically overestimate the real extinction towards the star by 1-2 mag (see Sec. 4.1 in Peterson et al. 2011). In order to make sure that the extinction values from Dunham et al. (2015) were accurate, we compared them to those derived from the spectra by Sicilia-Aguilar et al. (2011) for 8 tar- gets common to the two samples. We found that the extinctions derived with the two methods are consistent within the uncer- tainties. The effective temperatures and bolometric luminosities have finally been derived as explained in Sec. 5.1 and then used to determine the stellar masses as explained in Sec. 4.1. Appendix B: VLT/X-Shooter Spectra In this section we present the VLT/X-Shooter spectra obtained Fig. B.1). The Spectral Types derived from the different spec- tral indices calibrated are presented in Tab. B.1. In particular, SpT VIS was obtained from the average values from Riddick et al. (2007), as in Manara et al. (2013a, 2017); SpT TiO was obtained with the index by Jeffries et al. (2007); SpT NIR was obtained with the indices by Testi et al. (2001), as in Manara et al. (2013a); the uncertainties represent the spread between the different indices in the VIS and NIR arms. The adopted spectral types, reported in the last column of Tab. B.1, are taken from the indices calculated in the VIS arm of the spectrum. In addi- tion, the log of the our X-Shooter observations is presented in Tab. B.2 along with the Signal to Noise ratio achieved at differ- ent wavelengths. Article number, page 13 of 20 A&A proofs: manuscript no. Corona_Australis_arXiv Table A.1: Compilation of the most relevant stellar properties used in our analysis. Only the stars with known Spectral Type were included. Source Av Ref. log L(cid:63)/L(cid:12) J [mag] 10.99 13.98 10.77 12.92 10.38 14.19 12.95 12.83 14.85 14.45 13.90 14.91 12.33 14.08 15.58 13.41 9.31 10.59 12.03 14.57 11.61 10.46 11.91 13.67 14.06 10.82 13.38 7.60 9.78 12 12.31 8.49 8.49* 8.93 7.49 11.82 12.44 M(cid:63) [M(cid:12)] 0.10s CrA-1 0.04b CrA-4 0.21b CrA-6 0.05b CrA-8 0.45b CrA-9 0.16b CrA-10 0.09b CrA-12 0.38b CrA-13 0.24b CrA-15 0.32b CrA-16 0.41b CrA-18 0.41b CrA-21 0.14b CrA-22 0.04b CrA-23 0.04b CrA-26 0.12b CrA-28 0.53s CrA-30 0.23b CrA-31 0.12b CrA-35 0.14b CrA-36 0.18b CrA-40 0.40b CrA-41 0.24b CrA-45 0.05b CrA-47 0.08b CrA-48 0.52b CrA-52 0.09b CrA-53 0.76s CrA-54 0.87b CrA-55 0.20b CrA-56 0.14b CrA-57 0.69s SCrA N 0.69s SCrA S 2.25s TCrA 4.10s TYCrA 0.25s Halpha15 0.20s ISO-CrA-177 Av references. (1) This work (2) Dunham et al. (2015) (3) Sicilia-Aguilar et al. (2011) (4) Patten (1998) (5) López Martí et al. (2005) Evolutionary Tracks. (s) Siess et al. (2000) (b) Baraffe et al. (2015) Teff [K] 2860 2770 3190 2770 3720 3190 2980 3560 3300 3485 3640 3560 3085 2770 2770 3085 3810 3300 2980 2980 3085 3560 3300 2860 2980 3720 2980 4020 4210 3190 3085 3900 3900 7200 10500 3190 3085 [mag] 0.0 3.3 2.2 1.2 2.1 2.7 1.4 8.1 14.0 17.0 14.0 13.5 1.1 0.08 4.2 1.9 3.3 2.0 2.1 12.1 4.0 4.7 5.0 0.0 0.0 0.2 1.5 1.4 1.0 2.2 0.8 7.9 7.9 7.9 7.9 0.8 0.5 -0.90 -1.73 -0.52 -1.55 -0.29 -1.83 -1.51 -0.61 -0.78 -0.24 -0.35 -0.82 -1.28 -2.14 -2.26 -1.62 0.28 -0.45 -1.06 -0.93 -0.66 -0.05 -0.63 -1.97 -2.11 -0.69 -1.67 0.77 -0.11 -1.01 -1.31 0.97 0.97 1.46 2.47 -0.28 -0.54 1 2 2 2 2 3 2 3 3 3 3 3 1 3 1 3 2 1 2 1 1 3 1 1 1 3 1 3 3 3 1 2 2 2 2 4 5 Article number, page 14 of 20 P. Cazzoletti et al.: ALMA band 6 survey for Class II YSOs in CrA Table B.1: Spectral types derived from different spectral indices SpT TiO SpT VIS Source SpT NIR SpT CrA-1 M6.05±1.3 M5.46 M7.70±1.3 M6 CrA-22 M3.74±2.1 M4.48 M5.54±2.1 M4.5 CrA-26 M6.68±1.7 M0.64 L1 ±1.7 M7(cid:63) CrA-31 M3.66±3.0 M3.61 M7.96±3.0 M3.5 L1 ±2.3 CrA-36 M4.86±2.3 M2.84 M5(cid:63) CrA-40 M3.07±1.6 M6.42±1.6 M4.5(cid:63) CrA-42 M4.44±3.5 L2 ±3.5 ... CrA-45 M3.56±1.3 M2.22 M5.37±1.3 M3.5 CrA-47 M5.74±2.0 M5.87 L0.92±2.0 M6 CrA-48 M3.17±2.1 M5.22±2.1 M5(cid:63) CrA-53 M5.02±1.1 M5.13 M7.90±1.1 M5 CrA-57 M4.05±1.8 M4.57 M5.89±1.8 M4.5 L1.87±5.4 ... IRS10 (cid:63) Uncertain estimate of SpT due to the low S/N of the spectra. ... ... ... ... ... Table B.2: Night log and basic information on the spectra. In Column 1 is the name of the source, in Column 2 the date and time of the observations, in Column 3-5 the exposure times, in Column 6-8 the slit widths, in Column 9-11 the S/N measured at the indicated wavelengths, in Column 12-13 we show whether or not the Hα and Li lines have been detected. Source Date of observation [UT] Exp. Time [Nexp× s] NIR VIS UVB Slit width [(cid:48)(cid:48)] S/N @ λ [nm] Hα Li UVB VIS NIR 400 700 1000 Pr.Id. 299.C-5048 (PI Manara) CrA-31 CrA-36 CrA-42 CrA-45 2017-09-01T03:30:30.048 2017-09-17T02:22:50.221 2017-09-09T02:14:39.978 2017-09-06T00:37:53.044 4x215 4x600 4x630 4x440 4x135 4x690 4x700 4x340 4x3x75 4x3x250 4x3x250 4x3x150 1.0 1.0 1.0 1.0 Pr.Id. 0101.C-0893 (PI Cazzoletti) CrA-1 CrA-22 CrA-26 CrA-40 CrA-47 CrA-48 CrA-53 CrA-57 IRS10 2018-05-28T04:25:20.708 2018-06-14T06:36:52.359 2018-06-12T05:54:41.945 2018-05-26T08:08:42.457 2018-06-15T04:01:23.919 2018-05-26T05:05:48.019 2018-05-26T06:41:48.012 2018-05-27T05:18:45.651 2018-06-11T06:08:20.787 4x90 4x190 4x190 4x190 4x190 4x190 4x190 4x90 4x190 4x150 4x250 4x250 4x250 4x250 4x250 4x250 4x150 4x250 4x150 4x250 4x250 4x250 4x250 4x250 4x250 4x150 4x250 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 0.9 0.9 0.9 0.9 0.9 0.9 0.9 0.9 0.9 0.9 0.9 0.9 0.9 0.9 0.9 0.9 0.9 0.9 0.9 0.9 0.9 0.9 0.9 0.9 0.9 0.9 8 0 0 0 24 3 0 1 1 11 1 3 0 20 0 0 24 14 20 2 56 13 21 14 17 0 21 23 1 1110 35 58 13 69 309 20 52 83 2 Y Y N Y Y Y Y Y Y Y Y Y N Y N N Y Y ... N Y Y N Y Y N Article number, page 15 of 20 A&A proofs: manuscript no. Corona_Australis_arXiv Fig. B.1: Spectra observed in our X-Shooter programs (black) along with a template with the same Spectral Type (red). The name of the sources is in the title of each subfigure. The absolute flux of each template was normalized to the flux of the observation at λ = 1µm. Article number, page 16 of 20 1000Wavelength [nm]101810161014Flux [ergs1cm2nm1]CrA11000Wavelength [nm]1019101710151013Flux [ergs1cm2nm1]CrA221000Wavelength [nm]101910161013Flux [ergs1cm2nm1]CrA26P. Cazzoletti et al.: ALMA band 6 survey for Class II YSOs in CrA Fig. B.1: Continued Article number, page 17 of 20 1000Wavelength [nm]1018101610141012Flux [ergs1cm2nm1]CrA311000Wavelength [nm]101810141010106Flux [ergs1cm2nm1]CrA361000Wavelength [nm]10191017101510131011Flux [ergs1cm2nm1]CrA40A&A proofs: manuscript no. Corona_Australis_arXiv Fig. B.1: Continued Article number, page 18 of 20 1000Wavelength [nm]1014109104101Flux [ergs1cm2nm1]CrA421000Wavelength [nm]10181016101410121010Flux [ergs1cm2nm1]CrA451000Wavelength [nm]1020101810161014Flux [ergs1cm2nm1]CrA47P. Cazzoletti et al.: ALMA band 6 survey for Class II YSOs in CrA Fig. B.1: Continued Article number, page 19 of 20 1000Wavelength [nm]1021101910171015Flux [ergs1cm2nm1]CrA481000Wavelength [nm]1020101810161014Flux [ergs1cm2nm1]CrA531000Wavelength [nm]1020101810161014Flux [ergs1cm2nm1]CrA57A&A proofs: manuscript no. Corona_Australis_arXiv Fig. B.1: Continued Article number, page 20 of 20 1000Wavelength [nm]101810151012109Flux [ergs1cm2nm1]IRS10
1511.09401	1	1511	2015-11-30T17:38:13	Absolute magnitudes and phase coefficients of trans-Neptunian objects	[ "astro-ph.EP" ]	Context: Accurate measurements of diameters of trans-Neptunian objects are extremely complicated to obtain. Thermal modeling can provide good results, but accurate absolute magnitudes are needed to constrain the thermal models and derive diameters and geometric albedos. The absolute magnitude, Hv, is defined as the magnitude of the object reduced to unit helio- and geocentric distances and a zero solar phase angle and is determined using phase curves. Phase coefficients can also be obtained from phase curves. These are related to surface properties, yet not many are known. Aims: Our objective is to measure accurate V band absolute magnitudes and phase coefficients for a sample of trans-Neptunian objects, many of which have been observed, and modeled, within the 'TNOs are cool' program, one of Herschel Space Observatory key projects. Methods: We observed 56 objects using the V and R filters. These data, along with those available in the literature, were used to obtain phase curves and measure V band absolute magnitudes and phase coefficients by assuming a linear trend of the phase curves and considering magnitude variability due to rotational light-curve. Results: We obtained 237 new magnitudes for the 56 objects, six of them with no reported previous measurements. Including the data from the literature we report a total of 110 absolute magnitudes with their respective phase coefficients. The average value of Hv is 6.39, bracketed by a minimum of 14.60 and a maximum of -1.12. In the case of the phase coefficients we report 0.10 mag per degree as the median value and a very large dispersion, ranging from -0.88 up tp 1.35 mag per degree.	astro-ph.EP	astro-ph	Astronomy&Astrophysicsmanuscript no. ast_ph January 12, 2021 c(cid:13)ESO 2021 5 1 0 2 v o N 0 3 . ] P E h p - o r t s a [ 1 v 1 0 4 9 0 . 1 1 5 1 : v i X r a Absolute magnitudes and phase coefficients of trans-Neptunian objects⋆ A. Alvarez-Candal1, N. Pinilla-Alonso2, J.L. Ortiz3, R. Duffard3, N. Morales3, P. Santos-Sanz3, A. Thirouin4, and J.S. Silva1 1 Observatório Nacional / MCTI, Rua General José Cristino 77, Rio de Janeiro, RJ, 20921-400, Brazil e-mail: [email protected] 2 Department of Earth and Planetary Sciences, University of Tennessee, Knoxville, TN, 37996, United States 3 Instituto de Astrofísica de Andalucía, CSIC, Apt 3004, 18080, Granada, Spain 4 Lowell Observatory, 1400 W Mars Hill Rd, Flagstaff, 86001 Arizona, USA Received -- -- , -- -; accepted -- -- , -- - ABSTRACT Context. Accurate measurements of diameters of trans-Neptunian objects are extremely compli- cated to obtain. Thermal modeling can provide good results, but accurate absolute magnitudes are needed to constrain the thermal models and derive diameters and geometric albedos. The absolute magnitude, HV, is defined as the magnitude of the object reduced to unit helio- and geocentric distances and a zero solar phase angle and is determined using phase curves. Phase coefficients can also be obtained from phase curves. These are related to surface prop- erties, yet not many are known. Aims. Our objective is to measure accurate V band absolute magnitudes and phase coefficients for a sample of trans-Neptunian objects, many of which have been observed, and modeled, within the "TNOs are cool" program, one of Herschel Space Observatory key projects. ⋆Based in part on observations collected at the German-Spanish Astronomical Center, Calar Alto, operated jointly by Max-Planck-Institut für Astronomie and Instituto de Astrofísica de Andalucía (CSIC). Based in part on observations made with the Isaac Newton Telescope operated on the island of La Palma by the Isaac Newton Group in the Spanish Observatorio del Roque de los Muchachos of the Instituto de Astrofísica de Canarias. Partially based on data obtained with the 1.5 m telescope, which is operated by the Instituto de Astrofísica de Andalucía at the Sierra Nevada Observatory. Partially based on observations obtained at the Southern Astrophysical Research (SOAR) telescope, which is a joint project of the Ministério da Ciência, Tecnologia, e Inovação (MCTI) da República Federativa do Brasil, the U.S. National Optical Astronomy Ob- servatory (NOAO), the University of North Carolina at Chapel Hill (UNC), and Michigan State University (MSU). Based in part on observations made at the Observatório Astronômico do Sertão de Itaparica operated by the Observatório Nacional / MCTI, Brazil. Partially based on observations made with the Liverpool Tele- scope operated on the island of La Palma by Liverpool John Moores University in the Spanish Observatorio del Roque de los Muchachos of the Instituto de Astrofísica de Canarias with financial support from the UK Science and Technology Facilities Council. Article number, page 1 of 35 A. Alvarez-Candal et al.: Absolute magnitudes of TNOs Methods.We observed 56 objects using the V and R filters. These data, along with those available in the literature, were used to obtain phase curves and measure V band absolute magnitudes and phase coefficients by assuming a linear trend of the phase curves and considering magnitude variability due to rotational light-curve. Results. We obtained 237 new magnitudes for the 56 objects, six of them with no reported pre- vious measurements. Including the data from the literature we report a total of 110 absolute magnitudes with their respective phase coefficients. The average value of HV is 6.39, bracketed by a minimum of 14.60 and a maximum of -1.12. In the case of the phase coefficients we report 0.10 mag per degree as the median value and a very large dispersion, ranging from -0.88 up tp 1.35 mag per degree. Key words. Methods: observational -- Techniques: photometric -- Kuiper belt: general 1. Introduction The phase curve of a minor body shows how the reduced magnitude1 of the body changes with phase angle. The phase angle, α, is defined as the angle, measured at the location of the body, that the Earth and the Sun subtend. These curves show a complex behaviour, for phase angles between 5oand 30othey follow an overall linear trend, while at small angles a departure from linearity often occurs. In 1956 T. Gehrels coined the expression "opposition effect" and attributed it to the sudden increase of brightness at small α shown in the phase curve of asteroid 20 Massalia (Gehrels 1956), although no explanation was offered. Since then many works have modeled phase curves, with or without opposition effect, analysing the relationship between these curves and the properties of the surface: particle sizes, scattering properties, albedos, compaction, or composition, either by using astronomical or laboratory data, or theoretical modeling (e.g., Hapke 1693, Bowell et al.1989, Nelson et al. 2000, Shkuratov et al. 2002 and references therein). Besides providing information about surface properties, phase curves are also important be- cause using them we can measure the absolute magnitude, H, of an airless body. H is defined as the reduced magnitude of an object at α = 0o. Moreover, H is related to the diameter of the body, D, and its geometric albedo p. If we are considering magnitudes in the V band, then D [km] = 1.324 × 10(3−HV/5) √pV . (1) The first minor bodies to have their phase curves measured were asteroids (for instance the aforementioned work by Gehrels in 1956). Nowadays, we know that low-albedo (taxo- nomic classes D, P, or C) asteroids show lower opposition effect spikes than higher-albedo asteroids (S or M asteroids) (Belsakya and Shevchenko 2000). Also modern technologies al- lowed us to obtain incredible data of a handful of objects, see the recent work on comet 67P/Churyumov-Gerasimenko by Fornasier et al. (2015) using data from the ROSETTA 1 The observed standard magnitude normalized to the distance of the Sun and Earth. Article number, page 2 of 35 A. Alvarez-Candal et al.: Absolute magnitudes of TNOs spacecraft; or huge databases, such as the 250,000 absolute magnitudes of asteroids presented by Vereš et al. (2015) from Pan-STARRS. Unfortunately such data is not yet available for objects farther away in the solar system, with the exception of 134340 Pluto. Therefore, many of the physical characteristic of the trans- Neptunian population, for instance size, albedo, or density, are still hidden from us due to the limited quality of the information we can currently obtain: visible and/or near-infrared spectroscopy of about 100 objects (Barucci et al. 2011, and references therein), colors of about 300 (Hainaut et al. 2012) drawn from a known population of more than 1,400 objects. Moreover, these data belong to the largest known trans-Neptunian objects, TNOs, the most easily observed ones, or some Centaurs, which is a population of dynamically unstable objects orbiting among the giant planets, considered as coming from the trans-Neptunian region and therefore representative of this population. Nevertheless, considerable progress has been made in understanding the dynamical structure of the region, but the bulk of the physical characteristics of the bodies that inhabit it remains poorly determined. Several observational studies conducted in the last years show a vast heterogeneity on physical and chemical properties. With the objective of enlarging our knowledge of the TNO population, The Herschel Open Time Key Program on TNOs and Centaurs: "TNOs Are Cool" (Müller et al. 2007) was granted with 372.7 hours of observation on the Herschel Space Observatory (HSO). The observations are complete with a sample of 130 observed objects. The observed data are fed into thermal models (Müller et al. 2010) where a series of free parameters are fitted, among them are pV and D. These two quantities could be constrained using ground based data and thus fixing at least one of them in the modeling improving the accuracy of the results. Among the targets observed with Herschel there are several of them that do not have reliable HV magnitude, which is fundamental for the computation of D and pV (i.e. less uncertainties in HV mean less uncertainties in D and pV). Therefore, the HSO program "TNOs are Cool" needs support observations from ground-based telescopes. One critical problem that arises when studying phase curves of TNOs is the fact that α can only attain low values when observing from Earth-based facilities. For comparison: a typical main belt asteroid can be observed up to 20o or 30o, while a typical TNO can only reach up to 2o. This means that, for TNOs, we are observing well within the opposition effect region, which prevents us to use the full power of photometric models. On the other hand, the phase curves are very well approximated by linear functions within this restricted phase angle region (e.g., Sheppard and Jewitt 2002). Some effort has been made in this direction (see review by Belskaya et al. 2008, or the recent works by Perna et al. 2013 and Böhnhardt et al. 2014) but most of these use limited samples (usually one observation) assuming average values of the phase coefficients. With this in mind we started a survey in various telescopes to obtain V and R magnitudes for several TNOs at as many different phase angles as possible to measure phase curves and through them determine HV. The survey is being carried in both hemispheres using telescopes at different Article number, page 3 of 35 A. Alvarez-Candal et al.: Absolute magnitudes of TNOs locations. In the next section we describe the observations carried and the facilities where the data were obtained. In Sect. 3 we present the results, while their analysis is presented in Sect. 4. The discussion and some conclusions obtained from this work are presented in Sect. 5. 2. Observations and data reduction The data presented in this work were collected during several observing runs spanning between September 2011 and July 2015 for well over 40 nights. The instruments and facilities used were: The Calar Alto Faint Object Spectrograph at the 2.2-m telescope, CAHA2.2, and Multi Ob- ject Spectrograph for Calar Alto at the 3.5-m telescope, CAHA3.5, of the Calar Alto Ob- servatory2 sited at Sierra de Los Filabres (Spain); the Wide Field Camera at the 2.5-m Isaac Newton Telescope, INT, sited at the Roque de los Muchachos Observatory3 (Spain); the direct camera at the 1.5-m telescope, OSN, of the Sierra Nevada Observatory4 (Spain); the SOAR Optical Imager at the 4.1-m Southern Astrophysical Research telescope5 sited at Cerro Pachón (Chile); the direct camera at the 1-m telescope of the Observatório As- tronômico do Sertão de Itaparica6, OASI, Brazil; and the optical imaging component of the Infrared-Optical suite of instruments (IO:O) at the 2.0-m Liverpool telescope, Live, sited at the Roque de los Muchachos Observatory7 (Spain). Descriptions of instruments and telescopes can be found at their respective home pages. We always attempted to observe using the V and R filters sequentially, but in some cases this was not possible, either due to deterioration of weather conditions (i.e., no observation possible) or instrumental/telescope problems. The objects were targeted, whenever possible, at different phase angles aiming at the widest spread possible. Along with the TNOs we targeted several standard stars fields (from Landolt 1992 and Clem and Landolt 2013), or were provided by the observatory, as in the case of the Liverpool telescope, each night. We aimed at observing three different fields at three different airmasses per night covering the range of airmasses of our main targets. Most observations were carried by observing the target during three exposures of 600 s per filter, although in some cases shorter exposures (300 or 400 s) were used to avoid saturation from nearby bright stars or trailing by faster objects (a Centaur can reach up to 2 arcsecs in 10 min). We did not use differential tracking. The combination of the different images allowed us to increase signal-to-noise ratio while keeping trailing at reasonable values. In any case, we found this ap- proach better than, for instance, tracking at non-sidereal rate for 1800 s because during stacking of shorter exposures a better removal of bad pixels, cosmic ray hits, or background sources was obtained. 2 http://www.caha.es 3 http://www.ing.iac.es/Astronomy/telescopes/int/ 4 http://www.osn.iaa.es/content/15-m-telescope 5 http://www.soartelescope.org/ 6 http://www.on.br/impacton/ 7 http://telescope.livjm.ac.uk/ Article number, page 4 of 35 A. Alvarez-Candal et al.: Absolute magnitudes of TNOs Data reduction was performed using standard photometric methods with IRAF. Master bias frames were created from daily files, as well as master flat fields in both filters. Files including TNOs and standard stars fields were bias and flat field calibrated. Data from the Liverpool telescope were provided already calibrated. Identification of the targets was, for most of the objects, straight- forward by blinking different images or, in the most complicated cases, using Aladin8 (Bonnarel et al. 2000). Instrumental apparent magnitudes were obtained using aperture photometry selecting an aperture typically three times the seeing measured in the images, for TNOs and standard stars. Whenever a TNO was too close to another source, either by poor observing timing or by crowded fields, we performed instead aperture correction (see Stetson 1990). Using the standard stars we computed extinction coefficients and color terms to correct the magnitudes of the TNOs, thus m0 = m − χ[k1 + k2(v − r)], (2) where m0 is the apparent instrumental magnitude corrected by extinction (v0 or r0), m is the ap- parent instrumental magnitude (v or r), χ is the airmass, k1 and k2 are the zeroth and first order extinction coefficients, and (v − r) is the apparent instrumental color of the TNO. Next, we translated the m0 to the standard system. The transformation, to order zero, is M = m0 + ZP, (3) where M is the calibrated magnitude, and ZP is the zero point. Note that, as we had many runs in the same telescopes, we computed average extinction coefficients for each site that were used whenever the data did not allow us to compute the night value. The same is true for ZPs. In the particular case of the Liverpool telescope, we used the average extinction coefficient for the Roque de los Muchachos observatory. Table A.1 lists all observed objects, along with its calibrated V and R magnitudes, the night the object was observed, the heliocentric (r) and geocentric (∆) distances and the phase angle (α) at the moment of observation, the telescope used, and a series of notes indicating whether we used average extinction coefficients, or average zero points, or the object had not previous data reported. The errors in the final magnitudes include: (i) the error in the instrumental magnitudes, provided by IRAF (σi); (ii) the error due to atmospheric extintion, estimated as σe = m0 − (m − χk1); and the error in the calibration to the standard system, σZP. Therefore σ2 = σ2 ZP. Note 2. Where σi1 is the error that, whenever aperture correction was performed, σ2 provided by IRAF within the smaller aperture and σi2 is the error in the aperture correction, computed using the task mkapfile within IRAF. i + σ2 e + σ2 i = σi1 2 + σi2 8 http://aladin.u-strasbg.fr/ Article number, page 5 of 35 A. Alvarez-Candal et al.: Absolute magnitudes of TNOs 3. Analysis In total we obtained 237 new magnitudes for 56 objects, 6 of which did not have any magnitude reported before, to the best of our knowledge. The observed objects span from Centaurs up to Detached Objects (from 10 to more that 100 AU) while in eccentricity they reach values as high as 0.9. The inclinations are mostly below 40o with one object at about 80o and one in retrograde orbit (2008 YB3). Alongside our own data we made an extensive, although not complete, search in the literature of other published V and R magnitudes. We used as our primary reference database the MBOSS 2 article by Hainaut et al. (2012), but we did not take the data directly from their catalogue. Instead we took the data from each referenced article to be included in our list. We choose that approach because to make the phase curves we need reduced magnitudes (described in Sect. 3.2), which are computed using the heliocentric and geocentric distances at the moment of observation. At the same time we obtained information regarding the phase angle. It follows that we only used data that were reported along with the site and epoch of observation. We obtained the orbital information from JPL-Horizons9. In those cases where more than one magnitude was reported for the same night, for instance in light-curve analysis, we computed the average value and its standard deviation to use as input. In total we finished with over 1,800 individual measurements for over a hundred objects. Note that each individual measurement corresponds to one observing night, or entry. We did not reject any data based on their reported error bars. Before jumping to the results we stress three important issues: (i) we obtained data for 56 objects but these data alone cannot be used to create phase curves for all the objects, thus we recurred to the literature. In the remaining of the article this augmented set of data will be called our database; (ii) as can be seen in Eq. 1 we cannot split albedo and diameter only using HV, therefore whenever we speak about the brightness of an object we are referring exclusively to its magnitude and not to its albedo properties nor its size, unless explicitly mentioned, and (iii) the magnitudes for the phase curves should be averaged over the rotational period to remove effect of variability due to ∆m > 0, which is not the case for individual measurements. In the following subsections we describe first how we compute the colors for the complete database, and then the construction of the phase curves. 3.1. Colors Being a compilation from different sources our database is very heterogeneous. Some objects have many entries, in a few cases more than fifty, while most have less than ten entries (72% of the sample). Not all entries have data obtained with both filters, in some occasions only V filter was used, while in some others only the R filter magnitude is available. Whenever both magnitudes were available for the same night we computed (V − R). In this way, many objects have more than one measurement of (V − R). In those cases, we computed a weighted average color which we take 9 http://ssd.jpl.nasa.gov/horizons.cgi Article number, page 6 of 35 A. Alvarez-Candal et al.: Absolute magnitudes of TNOs as representative for the object. By doing so we are weighting up the most precise values of (V − R) instead of considering possible changes of color with phase angle, which is out of the scope of the present work. We show in Fig. 1 the color-magnitude diagram for all objects in our sample. If at least one entry for a given object was observed by us, we labeled that object as "This work", while if all observations for a given object were obtained from the literature, the label "Literature" was used. Note that the plot has more than 110 points. In fact we are showing the colors of objects that do not satisfy our criteria to construct the phase curve (see below). Fig. 1. Color - magnitude diagram for the objects in our database. We show in red the objects that have at least one color measured by us, while in blue appear those whose data come from the literature only. The (V − R) is shown for reference as a horizontal line. ⊙ Most objects shown in the figure are redder than the Sun, (V−R) ⊙ = 0.36, shown as a continuous horizontal line. Nevertheless, there are a few bluer objects, (V − R) ≈ 0. The great majority of the objects cluster at V ≈ 23, (V − R) ≈ 0.6. From the figure it is also clear that our observations have a clear cut-off at about V = 22.5, due to the size of the telescopes used, with only one object fainter than V = 23: 2003 QA91, having obvious large error bars. Article number, page 7 of 35 A. Alvarez-Candal et al.: Absolute magnitudes of TNOs 3.2. Phasecurves The main objective of this work is to compute absolute magnitudes, HV, and phase coefficients, β, of as many objects as possible. These data could be used as complement to the Herschel Space Observatory "TNOs are cool" key project. Several papers have already been published presenting HV of different TNOs (e.g., Sheppard and Jewitt 2002, Rabinowitz et al. 2006 and 2007, Perna et al. 2013, Böhnhardt et al. 2014, and others). We do not intend to repeat step by step these works, but to recompute the phase curves making the most of the increasing amount of data available nowadays. We are aware of the risks that arise due to the inhomogeneity of telescopes, instruments, detectors, and epochs. Nevertheless, we consider important to re-analyze the available data using, if not homogeneous inputs, at least homogeneous techniques. As mentioned above, we had to deal with the fact that not all entries (i.e., night of observation for a given object) were complete, in the sense that some objects for a given date were observed only in one filter, V or R. To avoid this problem we decided to construct the individual phase curves using magnitudes measured with the V filter. In those cases where V was not available, we used the average color measured above and the R magnitude to obtain V. We decided, for the scope of this work, to not analize separated the V and R data as we are more interested in having the larger possible quantity of phase curves. For instance, if we use only the V data, without the R data, we only obtain about 50 phase curves. A similar number of phase curves are obtained if using only R data, although not necessarily the same objects. The next step is to compute the reduced V whose notation is V(1, 1, α) which is the value used in the phase curves and represents the magnitude of the object if located at 1 AU from the Sun and being observed at a distance of 1 AU from the Earth. The reduced magnitude is computed from the values of V and the orbital information as V(1, 1, α) = V − 5 log (r∆). (4) We are now left with a set {V(1, 1, α), α} for each object. For the phase curves we only used data for objects that were observed at three different phase angles at least. We discarded a few objects that had a small coverage in α resulting in unreliable values of HV. We analyzed a total of 110 objects. For objects with no reported light-curve amplitude we assumed ∆m = 0 and performed a linear regression to measure HV via V(1, 1, α) = HV + α × β, (5) where β is the change of magnitude per degree, also known as phase coefficient. Each V(1, 1, α) was weighted by its error, assumed equal to that of the V magnitude, or propagated from the R magnitude and that of the average color, while α was assumed as having negligible error. By doing so we obtained HV as the y-intercept and β as the slope of Eq. 5. Article number, page 8 of 35 A. Alvarez-Candal et al.: Absolute magnitudes of TNOs 5.4 5.6 5.8 ) α , 1 , 1 ( V 6.0 0.2 5.5 5.3 5.1 V H 4.9 0.6 α (degrees) 1.0 1.4 0.3 0.7 0.1 β (mag per degree) 0.5 800 700 600 500 400 300 200 100 0 Fig. 2. Example of phase curve of 1996 TL66. Left: scatter plot of V(1, 1, α) versus α. The line represents the solution for HV and β as mentioned in the text. Right: density plot showing the phase space of solutions of Eq. 5 when ∆m , 0, in gray scale. Note that due to the effect of the ∆m values between 5.0 and 5.5 are possible for HV, while the same stands for β ∈ (0.041, 0.706) mag per degree. The continuous lines (red in electronic version) show the area that contain 68.3, 95.5, and 99.7 % of the solutions. We used the linear approach instead of using the full H-G system (Bowell et al. 1989) for simplicity, as we do want to add any more free parameters that will unnecessarily compli- cate the interpretation of results. We also make use of the results presented in Belskaya and Shevchenko (2000), and mentioned in the Introduction, who showed that the opposition effect, the major departure from linearity of the phase curve, is in fact more conspicuous among moderate albedo objects (pV > 0.25) which is not the case for most of the known TNOs (e.g., Lellouch et al. 2013, Lacerda et al. 2014). Some objects do have reported rotational light-curves with non-zero ∆m (we use in this work the data reported in Thirouin et al. 2010 and 2012). Note that ∆m can range up to half a magnitude in extreme, but rare, cases. As we are using reduced magnitudes obtained in different nights, and mostly individual measurements, we have to model the effect of light-curve variations on the value of V(1, 1, α). We proceeded as follows: for an object with ∆m , 0 we generated from {V(1, 1, α), α} new sets {Vi(1, 1, α), α}, with i running from 1 to 10,000, where Vi(1, 1, α) = V(1, 1, α) + randi × ∆m, (6) randi is a random number drawn from a uniform distribution within -1 and 1. By doing so, and feeding these values into Eq. 5, we finish with a set {HVi, βi}, from where we obtain HV and β as the average over the 10,000 realizations. In other words, for objects with ∆m > 0 we have 10,000 different solutions for Eq. 5. We com- puted the average of the solutions for HV and β and assumed these values as the most likely result. A graphical representation of the procedure is shown in Fig. 2. In the figure the left panel shows the representative phase curve along with the data points and their errors, while the right panel shows a two-dimensional histogram showing the phase-space covered by the 10,000 solutions. This method Article number, page 9 of 35 A. Alvarez-Candal et al.: Absolute magnitudes of TNOs y t i t n a u Q 14 12 10 8 6 4 2 0 14 12 10 8 6 HV 4 2 0 −2 Fig. 3. Histogram showing the distribution of obtained HV. allowed us to explore the solution space finding some interesting results, such as those unexpected cases with β < 0, which we will discuss in Sect. 5. All results are shown in Table A.2. The table reports observed object, HV and β, the number of points used in the fits, the light-curve amplitude, and the references to the works whose reported magnitudes were used. 4. Results We measured HV and β for a total of 110 objects. Figure 3 shows the distribution of HV resulting from applying our procedure. The distribution looks bi-modal, with one peak, the larger one, at HV ≈ 7 and a second one at HV ≈ 5. Our results cover a range from a minimum of HV = 14.6 (2005 UJ438) up to a maximum of -1.12 for Eris. The average value is 6.39, while the median is 6.58. The distribution of β is shown in Fig. 4. The average value is 0.09 mag per degree, while the median is 0.10 mag per degree, with a minimum of -0.88 mag per degree for 2003 GH55 and a maximum of 1.35 mag per degree for 2004 GV9. Almost 60 % of the values fall within 0.01 and 0.23 mag per degree. Curiously, the distribution shown in Fig. 4 seems to be the combination of two different distributions, one wide and shallow and a second one sharp and tall. To test this possibility we assumed that the distribution could be fitted by a sum of two Gaussian distributions F(β) = C1e− (β−β1)2 2σ2 1 + C2e− (β−β2)2 2σ2 2 , where Ci, βi, and σi are free parameters. We run a minimization script from python (scipy.optimize.leastsq) obtaining all six free parameters: C1 = 6.8, β1 = 0.27 mag per degree, σ1 = 0.10 mag per degree, and C2 = 26.2, β1 = 0.05 mag per degree, σ1 = 0.11 mag per degree. The best fitting F(β) is shown in Fig. 4, along with the two components. The two-Gaussian model describe rather well the Article number, page 10 of 35 A. Alvarez-Candal et al.: Absolute magnitudes of TNOs y t i t n a u Q 35 30 25 20 15 10 5 0 −1.0 −0.5 0.0 0.5 β (mag per degree) 1.0 1.5 Fig. 4. Histogram showing the distribution of obtained β. The dashed black line is the better fit to the distribution, modeled as the sum of two Gaussian distributions (see text). Each individual Gaussian distribution is shown in continuous green and dotted red lines. distribution of β, both with similar modes but different widths. In the Discussion we will come back to this model. Next, we compare our results with those of a few selected works: Rabinowitz et al. (2007), Perna et al. (2013), and Böhnhardt et al. (2014); and then we search for correlations among our results (HV, colors, β), orbital elements (semi-major axis, eccentricity, inclination), the absolute magnitudes used in the "TNOs are cool" Herschel Space Observatory key project and their mea- sured geometric albedos, and the light-curve amplitude ∆m. Orbital elements for each object were obtained from the Lowell Observatory10. 4.1. Comparisonwithselectedworks On one hand we selected Rabinowitz et al. (2007, Ra07) because it has the most dense phase curves reported for 25 outer solar system objects, while on the other hand Perna et al. (2013, Pe13) and Böhnhardt et al. (2014, Bo14) present results in support for the HSO "TNOs are cool" key project. The three works analyse their data following different criteria: Ra07 observed each target on many occasions, even attempting to obtain rotational properties. If a rotational light-curve could be determined, the data were corrected removing the short term variability, the remaining data were then rebinned in α, and then the phase curves were constructed. Pe13, using less dense data, computed phase curves for a few objects while for these objects with no enough data average values of β were assumed. Bo14 only used average values of β. We report in Table A.3 the values of HV along with the estimated β, if not used an average value. Note that our phase curves include the data reported in these three works. Overall, the agreement is very good among the four works. Nevertheless, there are some values that differ beyond three sigma. For clarity we report here these differences (shown in boldface in 10 ftp://ftp.lowell.edu/pub/elgb/astorb.html Article number, page 11 of 35 A. Alvarez-Candal et al.: Absolute magnitudes of TNOs Table A.3). With Ra07 Makemake (HV and β) and Sedna (HV); with Pe13 2005 UJ438 (HV) and Varda (HV); with Bo14 2003 GH55, 2004 PG115, and Okyrhoe. In this last case the differences are only in HV because these authors did not compute the phase curve, but used instead average values of β to obtain absolute magnitudes. Also, the errors in our data are somewhat larger than those in Ra07, Pe13 and Bo14. We will come back to this issue in the discussion. 4.2. Correlations We searched for possible correlations among pairs of variables. We define here a variable as any given set of quantities representing the population, for instance the variable β is the set of phase co- efficients of the TNOs sample. The correlations were explored using a Spearman test, which has the vantage of being non-parametric relying on ordering the data according to rank and running a linear regression through those ranks. The test returns two values rs, which gives the level of correlation of the tested variables, rs ≈ 1 indicates correlated quantities, while rs → 0 indicates uncorrelated data. The reliability of rs is given by Prs which indicates the probability of two variables being uncorrelated, in practical terms, the closer Prs is to zero, the more likely the result provided by rs becomes. One disadvantage of the Spearman test is that it does not consider the errors in the variables. To overcome this issue we proceeded as follows: let us assume we are trying to find the correlation among a set {x j, y j}, where each quantity x j (y j) has an error of σx j (σy j), j running from 1 up to N. Now we create 10,000 correlations by creating new sets {x ji, y ji}, where x ji = x j + randi × σx j, likewise for y j. In this case randi is a random number drawn from a normal distribution in [−1, 1]. The random number in x j is not necessarily the same as in y j. After performing the 10,000 correlations we finish with a set {rsi, Prs i}, which are displayed in form of density plots to show the likelihood of the correlation to held against the error bars. All relevant results are displayed in Figs. 5 - 11. Table 1 shows the result of the correlation tests: the first column shows the variables tested, second and third column show the nominal values of rs and Prs (those where the errors were not accounted for), while the last column reports our interpretation of the density plots of whether the correlation exists or not. For the scope of the present work we decided not to separate our sample into the sub-populations that appear among Centaurs and TNOs because dividing a sample of 110 objects into smaller sam- ples will just decrease the statistical significance of any possible result. Furthermore, should any real difference arise among any subgroup, this would clearly be seen in any of the tests proposed here. Such as the fact that no large Centaurs are known, or that the so-called Classic TNOs have low inclinations and then to be smaller in size than other subpopulations of TNOs. Be- low we report the most interesting findings of the search for correlations. Thereafter, we discuss on some individual cases that showed interesting or anomalous behaviours. HV vs. semi-major axis: Figure 5 shows the correlation between absolute magnitude and semi- major axis. This correlation is due to observational bias and accounts for the lack of faint objects Article number, page 12 of 35 A. Alvarez-Candal et al.: Absolute magnitudes of TNOs Table 1. Correlations rs Variables HV vs. a HV(ours) vs. HV(HSO) HV vs. pV β vs. HV -0.517 0.987 -0.509 -0.379 HV vs. ∆m 0.359 -0.335 HV vs. Inclination HV vs. e 0.207 β vs. ∆m -0.141 β vs. pV 0.011 0.185 HV vs. V-R 0.090 β vs. V-R 0.233 β vs. a β vs. e 0.137 0.140 β vs. Inclination Prs Correlation 7.6 × 10−9 yes * 7.1 × 10−51 yes 1.8 × 10−5 yes 4.5 × 10−5 weak weak 0.0020 weak 0.0003 no * 0.0299 no 0.2358 0.9341 no no 0.0532 no 0.3474 no 0.0142 0.1525 no no 0.1450 Notes. * Observational bias V H −2 0 2 4 6 8 10 12 14 16 1e−5 1.2 s r P 0.8 0.4 0.0 −0.6 −0.5 rs −0.4 10500 9000 7500 6000 4500 3000 1500 0 101 Semi-major axis (AU) 102 103 Fig. 5. Left: scatter plot of HV vs. semi-major axis. Right: Outcome of the 10,000 realizations in form of a two-dimensional histogram in rs and Prs which shows the phase space where the solutions lie. In a few cases it is relatively clear that a correlation might exists, while in some other cases large excursions are seen, which indicate that an false correlation could arise in the case of large errors. detected at large heliocentric distances, while no bright Centaur (defining loosely a Centaur as an object with semi-major axis below 30 AU) is known to exist. HV(ours) vs. HV(HSO): In this case we compared our computed magnitudes with those used by the Herschel Space Observatory "TNOs are cool" key project. The correlation is close to 1 (Fig. 6), although it is possible to see a small departure at the faint end with two objects with significantly smaller HV, they are (250112) 2002 KY14 (HV = 11.808 ± 0.763) and (145486) 2005 UJ438 (HV = 14.602 ± 0.617). In the first case we revised the data without finding any evident problem and we trust the value to be correct, while in the second case some care should be taken because the minimum value of phase angle is about 5.8o leaving most of the phase curve under sampled, which might be affecting the value of β. For the sake of comparison we fitted a linear function to the data according to HV(ours)= a + b×HV(HSO), obtaining b = 1.06 ± 0.03 and a = −0.27± 0.17. This indicates that, although Article number, page 13 of 35 −2 ) s r u o ( V H 2 6 10 14 A. Alvarez-Candal et al.: Absolute magnitudes of TNOs 1e−18 3 s r P 2 1 10 6 HV (HSO) 2 −2 0 0.84 0.88 0.92 0.96 1.00 rs 10500 9000 7500 6000 4500 3000 1500 0 Fig. 6. Left: scatter plot of HV as measured by us vs (ours). HV as used within the "TNO's are cool" program (HSO). Right: Two-dimensional histogram showing the most likely correlations. 1.2 1.0 0.8 V p 0.6 0.4 0.2 0.8 0.6 0.4 0.2 s r P 9000 7500 6000 4500 3000 1500 0.0 16 14 12 10 6 8 HV 4 2 0 −2 0.0 −0.697 −0.457 −0.218 0.022 0 rs Fig. 7. Left: scatter plot of HV vs. the geometric albedo measured by the "TNOs are cool" program. Right: Two-dimensional histogram showing the most likely correlations. HV(ours) are very similar to HV(HSO) they are not identical. This difference between our HV and those used by the "TNOs are cool" team are probably due to the fact that some of theirs were computed using single observations and assuming an average β. HV vs. pV: Figure 7 shows that there exists a correlation between the absolute magnitude and the geometric albedo, the brigher the object the largest the albedo. This is probably reflecting the fact that brighter objects tend to be the larger ones in size as well, and therefore are able to retain part of the original volatiles more reflective species that smaller objects cannot. β vs. HV: HV seems to have a weak anti-correlation with β indicating that brighter objects have larger positive slopes than fainter ones. From Fig. 8 one interesting details arise: there are a few objects with β < 0 (see also Fig. 4), even considering errorbars and light-curve amplitude (see Table A.2). This issue deserve further study and observations. From the density map the weak correlation seems quite consistent within the errors in HV and β. Article number, page 14 of 35 A. Alvarez-Candal et al.: Absolute magnitudes of TNOs 2 1 0 0.8 s r P 0.4 ) e e r g e d r e p g a m ( β −1 −2 14 10 6 HV 2 −2 0.0 −0.5 −0.3 −0.1 0.1 rs 10500 9000 7500 6000 4500 3000 1500 0 Fig. 8. Left: scatter plot of β vs. HV. Right: Two-dimensional histogram showing the most likely correlations. −2 V H 2 6 10 14 0.0 0.2 0.4 ∆m (mag) 0.6 0.8 0.12 s r P 0.08 0.04 0.00 0.15 0.25 0.35 0.45 rs 10500 9000 7500 6000 4500 3000 1500 0 Fig. 9. Left: scatter plot of HV vs. ∆m. Right: Two-dimensional histogram showing the most likely correla- tions. HV vs. ∆m: There is a weak correlation between absolute magnitude and ∆m pointing that brighter objects tend to have lower ∆m. Interestingly, among the faint object (fainter than HV = 10) no large (> 0.25) amplitudes are found (Fig. 9). It should be remembered that, although faint objects, these are usually in the range 50 to 100 km (see http://public-tnosarecool.lesia.obspm.fr/). In a previous work, Duffard et al. (2009) presented a similar value for this correlation. Also using their results (their Fig. 6) it is possible to see that objects with densities lower than 0.7 g cm−3 are unlikely in hydrostatic equilibrium and therefore could have large ∆m, which is not reflected in our Fig. 9. These density correspond to ≈ 400 km (from Fig. S7 in Ortiz et al. 2012) which is roughly HV ≈ 5.4. Brighter, possibly larger, objects are in hydrostatic equi- librium and their shapes are better described by Mclaurin spheroids whose ∆m are harder to measure due to their symmetry around the minor axis. HV vs. Eccentricity and Inclination There are two curious cases (Fig. 10 and 11). The first one, HV vs. eccentricity indicates that fainter objects tend to have higher eccentricities. This is Article number, page 15 of 35 A. Alvarez-Candal et al.: Absolute magnitudes of TNOs −2 V H 2 6 10 14 0.0 0.4 0.3 s r P 0.2 0.1 0.0 0.05 9000 7500 6000 4500 3000 1500 0 0.15 rs 0.25 0.2 0.4 0.6 Eccentricity 0.8 Fig. 10. Left: scatter plot of HV vs. eccentricity. Right: Two-dimensional histogram showing the most likely correlations. −2 V H 2 6 10 14 0 0.02 s r P 0.01 0.00 −0.45 −0.35 −0.25 rs 10500 9000 7500 6000 4500 3000 1500 0 40 Inclination (degrees) 80 120 Fig. 11. Left: scatter plot of HV vs. inclination. Right: Two-dimensional histogram showing the most likely correlations. an observational bias because faint objects are more easily observed close to perihelion, favoring objects with high eccentricties. The second one, HV vs. inclination, shows also a weak tendency of fainter objects having smaller inclinations. This might be reflecting the known fact that among the so-called "Classical" trans-Neptunian belt are found two subpopulations, the hot and cold (from dynamical considerations) where the cold, low-inclination population does not have objects as large as the hot, high-inclination, population. Note that although both tendencies seem significant over the 2-sigma level (> 95.5 %), only one seems closer to be a correlation having rs > 0.3. Other results: None of the other pairs of variables explored show any significant correlation, therefore their plots are not reported. Interesting objects In this paragraph we describe some objects that deserve more discussion. Article number, page 16 of 35 A. Alvarez-Candal et al.: Absolute magnitudes of TNOs 2060 Chiron, Meech and Belton (1989) detected a coma surrounding Chiron, this result probably influenced the interpretation of latter stellar occultations results (e.g. Bus et al. 1996) that detected secondary events which were associated to jets of material ejected from the surface. A recent re- analysis of all stellar occultation data, along with new photometric data, suggests that Chiron pos- sesses a ring system (Ortiz et al. 2015). Both phenomena, cometary-like activity and the pos- sible ring system, affect the photometric data obtained from Chiron, including the way the photometric measurements are performed, thus increasing the scattering in the phase curve. 10199 Chariklo, Braga-Ribas et al. (2014) detected a ring system around Chariklo using data from a stellar occulation. This result helped to interpret long-term changes in photometric and spectro- scopic data (Duffard et al. 2014) such as the secular variation in reduced magnitude (Belskaya et al. 2010) and the disappearance of a water-ice absorption feature in its near-infrared spectrum (Guil- bert et al. 2009). As with the case of Chiron, the phase curve of Chariklo does not follow a linear trend. Bright objects, those with HV brighter than 3 have β between 0.11 and 0.27 mag per degree. Spectroscopically it is known that these objects (2007 OR10, Eris, Makemake, Orcus, Quaoar, and Sedna) are very different, Eris and Makemake display CH4 absorption features while Orcus, Quaoar, and 2007 OR10 show water ice and probably some hydrocarbons. Therefore, particle size or compaction could play a more important role than composition on the phase curves. 5. Discussion and Conclusions We have observed 56 objects, six of them with no reported magnitudes in the literature, to the best of our knowledge. We combined these new V and R magnitudes with an extensive bibliographic survey to compute absolute magnitudes and phase coefficients. In total we report HV and β for 110 objects. Some of these objects had phase curves already reported, nevertheless it is important to include new data, always keeping in mind that we are combining data from different apparitions for the same object and that surface conditions might have changed between observations. Regarding the distribution of β, Fig. 4 clearly shows a quasi symmetric distribution. The max- imum and mode, coincides, to the second decimal place, with the average and median values: 0.10 mag per degree. We tested the hypothesis of having a two-population distribution, as- suming that each population could be described by a Gaussian function. The fit to the data is quite good, but does this indicate the existence of two real subpopulations? One possible ex- planation regards the quality of the data: A "high-quality" sub-sample clusters with a mode of β2 = 0.11 mag per degree within a sharp distribution, while "low-quality" data is more spread, but with a very similar mode (β2 = 0.10 mag per degree), considering as high-quality data these with small errors, precise β, and with (at least) an estimation of ∆m. Unfortunately, this is not strictly the case, as some of these objects fall within the wide and shallow distribu- tion, far from the sharp and tall distribution. Therefore, even if very tempting, we cannot use the sharp distribution as representative of the whole population as we could introduce unde- Article number, page 17 of 35 A. Alvarez-Candal et al.: Absolute magnitudes of TNOs sired biases in the results. Moreover, most of the objects fall within the wide distribution, 59 %, while 41 % fall within the sharp one. It is clear that there is not one representative value of β for the whole population. Therefore, the use of average values of β to compute HV should be regarded with caution. The phase coeffi- cients range from -0.88 up to 1.35 mag per degree. On the extreme positive side, the two objects (1996 GQ21 and 2004 GV9) have large errors associated. Among the extreme negative values, there are six objects (1998 KG62, 1998 UR43, 2002 GP32, 2003 GH55, 2005 UJ438, and Varda which have β < 0 even considering three times the error. Most of these cases are objects whose data are sparse and with few points. Two of them, UJ438 and Varda, have estimated light-curve amplitude while the rest has no reported value to the best of our knowledge. We are no aware of any physical mechanism that could explain a β < 0 using scattering models. There are some components of the light that could be negative, such as the incoherent second scattering order (Fig. 21 in Shkuratov et al. 2002) which is nonetheless non dominant, especially for the low values of α we can observe TNOs. These extremes values, either positive or negative, could be due to yet undetected phenom- ena such as rotational modulation poorly determined, ring systems, or cometary-like activity and deserve more observations. There are some phase curves that clearly do not follow a linear trend. Those of Chiron and Chariklo, in fact, do not follow any particular trend at all. It is convenient to bear in mind that the photometric models to understand the photometric behaviour of phase curves were thought for objects with nothing else than their bare surface to reflect/scatter/absorb photons. In the case of these possibly ringed systems the reflected light detected on Earth depends not only on the scattering properties of the material covering Chiron, or Chariklo, but as well of the particles in the rings and the geometry of the system. With this in mind, we propose that one criterion to seek candidates to bear ring systems is to search for these "non-linear" behaviour of the phase curve. As examples, based on the dispersion seen in their phase curves we propose that 1996 RQ20, 1998 SN165, or 2004 UX10 as candidates for further studies, among other objects. The correlations were discussed in their respective paragraphs. Overall, some of them are as- sociated to observational biases (HV and semi-major axis; HV and eccentricity), other can be inter- preted in terms of known properties of the TNO region (HV and inclination) while the rest can be considered as weak or non-existing and deserving more data, especially going deeper into the faint end of the population. We do not confirm a proposed anti-correlation between albedo and phase coefficient (see Belskaya et al. 2008 and references therein). One special note is deserved by the anti-correlation found between HV and pV, it would seem that the correlation is driven principally by the brighter objects. We ran the same test discarding those objects brighter than 3, and those associated to the Haumea dynamical group because they form a group that stands aside with par- ticular surface properties, and the relation still holds, rs = −0.356, Prs = 0.0092. Although the correlation does become weaker, without reaching a 3 − σ level, there seems to exists a trend of Article number, page 18 of 35 A. Alvarez-Candal et al.: Absolute magnitudes of TNOs brighter objects to have larger geometric albedos. An in-depth physical explanation remains yet to be formulated. Finally, the errors reported in HV are in some cases larger than previous works. This is reflecting the heterogeneity of the sample, how the effect of the rotational variability is considered, and the weighing of the data while performing the linear fits. Note, for instance, that all of the objects with σHV > 0.1 mag have either less than 10 data points or ∆m > 0.1 mag. Taking this into consideration our results are more accurate, although not as precise, than previous works and probably more real- istic, with the exception of the strategy followed by Rabinowitz et al. (2007). This work represents the first release of data taken at seven different telescopes in six observatories between late 2011 and mid 2015 which represents a large effort. It is important to mention that more observations are ongoing. Acknowledgements. AAC acknowledges support via diverse grants to FAPERJ and CNPq. JLO acknowledges support from the spanish Mineco grant AYA-2011-30106-CO2-O1, from FEDER funds and from the Proyecto de Excelencia de la Junta de Andalucía, J.A. 2012-FQM1776. R.D. acknowledges the support of MINECO for his Ramón y Cajal Contract. The authors would like to thank Y. Jiménez-Teja for technical support and P.H. Hasselmann for helpful discussions regarding phase curves. We are grateful to O. Hainaut whose comments helped to improve the quality of this manuscript. References Barucci, M.A., Romon, J., Doressoundiram, A., et al. 2000, AJ, 120, 496 Barucci„ M.A., Böhnhardt, H., Dotto, E., et al. 2002, A&A, 392, 335 Barucci, M.A., Cruikshank, D.P., Dotto, E., et al. 2005, A&A, 439, L1 Barucci, M.A., Alvarez-Candal, A., Merlin, F., et al. 2011, Icarus, 214, 297 Belskaya, I., & Shevchenko, V. 2000, Icarus, 147, 94 Belskaya, I.N., Levasseur-Regourd, A.-C., Shkuratov, Y. G. et al. 2008, Surface Properties of Kuiper Belt Objects and Centaurs from Photometry and Polarimetry, in The Solar System Beyond Neptune, ed. M.A. Barucci, H. Boehnhardt, D. P. Cruikshank, et al. (Tucson: Univ. of Arizona Press), 115 Belskaya, I.N., Bagnulo, S., Barucci, M.A., et al. 2010, Icarus, 210, 472 Bonnarel, F., Fernique, P., Bienaymé, O., et al. 2000, A&ASS, 143, 33 Böhnhardt, H., Tozzi, G.-P., Birkle, K., et al. 2001, A&A, 378, 653 Böhnhardt, H., Delsanti, A., Barucci, M.A., et al. 2002, A&A, 395, 297 Böhnhardt, H., Schulz, D., Protopapa, S., et al. 2014, EMP, October 2014 Bowell, E., Hapke, B., Domingue, D., et al. 1989, Application of photometric models to asteroids, in Asteroids II, ed. R.P. Binzel, T. Gehrels, & M. Shapely Matthews (Tucson: Univ. of Arizona Press), 524 Braga-Ribas, F., Sicardy, B., Ortiz, J.L., et al. 2014, Nature, 508, 72 Brown, W.R., & Luu, J.X. 1997, Icarus, 126, 218 Buie, M.W., & Bus, S.J. 1992, Icarus, 100, 288 Bus, S.J., Buie, M.W., Schleicher, D.G., et al. Icarus, 123, 478. Carraro, G., Maris, M., Bertin, D., et al. 2006, A&A, 460, L39 Clem, J.L., & Landolt, A.U. 2013, AJ, 146, 88 Davis, D.R., & Farinella, P. 1997, Icarus, 125, 50 Davies, J.K., Green, S., McBride, N., et al. 2000, Icarus, 146, 253 de Bergh, C., Delsanti, A., Tozzi, G.P., et al. 2005, A&A, 437, 1115 DeMeo, F., Fornasier, S., Barucci, M.A., et al. 2009, A&A, 493, 283 Delsanti, A., Böhnhardt, H., Barrera, L, et al. 2001, A&A, 380, 347 Doressoundiram, A., Barucci, M.A., Romon, J., et al. 2001, Icarus, 144, 277 Article number, page 19 of 35 A. Alvarez-Candal et al.: Absolute magnitudes of TNOs Doressoundiram, A., Peixinho, N., de Bergh, C., et al. 2002, ApJ, 124, 2279 Doressoundiram, A., Peixinho, N., Doucet,C., et al. 2005, Icarus, 174, 90 Dotto, E., Barucci, M.A., Böhnhardt, H., et al. 2003, Icarus, 162, 408 Duffard, R., Lazzaro, D., Pinto, S., et al. 2002, Icarus, 160, 44 Duffard, R., Ortiz, J.L., Thirouin, A., et al. 2009, A&A, 505, 1283 Duffard, R., Pinilla-Alonso, N., Ortiz, J.L., et al. 2014, A&A, 568, A79 Farnham, T.L., & Davies, J.K. 2003, Icarus, 164, 418 Ferrin, I., Rabinowitz, D., Schaefer, B., et al. 2001, ApJ, 548, L243 Fornasier, S., Doressoundiram, A., Tozzi, G.P., et al. 2004, A&A, 421, 353 Fornasier, S., Lazzaro, D., Alvarez-Candal, A., et al. 2014, A&A, 568, L11 Fornasier, S., Hasselmann, P.H., Barucci, M.A., et al. 2015, A&A, in press Gehrels, T. 1956, ApJ, 123, 331 Guilbert, A., Barucci, M.A., Brunetto, R., et al. 2009, A&A, 501, 777 Gil-Hutton, R., & Licandro, J. 2001, Icarus, 152, 246 Hainaut, O.R., Böhnhardt, H., & Protopapa, S. 2012, A&A, 546, A115 Hapke, B. 1963, JGR, 68, 4571 Jewitt, D., & Luu, J. 1993, Nature, 362, 730 Jewitt, D., & Luu, J. 1998, AJ, 115, 1667 Jewitt, D., & Luu, J. 2001, AJ, 122, 2099 Jewitt, D.C. 2002, AJ, 123, 1039 Jewitt, D.C., & Sheppard, S.S. 2002, AJ, 123, 2110 Lagerkvist, C.-I., Magnusson, P. 1990, A&ASS, 86, 119 Lellouch, E., Santos-Sanz, P., Lacerda, P., et al. 2013, A&A, 557, A60 Landolt, A.U. 1992, AJ, 104, 340 Lacerda, P., Fornasier, S., Lellouch, E., et al. 2014, ApJL, 793, L2 McBride, N., Davies, J.K., Green, S.F., et al. 1999, MNRAS, 306, 799 McBride, N., Green, S.F., Davies, J.K., et al. 2003, Icarus, 161, 501 Meech, K.J., & Belton, M.J.S. 1989, IAUC Circ. 4770, 1 Mueller, B.E.A., Tholen, D.J., Hartmann, W.K., et al. 1992, Icarus, 97, 150 Müller, T.G., Lellouch, E., Böhnhardt, H., et al. 2007, Earth Moon and Planets, 105, 209 Müller, T.G., Lellouch, E., Stansberry, J., et al. 2010, A&A, 518, 146 Nelson, R.M., Hapke, B.W., Smythe, W.D., et al. 2000, Icarus, 147, 545 Ortiz, J.L., Sota, A., Moreno, R., et al. 2004, A&A, 420, 383 Ortiz, J.L., Sicardy, B., Braga-Ribas, F., et al. 2012, Nature, 491, 566 Ortiz, J.L., Duffard, R., Pinilla-Alonso, N., et al. 2015, A&A, 576, A18 Pinilla-Alonso, N., Alvarez-Candal, A., Melita, M., et al. 2013, A&A 550, A13 Peixinho, N., Lacerda, P., Ortiz, J.L., et al. 2001, A&A, 371, 753 Peixinho, N., Böhnhardt, H., Belskaya, I., et al. 2004, Icarus, 170, 153 Peixinho, N., Delsanti, A., Guilbert-Lepoutre, A., et al. 2012, A&A, 546, A86 Perna, D., Barucci, M.A., Fornasier, S., et al. 2010, A&A, 510, A53 Perna, D., Dotto, E., Barucci, M.A., et al. 2013, A&A, 555, 49 Rabinowitz, D.K., Barkume, K., Brown, M.E., et al. 2006, ApJ, 639, 1238 Rabinowitz, D.L., Schaffer, B.E., & Tourtellotte, W. 2007, AJ, 133, 26 Romanishin, W., Tegler, S.C., Levine, J., et al. 1997, AJ, 113, 1893 Romanishin, W., & Tegler, S.C. 1999, Nature, 398, 129 Romanishin, W., Tegler, S.C., Consolmagno, G.J. 2010, AJ, 140, 29 Romon-Martin, J., Barucci, M.A., de Bergh, C., et al. 2002, Icarus, 160, 59 Santos-Sanz, P., Ortiz, J.L., Barrera, L., et al. 2009, A&A, 494, 693 Article number, page 20 of 35 A. Alvarez-Candal et al.: Absolute magnitudes of TNOs Schaefer, B.E., & Rabinowitz, D.L. 2002, Icarus, 160, 52 Schaller, E., & Brown, M. 2007, ApJ, 659, L61 Sheppard, S.S., & Jewitt, D.C. 2002, AJ, 124, 1757 Sheppard, S.S. 2010, AJ, 139, 1394 Shkuratov, Y., Ovcharenko, A., Zubko, E., et al. 2002, Icarus, 159, 396 Snodgrass, C., Carry, B., Dumas, C., et al. 2010, A&A, 511, A72 Stetson, P.B. 1990, PASP, 102, 932 Tegler, S.C., & Romanishin, W. 1997, Icarus, 126, 212 Tegler, S.C., Romanishin, W., Stone, A. et al. 1997, AJ, 114, 1230 Tegler, S.C., & Romanishin, W. 2000, Nature, 407, 979 Tegler, S.C., & Romanishin, W. 2003, Icarus, 161, 181 Tegler, S.C., Romanishin, W., Consolmagno, G.J. 2003, ApJ, 599, L49 Thirouin, A., Ortiz, J.L., Duffard, R., et al. 2010, A&A, 522, A93 Thirouin, A., Ortiz, J.L., Campo-Bagatin A., et al. 2012, MNRAS, 424, 3156 Vereš, P., Jedicke, R., Fitzsimmons, A., et al. 2015, Icarus, 261, 34 List of Objects Article number, page 21 of 35 A&A -- ast_ph, Online Material p 22 Appendix A: Tables Table A.1. Observations Object V R Night r (AU) ∆ (AU) α (degress) Telescope Notes 24835 1995 SM55 26181 1996 GQ21 26181 1996 GQ21 26181 1996 GQ21 40314 1999 KR16 40314 1999 KR16 47171 1999 TC36 47932 2000 GN171 82075 2000 YW134 82158 2001 FP185 82158 2001 FP185 2001 KD77 139775 2001 QG298 55565 2002 AW197 55565 2002 AW197 2002 GH32 2002 GP32 2002 GP32 95626 2002 GZ32 119951 2002 KX14 250112 2002 KY14 250112 2002 KY14 307261 2002 MS4 307261 2002 MS4 55637 2002 UX25 55637 2002 UX25 55637 2002 UX25 55637 2002 UX25 55638 2002 VE95 55638 2002 VE95 119979 2002 WC19 127546 2002 XU93 127546 2002 XU93 127546 2002 XU93 19.898±0.216 21.536±0.192 21.760±0.217 21.775±0.233 21.871±0.132 21.586±0.091 20.373±0.134 21.313±0.070 21.219±0.401 21.407±0.489 22.354±0.631 21.799±0.181 22.068±0.239 20.720±0.233 21.988±0.203 22.124±0.087 21.824±0.631 20.133±0.201 19.943±0.136 20.413±0.091 20.064±0.053 20.184±0.270 19.474±0.106 20.203±0.072 20.286±0.084 19.800±0.085 20.490±0.104 21.223±0.416 21.099±0.402 21.716±0.401 21.565±0.334 21.214±0.260 19.170±0.132 20.900±0.155 20.516±0.086 20.767±0.083 20.905±0.085 19.504±0.087 20.852±0.069 21.039±0.407 20.779±0.297 20.723±0.399 21.121±0.110 22.076±0.202 19.849±0.116 19.900±0.173 21.726±0.303 21.788±0.069 22.023±0.037 19.829±0.157 20.375±0.253 19.383±0.090 19.550±0.100 18.907±0.073 20.406±0.616 19.632±0.116 19.606±0.044 20.164±0.097 19.545±0.077 19.713±0.106 20.003±0.171 21.281±0.563 21.131±0.294 21.868±0.535 21.033±0.213 2012-12-09 38.4165 37.6015 2014-05-29 42.6600 41.6917 2013-06-10 42.3455 41.4599 2013-06-11 42.3464 41.4692 2013-06-03 35.3552 34.4522 2014-04-02 35.2260 34.4426 2013-09-03 30.5720 29.8969 2014-04-02 28.4086 27.6404 2012-12-09 44.6975 44.1306 2013-04-14 35.4526 34.4818 2013-05-11 35.4714 34.5972 2013-06-03 35.9812 35.0132 2013-07-17 31.7844 31.6575 2013-04-15 46.0579 45.5946 2013-04-17 46.0589 45.6247 2013-06-11 43.4742 42.6321 2013-06-11 32.3989 31.4001 2013-07-16 32.4100 31.6764 2013-04-14 18.5160 17.6439 2013-06-10 39.2847 38.2818 2012-12-08 9.5689 8.8098 2012-12-11 9.5725 8.8448 2013-06-03 47.0005 46.0946 2013-05-10 47.0046 46.3190 2011-10-31 41.4407 40.4513 2012-10-16 41.3080 40.3379 2012-12-11 41.2868 40.5805 2013-09-02 41.1853 40.6445 2012-12-11 28.8622 27.9090 2011-10-31 28.6992 27.9126 2012-12-09 41.7521 40.7709 2012-12-11 21.5317 20.9261 2012-12-10 21.5310 20.9301 2012-12-08 21.5296 20.9391 0.8285 0.4020 0.6737 0.6939 0.7536 1.0274 1.4249 1.3140 1.0432 0.4155 0.8229 0.4968 1.8231 1.1096 1.1283 0.7500 0.3222 1.2527 1.5778 0.2282 3.9111 4.1247 0.5670 0.9151 0.1105 0.3268 0.9553 1.1965 0.4964 1.2313 0.1179 2.0983 2.1084 2.1302 (1) (1) (1) (1) (1) (1) (1) CAHA2.2 SOAR INT INT CAHA3.5 SOAR OSN SOAR CAHA2.2 OSN CAHA3.5 (1) CAHA3.5 CAHA3.5 (1) (1,2) (1) (1) (3) (3) (1) OSN OSN INT INT CAHA3.5 OSN INT CAHA2.2 CAHA2.2 CAHA3.5 CAHA3.5 CAHA2.2 CAHA2.2 CAHA2.2 OSN (1) CAHA2.2 CAHA2.2 (1) CAHA2.2 (1,3) CAHA2.2 CAHA2.2 (1) CAHA2.2 A&A -- ast_ph, Online Material p 23 Table A.1. continued. Object V R Night r (AU) ∆ (AU) α (degress) Telescope Notes 127546 2002 XU93 120132 2003 FY128 120132 2003 FY128 120178 2003 OP32 120178 2003 OP32 120178 2003 OP32 120178 2003 OP32 120178 2003 OP32 120178 2003 OP32 2003 QA91 120181 2003 UR292 143707 2003 UY117 143707 2003 UY117 143707 2003 UY117 84922 2003 VS2 84922 2003 VS2 136204 2003 WL7 136204 2003 WL7 120216 2004 EW95 90568 2004 GV9 307982 2004 PG115 307982 2004 PG115 2004 PT107 2004 PT107 144897 2004 UX10 144897 2004 UX10 144897 2004 UX10 144897 2004 UX10 144897 2004 UX10 230965 2004 XA192 230965 2004 XA192 230965 2004 XA192 230965 2004 XA192 230965 2004 XA192 303775 2005 QU182 21.007±0.144 20.034±0.386 21.063±0.200 20.269±0.124 20.084±0.139 20.168±0.100 20.044±0.155 20.828±0.226 23.790±1.163 22.377±0.734 22.665±1.201 20.503±0.201 19.914±0.219 20.429±0.200 21.377±0.437 20.930±0.142 21.131±0.181 20.362±0.029 21.321±0.212 22.791±0.442 21.906±0.284 19.474±0.106 20.534±0.083 19.517±0.150 20.245±0.104 20.576±0.195 20.508±0.101 20.110±0.144 20.249±0.184 20.468±0.218 20.278±0.122 21.078±0.123 21.180±0.162 19.541±0.299 20.254±0.186 19.794±0.166 20.248±0.185 19.950±0.082 19.891±0.215 20.894±0.283 19.686±0.127 23.158±0.933 21.484±0.370 20.853±0.266 20.614±0.258 21.754±0.405 19.129±0.128 19.973±0.247 20.018±0.155 20.672±0.087 20.513±0.144 20.053±0.038 20.183±0.247 20.414±0.097 21.689±0.183 20.994±0.179 19.632±0.116 19.905±0.052 19.840±0.216 19.833±0.070 20.110±0.107 20.047±0.100 19.718±0.089 21.061±0.211 19.638±0.109 19.658±0.176 20.522±0.088 2012-10-16 21.4937 21.3872 2013-04-15 39.1861 38.1980 2013-04-16 39.1865 38.2004 2012-09-16 41.7560 40.8552 2012-09-17 41.7562 40.8621 2011-09-24 41.6606 40.8276 2011-09-25 41.6609 40.8367 2012-10-17 41.7642 41.1806 2011-10-31 41.6705 41.2996 2013-06-11 44.6206 44.3877 2012-09-19 26.7683 25.9857 2011-10-31 32.8608 31.8758 2011-09-25 32.8502 32.0036 2012-09-17 32.9610 32.2100 2012-12-09 36.5176 35.5619 2012-10-17 36.5148 35.8418 2011-10-31 14.9614 14.2362 2012-10-15 15.0067 14.5833 2014-05-29 27.0955 26.2375 2014-04-02 39.3485 38.5014 2013-09-03 37.3300 36.3912 2013-07-18 37.3077 36.4746 2012-10-16 38.2606 37.7248 2013-06-11 38.2517 37.8609 2011-10-31 39.0037 38.0264 2012-10-16 39.0466 38.0789 2011-09-25 38.9996 38.1471 2011-09-24 38.9994 38.1563 2012-09-19 39.0436 38.2544 2012-12-11 35.6167 34.8167 2012-12-10 35.6169 34.8209 2012-12-08 35.6187 34.8313 2011-10-31 35.6725 35.2058 2012-09-16 35.6294 35.8013 2013-07-18 50.1479 49.9291 2.6493 0.2555 0.2731 0.6137 0.6310 0.7714 0.7886 1.1117 1.2691 1.2746 1.3730 0.2149 0.9558 1.1793 0.3745 1.1682 2.6674 3.5000 1.1547 0.7870 0.5687 0.9090 1.2636 1.4142 0.2566 0.3597 0.7902 0.8129 0.9261 0.9358 0.9453 0.9649 1.4196 1.5890 1.1384 CAHA2.2 OSN OSN CAHA2.2 CAHA2.2 (1) CAHA2.2 CAHA2.2 CAHA2.2 CAHA2.2 (1) (1) INT (1,3) CAHA2.2 CAHA2.2 (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) CAHA2.2 CAHA2.2 CAHA2.2 CAHA2.2 CAHA2.2 CAHA2.2 SOAR SOAR OSN CAHA3.5 CAHA2.2 INT CAHA2.2 CAHA2.2 CAHA2.2 CAHA2.2 CAHA2.2 CAHA2.2 (3) CAHA2.2 (1,3) CAHA2.2 (3) CAHA2.2 (1,3) CAHA2.2 (3) CAHA3.5 A&A -- ast_ph, Online Material p 24 Table A.1. continued. Object V R Night r (AU) ∆ (AU) α (degress) Telescope Notes 145451 2005 RM43 145451 2005 RM43 145451 2005 RM43 145451 2005 RM43 145452 2005 RN43 145452 2005 RN43 145452 2005 RN43 145452 2005 RN43 145452 2005 RN43 145453 2005 RR43 145453 2005 RR43 145453 2005 RR43 145480 2005 TB190 145480 2005 TB190 145480 2005 TB190 145480 2005 TB190 145486 2005 UJ438 145486 2005 UJ438 145486 2005 UJ438 202421 2005 UQ513 202421 2005 UQ513 2007 OC10 225088 2007 OR10 225088 2007 OR10 225088 2007 OR10 225088 2007 OR10 225088 2007 OR10 225088 2007 OR10 309239 2007 RW10 342842 2008 YB3 2013 AZ60 2013 AZ60 65489 Ceto 65489 Ceto 10199 Chariklo 20.033±0.165 19.981±0.123 19.832±0.156 19.799±0.066 20.116±0.149 19.940±0.117 20.100±0.235 20.016±0.090 20.099±0.174 20.118±0.094 20.365±0.220 20.067±0.075 21.305±0.050 21.622±0.334 22.596±0.648 21.486±0.364 20.981±0.470 21.517±1.266 20.675±0.209 21.388±0.342 20.347±0.206 20.994±0.185 21.358±0.476 22.061±0.653 21.700±0.158 21.974±0.166 21.897±0.253 21.727±0.142 21.240±0.115 18.988±0.065 19.987±0.096 20.188±0.136 22.003±0.558 21.684±0.657 18.812±0.183 19.691±0.127 19.808±0.063 19.347±0.206 19.816±0.073 19.287±0.104 19.398±0.172 19.414±0.109 19.453±0.057 19.526±0.092 19.925±0.109 19.547±0.101 19.828±0.071 20.700±0.049 20.641±0.112 21.191±0.290 20.436±0.194 20.918±0.460 20.113±0.367 19.703±0.140 20.421±0.217 19.981±0.164 20.435±0.094 20.904±0.665 21.515±0.449 2011-10-31 35.5824 34.6917 2012-10-15 35.7194 34.9774 2011-09-25 35.5692 35.0655 2011-09-24 35.5688 35.0795 2013-09-02 40.6506 39.6582 2012-09-16 40.6611 39.7232 2012-09-19 40.6610 39.7405 2011-09-24 40.6726 39.7937 2011-10-31 40.6714 40.2388 2012-12-11 39.0136 38.1198 2011-10-31 38.8984 37.9956 2011-09-24 38.8883 38.3445 2013-09-03 46.2370 45.2423 2013-07-17 46.2395 45.6431 2012-09-17 46.2640 45.2744 2012-12-08 46.2569 46.2518 2013-05-05 9.3817 9.0079 2013-05-08 9.3875 9.0602 2012-12-08 9.1080 8.9316 2013-09-02 48.4570 47.7451 2012-12-09 48.5130 48.0673 2013-07-17 35.6259 34.7611 2013-09-03 86.8923 85.8991 2012-09-17 86.6596 85.7407 2015-07-20 87.3397 86.5163 2015-07-19 87.3391 86.5260 20.991±0.106 2013-07-17 86.8608 86.0473 2015-07-17 87.3381 86.5405 21.044±0.079 18.359±0.063 19.827±0.111 20.019±0.127 21.753±0.731 21.928±0.882 18.321±0.158 2013-07-17 28.3441 28.2058 2013-04-16 7.3259 7.5578 2013-04-15 8.6654 8.7576 2013-04-14 8.6676 8.7455 2013-05-11 34.1466 33.1959 2013-07-21 34.3227 34.0991 2014-05-29 14.8053 13.8611 0.7200 1.0843 1.4107 1.4241 0.2571 0.5098 0.5668 0.6834 1.2623 0.6082 0.6200 1.2540 0.2139 1.0304 1.1793 1.2193 5.8384 5.9318 6.1615 0.8554 1.0390 0.8738 0.1105 0.2647 0.3975 0.4068 0.4086 0.4199 2.0419 7.5275 6.5739 6.5840 0.5792 1.6579 1.4804 CAHA2.2 (1) CAHA2.2 CAHA2.2 CAHA2.2 OSN (1) CAHA2.2 CAHA2.2 CAHA2.2 CAHA2.2 (1) CAHA2.2 CAHA2.2 (1) CAHA2.2 OSN CAHA3.5 CAHA2.2 CAHA2.2 OSN OSN CAHA2.2 OSN CAHA2.2 CAHA3.5 OSN CAHA2.2 Live Live CAHA3.5 Live CAHA3.5 OSN OSN OSN CAHA3.5 CAHA3.5 SOAR (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (3) (3) (1) (1) (1) A&A -- ast_ph, Online Material p 25 Table A.1. continued. Object V R Night r (AU) ∆ (AU) α (degress) Telescope Notes 10199 Chariklo 2060 Chiron 2060 Chiron 136108 Haumea 136108 Haumea 136472 Makemake 136472 Makemake 136472 Makemake 5145 Pholus 5145 Pholus 120347 Salacia 120347 Salacia 120347 Salacia 120347 Salacia 88611 Teharonhiawako 42355 Typhon 174567 Varda 174567 Varda 174567 Varda 174567 Varda 18.505±0.056 18.263±0.132 17.429±0.107 17.580±0.078 16.966±0.104 17.241±0.070 21.677±0.075 21.765±0.181 20.482±0.121 20.558±0.231 20.800±0.146 20.795±0.267 22.538±0.517 20.504±0.358 20.387±0.057 20.517±0.290 20.474±0.150 20.116±0.084 18.379±0.035 18.349±0.037 17.843±0.080 17.161±0.108 17.250±0.073 16.421±0.105 16.469±0.091 16.796±0.060 20.911±0.035 21.221±0.190 20.212±0.086 20.713±0.326 20.309±0.076 20.128±0.146 22.609±0.623 19.920±0.365 19.722±0.060 19.948±0.616 20.235±0.132 19.918±0.089 2014-05-22 14.8000 13.8959 2012-10-16 17.3090 16.6359 2012-12-08 17.3610 17.5362 2013-05-06 50.8365 50.0445 2013-05-08 50.8362 50.0582 2013-05-06 52.3120 51.7204 2013-05-07 52.3122 51.7330 2013-05-08 52.3123 51.7449 2013-06-10 25.2536 24.2802 2013-06-11 25.2552 24.2817 2011-09-24 44.2535 43.3414 2011-09-25 44.2537 43.3440 2012-10-15 44.3415 43.5369 2011-10-31 44.2619 43.6166 2012-09-16 45.1115 44.1527 2013-05-11 19.0989 18.3380 2013-06-03 47.3363 46.3863 2013-05-10 47.3444 46.4795 2013-05-05 47.3456 46.5173 2013-04-14 47.3528 46.7400 1.8265 2.4772 3.1833 0.7069 0.7273 0.8969 0.9070 0.9162 0.6679 0.6685 0.5446 0.5505 0.7637 0.9799 0.3812 2.0288 0.4357 0.6414 0.7075 0.9715 OASI (1) CAHA2.2 CAHA2.2 (1,2) (1) (1,2) (1) (1) (1) (1) (3) (1) OSN OSN OSN OSN OSN INT INT CAHA2.2 CAHA2.2 CAHA2.2 CAHA2.2 CAHA2.2 CAHA3.5 CAHA3.5 CAHA3.5 OSN OSN Notes. (1) Average ext. coeff. (2) Average zero points. (3) Never reported before. Table A.2. Absolute Magnitudes Object HV 15760 1992 QB1 15788 1993 SB 15789 1993 SC 1994 EV3 16684 1994 JQ1 15820 1994 TB 19255 1994 VK8 1995 HM5 32929 1995 QY9 24835 1995 SM55 26181 1996 GQ21 1996 RQ20 1996 RR20 19299 1996 SZ4 1996 TK66 15874 1996 TL66 19308 1996 TO66 15875 1996 TP66 7.839 ± 0.097 7.995 ± 0.059 7.393 ± 0.020 8.183 ± 0.247 7.031 ± 0.078 8.017 ± 0.226 7.840 ± 0.923 8.315 ± 0.100 8.136 ± 0.515 4.584 ± 0.178 5.073 ± 0.050 7.201 ± 0.073 6.986 ± 0.128 8.564 ± 0.034 7.031 ± 0.086 5.257 ± 0.100 4.806 ± 0.144 7.461 ± 0.084 β (mag per degree) −0.193 ± 0.132 0.374 ± 0.066 0.050 ± 0.017 −0.803 ± 0.329 0.570 ± 0.125 0.133 ± 0.152 −0.173 ± 0.976 0.037 ± 0.074 −0.108 ± 0.459 0.139 ± 0.198 0.858 ± 0.124 −0.065 ± 0.075 0.391 ± 0.210 0.307 ± 0.054 −0.280 ± 0.115 0.375 ± 0.112 0.150 ± 0.197 0.127 ± 0.072 N ∆m References 3 5 8 3 5 9 3 5 4 8 6 5 3 4 3 5 7 5 -- -- TR00,JL01,Bo01 Da00,TR00,GH01,JL01,De01 0.04 JL98,Da00,Te97,JL01,TR97 -- -- Bo02,Bo01,GH01 Bo02,TR03,GH01 0.34 Da00,RT99,De01,JL01,TR97 0.42 TR00,Do01,RT99 -- RT99,GH01,Ba00 0.60 Da00,RT99,GH01 0.19 TR03,MB03,GH01,De01,Bo01,Do02,TW 0.10 MB03,Bo02,TW -- -- -- -- 0.12 0.33 0.12 RT99,De01,JL01,Sn10,Bo01 Bo02,TR00,JL01 TR00,Da00,JL01,Bo02 TR00,JL01,Do02 JL01,RT99,JL98,Da00,Bo01 JL98,Da00,RT99,GH01,Sh10,JL01,Bo01 RT99,JL01,JL98,Da00,Bo01 A & A -- a s t _ p h , O n l i n e M a t e r i a l p 2 6 Table A.2. continued. Object HV 118228 1996 TQ66 1996 TS66 33001 1997 CU29 1997 QH4 24952 1997 QJ4 91133 1998 HK151 385194 1998 KG62 26308 1998 SM165 35671 1998 SN165 1998 UR43 33340 1998 VG44 1999 CD158 26375 1999 DE9 1999 HS11 40314 1999 KR16 44594 1999 OX3 86047 1999 OY3 86177 1999 RY215 8.006 ± 0.422 6.535 ± 0.167 6.808 ± 0.057 7.216 ± 0.143 7.754 ± 0.113 7.340 ± 0.056 7.647 ± 0.194 5.938 ± 0.363 5.879 ± 0.109 9.047 ± 0.108 6.599 ± 0.205 5.289 ± 0.092 5.120 ± 0.024 6.843 ± 0.555 6.316 ± 0.139 7.980 ± 0.092 6.579 ± 0.044 7.097 ± 0.084 β (mag per degree) −0.415 ± 0.680 0.083 ± 0.220 0.075 ± 0.087 0.451 ± 0.142 0.290 ± 0.103 0.127 ± 0.088 −0.748 ± 0.205 0.446 ± 0.376 −0.031 ± 0.115 −0.764 ± 0.165 0.228 ± 0.158 0.092 ± 0.119 0.183 ± 0.032 0.233 ± 0.728 −0.126 ± 0.180 −0.086 ± 0.057 0.067 ± 0.042 0.341 ± 0.111 N ∆m References 4 4 4 4 5 5 3 3 6 3 3 3 0.22 0.16 -- -- -- RT99,JL01,GH01,Da00 JL01,RT99,JL98,Da00 Do01,TR00,Ba00,JL01 TR00,JL01,Bo02,De01 De01,GH01,Da00,JL01,Bo02 0.15 Bo01,Do01,MB03,Do02 -- Bo02,GH01,Do01 0.56 MB03,TR00,De01 0.16 De01,MB03,Do01,Fo04,GH01,JL01 -- GH01,De01 0.10 Do02,Bo01,Do01 -- Do02,Sn10,De01 36 0.10 Ra07,DM09,MB03,Te03,Do02,JL01,De01 3 4 -- Px04,TR03,Do01 0.18 Bo02,JL01,TW 12 0.11 Do01,Do02,De01,Pe10,TR00,Bo14,Do05,MB03,Px04,Sh10 4 3 -- -- Do02,TR00,Sn10,Bo02 Sn10,Bo02,Do01 A & A -- a s t _ p h , O n l i n e M a t e r i a l p 2 7 Table A.2. continued. Object HV β (mag per degree) 47171 1999 TC36 29981 1999 TD10 47932 2000 GN171 138537 2000 OK67 82075 2000 YW134 82158 2001 FP185 2001 KA77 2001 KD77 42301 2001 UR163 55565 2002 AW197 2002 GP32 95626 2002 GZ32 119951 2002 KX14 250112 2002 KY14 73480 2002 PN34 55636 2002 TX300 55637 2002 UX25 55638 2002 VE95 5.395 ± 0.030 9.105 ± 0.430 6.776 ± 0.243 6.629 ± 0.694 4.378 ± 0.687 6.420 ± 0.062 5.646 ± 0.090 6.299 ± 0.099 4.529 ± 0.063 3.593 ± 0.023 7.133 ± 0.027 7.419 ± 0.126 4.978 ± 0.017 11.808 ± 0.763 8.618 ± 0.054 3.574 ± 0.055 3.883 ± 0.048 5.813 ± 0.037 0.111 ± 0.027 0.033 ± 0.122 −0.101 ± 0.186 0.089 ± 0.518 0.377 ± 0.552 0.123 ± 0.052 0.130 ± 0.095 0.141 ± 0.082 0.364 ± 0.117 0.206 ± 0.029 −0.135 ± 0.036 0.043 ± 0.064 0.114 ± 0.031 −0.274 ± 0.193 0.090 ± 0.027 0.005 ± 0.044 0.159 ± 0.056 0.089 ± 0.024 N 45 21 29 3 3 5 3 3 3 39 4 29 20 4 57 37 42 43 ∆m References 0.07 Ra07,Te03,MB03,De01,Do03,DM09,Do01,Bo01,TW 0.65 MB03,Ra07,De01,TR03,Do02 0.61 Ra07,MB03,Bo02,DM09,TW -- Do02,De01 0.10 0.06 SS09,Do05,TW Te03,Px04,Do05,TW -- Do05,Px04,Do02 0.07 Px04,Do02,TW 0.08 Do05,SS09,Pe10 0.04 Ra07,DM09,Fo04,TW 0.03 Do05,TW 0.15 Ra07,Fo04,Do05,Te03,TW -- Ra07,Ro10,Bo14,DM09,TW 0.13 0.18 0.09 0.21 0.08 Bo14,Pe10,TW Ra07,Pe10,Te03 Ra07,Te03,Or04,Do05 Ra07,SS09,DM09,TW Pe10,Ra07,TW A & A -- a s t _ p h , O n l i n e M a t e r i a l p 2 8 Table A.2. continued. Object HV β (mag per degree) N ∆m References 127546 2002 XU93 208996 2003 AZ84 120061 2003 CO1 133067 2003 FB128 2003 FE128 120132 2003 FY128 385437 2003 GH55 120178 2003 OP32 143707 2003 UY117 2003 UZ117 2003 UZ413 136204 2003 WL7 120216 2004 EW95 90568 2004 GV9 307982 2004 PG115 120348 2004 TY364 144897 2004 UX10 230965 2004 XA192 7.031 ± 0.859 3.779 ± 0.114 9.146 ± 0.056 6.922 ± 0.566 7.381 ± 0.256 4.632 ± 0.187 7.319 ± 0.247 4.067 ± 0.318 5.830 ± 1.299 5.185 ± 0.054 4.361 ± 0.068 8.897 ± 0.149 6.579 ± 0.021 3.409 ± 0.357 4.874 ± 0.064 4.519 ± 0.137 4.825 ± 0.097 5.059 ± 0.085 0.498 ± 0.320 0.074 ± 0.118 0.092 ± 0.015 0.422 ± 0.469 −0.349 ± 0.209 0.535 ± 0.145 −0.880 ± 0.251 0.045 ± 0.280 −0.230 ± 1.329 0.214 ± 0.073 0.144 ± 0.096 0.089 ± 0.049 0.071 ± 0.024 1.353 ± 0.542 0.505 ± 0.051 0.146 ± 0.103 0.061 ± 0.103 −0.175 ± 0.070 5 5 5 3 5 7 3 -- Sh10,TW 0.14 DM09,Pe10,Fo04,SS09,Bo14 0.07 Pe10,Pe13,Te03 -- -- Pe13,Bo14 Pe13,Bo14 0.15 Pe10,Sh10,DM09,Bo14,TW -- Pe13,Bo14 11 0.26 Pe10,Bo14,Pe13,TW 3 3 3 4 4 3 8 32 8 5 -- -- -- TW Pe10,Bo14,DM09 Pe10 0.05 Pe13,TW -- Bo14,Pe13,TW 0.16 Bo14,DM09,TW -- Pe13,Bo14,TW 0.22 0.08 0.07 Ra07,Pe10 Ro10,Pe10,TW TW A & A -- a s t _ p h , O n l i n e M a t e r i a l p 2 9 Table A.2. continued. Object HV β (mag per degree) 303775 2005 QU182 145451 2005 RM43 145452 2005 RN43 145453 2005 RR43 145480 2005 TB190 145486 2005 UJ438 2007 OC10 225088 2007 OR10 281371 2008 FC76 342842 2008 YB3 55576 Amycus 8405 Asbolus 54598 Bienor 66652 Borasisi 65489 Ceto 19521 Chaos 10199 Chariklo 2060 Chiron 3.853 ± 0.028 4.704 ± 0.081 3.882 ± 0.036 4.252 ± 0.067 4.676 ± 0.084 14.602 ± 0.617 5.330 ± 0.825 2.316 ± 0.124 9.486 ± 0.078 11.024 ± 0.696 8.213 ± 0.621 9.138 ± 0.130 7.656 ± 0.443 6.032 ± 0.040 6.573 ± 0.126 4.987 ± 0.065 6.870 ± 0.055 6.399 ± 0.019 0.277 ± 0.034 −0.028 ± 0.064 0.138 ± 0.030 −0.003 ± 0.065 0.052 ± 0.106 −0.412 ± 0.098 0.223 ± 0.740 0.257 ± 0.505 0.101 ± 0.016 −0.104 ± 0.095 0.052 ± 0.351 0.042 ± 0.029 0.130 ± 0.170 0.231 ± 0.062 0.196 ± 0.096 0.102 ± 0.070 0.064 ± 0.016 0.083 ± 0.005 N 5 6 ∆m References -- Pe13,Bo14,TW 0.04 Bo14,DM09,TW 10 0.04 DM09,Pe13,TW 5 7 5 3 7 4 3 3 43 57 3 8 6 22 37 0.06 0.12 0.13 Pe10,DM09,TW Bo14,Pe13,TW Bo14,Pe13,TW -- -- -- 0.20 0.16 0.55 0.75 Pe13,TW Bo14,TW Pe10,Px12,Pe13 Pa13,Sh10,TW Px04,Fo04,Pe10 Ro97,BL97,Ra07,RM02 Ra07,DM09,Do02,Te03,De01 0.05 Do01,MB03 0.13 Bo14,Te03,Pe13,TW 0.10 De01,TR00,Do02,Bo01,Da00,Ba00 0.10 0.09 Pe01,Be10,DM09,Fo14,RT99,MB99,JL01,Pe10,TW Be10,Du02,Je02,TW A & A -- a s t _ p h , O n l i n e M a t e r i a l p 3 0 Table A.2. continued. Object HV β (mag per degree) 83982 Crantor 0.110 ± 0.149 0.184 ± 0.029 0.078 ± 0.031 0.119 ± 0.056 0.173 ± 0.026 0.194 ± 0.031 0.055 ± 0.057 0.202 ± 0.015 −0.023 ± 0.017 0.160 ± 0.022 −0.074 ± 0.026 0.153 ± 0.156 0.117 ± 0.221 0.132 ± 0.028 0.266 ± 0.008 0.095 ± 0.209 0.061 ± 0.034 0.128 ± 0.013 N 5 4 6 79 98 40 5 55 6 30 3 15 45 9 170 6 67 22 ∆m References 0.34 Px04,DM09,Fo04,Te03 -- De01,Te03,Do02,Bo01 0.24 TR03,De01,Do02,Pe01 0.10 DM09,Ra07,Ca06, 0.10 0.05 Fe01,SR02,TR03,Bo02,MB03,Do01,JL01 Ra07,DM09,Do02 -- Ba00,GH01,JL01,Bo01 0.03 Ra07,TW 0.07 Do03,De01,Pe10,TR03,Bo14,Do01 0.04 Ra07,Pe10,dB05 -- Do02,Bo02,TR00 0.60 Mu92,BB92,Be10,Ro97,TW 0.30 0.03 0.02 Ra07,Te03,DM09,Fo04 Sn10,Bo14,Pe13,TW Ra07,Ba05,Sh10,Pe10 0.22 Da00,Bo01,Ba00,JL01,RT99 0.34 0.07 Ra07,Te03,FD03,DM09,Ba02 Ra07,Te03,DM09,Pe10,Px04,TW A & A -- a s t _ p h , O n l i n e M a t e r i a l p 3 1 136472 Makemake 28978 Ixion 58534 Logos 52975 Cyllarus 52872 Okyrhoe 9.096 ± 0.405 9.064 ± 0.041 10.592 ± 0.171 31824 Elatus 136199 Eris −1.124 ± 0.025 38628 Huya 4.975 ± 0.037 3.774 ± 0.021 7.411 ± 0.041 0.009 ± 0.012 11.441 ± 0.062 2.280 ± 0.021 10.911 ± 0.069 7.474 ± 0.309 2.777 ± 0.250 4.151 ± 0.030 1.669 ± 0.004 79360 Sila-Nunam 5.573 ± 0.224 9.454 ± 0.137 7.670 ± 0.026 120347 Salacia 90482 Orcus 49036 Pelion 5145 Pholus 32532 Thereus 42355 Typhon 50000 Quaoar 90377 Sedna Table A.2. continued. Object HV 174567 Varda 20000 Varuna 3.988 ± 0.048 3.966 ± 0.233 β (mag per degree) −0.455 ± 0.071 0.104 ± 0.246 N 9 30 ∆m References 0.06 0.50 Bo14,Pe13,Pe10,TW Ra07,Do02,Pe10,JS02,TR03 Notes. BB92 = Buie and Bus (1992), Mu92 = Mueller et al. (1992), BL97 = Brown and Luu (1997), TR97 = Tegler and Romanishin (1997), Te97 = Tegler et al. (1997), Ro97 = Romanishin et al. (1997), JL98 = Jewitt and Luu (1998), MB99 = McBride et al. (1999), RT99 = Romanishin and Tegler (1999), Ba00 = Barucci et al. (2000), Da00 = Davies et al. (2000), TR00 = Tegler and Romanishin (2000), Bo01 = Boehnhardt et al. (2001), De01 = Delsanti et al. (2001), Do01 = Doressoundiram et al. (2001), Fe01 = Ferrin et al. (2001), GH01 = Gil-Hutton and Licandro (2001), JL01 = Jewitt and Luu (2001), Pe01 = Peixinho et al. (2001), Ba02 = Barucci et al. (2002), Bo02 = Boehnhardt et al. (2002), Do02 = Doressoundiram et al. (2002), Du02 = Duffard et al. (2002), Je02 = Jewitt (2002). JS02 = Jewitt and Sheppard (2002), RM02 = Romon-Martin et al. (2002), SR02 = Schaefer and Rabinowitz (2002), FD03 = Farnham and Davies (2003), MB03 = McBride et al. (2003), Do03 = Dotto et al. (2003), TR03 = Tegler and Romanishin (2003), Te03 = Tegler et al. (2003), Fo04 = Fornasier et al. (2004), Or04 = Ortiz et al. (2004), Px04 = Peixinho et al. (2004), Ba05 = Barucci et al. (2005), dB05 = de Bergh et al. (2005), Do05 = Doressoundiram et al. (2005), Ca06 = Carraro et al. (2006), Ra07 = Rabinowitz et al. (2007), DM09 = DeMeo et al. (2009), SS09 = Santos-Sanz et al. (2009), Be10 = Belskaya et al. (2010), Pe10 = Perna et al. (2010), Ro10 = Romanishin et al. (2010), Sh10 = Sheppard (2010), Sn10 = Snodgrass et al. (2010), Px12 = Peixinho et al. (2012), Pe13 = Perna et al. (2013), PA13 = Pinilla-Alonso et al. (2013), Bo14 = Böhnhardt et al. (2014), Fo14 = Fornasier et al. (2014), TW = This work. A & A -- a s t _ p h , O n l i n e M a t e r i a l p 3 2 Table A.3. Comparison between the values of HV and β from this work with those from three selected references Ra07 HV 5.103 ± 0.029 5.272 ± 0.055 8.793 ± 0.029 6.368 ± 0.034 3.568 ± 0.030 7.389 ± 0.059 4.862 ± 0.038 8.660 ± 0.017 3.365 ± 0.044 3.873 ± 0.020 5.748 ± 0.058 This work Object HV β (mag per degree) 1999 DE9 1999 TC36 1999 TD10 2000 GN171 2002 AW197 2002 GZ32 2002 KX14 2002 KY14 2002 PN34 2002 TX300 2002 UX25 2002 VE95 2003 AZ84 2003 CO1 2003 FB128 2003 FE128 2003 FY128 5.120 ± 0.024 5.395 ± 0.030 9.105 ± 0.430 6.776 ± 0.243 3.593 ± 0.023 7.419 ± 0.126 4.978 ± 0.017 11.808 ± 0.763 8.618 ± 0.054 3.574 ± 0.055 3.883 ± 0.048 5.813 ± 0.037 3.779 ± 0.114 9.146 ± 0.056 6.922 ± 0.566 7.381 ± 0.256 4.632 ± 0.187 0.183 ± 0.032 0.111 ± 0.027 0.033 ± 0.122 −0.101 ± 0.186 0.206 ± 0.029 0.043 ± 0.064 0.114 ± 0.031 −0.274 ± 0.193 0.090 ± 0.027 0.005 ± 0.044 0.159 ± 0.056 0.089 ± 0.024 0.074 ± 0.118 0.092 ± 0.015 0.422 ± 0.469 −0.349 ± 0.209 0.535 ± 0.145 β (mag per degree) Pe13 HV β (mag per degree) Bo13 HV 0.209 ± 0.035 0.131 ± 0.049 0.150 ± 0.014 0.143 ± 0.030 0.128 ± 0.040 -0.025 ± 0.041 0.159 ± 0.044 0.043 ± 0.005 0.158 ± 0.053 0.158 ± 0.025 0.121 ± 0.039 A & A -- a s t _ p h , O n l i n e M a t e r i a l p 3 3 5.07 ± 0.03 10.50 ± 0.08 3.54 ± 0.03 7.26 ± 0.05 6.94 ± 0.07 5.36 ± 0.08 0.09 ± 0.01 9.07 ± 0.05 7.09 ± 0.20 6.74 ± 0.18 Table A.3. continued. This work Object HV 2003 GH55 2003 OP32 2003 UZ117 2003 WL7 2004 EW95 2004 GV9 2004 PG115 2004 TY364 2005 QU182 2005 RM43 2005 RN43 2005 TB190 2005 UJ438 2007 OC10 2007 OR10 2008 FC76 Asbolus 7.319 ± 0.247 4.067 ± 0.318 5.185 ± 0.054 8.897 ± 0.149 6.579 ± 0.021 3.409 ± 0.357 4.874 ± 0.064 4.519 ± 0.137 3.853 ± 0.028 4.704 ± 0.081 3.882 ± 0.036 4.676 ± 0.084 14.602 ± 0.617 5.330 ± 0.825 2.316 ± 0.124 9.486 ± 0.078 9.138 ± 0.130 β (mag per degree) −0.880 ± 0.251 0.045 ± 0.280 0.214 ± 0.073 0.089 ± 0.049 0.071 ± 0.024 1.353 ± 0.542 0.505 ± 0.051 0.146 ± 0.103 0.277 ± 0.034 −0.028 ± 0.064 0.138 ± 0.030 0.052 ± 0.106 −0.412 ± 0.098 0.223 ± 0.740 0.257 ± 0.505 0.101 ± 0.016 0.042 ± 0.029 Ra07 HV β (mag per degree) 4.434 ± 0.074 0.184 ± 0.070 β (mag per degree) 0.05 ± 0.08 0.18 ± 0.03 Pe13 HV 6.32 ± 0.13 3.99 ± 0.11 8.75 ± 0.16 6.39 ± 0.15 5.23 ± 0.15 3.82 ± 0.12 3.72 ± 0.05 4.62 ± 0.15 11.14 ± 0.32 5.36 ± 0.13 9.107 ± 0.016 0.050 ± 0.004 9.43 ± 0.13 0.09 ± 0.02 Bo13 HV 6.18 ± 0.04 3.79 ± 0.08 5.27 ± 0.02 6.52 ± 0.01 4.03 ± 0.03 5.53 ± 0.05 3.99 ± 0.02 4.52 ± 0.01 4.56 ± 0.02 2.34 ± 0.01 A & A -- a s t _ p h , O n l i n e M a t e r i a l p 3 4 Table A.3. continued. This work Object HV β (mag per degree) Ra07 HV β (mag per degree) Pe13 HV β (mag per degree) Bo13 HV Bienor Orcus Ixion 7.656 ± 0.443 Ceto 6.573 ± 0.126 Eris −1.124 ± 0.025 Huya 5.015 ± 0.021 3.774 ± 0.021 0.009 ± 0.012 11.441 ± 0.062 2.280 ± 0.021 2.777 ± 0.250 4.151 ± 0.030 1.443 ± 0.003 9.454 ± 0.137 7.670 ± 0.026 3.988 ± 0.048 3.966 ± 0.233 Varda Quaoar Salacia Sedna Makemake Okyrhoe Thereus Typhon Varuna 0.130 ± 0.170 0.196 ± 0.096 0.119 ± 0.056 0.124 ± 0.015 0.194 ± 0.031 0.202 ± 0.015 −0.023 ± 0.017 0.160 ± 0.022 0.117 ± 0.221 0.132 ± 0.028 0.200 ± 0.006 0.061 ± 0.034 0.128 ± 0.013 −0.455 ± 0.071 0.104 ± 0.246 7.588 ± 0.035 0.095 ± 0.016 -1.116 ± 0.009 5.048 ± 0.005 3.766 ± 0.042 0.091 ± 0.015 0.105 ± 0.020 0.155 ± 0.041 0.133 ± 0.043 0.054 ± 0.019 2.328 ± 0.028 2.729 ± 0.025 0.114 ± 0.030 0.159 ± 0.027 1.829 ± 0.048 9.417 ± 0.014 7.676 ± 0.037 0.072 ± 0.004 0.126 ± 0.022 3.760 ± 0.032 0.278 ± 0.047 6.58 ± 0.10 0.10 ± 0.06 6.60 ± 0.01 10.83 ± 0.01 4.26 ± 0.06 0.04 ± 0.06 4.01 ± 0.02 3.51 ± 0.06 3.93 ± 0.07 A & A -- a s t _ p h , O n l i n e M a t e r i a l p 3 5
1704.05881	1	1704	2017-04-19T18:20:16	The long-term evolution of known resonant trans-Neptunian objects	[ "astro-ph.EP" ]	Aims. Numerous trans-Neptunian objects are known to be in mean-motion resonance with Neptune. We aim to describe their long-term orbital evolution (both past and future) by means of a one-degree-of-freedom secular model. In this paper, we focus only on objects with a semi-major axis larger than 50 astronomical units (au). Methods. For each resonant object considered, a 500 000-year numerical integration is performed. The output is digitally filtered to get the parameters of the resonant secular model. Their long-term (Giga-year) orbital evolution is then represented by the level curves of the secular Hamiltonian. Results. For the majority of objects considered, the mean-motion resonance has little impact on the long-term trajectories (the secular dynamics is similar to a non-resonant one). However, a subset of objects is strongly affected by the resonance, producing moderately-high-amplitude oscillations of the perihelion distance and/or libration of the argument of perihelion around a fixed centre. Moreover, the high perihelion distance of the object 2015 FJ345 is plainly explained by long-term resonant dynamics, allowing us to also deduce its orbital elements at the time of capture in resonance (at least 15 million years ago). The same type of past evolution is expected for 2014 FZ71.	astro-ph.EP	astro-ph	Astronomy & Astrophysics manuscript no. AeA_2017 November 13, 2018 c(cid:13)ESO 2018 The long-term evolution of known resonant trans-Neptunian objects M. Saillenfest1, 2 and G. Lari2 1 IMCCE, Observatoire de Paris, 77 av. Denfert-Rochereau, 75014 Paris, France e-mail: [email protected] 2 Dipartimento di Matematica, Università di Pisa, Largo Bruno Pontecorvo 5, 56127 Pisa, Italia e-mail: [email protected] 7 1 0 2 r p A 9 1 . ] P E h p - o r t s a [ 1 v 1 8 8 5 0 . 4 0 7 1 : v i X r a Received January 30, 2017; accepted April 10, 2017 ABSTRACT Aims. Numerous trans-Neptunian objects are known to be in mean-motion resonance with Neptune. We aim to describe their long- term orbital evolution (both past and future) by means of a one-degree-of-freedom secular model. In this paper, we focus only on objects with a semi-major axis larger than 50 astronomical units (au). Methods. For each resonant object considered, a 500 000-year numerical integration is performed. The output is digitally filtered to get the parameters of the resonant secular model. Their long-term (Giga-year) orbital evolution is then represented by the level curves of the secular Hamiltonian. Results. For the majority of objects considered, the mean-motion resonance has little impact on the long-term trajectories (the secular dynamics is similar to a non-resonant one). However, a subset of objects is strongly affected by the resonance, producing moderately- high-amplitude oscillations of the perihelion distance and/or libration of the argument of perihelion around a fixed centre. Moreover, the high perihelion distance of the object 2015 FJ345 is plainly explained by long-term resonant dynamics, allowing us to also deduce its orbital elements at the time of capture in resonance (at least 15 million years ago). The same type of past evolution is expected for 2014 FZ71. Key words. trans-Neptunian object -- mean-motion resonance -- secular model 1. Introduction The orbital distribution of the trans-Neptunian objects is still an open problem, leading to numerous conjectures about the origin and evolution of the external Solar System. In this context, mean- motion resonances with Neptune are known to have played an important role in sculpting this population from its initial dis- tribution (see for instance Malhotra 1993; Duncan et al. 1995; Gomes et al. 2005). According to numerical simulations, tem- porary and chaotic resonance captures are the most frequent, but some trappings are sustainable on a Giga-year timescale, in particular when the resonant link leads the small body towards a dynamical region where the semi-major axis cannot diffuse anymore. This happens for instance if the perihelion distance increases during the resonant motion, because this reduces the width of every neighbouring resonance, preventing the object from undergoing chaotic diffusion driven by resonance overlap. The resulting trajectory is quasi-integrable and it can be mod- elled by a secular-like theory. Recently, Saillenfest et al. (2016) presented a semi-analytical model designed to accurately describe, and in a plain way, the long-term dynamics of such objects. This model can be seen as a generalisation of the secular theory by Kozai (1985) and its ex- tension by Gallardo et al. (2012). In order to reduce the system to a single degree of freedom, the key element of the resolution consists in the large separation of timescales between the oscilla- tion of the resonant angle and the precession of the argument of perihelion (adiabatic approximation). Up to now, the application of this model has been made exclusively on distant small bodies (a > 100 au), for which the adiabatic approximation is more rel- evant. However, only a few trans-Neptunian objects are known with semi-major axes that large, and their orbital uncertainties are necessarily rather high since only a fraction of their orbits has been observed (orbital period > 1000 years). Consequently, the application of a secular model to these objects can only give the long-term dynamics these objects would have if they were in mean-motion resonance with Neptune (Saillenfest et al. 2017). In this paper, our goal is to study the long-term evolution of the trans-Neptunian objects known to be in resonance, that is, those for which the uncertainties of the orbital elements are pre- dominantly compatible with a resonant trapping. As presented for instance by Lykawka & Mukai (2007), these objects are quite numerous and most of them have small semi-major axes (because of the observational constraints). We show here that, fortunately, the adiabatic approximation is also viable for semi- major axes as small as 50 au, even if the two timescales are not as strikingly separated as for a > 100 au. For even closer ob- jects, individual analyses would be required, not only to check the validity of the adiabatic approximation, but also the possible occurrence of secular resonances with the giant planets (Kneze- vic et al. 1991), which would also invalidate this model. Conse- quently, this article focusses on the ∼ 40 currently known reso- nant trans-Neptunian objects with a > 50 au. We selected the known resonant trans-Neptunian objects from three reference classifications (Lykawka & Mukai 2007; Gladman et al. 2008, 2012), retaining only the "securely clas- sified" ones with a semi-major axis larger than 50 au. Two re- cent observation reports were also added: Bannister et al. (2016) Article number, page 1 of 9 A&A proofs: manuscript no. AeA_2017 present the discovery of 2015 RR245, which has a confirmed res- onant classification; on the other hand, Sheppard et al. (2016) introduce some objects that are probably resonant or have been strongly affected by a mean-motion resonance with Neptune. This last classification is slightly different, but we chose to add these objects to this study to emphasise the link with recent re- sults focussed on high-perihelion objects. Indeed, the secular model used in this paper proved to be an efficient tool to re- veal dynamical paths leading to high perihelion distances and large inclinations. The objects studied in this paper are listed in Tab. 1. In Sect. 2, we recall briefly the resonant secular model used (Sect. 2.1) and we explain how suitable secular parameters can be obtained from known objects (Sect. 2.2). Then, Sect. 3 sum- marises our results using a selection of the most representative phase portraits obtained. Most of the objects considered have quite ordinary secular dynamics (Sect. 3.1), but a subset of them evolve near or inside secular libration islands (Sect. 3.2) and one object follows a regular-by-part secular trajectory (Sect. 3.3). 2. Model and method 2.1. The resonant secular model In this section, we outline the semi-analytical model introduced by Saillenfest et al. (2016). We consider a massless particle af- fected by the Sun and N planets, which are supposed on circular and coplanar orbits (typically Jupiter, Saturn, Uranus and Nep- tune in their invariable plane). The dynamics of the particle are studied in the exterior mean-motion resonance kp : k with the planet p. We introduce the resonant angle: σ = k λ − kp λp − (k − kp) (1) with k, kp ∈ N and k > kp. In this expression, λ and are the mean longitude and the longitude of perihelion of the small body, and λp is the mean longitude of the planet p. From the Hamilto- nian in osculating coordinates, we get the "semi-secular" dynam- ical system by a perturbative method to first order. The resulting Hamiltonian can be written as: K(cid:0)Σ, U, V, σ, u(cid:1) = − µ2 2 (kΣ)2 − npkpΣ + εK1 (cid:0)Σ, U, V, σ, u(cid:1) (2) where µ is the gravitational parameter of the Sun and np is the mean motion of the planet p. In that expression, εK1 is the nu- merically computed average of the perturbative part of the oscu- lating Hamiltonian over the short-period angles. The canonical coordinates used can be written in terms of the usual heliocen- tric Keplerian elements (a, e, I, ω, Ω, M) of the particle:  √ √ √ Σ = U = V = (cid:16)√ (cid:16)√ (cid:17) 1 − e2 − kp/k 1 − e2 cos I − kp/k µa/k µa µa (cid:17) (3) along with their conjugated angles σ and (u, v) = (ω, Ω). Since the Hamiltonian (2) does not depend on v, the quantity V is a semi-secular constant of motion. The pairs of coordinates (Σ, σ) and (U, u) evolve generally on very different timescales, so we can use the adiabatic approximation to reduce the system to a single degree of freedom. The "secular" Hamiltonian function writes finally: F(cid:0)J, U, V, u(cid:1) = K(Σ0, U, V, σ0, u) (4) Article number, page 2 of 9 where (Σ0, σ0) represents any point of the level curve of K for (U, u) fixed which encloses the signed area 2πJ. The momentum J is a secular constant of motion, related to the oscillation am- plitude of (Σ, σ). It is negative if the particle evolves inside the resonance island. Since the secular system has only one degree of freedom, the dynamics can be described by the level curves of (4) in the plane (U, u) with V and J as parameters. The interpretation of these level curves is helped by the choice of a "reference semi-major axis" a0 for the resonance considered. Indeed, the constant V can be replaced by: √ √ 1 − e2 cos I µa0 + kp/k = η0 = V/ (5) where e and I are called the "reference" eccentricity and incli- nation. In simple words, these variables are equal to the secular elements of the particle when its semi-major axis is equal to a0. Then, the variable U is equivalent to the reference perihelion distance q = a0(1 − e). The level curves of (4) can be plotted in the plane (q, ω) which is much more self-explanatory than its canonical counterpart (U, u). In the following, this model is applied considering the four giant planets of the Solar System, the mass of the inner ones being added to the Sun. The secular parameters (η0, J) of the objects studied in this paper are computed using the method de- scribed in the following section. They are gathered in Tab. 1, along with the constants a0 chosen. The current positions (ω, q) of the objects are also given. 2.2. Determination of the secular parameters In the previous section, we described the changes of coordinates leading to the resonant secular model. The problem is to com- pute this set of coordinates for a given object, for which only the osculating coordinates are known. The standard procedure consists in digitally filtering the output of a medium-term nu- merical integration, thus removing the short-period component of the trajectory as we did in the semi-analytical theory. The de- tailed procedure was introduced by Carpino et al. (1987) and implemented in the OrbFit Software Package1. For all the ob- jects studied in this article, a 500 000-year numerical integration was found more than enough to cover the semi-secular oscilla- tions (related to the resonant angle). The numerical integrations were made using the same software package, taking the nominal orbits given by AstDyS database2 as initial conditions. The four giant planets were integrated consistently and the masses of the inner ones were added to the Sun. The invariable plane of the Solar System was taken as reference plane for the output. We used heliocentric coordinates, as this is the system on which is based the semi-analytical theory3. The secular values of (U, V, u) can be directly picked up from the filtered series, whereas the parameter J requires the compu- tation of the area enclosed by the trajectory in the (Σ, σ) plane. Fig. 1 shows some examples of filtered output, with the graph- ical representation of the corresponding areas 2πJ. In the back- ground, the level curves of (2) with (U, u) fixed at their secular (that is filtered) current values are also plotted in order to as- sess the validity of the semi-analytical procedure. The filtered trajectories were always found to follow pretty well the level curves, showing that our simplified and averaged model captures 1 http://adams.dm.unipi.it/orbfit/ 2 http://hamilton.dm.unipi.it/astdys/ 3 Since the temporal series obtained are meant to be filtered, the short- period oscillations of the barycentre of the Solar System, usually re- moved by using barycentric coordinates, are not a problem. M. Saillenfest and G. Lari: The long-term evolution of known resonant trans-Neptunian objects 3. General results 3.1. Typical resonant secular evolutions Most of the objects studied in this paper are not very affected by their resonant relation with Neptune: this results in a "flat" sec- ular evolution, with circulating ω and almost constant q. Such a secular behaviour is similar to what we would obtain for a non-resonant dynamics (where the only notable features are lo- cated at high inclinations). Some examples of level curves of the secular Hamiltonian (4) are shown in Fig. 2. The graphs are π- periodic in ω. The vast majority of objects studied in this paper follow these typical secular trajectories, with the exception of the objects 135571, 2004 KZ18, 2008 ST291, 82075, 2015 FJ345 and 2014 FZ71, which we describe in Sects. 3.2 and 3.3. In or- der to check the validity of the adiabatic approximation, each graph also features a numerical integration of the two-degree-of- freedom semi-secular system. As required, the mean trajectory follows a level curve of the secular Hamiltonian. The extra os- cillations due to the second degree of freedom are even almost undistinguishable, hidden in the curve width. Fig. 2a presents the most common case: the parameters of the secular model do not allow any peculiar geometry, and the specific level curve fol- lowed by the particle has nothing special either. In Fig. 2b, the particle is pushed outside of the resonance island in some por- tions of its trajectory. This does not affect the overall secular dynamics though, since it re-enters the resonance with a simi- lar parameter J. Finally, Fig. 2c presents a "missed" interesting case: the parameters of the secular model do allow important variations of the orbital elements but the particle is located on a flat level curve, outside of any libration island. 3.2. Objects affected by secular libration islands The object 135571, presented in Fig. 3a, is the only one show- ing a retrograde circulation of ω. Indeed, it follows a level curve located above the two maxima of the secular Hamiltonian (4). It is also very close to the separatrix: a small chaotic diffusion (or a refinement of the orbital solution) could easily put it inside the secular libration island. The object 2004 KZ18, on the other hand, is clearly located in the island (Fig. 3b). This smooth quasi- periodic trajectory may not be "permanent", however, since it in- volves small perihelion distances. Indeed, the neighbouring reso- nances have a large width and a chaotic drift out of the resonance is a serious risk. Figure 3c shows the case of 2008 ST291 which is associated with a much more distant resonance. The resonant angle σ oscil- lates in an asymmetric libration island (Fig. 1c), which results in a secular Hamiltonian with level curves that are also asymmetric. 2008 ST291 is located inside a secular libration island, confining ω around a fixed value and producing large-amplitude oscilla- tions of the perihelion distance. The object is on the descend- ing part of the trajectory, implying that it has completed at least one cycle (period of about 115 Myrs). However, we note that 2008 ST291 has still uncertain orbital elements: Sheppard et al. (2016) reported clones evolving from q ≈ 35 to 60 au. These limits can be roughly measured in Fig. 3c, according to the po- sition of the separatrix with respect to the object. Finally we can also mention the object 82075, presented in Fig. 3d, which does not interact directly with a secular libra- tion island but which is located on a level curve producing rather large variations of orbital elements (circulation of ω and oscilla- tions of q from about 38 to 44 au). Article number, page 3 of 9 Fig. 1. Computation of the secular constant of motion J from the filtered numerical integration. The examples shown are (from top to bottom): 82075, 119068, 2008 ST291 and 136120. See Tab 1 for the correspond- ing resonances. The axes are the resonant angle σ from Eq. 1 (rad), and its conjugated momentum Σ from Eq. 3 (au2rad/yr). The blue crosses come from the filtered output, whereas the red hatched area is equal to the quantity −2πJ used to construct the secular model. The black lines in the background are the level curves of (2) with (U, u) fixed. Various cases are shown: small (a,c) or large (b,d) area, single (a,b) or double (c,d) resonance island, simple (a,b), asymmetric (c) or horseshoe (d) oscillations. the essence of the dynamics. The characteristics of the corre- sponding secular models are gathered in Tab. 1. The oscillation amplitudes of the resonant angles were generally found to be in very good agreement with the values given in Lykawka & Mukai (2007) and Gladman et al. (2012). In our numerical integrations, some objects were found out- side the expected resonances (blank lines in Tab. 1). Apart from two exceptions, this concerns small bodies reported by Shep- pard et al. (2016) which are not classified as "resonant" from their current orbit but from their probable dynamical history. For these objects, a secular model near the resonance could still be developed (in which 2πJ is the area under the curve), or the un- certainties of the nominal orbit could be taken into account to develop a secular model for a "potential" resonant orbit. This will be realised for 2014 FZ71. a)b)c)d)5.975.9755.985.985Σ9.339.359.379.39Σ10.4210.4310.4410.4510.46Σ0π/2π3π/22πσ16.516.5516.616.65ΣA&A proofs: manuscript no. AeA_2017 Fig. 2. Typical examples of phase portraits for the objects studied in this paper. The colour shades represent the value of the secular Hamiltonian (dark/light for low/high), from which can be deduced the direction of motion along the level curves (black lines). The white zones are forbidden by the value of the parameter η0 (it would require a cosine of inclination larger than 1), and the grey zones are forbidden by the value of the parameter J (the resonance island is too narrow to contain the signed area 2πJ). The green cross represents the current position of the object, and the red curve comes from a numerical integration of the two-degree-of-freedom semi-secular system. The names are written above the graphs (see Tab. 1 for the parameters of the models). On the graph b, the secular trajectory of 79978 forces periodically σ to circulate when the curve enters the grey zone (separatrix crossing). The numerical integration shows that this does not significantly affect the overall dynamics (red curve). The graph c shows that 60621 follows a "flat" trajectory even if the parameters of the model could allow interesting variations of orbital elements. The behaviour of the object 2004 EG96 (not shown) is very similar to that of 60621, but it follows a level curve much closer to the secular separatrix (lying just above the red curve). A small diffusion of its orbital elements (or a refinement of its orbital solution) could thus put it on a level curve leading to much higher perihelion distances. 3.3. Regular-by-part secular dynamics The object 2015 FJ345 presents the most interesting secular dy- namics. Figure 4a shows that it is currently located in the single- island region of the 1 : 3 resonance (that is above the "green line"), but on a level curve coming from and leading to discon- tinuous transitions. Hence, its perihelion distance was probably raised from the double-island region (below the green line), start- ing much closer to Neptune, on a secular trajectory leading to the transition. The resonant secular model can be used to recon- struct the detailed scenario: following the secular level curve of 2015 FJ345 backwards (toward the left), its intersection with the grey zone gives its location when it was ejected from the vanish- ing resonance island. At this point, the total area available in this island gives the parameter J to be used in a secular model de- scribing the previous portion of the secular dynamics. Proceed- ing this way, we suppose that on a secular timescale, the sep- aratrix crossings can be considered as instantaneous. A chaotic extra change of J should be expected in the real case (depending on the exact phase of Σ and σ at the moment of the crossing), but it is expected to be negligible if the adiabatic approximation is well verified. The area measured is 2πJ ≈ −0.0115 au2rad2/yr, which is quite small since the separatrix crossing happened near the green line (blue spot on Fig. 4a). The corresponding previ- ous secular trajectory is shown in Fig. 4b. Surprisingly, we find once again a trajectory coming from and leading to a transition, without any passage to smaller perihelion distances, where the resonance capture with Neptune could have occurred. We use the same method to get the secular trajectory before this sec- ond transition (green spot in Fig. 4b). The level curves of the corresponding secular Hamiltonian are presented in Fig. 4c: we finally get a secular trajectory leading to much lower perihelion distances (at least below q = 35 au, depending on the exact level curve followed by the particle), where the capture in resonance with Neptune probably occurred. Since the overall trajectory is rather complicated, Fig. 5 sums up the three components by showing a numerical integration Article number, page 4 of 9 0.0◦11.8◦20.9◦26.0◦29.4◦I(deg)a)2015RR2450.0◦15.0◦20.6◦23.5◦I(deg)b)799780.0◦12.1◦18.8◦22.3◦24.0◦24.7◦I(deg)c)60621π5π/43π/27π/42πω(rad)3035404550q(au)0π/4π/23π/4πω(rad)30354045q(au)0π/4π/23π/4πω(rad)303540455055q(au)M. Saillenfest and G. Lari: The long-term evolution of known resonant trans-Neptunian objects Fig. 3. Phase portraits of the objects affected by secular libration islands. On the graph a, the object 135571 is the only one found to have a retrograde circulation of ω. On the graph b, the secular evolution of 2004 KZ18 inside the mean-motion resonance makes ω oscillate around 0. On the graph c, there are two distinct regions according to the topology of the semi-secular phase-space, since the resonance of 2008 ST291 is of type 1 : k. Below the green line, the resonance island is doubled (as in Fig. 1c,d) resulting in asymmetric level curves. Above the green line, only one resonance island remains (as in Fig. 1a,b) and the level curves are symmetric. If the line is crossed, there can be either a discontinuity (thick green line) if the particle is located in the vanishing island, or a smooth transition (dotted green line) if it is located in the persisting island. of the two-degree-of-freedom semi-secular system. We started from a position (ω, q) on the level curve of Fig. 4c, as a poten- tial position of capture in resonance with Neptune (black spot). The two separatrix crossings were found to be a bit sensitive to the initial phase chosen for σ and Σ, so we tried different values distributed all along the level curve of the semi-secular Hamil- tonian. The grey trajectories in Fig. 5 give an idea of the range of possible behaviours. Roughly half of the integrated trajecto- ries were found to follow qualitatively the scenario from Fig. 4 (the red curve is one example of them). As predicted, the tra- jectory ends up near the observed position of 2015 FJ345, show- ing that this scenario is dynamically possible. This also gives the timescale involved, which counts in Myrs. In the future, 2015 FJ345 is expected to go on with these types of transitions, possibly turning back to small perihelion distances. In partic- ular, it is possible that several loops already occurred between its capture in resonance and its observed position. It could also stay in a resonant high-perihelion state for Gyrs if it has trig- gered a "high-perihelion trapping mechanism" (Saillenfest et al. 2017). The orbital uncertainties, though, prevent from any defini- tive conclusion. As discussed by Sheppard et al. (2016), such a mechanism is probably also responsible for the high perihelion distance of 2014 FZ71 even if this object is not found currently in resonance (according to its nominal orbit). Consequently, it could have been left in its current position at the end of Neptune's migra- tion (Gomes et al. 2005, 2008), or its resonant secular dynamics itself could have pushed it out of the resonance separatrices. We studied this last scenario by a close-to-resonance secular model (Saillenfest et al. 2016), but no interaction at all with the reso- Article number, page 5 of 9 0.0◦12.7◦19.2◦22.6◦24.3◦24.9◦I(deg)a)13557117.5◦24.3◦28.1◦30.4◦31.6◦32.1◦I(deg)b)2004KZ180.0◦18.4◦28.8◦34.1◦37.2◦39.1◦I(deg)c)2008ST2910.0◦8.3◦17.2◦21.2◦23.5◦24.5◦I(deg)d)82075π5π/43π/27π/42πω(rad)303540455055q(au)-π/2−π/40π/4π/2ω(rad)303540455055q(au)π5π/43π/27π/42πω(rad)304050607080q(au)π5π/43π/27π/42πω(rad)303540455055q(au)A&A proofs: manuscript no. AeA_2017 Fig. 4. Level curves of the secular Hamiltonian for the current and past evolution of the object 2015 FJ345. Three smooth trajectories are considered, divided by separatrix crossings. The constant parameter η0 is given in Tab. 1, whereas 2πJ is given above the graphs (au2rad2/yr). The current orbital elements of 2015 FJ345 result in the leftmost phase portrait a. Contrary to previous graphs, the secondary oscillations of the semi-secular trajectory (red curve) are noticeable. The secular trajectory (mean curve), though, follows still pretty well the level curves given by the model. Starting from the current position of 2015 FJ345 (blue cross), the red curve is drawn until it reaches a change of the resonance topology (from a single to a double island), on the green line. Following the level curve toward the left (past evolution), the blue spot represents the position of the previous separatrix crossing undergone by 2015 FJ345. The middle graph b shows the secular model for the past evolution of 2015 FJ345. The blue spot represents its position at the transition to its current state (same as graph a). Following the level curve downwards (past evolution), the green spot shows the position of the previous separatrix crossing undergone by 2015 FJ345. The rightmost graph c presents the secular model for the earlier evolution of 2015 FJ345. The green spot shows its position at the transition to the next part of the trajectory (same as graph b). Following the level curve downwards (past evolution), the black spot is taken as initial condition for a numerical integration (see text and Fig. 5). Fig. 5. Numerical integration of the two-degree-of-freedom semi-secular system. The initial conditions (origin of the time and black spot on the right) are taken according to the level curves plotted in Fig. 4c. On the right, the evolution of the couple (U, u) is drawn (with q instead of U, obtained by inverting Eq. 3). Several initial phases for (σ, Σ) were tried, leading to secular trajectories diverging at the first separatrix crossing (grey curves). Among them, the red trajectory is in agreement with our backward reconstruction: its second transition occurs on the blue spot, and it passes near the current position of 2015 FJ345 (to be compared with the level curves from Fig. 4). On the left, the evolution of the couple (σ, Σ) is represented for the red trajectory (with a instead of Σ). The two separatrix crossings, leading to a change of oscillation island, are easily noticeable. The bottom graph presents the evolution of the area 2πJ for all the trajectories plotted on the right, showing the divergence at the first transition. Even if 2πJ is plotted as a continuous line, remember that its definition changes after each transition (different resonance island). nance was detected (flat level curves). The migration of Neptune set apart, 2014 FZ71 is thus probably much closer to the reso- nance than its current best-fit orbit seems to indicate. Actually, the uncertainties are even compatible with a trapping in reso- nance (as already reported by Sheppard et al. 2016). As shown in Fig. 6, the semi-major axis of 2014 FZ71 can be slightly modified (at the 1-sigma level) to enter the resonance. If we change only the value of a, the minimum area 2πJ in the case of resonance is equal to −0.035 au2rad2/yr (red hatched region in Fig. 6). The corresponding resonant secular model is presented in Fig. 7a: we Article number, page 6 of 9 I(deg)a)leftisland2πJ=−0.065I(deg)b)rightisland2πJ=−0.011518.6◦25.6◦29.8◦32.5◦34.3◦35.4◦35.9◦I(deg)c)leftisland2πJ=−0.0370π/4π/23π/4πω(rad)30354045505560q(au)0π/4π/23π/4πω(rad)0π/4π/23π/4πω(rad)0π/2π3π/22πσ(rad)62.562.662.762.8a(au)-0.1-0.0500510152025302πJ(au2rad2/yr)time(Myrs)3035404550550π/4π/23π/4π5π/43π/27π/42πq(au)ω(rad)M. Saillenfest and G. Lari: The long-term evolution of known resonant trans-Neptunian objects -- Locked in the 2 : 5 resonance, 135571 evolves very close to secular libration islands and shows a retrograde (or possibly oscillatory) evolution of ω. -- In the same resonance, 2004 KZ18 is located near the centre of a libration island at ω = 0 (a perfect secular evolution would produce very small oscillations of ω around 0). -- The object 82075 (3 : 8 resonance) shows oscillations of the perihelion distance from 38 to 44 au (with circulating ω). -- In the more distant 1 : 6 resonance, 2008 ST291 evolves in a wide asymmetric libration island centred at ω ≈ 117o, result- ing in oscillations of q from 39 to 55 au. Fig. 6. Semi-secular phase portrait of 2014 FZ71. On the right axis, the values of a corresponding to Σ are shown. The blue crosses come from the filtered output of the full non-averaged numerical integration (same as Fig. 1), with the current position of 2014 FZ71 on the big spot. For comparison, the 1-sigma error bar of the (osculating) orbital solution is added. The uncertainty is compatible with the red hatched area inside the resonance, from which we get a possible value of the parameter J. The same method was used in Saillenfest et al. (2017), but it was not restricted to the uncertainty ranges. get a geometry very similar to what we obtained for 2015 FJ345, indicating that an analogous long-term resonant evolution prob- ably occurred. Following the level curve toward the past, the sec- ond resonance island has a total area of −0.00205 au2rad2/yr on the transition line, which gives the secular model represented in Fig. 7b. This time, we directly obtain a trajectory leading to much smaller perihelion distances (∼ 38.5 au), compatible with the numerical experiments by Sheppard et al. (2016). The pa- rameter η0 puts a lower limit on the reachable perihelion dis- tance (independently of the separatrix crossings encountered): this limit is equal to about 37 au and corresponds to I = 0. This is quite high for a resonant capture (since a diffusion of a is nec- essary), but that argument was used by Sheppard et al. (2016) to state that 2014 FZ71 possibly originated in the Scattered Disc. 4. Discussion and conclusion A one-degree-of-freedom secular model was used to study the long-term dynamics of the known trans-Neptunian objects in res- onance with Neptune, with semi-major axes larger than 50 au. This method allows us to visualise their long-term trajectories in the plane (ω, q), in which the notable features (equilibrium points, libration islands, separatrices) can be located. Accord- ing to the model, most of them experience virtually no change of perihelion distance, indicating a small influence of the res- onance on their long-term dynamics. Indeed, the resonant link can never bring them away from their capture configurations, re- sulting in unstable transient resonances. This is confirmed for 2015 RR245 by the detailed dynamical study by Bannister et al. (2016). If an object was left on a distant resonant trajectory by the late migration of Neptune, on the contrary, the constancy of its secular perihelion distance would be a guarantee of stability (Gomes et al. 2008). On the other hand, four objects are located near notable fea- tures of the phase portraits: The resonance captures of 135571 and 2004 KZ18 probably oc- curred in the lowest part of their secular trajectories (smallest perihelion distances). However, since their dynamics are quasi- periodic, they will regularly return to their entrance configura- tions, leading to possible expulsions. On the other hand, the min- imum perihelion distances reachable by 82075 and 2008 ST291 seem a bit high to be considered as capture locations. This led Lykawka & Mukai (2007) to classify 82075 as "detached", whereas the orbital uncertainties of 2008 ST291 prevent us from reaching a definitive conclusion. The cases of 2015 FJ345 and 2014 FZ71 are more complicated (1 : 3 and 1 : 4 resonances). Indeed, they follow secular tra- jectories leading to separatrix crossings, which makes necessary the use of piecewise secular models. Following the level curves backwards, possible scenarios can be retraced from their capture into resonance to their current positions. They require one or more resonant transitions (passage from one resonance island to the other) and lead to much smaller perihelion distances. How- ever, the separatrix crossings are actually not instantaneous (their exact outcomes depend on the phase at the time of transition), leading to potential trajectories significantly different than the reconstructed ones. This holds especially for these two objects, which present a substantial departure from the adiabatic approx- imation. Nevertheless, the scenarios presented here are dynami- cally possible (as shown by numerical integrations): they can be considered as the ones producing the shortest paths from capture into resonance to the observed positions. The method gives also a precise idea of what type of trajec- tory can be followed by a given object. This allows us to distin- guish which orbits could have been created by a resonant link with Neptune on its current orbit, and which ones have a more complex history (involving the planetary migration or another √ source of perturbation). The fixed value of V, or equivalently the 1 − e2 cos I used by some authors approximate constancy4 of (Gomes et al. 2008; Sheppard et al. 2016), gives only an upper bound for the variations in perihelion distance (white regions in our figures). Indeed, an object could very well be detached, but with a parameter V allowing in principle any eccentricity and in- clination. Consider for instance, a particle located on the highest level curve drawn on Fig. 4c. We did not try to obtain similar graphs for 2004 XR190 since its location out-of-resonance is securely assessed. Its current po- sition is pretty well explained by an analogous resonant evolu- tion, but occurring during the late migration of Neptune (Gomes 2011). Finally, we would like to point out that for all objects except 2014 FZ71, the secular parameters were obtained from the nomi- nal orbital solutions (through the filtered numerical integration). (cid:16)√ 1 − e2 cos I is constant only in a non- 4 Strictly speaking, the quantity resonant secular model. The analogous constant in the resonant case is 1 − e2 cos I − kp/k V = , but since the particle is trapped in resonance, a is never far from some given value. (cid:17) √ √ µa Article number, page 7 of 9 75.6375.7475.8575.96a(au)0π/2π3π/22πσ(rad)13.6613.6713.6813.69Σ(au2rad/yr)A&A proofs: manuscript no. AeA_2017 Fig. 7. Level curves of the secular Hamiltonian for the current and past evolution of the object 2014 FZ71. Two smooth trajectories are considered, divided by a separatrix crossing. The constant parameter η0 is given in Tab. 1, whereas 2πJ is given above the graphs (au2rad2/yr). The current orbital elements of 2014 FZ71 result in the left phase portrait a. Following the level curve toward the left (past evolution), the blue spot shows the position of the previous separatrix crossing underwent by 2014 FZ71. The right graph b presents the secular model for the past evolution of 2014 FZ71. The blue spot shows its position at the transition to its current state (same as graph a). Malhotra, R. 1993, Nature, 365, 819 Saillenfest, M., Fouchard, M., Tommei, G., & Valsecchi, G. B. 2016, Celestial Mechanics and Dynamical Astronomy, 126, 369 Saillenfest, M., Fouchard, M., Tommei, G., & Valsecchi, G. B. 2017, Celestial Mechanics and Dynamical Astronomy, 127, 477 Sheppard, S. S., Trujillo, C., & Tholen, D. J. 2016, ApJ, 825, L13 Consequently, the actual parameters could be quite different, es- pecially for the recently discovered objects. The corresponding changes concern mostly the parameter J, which is sensitive to the initial conditions. For instance, a slight increase of the area 2πJ leads to a wider coverage of the "grey" zones (on the sides of which separatrix crossings occur). This should not affect our general conclusions, though: for instance, a slight change of or- bital elements would leave 2015 FJ345 in the high-perihelion re- gion and on a trajectory coming from a much lower perihelion distance. If ever it turns out to be currently out of the resonance, the migration of Neptune could also be invoked (same type of trajectory, but in an earlier stage of the Solar System). Acknowledgements. This work was partly funded by Paris Sciences et Lettres (PSL). We thank the anonymous referee for her/his detailed comments which led to a much better version of the article. References Bannister, M. T., Alexandersen, M., Benecchi, S. D., et al. 2016, AJ, 152, 212 Carpino, M., Milani, A., & Nobili, A. M. 1987, A&A, 181, 182 Duncan, M. J., Levison, H. F., & Budd, S. M. 1995, AJ, 110, 3073 Gallardo, T., Hugo, G., & Pais, P. 2012, Icarus, 220, 392 Gladman, B., Lawler, S. M., Petit, J.-M., et al. 2012, AJ, 144, 23 Gladman, B., Marsden, B. G., & Vanlaerhoven, C. 2008, Nomenclature in the Outer Solar System, ed. M. A. Barucci, H. Boehnhardt, D. P. Cruikshank, A. Morbidelli, & R. Dotson, 43 Gomes, R. S. 2011, Icarus, 215, 661 Gomes, R. S., Fern Ndez, J. A., Gallardo, T., & Brunini, A. 2008, The Scattered Disk: Origins, Dynamics, and End States, ed. M. A. Barucci, H. Boehnhardt, D. P. Cruikshank, A. Morbidelli, & R. Dotson, 259 Gomes, R. S., Gallardo, T., Fernández, J. A., & Brunini, A. 2005, Celestial Me- chanics and Dynamical Astronomy, 91, 109 Knezevic, Z., Milani, A., Farinella, P., Froeschle, C., & Froeschle, C. 1991, Icarus, 93, 316 Kozai, Y. 1985, Celestial Mechanics, 36, 47 Lykawka, P. S. & Mukai, T. 2007, Icarus, 189, 213 Article number, page 8 of 9 I(deg)a)leftisland2πJ=−0.0350.0◦12.7◦19.8◦23.9◦26.6◦28.5◦29.7◦30.4◦I(deg)b)rightisland2πJ=−0.00205π5π/43π/27π/42πω(rad)303540455055606570q(au)π5π/43π/27π/42πω(rad)M. Saillenfest and G. Lari: The long-term evolution of known resonant trans-Neptunian objects 2 1 L G 6 1 H S 6 1 A B Name (42301) (95625) (131696) (181867) (79978) (119878) (84522) (26375) (60621) (38084) (143707) (135571) (69988) (119068) (82075) (136120) (160148) (126619) (26181) (184212) 2001 KG76 2001 UR163 2002 GX32 2002 CZ248 2001 XT254 1999 CV118 1999 CC158 2002 CY224 2002 TC302 1999 DE9 2000 FE8 1999 HB12 2004 KZ18 2004 EG96 2002 GP32 2000 SR331 2003 UY117 2002 GG32 1998 WA31 2001 KC77 2004 HO79 2001 XQ254 2015 GP50 2012 FH84 2000 YW134 2004 XR190 2015 FJ345 2003 LG7 2015 KH162 2013 FQ28 2001 KV76 2002 CX154 2014 FC69 2014 FZ71 2015 RR245 1996 GQ21 2008 ST291 2004 PB112 (cid:63) (cid:63) (cid:63) (cid:63) (cid:63) (cid:63) (cid:63) (cid:63) (cid:63) (cid:63) (cid:63) (cid:63) 7 0 Y kp :k L 4:9 (cid:63) 4:9 (cid:63) 3:7 (cid:63) 3:7 3:7 3:7 5:12 5:12 2:5 2:5 2:5 2:5 2:5 2:5 2:5 2:5 2:5 2:5 2:5 2:5 2:5 2:5 2:5 2:5 3:8 3:8 1:3 1:3 1:3 1:3 2:7 3:11 3:11 1:4 2:9 2:11 1:6 5:27 (cid:63) (cid:63) (cid:63) 8 0 L G × × × × × × × × × × × × × × × × × × × × × × × × (cid:63) (cid:63) (cid:63) (cid:63) (cid:63) (cid:63) (cid:63) × × × × × × × × × a0 51.718 51.719 52.987 52.987 52.988 52.988 53.991 53.991 55.474 55.478 55.479 55.480 55.480 55.481 55.481 55.481 55.481 55.481 55.481 55.482 55.482 55.484 η0 0.939 0.958 0.901 0.919 0.946 0.949 0.904 0.896 0.784 0.897 0.909 0.890 0.847 0.871 0.907 0.896 0.899 0.907 0.890 0.913 0.907 0.894 −2πJ 0.02 0.04 0.07 0.026 0.04 0.04 0.017 0.011 0.13 0.18 0.05 0.105 0.0015 0.045 0.0013 0.023 0.064 0.03 0.12 0.24 0.08 0.19 ω 2.668 6.148 3.359 5.252 3.294 2.650 1.841 2.689 1.585 2.884 2.739 1.052 0.136 0.185 0.463 3.812 1.895 4.140 5.571 3.239 2.359 0.189 q 34.017 36.974 33.146 32.413 35.910 37.493 39.096 35.275 39.002 32.283 33.109 32.615 34.310 32.110 32.081 31.157 32.510 35.905 31.581 35.455 32.507 31.089 57.920 0.909 0.00035 5.489 41.146 62.649 62.660 0.809 0.817 0.065 0.5 1.391 5.913 50.905 32.483 69.430 0.835 0.076 4.098 34.421 (cid:63) 75.900 82.100 99.453 107.575 0.860 0.802 0.761 0.713 0.035 0.075 0.003 0.0055 4.277 4.360 5.721 0.158 56.117 33.942 42.566 35.392 Table 1. List of the trans-Neptunian objects studied in this paper. The resonance ratios kp : k are all with Neptune. The references used are abbreviated in LY07 (Lykawka & Mukai 2007), GL08 (Gladman et al. 2008), GL12 (Gladman et al. 2012), SH16 (Sheppard et al. 2016) and BA16 (Bannister et al. 2016). A cross indicates that the object is classified as "resonant"; a star indicates that the article provides also the oscillation amplitude of the resonant angle with its uncertainty. In LY07 and GL12, the star is omitted if the classification is said "insecure". The right part of the table gives the characteristics of the resonant secular models obtained in this paper: a0 is the reference semi-major axis chosen (au) whereas η0 (no unit) and −2πJ (au2rad2/yr) are the fixed parameters. Finally, ω (rad) and q (au) are the current secular argument of perihelion and reference perihelion distance of the object (see Sect. 2.1). Blank fields mean that the object was found non-resonant in the numerical integration starting from the nominal orbit given by AstDyS database. In the particular case of 2014 FZ71, a secular model was developed although the nominal orbital solution was not resonant (see Sect. 3.3). In theory, the 1-sigma error bars would allow it also for 2015 GP50, 2012 FH84 and 2013 FQ28. Article number, page 9 of 9
1210.8087	3	1210	2012-12-12T12:49:42	GJ 1214b revised. Improved trigonometric parallax, stellar parameters, orbital solution, and bulk properties for the super-Earth GJ 1214b	[ "astro-ph.EP", "astro-ph.SR" ]	GJ 1214 is orbited by a transiting super-Earth-mass planet. It is a primary target for ongoing efforts to understand the emerging population of super-Earth-mass planets around M dwarfs. We present new precision astrometric measurements, a re-analysis of HARPS radial velocity measurements, and medium-resolution infrared spectroscopy of GJ 1214. We combine these measurements with recent transit follow-up observations and new catalog photometry to provide a comprehensive update of the star-planet properties. The distance is obtained with 0.6% relative uncertainty using CAPScam astrometry. The new value increases the nominal distance to the star by ~10% and is significantly more precise than previous measurements. Updated Doppler measurements combined with published transit observations significantly refine the constraints on the orbital solution. The analysis of the infrared spectrum and photometry confirm that the star is enriched in metals compared to the Sun. Using all this information, combined with empirical mass-luminosity relations for low mass stars, we derive updated values for the bulk properties of the star-planet system. We also use infrared absolute fluxes to estimate the stellar radius and to re-derive the star-planet properties. Both approaches provide very consistent values for the system. Our analysis shows indicates that the favoured mean density of GJ 1214b is 1.6 +/-0.6 g cm^{-3}. We illustrate how fundamental properties of M dwarfs are of paramount importance in the proper characterization of the low mass planetary candidates orbiting them. Given that the distance is now known to better than 1%, interferometric measurements of the stellar radius, additional high precision Doppler observations, and/or or detection of the secondary transit (occultation), are necessary to further improve the constraints on the GJ 1214 star-planet properties.	astro-ph.EP	astro-ph	Astronomy&Astrophysicsmanuscript no. angladaGJ1214 October 29, 2018 c(cid:13) ESO 2018 GJ 1214b revised Improved trigonometric parallax, stellar parameters, orbital solution, and bulk properties for the super-Earth GJ 1214b Guillem Anglada-Escud´e1,2, B´arbara Rojas-Ayala3, Alan P. Boss2, Alycia J. Weinberger2, and James P. Lloyd4 2 1 0 2 c e D 2 1 . ] P E h p - o r t s a [ 3 v 7 8 0 8 . 0 1 2 1 : v i X r a 1 Universitat Gottingen, Institut fur Astrophysik, Friedrich-Hund-Platz 1, 37077 Gottingen, Germany 2 Carnegie Institution of Washington, Department of Terrestrial Magnetism, 5241 Broad Branch Rd. NW, Washington D.C., 20015, USA 3 Department of Astrophysics, American Museum of Natural History, Central Park West and 79th Street, New York, NY 10024 4 Department of Astronomy, Cornell University, 122 Sciences Drive, Ithaca, NY 14853, USA submitted March 2012 ABSTRACT Context.GJ 1214 is orbited by a transiting super-Earth-mass planet. It is a primary target for ongoing efforts to understand the emerg- ing population of super-Earth-mass planets around M dwarfs, some of which are detected within the liquid water (habitable) zone of their host stars. Aims. We present new precision astrometric measurements, a re-analysis of HARPS radial velocity measurements, and new medium- resolution infrared spectroscopy of GJ 1214. We combine these measurements with recent transit follow-up observations and new catalog photometry to provide a comprehensive update of the star -- planet properties. Methods. The distance is obtained with 0.6% relative uncertainty using CAPScam astrometry. The new value increases the nominal distance to the star by ∼10% and is significantly more precise than previous measurements. Improved radial velocity measurements have been obtained analyzing public HARPS spectra of GJ 1214 using the HARPS-TERRA software and are 25% more precise than the original ones. The Doppler measurements combined with recently published transit observations significantly refine the constraints on the orbital solution, especially on the planet's eccentricity. The analysis of the infrared spectrum and photometry confirm that the star is enriched in metals compared to the Sun. Results. Using all this new fundamental information, combined with empirical mass -- luminosity relations for low mass stars, we derive updated values for the bulk properties of the star -- planet system. We also use infrared absolute fluxes to estimate the stellar radius and to re-derive the star -- planet properties. Both approaches provide very consistent values for the system. Our analysis shows that the updated expected value for the planet mean density is 1.6±0.6 g cm−3, and that a density comparable to the Earth (∼ 5.5 g cm−3) is now ruled out with very high confidence. Conclusions. This study illustrates how the fundamental properties of M dwarfs are of paramount importance in the proper charac- terization of the low mass planetary candidates orbiting them. Given that the distance is now known to better than 1%, interferometric measurements of the stellar radius, additional high precision Doppler observations, and/or or detection of the secondary transit (oc- cultation), are necessary to further improve the constraints on the GJ 1214 star -- planet properties. Key words. Stars: individual :GJ 1214 -- Astrometry -- Techniques : radial velocities -- Stars: late-type 1. Introduction In the last few years, it has become clear that a large frac- tion of low mass stars (> 30% Bonfils et al. 2011) are or- bited by super-Earth-mass planets on relatively short period or- bits, including a handful detected in the liquid water (habit- able) zones (e.g., GJ 581d and GJ 667Cc: Mayor et al. 2009; Anglada-Escud´e et al. 2012a, respectively). These planets have small but non-negligible probabilities of transiting in front of their host stars. Given the favorable star -- planet size ratio, we already have the technical means to begin spectroscopic char- acterization of their atmospheres. GJ 1214b was the first super- Earth found to transit in front of an M dwarf (Charbonneau et al. 2009) and, as a consequence, it has received significant obser- vational and theoretical attention in the past three years. While the orbit is too close to the star to support hospitable oceans of liquid water, GJ 1214b offers the first opportunity to attempt at- mospheric characterization of a super-Earth (Bean et al. 2010; Croll et al. 2011; Berta et al. 2012), but apparently contradictory results have emerged from these studies. Different gases in the atmosphere of a planet absorb light more efficiently at certain wavelength ranges. As a result, one should be able to measure a different effective transit depth as a function of wavelength. For example, Croll et al. (2011) reported excess absorption in the K band, which would be compatible with an extended and mostly transparent atmosphere with a strong absorber in the near in- frared. On the other hand, Bean et al. (2010) could not detect significant features in the optical transit depths. This result was confirmed by the same group in a more extended wavelength range (Bean et al. 2011; D´esert et al. 2011) and has also been confirmed using HST spectrophotometric observations in the near infrared (Berta et al. 2012). Such a flat transmission spec- trum favors a very opaque atmosphere with high concentrations Send [email protected] requests offprint to: G. Anglada-Escud´e, e-mail: 1 Guillem Anglada-Escud´e et al.: GJ 1214b revised of water vapor as the main source of opacity, indicating that the planet could be mainly composed of water. All these interpretations rely on atmospheric models that are strongly dependent on the planet's bulk properties, especially its mass, radius, mean density, surface gravity, and stellar irradia- tion. Prior to the detection of the planet candidate, GJ 1214 was a largely ignored M dwarf. As a result, some of its fundamen- tal properties had significant uncertainties (e.g., its distance). Uncertainties in the fundamental parameters of GJ 1214 prop- agate strongly into the planet's bulk properties, adding an extra element of uncertainty in discussions about the possible nature of its atmosphere. Probably the most significant measurement we provide here is the new measurement of its trigonometric parallax at 0.6% precision (previous parallax measurement had a ∼ 10% uncer- tainty, van Altena et al. 2001). As a consequence, the lumi- nosity and mass of GJ 1214 have experienced significant up- dates as well. The orbital solution for GJ 1214 can also be up- dated using recently published transit observations and refined Doppler measurements obtained with our newly developed soft- ware (HARPS-TERRA, Anglada-Escud´e & Butler 2012). In ad- dition, the WISE catalog (Cutri et al. 2011) has also been re- cently released, adding four more absolute flux measurements of GJ 1214, thus enabling a comprehensive spectral energy dis- tribution adjustment and a more secure determination of Teff. In Section 2, we present the new CAPSCam astrometric mea- surements and our re-analysis of public HARPS Doppler mea- surements. A new Keplerian solution for GJ 1214b is presented in Section 2.3. In Section 3, we give an overview of the stellar properties in the light of the new distance measurement, its near infrared spectrum, and updated absolute magnitudes. Finally, Section 4 combines all of the transit observables with the new orbital solution and provides the posterior probability distribu- tions for the updated star -- planet parameters. Our conclusions are summarized in Section 5. 2. Observations and Data Reduction 2.1. Astrometryandtrigonometricparallax The trigonometric parallax of GJ 1214 has been obtained us- ing the Carnegie Astrometric Planet Search Camera (CAPSCam, Boss et al. 2009) installed in on the 2.5m duPont telescope of the Las Campanas Observatory (Chile). The observations span 26 months (June 2009 to Sep 2012) and 10 epochs have been ob- tained at an average precision of 1.0 milliarcsec (mas) per epoch. Each epoch consists of 20 or more exposures of 45 seconds each. GJ 1214 is significantly brighter than the average back- ground sources and would saturate the detector in less than 10 seconds. However CAPSCam can read out a small part of the array much faster than the full field (Boss et al. 2009). For GJ 1214, a window of 64×64 pixels is read out every 5 seconds, and all subrasters are added to the final full field image. The field of view of CAPSCam is 6.6 arcminutes wide and is typically rich in background stars, so a very robust reference frame with more than 30 objects can be used to correct for field distortions. Centroid extraction, source crossmatching, field distortion cor- rection, and astrometric solutions for all the stars in the field have been obtained using the ATPa astrometric software developed within the CAPS project (available upon request). The meth- ods and algorithms applied are outlined in Boss et al. (2009) and Anglada-Escud´e et al. (2012b). A measurement of the parallax and proper motion requires fitting 5 astrometric parameters simultaneously: initial offset in 2 Table 1. Local astrometric measurements of GJ 1214 JD (days) 2455368.781418 2455408.647011 2455638.911225 2455664.846719 2455779.634773 2455782.624045 2456018.887788 2456084.702758 2456134.624810 2456196.509776 R.A. (mas) 0.45 25.62 511.34 542.31 611.31 614.41 1111.53 1159.22 1184.17 1264.98 σR.A. (mas) 0.91 1.09 0.79 0.74 0.45 0.67 0.81 0.52 0.39 0.82 Dec. (mas) 12.30 -78.97 -567.61 -604.85 -826.25 -834.84 -1325.39 -1433.52 -1534.43 -1692.83 σDec. (mas) 3.86 2.22 1.88 1.64 0.93 1.07 1.32 1.08 0.74 0.88 R.A. and Dec., proper motion in R.A. and Dec., and the parallax itself. The formal uncertainties from a classic least-squares anal- ysis (e.g., from the diagonal of the covariance matrix) are usu- ally overoptimistic and do not properly account for correlations between parameters. To overcome these issues, we estimate the uncertainties in the astrometric parameters using a Monte Carlo approach. This is done by generating many realistic sets of mea- surements and measuring the standard deviation of the resulting derived parameters for the entire Monte Carlo-generated sample. To do this properly, a first realistic estimate of the epoch-to- epoch accuracy is needed. The outputs of the astrometric pro- cessing are the astrometric parameters of the target star and of all the other objects in the field. By combining the residuals of all the reference stars from all the epochs, we can compute the expected uncertainty per epoch. Doing this, we obtain an epoch- to-epoch accuracy of ∼ 1.3 mas on both R.A. and Dec.. Since the reference stars are fainter than the target, this is a conservative estimate of the real precision for the target. The star itself, which is not included in the reference frame, shows a standard devia- tion in the residuals of 1.0 mas/epoch, which is also consistent with the expected CAPSCam performance, assuming 20 minutes of on -- sky observations per epoch (see Boss et al. 2009). Then, we simulate 105 synthetic sets of astrometric observations (same format as Table 1) using the nominal parallax and proper mo- tion at the same epochs of observation. Random Gaussian noise with a single epoch uncertainty of 1.3 mas is injected into each synthetic data set and the 5 astrometric parameters are derived. The standard deviation of each parameter over the 105 solutions is the corresponding uncertainty. The obtained uncertainty for the differential parallax is 0.44 mas and corresponds to a rela- tive precision of ∼ 0.6%. We note that this Monte Carlo method implicitly accounts for parameter correlation (see discussion in Faherty et al. 2012). The differential astrometric measurements used to measure the differential parallax and proper motion of GJ 1214 are given in Table 1. The best fit solution over-plotted with the astrometric epochs is shown in Figure 1. The measured parallax is relative to the reference stars. Since they are not at infinite distances, these reference stars also have parallactic motion. As a result, the average paral- lax of the reference frame cannot be derived from the as- trometric observations alone. We obtain this average paral- lax (also called the parallax zero-point correction) using cat- alog photometry for the reference stars using the procedure described in Anglada-Escud´e et al. (2012b) with 19 reference stars with good CAPSCam astrometry (RMS of the residuals below 2.0 mas/epoch) and reliable catalog photometry (B<18 NOMAD and JHK s 2MASS photometry; Zacharias et al. 2005; Skrutskie et al. 2006). For this set of observations, we determine a zero -- point correction of -0.40 ± 0.3 mas. This value, when Guillem Anglada-Escud´e et al.: GJ 1214b revised Table 3. HARPS-TERRA radial velocity measurements. 1500 1000 500 ] s a m [ . A R . 0 6 3 0 -3 -6 ] s a m [ - C O ] s a m [ . c e D 0 -500 -1000 -1500 6 3 0 -3 -6 5400 5600 6000 JD-2450000 [days] 5800 6200 5400 5600 6000 JD-24000000 [days] 5800 6200 Fig. 1. Astrometric motion of GJ 1214 as a function of time. Top panels represent the motion in R.A.(left) and Dec. (right). Bottom panels show the residuals to the astrometric fit. The RMS of the residuals is below 1 mas. Table 2. Basic astrometric information and results from the anal- ysis of the astrometry. Parameter R.A. Dec. Catalog µ∗R.A. [mas yr−1] Catalog µDec. [mas yr−1] Relative µ∗R.A. Relative µDec. Relative parallax [mas] Zero-point correction [mas] Absolute parallax [mas] Distance [pc] Value 17 15 18.941 +04 57 49.7 1 585 2 -752 2 581.88 ± 0.5 -734.6 ± 0.8 69.11 ± 0.4 -0.4 ± 0.3 68.71 ± 0.6 14.55 ± 0.13 (1) 2MASS catalog (Skrutskie et al. 2006) (2) LSPM-NORTH Notes. catalog L´epine & Shara (2005) added to the relative parallax, produces the final absolute paral- lax measurement in Table 2. The inverse of the parallax in arc- seconds is the distance in parsecs (pc) and amounts to 14.55 ± 0.13 pc. The updated BVRIJHKs, W1, W2, W3 and W4 absolute magnitudes for GJ 1214 are presented and discussed in Section 3. We note that the measured proper motions are also differ- ential and contain an unknown offset due to the unknown aver- age motion of the background stars (e.g., all galactic plane stars move roughly in the same direction). Even though catalog val- ues are less precise than the ones we obtain, catalog proper mo- tions are usually corrected for proper motion zero -- point ambi- guities. Differential measurements as well as the suggested val- ues for the proper motion of GJ 1214 (LSPM-NORTH catalog, L´epine & Shara 2005) are also provided in Table 2. 2.2. Radialvelocitymeasurements The ESO public archive1 contains 21 reduced HARPS spectra of GJ 1214 obtained by Charbonneau et al. (2009). We rean- alyzed these public HARPS spectra using our software called HARPS-TERRA (HARPS Template Enhanced Radial veloc- ity Re-analysis Application, Anglada-Escud´e & Butler 2012). HARPS-TERRA derives differential RV measurement by con- structing a high signal-to-noise ratio template from the obser- vations and matching it to each spectrum using a least-squares approach. GJ 1214 is faint at optical wavelengths (V∼14). As a consequence, very long exposures (2400 sec) were required to obtain any signal at all. Still, the typical S/N at 6100 Å is only ∼10 so careful construction of the template is a key element in 1 http://archive.eso.org/wdb/wdb/eso/repro/form JD (days) RV (m s−1) σRV (m s−1) 2455036.57372 2455036.65153 2455037.58578 2455037.65309 2455038.53985 2455038.63702 2455039.55202 2455039.63876 2455040.56221 2455040.63961 2455041.57417 2455042.52391 2455042.54566 2455042.63521 2455045.55962 2455045.64403 2455046.55684 2455046.63141 2455047.55042 2455048.54997 2455048.61096 -10.82 -4.77 6.86 -3.37 5.67 5.61 -14.88 -13.53 14.15 5.90 5.68 -2.11 -5.57 -4.15 3.43 1.61 10.59 8.28 -13.26 1.64 3.03 1.80 2.08 3.63 1.72 1.41 1.78 1.69 1.61 1.84 2.49 1.70 3.91 1.37 1.42 2.92 2.52 1.97 1.94 2.19 1.93 1.63 order to achieve the maximum precision. The standard setup of HARPS-TERRA for M dwarfs uses all the echelle apertures red- der than the 22nd one (4400Å< λ < 6800Å) and adjusts a cubic polynomial to correct for the variability of the blaze function across each echelle order. Detailed algorithms and performance of HARPS-TERRA on a representative sample of stars is given in Anglada-Escud´e & Butler (2012). It is worth noticing that er- ror bars listed in 3 are smaller than those in Charbonneau et al. (2009), suggesting a more optimal usage of the Doppler informa- tion in the spectra. However, reduced error bars do not guarantee a more accurate orbital fit when the RV variability is dominated by systematic noise (either instrumental or stellar). To account for this, we included an unknown noise term in our Monte Carlo Markov Chain runs which had a substantial effect on the derived distributions, especially on the orbital eccentricity (see discus- sion in Section 2.3). 2.3. Orbitalsolution As discussed in Carter et al. (2011), uncertainties in the orbital parameters (especially the eccentricity) are the main limitation in the derivation of precise star -- planet parameters. Therefore, we are interested not only in the favoured orbital solution, but also in obtaining a realistic numerical representation of the pos- terior density function of the parameters to propagate them into the estimates of the star -- planet properties. These samples are generated using a Bayesian Monte Carlo Markov chain (MCMC) method. We use custom-made MCMC software to combine RV with transit observations and to obtain such dis- tributions. The software is based on the MCMC tools used in Anglada-Escud´e et al. (2012b) adapted to the Doppler plus tran- sit problem, and uses a Gibbs sampler. Jump scales of the Gibbs sampler are initialized using the diagonal elements of the co- variance matrix at the maximum likelihood solution, and they are adjusted in the first 105 steps (the "burn-in" period) to accept between 10% and 30% of the jump proposals for each param- eter. As discussed below, we assume flat priors for all the sam- pling parameters. As a consequence, the probability of accepting 3 Guillem Anglada-Escud´e et al.: GJ 1214b revised a new state is just the ratio of likelihoods at the current and pro- posed jump position. An outline of the general MCMC method applied to the Keplerian problem can be found in Ford (2006). The Keplerian model for the radial velocities is the same as in Anglada-Escud´e & Tuomi (2012) and the transit observations are predicted using the recipes given in the NASA Exoplanet Archive webpage 2. Transit observations strongly constrain the orbital period, but also put strong restrictions on the sum of the initial mean anomaly M0, the argument of the node ω (also called, mean longitude λ0 = M0+ω), and the product e cos ω (ec- centricity and argument of the node). Accordingly, we use λ0 as a free parameter. Since the transit instants explicitly depend on e cos ω, e sin ω and e cos ω are sometimes used as the MCMC sampling parameters. We also tested this approach, and, while it has some desirable properties (e.g., symmetry around zero eccentricity) -- this parameterization imposes an implicit prior 3 that severely biases the estimates of e to artificially larger values (e.g., see Ford 2006; Barros et al. 2011, for similar discussions). Therefore, at the sampling level, we choose those free parame- ters that, from our point of view, should have flat prior distribu- tions. These are: orbital frequency 1/P, RV semi-amplitude K, mean longitude λ0, argument of the node ω, and eccentricity e. Despite ω being poorly defined and uncertain at low eccentrici- ties, the convergence of the MCMC was substantially improved thanks to the use of λ0 instead of M0, so we find this parametriza- tion convenient and sufficient for our purposes. All the approxi- mate parameter values are already known from previous studies and, therefore, we initialize the MCMCs within 3 standard de- viations of the orbital solution proposed by Charbonneau et al. (2009). Real uncertainties in radial velocity measurements are diffi- cult to estimate properly, especially when the star is active. To account for that, we model the uncertainty of the i-th measure- i + s2, where ǫi are the formal uncertainties de- ment as σ2 rived from HARPS-TERRA (third column in Table 3), and s is a jitter parameter that accounts for the extra systematic noise. In this context, the jitter parameter s is treated as any other free parameter. i = ǫ2 We assign a constant uncertainty to each transit instant of 400 sec. This uncertainty is 10 times larger than the typical re- ported formal errors but it is more consistent with the differ- ent instants of transit measurements obtained simultaneously by different groups (see Table 8). While the MCMC convergence properties are greatly improved, 400 sec is a small fraction of the time-baseline covered by the transits and, as a consequence, the period is still strongly constrained. The transit time measure- ments have been extracted from Sada et al. (2010), Carter et al. (2011), and Kundurthy et al. (2011). In table 4, we provide the expected values for the most rel- evant combination of parameters for physical and transit pre- diction computations as derived from the combination of 10 MCMC runs with 107 steps each. Figure 2 is an illustration of the maximum a posteriori probability (MAP) solution and is given only for illustration purposes. Since the amplitude of 2 http://exoplanetarchive.ipac.caltech.edu/applications/TransitSearch/ guide/algorithms.html 3 The implicit prior comes from the fact that x = e sin ω and y = e cos ω are cartesian coordinates derived from the polar radial coordi- nate e and the angle ω. Because MCMCs are based on the integration of a probability distribution, to preserve the flat prior choices for e and ω, one would need to include the Jacobian of the transformation (1/e) as a prior when using x and y as the Markov chain sampling parameters. The use of such a prior is undesirable due to numerical stability issues (divergence for e = 0). 4 30 20 10 0 -10 -20 Transit P = 1.58040 d e = 0.11 K = 10.68 m s-1 RMS 3.40 m s-1 -0.8 -0.6 -0.4 -0.2 0 Orbital phase [days] 0.2 0.4 0.6 0.8 Fig. 2. HARPS-TERRA radial velocity measurements folded to the orbital period. The instant of transit is depicted as a dashed vertical line. ω n i s e 0.6 0.4 0.2 0 -0.2 -0.4 -0.6 10 combined chains 1 chain (107 steps) 1 y t i l i b a b o r P 0.5 Distribution for e (arbitrary normalization) Cumulative distribution for e e < 0.23 (95% c.l.) 0.2 0.4 0.6 0 0 0.1 0.2 Eccentricity 0.3 0.4 -0.4 -0.2 0 e cos ω Fig. 3. Left. Distribution of MCMC states for the derived quan- tities e cos ω and e sin ω as obtained from 1 MC chain of 107 steps (red) and the combination of 10 chains of 107 steps each (black). Only 1% of the steps are included in this plot to improve visualization. Right. Marginalized probability distribution of the eccentricity is shown in black (arbitrary normalization). The cor- responding cumulative distribution (red histogram) shows that eccentricities higher than 0.23 are ruled out at a 95% confidence level. the signal is small compared to the noise, the eccentricity is still poorly constrained. The posterior distribution of the eccen- tricity has a maximum very close to 0 and monotonically de- creases with e (see Figure 3) implying that close to circular orbits are favored. Because of their significance for the transit observables (e.g., see Section 4), an illustration of the MCMC samples of the derived parameters e cos ω and e sin ω is also provided in figure 3. Previous estimates of the eccentricity of GJ 1214b Charbonneau et al. (2009); Carter et al. (2011, e.g., ) seemed to favor eccentricities close to 0.1, but still compatible with 0. In initial tests, we recovered this same slightly higher eccentricity value if the noise parameter s was fixed to 0. This indicates that previous estimates of the orbital elements were bi- ased due to a non-consistent treatment of the RV uncertainties. Even though stellar jitter dominates the error budget (s ∼ 2.9 m s−1), the RMS of the MAP solution using the new HARPS- TERRA measurements is lower (3.4 m s−1) than the one reported by Charbonneau et al. (2009, 4.4 m s−1). Given the strong ef- fect of the jitter parameter s on the result, and until more RV measurements become public, we strongly recommend the use of the updated solution in Table 4 for any future work (e.g., in searching for secondary transits). Applying the same 95% con- fidence level used by Charbonneau et al. (2009), we obtain an upper limit to the eccentricity of 0.23. Guillem Anglada-Escud´e et al.: GJ 1214b revised Table 4. Expected values of the most useful parameter combi- nations obtained using RV and transit observations. Numbers in parentheses represent the standard deviation of the distributions (last two significant digits of the expected values). Parameter P [days] K [m s−1] λ0 [deg] ecos ω esin ω Jitter s [m s−1] γ Other parameter combinations e M sin i [M jup] M sin i [M ] ⊕ a [AU] Ntransits NRV RMSRV [m s−1] 1.580400(14) 10.9(1.6) 210.9(6.2)a -0.033(55) -0.044(90) 3.6(1.1) 0.35(1.2) <0.23b 0.0195(28)c 6.20(91)c 0.0148c 30 21 3.5 (a) Julian date of the reference epoch on which λ0 is computed Notes. is the first epoch of the RV data at T0 = 2455036.57372 days. For a circular orbit λ0 is equal to the mean anomaly M0 at the reference (b) Distribution of the eccentricity peaks close to 0 (see Figure epoch. 3). (c) Assumes M . The uncertainty in the mass of GJ 1214 is considered in the final proposed parameters for the star -- planet system given in Section 4. ∗=0.176 M ⊙ Another interesting parameter is the minimum mass of GJ 1214b. Even though the new RV amplitude K is smaller (∼ 10.9 m s−1) than the previously reported one (∼12.5 m s−1), the updated minimum mass does not change significantly com- pared to Charbonneau et al. (2009) due to the similar relative increase in the updated stellar mass derived from the new dis- tance determination (see Section 4). The origin of the slightly smaller amplitude is likely caused by stellar activity. By analyz- ing Doppler measurements on known active M dwarfs (e.g., AD Leo, Reiners et al. 2012), we found that the classic HARPS- CCF approach always produces a larger Doppler amplitude than the one derived using HARPS-TERRA (between 1.1 to 1.5 times larger) if a candidate signal is caused by activity. The likely ex- planation for this is the different sensitivities of the two meth- ods to the changes in the stellar line-profiles. While one could use this effect to obtain a further diagnostic to check the reality of low-mass candidate planets (under investigation), in this case it means that the most likely source of systematic noise comes from the star rather than from the instrument. Doppler follow-up of the star is needed to better understand the origin of the extra noise and to further constrain the orbital solution. 3. Updated properties for GJ 1214 3.1. Metallicity techniques, Two independent the photometric method by Schlaufman & Laughlin (2010) and the K-band spectroscopic [Fe/H] index by Rojas-Ayala et al. (2012), agree on the metal- richness of GJ 1214, with [Fe/H] = +0.28 dex and [Fe/H] = +0.20 dex, respectively. However, since the [Fe/H] photomet- Table 5. Updated absolute photometry for GJ 1214. Johnsons- Cousins BVRI photometry is from (Dawson & Forbes 1992). JHKs photometry is from the 2MASS catalog (Skrutskie et al. 2006). The W1, W2, W3 and W4 are the mid-infrared bands from the WISE Preliminary Data Release (Cutri et al. 2011) (central wavelengths are 3.4 µ, 4.6µm, 12µm and 22µm). Properties derived from the fits of the absolute photometry to the BT-Settl-2010 model grid are also provided. MB MV MR MI MJ MH MKs MW1 MW2 MW3 MW4 Teff [K] L ∗ R ∗ [Fe/H]phot [Fe/H]spec [M/H]spec [10−3 L [R ⊙ ] 15.532 ± 0.080 13.822 ± 0.041 12.441 ± 0.043 11.608 ± 0.042 8.934 ± 0.041 8.274 ± 0.041 7.964 ± 0.038 7.781 ± 0.041 7.614 ± 0.039 7.407 ± 0.041 7.204 ± 0.178 3252 ± 201 4.05 ± 0.19 0.201 ± 0.010 +0.13, +0.052 +0.203 +0.153 ] ⊙ Notes. rors. (3) Rojas-Ayala et al. (2012) (2) (1) Uncertainty in Te f f only accounts for statistical er- (2012) Schlaufman & Laughlin (2010); Neves et al. ric calibrations depend on the distance of the star (to obtain MKs), the photometric value of GJ 1214 needs to be re-calculated using its updated distance. The updated absolute magnitudes are given in Table 5. The updated photometric [Fe/H] values are [Fe/H] = +0.13 dex and [Fe/H] = +0.05 dex, using the Schlaufman & Laughlin (2010) calibration and the recent cali- bration by Neves et al. (2012), respectively. Considering the dis- persions associated with the calibrations (σ 0.1-0.15 dex), the estimates are consistent with each other, corroborating that GJ 1214 is a solar or super-solar [Fe/H] star. A comparative approach can also be performed to confirm the metal-richness of GJ 1214. Figure 4 shows the K-band spec- tra of Gl 699 (Barnard's star), Gl 231.1B, and GJ12144. Gl 699 is the second nearest M star to the solar system and its kinematics, Hα activity, and [Fe/H] measurements are consistent with Gl 699 being an old disk/halo star (Gizis 1997; Rojas-Ayala et al. 2012). Gl 231.1B is the low-mass companion of a nearly solar metallic- ity G0V star ([Fe/H]=-0.04 dex, Valenti & Fischer (2005)). The overall shapes of the K-band spectra of all three stars are quite similar (same spectral type), and the morphology of the surface- gravity-sensitive CO bands corresponds to M dwarf stars (log g ∼ 5). However, all the absorption features of GJ 1214 are stronger than the ones exhibited by metal-poor Gl 699 ([Fe/H] = -0.39 dex, Rojas-Ayala et al. (2012)) and solar metallicity Gl 231.1B. Therefore, also in a relative sense, the strong absorption features favor a high metal content for GJ 1214. 4 These spectra are part of the K-band spectral atlas by Rojas-Ayala et al. (2012), and available to the community in the online version of that article. 5 Guillem Anglada-Escud´e et al.: GJ 1214b revised ] 1 - m µ 2 - m c W [ λ F d e z i l a m r o N Na I Ca I H2O CO CO Na I CO 1.4 1.2 1.0 0.8 GJ 1214 Gl 231.1 B 0.6 Gl 699 0.4 0.2 0.0 2.1 2.2 λ [µm] 2.3 2.4 Fig. 4. K-band spectra of GJ 1214 (top), Gl 231.1B (middle), and Gl 699 (bottom). All the stars have the same K-band spectral type (same overall shape of their spectra), but different metallic- ities, as indicated by the strengths of their absorption features. GJ 1214's metallicity should be at least equal to or higher than solar. 3.2. Stellartemperature,luminosityandradiusfrommodels We first provide an overview of previous methods and de- terminations of the temperature, luminosity, and radius of GJ 1214. In the discovery paper of GJ1214b, (Charbonneau et al. 2009) estimated a Teff=3026K and a R⋆= 0.211R for GJ 1214. ⊙ Kundurthy et al. (2011) obtained Teff =2949K and R⋆= 0.211R . ⊙ Constraining parameters to stellar evolution isochrones by Baraffe et al. (1998), Carter et al. (2011) led to estimates of Teff=3170K and R⋆= 0.179R , assuming that GJ 1214 is an "old ⊙ star". However, all these results were based on the distance deter- mination of ∼ 13 pc by van Altena et al. (2001), and reveal that the effective temperature of GJ 1214 is quite uncertain, given all the degeneracies in the evolutionary model used and a low precision distance estimate. We attempted to get an estimate of the stellar radius using up-to-date interferometric empirical calibrations. For example, Kervella et al. (2004) provides an empirical luminosity -- radius calibration that extends to low mass stars and uses absolute mag- nitudes as the only inputs. We tried different combinations of photometric bands obtaining inconsistent results. All the rela- tions involving optical bands (BVRI) gave values larger than 0.4 R , which are very unrealistic considering the spectral type of ⊙ GJ 1214. The relations restricted to infrared colors(J, H, and K) provided estimates a bit more realistic (between 0.15 to 0.2 R ) ⊙ but still not fully compatible with each other. We also tried the calibration provided by Demory et al. (2009) that used K band photometry to minimize the metallicity effect on optical magni- tudes. In Demory et al. (2009), new measurements of 6 new M dwarfs were presented and a new radius -- luminosity calibration was derived. They found a remarkable agreement with the evolu- tionary models of Baraffe et al. (1998) if the measured radii was plotted against the K absolute magnitude. Using this approach, we obtain a stellar radius of 0.193 R which, at least, seems to ⊙ be in the expected range. Given that the state-of-the-art empirical relations are not en- tirely self consistent, we also obtained a new estimate of the ef- fective temperature, luminosity, and radius of GJ 1214 using up- dated BVRIJHKsW1W2 absolute photometry (Table 5) together with the model atmosphere grid BT-Settl-2010 (Allard et al. 6 2011)5. The lack of indicators of youth in GJ 1214, such as Hα emission, along with its space motion, gives an age estimate of 3 -- 10 Gyr for the star (Reid & Gizis 2005). This age estimate is also supported by its long rotation period (Charbonneau et al. 2009). Therefore, we fixed the surface gravity of the syn- thetic spectral models to log g = 5.0 and [M/H]=0.0. Figure 5 shows the bolometric luminosity as function of absolute mag- nitude for the BT-Settl-2010 grid. The inferred bolometric lu- minosities with the JHKsW1W2 photometry are consistent with each other, and higher than the luminosities obtained with the BVRI photometry. Previous synthetic models (e.g. NextGen; Hauschildt et al. 1999) showed a lack of flux in the K band when compared with observed spectra of low-mass stars. New solar abundances and the inclusion of dust grain formation seems to have solved most of the previous discrepancy, allowing the BT- Settl-2010 models to reproduced fairly well the infrared spectral energy distribution (SED) of M-dwarfs (Allard et al. 2011, how- ever, the FeH opacity data is still incomplete for this region). Although models have also been improved at shorter wave- lengths (BVRI), they remain too bright in the ultraviolet and vis- ible part of the M dwarf spectra, possibly due to missing sources of opacity in the modeling process (Allard et al. 2011). The ef- fective temperatures and radii corresponding to the best SEDs fit to the 9 wavebands BVRIJHKsW1W2, only BVRI, and only JHKsW1W2, obtained from their respective bolometric lumi- nosities in Figure 5, are shown in Figure 6. Given that 1) bolo- metric luminosities obtained with JHKsW1W2 photometry are consistent with each other, 2) the models provide a better fit at longer wavelengths, and 3) the empirical luminosity -- radius pro- vides more self-consistency using nIR colors, the fit with only JHKsW1W2 provides the most reliable results and is the one to be used in deriving further properties of the star -- planet system. The SED fitting using only these nIR bands is further sup- ported by other results. For example in Section 4, we will ob- ) tain an independent stellar radius estimate (0.211 ± 0.011 R ⊙ combining the stellar mass with direct observables from the light curve and Doppler data. Such an estimate is compatible ). Also, with the value we discussed in this section (∼0.2 R ⊙ the estimate of the effective temperature (Teff ∼ 3250K) is in excellent agreement with the effective temperature derived from water absorption in K-band by Rojas-Ayala et al. (2012, Teff =3245 K). Furthermore, next Section shows this Teff also agrees with the one derived from SED fits to the evolutionary models (Baraffe et al. 1998). 3.3. Stellarmass Although most of the previously reported mass-radius estimates of GJ 1214 agree, we note that they all assume the trigonomet- ric distance estimate from van Altena et al. (2001). This distance determination has an uncertainty of 10% which is not always ac- counted for in the aforementioned predictions. We also note that the M1V-M4V spectral type range spans a broad range of stellar masses (from 0.4 to 0.15 M ) and effective temperatures (3500 to 2800 K). This makes any mass estimate very sensitive to un- certainties in the parallax mesurement. Given the updated dis- tance, we can apply the Delfosse et al. (2000) (hereafter DF00) relations to derive a new mass for GJ 1214 using the absolute J, H and K magnitudes. The masses derived from each photomet- ric band are M(J) = 0.174 M and M(K)= 0.177 , which are in very good agreement. ⊙ ,M(H)= 0.177 M ⊙ ⊙ ⊙ 5 http://phoenix.ens-lyon.fr/simulator/index.faces Guillem Anglada-Escud´e et al.: GJ 1214b revised MB MB MV MV MR MR MI MI MJ MJ MH MH MKs MKs MW1 MW1 MW2 MW2 ) N U S L / L ( G O L -2.2 -2.3 -2.4 -2.5 -2.6 -2.7 -2.8 -2.9 18 16 14 ABSOLUTE MAGNITUDE 12 10 8 Fig. 5. Luminosity as function of absolute magnitude for all the wavebands listed in Table 5. Open diamonds represent the pho- tometry of GJ 1214. The color asterisks represent the BT-Settl- 2010 values, and the color dotted lines the linear interpolations between them. The dashed line indicates the mean bolomet- ric luminosity estimated using all the photometry (log(L/L )= - 2.57). The dotted-dashed line indicates the mean luminosity with only the BVRI photometry (log(L/L )= -2.73), and the 3 dotted- dashed line the adopted mean luminosity for GJ 1214, derived using only the JHKsW1W1 absolute magnitudes (log(L/L )= -2.44). The lowest bolometric luminosity corresponds to the I magnitude, where the SED is very steep. ⊙ ⊙ ⊙ Another way to obtain an estimate of the mass for GJ 1214 is to compare the absolute magnitude with the synthetic colors from the evolutionary models in Baraffe et al. (1998). Unfortunately, Baraffe et al. (1998) does not contain models for stars with [Fe/H]>0.0. As a result the optical colors (e.g., B, V, R, I), cannot be properly adjusted to a model. The fit in- cluding the BVRIJHK magnitudes has Te f f = 2987 K (M = and is of very low quality (χ2 = 45 for 6 degrees of 0.12 M ⊙ freedom). The fit to BVR alone give Te f f = 2880 K, M = 0.11 and a χ2 = 10.0 for 3 degrees of freedom. In contrast, the best M ⊙ fit model obtained from adjusting J, H, and K has a χ2 of 1.92 for 3 degrees of freedom, an effective temperature of 3225 K and a mass of M = 0.172 M . Note that the mass value is very good ⊙ agreement with the mass estimate derived from the DF00, and the effective temperature is surprisingly close to one from the at- mospheric transfer models in the previous Section 3.2. How the uncertainty in the stellar mass affects the planet parameters will be discussed and accounted for in Section 4. 4. Star -- planet parameters from direct observables 4.1. Usingthestellarmassastheinput Information from the transit light curves together with the orbital solution can be combined with the stellar mass (or the stellar radius) to fully characterize the bulk properties of the star -- planet system. These methods have been devel- oped over the years by different authors and we suggest read- ing Seager & Mall´en-Ornelas (2003), Southworth (2008) and Carter et al. (2011) for more information on the derivations. This subsection describes the relevant relations and the general ap- proach used to derive the star -- planet parameters using the stellar mass as the input (mass -- input approach). ] 1 - m µ 2 - m c 1 - s g r e [ λ F 5•10-10 4•10-10 3•10-10 2•10-10 1•10-10 0 0 Teff=3250K, R=0.20Rsun Teff=3173K, R=0.18Rsun Teff=3077K, R=0.16Rsun 1 2 3 µm 4 5 Fig. 6. Spectral energy distribution fits for GJ 1214 for the ef- fective temperatures and radii derived from photometry. The magenta, green, and blue SEDs are the spectral templates that represent the values obtained with only BVRI photometry, only JHKsW1W2 photometry, and the 9 wavebands, respectively. All spectral templates have solar metallicity, log g = 5.0 and distance equal to 14.47 pc. The photometry data of GJ 1214 are plotted as black dots. Color dots indicate flux levels of each spectral template integrated over the corresponding filter bandwidth, de- picted by the horizontal black lines. The spectral template with Te f f =3250K and R=0.2R provides the best fit to the infrared ⊙ photometry of GJ 1214. Assuming a circular orbit for the planet, the transit light curve alone allows one to obtain a direct measurement of the mean stellar density ρ∗,circ. Given a fully Keplerian solution from the Doppler data, the stellar density also depends on the eccen- tricity of the orbit and can be written as ρ∗ = ρ∗,circ √1 − e2 1 + e sin ω 3     . (1) Because no prior information is required from the orbital fit, ρ∗,circ is a quantity typically provided by the studies analyz- ing transit light curves of GJ 1214 (e.g., Carter et al. 2011; Kundurthy et al. 2011). These two studies present the analy- sis of new light curves and combine them with previous light curves from Charbonneau et al. (2009) and Sada et al. (2010). A weighted mean of ρ∗,circ from these two studies will be used in all that follows. The detailed derivation of Eq. 1 was first given by Seager & Mall´en-Ornelas (2003) for circular orbits, while Carter et al. (2011) included the dependence on the eccentricity. With the mean stellar density from Equation 1 and the stellar mass from the empirical calibrations discussed in Section 3.3, one can trivially derive the radius of the star. Using this radius and the transit depth Rp/R (again, a direct observable from the light curves) one then obtains the radius of the planet. The com- bination of the minimum mass (from RV) and the inclination (from transit light curve) provides the planet's true mass, which is then combined with the planet radius to finally derive the mean planet density. To obtain realistic a posteriori distributions, one needs to assume realistic distributions for the input parameters of the model. The complete list of input parameters used at this point are given in Table 6. For all the values borrowed from the lit- erature, a Gaussian distribution N (cid:2)µ, σ(cid:3) is assumed where µ is ∗ 7 Guillem Anglada-Escud´e et al.: GJ 1214b revised Table 6. Input parameters used to compute the Monte Carlo dis- tributions of the star -- planet parameters Parameter name Distribution type Expected value Standard deviation Ref.A ∗ ] [M ⊙ M ∗ ρ∗,circ [g cm−3] a/R ∗ Rp/R Inc. [deg] Period [days] K [m s−1] λ0 [deg] e cos ω e sin ω Gaussian Gaussian Gaussian Gaussian Gaussian Bayesian Bayesian Bayesian Bayesian Bayesian 0.176 23.695 14.62 0.01178 89.19 1.580400 10.9 210.9 -0.033 -0.044 0.009 1.7 0.3 0.001 0.5 1.6 6.2 0.055 0.090 1.4 10−5 1,2 3,4 3,4 3,4 3,4 1 1 1 1 1 Notes. (A) (1) This work, (2) Delfosse et al. (2000), (3) Kundurthy et al. (2011), (4) Carter et al. (2011) the preferred value and σ is its published uncertainty. For the parameters derived from the orbital solution in Section 2.3, we directly draw samples from the MCMC distributions generated during the orbital analysis. The uncertainty in the stellar mass has to be accounted for at this point. Given the precision in the distance and in the J, H and K photometry, the major source of uncertainty in the mass is due to the actual accuracy of the DF00 calibration. This accuracy is not very well known, but other stud- ies suggest that it should be correct at the 5% level, which is the relative uncertainty we will use for the stellar mass. The process of solving for all the star -- planet parameters is done for the 106 synthetic input sets. The result is a numerical representation of the empirical probability distributions for the derived star -- planet properties. In Figure 7 we show the obtained distributions concerning the planet properties only (mass, radius, density and surface gravity). The surface gravity of the planet can also be derived from observables only using the prescrip- tions given by Southworth (2008) (combination of light curve parameters and parameters from the orbital solution). Table 7, provides the expected values and standard deviations of the dis- tributions for each derived star -- planet parameter. We find that the mean density of GJ 1214b has to be smaller than 2.40 g cm−3 with a 95 % confidence level (c.l.), with 1.69 g cm−3 being the expected value of the distribution. An Earth- like density (e.g., ρ > 5.5 g cm3) is thus ruled out at a > 99.9 % c.l. As illustrated in Fig. 7 this density range confirms that the planet is more similar to a small version of Neptune, or a water dominated body (Berta et al. 2012, e.g., ocean planets, ), rather than a scaled-up version of the Earth with a solid surface. This information by itself does not solve the question about the atmo- spheric composition discussed in the Introduction, but, at least, eliminates the uncertainty in the distance of the star in modeling of the possible planetary structure. We want to stress that, thanks to the new distance estimate, we now find a remarkably good agreement of the derived stellar radius compared to the one obtained from the SED fit in Section 3.2. As a final test, we reproduced the star -- planet parameters us- ing the stellar radius instead of the stellar mass. This consists of doing the following: the stellar radius (SED fit) combined with the mean stellar density (light curve plus RV) gives the stellar mass. Then the stellar mass combined with the RV ob- servables and the inclination (light curve) provides the planet mass. In parallel, the stellar radius with the transit depth (light 8 0.30 0.25 0.20 0.15 0.10 0.05 0.00 0.20 0.15 0.10 0.05 y t i l i b a b o r P y t i l i b a b o r P 0 5 10 15 Planet mass [MEarth] 20 0.30 0.25 0.20 0.15 0.10 0.05 0.00 0.25 0.20 0.15 0.10 0.05 y t i l i b a b o r P y t i l i b a b o r P 0 1 2 3 4 5 Planet radius [REarth] Uranus Neptune Europa (Jovian moon) Earth 0.00 0 5 10 15 Planet surface gravity [m s-2] 20 0.00 0 1 2 4 Planet density [g cm-3] 3 5 Fig. 7. Final probability distributions for most significant bulk properties of GJ 1214b. The mean densities of a few represen- tative Solar system objects are marked as black vertical bars for reference. Table 7. Suggested new star -- planet parameters as derived using the stellar mass as the only input parameter. Note that some of the distributions in Fig. 7 are strongly asymetric. ] Parameter name Star mass [M ⊙ Star radius [R Star surface gravity [m s−1] log g [g in cm s−1]b Star mean density [g cm−3] ⊙ ] Planet radius [R ] ⊕ Planet mass [M ] ⊕ Planet surface gravity [m s−2] Planet mean density [g cm−3] Planet Teq[K] (Bond albedo = 0) Planet Teq[K] (Bond albedo = 0.75) Maximum Expected value - probability† 0.176 a Standard deviation 0.0087 0.213 1032 5.01 26.36 2.80 6.26 7.66 1.56 0.211 1109 5.04 27.7 2.72 6.19 7.68 1.69 576 407 0.011 220 0.07 8.9 0.24 0.91 1.19 0.61b 14 11 Notes. (†) Maximum of the marginalized posterior distribution (a) Used as input (b) From light curve and Doppler analysis (no spectral adjust- ment) (c) ρp < 2.40 g cm−3 with a 95% c.l. curve) gives the planet radius, which combined with the mass finally provides the planet density. This approach, however, has a non-obvious drawback. The mass obtained through the stellar density formula depends as the third power on the stellar radius ∗ ∝ R3), making the stellar mass determination very sensitive (M to small radius changes. Going the other way around (starting from the mass, derive the radius), the stellar radius depends as M−1/3, resulting in a much weaker dependence. For example, a 5% uncertainty in the mass translates to a 1.6% uncertainty in the radius. On the other hand, a 5% uncertainty in the radius trans- lates to a 15% uncertainty in the mass. When the uncertainty in the stellar density is also included, this difference is even more pronounced. Therefore, we consider the mass input method a much more robust way of providing a consistent picture. The updated values using the mass input approach for the star -- planet parameters are in Table 7 and the marginalized distributions for the relevant planet parameters are depicted in Fig. 7 Guillem Anglada-Escud´e et al.: GJ 1214b revised 5. Conclusions The metallicity of GJ 1214 (and of M dwarfs in general) and its effects on the observable fluxes (e.g., optical fluxes) still require a better understanding, both observationally and theoretically. Even though spectroscopic methods in the nIR are getting bet- ter at determining the metal content of cool stars, precise direct measurements of their distances are still required for the charac- terization of exoplanets around M dwarfs. At least for the stel- lar masses, the model predictions from the near-infrared fluxes seem to be in good general agreement with current empirical relations derived from measured masses of M dwarfs in bina- ries (e.g., DF00). This is fortunate because the stellar mass is the only input parameter required to derive the star -- planet bulk properties from direct observables. The fit of the stellar spec- tral energy distributions to the BT-Settl-2010 model grid also provides consistent estimates in the stellar properties if infrared photometry is used. Even with the remaining ambiguities in the stellar parameters, both approaches (using a stellar mass from empirical relations, or deriving a radius from the absolute fluxes) lead to consistent results for the star -- planet bulk properties, but the mass input approach is preferred due to the lower sensitivity of the method to uncertainties in the input parameters. Several studies have been published in recent years trying to better char- acterize the GJ 1214 star -- planet system. The major source of the reported uncertainties came from directly observable quantities that we have improved, collected, and combined here: trigono- metric parallax and corresponding absolute fluxes, improved RV measurements, inclusion of all of the transit observations in the derivation of the orbital solution, additional infrared flux mea- surements, and light curve observables derived from photomet- ric follow-up programs. We now find remarkable agreement of the derived star prop- erties obtained by comparing the JHK fluxes to up-to-date at- mospheric and evolutionary models. All previous studies had to invoke some mechanism (e.g., spot coverage) to justify the mis- match between the predicted versus observed properties of the star. This alone highlights the importance of obtaining direct and accurate distance measurements of low mass stars. At this time, the quantity that most requires further improve- ment and/or independent determination is the radius of the star. To our knowledge, this can be observationally achieved with 3 different methods: 1) additional RV measurements to fur- ther constrain the orbital eccentricity, 2) direct measurement of the stellar diameter using optical/nIR interferometry (e.g., von Braun et al. 2012), or 3) detection of the secondary tran- sit (whose instant strongly depends on e and ω). We note that GJ 1214 is faint at optical wavelengths compared to other stars observed by high precision RV instruments (e.g., HARPS or HIRES Bonfils et al. 2011; Vogt et al. 2010). In the case of HARPS, integrations of 45 min were required to obtain a pre- cision of ∼3.4 m s−1, so a refinement of the orbital eccentric- ity through additional RV measurements is time-consuming but quite possible. On-going space -- based photometric observations in the mid-infrared (e.g., using Warm Spitzer/NASA) should be able to detect the secondary transit soon and pin down the or- bital eccentricity to greater precision. We hope that the updated orbital solution provided here facilitates the task of finding such a secondary transit. Acknowledgements. We thank the referee D. S´egransan for useful comments that helped improving the manuscript. GA has been partially supported by a Carnegie Postdoctoral Fellowship and by NASA Astrobiology Institute grant NNA09DA81A. BR thanks the staff and telescope operators of Palomar Observatory for their support. The CAPS team (APB, AJW, GA) thanks the staff and telescope operators of Las Campanas Observatory for the very succesful ob- serving runs and the Carnegie Observatories for continuous support of the CAPS project. We thank France Allard for helpful discussions about various topics. Part of this work is based on data obtained from the ESO Science Archive Facility. This research has made use of NASA's Astrophysics Data System Bibliographic Services, the SIMBAD database, operated at CDS, Strasbourg, France. This pub- lication makes use of data products from the Two Micron All Sky Survey, which is a joint project of the University of Massachusetts and the Infrared Processing and Analysis Center/California Institute of Technology, funded by the National Aeronautics and Space Administration and the National Science Foundation. This publication makes use of data products from the Wide-field Infrared Survey Explorer, which is a joint project of the University of California, Los Angeles, and the Jet Propulsion Laboratory/California Institute of Technology, funded by the National Aeronautics and Space Administration. References Allard, F., Homeier, D., & Freytag, B. 2011, arXiv:1112.3591 Anglada-Escud´e, G., Arriagada, P., Vogt, S. S., et al. 2012, ApJ, 751, L16 Anglada-Escud´e, G., Boss, A. P., Weinberger, A. J., et al. 2012b, ApJ, 746, 37 Anglada-Escud´e, G. & Butler, R. P. 2012, ApJS, 200, 15A Anglada-Escud´e, G. & Tuomi, M. 2012, A&A, 548A, 58A Baraffe, I., Chabrier, G., Allard, F., & Hauschildt, P. H. 1998, A&A, 337, 403 Barros, S. C. C., Faedi, F., Collier Cameron, A., & et al. 2011, A&A, 525, A54 Bean, J. L., D´esert, J.-M., Kabath, P., et al. 2011, ApJ, 743, 92 Bean, J. L., Miller-Ricci Kempton, E., & Homeier, D. 2010, Nature, 468, 669 Berta, Z. K., Charbonneau, D., D´esert, J.-M., et al. 2012, ApJ, 747, 35 Bonfils, X., Delfosse, X., Udry, S., et al. 2011, arXiv:1111.5019 Boss, A. P., Weinberger, A. J., Anglada-Escud´e, G., et al. 2009, PASP, 121, 1218 Carter, J. A., Winn, J. N., Holman, M. J., et al. 2011, ApJ, 730, 82 Charbonneau, D., Berta, Z. K., Irwin, J. et al. 2009, Nature, 462, 891 Croll, B., Albert, L., Jayawardhana, R., et al. 2011, ApJ, 736, 78 Cutri, R. M., Wright, E. L., Conrow, T., et al. 2011, Explanatory Supplement to the WISE Preliminary Data Release Products, Tech. rep., IPAC/CalTech Dawson, P. C. & Forbes, D. 1992, AJ, 103, 2063 Delfosse, X., Forveille, T., S´egransan, D., et al. 2000, A&A, 364, 217 Demory, B.-O., S´egransan, D., Forveille, T., et al. 2009, A&A, 505, 205 D´esert, J.-M., Bean, J., Miller-Ricci Kempton, E., et al. 2011, ApJ, 731, L40 Faherty, J. ,Burgasser, A. J., Walter, F. M., et al. 2012, ApJ752, 56F Ford, E. B. 2006, ApJ, 642, 505 Gizis, J. E. 1997, AJ, 113, 806 Hauschildt, P. H., Allard, F., Ferguson, J., et al. 1999, ApJ, 525, 871 Kervella, P., Th´evenin, F., Di Folco, E., & S´egransan, D. 2004, A&A, 426, 297 Kundurthy, P., Agol, E., Becker, A. C., et al. 2011, ApJ, 731, 123 L´epine, S. & Shara, M. M. 2005, AJ, 129, 1483 Mayor, M., Bonfils, X., Forveille, T., et al. 2009, A&A, 507, 487 Neves, V., Bonfils, X., Santos, N. C., et al. 2012, A&A, 538, A25 Reid, I. N. & Gizis, J. E. 2005, PASP, 117, 676 Reiners, A., Shulyak, D., Anglada-Escud´e, G., et al. 2012, A&Asubmitted Rojas-Ayala, B., Covey, K. R., Muirhead, P. S., & Lloyd, J. P. 2012, ApJ, 748, 93 Sada, P. V., Deming, D., Jackson, B., et al. 2010, ApJ, 720, L215 Schlaufman, K. C. & Laughlin, G. 2010, A&A, 519, A105 Seager, S. & Mall´en-Ornelas, G. 2003, ApJ, 585, 1038 Skrutskie, M. F., Cutri, R. M., Stiening, R., et al. 2006, AJ, 131, 1163 Southworth, J. 2008, MNRAS, 386, 1644 Valenti, J. A. & Fischer, D. A. 2005, ApJS, 159, 141 van Altena, W. F., Lee, J. T., & Hoffleit, E. D. 2001, VizieR Online Data Catalog, 1238, 0 Vogt, S. S., Butler, R. P., Rivera, E. J., et al. 2010, ApJ, 723, 954 von Braun, K., Boyajian, T. S., Kane, S. R., et al. 2012, ApJ, 753, 171V Zacharias, N., Monet, D. G., Levine, S. E., et al. 2005, VizieR Online Data Catalog, 1297, 0 9 Guillem Anglada-Escud´e et al.: GJ 1214b revised Table 8. Transit time observations used in the orbital fitting BJD (days) Uncertainty (days) Source 2455307.892689 2455353.724652 2455383.752334 2454980.748942 2454983.909686 2454964.944935 2454980.748976 2454983.909689 2454999.713690 2454980.748570 2454983.909820 2454988.650808 2455002.874670 2455269.962990 2455288.928200 2455296.830130 2455315.794850 2455315.794693 2455318.955230 2455353.723870 2455356.884950 2455364.787000 2455375.849970 2455383.752050 2455391.654105 2455315.793430 2455345.821260 2455345.821330 2455353.723320 2455364.786690 0.000263 0.000311 0.000264 0.000417 0.000401 0.000788 0.000264 0.000228 0.000253 0.000150 0.000160 0.000049 0.000190 0.000160 0.001100 0.000230 0.000230 0.000080 0.000170 0.000180 0.000150 0.000150 0.000130 0.000130 0.000059 0.000420 0.000140 0.000370 0.000360 0.000290 1 1 1 1 1 2 2 2 2 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 4 4 4 4 4 Notes. (1)(Kundurthy et al. 2011) (2)(Kundurthy et al. 2011) using re- analized light curves from Charbonneau et al. (2009) (3)(Carter et al. 2011) (4)(Sada et al. 2010) 10
1711.05428	1	1711	2017-11-15T06:53:58	A Model of the H$\alpha$ and Na Transmission Spectrum of HD 189733b	[ "astro-ph.EP" ]	This paper presents a detailed hydrostatic model of the upper atmosphere of HD 189733b, with the goal of constraining its temperature, particle densities, and radiation field over the pressure range $10^{-4}-10\, \mu \rm bar$, where the observed H$\alpha$ transmission spectrum is produced. The atomic hydrogen level population is computed including both collisional and radiative transition rates. The Ly$\alpha$ resonant scattering is computed using a Monte-Carlo simulation. The model transmission spectra are in broad agreement with the data. Excitation of the H(2$\ell$) population is mainly by Ly$\alpha$ radiative excitation due to the large Ly$\alpha$ intensity. The density of H(2$\ell$) is nearly flat over two decades in pressure, and is optically thick to H$\alpha$. Additional models computed for a range of the stellar Lyman continuum (LyC) flux suggest that the variability in H$\alpha$ transit depth may be due to the variability in the stellar LyC. Since metal lines provide the dominant cooling of this part of the atmosphere, the atmosphere structure is sensitive to the density of species such as Mg and Na which may themselves be constrained by observations. Since the H$\alpha$ and Na D lines have comparable absorption depths, we argue that the center of the Na D lines are also formed in the atomic layer where the H$\alpha$ line is formed.	astro-ph.EP	astro-ph	Draft version November 16, 2017 Preprint typeset using LATEX style emulateapj v. 12/16/11 A MODEL OF THE Hα AND Na TRANSMISSION SPECTRUM OF HD 189733b Chenliang Huang ( )1,2, Phil Arras1, Duncan Christie3, and Zhi-Yun Li1 1Department of Astronomy, University of Virginia, Charlottesville, VA 22904, USA; [email protected], [email protected] 2Dept. of Physics and Astronomy, University of Nevada, Las Vegas, NV 89154 and 3Department of Astronomy, University of Florida, Gainesville, FL 32611, USA 7 1 0 2 v o N 5 1 . ] P E h p - o r t s a [ 1 v 8 2 4 5 0 . 1 1 7 1 : v i X r a Draft version November 16, 2017 ABSTRACT This paper presents a detailed hydrostatic model of the upper atmosphere of HD 189733b, with the goal of constraining its temperature, particle densities, and radiation field over the pressure range 10−4−10 µbar, where the observed Hα transmission spectrum is produced. The atomic hydrogen level population is computed including both collisional and radiative transition rates. The Lyα resonant scattering is computed using a Monte-Carlo simulation. The model transmission spectra are in broad agreement with the data. Excitation of the H(2(cid:96)) population is mainly by Lyα radiative excitation due to the large Lyα intensity. The density of H(2(cid:96)) is nearly flat over two decades in pressure, and is optically thick to Hα. Additional models computed for a range of the stellar Lyman continuum (LyC) flux suggest that the variability in Hα transit depth may be due to the variability in the stellar LyC. Since metal lines provide the dominant cooling of this part of the atmosphere, the atmosphere structure is sensitive to the density of species such as Mg and Na which may themselves be constrained by observations. Since the Hα and Na D lines have comparable absorption depths, we argue that the center of the Na D lines are also formed in the atomic layer where the Hα line is formed. Keywords: line: formation – planets and satellites: atmospheres – planets and satellites: individual (HD 189733b) – radiative transfer 1. INTRODUCTION The first detection of an exoplanetary atmosphere was accomplished via measuring the sodium doublet transit signal of HD 209458b (Charbonneau et al. 2002). The at- mosphere of HD 189733b has also been detected by Lyα transit (Lecavelier Des Etangs et al. 2010; Bourrier et al. 2013). The species O I, Na I, and possibly K I (Pont et al. 2013; Jensen et al. 2011) were detected by the Hub- ble Space Telescope (HST) (Ben-Jaffel & Ballester 2013; Huitson et al. 2012; Pont et al. 2013). An indication of an extended atmosphere was also found in X-ray by Chandra (Poppenhaeger et al. 2013). The Hα, Hβ, and Hγ hydrogen lines, and absorption lines from Na I and possibly Mg I were detected in HD 189733b's atmosphere (Jensen et al. 2011, 2012; Cauley et al. 2015, 2016; Red- field et al. 2008; Wyttenbach et al. 2015; Khalafinejad et al. 2017), which shows the promise of ground based telescopes in studying exoplanet atmospheres. An H2O feature has been identified at 3.2µm during the secondary transit (Birkby et al. 2013). The atmosphere of HD 209458b has been modeled in order to compare to the observed H Lyα, O, Si III, and Na I lines (Fortney et al. 2003; Koskinen et al. 2013a,b; Lavvas et al. 2014). For the purpose of studying the Lyα emission spectrum, Menager et al. (2013) calculated the Lyα resonant scattering process in the atmosphere of HD 209458b, based on the atmospheric structure model in Koskinen et al. (2013a), and HD 189733b based on an unpublished model (Koskinen et al. 2011). A sim- ulation of HD 189733b's escaping atmosphere has been performed by Salz et al. (2016). As the hot gas in the upper thermosphere is more weakly bound to the planet, conditions there set the boundary condition for the rate of gas escape (Yelle 2004; Garc´ıa Munoz 2007; Murray-Clay et al. 2009) . Among detected species, Hα is a sensitive probe of the planet's upper atmosphere because the excitation and de- excitation processes for H(2(cid:96)), the absorber of Hα, are strongly dependent on the local particle densities, tem- perature, and radiation field. In addition, unlike Lyα, the interstellar medium is transparent to Hα and this op- tical line can be observed with large ground based tele- scopes. Therefore, the Hα transmission spectrum is a powerful and economical method to probe the structure of the planet's upper atmosphere. As this work shows, the temperature, and hence scale height, in the region optically thick to Hα is (approx- imately) set by a balance of photoelectric heating and line cooling by metal species, mainly Mg I and Na I. If only photoelectric heating and line cooling from hydro- gen were included, the atmosphere would be hotter by (cid:39) 2000 − 3000 K (Christie et al. 2013), giving transit depths far too large in comparison to observations. Fur- thermore, several studies (Garc´ıa Munoz 2007; Koskinen et al. 2013a; Lavvas et al. 2014) suggested that the tran- sition from atomic to molecule hydrogen occurs at pres- sures P (cid:39) 10 µbar. These studies included detailed heat- ing and cooling physics in the molecular layer. But trans- mission spectra for the Na D doublet and Mg lines may in principle provide further constraints on atmosphere models around this transition altitude. In addition, the atmospheric temperature of HD 189733b derived from the Na doublet transmission spectrum by Huitson et al. (2012) and Wyttenbach et al. (2015) is significantly lower than the modeled upper atmosphere temperature found in Salz et al. (2016) and Christie et al. (2013). A model of the Na transmission spectrum is required to understand these contradictory results. Both the HD 189733b Hα transmission spectrum ob- 2 Adopted values for the orbital and physical parameters of HD 189733 and HD 189733b Table 1 Stellar type Star mass Star radius Semi-major axis Planet mass Planet radius Planet surface gravity K2V M(cid:63) = 1.60 × 1033 g R(cid:63) = 5.60 × 1010 cm Mp = 2.17 × 1030 g Rp = 8.137 × 109 cm a = 0.031 AU gp = GMp/R2 p = 2.2 × 103 cm s−2 Note. - Source: exoplanets.org served by Cauley et al. (2016) and the Na D transmission spectrum presented by Wyttenbach et al. (2015) have the spectral resolution to resolve the line core. The line center transit depths of both observations are about 1- 1.5%, which means the line core absorption features of both species are mostly contributed by the same region in the atmosphere. Since the temperature of the molec- ular layer is below 3000 K, and the molecular hydro- gen has a large absorption cross section to Lyα photon (Black & van Dishoeck 1987), both the collisional exci- tation rate and radiative excitation rate are too small to create enough H(2(cid:96)) to absorb the Hα in the molecular layer. These results suggest that the absorption features of both Hα and Na near line center are tracing the atomic layer in the HD 189733b atmosphere. The HD 189733b Hα transmission spectrum was mod- eled by Christie et al. (2013), who constructed a hydro- static atmosphere model similar to the one considered here. In that work, a detailed treatment of Lyα radia- tion transfer was not included, and hence the role of Lyα excitation deep in the atmosphere was not appreciated. Christie et al. (2013) showed that if collisional excita- tion dominates, it would lead to a fairly constant H(2s) density within the atomic layer, because of the combi- nation of increasing temperature and decreasing H(1s) density with radius. In attempting to improve on their model, it was found that Lyα, especially from recom- binations occurring within the atmosphere, could give radiative excitation rate to H(2p) much larger than the collisional excitation rate, and that this excitation could occur deeper in the atmosphere where the H(1s) density is higher, even though the temperature is much lower there. This key insight motivated the detailed Lyα ra- diative transfer treatment in the present work. 2. ANALYTIC ESTIMATES Observations of the Hα transmission spectrum for HD 189733b show a ∼ 1% line center transit depth, and a half-width of ∼ 0.4 A (e.g. Cauley et al. (2015)). Wyt- tenbach et al. (2015) measured a nearly ∼ 1% transit depth for both Na D lines, among which the Na I 5890 A is slightly deeper. This section contains analytic esti- mates for the conditions in the atmosphere required to generate the observed Hα and Na D absorption lines. The outermost reaches of HD 189733b's atmosphere are highly ionized by stellar photons. Moving inward, the radiative recombination rate eventually increases to the point that the atmosphere is dominated by atomic hydrogen, at a pressure level P (cid:39) 10−3 µbar. The tem- perature near the transition from ionized to atomic is regulated to be near T (cid:39) 10, 000 K, which is far too hot for molecular hydrogen to form, and hence there must be a layer of atomic hydrogen extending over the tem- perature range T (cid:39) 2, 500− 10, 000 K. In the hydrostatic model presented in this paper, the temperature in the atomic layer is set by a balance of photoelectric heat- ing and atomic line cooling, for which H2 will dominate at P (cid:38) 10 µ bar. In terms of size, the atomic hydro- gen layer extends over ∼ 10 pressure scale heights, has a mean molecular weight µ (cid:39) 1.3 and mean temperature T (cid:39) 5, 000 K. As compared to the underlying molecular layer, with mean molecular weight µ (cid:39) 2.3 and temper- ature T (cid:39) 1, 000 K, the scale height in the atomic layer is larger by a factor of ∼ 10 compared to the molecu- lar layer, and hence can give rise to absorption to much larger altitudes. The origin of the H(2(cid:96)) population requires a detailed level population calculation. In the present model of the atomic layer, it is found that radiative excitation by Lyα creates a nearly constant H(2(cid:96)) density over ∼ 6 pressure scale heights near the base of the atomic layer. This is the cause of the Hα absorption. The underlying molecular layer is expected to be optically thin to Hα for two rea- sons. First, the density of atomic hydrogen drops rapidly into the molecular layer (e.g. Lavvas et al. 2014), as com- pared to the base of the atomic layer, due to the much lower temperature there. Second, the mean free path to true absorption of Lyα by H2 (Black & van Dishoeck 1987) rapidly decreases as the H2 density increases, and hence the Lyα intensity is expected to drop rapidly in the molecular layer, with an associated decrease in exci- tation to the n=2 state. The measured line center transit depth ∆F/F ∼ 1% requires a certain area to be optically thick. The scale height in the atomic layer is H = kBT µmpg = 1500 km , (1) (cid:18) T (cid:19) 5000 K where the mean molecular weight has been assumed to be µ = 1.3. If N scale heights are optically thick, as compared to the neighboring continuum radiation, this gives an extra absorption depth ∆F F (cid:39) N 2πRpH πR2 (cid:63) (cid:18) N = 8 × 10 (cid:19)(cid:18) T −4N (cid:19) 6 5000 K = 0.5% . (2) (cid:18) T (cid:19) 5000 K Hence the measured line center depth can only be ex- plained by a layer extending many pressure scale heights, and with high temperature T (cid:38) 5, 000 K. The line center optical depth must be greater than unity over the above annulus. As will be shown, n2(cid:96) is nearly constant over a large pressure range. Then for an effective path length 2(cid:112)2N RpH (cid:39) 7 × 109 cm (N/6)1/2(T /5, 000K)1/2 and Hα line center cross section σ0 = 5×10−13 cm2 (5, 000K/T)1/2, the maximum line center optical depth is (cid:16) (cid:17) 104 cm−3 for the fiducial value n2(cid:96) = 104 cm−3. τ0 (cid:39) 35 × n2(cid:96) (3) The line width is mainly set by the temperature and the maximum line center optical depth. For sightlines optically thick to Hα at a line center, the optical depth at distance x = (ν − ν0)/∆νD from line center is τ (x) = τ0 exp(−x2), when Doppler broadening dominates. All frequencies out to x (cid:39) ln(τ0) are then optically thick. In velocity units, the line width is then (cid:112) (ln τ0)1/2 = 9.1 km s −1 (ln τ0)1/2(4) (cid:18) 2kBT (cid:19)1/2 ∆v = mp or in wavelength units ∆λ = 0.2 A (ln τ0)1/2 . (5) Since τ0 (cid:29) 1 for Hα over a large region, the width will be larger than the thermal width. If Lyα excitation is balanced by radiative de- excitation, and (cid:96)-mixing populates the 2s state, the abun- dances relative to the ground state are n2p n1s (cid:39) 3 J12B12 A21 = (cid:18) n2s n1s −8 (cid:39) 10 10−9 erg cm−2 s−1 Hz−1 J12 (cid:19) , (6) where A21 and B12 are the Einstein A and B (absorp- tion) coefficient respectively, following the definition in Rybicki & Lightman (1979). The peak Lyα intensity found near the peak in photoionization of H(1s) near J12 (cid:39) 0.1FLyC/∆νD (cid:39) 10−8 erg cm−2 s−1 Hz−1 is P (cid:39) 10−3 µbar. Here, FLyC (cid:39) 104 erg cm−2 s−1 is the LyC flux deposited in that region, and it is assumed that each ionization is balanced by a recombination produc- ing a Lyα photon. While the ratio of excited state to ground state is high near the peak in Lyα, the H(1s) density there is too small for significant H(2(cid:96)) density. The key question for the H(2(cid:96)) population is how fast the Lyα intensity decreases moving deeper into the at- mosphere. If Lyα intensity does not drop off too fast, the rapid increase in H(1s) density with depth will lead to higher H(2(cid:96)) density deeper in the atmosphere. One can imagine two limits to answer this question. In the first limit, there is a shallow source of Lyα at optical depth τs and the intensity Jν(x, τ ) is desired at τ (cid:29) τs. An analytic solution based on the Fokker-Planck equation given in Harrington (1973) is Jν(x, τ )(cid:39) 0.1 . (7) This expression is valid in the plateau of the intensity near line center. Since Lyα optical depth τ ∝ n1s, this scaling for Jν would give n2(cid:96) ∝ n1sJν (cid:39) constant with depth. The second limit to imagine is where radia- tion is emitted and absorbed locally, which is appropri- ate deep in the atmosphere where τ ∼ 108 scatterings are required to escape the atmosphere. For a constant, frequency-integrated source function S, and true absorp- tion by metal species, the frequency-integrated photon energy equation becomes S (cid:39) nmσmJ, where J is the frequency-integrated intensity, nm is the metal number density and σm is the metal photoionization cross sec- tion at Lyα. For a constant mixing ratio, nm ∝ n1s, and again n2(cid:96) (cid:39) constant. While the scaling found by these two estimates, constant 2(cid:96) density, is the same, it is found that the local balance of sources and sinks is the applicable limit in the present atmosphere model. (cid:18) F0 (cid:19)(cid:16) τs (cid:17) ∆νD τ 3 Up to this point, the estimates have been concerned with the Hα transmission line, however, the center of each line in the Na doublet may also be formed in the atomic layer. Because the cross section of Na I 5890 is larger than Na I 5896 by a factor of 2, the difference in transit radius between the two resolved line centers corre- sponds to ln(2)H, assuming a constant Na I number frac- tion. Keeping in mind the error bars in the measurement, according to Wyttenbach et al. (2015), the difference in transit radius between the Na D lines is ∼ 3000 km, which gives a local scale height H (cid:39) 4300 km. Plugging in the transit radius R(Na) = 9.4 × 109 cm to compute the gravity, assuming µ = 1.3, the temperature derived from the line centers is T (cid:39) 11, 000 K. In order to explain this scale height with a molecular gas, the temperature has to be higher by a factor of 2. But this high tem- perature is inconsistent with the gas being in molecular form. 3. HYDROGEN LEVEL POPULATION Balmer line photons are absorbed by the 2(cid:96) excited states of H. Due to ionization and the subsequent recom- bination cascades, and a radiation excitation tempera- ture different from the gas temperature, the level popu- lations are not set by the Boltzmann distribution at the local gas temperature. Therefore, a study of the H level population over the range of densities, temperatures, and intensities found in hot Jupiter atmosphere is required. The following processes are considered. 1. Hydrogen radiative (de-)excitation of all possible electric dipole transitions between multiplets up to n = 6 (Wiese & Fuhr 2009); 2. Electron collisional (de-)excitation for transitions from 1s to each sub-state (cid:96) up to n = 5, from 2s to each sub-state (cid:96) up to n=5, and from 2p to each sub-state (cid:96) up to n=3 (CHIANTI database, Dere et al. (1997); Del Zanna et al. (2015)); 3. Electron collisional (cid:96)-mixing between 2s and 2p (Seaton 1955); 4. Proton collisional (cid:96)-mixing between 2s and 2p (Seaton 1955), and (cid:96)-mixing for levels 3 ≤ n ≤ 6 (Vrinceanu et al. 2012); 5. Electron collisional ionization and three-body re- combination for each sub-state (cid:96) up to n=4 and total ((cid:96)-unresolved) rates for n =5 and 6 (Janev et al. 2003). The cross sections are assumed to be equal for all sub-states in the same level for n =5 and 6; 6. Photoionization of each sub-state (cid:96) up to n=6 (The Opacity Project 1995). The corresponding recom- bination rate can be calculated using the Milne re- lation. The stellar spectrum of Eri, another K2V star, given by the MUSCLES treasury survey version 2.1 (France et al. 2016; Youngblood et al. 2016; Loyd et al. 2016) is applied in the model. Its energy flux (Fλ) is normal- ized to the distance a = 0.031 AU of HD 189733b from its parent star, shown in Figure 1. The Lyα fluxes are 4 Figure 1. Stellar energy flux spectrum (Fλ) against wavelength (λ) used in the model at orbital distance of HD 189733b. combined with models of solar active regions (Fontenla et al. 2011) to estimate the EUV luminosity in the wave- length region 100 – 1170 A in 100 A bins. The stellar flux attenuation due to bound-bound transitions is not included in this calculation because it is relatively unim- portant in setting the level populations, and performing radiation transfer calculations in addition to Lyα would greatly complicate the analysis. Bound-bound transi- tions due to Lyβ, Lyγ etc. are ignored as it is assumed that these photons rapidly down-convert into a Lyα pho- ton and lower series photons at the large optical depths of interest for these lines. The level population is determined by the kinetic equi- librium between production and loss processes. The equation of rate equilibrium for the state j is αjnenp + βjn2 enp + C (e) k→jnenk (cid:88) (cid:88) (cid:88) k<j C (p) + C (p) k→jnpnk + Ak→jnk + Bk→j ¯Jk→jnk Γj + j→kne + j→knp + C (e) j→∞ne + k Bj→k ¯Jj→k + Aj→k (cid:17) nj, (8) k>j k<j where αj and βj denote radiative and three-body recom- bination rate coefficients for state j, respectively. Case B recombination with α1s ≡ 0 is employed. The nk is the number density of hydrogen in sub-state k. The in- equality k > j denotes a downward transition from a state with principle quantum number of state k larger than that of state j. The rate coefficients C (e/p) j→k are for electron/proton impacts causing a transition from state j to state k, and the state ∞ represents ionization. The rate for proton collisions is only included for the (cid:96)- mixing transitions at fixed principal quantum number. The spontaneous radiative decay rates are Aj→k. The photoionization rate from state j is denoted Γj, and at- tenuation from the overlying gas is included in the ground (cid:88) (cid:16) k = k (cid:88) (cid:88) (cid:88) (cid:88) k k>j C (e) state photoionization calculation. The optical depth for ionizing photons is computed as τatt,ν = σ1s(ν)N1s + σm(ν)Nm, (9) (cid:88) m where Nk(r) =(cid:82) ∞ r dr(cid:48)nk(r(cid:48)) is the column of species k above the layer under consideration. The sum over the subscript m stands for all metal species considered. As discussed in Section 4, this is m =neutral and first ion- ized C, O, Mg, Si, and S, and neutral Na and K. Including the attenuation factor e−τatt,ν, the photoionization rate is (cid:90) ∞ Γj,pi = σj,pi(ν) νj,th −τatt,ν dν, e Fν hν (10) where νth is the corresponding photoionization threshold frequency. The photoionization rates and photoelectric heating rates (see Equation 28 and 31) at τatt = 0 are listed in Table 2. The H(1s) state can also undergo "secondary ioniza- tion" by photoelectrons generated when a photon with much higher energy than the ionization threshold ionizes a hydrogen or metal atom. If ne/n1s is small, photo- electrons can cause H(1s) collisional ionization and ex- citation before sharing their energy with other electrons through Coulomb collisions, which would increase the photoionization rate and reduce the photoelectron heat- ing efficiency. Different from the treatment of constant efficiency applied in Yelle (2004) and Murray-Clay et al. (2009), or the ionization fraction xe = ne/n1s indepen- dent efficiency applied in Koskinen et al. (2013a), an ef- ficiency dependent on local xe and photoelectron energy E = h(ν − νth) is used here. Given the incoming photon frequency ν and ionization threshold energy Eth, Draine (2011) gives the number of secondary ionizations per ion- izing photon, in the case of E > 50 eV and xe < 1.2, to be (cid:18) E − 15 eV 35 eV (cid:19)(cid:18) ke(E) = 1 − xe/1.2 e / ln(E/35 eV) 1 + 18x0.8 (cid:19) . (11) After correcting for the secondary ionizations, the form for the H(1s) ionization rate which is used in the rate equations becomes Γ1s = Γ1st 1s + Γ2nd 1s (cid:90) ∞ = σ1s(ν)(1 + ke) νth −τatt,ν dν, e Fν hν (12) 1s and Γ2nd 1s where Γ1st stand for primary and secondary photoionization rate from H(1s) respectively. Secondary photoionization becomes an important consideration at pressures P (cid:38) 0.1 µbar, where the initially more abun- dant lower energy photons have already been absorbed higher in the atmosphere, and the dominant photons be- ing absorbed can induce at least one secondary ionization on average. Lastly, the bound-bound radiative excitation rates are given by Bl→u ¯Jl→u = gu gl c2Au→l 2hν3 (cid:90) dνJνφ(ν), (13) where ¯Jl→u is line profile weighted mean intensity and 100101102103104105λ(A)10−210−1100101102103104105Fλ(ergcm−2s−1A−1)Photoionization rates and photoelectric heating rates Table 2 Species H(1s) H(2s) H(2p) O I O II C I C II Mg I Mg II Si I Si II S I S II Na I K I Γpi (s−1) 1.61 × 10−3 25.7 21.5 3.06 × 10−3 7.10 × 10−4 7.43 × 10−3 3.90 × 10−4 6.12 × 10−4 2.23 × 10−4 2.18 × 10−2 2.07 × 10−4 2.21 × 10−2 2.30 × 10−4 9.92 × 10−4 2.08 × 10−3 Qpi (erg s−1) 5.91 × 10−15 1.42 × 10−11 1.13 × 10−11 6.14 × 10−14 1.51 × 10−14 5.91 × 10−14 1.07 × 10−14 2.96 × 10−14 2.14 × 10−14 1.02 × 10−14 1.07 × 10−14 1.56 × 10−13 1.04 × 10−14 5.36 × 10−14 2.60 × 10−14 Note. - Atmospheric attenuation, secondary ionization effect and the contribution by Lyα photon are not included. Rate which may be important to n2p population and de-population Table 3 Rates (cm−3s−1) Process 1.7 × 1011(n1s/1010 cm−3) Radiative excitation 1s → 2p 8.4 × 104(nen1s/1019 cm−6) Collisional excitation 1s → 2p 6.2 × 104(nenp/1018 cm−6) Radiative recombination to 2p 6.3 × 1011(n2p/103 cm−3) Spontaneous decay 2p → 1s 1.9 × 108(n2pnp/1012 cm−6) p collisional (cid:96)-mixing 2p → 2s 2.2 × 104(n2p/103 cm−3) 7.4 × 103(n2pne/1012 cm−6) Collisional de-excitation 2p → 1s 7.4 × 102(n2pne/1012 cm−6) Collisional ionization from 2p Note. - In the table, ¯JLyα = 2 × 10−9 erg cm−2 s−1 Hz−1 and T =8000 K are applied. The reference numbers are n1s = 1010 cm−3, ne = np = 109 cm−3 and n2p = 103 cm−3. Photoionization from 2p φ(ν) is the Voigt profile. Equation 8 is evaluated for all 0 ≤ (cid:96) ≤ n − 1 and 1 ≤ n ≤ 6, resulting in a linear system of 21 equations in total for the number density of each (n, (cid:96)) state, nn(cid:96). The quantities ne, np, and T are treated as given param- eters in the equations. The linear system is solving using Gauss-Jordan elimination (Press et al. 2007). The rates of important processes related to the 2p state population are listed in Table 3. The rates related to higher excited states are not listed in the table because they cannot have a large net effect on 2p in steady state, unless the higher state itself has a large source or sink, which is not the case. The stellar Lyα mean intensity (see Section 5) is applied for the estimate. The table shows that all other rates except the (cid:96)-mixing rates between 2s and 2p are negligible compared to the radiative rates between 2p and 1s. The proton collisional (cid:96)-mixing rate is much larger than the electron rate. The (cid:96)-mixing rates between 2s and 2p nearly cancel each other and 2(cid:96) states are in collisional equilibrium due to the large (cid:96)-mixing rates at the densities of interest. Thus, A2p→1sn2p and B1s→2p ¯JLyαn1s completely dominate the 2p generation and destruction rate, and the n=2 state number densities are simply set by ¯JLyα, so we have, n2p ≈ 3n2s ≈ n1s gu gl c2 ¯JLyα 2hν3 . 5 (14) In Section 6, it will be shown that this is a good ap- proximation for the whole simulation region. To obtain the intensity of Lyα, a resonant scattering study of Lyα photon will be discussed in Section 5. 4. THE ATMOSPHERE MODEL 4.1. Basic Structure et al. (2013), Following Christie a spherically- symmetric, hydrostatic atmosphere model is constructed for the region composed of ionized and atomic gas sit- ting above the molecular atmosphere. The transit radius measured in broadband optical wavelengths is Rp, and the base of the atomic layer is at radius Rb > Rp. The thickness of the molecular layer below the atomic layer is then Rb − Rp. Assuming an isothermal molecular layer with equilibrium temperature T = Teq = 1140 K (Wyt- tenbach et al. 2015), the thickness is Rb − Rp ≈ kBTeqR2 p µmHGMp ln , (15) (cid:18) Pp (cid:19) Pb where µ ≈ 2.3 is the mean molecular weight, Pp = 1 bar is the pressure of the optical photosphere suggested by Sharp & Burrows (2007). It is assumed that the pres- sure at the base of the atomic layer is Pb = 10 µbar. It will be shown that the temperature in the atomic layer becomes small enough for molecular hydrogen to domi- nate there, for the assumptions used here. 4.2. Differential Equations Given the temperature and number density of each species at one level in the atmosphere, the equation of hydrostatic balance and equations for the column of each species must be integrated inward to find pressure and columns at the next step inward. The hydrostatic bal- ance equation is r2 and the columns are integrated as ρGMp dP dr = − dNi dr = −ni. (16) (17) The subscript i stands for each species, including H(1s) and the individual neutral and first ionized metal element considered. The ideal gas law for the gas pressure is where fz = (cid:80) fm is the sum of metal species relative P = (ne + (np + nH)(1 + fz + fHe))kBT, (18) number density abundance to hydrogen, and fHe is the fraction of He by number assuming solar abundance (As- plund et al. 2009). Ionization of He is ignored in this paper. The gas density is written (cid:88) ρ = (np + nH)(1 + 4fHe + mmfm)mp, (19) where mm is the metal atomic mass in atomic units. The pressure and the column density are integrated in- ward, with a starting value Ptop = 5 × 10−5µbar on the 6 outside, where the atmosphere above becomes optically thin to Lyα. The starting value of each Ni = 0. The solution is integrated inward from a starting radius Rtop and pressure Ptop, until the base radius Rb is reached. The boundary condition imposed there is that P = Pb. This boundary condition is satisfied in practice by vary- ing the starting radius Rtop until P (Rb) = Pb, the desired value, using Brent's method (Press et al. 2007). The hydrostatic model will be inaccurate near the outer boundary, as a number of physical effects have been neglected, such as: outflowing gas from the planet, inter- action with the stellar wind, strong magnetic forces, radi- ation pressure, and stellar tidal forces. The region where these effects may be appreciable will be estimated in Sec- tion 8.1. However, in the region where this model shows the dominant absorption by H(2(cid:96)), the density is so high that these effects are negligible. Hence the hydrostatic model is sufficient for the purposes of this study. 4.3. Ionization State and Temperature At each level of the atmosphere, the pressure P and columns Ni are given by the boundary conditions or the integration of Equations 16 and 17. The temperature and particle densities must then be updated to continue the integration. Since the gas is not in local thermody- namic equilibrium (LTE), these quantities must be de- termined by solving rate equilibrium equations for ion- ization/recombination, heating/cooling, a charge balance equation, and the equation of state. The equations used are as following. Terms related to metal species will be discussed in more detail in Section 4.4. The charge balance equation is (cid:88) (cid:88) np + nM II + 2 nM III = ne, (20) where nM II and nM III are the number density of first ionized and second ionized metal species respectively. Higher ionization states are ignored as their abundance would be negligible for the given ionizing flux and parti- cle densities. The hydrogen ionization and recombination balance equation is (cid:88) (αBne +k(O) + C (e) ion nO I)np = n1s(Γ1s +C (e) 2→∞nen2 + Γ2pn2p + Γ2sn2s + 1s→∞ne +k(O) m nm, Γ2nd rec nO II) (21) m where αB is the case B recombination rate, which is a good approximation for region deeper than 3×10−3 µbar where the atmosphere is optically thick to Lyman con- k(O) tinuum photons near the ionization threshold. ion and k(O) rec are rates of whichoxygen ionizes and recom- bines through charge exchange with hydrogen respec- tively. The n=2 state has separate contributions from H(2s) and H(2p) as C (e) 2→knen2 = C (e) 2s→knen2s + C (e) 2p→knen2p. (22) The last term in Equation 21 represents ionization from the H(1s) state due to photoelectrons created by metal ionization. Hence high energy photoelectrons created through ionization of metal species can have the same secondary ionization effect as Equation 11. The sec- ondary ionization rate due to metal species is then Γ2nd m = σm,pi(ν)ke νth −τatt,ν dν, e Fν hν (23) (cid:90) ∞ where σm,pi(ν) is the metal photoionization cross section. When evaluating the excited state H abundance in Equa- tion 21, the approximation in Equation 14 is applied. The heating and cooling balance equation is (cid:104)(cid:0)13.6 eVC (e) 1s→∞ + 10.2 eVC (e) + 3.4 eVC (e) 2→∞n2 + Λmnm + Λf f np (cid:1)n1s m 1s→2 (cid:88) (cid:88) (cid:88) m (cid:105) + (cid:104)Err(cid:105)αBnp + kBT αmnion m ne (24) = n1sQ1s(N1s) + Q2pn2p + Q2sn2s + 10.2 eV nen2C (e) 2→1 + Qmnm, where Λ stands for cooling function. Osterbrock & Fer- land (2006) give the free free cooling rate m Λf f = 1.85 × 10 −27 T 1/2( erg cm3 s −1), (25) where T is in units of Kelvin. The mean kinetic energy of the recombining electrons is (Draine 2011) (cid:104)Err(cid:105) = [0.684 − 0.0416 ln(T4)]kBT, where T4 = T /104 K. In Equation 24, the symbol Q rep- resents the photoelectric heating rate, per photoioniza- tion, corrected for the secondary ionization effect. Dal- garno et al. (1999) find that secondary electrons give rise to approximately the same number of ionizations as 1s→2p excitations, so the heat deposited into the atmo- sphere by one photoelectron with energy E is taken to be (26) η(E) = E − (13.6eV + 10.2eV)ke, (27) where the second term represents the energy lost by the photoelectron to ionizations and Lyα excitations. Thus, the net photoelectric heating rate is (cid:90) ∞ Q = ησpi(ν) νth −τatt,ν dν. e Fν hν (28) Given P and the columns Ni, Equations 18, 20, 21, and 24 give four algebraic equations to solve for T , ne, np, and n1s at this level in the atmosphere. A globally convergent Newton's method (Press et al. 2007) is applied to solve the set of equations. 4.4. Radiative Cooling Due to Metal Species Although H and He are by far the most abundant ele- ments, their electron-impact line cooling rates are heavily suppressed at temperatures T (cid:46) 104 K due to the high excitation energies. Metal line cooling due to electron impact followed by radiative de-excitation is an impor- tant coolant, especially near the base of the atmosphere at T (cid:46) 8000 K. The ionization/recombination rate equi- librium equation is included to determine the relative abundance of each ionization state. Transitions yielding large cooling rates are chosen from abundant elements, and by striking a balance between low excitation ener- gies, ∆E, and large radiative decay rate Aul. Solar abun- dance is assumed (Asplund et al. 2009). The elements considered are O, C, Mg, Si, S, Na, and K. Although Mg was not a priori expected to be abundant in the upper atmosphere due to condensation (Visscher et al. 2010; Koskinen et al. 2013b), Mg I is in fact detected in HD 209458b (Vidal-Madjar et al. 2013) and marginally de- tected in HD 189733b (Cauley et al. 2016). The abundance of each ionization state is set by solving for rate equilibrium between ionization and recombina- tion. Only neutral, first, and second-ionized atoms are included. As a special case, rather than photoionization, colli- sional, and radiative recombination, the ionization state of oxygen is determined by charge exchange with hydro- gen. Considering that nH (cid:38) 105 cm−3 everywhere in the model, the charge exchange rates in the high-density limit in Draine (2011) are applied. The energy differences between three fine-structure levels of neutral oxygen are ignored because they are much smaller than kBT . The rate equilibrium equations are (ΓM I + C (e) M I,∞ne + k(O) ion np)nM I = (αM IIne + k(O) ΓM IInM II = αM III nenM III nM I + nM II + nM III = (nH + np)fm, rec n1s)nM II (29) where C (e) M I,∞ is the electron collisional ionization rate, which is only considered for Na and K atoms. The col- lisional ionization of other metal species are ignored be- cause of the much higher ionization potential. The sec- ondary ionization states of Na and K are ignored. The photoionization rates of all species from Verner et al. (1995, 1996). The Na and K collisional ionization rates are given by Lennon et al. (1988). The rates of Pequignot et al. (1991) are used to de- scribe the recombination of C and O, of Shull & van Steenberg (1982) for that of Mg, Si, and S, of Verner & Ferland (1996) and Landini & Fossi (1991) for that of Na and K respectively. Because the ionization potentials of Mg I, Si I, Na I, and K I are smaller than the Lyα energy, both contin- uum and Lyα photons contribute to their photoioniza- tion and photoelectric heating rates. The metal species photoionization rate is where the first integral excludes the stellar flux contribu- tion near the Lyα line, and the metal species photoelec- tric heating rate Qm,pi = ησm,pi(ν) νth −τatt,ν dν e Fν hν + 4πσm,pi(νLyα) νLyα − νth νLyα (cid:90) Jνdν, (31) Γm,pi = σm,pi(ν) νth Fν hν + −τatt,ν dν e 4πσm,pi(νLyα) hνLyα (cid:90) Jνdν, (30) (cid:90) ∞ (cid:90) ∞ 7 the intensity in the vicinity of the Lyα line, found as a result of the resonant scattering calculation in Section 5. The metal line cooling rates require a model for the excited state densities. For a two-level system, rate equi- librium between upward and downward rates gives nl(neC (e) lu + Blu ¯Jlu) = nu(Aul + neC (e) ul ). (32) Stimulated emission is ignored due to dilution of the stel- lar flux. Collisional excitation is a sink of thermal trans- lation energy while collisional de-excitation is a source. Thus the cooling rate of this two levels system is ∆E ne(C (e) (cid:32) lu nl − C (e) = nenl ∆E C (e) lu ul nu) Aul − Blu ¯Jlu(nl/nu)eq Aul + neC (e) ul (cid:33) ≡ Λnenl, (33) where (nl/nu)eq = gl/gu exp(∆E/(kBT )), and the last equality defines Λ(T ), the cooling function. Permitted transitions in the optical and near UV bands are always associated with strong absorption features in the stellar spectrum, which lead to a very small radiative excitation rate Blu ¯Jlu(nl/nu)eq for the transitions used. In the case of forbidden transitions, to compensate for the small Aul, only small ∆E transitions give rise to significant cooling. At the long wavelength end, the dilution of radiation flux due to the solid angle of the star overcomes the effect of higher brightness temperature of the star. As a result, radiative excitation is negligible in Equation 33 for both permitted and forbidden transitions. For forbidden transitions, the electron number density ne is much larger than the critical density ncrit above which collisional de-excitation dominates radiative de- excitation. In this limit, the level population is given by the Boltzmann distribution, and the cooling rate be- comes (cid:18) nu (cid:19) Λnenl = ∆EnlAul nl , eq (34) which is independent of ne and collisional rates. The emitted metal line photons are assumed to escape freely from the atmosphere. In reality, the atmosphere may be optically thick to permitted emissions near the base of the atomic layer. The major cooling processes are listed in Table 4, while Table 5 in the appendix contains addition transitions from O, C, S, and Si lines which are included in the model but only have a minor effect on the temperature. The lower state of some transitions may not be the ground state. In this case, the lower state is assumed to reach collisional equilibrium with ground state in the cooling rate calculation. Mg I 5184 A is another line that may contribute to cooling. The lower state of this transition is not the ground state, and the upper state is associated with the ground state of the strong resonance line Mg I 2852 A. A Mg level population calculation and Mg I 2852 A radia- tion transfer study are required to accurately model the cooling effect, which is beyond the scope of this work. Mg I 5184 A is not included in this work as a result. where σm,pi(νLyα) is only nonzero for Mg I, Si I, Na I, and K I. The mean intensity Jν in these formulas denotes 4.5. Molecular Hydrogen 8 Transition Mg I 2852 Mg I 4571 Mg II 2803 Mg II 2796 Na I 5890+5896 K I 7665+7699 ∆E (eV) 4.35 2.712 4.422 4.434 2.104 1.615 List of major metal cooling transitions Table 4 a Aul (s−1) 4.91 × 108 254 2.57 × 108 2.60 × 108 6.16 × 107 3.78 × 107 b ncrit (cm−3) 6.7 × 1015 8.5 × 109 8.6 × 1014 2.2 × 1014 4.4 × 1014 1.8 × 1014 Λ ( erg cm3 s−1) 3.4 × 10−19T 0.18e−5.04/T4 c 1.0 × 10−16e−3.15/T4 /(254 + 3.0 × 10−8ne) 7.1 × 10−12C(e) 7.1 × 10−12C(e) 3.4 × 10−12C(e) lu lu 3.7 × 10−19T 0.18e−1.87/T4 lu Source of collision rate Van Regemorter formula d Osorio et al. (2015) CHIANTI e CHIANTI Igenbergs et al. (2008) Van Regemorter formula (2015) aKramida et al. bncrit = Aul/C(e) cT4 = T /104 K dvan Regemorter (1962) e Dere et al. (1997); Del Zanna et al. (2015) ul (4000 K) Near the base of the atomic layer, where the temper- ature drops below T ∼ 2000 − 3000 K, it is expected that the density of molecules will increase rapidly and come to dominate over the atomic species. This is the base of the atomic layer and the top of the molecular layer. The atomic-to-molecular transition is not self- consistently modeled in this work, as the rate equations to determine molecular densities (e.g. Yelle 2004; Garc´ıa Munoz 2007), and the strong effect of molecular coolants from e.g. H2O rotation-vibration bands, are not taken into account. Although the details of molecule formation are beyond the scope of this paper, a rough estimate of the H2 num- ber density is made to verify that the temperature does indeed become low enough to form molecules as the base is approached. Lenzuni et al. (1991) noted that for a wide range of radiation intensity, that dissociation of H2 is due to collisional processes, rather than photo-dissociation. Yelle (2004) found the same result for a model for the thermosphere of HD 209458b. In this limit, an estimate of the H2 density can be found using the Saha equation. Applying the H2 partition function from Borysow et al. (1989), the H2 number density, nH2 can be computed from the atomic hydrogen density, nH, and the temper- ature T . The buildup of a significant column of H2 will shield the lower atmosphere from stellar UV, and allow the for- mation of other molecules, e.g. CO and H2O which may be important coolants. Another effect more relevant to this work is that H2 may act as a strong absorber of Lyα photons, effectively setting a lower boundary to the region of the atmosphere that may have a large Lyα intensity. Black & van Dishoeck (1987) discussed nu- merous "accidental resonances" between Lyα and elec- tronic transitions in H2. The strongest of these can have oscillator strength f ∼ 10−2, implying that a column NH2 (cid:39) 1014 cm−2 is required to give unit optical depth for these transitions. For a scale height H (cid:39) 108 cm, this gives a critical number density nH2 (cid:39) 106 cm−3 for true absorption optical depth unity over a scale height. In practice, Lyα photons near the optically-thick base of the atomic layer would require a large number of scatter- ings to escape the atmosphere, implying a total distance traversed even larger than a scale height. This will ef- fectively set an absorbing lower boundary for the atomic layer in the radiation transfer described in Section 5. 5. Lyα RADIATION TRANSFER Lyα photons can excite hydrogen from the 1s to the 2p state, providing a population of absorbers that may be detected in Balmer line transmission spectra. Lyα may also play a role in the heating/cooling and ioniza- tion/recombination balance, so a detailed Lyα radiation transfer calculation is crucial. Two sources of Lyα are included in the model, the stellar Lyα incident through the top of the planet's at- mosphere, and also Lyα generated within the planet's atmosphere by electron impact excitation or a recom- bination cascade. The results of particle densities and temperature versus radius from the hydrostatic model in Section 4 are used to specify the Lyα source function, as well as the mean free paths to scattering and true ab- sorption. In the transfer calculation it is convenient to use Lyα line center optical depth, τ , as the vertical coor- dinate. A plot of pressure P and radius r versus τ will be shown in Figure 3. A major simplifying assumption is to use plane-parallel geometry, so that mean intensity Jν(r) is tabulated as a function of altitude. The radius in the hydrostatic model varies by a factor of 2 from base to top, and by a factor of (cid:39) 20% near the base of the layer where the Hα line is formed. The plane-parallel assumption simplifies the calculation, and is consistent with uniform irradiation assumed in the hydrostatic model. At the outer boundary (τ = 0), the unpolarized stel- lar Lyα intensity enters the slab vertically. The line is parameterized by a double Gaussian line profile with width (in velocity units) σ = 49 km s−1 and centers at µ = ±74 km s−1 (Gladstone 1988; Curdt et al. 2010; Instead of using the value of Eri Tian et al. 2009). from MUSCLES, the integrated line flux at the top of the HD 189733b's atmosphere given by Linsky et al. (2013), F0 = 2.0 × 104 erg cm−2 s−1, is applied. Outgoing pho- tons can escape from the top boundary freely. An absorption bottom boundary is applied at the base of the slab, which represents the base of the atomic layer. Physically, this boundary condition is imposed to repre- sent the short mean free path to true absorption of Lyα on H2 (Shull 1978; Black & van Dishoeck 1987) and H2O (Miguel et al. 2015). Since nH2 increases inward much faster than nH, the mean free path to resonant scatter- ing will become longer than that to true absorption in the molecular layer, greatly decreasing the Lyα intensity compared to the atomic layer. As will be shown, even in the atomic layer, the intensity falls rapidly toward the base. Lyα generated inside the atmosphere by collisional ex- citation or a recombination cascade is assumed to be un- polarized and have an initial frequency distribution given by a Gaussian distribution with the Doppler width set by the local temperature. In practice, this initial distribu- tion is much narrower than the resultant mean intensity, so it is the same effect as initializing photons at line cen- ter. The source function inside the atmosphere is (cid:16) Sν = 10.2 eV 4π αBnpne + C (e) 1s→2n1sne + Γ2nd 1s n1s + Γ2nd m nm (cid:88) (cid:17) φν, (35) m where φν is the Doppler profile evaluated at the local temperature. The first term in Equation 35 represents recombinations, each of which is assumed to produce one Lyα photon. The following three terms represent col- lisional excitation by thermal electrons, photoelectrons from ionizing H(1s), and photoelectrons generated from ionizing metals. Recombination and collisional excita- tion to H(2s) are also included because H(2s) and H(2p) are well coupled by the (cid:96)-mixing. The line center optical depth to scattering reaches val- ues as large as τ (cid:39) 108 near the base of the model, and photons will scatter ∼ τ times before exiting the atmosphere (Harrington 1973) if no other process inter- venes. Lyα photons can leave this scattering cycle by two categories of processes. First, the radiative excitation may not be followed by radiative de-excitation of a Lyα photon some fraction of the time. The dominant pro- cesses are: photoionization from the n=2 state by stel- lar Balmer continuum emission; collisional de-excitation; and two photon decay from 2s→1s. The former two pro- cesses also contribute to ionizing and heating rates. The second kind of process is "true absorption", in which a Lyα photon is absorbed by some other species besides H(1s), for example by photoionizing an atom with low ionization potential less than 10.2eV. Lyα photons can also leave this scattering cycle by the (cid:96)-mixing or radiative excitation processes from the 2p state. Because it will be followed by the reverse pro- cess immediately, these processes are equivalent to Lyα photon redistribution that put the line wing photons fre- quently back to the line center and stop the photons from escaping. These processes are not included in this model because a careful consideration required the Hα mean intensity in the atmosphere, which is not available right now. The plane-parallel transfer equation including resonant scattering, true absorption, the source function for pho- ton creation, and excitation followed by de-excitation a fraction 1 − of the time is µ dIν(z, n) dz = −(αsc + αabs)Iν(z, n) + Sν(z) (cid:48) αsc )dΩ φν )Iν(cid:48)(z, n(cid:48) , n(cid:48) (cid:48) R(ν, n; ν + (1 − )4π where µ = cos θ, φν is the Voigt line profile, αsc is reso- nant scattering coefficient, αabs is true absorption coef- ficient, and R is the Hummer (1962) case II-B redistri- (36) dν , (cid:48) (cid:90) 9 bution function, with dipole angular dependence. The redistribution function R gives the probability that the photon is scattered from incident frequency ν(cid:48) and direc- tion n(cid:48) to frequency ν and direction n. Case II-B redis- tribution assumes that, in the rest frame of the atom, the line profile for absorption of the photon (excitation) is a Lorentzian profile, and that in the rest frame the outgo- ing photon has the same energy as the ingoing photon. Hummer (1962) presents formulae for the resulting redis- tribution function thermally averaged over a Boltzmann distribution of atom velocities. Case II-B with dipole angular dependence results for resonant scattering when the initial and final states are the H(1s) state, and the fi- nite lifetime of the intermediate H(2p) state is included. Fine structure splitting of the excited state is ignored. This is a good approximation in the present application where the mean intensity is much broader than the fine structure separation of the J = 1/2 and 3/2 states. The transfer equation is solved numerically with the Monte Carlo method (e.g. Whitney 2011). First, unpo- larized photon packets are initialized at a point randomly generated from the source function (see Equation 35) and with a randomly chosen direction. Second, the optical depth τ that the photon will travel through before it is scattered or absorbed is randomly generated with a prob- ability density e−τ . The spatial location of the scattering or "true absorption" at optical depth τ from the point of emission is then determined with the knowledge of the densities and temperature, and cross sections of nH, nMgI, nSiI, nNaI, nKI. The Lucy method (Lucy 1999) is used to tabulate intensity and flux from the motion of the photon packets. This was crucial in optically thick regions, and worked much better than accumulating pho- ton statistics only when they pass through the face of a cell. Based on the radiative excitation and absorption optical depth, the rejection method is used to determine whether the photon is absorbed. Whether the H(2p) emits another Lyα photon is determined by comparing a random number with 1 − . Third, at each scatter- ing, the outgoing photon direction is chosen including polarization accumulated during prior scatterings. The Stokes matrix for Rayleigh scattering is used. However, rather than using the Stokes matrix in scattering-plane coordinates, as discussed in Chandrasekhar (1960), a scattering matrix was derived in terms of the incoming and outgoing photon direction without reference to the scattering plane, which was found to be more convenient for numerical calculations. Then, the velocity of scatter- ers along the direction of the incident photon is randomly generated by the method described in Zheng & Miralda- Escud´e (2002), with small modifications to improve the efficiency. Recoil is included in computing the new fre- quency of the photon after scattering. The process of propagation and scattering is repeated until the photon escapes the modeled system or leaves the scattering cy- cle. By knowing the n1s from the hydrostatic atmosphere model and ¯JLyα from the Lyα radiation transfer calcu- lation, the n2(cid:96) population is given by Equation 14. Note that the ionization, heating, and cooling rates in the at- mosphere depend on the n2(cid:96) and Lyα mean intensity. Therefore, the hydrostatic model and radiation transfer calculation are performed iteratively, until the n2(cid:96) con- verged, which takes typically ∼ 8 iterations. 10 Figure 2. Lyα mean intensity spectrum (Jν ) at six different depths in the atmosphere. At line center the spectrum is flat, much wider than the thermal line width, and becomes broader with depth. The intensity decreases rapidly on the line wing. The fluctuations in the spectra are due to the statistical noise in the Monte Carlo simulation. Figure 3. Temperature (T ) and Lyα line center optical depth (τ ) versus pressure (P , bottom axis) and radius (r, top axis). 6. THE FIDUCIAL ATMOSPHERE MODEL In this section, results are presented for the fiducial model with solar abundance for all species and extreme ultraviolet (EUV) and X-ray flux set by the day-side value. The Lyα mean intensity spectrum in the atmo- sphere is shown in Figure 2. Temperature versus height for the fiducial model is shown in Figure 3. The heating and cooling rates per unit volume are given in Figure 4, and the number densities of each species are given in Fig- ure 5. Figure 6 shows the Lyα photon sources, sinks, and H ionization rates per unit volume. Line profile weighted Lyα mean intensity ¯JLyα of the fiducial model are given in Figure 7. Figure 2 shows the Lyα mean intensity at six differ- ent depths in the atmosphere. The line center of the resonant scattering spectrum has a flat part with width ∼ (aτ )1/3νD (Harrington 1973). The dip at line center near the surface arises because the photons have to dif- fuse away from line center in order to escape from the slab, due to the extremely short mean free path at line At the top of the model, center. The mean intensity near line center has a peak around τ ∼ 104, where the H(1s) photoionization, as well as the subsequent recombination, which is the major Lyα photon source, is the strongest. The spectrum decreases rapidly on the line wing. The fluctuations in the spec- tra are due to the statistical noise in the Monte Carlo simulation. Since the number densities of H(2(cid:96)) are pro- portional to the ¯JLyα, this fluctuation also appears in the H(2(cid:96)) number densities curve, and other curves such as heating rates. for pressures P < 3 × 10−3 µbar, the gas is fully ionized and the contribution to the number density of electrons by metal species is negli- gible. The gas is optically thin to LyC photons, thus the ionization rate and heating rate per particle are nearly eαB/Γ1s ∝ P 2. The constant. gas temperature is set by the balance of H(1s) photo- electric heating and line cooling by Mg II for solar Mg abundance or Lyα for low Mg abundance. The cooling rate nMgIIneΛMg ≈ n1sQ1s ∝ P 2. Because the nMgII increases faster than P (see Figure 5), the gas temper- ature is regulated to T ∝ 1/ log P near temperatures T = 9000 − 13, 000 K, shown in Figure 3. Note that the ionized region of the hydrostatic model at P (cid:46) 3 × 10−4 µbar may have an unphysically high In this region n1s ≈ n2 temperature, as the inclusion of a hydrodynamic outflow and adiabatic cooling may be important in this region, as in the well-studied case of HD 209458b (e.g. Yelle 2004). However, inspection of Figure 5 shows that the H(2(cid:96)) and Na I densities are negligible in this region, and hence errors in the temperature profile there will not affect the Hα and Na transit depth. H(1s) is the main absorber of the stellar LyC flux over the majority of the energy range. Besides H(1s), the 2s- shell of O I absorbs most photons with energy above 538 eV, and C and Si are the main absorbers of the pho- tons below 13.6 eV in the atomic layer. The atmosphere becomes optically thick to 400 A photons at the pres- sure of ∼ 5× 10−3 µbar. The strong stellar flux between 100 and 400 A causes the local maximum in the T − P profile. Photoelectric heating from ionization of H(1s) contributes the large peak in Figure 4 over the pressure range P = 10−3 − 10−1 µbar. Below that, it continues to be an important source of heating, with absorption of successively higher energy photons with depth, and ionization by secondary electrons becoming important. There is a narrow region near P ∼ 1 µbar where heat- ing due to electron impact de-excitation of H(2(cid:96)) domi- nates. Ultimately the heating in this region is due to the Lyα radiation, which excites the atoms to the n=2 state. Near the base, at P = 1 − 10 µbar, metal photoelectric heating from ionization of Si, O, and C dominates the heating. Na line cooling is the dominant coolant below P = 0.5 µbar. Above that, assuming solar Mg abun- dance, Mg line cooling dominates, among which Mg I 4571 contributes most at P > 0.4 µbar while Mg II contributes most above. Lyα line would be the major coolant instead if the Mg abundance is low. Hence near the base of the model, both heating and cooling are controlled by metal species, and the tem- perature is not sensitive to an overall shift in metallicity. Above P = 1 µbar, the temperature is sensitive to metal- licity, and eliminating a single important coolant could v (km/s)400-300-200-100-0100200300400)-1 Hz-1 s-2(erg cmnJ11-1010-109-108-10barmP=5.1e-05 barmP=3.8e-04 barmP=2.9e-03 barmP=6.9e-02 barmP=0.51 barmP=1.8 10−510−410−310−210−1100101102P(µbar)2000400060008000100001200014000T(K)Tτ10−210−1100101102103104105106107108109τ1512109.28.78.4r(104km)11 Figure 4. Major heating (Qn, solid lines) and cooling (Λnen) rates versus pressure (P ). The line cooling profiles (dashed lines) present the radiative cooling of Lyα and the metal lines listed in Table 4. make a difference, with Mg and Na-poor atmospheres ex- pected to be hotter and more extended. Since this region is important for the Hα and Na transmission spectrum, the temperature, with it's dependence on the metallic- ity, may be an important parameter in understanding the transmission spectrum. Lyα line center optical depth versus height for the fidu- cial model is also shown in Figure 3. Near the base of the atomic layer, τ ∝ P ∝ n(H(1s)), since ground state hydrogen is the dominant species. However, above P = 2 × 10−3 µbar, hydrogen is predominantly ionized, and the optical depth decreases outward approximately as τ ∝ n(H(1s)) ∝ n2 p ∝ P 2 in ionization equilibrium. At the outer boundary of the model, τ drops abruptly to zero due to the τ = 0 boundary condition there. The most abundant species in Figure 5 are electrons and protons above P = 5 × 10−3 µbar, and H(1s) be- low. Recall that ionization of He is ignored in this pa- per, and that He is assumed to be neutral and have solar abundance. The electron number density stays nearly constant in the deeper part of the atomic layer because both the ionization and the recombination rates are in- sensitive to altitude and temperature. The ionization is dominated by the photoionization from H(2(cid:96)) in this region and a flat H(2(cid:96)) number densities lead to a flat ionization rate (see Figure 6). From the near equality ne (cid:39) np, ionization of hydrogen supplies most of the electrons down to 1 µbar, and first ionized Mg and Si are important below. The ionization state of O closely fol- lows that of H because of the very large O and H charge exchange rate. The atmosphere becomes opaque to the stellar flux above 10.4 eV due to the absorption by S I and C I. The atmosphere is transparent to stellar flux below the S I ionization threshold throughout the model. The ionization of Na I and K I is dominated by collisional ionization at the level above 0.1 µbar. Near the base of the model, the density of H2 rises rapidly, and is only slightly less abundant than H(1s). In a more complete model it is expected that the inclusion of strong molecular cooling due to e.g. H+ 3 and H2O would cause the temperature near the base to be even lower and nH2 to be even larger. The combination of large Lyα intensity ¯JLyα ∝ P −1 (see Figure 7) and increasing H(1s) density with depth gives rise to an approximately flat H(2(cid:96)) densities around 104 cm−3 over two decades in pressure near the base of atomic layer. The number densities of Na I and H(2(cid:96)) are similar in the pressure region between 10−2µbar and 10−1µbar. In Section 7, it will be shown that τ ∼ 1 for Na D and Hα in this pressure region. This leads to similar transit depths for the Hα and Na D transmission lines, in agree- ment with the observations of Cauley et al. (2016) and Wyttenbach et al. (2015). Shown in Figure 6, the photoionization rate from H(1s) dominates the H ionization rate in the top layer of the bar) mP (4-103-102-101-10110)-1 s-3 (erg cm ne nL & Q n 8-107-106-105-10H(1s) photoelectric heatingH(n=2) photoelectric heatingMetal photoelectric heatingH(n=2) collisional de-excitation heatingFree free coolingH line coolingH recombination coolingC recombination coolingSi recombination coolingMg line coolingNa line coolingK line cooling12 Figure 5. Number density (n) of main species against the pressure (P ). The solid lines present the profiles for neutral atoms and the dashed lines for first ionized ions. The combination of large Lyα intensity ¯JLyα ∝ P −1 and increasing H(1s) density with depth give rise to an approximately flat H(2(cid:96)) density around 104 cm−3 over two decades in pressure near the base of thermosphere. The number densities of Na I and H(2(cid:96)) are comparable in the pressure region between 10−2µbar and 10−1µbar, where the line center optical depths of Hα and Na D are near unity (see Figure 11). main, is FLyC = 2.6 × 104 erg cm−2 s−1. Besides the 2.0 × 104 erg cm−2 s−1 stellar Lyα flux that mostly re- flects back out, a net 9.6 × 103 erg cm−2 s−1 flux leav- atmosphere, and becomes constant after the atmosphere becomes optically thick to the LyC photons. The pho- toionization from H(2(cid:96)) takes over for the region P (cid:38) 10−1 µbar. The rate of charge exchange between O and H can be very high, but they almost cancel each other and leave a small net effect. The collisional ionization by secondary e generated by the photoionization of metals ing the atmosphere originated from the Lyα emission inside the atmosphere. In comparison, integrating the source function in Equation 35 over height, solid angle and frequency, the total column Lyα internal emission is 1.6 × 104 erg cm−2 s−1. And the flux of Lyα that hits the bottom boundary is 44 erg cm−2 s−1. ((cid:80) Because the photoionization of an H(2(cid:96)) is followed by a radiative recombination cascade, emitting another Lyα photon, these two processes taken together can be thought of not as sources or sinks, but as a redistribu- tion in photon energy. The absorption of a photon on the wing, and its subsequent re-emission at line center make it harder for the photon to escape the atmosphere. The most important remaining "real" photon sources are excitation by secondary e, and radiative recombination cascade of a p which was collisionally ionized by high energy photoelectron. The "real" photon sinks are col- lisional de-excitation, photoionization of metals by Lyα, and 2 photon decay. A breakdown of line profile weighted mean intensity, ¯JLyα, in terms of the different sources is given in Fig- ure 7. The external Lyα directly from star stays nearly m nm) becomes large at P > 1 µbar. 1s n1s) and metals ((cid:80) m Γ2nd The H radiative recombination cascade process dom- inates the Lyα photon production throughout the sim- ulation domain. The radiative decay after a thermal e collisional excitation from 1s state has a narrow peak in creating Lyα photon near 10−2 µbar. The collisional ex- citation by e generated by the photoionization of H(1s) (Γ2nd m nm) becomes important in creating Lyα photons near the top and base of the atomic layer respectively. The stellar Lyα photons incident through the surface and the photons generated above 10−2 µbar can mostly escape through the top boundary. In contrast, the pho- tons emitted below 0.1 µbar are mostly absorbed during the resonant scattering processes due to the high optical depth. The Lyα flux at the bottom boundary is about 40 erg cm−2 s−1. m Γ2nd The total LyC flux absorbed inside the simulation do- bar)mP (4-103-102-101-10110)-3 (cmn2103104105106107108109101010111012101310 IO IIO IC IIC IMg IIMg ISi IISi INa IINa IK IIKH(1s)pe2HH(n=2)constant above 0.05 µbar. Because the incident stel- lar Lyα is peaked at the frequency that is more than 5 Doppler widths away from the line center, the cross sec- tion to these photons in the Lorentzian wing is about 105 smaller than the cross section in the line center. Most stellar Lyα photons can directly penetrate to the intensity above this layer. Radiative recombination cas- P ∼ 10−2µbar level, which leads to a nearly constant cades are important at 10−3 − 10−2 µbar and below 0.1 µbar, with secondary e excitation becoming the second largest source deeper than 1 µbar. The Lyα mean intensity near the base can be estimated by assuming a local balance of frequency-integrated sources and sinks, giving J (cid:39) SL. Here J is the fre- quency integrated mean intensity, S is the frequency in- tegrated source function, and L is the total path length traversed by the photon before it is destroyed. The pho- ton source is insensitive to altitude as shown in Figure 6. For a sink given by true absorption due to photoioniza- tion of metals, the mean intensity is J (cid:39) SL = S nmσm,pi(νLyα) . (37) The photoionization cross section of metals by Lyα pho- tons, σm,pi(νLyα), is independent of altitude and fre- quency. The number densities of Mg I, Si I, Na I, and K I scale proportional or slightly steeper than P . As a result, the J ∝ P −1. Next consider collisional de-excitation and 2 photon decay, for which −1lmfp (cid:18) J (cid:39) S = S (cid:19)(cid:18) (cid:19) 1 , (38) 3A2p→1s 4neC2→1s(T ) + A2s→1s n1sσ1s(ν0) where is the probability of excitation not followed by de-excitation, and lmfp is the line center mean free path. Because ne, C2→1s(T ) are insensitive to altitude, J ∝ P −1 in this case also. To show the reliability of the approximations given in Equation 14, which are used in the hydrostatic model, the results for the H(1s) number density and temperature were plugged back to the full hydrogen level population code described in Section 3. Figure 8 compares the ap- proximate result and the full calculation for the ne, n2s, and n2p. The approximation holds for the whole simula- tion region. The n2s and n2p obtained from both meth- ods are almost identical except the approximate method slightly underestimates the n2s at the very top of the at- mosphere, due to the fact that the contribution from re- combination and collisional excitation is not completely negligible there. The ne obtained from the hydrogen level population calculation is slightly larger in the majority of the regions except the part that is close to the inner boundary. It is because that, in the hydrogen level popu- lation calculation, the secondary ionization due to metal species (see Equation 23) is not included, and the recom- binations are only included up to n = 6, while the case B recombination rate used in the hydrostatic model sums over all levels. The number density of H(n = 3) obtained from the level population is also shown in the Figure 8. Because an optically thin stellar Hα intensity is applied and Lyβ radiation transfer is not carefully considered, the number 13 density of H(n = 3) shown here is only a rough estimate. Nevertheless, the low number density indicates that the Paschen series absorption features are unlikely to be ob- served. 7. TRANSMISSION SPECTRUM For a planet at distance d from the observer, with uni- form intensity Iν over the stellar disk, the measured flux is Fν = 2πIν d2 −τν (b)bdb, e (39) (cid:90) R(cid:63) 0 where τν(b) is the optical depth along a trajectory asso- ciated with the impact parameter b. The optical depth can be divided into a continuum part, τc(b), which is in- dependent of frequency over the line, and the line opacity part due to absorption by H(2(cid:96)), (cid:90) √R2 top−b2 τl,ν(b) = 2 0 (n2sσ2s + n2pσ2p)ds, (40) where s is the line of sight distance. The continuum ab- sorption is then approximated as complete for b < Rp and zero for b > Rp. The continuum integral then be- comes as (cid:90) R(cid:63) ν ≡ Fν − F (c) F (c) ν = 2πIν d2 e 0 −τcbdb = Iν π(R2 (cid:63) − R2 p) d2 , (41) where Rp is the radius of the planet due to the continuum opacity. The difference in flux due to total opacity and continuum opacity is then ∆Fν = 2πIν d2 0 e −τc(b)−τl,ν (b) − e bdb. (42) The contribution from both terms is zero for b < Rp due to the continuum opacity, and there is no continuum absorption outside that range, and so this expression can be rewritten (cid:90) R(cid:63) (cid:16) Rp (cid:17) −τl,ν (b) − 1 e ∆Fν = 2πIν d2 bdb. (43) Equivalent to the transmission spectrum defined in the observations (e.g. Cauley et al. (2015)), the fractional change in flux, relative to the continuum integral at the same frequency, is then ∆F F (ν) ≡ = ∆Fν F (c) ν 2 (cid:63) − R2 R2 p (cid:90) R(cid:63) (cid:16) Rp (cid:17) e −τl,ν (b) − 1 bdb. (44) The ratio ∆F/F will be referred to as the model trans- mission spectrum. 7.1. Hα and Hβ Transmission Spectrum From the discussion in Section 5, the Lyα intensity is small in the molecular layer at r < Rb. As a result, the H(2(cid:96)) density there is small, and the region between Rp and Rb is transparent to Hα. (cid:90) R(cid:63) (cid:16) −τc(b)(cid:17) 14 1s n1s) and metals ((cid:80) m Γ2nd Figure 6. Lyα photon sources, sinks, and H ionization rates per unit volume as a function of pressure (P ). Each Lyα photon source in Equation 35 is plotted with a solid line. The secondary e excitation stands for the collisional excitation by e generated by the photoionization of H(1s) (Γ2nd m nm). The sink rates are the output from the Lyα Monte Carlo simulation. The process that a Lyα photon photoionizes a low ionization potential metal atom is referred as the photoionization of metals under the Lyα photon sinks section. A breakdown of the H radiative recombination cascade rate into individual H ionization processes based on Equation 21 is also plotted. The output rate of photoionization of H(2(cid:96)) in the Lyα Monte Carlo simulation as a photon sink recovers the rate of the same process as a H ionization given by the ionization equation. O charge exchange stands for the difference between recombination of an O II and ionization an O I by charge exchange with H (k(O) ion nO Inp). The photoionization of metals listed here stands for the H collisional ionization by e generated by the photoionization of metals ((cid:80) rec nO IIn1s − k(O) m Γ2nd m nm). Figure 7. Line profile weighted mean intensity ¯JLyα from differ- ent initial Lyα photon generation mechanisms against the pressure (P ). The large ¯JLyα stays nearly constant deep into the atmo- sphere and ¯JLyα ∝ P −1 near the base of the atomic layer. Both Lyα photons created inside the atmosphere and incident from the star are important in the Hα line formation region. Figure 8. Comparison of the hydrogen sub-states and electron number densities (n) obtained from the full hydrogen level pop- ulation code (solid line) and the approximation in Equation 14 (dashed line), which mostly overlaps with the solid lines. A rough estimate of the H(n = 3) number density is also shown in green. bar) mP (4-103-102-101-10110)-1 s-3 Sources and Sinks (cma Ly210310410510610 photon sinksaLyPhotoionization of H(n=2)Photoionization of metalsCollisional de-excitation2 photon decay photon sourcesaLyRadiative recombination cascadeThermal e excitationSecondary e exciationContributions to H ionizationPhotoionization of H(1s)Photoionization of H(n=2)O charge exchangePhotoionization of metalsCollisional ionization of H(n=2)Collisional ionization of H(1s)bar) mP (4-103-102-101-10110)-1 Hz-1 s-2 (erg cmaLyJ 10-109-108-10TotalStellarRecombination cascade excitation-Thermal ePhotoelectric excitationbar) mP (4-103-102-101-10110)-3 (cmn 3-102-101-101102103104105106107108109101010111012101310H(1s)H(2s)H(2p)eH(n=3)H(2s) approxH(2p) approxe approx15 Figure 9. Hα transmission spectrum. The black solid line shows the transmission spectrum of the fiducial model. The blue solid line shows the model with atomic layer base pressure Pb = 1 µbar. Black and red dashed lines show the model with stellar LyC mul- tiplier factor ξ = 1/4 and ξ = 4 respectively. The red solid line shows the model without Mg. Circles in the plot are observational data from Cauley et al. (2015, 2016). Figure 11. Hα, Hβ, Na D doublet, and Mg II 2795 line center optical depth (τ0) versus impact parameter (b). The vertical solid lines show the location of pressure levels 10 µbar, 1 µbar, 0.1 µbar, 0.01 µbar, and 0.001 µbar from left to right. Figure 10. Comparison of the Hβ transmission spectrum ob- served by Cauley et al. (2015) and Cauley et al. (2016), to the calculated transmission spectrum of the fiducial model, as well as models with different LyC boost factor ξ and Mg abundance. Figure 9 shows the model Hα transmission spectrum and the data from Cauley et al. (2015, 2016). We do not include the results of Jensen et al. (2012) since these observations were not performed across a single transit. The fiducial model discussed in Section 6 is given by the black line labeled "ξ = 1, Pb = 10 µbar". Given the noise in the data, the fiducial model is in broad agreement for both the line center absorption depth and the line width. The double-peak feature in Cauley et al. (2016)'s observations, whose amplitude is similar to the fluctua- tion in the continuum wavelength, cannot be explained by the model. The wavelengths have been corrected for the index of refraction of air at "standard condition", nHα = 1.0002762, according to Cox (2000). The plot also shows the effect of a different LyC flux, as denoted by lines with a different value of the factor ξ and the metallicity, which will be discussed in the Section 7.5 Figure 12. Hα and Na D transmission spectrum equivalent width per unit impact parameter (dWλ/db) (defined in Equation 45) ver- sus b. The contribution to Wλ by Na D doublet are summed to- gether. The vertical solid lines show the location of 10 µbar, 1 µbar, 0.1 µbar, 0.01 µbar, and 0.001 µbar from left to right. and 7.6, respectively. The base pressure, Pb, is not self-consistently deter- mined in this study. In order to investigate the depen- dence of the transmission spectrum on this parameter, the blue solid line labeled "ξ = 1, Pb = 1 µbar" shows a model with the base of the atomic layer at Pb = 1 µbar. The line center transit depth is smaller by (cid:39) 20% for Pb = 1 µbar. Changing this boundary causes only small changes the atmosphere properties, so the Hα be- comes optically thick at approximately the same pres- sure. However, because the scale height between 1 µbar and 10 µbar significantly decreases after switching to lower temperature and larger mean molecular weight, the radius in the atomic layer that corresponds to the same pressure becomes smaller, which leads to a smaller transit depth. Although we do not expect the transition from atomic layer to molecular layer to be as high up as 1 µbar based on Figure 5, a more physical molecular model is required to produce a more precise transmission )Å (l6562656365646565F/FD 0.015-0.01-0.005-0barm=10 b=1, Pxbarm=1 b=1, Pxbarm=10 b=1/4, Pxbarm=10 b=4, Pxbar, Low Mgm=10 b=1, PxCauley et al. 2015Cauley et al. 2016)Å (l486148624863F/FD 0.015-0.01-0.005-00.005=1x=1/4x=4x=1, Low MgxCauley et al. 2015Cauley et al. 2016 (1000 km)b80901001101201301401501600t 5-104-103-102-101-10110210310410510aHbH 5896 INa 5890 INa 2795 IIMg (1000 km)b8090100110120130140150160b/dλW d24−1023−1022−1021−1020−1019−10αHβH INa 2795 IIMg16 spectrum. Figure 10 shows the model Hβ transmission spectrum and the data from Cauley et al. (2015, 2016). While the model roughly agrees with the observation, there is an extra absorption on the blue side of the line, which cannot explained by the model. Because of the smaller cross section, Hβ line probes a deeper region in the at- mosphere compared to Hα. The lower temperature there leads to a narrower line width. Compared to Hα, the Hβ observations have larger uncertainty and less significant transit depth variation. Figure 11 shows the line center optical depth of the fiducial model versus impact parameter b. The Hα line center optical depth reaches the maximum value ∼ 70 for b = Rb, the base of the atomic layer. Although the optical depth of Hα slightly decreases inward, due to the (assumed) transparent molecular layer, the opti- cal depth is still much larger than 1 all the way to the continuum radius b = Rp. Lecavelier Des Etangs et al. (2008) showed that the optical depth at the effective ra- dius is τeq (cid:39) 0.561, and is not sensitive to the details of the atmospheric structure. In the fiducial model, the effective radius is 9.93× 109 cm, corresponding to an Hα optical depth τ = 0.55 and pressure P = 5.2×10−3 µbar. The optical depth drops to below ∼ 10−2 at a pressure 10−3 µbar, which means the contribution to Hα absorp- tion from the atmosphere above this level is small. Deter- mined by the ratio of oscillator strength and wavelength, the ratio between the optical depth of Hα and Hβ is 7.3. To indicate the vertical distribution of Hα and Na D absorption by the atmosphere, Figure 12 shows the equivalent width contributed by an annulus of atmo- sphere with radius b, defined as (cid:90) (cid:16) −τl,ν (b)(cid:17) dWλ db = 2b R2 (cid:63) − R2 p 1 − e dλ. (45) For impact parameters in the range Rp < b < Rb, which go through the molecular layer, the Hα line cen- ter optical depth is (cid:39) 70, and the absorption in the Lorentzian damping wing is negligible. Therefore, the base of the atomic layer is in the flat portion of the curve of growth (Draine 2011). The contribution to the equiv- alent width decreases slowly inward in this part of the atmosphere because of the smaller annulus radius and the lower temperature. It is shown that the atomic layer of the atmosphere has an approximately uniform contri- bution to Hα absorption, while the absorption of Hβ is dominated by the region of P (cid:38) 0.1 µbar. In princi- ple, high quality Hα and Hβ observations can be a good tracer for the vertical structure of the atomic layer. 7.2. Na D Transmission Spectrum Unlike Hα, the Na D doublet lines are absorbed by ground state Na, which has high density deep in the at- mosphere. The molecular layer is not treated in detail (cid:16) 0 du 1 With the approximation of the uniform mixing ratio and isothermal thin atmosphere, the integral in Equation 43 may be 1 − e−βe−u(cid:17) (cid:39) ln(β)+γ +O(β−2), expanded in a series as(cid:82) ∞ n(Rp)σ(cid:112)2πRpH. This formula is valid for β (cid:29) 1. This expansion where γ (cid:39) 0.577 is the Euler-Mascheroni constant and β (cid:39) then gives τeq = e−γ (cid:39) 0.561, in good agreement with Lecavelier Des Etangs et al. (2008). Figure 13. Comparison of the Na D doublet transmission spec- trum observed by Huitson et al. (2012) to the transmission spec- trum of the fiducial model (ξ = 1, Pb = 10 µbar). The histogram shown as a black line is the modeled spectrum binned to the instru- ment resolution. The blue curve shows the result of the modeled spectrum convolved with a Gaussian profile with FWHM matching the instrument resolution. Figure 14. Comparison of the Na D doublet transmission spec- trum observed by Wyttenbach et al. (2015) binned by 5×, to the calculated transmission spectrum of the fiducial model, as well as models with different atomic layer base pressures Pb, LyC boost factor ξ and metallicity. Note the break in the x-axis. The spectra of the three models with Pb = 10 µbar and solar abundance are nearly overlap on each other. here, rather a simple model with constant (solar abun- dance) mixing ratio and temperature T = 1140 K is used. Because Na and Mg are extremely optically thick in the molecular layer, their number density is not important in determining the line profile. For simplicity, number density of Na and Mg are assumed to be equal to the value at the very base layer of the model. Similar to Lyα, Na D photons undergo a resonant scattering pro- cess in the atmosphere, at least at pressures sufficiently low that collisional de-excitation and collisional broaden- ing are negligible. An accurate model of the Na D trans- mission spectrum requires treatment of resonant scatter- ing by Na I, as well as true absorption and emission by the atmosphere, which is beyond the scope of this pa- per. Instead, as was done for Hα in this work, a simple )Å (l588058855890589559005905F/FD0.0025-0.002-0.0015-0.001-0.0005-0Binned modelConvolved modelHuitson et al. 2012)Å (l588958905891F/FD 0.012-0.01-0.008-0.006-0.004-0.002-00.0020.004589558965897barm=10 b=1, Pxbarm=10 b=1/4, Pxbarm=10 b=4, Pxbarm=1 b=1, Pxbar, Low Mgm=10 b=1, PxWyttenbach et al. 2015e−τl,ν (b) absorption will be used to compute the trans- mission spectrum. Because the line wings of the Na D doublet are overlapping, τl,ν(b) uses the sum of the cross sections for each line of the doublet, evaluated at fre- quency ν. Figure 11 shows that the optical depths of the Na D doublet lines reach τ ∼ 0.5 at P ∼ 10−2 µbar, compa- rable to that of Hα, agreeing with the inference made previously based on the similar transit depth of Hα and Na D. At this altitude, the temperature is ∼ 8500 K, comparable to the analytic estimate using the difference between the Na D doublet transit depths discussed in Section 2. The optical depth of the Na D doublet become much larger than unity below 1 µbar. Thus Na D absorp- tion by the atmosphere near the base of the atomic layer is in the damped portion of the curve of growth, which explains the large contribution to the equivalent width shown in Figure 12. However, because of the slow tran- sit depth variation with frequency on the damping wing, it is difficult to distinguish the Na D damping wings from possible additional sources of continuum opacity or ob- servational error bars. The presence of clouds or hazes would further complicate the detection of this lower por- tion of the atmosphere, in spite of its large equivalent width contribution. The portion of the curve that is deeper than 10 µbar has no practical meaning because the value is limited by the wavelength integration range of the equivalent width. Figure 13 compares the observed Na D transmission spectrum from Huitson et al. (2012) and the fiducial model. The wavelength has been corrected for the index of refraction of air, nNaD = 1.0002771 according to Cox (2000). The spectral resolution of the Space Telescope Imaging Spectrograph (STIS) G750M grating aboard the HST used in this observation is ∼ 6 times broader than the full width at half maximum (FWHM) of each line in the Na D doublet. Two methods are used to compare the model spectrum with the low spectral resolution ob- servation. The first method is described in Huitson et al. (2012), and is shown as a black histogram. The model spectrum is binned to the STIS instrument resolution, 2 pixels, since the G750M grating gives a resolution ∼2 pixels at 5893 A. Care is required since the absorption depth of the binned spectrum near each line center de- pends on the wavelength range used for binning. In view of this, the second method, shown as a blue solid curve, convolves the model spectrum with a Gaussian profile with FWHM matching the instrument resolution (2 pixel widths). For both methods, to imitate the process of normalization to the continuum outside the regions of interest which been done in observational data reducing, the Na absorption depth at 5912 A is subtracted out and treated as the continuum. Figure 14 compares the high resolution Na D trans- mission spectrum observed by Wyttenbach et al. (2015) with the fiducial model, as well as models with different atomic layer base pressures Pb, ξ and metallicity, equiv- alent to the models shown in Figure 9. In comparing to the Hα transmission spectrum, recall that the line width of Na is narrower as compared to Hα due to the larger mean atomic weight of Na. To reduce the noise, the data plotted are binned by 5×. The resulting 0.05 A bin width is equal to the FWHM of the average spectrograph line 17 spread function, and is ∼ 3 times narrower than the Na D FWHM. On top of the 2.3 km s−1 shift to the red which accounts for the systemic velocity, the data were shifted by 10 km s−1 to the red to cancel the observed blueshift from an unknown source described in Wyttenbach et al. (2015). Similar to the treatment in Figure 13, the cor- rection from the index of refraction and continuum flux are made to the simulated spectrum. No binning or con- volving is required because the spectral features are well resolved. The line center absorption depths generated by the models agree with the Na D spectrum in both observa- tions roughly to the level of the observational error bars. Note that there is a strong absorption feature on the red side of the line center which cannot be explained by the model. 7.3. Retrieval of the Temperature Profile from Na D Transmission Spectra To measure the temperature versus altitude profile from the Na D transmission spectrum, r(λ), Lecavelier Des Etangs et al. (2008) applied the analytic model r(λ) = H ln(σ(λ)) + constant, (46) which is derived for a plane-parallel isothermal atmo- sphere with uniform mixing ratio of Na I and scale height H = kBT /µmpg. Here σ(λ) is the summed cross section from each line of the Na D doublet, and a Voigt pro- file at the local temperature is used. The wavelength- independent constant term is determined by the radius at continuum wavelengths. If the temperature and abun- dance vary slowly with altitude, an approximate scale height H of the atmosphere at a certain radius can be derived from H = (dr/dλ)/(d ln σ/dλ). Then, applying a mean molecular weight µ, the local temperature T (r) at r(λ) can be computed from the fitted value of H. Methods similar to this were applied by Huitson et al. (2012) and Wyttenbach et al. (2015) to measure the upper atmosphere temperature from their observed Na transmission spectra. The atmosphere was assumed to be molecular with µ = 2.3 in both studies. To decrease the uncertainty of the temperature measurement due to noise in the observed r(λ) profile, Huitson et al. (2012) broke the spectrum into small wavelength intervals, and fit r(λ) in each interval by varying H and the constant. Wyttenbach et al. (2015) also broke the spectrum into intervals, and fit for H in each spectral region, but with a fixed value for the constant term in Equation 46 in each interval. As a result, if connecting the fitting curves from separate wavelength ranges together, the joined curve is not continuous in both the slope and value of the tran- sit depth at the boundary between adjacent wavelength ranges. The measured temperatures near line center in both studies are more than a factor of 2 lower than the model temperature here (see Figure 16) over the relevant re- gion of the atmosphere. This is in spite of their using a mean molecular weight µ = 2.3, which assumes molec- ular hydrogen, while here the mean molecular weight in the atomic layer is closer to µ (cid:39) 1.3, smaller by a factor of 2. Such a large difference in temperature cannot be explained by the abundance variations due to ionization seen in Figure 5. 18 Figure 15. Comparison of the fiducial model transmission spec- trum and two different fits at a single point (circled) using the isothermal transit radius approximation in Equation 46. The black line shows the transit radius of the Na I 5890 line for the fiducial "ξ = 1, Pb = 10µ bar" model (see Figures 13 and 14). The red curve is chosen to agree with the value of the transit radius at r(λ) = 1.057Rp, and with the continuum r(λ) on the line wing. This requires a temperature T = 2830 K, assuming µ = 2.3. The blue curve is chosen to have the same slope as the black curve at the circled point, which requires T = 6010 K for µ = 2.3. Figure 16. Retrieved temperature profile using two different fit- ting strategies shown in Figure 15. Temperature profile fitted from the observed Na D transmission spectrum obtained by Wyttenbach et al. (2015) is shown with red circles. One indication that higher temperatures than found in Huitson et al. (2012) and Wyttenbach et al. (2015) are required comes from the inferred range of density be- tween line center and line wing. A lower temperature means a larger density difference between the base of the atmosphere, where the continuum forms, and higher al- titudes where the line center forms. An underestimate of this density decrease can be found by using the high- est fitted temperature used in Wyttenbach et al. (2015), T = 3270 K, and mean molecular weight µ = 2.3. As- suming the pressure is large, 1 bar at the continuum al- titude, also errs on the side of high density higher in the atmosphere. With these two assumptions, an isother- mal atmosphere gives the pressure 5 × 10−5 µbar at the line center altitude 1.27 × 104 km. To be optically thick to the Na I 5890 line, the ground state density must be nNaI the line center cross section of Na I 5890. This requires (cid:38) (σNaD2√2πrH)−1 (cid:39) 102 cm−3, where σNaD2 is a ∼ 10−6 mixing ratio of Na I which means Na has to be mostly neutral at this altitude if the atmosphere is in solar abundance. However, because the atmosphere is optically thin to the stellar flux at this pressure, Na is significantly ionized (see the left hand side of Figure 5). Therefore, the highest temperature measured in Wyt- tenbach et al. (2015) may underestimate the line center temperature formed high in the atmosphere. For sufficiently high spectral resolution data, r(λ), with high signal to noise for each data point, and for an atmo- sphere which is nearly isothermal and with small abun- dance gradients, the temperature of the atmosphere will be accurately recovered using Equation 46 in the plane- parallel limit. However, in an atmosphere where temper- ature increases upward rapidly, this method tends to un- derestimate the temperature (Wyttenbach et al. 2015). The problem is exacerbated when the opacity is provided by the line's Doppler core at the altitudes of interest. To better understand the retrieval of a temperature profile for the non-isothermal, non-constant abundance case, an example is given here to fit the fiducial model r(λ) (see Figures 13 and 14) with the isothermal profile in Equation 46. This eliminates measurement errors in the data, and a fine enough grid of points is used so that nu- merical error is negligible. The black curve in Figure 15 shows the fiducial model for the Na I 5890 absorption profile, the same as in Figure 14. Two methods are used to fit Equation 46 to the fiducial model. "Method 1" is equivalent to that in Wyttenbach et al. (2015). The constant term in Equation 46 is chosen in order that the absorption depth is 0 at 5912 A, on the line wing. The temperature T (r) at each radius r(λ) is determined by matching the value of the transit radius using Equa- tion 46 to the fiducial model. A mean molecular weight µ = 2.3 is used, as in Wyttenbach et al. (2015). Fig- ure 15 shows an example of a Method 1 fit with T = 2830 K, which matches the value of the absorption depth at R = 1.057Rp. When the value is fitted, the slope will be smaller than that of the fiducial model. "Method 2" is equivalent to that in Huitson et al. (2012). By adjusting H and the constant term at each r(λ), Equation 46 is used to match the slope of the fiducial model transit ra- dius. Again µ = 2.3 is used. Figure 15 shows an example of a Method 2 fit with T = 6010 that is tangent to the fiducial model r(λ) curve at R = 1.057Rp. The retrieved temperature profiles for Method 1 and Method 2 are compared to the fiducial model tempera- ture profile in Figure 16. For comparison, the Method 1 temperature profile estimated from the data by Wyt- tenbach et al. (2015) is shown as the points with error bars. Given that there is no numerical noise in this ex- ample, as the isothermal r(λ) is fit to a theoretical model, the disagreement between the Method 1 and Method 2 fits and the true temperature profile is quite large. The disagreement would be even larger if the more appro- priate µ (cid:39) 1.3 was used near line center. Method 1 can reasonably retrieve the temperature in the molecular layer, where the fiducial model temperature is constant and µ = 2.3. However, the retrieved temperature in the atomic layer is lower than the fiducial model where the )Å (l5889.65889.858905890.2p/Rl R0.9511.051.11.151.2ModelValueSlopeAltitude (1000 km)0246810121416 T (K)010002000300040005000600070008000900010000Wyttenbach et al 2015ModelValueSlope19 Figure 17 shows the model predicted transmission spectrum of Mg II 2795, one of the Mg II doublets. No correction of index of refraction has been applied. As- suming solar metalicity, almost the entire atmosphere in the simulation region is optically thick to Mg lines at the line center, because of the high abundance and shallower dependence on pressure (see Figure 5). In this case, the line center transit depth is ∼ 5%, 3 times larger than the planet transit depth in continuum. In contrast, assum- ing the abundance of Mg is 10−4 of the solar value, the transit depth is ∼ 0.8%. The model predicts very sim- ilar transmission spectra for Mg II 2803 and Mg I 2852 line. Because of the probable large transit depth and its strong impact on the physical properties of the atmo- sphere, the transmission spectra of Mg resonant lines in middle UV might be a good target to constrain the hot Jupiter upper atmosphere. 7.5. Impact of LyC Flux on Transit Depth HD 189733 is known to be an active star (Boisse et al. 2009; Pillitteri et al. 2014, 2015). To show the de- pendence of the transmission spectrum on the stellar EUV/X-ray flux, two more models with an extra LyC flux multiplier factor ξ = 1/4 and ξ = 4 are applied. Figure 9 and 10 show that a stronger LyC flux will make the Hα and Hβ transit depth deeper. This is consistent with the conclusion in Cauley et al. (2017) that the Hα transit showing the largest absorption value occurs when the star is the most active. In comparison, Figure 14 shows that LyC flux has no effect on the NaD transmis- sion spectrum. The reason of this difference is that the Na D transmission spectrum depends on the temperature or scale height of the atmosphere as well as Na ionization fraction below 10−2 µbar. Above the level of 10−1 µbar, collisional ionization dominate Na ionization. Although the temperature of the atmosphere increases with the ξ, the effect of increasing scale height is canceled by the Na higher ionization fraction. In contrast, the ionization of H is not sensitive to the temperature. Instead, Balmer lines depend on the Lyα intensity in this region, which is larger for a strong LyC flux environment. The Balmer lines transmission spectra also become slightly broader in the strong LyC flux case, because the lines become optically thick at higher and hotter part of the atmo- sphere in this case. This indicates that Balmer lines are the better tracer of atmosphere temperature compared to Na doublets. Comparing the variability of the transit depths of the Hα and Na D lines is a possible method to break the degeneracy between the transit depth variabil- ity due to blocking an active region on the star surface and the change in the atmosphere due to stellar activity. 7.6. Impact of Metallicity on Transit Depth The metallicity is crucial in the model presented in this paper, but its value is uncertain. Since Mg is the dom- inant coolant in the model, a model that reduces the Mg abundance to 10−4 of the solar value is calculated to assess the effect of metallicity on the transmission spec- trum. Because the atmosphere is warmer and more ex- tended without Mg cooling, the transit depths of Hα and Hβ become deeper, as shown in Figures 9 and 10. The transit depth of Na is insensitive to the Mg abundance because of the trade off between atmosphere scale height Figure 17. The calculated Mg II 2795 transmission spectrum of the fiducial model, as well as models with different LyC boost factor ξ and Mg abundance. temperature increases outward. Although the retrieved temperature is still higher than the points from Wytten- bach et al. (2015), this example partially explains the lower inferred temperature in that work as compared to the fiducial model in this work. Retrieving the temperature from a high resolution spectrum using Method 2 will significantly overestimate the temperature. In the Doppler core, Equation 46 gives r(λ) − r(λ0) =−H (cid:18) ∆ν ∆νD (cid:19)2 (cid:18) ∆λ (cid:19)2 , (47) (cid:39)−6000 km 0.1 A independent of temperature. Equation 47 shows that the slope gets steeper further from line center. This contin- ues until the damping wing is reached, where the slope becomes more shallow. Therefore, Equation 47, evalu- ated near the core-wing boundary, gives the maximum slope of transit radius with ∆λ for the isothermal pro- file. By contrast, the fiducial model r(λ) is steeper than Equation 47 in the line core because the absorption by a higher and hotter atmosphere layer produces a broader absorption than if the temperature is constant. As the result, the slope of a section of the fiducial model near the line core is too steep and Equation 46 cannot produce such a steep slope for any temperature. In addition, near the line core, the slope of r(λ) for the fiducial model also depends on the Na I abundance gradient. Collisional ion- ization by thermal electrons and Lyα photoionization de- crease the Na ionization fraction at level above 0.1 µbar (see Figure 5), which can decrease the slope near the line center. These results suggest that the isothermal model may not accurately retrieve a rapidly rising temperature pro- file (see also Heng et al. (2015)). In comparison, fitting the whole wavelength range with an atmosphere model contains several isothermal layers, or a single layer with a continuous temperature profile may better constrain the atmosphere temperature. 7.4. Mg Transmission Spectrum )Å (l27962796.52797F/FD 0.06-0.04-0.02-0=1x=1/4x=4x=1, Low Mgxbarm=1 b=1, Px20 and ionization fraction discussed in 7.5. The relatively large transit depth difference between two models with and without Mg indicates that high precision Balmer and Mg transmission spectrum measurements can constrain the metallicity in the upper atmosphere. 8. DISCUSSION 8.1. Other Possible Cooling Mechanisms Adiabatic cooling is another potential cooling mecha- nism discussed in the literature. Koskinen et al. (2013a) constructed an atmosphere model for a similar hot Jupiter HD 209458b. In their model, the stellar heat- ing is mainly balanced by adiabatic cooling. Compared with HD 189733b system, the LyC flux of HD 209458 is weaker and the orbit of HD 209458b is further away from the star. They also introduced a factor of 1/4 re- duction on stellar flux to account for uniform day-night heat redistribution. As a result of these differences, the heating rate of Koskinen et al. (2013a) HD 209458b at- mosphere model is more than 20 times smaller than the rate in our HD 189733b model. On the other hand, the adiabatic cooling does not differ much in two systems because of the similar mass loss rate (Murray-Clay et al. 2009). Therefore, the adiabatic cooling is unlikely to be the answer in the case of HD 189733b. The effect of the adiabatic cooling of the model here can be estimated with an assumed mass loss rate. In- cluding the adiabatic cooling in the pressure coordinate system according to Bildsten (1998), the entropy equa- tion takes the following form, (cid:88) where (cid:80) H and (cid:80) C stand for the sum of heating and Cp M T ρgp d ln T d ln P (cid:88) ∇ad − , (48) H − C = 4πR2 pP (cid:18) (cid:19) cooling rates respectively as shown in the right and left hand side of Equation 24, ∇ad = (d ln T /d ln P )S is the M adiabatic temperature gradient at constant entropy, is the mass loss rate, and Cp is the specific heat per unit mass at constant pressure. In the fiducial model, the at the radius that 13.6 eV photon becomes optically thick, where most LyC photons get absorbed. Ap- plying the mass-loss rate gested by Salz et al. (2016) using a hydrodynamic es- caping atmosphere model, the adiabatic cooling rate is temperature gradient is about dT /dr = 5×10−7 K cm−1 M = 4 × 109 g s−1 sug- 1.1 × 10−8 erg cm−3 s−1 and the first term in the Equa- tion 48 is the dominant source. Compared to the heat- ing and cooling rates shown in Figure 4, the adiabatic cooling is more than two orders of magnitude smaller in the region where Hα mostly absorbed, and may only be- come important in the region above P ∼ 3 × 10−4 µbar. The mass-loss rate may be model dependent. An up- per bound for the mass loss rate can be found using the energy-limited escape rate (Murray-Clay et al. 2009), which assumes all LyC flux converts to unbinding the at- corresponds to the energy-limited mass-loss rate mosphere. The LyC flux FLyC = 2.6 × 104 erg cm−2 s−1 M = 3 × 1011 g s−1. In this case, the adiabatic cooling rate is 8 × 10−7 erg cm−3 s−1 and still has a less than 15% effect in the region mainly concerned. 8.2. Comparing with Other Hot Jupiter Upper Atmosphere Models Christie et al. (2013) - The present study agrees with the conclusion in Christie et al. (2013) that the 2p occupa- tion is set by radiative excitation and de-excitation, and as an improvement, we include a Lyα radiation trans- fer inside the atmosphere instead of applying a constant solar Lyα intensity. Because of the ∼ 30 times stronger stellar Lyα intensity of HD 189733 comparing to the Sun and considering the Lyα photons generated inside the at- mosphere due to collisional excitation and recombination cascades, the Lyα mean intensity should be ∼ 100 times larger for the majority of the atomic layer. In addition, because Christie et al. (2013) underestimates the n2p by a factor of 20 due to a math error, the n2p should be signif- icantly larger in the whole simulation domain, and thus the atmosphere is optically thick to Hα mainly due to the absorption of H(2p). The observed Hα absorption width agrees well with an optically thick atmosphere model. Because of the much larger n2(cid:96), the photoionization of H(2(cid:96)) is larger compared to photoionization of the ground state in the atomic layer. Hence the ne and np in this work is ∼ 10 times larger. Considering the proton collisional (cid:96)-mixing process with rate ∼ 10 times larger than electron collisional pro- cess, as well as the large H(2p) population, the creation of 2s hydrogen is dominated by (cid:96)-mixing rather than col- lisional excitation considered in Christie et al. (2013). As a result, 2s and 2p reach collisional equilibrium. In addition, it is shown that the metal lines are crucial in cooling the atmosphere. Assuming solar abundance, lines of Mg and Na can cool the atomic layer by (cid:39) 2000− 3000 K. Menager et al. (2013) - Menager et al. (2013) investi- gate the Lyα emission and reflection by the atmosphere of HD 189733b. The temperature and electron, hydro- gen, and helium number density profiles of HD 189733b from the Koskinen et al. (2011) unpublished model were applied. The temperature, ne, np, and nH are in broad agreement with the profiles presented in this paper in the corresponding pressure range. According to a sim- ilar model of HD 209458b presented in Koskinen et al. (2013a), it should be a one-dimensional hydrodynamic model of the upper atmosphere considering hydrogen and helium constructed on top of a full photochemical model of the lower atmosphere. They chose the average solar flux as their stellar spectrum, which is ∼ 10 times smaller than the synthetic spectrum from MUSCLES. Different from the photoelectron heating efficiency η(E) calculated at the fixed ionization fraction xe = 0.1 throughout the model, or a constant η applied in the Koskinen et al. (2013a), a η based on the local xe is used in this work. The temperature in the HD 189733b model of Kosk- inen et al. (2011) reachs a peak of about 13000 K at a pressure of 3 × 10−4 µbar. The adiabatic cooling lowers the temperature at higher altitude. Their temperature at pressure range 10−3 to 1 µbar is higher by about 3000 K. Two possible reasons of this difference are Koskinen et al. (2013a) do not consider metal lines cooling, which are the dominant cooling mechanisms in our model, and conduction is not included in this work, which is a net heating in this pressure range according to their result. Their temperature decreases much faster with pressure above 1 µbar comparing to Figure 3. The lack of molec- ular cooling in this work is the possible reason. Their ne decreases slowly with pressure in the atomic layer and is ∼ 10 times smaller at 10 µbar compared to this work. This difference is the result of missing the photoioniza- tion from H(2s), which is the dominant ionization mech- anism in this region. In the Lyα radiation transfer simulation of Menager et al. (2013), the Lyα photons emitted by the star and by the planetary atmosphere are considered. However, when considering the Lyα photon emitted by the plane- tary atmosphere, the Lyα photon created through recom- bination cascades is not included. In order to calculate the Lyα thermal excitation in the atmosphere, the num- ber densities of 2s and 2p states hydrogen are modeled with a level population study. In the 2p state level equa- tion, Lyα excitation, which completely dominates the 2p state, is not included. In addition, because the p colli- sional (cid:96)-mixing process is missing in their model, the 2(cid:96) state number densities result shown are affected. Menager et al. (2013) claimed that the thermal emis- sion of HD 189733b contributes to 6% of the total intensity of the Lyα line. In the fiducial model of this work, this ratio is 9.6 × 103 erg cm−2 s−1/(2.0 × 104 erg cm−2 s−1 + 9.6× 103 erg cm−2 s−1) = 32%. This ratio strongly depends on the metallicity as well as the stellar LyC to Lyα flux ratio. 9. CONCLUSION A detailed one-dimensional hydrostatic atmosphere model is constructed over the region dominated by atomic hydrogen and comparison of model transmission spectra to the data has been made. An atomic hydrogen level population calculation and a Monte-Carlo Lyα ra- diation transfer are done to model the abundance of 2(cid:96) state hydrogen. The model transmission spectra of Hα, Hβ, and Na are in broad agreement with the HD 189733b data for both the line center absorption depth and the line width, although the comparison is complicated by the observed variability. The Lyα radiation transfer shows that the Lyα has a very broad line width with a flat top due to the resonant scattering process. The line profile weighted mean inten- sity ¯JLyα is large and approximately constant down to the P = 0.1 µbar level of the atmosphere. Lyα photons created inside the atmosphere and incident from the star are both important. The Lyα source function extends deep into the atmosphere due to ionization from pro- gressively higher energy stellar LyC photons. The stel- lar Lyα photon can penetrate into very large line center optical depth because the stellar Lyα intensity is much boarder than the Doppler width inside the atmosphere. The stellar Lyα photons incident through the surface and the photons generated above 10−2 µbar can mostly es- cape through the top boundary. In contrast, the pho- tons emitted below 0.1 µbar are mostly absorbed during the resonant scattering processes due to the high optical depth. For P (cid:38) 0.1 µbar, ¯JLyα ∝ P −1. The n2p is determined by the radiative rates between 1s and 2p throughout the simulation domain because of the large Lyα intensity. The 2s and 2p states reach col- lisional equilibrium by the large p collisional (cid:96)-mixing rate, which was overlooked in this context. The com- 21 bination of the decreasing Lyα excitation rates and the increasing hydrogen density gives rise to a nearly flat n2(cid:96) over two decades in pressure. This layer is opti- cally thick to Hα, and the temperature is in the range T (cid:39) 3000 − 8500 K. Both Hα and NaD are optically thick up to the level P ∼ 10−2 µbar, which corresponds to the atomic layer of the atmosphere. Assuming solar abundance, radiative cooling due to metal species domi- nates over the entire model, with Mg and Na being the two most important species. The model shows that Mg II may have a very large transit depth assuming solar abun- dance, which might be a good target to constrain the atmospheric properties. Additional models computed for a range of the stel- lar LyC flux find transit depth of Hα changes with LyC level, suggesting that the variability in Hα transit depth may be due to variability in the stellar LyC. In contrast, the Na absorption profile is insensitive to the LyC level. Since metal lines provide the dominant cooling of this part of the atmosphere, the atmosphere structure is sen- sitive to the density of species such as Mg and Na, which may themselves be constrained by observations. Lastly, since the Hα and Na D lines have comparable absorp- tion depths for the same spectral resolution, we argue that the center of the Na D lines are also formed in the atomic layer where the Hα line is formed. The present model is in agreement with the observed Na D transmission spectrum by Huitson et al. (2012) and Wyttenbach et al. (2015), although the inferred at- mospheric temperature is significantly larger than that found assuming an isothermal profile and molecular com- position. It is shown that the temperature achieved by fitting each wavelength interval in the observed transmis- sion spectrum with an isothermal atmosphere model may not accurately retrieve the original temperature profile, if the temperature increases rapidly with the altitude. 10. ACKNOWLEDGMENTS We are grateful to Wilson Cauley and Aur´elien Wyt- tenbach for kindly providing the data. We thank Wil- son Cauley, Seth Redfield, and Adam Jansen for useful discussions regarding Hα observations. We also thank the helpful conversations with Ira Wasserman, Craig Sarazin, Shane Davis, Roger Chevalier, and Remy Inde- betouw regarding the model construction, and the help from Katherine Holcomb in improving the code efficiency. The simulations in this work were carried out on the Ri- vanna computer cluster at the University of Virginia. We also thank the referee for providing constructive com- ments and suggestions, especially in suggesting use of the MUSCLES spectrum. This research was supported by NASA grants NNX14AE16G, NNX10AH29G, and NNX15AE05G. APPENDIX Table 5 contains the transitions from O, C, S, and Si lines, which are included in the model as cooling pro- cesses but only have a minor effect on the temperature. Asplund, M., Grevesse, N., Sauval, A. J., & Scott, P. 2009, REFERENCES ARA&A, 47, 481 Ben-Jaffel, L., & Ballester, G. E. 2013, A&A, 553, A52 List of minor metal cooling transitions Table 5 22 Transition O I 6300+6363 C I 9850+9824 C I 8727 Si II 2350+2334 Si II 2344 Si II 2335 Si II 1817(2D3/2)+1808 Si II 1817(2D5/2) Si I 16454+16068 Si I 10991 Si I 6527 Si I 3020+3007 Si I 2988 Si I 2882 Si I 2529+2519+2514 Si I 2524 Si I 2516+2507 S I 25.25 µm S I 11306+10821 S I 7725 S II 10336 S II 10320+10287 S II 6731 S II 6716 S II 4076 S II 4069 aKramida et al. bT4 = T /104 K (2015) ∆E (eV) 1.957 1.260 1.420 5.294 5.287 5.309 6.839 6.823 0.762 1.128 1.899 4.108 4.149 4.301 4.917 4.911 4.935 0.0491 1.121 1.605 1.199 1.203 1.842 1.845 3.041 3.046 Eu (eV) 1.967 1.264 2.684 5.310 5.323 5.345 6.857 6.859 0.781 1.909 1.909 4.132 4.930 5.082 4.930 4.920 4.954 0.0491 1.145 2.750 3.041 3.046 1.842 1.845 3.041 3.046 a Aul (s−1) 7.45 × 10−3 1.45 × 10−4 0.599 1.00 1.02 × 103 1.31 × 103 2.44 × 103 2.86 × 106 2.65 × 106 2.72 × 10−3 2.74 × 10−2 4.4 × 103 2.66 × 106 2.17 × 108 2.19 × 108 2.22 × 108 2.23 × 108 1.40 × 10−3 2.75 × 10−2 1.38 0.142 0.272 6.84 × 10−4 2.02 × 10−4 7.72 × 10−2 0.192 3.6 × 10−21e−3.12/T4 /(0.599 + 2.39 × 10−8ne) Pequignot & Aldrovandi (1976) Λ ( erg cm3 s−1) 1.3 × 10−14/nee−2.28/T4 b 3.3 × 10−16/nee−1.47/T4 lu 8.48 × 10−12C(e) 8.47 × 10−12C(e) 8.51 × 10−12C(e) 1.10 × 10−11C(e) 1.10 × 10−11C(e) lu lu lu lu 1.84 × 10−15/nee−0.906/T4 2.01 × 10−13/nee−2.22/T4 9.26 × 10−15/nee−2.22/T4 9.3 × 10−24T 0.18e−4.79/T4 4.2 × 10−22T 0.18e−5.72/T4 3.1 × 10−20T 0.18e−5.90/T4 5.5 × 10−20T 0.18e−5.72/T4 1.2 × 10−20T 0.18e−5.71/T4 4.3 × 10−20T 0.18e−5.75/T4 3.67 × 10−17/nee−0.0570/T4 2.78 × 10−14/nee−1.33/T4 3.94 × 10−13/nee−3.19/T4 5.46 × 10−14/nee−3.53/T4 2.09 × 10−13/nee−3.53/T4 8.07 × 10−16/nee−2.14/T4 3.59 × 10−16/nee−2.14/T4 8.27 × 10−14/nee−3.53/T4 3.75 × 10−13/nee−3.53/T4 Source of collision rate or cooling rate Equation 34 Equation 34 CHIANTI CHIANTI CHIANTI CHIANTI CHIANTI Equation 34 Equation 34 Equation 34 Van Regemorter formula Van Regemorter formula Van Regemorter formula Van Regemorter formula Van Regemorter formula Van Regemorter formula Equation 34 Equation 34 Equation 34 Equation 34 Equation 34 Equation 34 Equation 34 Equation 34 Equation 34 Bildsten, L. 1998, NATO Advanced Science Institutes (ASI) France, K., Parke Loyd, R. O., Youngblood, A., et al. 2016, ApJ, Series C, 515, 419 Birkby, J. L., de Kok, R. J., Brogi, M., et al. 2013, MNRAS, 436, L35 Black, J. H., & van Dishoeck, E. F. 1987, ApJ, 322, 412 Boisse, I., Moutou, C., Vidal-Madjar, A., et al. 2009, A&A, 495, 959 Borysow, A., Frommhold, L., & Moraldi, M. 1989, ApJ, 336, 495 Bourrier, V., Lecavelier des Etangs, A., Dupuy, H., et al. 2013, A&A, 551, A63 Cauley, P. W., Redfield, S., Jensen, A. G., et al. 2015, ApJ, 810, 13 Cauley, P. W., Redfield, S., Jensen, A. G., & Barman, T. 2016, AJ, 152, 20 820, 89 Garc´ıa Munoz, A. 2007, Planet. Space Sci., 55, 1426 Gladstone, G. R. 1988, J. Geophys. Res., 93, 14623 Harrington, J. P. 1973, MNRAS, 162, 43 Heng, K., Wyttenbach, A., Lavie, B., et al. 2015, ApJ, 803, L9 Huitson, C. M., Sing, D. K., Vidal-Madjar, A., et al. 2012, MNRAS, 422, 2477 Hummer, D. G. 1962, MNRAS, 125, 21 Igenbergs, K., Schweinzer, J., Bray, I., Bridi, D., & Aumayr, F. 2008, Atomic Data and Nuclear Data Tables, 94, 981 Janev, Ratko K., Reiter, D., & Samm, U. 2003, Collision processes in low-temperature hydrogen plasmas, Forschungszentrum Julich, Zentralbibliothek Cauley, P. W., Redfield, S., & Jensen, A. G. 2017, AJ, 153, 217 Chandrasekhar, S. 1960, New York: Dover, 1960, Charbonneau, D., Brown, T. M., Noyes, R. W., & Gilliland, R. L. Jensen, A. G., Redfield, S., Endl, M., et al. 2011, ApJ, 743, 203 Jensen, A. G., Redfield, S., Endl, M., et al. 2012, ApJ, 751, 86 Khalafinejad, S., von Essen, C., Hoeijmakers, H. J., et al. 2017, 2002, ApJ, 568, 377 A&A, 598, A131 Curdt, W., Tian, H., Teriaca, L., & Schuhle, U. 2010, A&A, 511, Koskinen, T. T., Harris, M., Yelle, R. V., Lavvas, P., & Lewis, N. L4 Christie, D., Arras, P., & Li, Z.-Y. 2013, ApJ, 772, 144 Cox, A. N. 2000, Allen's Astrophysical Quantities, Dalgarno, A., Yan, M., & Liu, W. 1999, ApJS, 125, 237 Del Zanna, G., Dere, K. P., Young, P. R., Landi, E., & Mason, H. E. 2015, A&A, 582, A56 Dere, K. P., Landi, E., Mason, H. E., Monsignori Fossi, B. C., & Young, P. R. 1997, A&AS, 125, Draine, B. T. 2011, Physics of the Interstellar and Intergalactic Medium by Bruce T. Draine. Princeton University Press, 2011. ISBN: 978-0-691-12214-4 Ehrenreich, D., Tinetti, G., Lecavelier Des Etangs, A., Vidal-Madjar, A., & Selsis, F. 2006, A&A, 448, 379 Fontenla, J. M., Harder, J., Livingston, W., Snow, M., & Woods, T. 2011, Journal of Geophysical Research (Atmospheres), 116, D20108 Fortney, J. J., Sudarsky, D., Hubeny, I., et al. 2003, ApJ, 589, 615 2011, EPSC-DPS Joint Meeting 2011, 1169 Koskinen, T. T., Harris, M. J., Yelle, R. V., & Lavvas, P. 2013, Icarus, 226, 1678 Koskinen, T. T., Yelle, R. V., Harris, M. J., & Lavvas, P. 2013, Icarus, 226, 1695 Kramida, A., Ralchenko, Yu., Reader, J. & NIST ASD Team 2015, NIST Atomic Spectra Database (version 5.3), [Online]. Available: http://physics.nist.gov/asd Landini, M., & Fossi, B. C. M. 1991, A&AS, 91, 183 Lavvas, P., Koskinen, T., & Yelle, R. V. 2014, ApJ, 796, 15 Lecavelier Des Etangs, A., Pont, F., Vidal-Madjar, A., & Sing, D. 2008, A&A, 481, L83 Lecavelier Des Etangs, A., Ehrenreich, D., Vidal-Madjar, A., et al. 2010, A&A, 514, A72 Lennon, M. A., Bell, K. L., Gilbody, H. B., et al. 1988, Journal of Physical and Chemical Reference Data, 17, 1285 23 Lenzuni, P., Chernoff, D. F., & Salpeter, E. E. 1991, ApJS, 76, Salz, M., Czesla, S., Schneider, P. C., & Schmitt, J. H. M. M. 759 Linsky, J. L., France, K., & Ayres, T. 2013, ApJ, 766, 69 Loyd, R. O. P., France, K., Youngblood, A., et al. 2016, ApJ, 824, 102 Lucy, L. B. 1999, A&A, 344, 282 Menager, H., Barth´elemy, M., Koskinen, T., et al. 2013, Icarus, 226, 1709 2016, A&A, 586, A75 Seaton, M. J. 1955, Proceedings of the Physical Society A, 68, 457 Sharp, C. M., & Burrows, A. 2007, ApJS, 168, 140 Shull, J. M., & van Steenberg, M. 1982, ApJS, 48, 95 Shull, J. M. 1978, ApJ, 224, 841 The Opacity Project Team. 1995, The Opacity Project Vol. 1, Institute of Physics Publications, Bristol, UK Miguel, Y., Kaltenegger, L., Linsky, J. L., & Rugheimer, S. 2015, Tian, H., Curdt, W., Marsch, E., & Schuhle, U. 2009, A&A, 504, MNRAS, 446, 345 Murray-Clay, R. A., Chiang, E. I., & Murray, N. 2009, ApJ, 693, 23 Osorio, Y., Barklem, P. S., Lind, K., et al. 2015, A&A, 579, A53 Osterbrock, D. E., & Ferland, G. J. 2006, Astrophysics of gaseous nebulae and active galactic nuclei, 2nd. ed. by D.E. Osterbrock and G.J. Ferland. Sausalito, CA: University Science Books, 2006, Pequignot, D., & Aldrovandi, S. M. V. 1976, A&A, 50, 141 Pequignot, D., Petitjean, P., & Boisson, C. 1991, A&A, 251, 680 Pillitteri, I., Wolk, S. J., Lopez-Santiago, J., et al. 2014, ApJ, 785, 145 Pillitteri, I., Maggio, A., Micela, G., et al. 2015, ApJ, 805, 52 Pont, F., Sing, D. K., Gibson, N. P., et al. 2013, MNRAS, 432, 2917 239 van Regemorter, H. 1962, ApJ, 136, 906 Verner, D. A., Ferland, G. J., Korista, K. T., & Yakovlev, D. G. 1995, Bulletin of the American Astronomical Society, 27, 34.02 Verner, D. A., & Ferland, G. J. 1996, ApJS, 103, 467 Verner, D. A., Ferland, G. J., Korista, K. T., & Yakovlev, D. G. 1996, ApJ, 465, 487 Vidal-Madjar, A., Huitson, C. M., Bourrier, V., et al. 2013, A&A, 560, A54 Visscher, C., Lodders, K., & Fegley, B., Jr. 2010, ApJ, 716, 1060-1075 Vrinceanu, D., Onofrio, R., & Sadeghpour, H. R. 2012, ApJ, 747, 56 Whitney, B. A. 2011, Bulletin of the Astronomical Society of India, 39, 101 Poppenhaeger, K., Schmitt, J. H. M. M., & Wolk, S. J. 2013, Wiese, W. L., & Fuhr, J. R. 2009, Journal of Physical and ApJ, 773, 62 Chemical Reference Data, 38, 565 Press, W. H., Teukolsky, S. A., Vetterling, W. T., & Flannery, Wyttenbach, A., Ehrenreich, D., Lovis, C., Udry, S., & Pepe, F. B. P. 2007, Numerical recipes 3rd edition : the art of scientific computing by William H. Press. Cambridge University Press. ISBN : 0521880688, 2015, A&A, 577, A62 Yelle, R. V. 2004, Icarus, 170, 167 Youngblood, A., France, K., Parke Loyd, R. O., et al. 2016, ApJ, Redfield, S., Endl, M., Cochran, W. D., & Koesterke, L. 2008, 824, 101 ApJ, 673, L87 Rybicki, G. B., & Lightman, A. P. 1979, New York, Wiley-Interscience, 1979. 393 p., Zheng, Z., & Miralda-Escud´e, J. 2002, ApJ, 578, 33
1510.05690	1	1510	2015-10-19T21:06:29	Saturn's Seasonally Changing Atmosphere: Thermal Structure, Composition and Aerosols	[ "astro-ph.EP" ]	The longevity of Cassini's exploration of Saturn's atmosphere (a third of a Saturnian year) means that we have been able to track the seasonal evolution of atmospheric temperatures, chemistry and cloud opacity over almost every season, from solstice to solstice and from perihelion to aphelion. Cassini has built upon the decades-long ground-based record to observe seasonal shifts in atmospheric temperature, finding a thermal response that lags behind the seasonal insolation with a lag time that increases with depth into the atmosphere, in agreement with radiative climate models. Seasonal hemispheric contrasts are perturbed at smaller scales by atmospheric circulation, such as belt/zone dynamics, the equatorial oscillations and the polar vortices. Temperature asymmetries are largest in the middle stratosphere and become insignificant near the radiative-convective boundary. Cassini has also measured southern-summertime asymmetries in atmospheric composition, including ammonia (the key species for the topmost clouds), phosphine and para-hydrogen (both disequilibrium species) in the upper troposphere; and hydrocarbons deriving from the UV photolysis of methane in the stratosphere (principally ethane and acetylene). These chemical asymmetries are now altering in subtle ways due to (i) the changing chemical efficiencies with temperature and insolation; and (ii) vertical motions associated with large-scale overturning in response to the seasonal temperature contrasts. Similarly, hemispheric contrasts in tropospheric aerosol opacity and coloration that were identified during the earliest phases of Cassini's exploration have now reversed, suggesting an intricate link between the clouds and the temperatures. [Abridged]	astro-ph.EP	astro-ph	10 Saturn’s seasonally changing atmosphere: thermal structure, composition and aerosols L.N. FLETCHER1, T.K. GREATHOUSE2, S. GUERLET3, J.I. MOSES4 AND R.A. WEST5 5 1 0 2 t c O 9 1 . ] P E h p - o r t s a [ 1 v 0 9 6 5 0 . 0 1 5 1 : v i X r a 1 Department of Physics and Astronomy, University of Leicester, University Road, Leicester, LE1 7RH, UK. 2 Southwest Research Institute, Department of Space Science, Space Science and Engineering Division, 6220 Culebra Road, San Antonio, TX 78238-5166, USA. 3 Laboratoire de Meteorologie Dynamique / CNRS / Univ. Paris 6, 4 Place Jussieu, 75252 Paris, France. 4 Space Sciences Institute, 4750 Walnut St, Suite 205, Boulder, CO 80301, USA. 5 NASA Jet Propulsion Laboratory, 4800 Oak Grove Drive, Pasadena, CA 91109, USA. Copyright Notice J.I. Moses and R.A. West The Chapter, Saturn’s Seasonally Changing Atmosphere: Ther- mal Structure, Composition and Aerosols, is to be published by Cambridge University Press as part of a multi-volume work edited by Kevin Baines, Michael Flasar, Norbert Krupp, and Thomas Stallard, entitled “Saturn in the 21st Century” (‘the Volume’) c(cid:13) in the Chapter, L.N. Fletcher, T.K. Greathouse, S. Guerlet, c(cid:13) in the Volume, Cambridge University Press NB: The copy of the Chapter, as displayed on this website, is a draft, pre-publication copy only. The final, published version of the Chapter will be available to purchase through Cambridge University Press and other standard distribution channels as part of the wider, edited Volume, once published. This draft copy is made available for personal use only and must not be sold or re-distributed. Abstract The longevity of Cassini’s exploration of Saturn’s atmo- sphere (a third of a Saturnian year) means that we have been able to track the seasonal evolution of atmospheric temper- atures, chemistry and cloud opacity over almost every sea- son, from solstice to solstice and from perihelion to aphelion. Cassini has built upon the decades-long ground-based record to observe seasonal shifts in atmospheric temperature, find- ing a thermal response that lags behind the seasonal insola- tion with a lag time that increases with depth into the at- mosphere, in agreement with radiative climate models. Sea- sonal hemispheric contrasts are perturbed at smaller scales by atmospheric circulation, such as belt/zone dynamics, the equatorial oscillations and the polar vortices. Temperature asymmetries are largest in the middle stratosphere and be- come insignificant near the radiative-convective boundary. 1 Cassini has also measured southern-summertime asymme- tries in atmospheric composition, including ammonia (the key species for the topmost clouds), phosphine and para- hydrogen (both disequilibrium species) in the upper tropo- sphere; and hydrocarbons deriving from the UV photoly- sis of methane in the stratosphere (principally ethane and acetylene). These chemical asymmetries are now altering in subtle ways due to (i) the changing chemical efficiencies with temperature and insolation; and (ii) vertical motions associated with large-scale overturning in response to the seasonal temperature contrasts. Similarly, hemispheric con- trasts in tropospheric aerosol opacity and coloration that were identified during the earliest phases of Cassini’s ex- ploration have now reversed, suggesting an intricate link be- tween the clouds and the temperatures. Finally, comparisons of observations between Voyager and Cassini (both observ- ing in early northern spring, one Saturn year apart) show tantalising suggestions of non-seasonal variability. Disentan- gling the competing effects of radiative balance, chemistry and dynamics in shaping the seasonal evolution of Saturn’s temperatures, clouds and composition remains the key chal- lenge for the next generation of observations and numerical simulations. 10.1 Introduction We can achieve a greater understanding of any complex sys- tem by studying how that system evolves with time. Sat- urn, with its 26.7◦ Earth-like axial tilt, 29.5-Earth-year or- bital period and orbital eccentricity of 0.057 (perihelion near northern winter solstice, aphelion near northern summer solstice), is our closest and best example of a seasonally- variable giant planet atmosphere, in contrast with Jupiter (negligible seasonal influences from the 3.1◦ tilt), Uranus (extreme seasonal contrasts from the 98◦ tilt) and Neptune (slow evolution due to the 165-year period). Furthermore, Cassini has provided our best opportunity in a generation to study the seasonal evolution of a giant planet atmo- sphere and the influence of temporal variations in sunlight on the atmospheric temperatures, clouds and chemistry. Sat- urn completed 13.5 orbits of the Sun between Galileo’s first glimpses of the ringed planet and its “strange appendages” (July 1610) and the Cassini spacecraft’s arrival at Saturn orbit in June 2004. Studies of seasonally-changing atmo- spheric properties have only been possible for the past four 2 Fletcher, Greathouse, Guerlet, Moses & West decades (1.3 Saturn years, coinciding with the improved ca- pabilities of ground-based infrared remote sensing), with the Cassini orbiter providing reconnaissance for the last third of a Saturn year (Fig. 10.1). In this chapter we review our current knowledge of Saturn’s temporally-variable thermal structure, composition and aerosols, from the churning tro- pospheric cloud decks to the middle atmosphere. What properties might we expect to be time-variable on a giant planet? Atmospheric temperatures are governed by a delicate balance between Saturn’s internal heat source and heating from the Sun. Both diabatic (balance between ra- diative heating and cooling) and adiabatic (atmospheric mo- tions redistributing energy vertically and horizontally) forc- ings govern the spatial variations of temperature, so that we might expect warm summers and cool winters albeit with a phase lag compared to the solstices due to the atmospheric inertia. The seasonal diabatic heating will generate hemi- spheric temperature gradients that are superimposed on the belt/zone structure, which then drive atmospheric transport to redistribute excess energy. These hemispheric tempera- ture contrasts should remain in balance with Saturn’s zonal wind system (via the thermal wind relation), such that ver- tical shears on the zonal jets could vary with time. Sat- urn’s chemical composition may also vary with season, as variations in ultraviolet insolation drive ionization and pho- todissociation rates, which in turn govern the populations of hydrocarbons and hazes derived from methane photoly- sis in the stratosphere, and the distribution of key volatiles (e.g., NH3) and disequilibrium species (e.g., PH3) in the upper troposphere. Photochemically-produced hazes could sediment downward to serve as cloud-condensation nuclei for condensible volatiles, which in turn could cause aerosol and cloud properties to vary with time (e.g., Fig. 10.1). Finally, dynamic phenomena in the weather layer (vortices, storms, plumes and waves) respond to modifications of atmospheric stratification, so that seasons could help modulate meteoro- logical activity. Cassini’s longevity, coupled with a long baseline of ground-based observing, allows us to monitor each of these processes during a Saturn year to provide a four-dimensional understanding of Saturn’s troposphere and stratosphere. Seasons are indicated by the planetocentric solar longitude (Ls), from 0◦ at northern spring equinox (1980, 2009), to 90◦ at the northern summer solstice (1987, 2017), 180◦ at northern autumnal equinox (1995) and 270◦ at the north- Figure 10.1 Saturn’s changing insolation from 2006 to 2012, three years on either side of the northern spring equinox. The colors of Saturn’s tropospheric clouds and hazes can be seen shifting as northern winter becomes northern spring. Compiled from Cassini images courtesy of NASA/JPL-Caltech. ern winter solstice (2002). Saturn’s southern summers re- ceive greater insolation than northern summers (perihelion occurs near Ls = 280◦, July 2003; aphelion at Ls = 100◦, April 2018). In addition to ground-based remote sensing since the mid-1970s (southern summer), Saturn’s seasonal asymmetries have been observed by four visiting spacecraft: observations by Pioneer 11 and Voyagers 1 and 2 were clus- tered around northern spring equinox (1979-1981); Cassini entered Saturn orbit in June 2004, two years after southern summer solstice (Ls = 293◦) and aims to complete its mis- sion at northern summer solstice (Ls = 93◦ in September 2017). This chapter is organized as follows: Section 10.2 reviews investigations of Saturn’s tropospheric and stratospheric temperature field, comparing them to climate models to un- derstand the influence and variability of atmospheric circu- lation. In Section 10.3 we review observations and chemical modeling of the spatial distributions and variability of key atmospheric species. Section 10.4 reviews the characteris- tics of Saturn’s clouds and hazes, focusing on their time- variable properties, before we review unanswered questions in Section 10.5. Planetographic latitudes are assumed un- less otherwise stated. We confine our discussion to Saturn’s seasonally variable troposphere and stratosphere; the ther- mosphere and ionosphere will be discussed in Chapter 9. 10.2 Seasonally-evolving thermal structure 10.2.1 Pre-Cassini studies Saturn’s temperature structure in the cloud-forming re- gion and the lower troposphere is expected to follow an adiabatic gradient, with the lapse rate dominated by the heat capacity of the hydrogen-helium atmosphere but with small contributions from latent heat released by the con- densation of volatile species (NH3, NH4SH and H2O) and lagged conversion between the two different spin isomers (ortho- and para-H2) of molecular hydrogen (see the re- view by Ingersoll et al., 1984, and Section 10.3.2). At lower 16 MAR 2006 25 FEB 2011 17 JUL 2007 18 OCT 2009 6 MAY 2012 5 DEC 2010 Saturn’s seasonally changing atmosphere: thermal structure, composition and aerosols 3 pressures above the radiative-convective boundary (350-500 mbar, Fletcher et al., 2007a), the atmospheric opacity drops sufficiently to allow efficient cooling by radiation and the temperatures deviate from the adiabat, becoming more sta- bly stratified in the upper troposphere towards the tem- perature minimum (the tropopause near 80 mbar). Atmo- spheric heating due to short-wavelength sunlight absorption by methane and aerosols causes temperatures to rise again in the stratosphere, balanced by long-wavelength cooling from methane (upper stratosphere), ethane and acetylene (middle and lower stratosphere), and the collision-induced hydrogen-helium continuum (upper troposphere and lower stratosphere). It is the seasonal dependence of the atmo- spheric heating (and cooling, via chemical changes) that causes Saturn’s upper tropospheric and stratospheric tem- peratures to vary considerably with season. Seasonal temperature variations were observed as asym- metries in emission measured at thermal infrared wave- lengths, first detected in ground-based images from strato- spheric ethane near 12 µm (Gillett and Orton, 1975; Rieke, 1975) and methane near 8 µm (Tokunaga et al., 1978) during southern summertime conditions, revealing enhanced emis- sion from the southern pole. Tropospheric contrasts in the 17-23 µm region were still present but more muted (Cald- well et al., 1978; Tokunaga et al., 1978), consistent with a seasonal response that becomes weaker with depth. The first spacecraft measurements of Saturn’s temperatures by Pioneer 11 (1979, Ls = 354◦) revealed no tropospheric ther- mal asymmetries between 10◦N and 30◦S just before the northern spring equinox (Orton and Ingersoll, 1980). How- ever, inversions of Voyager 1 (1980, Ls = 8.6◦) and 2 (1981, Ls = 18.2◦) IRIS 14-50 µm spectra revealed tropospheric temperature asymmetries (Hanel et al., 1981, 1982; Con- rath and Pirraglia, 1983; Conrath et al., 1998), with the north cooler than the south at 150-200 mbar shortly after northern spring equinox. The equinoctial timing of these observations suggested a delay between increased insolation (the solar forcing) and the atmospheric response (Cess and Caldwell, 1979) as a consequence of the high thermal inertia of the atmosphere. At tropospheric depths of 500-700 mbar this inertia, and hence the lagged response to the solar forc- ing, becomes so large that Saturn’s northern and southern hemisphere temperatures do not show significant asymme- tries (Conrath and Pirraglia, 1983). Seasonal amplitudes are largest in the stratosphere, as we shall describe below. Twenty-three years would pass before another spacecraft reached the Saturn system, but ground-based studies con- tinued to provide insights into Saturn’s seasons, particu- larly with the advent of 2D mid-infrared detector technolo- gies and high-resolution spectroscopy (see the review by Or- ton et al., 2009). Saturn’s north polar regions exhibited en- hanced methane and ethane emission by early northern sum- mer (observations by Gezari et al., 1989, in March 1989, Ls = 104◦, Fig. 10.2) just like the south pole had in south- ern summer, although this enhanced emission was not read- ily observable in the troposphere (observations in December 1992, Ls = 146◦ by Ollivier et al., 2000). When southern summer returned in the early 2000s, Greathouse et al. (2005) used high-resolution CH4 emission spectroscopy to derive Figure 10.2 Stratospheric thermal emission imaged by ground-based facilities in March 1989 (near northern summer solstice, Gezari et al., 1989) and February 2004 (near southern summer solstice, Orton and Yanamandra-Fisher, 2005), showing enhanced emission from the summer pole and similarities in the seasonal response. a stratospheric temperature asymmetry at southern sum- mer solstice (2002, Ls = 268◦, the south pole 10 K warmer than the equator) and an asymmetry that weakened with in- creasing depth; while Orton and Yanamandra-Fisher (2005) presented high resolution 7-25 µm images that revealed the southern summer hemisphere in exquisite detail (February 2004, Ls = 287◦, Fig. 10.2). Specifically, they observed a 15 K temperature contrast from the equator to the south pole at 3 mbar, a sharp temperature gradient near 70◦S (the edge of the south polar warm hood) and an intense tropospheric hotspot associated with the south polar cyclone poleward of 87◦S. These observations confirmed that Cassini would ob- serve a stark asymmetry between the northern winter and southern summer hemispheres upon arrival in 2004. 10.2.2 Cassini observations in southern summer Cassini’s great advantage is the ability to view both the northern and southern hemispheres near-simultaneously, whereas Saturnian winter is forever hidden to an Earth- based observer. Tropospheric and stratospheric tempera- tures are provided via a combination of infrared remote sens- ing, radio occultation and ultraviolet stellar occultations. Nadir 7-1000 µm spectroscopy with the Composite Infrared Spectrometer (CIRS, Flasar et al., 2004) measures tropo- spheric temperatures from the tropopause down to the top- most cloud decks (approximately 80-800 mbar) and strato- spheric temperatures (from methane emission) in the 0.5-5.0 mbar range (Flasar et al., 2005; Fletcher et al., 2007a). CIRS limb observations (Fouchet et al., 2008; Guerlet et al., 2009) complement the vertical coverage of nadir observations by constraining the stratospheric temperature profile between Northern Summer, March 1989, Ls=104o Southern Summer, February 2004, Ls=287o 7.8 µm 11.6 µm 12.4 µm Gezari et al., (1989), NASA/IRTF 17.6 µm 8.0 µm Orton & Yanamandra-Fisher (2005), Keck I/LWS 4 Fletcher, Greathouse, Guerlet, Moses & West 20 mbar and 1 µbar with a vertical resolution of 1-2 scale heights, at the expense of poorer spatial and temporal cover- age. Radio-occultations of Cassini by Saturn (Schinder et al., 2011), available at a limited number of latitudes, achieve the highest vertical resolution (6-7 km, a tenth of a scale height) and constrain the temperature-pressure profile between 1500 mbar and 0.1 mbar. Cassini’s prime mission provided a snapshot of Saturn’s hemispheric temperature asymmetries in late southern sum- mer (2004-2008, Ls = 293 − 345◦, Flasar et al., 2005; Fletcher et al., 2007a, 2008; Guerlet et al., 2009), as shown in Fig. 10.3(a). The summer pole was found to be 40 K warmer than the winter pole at 1 mbar, with the contrast de- creasing with increasing pressure (Fletcher et al., 2007a). In- triguingly, the latitudinal asymmetry appeared to be smaller (≈ 24 K) at 0.1 mbar and smaller still at 0.01 mbar(Guerlet et al., 2009). The tropopause was around 10 K cooler (and slightly higher in altitude) in the winter hemisphere than the summer hemisphere (Fletcher et al., 2007a). Below the tropopause, the lapse rate increased with depth until reach- ing the dry adiabat at approximately 350-500 mbar (Lindal et al., 1985; Fletcher et al., 2007a), likely indicating the lo- cation of the radiative-convective boundary that separates the stably stratified upper troposphere from the convec- tive deeper troposphere. This lapse rate change occurred at higher pressures (400-500 mbar) in the summer hemi- sphere than in the northern hemisphere (350-450 mbar), due to the greater penetration of solar heating in the southern hemisphere. Temperature asymmetries became negligible for p > 500 mbar. Between the radiative-convective boundary and the tropopause, inversions of far-infrared spectra revealed an inflection point in the tropospheric temperature structure in the 100-300 mbar region. This perturbation was referred to as the ‘knee’ (Fletcher et al., 2007a), suggestive of heating that was localised in altitude within Saturn’s tropospheric haze. It was first noted in Voyager/IRIS retrievals at equa- torial and southern latitudes (Hanel et al., 1981) and in Voyager radio occultations at 3◦S and 74◦S (Lindal et al., 1985). The pressure level and magnitude of this temperature perturbation varied strongly with latitude, being enhanced in southern summer and weak or absent in the northern winter hemisphere. The knee was higher and weaker over the equator, and showed local maxima at 15◦N and 15◦S associated with the warm equatorial belts (see Fig. 10.20 at the end of this chapter). Fletcher et al. (2007a) concluded that this was a radiative effect due to solar absorption by aerosols in the upper troposphere, and explained the asym- metry in terms of both seasonal insolation and the variable distributions of tropospheric aerosols (later confirmed by radiative climate modeling by Friedson and Moses, 2012; Guerlet et al., 2014). The relationship between the ‘knee’ and the upper tropospheric haze is discussed in Section 10.4. Saturn’s temperature distribution in Fig. 10.3 reveals the influence of dynamics as well as radiative balance. The global temperature asymmetries are superimposed onto small-scale latitudinal contrasts between the cool zones (anticyclonic shear regions equatorward of prograde jets) and warmer Figure 10.3 Zonal mean temperatures derived from Cassini/CIRS nadir infrared spectroscopy at three different epochs, representing a snapshot of the temperature field in (a) late southern summer (2005-2006, Ls ≈ 313◦); (b) northern spring equinox (2009-2010, Ls ≈ 5◦) and (c) early northern spring (2013-2014, Ls ≈ 52◦). Adapted and updated from Fletcher et al. (2010) and Fletcher et al. (2015). -90-75-60-45-30-150153045607590Planetographic Latitude1.0000.1000.0100.001Pressure [bar](c) 2012-201390909090100100100110110110110120120120120130130130130140 -90-75-60-45-30-150153045607590Planetographic Latitude1.0000.1000.0100.001Pressure [bar](b) 2009-201080909090100100100100110110110110120120120120130130130130140140150 -90-75-60-45-30-150153045607590Planetographic Latitude1.0000.1000.0100.001Pressure [bar](a) 2005-200680909090100100100110110110120120120120130130130130140140150160 Saturn’s seasonally changing atmosphere: thermal structure, composition and aerosols 5 belts (cyclonic shear regions poleward of prograde jets) (Conrath and Pirraglia, 1983; Fletcher et al., 2007a). Note that the correlations between this belt/zone structure (de- fined in terms of the zonal jets and tropospheric tempera- tures) and the cloud reflectivity is not as clear-cut as for Jupiter. For example, bands of low reflectivity are often narrow and located close to the prograde jet peaks (e.g., Vasavada et al., 2006), and do not appear correlated with the thermal structure. Saturn’s polar troposphere features long-lived cyclonic ‘hot spots’ located directly at each pole irrespective of season, and the northern hexagonal jet at 77◦N is related to a hexagonal warm polar belt poleward of this latitude (Fletcher et al., 2008). At lower pressures, the polar stratosphere features a warm ‘polar hood’ in summer that is absent in winter, one of the most extreme examples of a seasonal phenomenon on Saturn (Fletcher et al., 2008). In Saturn’s tropical stratosphere, contrasts between the equa- tor and neighboring latitudes are observed to oscillate with time and altitude. This ‘semi-annual oscillation’ in the ther- mal structure implies a strong vertical shear of the zonal wind at the equator and is reminiscent of the Quasi-Biennal Oscillation in the Earth’s stratosphere, a dynamical phe- nomenon driven by wave-zonal flow interactions (Fouchet et al., 2008; Orton et al., 2008). 10.2.3 Seasonal evolution of temperatures Saturn’s thermal structure during southern summer was re- viewed by Del Genio et al. (2009), but Cassini has since revealed how the global temperature structure has evolved with season through northern spring equinox. Fletcher et al. (2010) used CIRS observations from 2004 to 2009 (Ls = 297− 358◦) to determine Saturn’s upper tropospheric and stratospheric temperature variability, finding (i) strato- spheric warming of northern mid-latitudes by 6-10 K at 1 mbar as they emerged from ring shadow into springtime con- ditions; (ii) southern cooling by 4-6 K (both at mid-latitudes and within the south polar stratospheric hood poleward of 70◦S) and a resulting decrease in the 40-K asymmetry be- tween the hemispheres that was present in 2004; and (iii) a tropospheric response to the seasonal insolation shifts that seemed to be larger at the locations of the broadest ret- rograde jets. The ‘flattening’ of the summertime tempera- ture asymmetry (by Saturn’s equinox, northern and south- ern mid-latitude 1-mbar temperatures were both in the 140- 145 K range) followed the expectations of a radiative climate model (Greathouse et al., 2010, and see below), albeit per- turbed by the equatorial oscillation, polar vortex dynamics and the belt/zone structure. Sinclair et al. (2013) extended this analysis to include observations in 2010 (Ls = 15◦) and observed the continued northern warming and south- ern cooling, particularly intense within Saturn’s south polar region (≈ 17 K between 2005 and 2010). At higher strato- spheric altitudes, Sylvestre et al. (2015) extended the analy- sis of Guerlet et al. (2009) by considering limb spectroscopy in 2010-2012, finding a seasonal trend consistent with the nadir data at 1 mbar, but with smaller variations at lower pressures. For example, mid-latitude temperatures near 0.1 Figure 10.4 Zonal mean temperatures in the stratosphere (1 mbar) and troposphere (330 mbar) as derived from nadir Cassini/CIRS spectra over the duration of the Cassini mission. Note that the absence of high-latitude measurements between 2010 and 2012 is due to Cassini’s near-equatorial orbit at that time, preventing nadir observations of the poles. Updated from Fletcher et al. (2010) and Fletcher et al. (2015). mbar have remained approximately constant between 2005 and 2010. Cassini’s monitoring of the seasonal temperatures has now extended well into northern spring, and most recently Fletcher et al. (2015) extended the work of Fletcher et al. (2010) and Sinclair et al. (2013) to 2014 (Ls = 56◦). The zonal mean temperature as a function of latitude and pres- sure is shown for three epochs in Fig. 10.3, and all CIRS nadir retrievals are shown as a function of time in Fig. 10.4. Focusing on the highest latitudes, tropospheric contrasts be- tween the cool polar zones (80 − 85◦ latitude) and warm polar belts (near 75 − 80◦ latitude) have varied over the ten-year span of observations, indicating changes to the ver- tical shear on the zonal jets via the thermal wind equation (Fig. 10.3). The warm south polar stratosphere has cooled dramatically by ≈ 5 K/yr, mirrored by warming of a sim- ilar magnitude in the north. However, while the south po- lar region was isolated by a strong thermal gradient near (a) Stratospheric Temperatures (p=1 mbar)-90-75-60-45-30-150153045607590Planetographic Latitude120140160180Temperature [K](b) Tropospheric Temperatures (p=330 mbar)-90-75-60-45-30-150153045607590Planetographic Latitude9092949698100102104Temperature [K] 300320340360380400Heliocentric Longitude 2005200620072008200920102011201220132014Year 300320340360380400Heliocentric Longitude 2005200620072008200920102011201220132014Year 6 Fletcher, Greathouse, Guerlet, Moses & West 75◦S during the height of summer, no similar boundary was apparent near 75◦N during spring despite rising tempera- tures towards the north pole at Ls = 56◦ (the last published data), suggesting that the northern summer vortex has yet to form (see Fig. 10.5, Chapter 12 and Fletcher et al., 2015). The peak stratospheric warming in the north was occurring at lower pressures (0.5-1.0 mbar) than the peak stratospheric cooling in the south (1-3 mbar). Fig. 10.5 demonstrates that north polar minima in stratospheric temperatures were de- tected in 2008-2010 (lagging one season, or 6-8 years, be- hind winter solstice); south polar maxima appear to have occurred before the start of the Cassini observations (1-2 years after summer solstice). 10.2.4 Seasonal climate modeling Seasonal climate modeling is an attempt to create models that accurately predict the temporal evolution of a planet’s thermal structure as a function of altitude, latitude and lon- gitude. Diurnal temperature variations are not expected due to Saturn’s high thermal inertia, the relatively low amount of solar forcing at Saturn’s distance (≈ 100 times less than at Earth), and Saturn’s fast rotation (10 hour 39 minute long days), and current models bear this out (Greathouse et al., 2008; Guerlet et al., 2014). This fact allows modelers to reduce the complexity of the problem and focus on the temporal evolution of temperatures as a function of altitude and latitude only, calling then for detailed 2-D time variable models. The cooling of Saturn’s stratosphere is dominated by the radiative emissions from C2H2 and C2H6 at pressures lower than 5 mbar, along with some cooling due to the ν4 band of CH4 near 8 µm. Cooling due to other hydrocarbons is esti- mated to account for no more than 5% of the total radiative cooling rate (Guerlet et al., 2014). Tropospheric and lower stratospheric cooling occurs via emission from the H2-H2 and H2-He collision-induced continuum. Atmospheric heat- ing is due to the absorption of sunlight primarily by CH4 and aerosols. C2H2 and C2H6 are both byproducts of CH4 pho- Figure 10.5 Temperatures at 1 mbar derived in the polar regions of Saturn (Fletcher et al., 2015), exposed to the most severe insolation changes over a saturnian year. Cassini measurements are compared to the model predictions of Greathouse et al. (2010), Guerlet et al. (2014) and Friedson and Moses (2012). Although discrepancies in the absolute temperatures are evident, the timing of the polar maxima/minima in temperatures is reasonably well reproduced. tochemistry (for details see Section 10.3) which is initiated at the top of the atmosphere just below the CH4 homopause. Since C2H2 and C2H6 are the dominant radiative coolants, and their distribution controls the extent and direction of radiant energy to space, their distribution can alter strato- spheric temperatures, which in turn can induce circulation patterns that serve to redistribute the molecules. While this complicated interchange continues, ongoing photochemical processes are constantly making their own adjustments to the abundances of C2H2 and C2H6. This interconnected- ness means that to produce a truly accurate seasonal cli- mate model one needs to include accurate calculations of the absorption and emission of radiation, photochemistry, and dynamics. Any one of these would make for a com- plex model on its own and the combination of all three is a worthy goal. While this level of complexity is currently not achieved, we show below that much has been accomplished in all three disciplines, and the most recent incarnations of seasonal models are moving ever closer to this goal. The first generation of seasonal climate models for Saturn were inspired by the first observations of seasonal tempera- ture asymmetries in the 1970s and Pioneer-Voyager epoch. Radiative-convective equilibrium models in the 1970s (Cald- well, 1977; Tokunaga and Cess, 1977; Appleby and Hogan, 1984) showed that the solar absorption by methane in the visible and near-infrared could explain the temperature in- version identified from ethane limb brightening (Gillett and Orton, 1975). Early models of Saturn’s stratospheric tem- perature response (Cess and Caldwell, 1979; Carlson et al., 1980) were extended into the troposphere by B´ezard et al. (1984), and indicated that an optimal fit to the measured 20062008201020122014Time-90-80-70-60Planetographic Latitude13113513513913914314314714715115515560708090Planetographic Latitude30032034002040Planetocentric Solar Longitude [Ls/deg](c) Guerlet Model11511911912312312712713113113513513913914360708090Planetographic Latitude30032034002040Planetocentric Solar Longitude [Ls/deg](b) Greathouse Model12012012412412412812813213213613614014014414414814820062008201020122014Time-90-80-70-60Planetographic Latitude12813213613614014014414414814815215215615616020062008201020122014Time-90-85-80-75-70-65-60Planetographic Latitude13013413813814214214614615015015415415816260657075808590Planetographic Latitude30032034002040Planetocentric Solar Longitude [Ls/deg](a) Cassini/CIRS Measured Temperatures11911912312312312712712713113113513513913914320062008201020122014Time-90-85-80-75-70-65-60Planetographic Latitude14214614615015415415860657075808590Planetographic Latitude30032034002040Planetocentric Solar Longitude [Ls/deg](d) Friedson & Moses Model118122122126130130134134138138142146Saturn’s seasonally changing atmosphere: thermal structure, composition and aerosols 7 tropospheric temperatures required an additional source of opacity (potentially from tropospheric aerosols). B´ezard and Gautier (1985) incorporated a non-gray treatment of radia- tive transfer to construct a more sophisticated model, ac- counting for seasonal changes in solar forcing, ring obscu- ration and planetary oblateness. Though they were limited by both the lack of some lab measurements (several near- infrared bands of methane had yet to be measured in the lab, B´ezard and Gautier, 1985) and data on Saturn (detailed measurements of temperatures versus altitude, latitude, and time along with the variations of mixing ratio for C2H2 and C2H6 as a function of time, altitude and latitude), they were able to produce models that to first order reproduced early ground based observations and much of the data retrieved by Voyager 1 and 2, including the observation that the thermal inertia increased with depth such that there was no seasonal modulation of the deepest atmospheric temperatures. These models were one-dimensional in nature, requiring the user to model each latitude of interest and then compile the results to show how the global temperatures changed over time. Conrath et al. (1990) produced more sophisticated radiative- convective-dynamical models of Saturn’s troposphere and stratosphere for a comparison to Voyager results, and Bar- net et al. (1992) included the effects of ring shadowing and ring thermal emission. The radiative climate models have evolved with time as (i) the spectroscopic parameters of opacity sources (methane, hydrocarbons and hazes) have become better constrained; and (ii) the spatial distributions of the key infrared coolants (ethane and acetylene) have become better known; and (iii) the vertical distribution of clouds and hazes, which have a substantial contribution to the radiative budget, have been revealed. The richness and complexity in the Cassini sea- sonal dataset prompted the evolution of a new generation of radiative climate models, taking advantage of improvements in laboratory measurements and computational capabilities. The first such model (Greathouse et al., 2010, hereafter TG) was a purely radiative seasonal model of Saturn’s strato- sphere, with the capability of assuming any vertical, merid- ional, or temporal variation of the key hydrocarbon coolants. This model is one dimensional in nature, one latitude over time, requiring multiple runs to compile results from differ- ent latitudes into a 2-D representation of Saturn’s strato- spheric seasonal evolution. However, as initially planned, this model was absorbed as a module within the Explicit Planetary Isentropic-Coordinate Global Circulation Model (EPIC GCM), allowing EPIC to accurately calculate the radiative heating and cooling rates while accounting for dy- namics (Dowling et al., 2010). The second model, the Outer Planet General Circulation Model (OPGCM), is a strato- spheric and tropospheric seasonal dynamical model which efficiently and accurately accounts for the radiative heating and cooling while also tracking circulation caused by the seasonal forcing in 3-dimensions (Friedson and Moses, 2012, hereafter FM). Finally, the most recent seasonal model is the stratospheric and tropospheric radiative-convective model produced by Guerlet et al. (2014) (hereafter SG). These three modern models are based largely on the same underlying assumptions, and so we compare each to Cassini measurements of the 1-mbar temperature contrasts in Fig. 10.6 to illuminate interesting seasonally forced effects and lo- calized dynamical effects. The TG model assumes the merid- ional and vertical mixing ratios for C2H2 and C2H6 as mea- sured by Guerlet et al. (2009), and assumes that this distri- bution is fixed with time. The TG model is the coarsest (in latitude sampling) of the three models, and does not include the effects of dynamics (dashed line in Fig. 10.6). The FM model uses the latitudinal average of the vertical profiles of C2H2 and C2H6 as measured by Guerlet et al. (2009) and holds these vertical profiles constant with time and latitude, even though they are accounting for dynamical redistribu- tion of heat via diffusion and advection (dotted line in Fig. 10.6). The SG model uses the average of the vertical mixing ratio profiles for C2H2 and C2H6 between 40◦N and 40◦S (planetocentric) as taken from Guerlet et al. (2009) (solid line in Fig. 10.6). These mean vertical profiles are assumed constant with latitude and time. None of the models feature the sharp upturns in C2H2 and C2H6 abundances at high latitudes (nor the added contribution of polar stratospheric aerosols), and so are unlikely to reproduce the radiative en- ergy balance near the poles (Fletcher et al., 2015; Guerlet et al., 2015). A comparison between the three models reveals how the different assumptions manifest themselves in the final pre- dicted temperatures and how those final temperatures com- pare to the measurements. Given these assumptions, one would expect that the TG and SG models should return very similar results with the differences between the two being due to the different C2H2 and C2H6 distributions assumed, which is in fact what we see in Fig. 10.6. Similarly the com- parison between the FM and the SG/TG models show the effects of dynamical diffusion and advection of heat on tem- peratures, as the radiative heating and cooling scheme in all three models are quite similar. Where the FM model (dot- ted) is significantly cooler/hotter than the SG model (solid), this could be related to upwelling/downwelling in the FM model causing adiabatic expansion/compression. However, this assumes identical treatment of radiative balance be- tween the three models, and neglects other sources of heating and cooling (e.g., interactions between waves and the mean flow). As can be seen in Fig. 10.6, all three models do an im- pressive job of reproducing the main pole to pole temper- ature variations as measured by Cassini/CIRS nadir obser- vations (Fletcher et al., 2010, 2015). The strong southern summer temperature gradient seen in the top panel of Fig. 10.6 (Ls = 306◦) transitioning to a more equal north/south temperature profile by mid-northern spring (Ls = 46◦) is well tracked by the models, with the exception of the FM model which does not cool as fast at high southern lati- tudes. However, several discrepancies stand out. The tem- peratures between ±20◦ planetocentric latitude are com- pletely different compared to the three model predictions. Temperatures in this region are significantly altered by the dynamical phenomenon known as Saturn’s Semi-Annual Os- cillation (SSAO) (Fouchet et al., 2008; Orton et al., 2008; Guerlet et al., 2011). The TG and SG models are purely radiative and thus could not possibly match the temper- 8 Fletcher, Greathouse, Guerlet, Moses & West Figure 10.6 Saturn’s meridional temperature variations at 1 mbar at 3 different epochs derived from nadir Cassini/CIRS observations, compared with model predictions from Guerlet et al. (2014) as the solid line, Friedson and Moses (2012) as the dotted line, Greathouse et al. (2010) as the dashed line. ature variation of this dynamical phenomenon. While the FM model could possibly reproduce such a dynamical event, they suggest that the coarseness of their vertical grid may have made it impossible to resolve the waves needed to force the SSAO. Another complication is the heating associated with Saturn’s stratospheric vortex (Fletcher et al., 2012b). The nominal seasonal trend of stratospheric temperatures at mid-northern latitudes was disrupted by the production of the stratospheric vortex (known as the ‘beacon’, see Chap- ter 13) in late 2010, and although this region was avoided to produce the measured temperatures in Figure 10.6, the storm-induced heating of the northern mid-latitudes is the likely cause for the mismatch with the model predictions in 2013. One of the starkest discrepancies between the models and the data in Fig. 10.6 appears at the polar latitudes. The 1-mbar temperatures measured by Cassini at the northern and southern poles are compared to the three models in Fig. 10.5, to assess their capabilities for reproducing the absolute temperatures and the timing of the polar minima/maxima in temperatures. Although the timing of the south polar max- imum is reproduced in the models (1-2 years after summer solstice), the south polar temperatures are elevated over the predictions of all three models, and the temperature range is larger in the data than in the model predictions. At the north pole, the models are unable to reproduce the obser- vation that the coldest temperatures are identified between 75− 80◦N, rather than at the north pole itself, although the 6-8 year phase lag between the winter solstice and the coldest temperatures is consistent with the models. It is suspected that these differences are due to circulations associated with polar vortices, or the increasing importance of stratospheric aerosols or enhanced hydrocarbons in the radiative budget at high latitudes (Fletcher et al., 2015). Interestingly, the FM model stays warmer for longer at the south pole, over- predicting the southern temperatures in 2013-14. This large temperature increase poleward of 60◦S seen in the FM model is the result of extensive downwelling of gas, whereas obser- vations show that this region has in fact cooled substantially for the duration of Cassini’s observations (Fig. 10.5). While the downward advection of material naturally heats the gas by adiabatic compression, one would also expect an increase in cooling due to an increase in C2H2 and C2H6 mixing ra- tio. If advection of material to serve as coolants (i.e., C2H2 and C2H6 concentrations, stratospheric aerosols) could be included in the FM model, it might help explain the mis- match between the data and model. We now return to the curious fact that the seasonal response measured at higher stratospheric altitudes using CIRS limb observations (0.01-0.1 mbar) by Guerlet et al. (2009) and Sylvestre et al. (2015) appears to be smaller than that at 1 mbar. Fig. 10.7 shows the comparison of the three models to the limb observations in southern summer (Ls = 313◦) at 0.1 and 0.01 mbar. Although there is remark- able agreement with the temperatures in southern summer (particularly at 0.01 mbar), the measured temperatures are much higher than model predictions in the north. This could be the result of several processes - near 25◦N the effect of ring shadowing may have been overestimated, or significant mixing/advection might wash out the ring shadow effects seen in the purely radiative models. Although the FM model does not offer useful results at the 0.01-mbar level, it does at the 0.1-mbar level. There we can see that the cool region of temperatures due to the ring shadow (15-30◦N) seen in the TG and SG models is warmed significantly by subsidence in the FM model. It is likely this effect occurs at 0.01 mbar as well, and may help to explain the model-data mismatch at the higher altitudes. Away from the ring-shadowed re- gion, additional sources of upper atmospheric heating, such as gravity wave breaking or from some interaction with the thermosphere, might be the culprit for the warmer temper- atures over the rest of the northern hemisphere. Figs. 10.5, 10.6 and 10.7 demonstrate how well the mod- els perform relative to the measured data and at specific instances in time. It is also useful to compare how the mod- els behave over an entire Saturn year at a single latitude to learn about seasonal phase lags. In Fig. 10.8 we plot the pre- dicted temperatures of Saturn from all three models at 75◦S planetocentric latitude. The solid lines represent the temper- atures at 1 mbar while the dashed are from 0.1 mbar. All of the models show that the peak temperature in southern summer occurs about 30◦ of Ls after the southern summer 120130140150160170Temperature (K) 120130140150160170Temperature (K)2005−2006 (LS=306) 120130140150160170Temperature (K)2010−2011 (LS= 11)−90−60−300306090Planetocentric Latitude (deg)120130140150160170Temperature (K)2013−2014 (LS= 46)Saturn’s seasonally changing atmosphere: thermal structure, composition and aerosols 9 Figure 10.8 Model/model comparison at 75◦S between the Guerlet et al. (2014) model in red, Friedson and Moses (2012) model in blue and the Greathouse et al. (2010) model in green. The solid lines represent the change in temperature over time at the 1.0-mbar level while the dashed lines are for the 0.1-mbar level. The vertical black lines represent the times of Northern Summer Solstice, Northern Autumnal Equinox, and Northern Winter Solstice, from left to right respectively. complicated dynamical model of FM. Advection of the hy- drocarbon abundances along with the heat may be signifi- cant, in addition to the seasonally shifting contribution of tropospheric and stratospheric aerosols and other heating sources (e.g., wave activity) not currently accounted for in the theoretical models. Hopefully, with the advent of new 3D GCMs with proper radiative forcing and higher spatial resolution (required for resolving waves and instabilities that influence the circulation), a model will be produced in the near future with a self consistent radiative, chemical and dynamical system. 10.2.5 Implications of seasonal temperature change The seasonal shifting of hemispheric temperature asymme- tries has implications for both Saturn’s global energy budget and atmospheric circulation (e.g., polar atmospheric circu- lations, wave phenomena, tropospheric overturning, storm perturbations, Hadley circulations). Li et al. (2010) demon- strated that the shifting thermal asymmetries measured by CIRS had implications for Saturn’s emitted power (average of 4.952±0.035 W/m2), which decreased by 2% from 2005 to 2009 and was found to be larger in the south (5.331 ± 0.058 W/m2) than in the north (4.573 ± 0.014 W/m2). The peak contribution to the outgoing emission (near 320 mbar) was found to be shallower in the northern winter hemisphere than in southern summer, potentially due to seasonal asym- metries in aerosol content in the upper troposphere, or the details of the upper tropospheric temperature structure. Thus emitted power measurements over a full Saturn year are desirable to properly constrain the energy budget. Meridional temperature gradients are related to the ver- tical shear on the zonal winds via the thermal wind equa- tion, as the Coriolis forces on the winds should remain in geostrophic balance with the horizontal pressure gradients. Figure 10.7 Model-data comparisons for temperatures at the 0.01 and 0.1 mbar level. The observed temperatures were retrieved using Cassini/CIRS limb observations in late southern summer Guerlet et al. (2009). Models are identified as the solid line for Guerlet et al. (2014), the dotted line for Friedson and Moses (2012), and the dashed line for Greathouse et al. (2010). Note the Friedson & Moses model is not shown at the 0.01 mbar level as it does not accurately model Saturn’s atmosphere at that pressure level. solstice (Ls = 270◦) at 1 mbar, but only about 15◦ of Ls at 0.1 mbar. This is due to the lower-pressure level hav- ing less mass, and thus less thermal inertia, in addition to the fact that the atmospheric coolants are more abundant at lower pressures. The models all suggest that the maxi- mum and minimum in temperatures are more extreme and can change more rapidly at lower pressures due to the lower thermal inertia found there, which appears to be oddly in- consistent with the limb-data analyses from Cassini. Finally, the model of FM is significantly different from those of TG and SG. The reduction of the wintertime temperatures dur- ing polar night is not surprising as the purely radiative sea- sonal models continually radiate to space whilst in darkness without accounting for any heat advection, but it is prob- able that significant diffusion and advection of heat would occur in this region. However, it is interesting to note that the models that do the best job of reproducing the measured temperatures at Ls = 46◦ (i.e., southern autumn, Fig. 10.4) and 75◦S latitude are those without dynamics. In summary, it appears that the dominant driver of Sat- urn’s stratospheric temperatures is solar forcing, albeit per- turbed by circulation and dynamics. While purely radiative seasonal models do not reproduce the measured tempera- tures perfectly, they seem to do as good a job as the more 110120130140150160170Temperature (K) 110120130140150160170Temperature (K)2005−2007 (LS=313) 0.01 mbar−90−60−300306090Planetocentric Latitude (deg)110120130140150160170Temperature (K)2005−2007 (LS=313) 0.1 mbar060120180240300360LS (deg)110120130140150160Temperature (K)060120180240300360LS (deg)110120130140150160Temperature (K)10 Fletcher, Greathouse, Guerlet, Moses & West The stratospheric dT /dy is indicative of positive vertical shear in the winter and negative vertical shear in the sum- mer (Friedson and Moses, 2012). Cassini’s seasonal monitor- ing has shown that dT /dy has become increasingly positive in the mid-stratosphere over the ten years of the mission (see Fig. 10.3), meaning that northern middle-atmospheric pro- grade jets should become more retrograde (i.e., decreasing eastward velocities) and southern prograde jets should be- come more prograde (increasing eastward velocities). These modifications to the stratospheric wind field, although in- ferred indirectly from temperatures, can have implications for the transmissivity of waves upwards from the convec- tive troposphere (e.g., the wave transport of energy from Saturn’s storm regions in 2010-11, Fletcher et al., 2012b). Despite the large temperature fluctuations, thermal wind variability in the troposphere has been small over the dura- tion of the Cassini mission (variations smaller than 10 m/s per scale height at the tropopause, Fletcher et al., 2016), with the largest changes in the northern flank of the equa- torial jet. Finally, although radiative climate models successfully re- produce the magnitude and scale of the observed asymme- tries, they lack the dynamic perturbations that govern the temperatures on a smaller scale. When purely radiative cal- culations are insufficient to reproduce the observed temper- ature changes, we can use the thermodynamic energy re- lationship (Hanel et al., 2003; Holton, 2004) to relate the change in the temperature field to the residual mean cir- culation causing net heating (subsidence) and cooling (up- welling). Conrath et al. (1990) used these techniques to show that a diffuse inter-hemispheric circulation was expected at solstice, with rising motion in the summer hemisphere and downward motion in the winter hemisphere. However, at equinox the flow consisted of two cells, with rising motion at low latitudes and subsidence poleward of ±30◦ latitude, which does not appear to correspond to the equinoctial ob- servations of Cassini. Friedson and Moses (2012) produced a more complex model to predict these inter hemispheric transports, finding a seasonally-reversing Hadley-like circu- lation at tropical latitudes and cross-equatorial flow from the summer into the winter hemisphere twice per year. Fletcher et al. (2015) used the deviation of Saturn’s mea- sured temperatures from radiative equilibrium to show that upwelling/downwelling winds with zonally-averaged vertical velocities of w ≈ 0.1 mm/s could account for the cool- ing/warming of the southern/northern polar stratospheres at 1 mbar, respectively. In summary, the general warm- ing of the northern hemisphere and cooling of the southern hemisphere over the duration of the mission could be re- produced by an inter-hemispheric transport into the spring hemisphere in the stratosphere and upper troposphere, per- turbed at low latitudes by a Hadley-type circulation, and at high latitudes by the formation and dissipation of polar vortices. 10.2.6 Non-seasonal phenomena At low latitudes, the signature of Saturn’s evolving sea- sons appear to be overwhelmed by the phenomenon known Figure 10.9 Zonal mean temperature retrievals from Voyager/IRIS (Ls = 8◦) and Cassini/CIRS (Ls = 3◦) as presented by Sinclair et al. (2014). Although these differ by ∆Ls = 5◦, the comparison shows that the zonal mean stratospheric temperatures differ considerably at the equator between the two epochs. The upper troposphere also appears to be warmer in the southern autumn hemisphere (10-45◦S) in the Cassini measurements compared to the Voyager measurements, as discussed by Li et al. (2013), suggesting non-seasonal variability. The hashed areas are regions of low retrieval confidence. (perhaps incorrectly) as Saturn’s Semi-Annual Oscillation (SSAO). Originally identified by Fouchet et al. (2008) and Orton et al. (2008) as temporally-evolving oscillations in the vertical and latitudinal temperature structure, the down- ward propagation of the wave structure has been studied by Cassini infrared spectroscopy and radio occultation data (Fletcher et al., 2010; Guerlet et al., 2011; Li et al., 2011; Schinder et al., 2011). These studies revealed that the local temperature extrema observed in 2005 (an equatorial local maximum located at the 1-mbar pressure level and a local minimum located at 0.1 mbar) descended by approximately 1.3 scale heights in 4.2 years (see Fig. 10.4). The down- ward propagation of the oscillation was consistent with it being driven by absorption of upwardly propagating waves and suggested a ∼15-year period for Saturn’s equatorial os- cillation, as already derived from long-term ground based observations Orton et al. (2008). Saturn’s seasonally changing atmosphere: thermal structure, composition and aerosols 11 Cassini observations in early northern spring might be ex- pected to replicate Voyager 1 and 2 observations taken one Saturnian year earlier (Ls = 8 − 18◦). These datasets have been compared by Li et al. (2013), Sinclair et al. (2014) and Fletcher et al. (2016), and mid-latitude temperatures were largely the same between the two epochs, confirming a re- peatability of the broad seasonal cycle. In the stratosphere, Sinclair et al. (2014) showed that CIRS 2009-10 tempera- tures were 7.1± 1.2 K warmer than IRIS 1980 temperatures near 2 mbar in the equatorial region (Fig. 10.9), implying that Voyager and Cassini are capturing the equatorial oscil- lation in slightly different phases, which is inconsistent with the previously-identified period of the SSAO. Indeed, this suggests that the oscillation may be quasi-periodic, as on Earth. It is possible that seasonal storm activity, such as the equatorial eruption in 1990 and the northern mid-latitude eruption of 2010 (see Chapter 13), may have disrupted this equatorial oscillation. Evidence for non-seasonal variability was also observed in the troposphere (Li et al., 2013; Fletcher et al., 2016), potentially as a result of the wave structure impinging on the stably-stratified upper troposphere. Li et al. (2013) sug- gested that CIRS results were 2-4 K warmer than IRIS results at ±15◦ latitude near 300 mbar, and that the tropopause (50-100 mbar) is 8-10 K warmer in 2009/10 compared to 1980. Although low-pressure (p < 100 mbar) changes were not confirmed by Sinclair et al. (2014) and Fletcher et al. (2016), they do confirm the changing lati- tudinal thermal gradients near the equator (and hence the vertical shear on zonal winds), which were rather different between the Voyager and Cassini epochs. The distribution of para-H2 also suggested substantial differences in the tropical region between Voyager and Cassini (see Section 10.3.2 and Fletcher et al., 2016). Taking the tropospheric and strato- spheric results together, it appears that Saturn’s equato- rial oscillation may not be strictly ‘semi-annual’ at all, and possibly changes from year to year in response to dynamic phenomena such as storm eruptions. 10.3 Distribution of chemical species 10.3.1 Overview of Saturn’s atmospheric composition The composition and 3D distribution of gas-phase con- stituents in Saturn’s atmosphere is controlled by the cou- pled influence of thermochemical equilibrium, photochem- istry and other disequilibrium chemical mechanisms, global circulation and regional atmospheric dynamics, and aerosol microphysical processes. In the absence of in situ sounding of Saturn’s atmosphere (see Chapter 14), we rely on remote sensing measurements (Fig. 10.10) to determine the spatial distribution of these species, and theoretical chemical mod- els to explain why particular species are observed or not observed. Saturn’s hydrogen-helium atmosphere contains a wealth of reduced trace species deriving from the dominant chemical elements (carbon, nitrogen, sulfur, phosphorus and oxygen), and thermochemical equilibrium in the deep hot at- mosphere steers the bulk atmospheric abundances toward the dominant H2, He, H2O, CH4, NH3, and H2S atmo- spheric composition at the low temperatures and pressures found in the observable upper atmosphere (e.g., Fegley and Prinn, 1985). Metals, silicates, and other refractory species condense out deep in the atmosphere and remain unobserv- able. Nitrogen, sulfur and oxygen compounds are largely hid- den from the reach of remote sensing due to condensation into cloud decks (Section 10.4), with the exception of NH3, which is volatile enough that its presence can be detected in the upper troposphere within its condensation region. Oxy- gen compounds can be detected in the upper atmosphere when delivered from sources external to Saturn. Tempera- tures at Saturn’s tropopause remain too warm for methane condensation, so that this principal carbon-bearing molecule remains abundant throughout the observable upper tropo- sphere and stratosphere. Chemical equilibrium is not always maintained, how- ever, as the gases at depth are convected upward into cooler regions where chemical-kinetic reactions can become more sluggish. Species mixing ratios “quench” when ver- tical transport time scales fall below the chemical kinetic time scales for conversion between different major forms of an element (e.g., Prinn and Barshay, 1977; Lewis and Feg- ley, 1984). Such quenching is responsible for the presence of CO, PH3, GeH4, and AsH3 in Saturn’s upper tropo- sphere, whereas equilibrium arguments suggest that phos- phorus, germanium and arsenic should be sequestered in condensates at depth (e.g., Fegley and Lodders, 1994). When a species is able to avoid cold-trapping by conden- sation, it can be transported to sufficiently high altitudes to interact with solar UV radiation, and can serve as a par- ent molecule for chains of chemical reactions that produce additional disequilibrium species in the upper troposphere and stratosphere (e.g., N2H4, P2H4 and complex hydrocar- bons). Saturn’s axial tilt and seasonally variable solar insola- tion as a function of latitude result in temporal and merid- ional variations in the photochemical production and loss rate of atmospheric species. This variable photochemistry, combined with atmospheric dynamics, controls the distribu- tion of the atmospheric constituents in the stratosphere and upper troposphere, as shown in Fig. 10.11. In this section, we review our current knowledge of Saturn’s seasonally- variable composition, building upon the extensive reviews by Prinn et al. (1984) and Fouchet et al. (2009), and showing how the species distributions are intricately connected with Saturn’s radiative budget (Section 10.2) and aerosols (Sec- tion 10.4). We briefly describe the dominant processes and discuss the advances in our understanding of atmospheric chemistry that have occurred since the review by Fouchet et al. (2009). We confine our discussion to Saturn’s tropo- sphere and stratosphere; the thermosphere and ionosphere are covered in Chapter 9. 12 Fletcher, Greathouse, Guerlet, Moses & West Figure 10.11 Predicted vertical distributions of several tropospheric species of interest on Saturn for 27◦S planetographic latitude under summer conditions, based on the photochemical model described in Moses et al. (2010). The sharply decreasing abundance of H2O, H2S, NH3, P2H4, and N2H4 with increasing altitude in the 0.1-10 bar region is due to condensation, whereas the sharp decrease in the PH3 abundance near 0.2 bar is due to photochemistry. Molecules whose mixing ratios increase with increasing altitude have a source at higher altitudes, either from stratospheric hydrocarbon photochemistry (e.g., C2H6, C2H2, C3H8, HCN) or from influx of external material combined with photochemistry (e.g., H2O, CO, CO2). 10.3.2 Tropospheric composition Condensible volatiles The vertical structure of Saturn’s cloud-forming region is determined by thermochemical equilibrium, condensation chemistry and vertical mixing of quenched species. Thermo- chemical equilibrium models for Saturn have been developed by Fegley and Prinn (1985), Fegley and Lodders (1994), Lod- ders and Fegley (2002), Visscher and Fegley (2005) and Viss- cher et al. (2006). Water, the dominant equilibrium form of oxygen on Saturn, will condense in the upper troposphere to form a liquid aqueous solution cloud at depth (around 20 bar, although local temperatures, meteorology and the bulk abundances will affect these cloud condensation lev- els), trending to water ice at higher altitudes (e.g., Weiden- schilling and Lewis, 1973). The pressure at the cloud base will depend on the unknown bulk oxygen abundance on Sat- urn, as well as on the details of regional atmospheric dynam- ics (e.g., Sugiyama et al., 2011, 2014; Palotai et al., 2014). Because some of the planet’s oxygen is tied up in condensed silicates at even deeper levels, the water abundance at the base of the aqueous solution cloud represents an already de- pleted fraction of the bulk oxygen abundance. Visscher and Fegley (2005) determine that roughly 20% of Saturn’s bulk oxygen will be tied up in these condensates if the oxygen-to- silicate-and-metal fraction in the atmosphere is in solar pro- portions. Indeed, signatures of Saturn’s tropospheric water have only been detected near 5 µm by ISO (probing the 2-4 bar level above the cloud, de Graauw et al., 1997), and the mixing ratios were highly sub-solar. Cassini/VIMS spectra Figure 10.10 Examples of remote sensing data from the UV to the sub-millimeter as measured by Cassini/VIMS (top two panels) and Cassini/CIRS (bottom two panels), demonstrating the rich variety of absorption and emission features. Note that the VIMS panels are provided in wavelength and reflectivity units, whereas the CIRS panels are provided in wavenumber and brightness temperature units, to maximize the visibility of features. All spectra were averaged over Saturn’s equatorial region, and can be expected to vary from location to location on the planet. Not shown are Cassini/UVIS observations at the shortest wavelengths (55-190 nm) and Cassini/RADAR observations at the longest wavelengths (2.2 cm). 0.40.50.60.70.80.91.0Wavelength [μm]0.000.050.100.150.20I/F 10000. 12000. 15000. 20000.123450.00.10.20.3I/F 2000. 2500. 4000. 10000.100200300400500600Wavenumbers [cm-1]859095100105110115120Brightness T [K] 17.0 20.0 25.0 50.0100.0500.0600800100012001400Wavenumbers [cm-1]90100110120130140Brightness T [K] 7.5 8.0 9.010.012.014.016.0Wavelength [μm]Wavelength [μm](a) Visible (VIMS-V)(b) Near-Infrared (VIMS-IR)Wavenumbers [cm-1]Wavenumbers [cm-1](d) Far-Infrared/Sub-mm (CIRS, 0.5 cm-1)Wavelength [μm](c) Mid-Infared (CIRS, 2.5 cm-1)C2H2C2H6PH3PH3CH4H2-HeH2-HePH3NH3CH4CH4CH4CH4CH4PH3PH3GeH4, CO, CH3D, AsH3CH3DCH4CH4CH4CH4UV Haze SlopeH2CH4C2H6Mole FractionPressure (bar)CH4HeH2NH3H2SH2OH2OHCNHPH3AsH3CO2C2H2C2H2N2H4C3H8COP2H4N2C2H6Saturn’s seasonally changing atmosphere: thermal structure, composition and aerosols 13 of the same region were unable to provide sufficient sensitiv- ity to measure the water abundance (Fletcher et al., 2011). Ammonia and hydrogen sulfide are the dominant equi- librium forms of nitrogen and sulfur, respectively, in Sat- urn’s atmosphere. Some NH3 will be dissolved in the upper- tropospheric water solution cloud. Above that level, gas- phase NH3 and H2S are expected to react to form a crys- talline NH4SH cloud, followed at even higher altitudes by a cloud of NH3 ice. Briggs and Sackett (1989) used 1.3-70 cm radio observations from the VLA and Arecibo in 1980 to measure a depletion in NH3 from 25 bar to 2 bar, suggest- ing that the formation of the NH4SH cloud was the main NH3 sink, which indirectly required H2S to be ten times so- lar composition (approximately 400 ppm, following van der Tak et al., 1999, but depending on the solar sulfur abun- dances chosen). However, uncertainties in the spectral line shapes and data calibration in the radio region, combined with the relatively featureless spectra of the various con- stituents, make this result highly uncertain, and H2S has never been directly detected. The deep formation of the H2O and NH4SH clouds means that Saturn’s bulk oxygen and sulfur abundances therefore remain unknown, and that H2O at least is unlikely to contribute significantly to seasonally- variable photochemistry in the upper troposphere. Ammonia, on the other hand, is volatile enough to main- tain a presence at the uppermost tropospheric altitudes (see Fig. 10.11). Signatures of NH3 were first tentatively iden- tified at visible wavelengths by Giver and Spinrad (1966) and positively identified by Encrenaz et al. (1974), and has since been identified in the near-infrared (Larson, 1980; Fink et al., 1983), far-infrared (Courtin et al., 1984) and mi- crowave wavelengths (the 1.2-cm inversion band, de Pater and Massie, 1985). Measurements at infrared wavelengths have provided globally-averaged abundances at a range of altitudes: microwave and radio observations indicate mole fractions exceeding 500 ppm at p > 3 bar (de Pater and Massie, 1985; Briggs and Sackett, 1989; van der Tak et al., 1999), declining to abundances of around 100 ppm near the condensation altitude (Courtin et al., 1984; Grossman et al., 1989; Briggs and Sackett, 1989; Noll and Larson, 1990; de Graauw et al., 1997; Orton et al., 2000; Burgdorf et al., 2004), and becoming strongly sub-saturated above the clouds (approximately 0.1 ppm at 500 mbar, Kerola et al., 1997; Kim et al., 2006; Fletcher et al., 2009). See Orton et al. (2009) for a historical overview. These globally-averaged val- ues suggest that ammonia decreases with altitude above the clouds due to condensation (relative humidities of approx- imately 50% de Graauw et al., 1997) and photochemical processes. Both of these processes depend on temperature and seasonal insolation cycles, so we might expect ammonia to be spatially and temporally variable. Spatially-resolved distributions of NH3 have been pro- vided by the CIRS (Hurley et al., 2012), VIMS (Fletcher et al., 2011) and RADAR instruments (Janssen et al., 2013; Laraia et al., 2013) on the Cassini spacecraft (Fig. 10.12). Hurley et al. (2012) modelled CIRS far-IR rotational NH3 lines to determine the 600-mbar NH3 distribution (i.e., above the clouds), finding that NH3 was largely uncorre- lated with the belt/zone structure of the thermal field, and Figure 10.12 Zonal mean ammonia distributions measured (a) in the upper troposphere by Cassini/CIRS (Hurley et al., 2012) and in the cloud-forming region by (b) Cassini/VIMS (Fletcher et al., 2011) and (c) Cassini/RADAR (Laraia et al., 2013). The solid line in panel (b) represents a VIMS retrieval in the absence of aerosol scattering, the dotted line shows the effects of scattering. The isolated point with error bars at 60◦S shows the typical uncertainty on the VIMS inversion. In panel (c), large negative values of the 2.2-cm brightness temperature residuals imply a low NH3 humidity, and these regions are aligned with regions of depleted ammonia vapor in the VIMS results. The region between ±4◦ is from a single RADAR scan, whereas the rest is an average over five RADAR maps. instead showed a weak hemispheric asymmetry (with higher abundances in the southern summer hemisphere) and no ev- idence for peak equatorial abundances (Fig. 10.12a). They concluded that condensation and photolytic processes were shaping the distribution, and that NH3 does not trace the local circulation in this region. Fletcher et al. (2011) used VIMS observations of the 5-µm region to identify a nar- row equatorial peak (within 5◦ of the equator) in NH3 at p > 1 bar in the cloud-forming region, with peak values of 500 ppm similar to the deep abundances observed in the microwave (de Pater and Massie, 1985). Elsewhere the NH3 abundances fell to 100-200 ppm in the 1-4 bar region, consis- tent with previous studies of the condensation region, and there was no indication of a strong hemispheric asymme- try (Fig. 10.12b). 2.2-cm brightness temperature maps mea- sured by RADAR (Janssen et al., 2013; Laraia et al., 2013) 0100200300400500NH3 Mole Fraction [ppm]-90-75-60-45-30-150153045607590Planetographic Latitude-6-5-4-3-2-10Residual Brightness Temperature [K](b) Cassini/VIMS mole fraction in the 1-4 bar range(c) Cassini/RADAR TB residuals(a) Cassini/CIRS mole fraction near 600 mbar 0.00.10.20.30.4NH3 Mole Fraction [ppm]14 Fletcher, Greathouse, Guerlet, Moses & West were used to study NH3 humidity near 1.5 bar, finding low brightnesses near the equator (consistent with the narrow NH3 peak observed by VIMS) that were flanked by regions of high brightness near ±9◦ latitude (Fig. 10.12c), and a subtle asymmetry between the northern and southern hemi- sphere (Janssen et al., 2013). Furthermore, Janssen et al. (2013) demonstrated a remarkable stability in the latitu- dinal distribution of NH3 measured by RADAR over the 2005-2011 time span of the Cassini data, in contrast to the time-variability suggested by VLA data in the 1980s and 1990s by van der Tak et al. (1999). This raises the possi- bility of long-term variability in Saturn’s banded structure similar to Jupiter’s ‘global upheavals’. However, during the Cassini epoch, the equatorial maximum in NH3 observed by both VIMS and RADAR, flanked by local minima near ±9◦, may be consistent with strong equatorial upwelling flanked by regions of subsidence. Laraia et al. (2013) deduced an av- erage humidity of 70 ± 15% in the cloud region, and showed that their data were consistent with a 3 − 4× solar enrich- ment (360-480 ppm) at depth, but depleted for p < 2 bar. Fig. 10.12c may show a slight hemispheric asymmetry, with higher brightnesses (lower NH3 abundances) in the northern, winter hemisphere, qualitatively consistent with the asym- metry in the upper troposphere identified by Hurley et al. (2012). The possible mechanisms creating this asymmetry in Saturn’s southern summer are discussed later in this section. Quenching and disequilibrium Phosphine: Thermochemical equilibrium arguments pre- dict the absence of several species in Saturn’s upper tropo- sphere (phosphine, arsine, germane and carbon monoxide) due to sequestration in the hotter deep interior. For example, equilibrium models suggest that phosphorus would be tied up in PH3 at great depths on Saturn, but reactions with water should convert the phosphine to P4O6 at tempera- tures below ∼900 K, with condensed NH4H2PO4 becoming the dominant phase at temperatures below ∼400 K (Feg- ley and Prinn, 1985). Nevertheless, PH3 has been shown to have a significant effect on Saturn’s spectrum, providing a clear indication that phosphorus chemical equilibrium is not achieved on Saturn. The presence of PH3 is most likely due to rapid vertical mixing, transporting PH3 upwards with a dynamical timescale shorter than the chemical depletion timescale, so that the observed abundances are represen- tative of the “quenched” equilibrium conditions at deeper atmospheric levels (Prinn et al., 1984; Fegley and Lodders, 1994). Furthermore, the PH3 profile will be very sensitive to vertical winds, with downwelling winds suppressing the PH3 abundance at the ∼100-400 mbar level and upwelling winds enhancing it, making this species a useful tracer of tropospheric dynamics. Before the arrival of Cassini, PH3 studies focused on disk- integrated abundances and the vertical distribution (evi- dence for PH3 photolysis). Following its first detection near 10 µm (Gillett and Forrest, 1974; Bregman et al., 1975), PH3 has been studied in the mid-IR from ground- and space-based observatories (Tokunaga et al., 1980; Courtin et al., 1984; de Graauw et al., 1997; Lellouch et al., 2001); in the 5-µm opacity window (Larson, 1980; Noll and Larson, 1991; de Graauw et al., 1997); the 3-µm reflected compo- nent (Larson, 1980; Kerola et al., 1997; Kim and Geballe, 2005; Kim et al., 2006); the sub-millimeter (Weisstein and Serabyn, 1994; Davis et al., 1996; Orton et al., 2000, 2001; Burgdorf et al., 2004; Fletcher et al., 2012a) and the ul- traviolet (Winkelstein et al., 1983; Edgington et al., 1997). Vertical distributions were presented by Noll and Larson (1991); de Graauw et al. (1997); Orton et al. (2001); Lel- louch et al. (2001); Kim et al. (2006); Fletcher et al. (2012a) to show that PH3 declined with altitude above the 500-700- mbar level and was not present in the stratosphere, but none of these studies provided assessments of the spatial distribu- tion, which would be required to understand seasonal con- trasts and regions of powerful convective uplift. To date, Cassini is the only facility to have provided the spatial distribution of PH3 from Cassini/VIMS (5-µm) and Cassini/CIRS (10 µm and sub-millimeter), as shown in Fig. 10.13. Although there are quantitative differences between the mole fractions derived from the two instruments, the results agree qualitatively in the upper troposphere. A key finding was the enhancement in upper tropospheric PH3 in the equatorial region (Fletcher et al., 2007b, 2009, 2011), suggestive of rapid vertical mixing consistent with the cold equatorial temperatures (Section 10.2) and elevated hazes (Section 10.4) observed there. This PH3-enriched zone is flanked at tropical latitudes (±23◦ planetographic) by re- gions of depletion, consistent with a Hadley-style circula- tion. However, the equatorial enhancement is only identifi- able in the upper troposphere, and not in the deeper adia- batic region (i.e., below the radiative-convective boundary), where the PH3 distribution suggests depleted abundances at the equator and enhanced abundances in the neighbor- ing belts (Fig. 10.13(c) and Fletcher et al., 2011). CIRS does not have the altitudinal sensitivity to probe pressures greater than 1 bar, so this ‘deep’ distribution of PH3 comes solely from inversions of VIMS spectra in the 5-µm window, where measurements of composition are rather degenerate with assumptions about tropospheric aerosol properties. As a result, this difference in the zonal mean PH3 distribu- tion between the upper and lower troposphere has yet to be explained, but could be related to a transition in the tropospheric circulation above and below the main cloud decks. In addition, PH3 was found to be depleted within the cyclonic polar vortices identified at both the summer and winter poles (Fletcher et al., 2008). Finally, both CIRS and VIMS observations suggest a hemispheric asymmetry in the PH3 abundance, being more abundant in the southern summer hemisphere. Whether this asymmetry is permanent or will reverse with Saturn’s seasons remains to be seen - (Fletcher et al., 2010) suggested that the zonal mean PH3 distribution measured by CIRS had not changed during the 2004-2009 period. The possible seasonal nature of this con- trast is discussed below. Other disequilibrium species: Ground-based observa- tions of Saturn’s 5-µm opacity window in 1986-1989 pro- vided the first detections of germane (GeH4, Noll et al., 1988), arsine (AsH3, Noll et al., 1989; Noll and Larson, 1990; B´ezard et al., 1989) and carbon monoxide (CO, Noll Saturn’s seasonally changing atmosphere: thermal structure, composition and aerosols 15 mum flanked by local maxima at ±7◦ and a possible north- south asymmetry (the size of which depended on the as- sumed scattering properties of Saturn’s aerosols in the 5-µm window). The global mean abundances of AsH3 (2.2 ± 0.3 ppb with fully-scattering aerosols, Fletcher et al., 2011) were consistent with ground-based determinations of 3.0 ± 1.0 ppb (Noll and Larson, 1990) and 2.4+1.4−1.2 ppb (B´ezard et al., 1989). Saturn’s CO can originate from two sources: an exter- nal source of oxygen-bearing molecules in the upper strato- sphere (detected in the sub-millimeter by Cavali´e et al., 2009, 2010), and an internal source as a quenched disequi- librium species (measured at ppb-levels in the upper tropo- sphere by Noll et al., 1986; Noll and Larson, 1991; Cavali´e et al., 2009). These measurements are complex, as the in- ternal and external CO distributions are difficult to sepa- rate spectrally, meaning that measurements of the intrinsic CO abundance remain highly uncertain. For the intrinsic, deep-tropospheric source of CO, recent ab initio transition- state theory rate-coefficient calculations and numerical mod- eling by Moses et al. (2011) and Visscher and Moses (2011) have advanced our understanding of the CH4-CO quench process on hydrogen-rich planets. These authors find that thermal decomposition of methanol to form CH3 + OH is the rate-limiting step in the CO → CH4 conversion pro- cess under solar-system giant-planet conditions. The result- ing thermochemical kinetics and transport models can reli- ably predict the resulting quenched CO abundance in the upper tropospheres of the giant planets, and possibly pro- vide indirect constraints on the deep water abundance and transport rates if the kinetic conversion between CH4 and CO at high temperatures and pressures were well character- ized (e.g., Prinn and Barshay, 1977; Lewis and Fegley, 1984; Yung et al., 1988; Fegley and Lodders, 1994; Lodders and Fegley, 2002; B´ezard et al., 2002; Visscher and Fegley, 2005; Visscher et al., 2010). The external source of CO (cometary impacts, continuous interaction with the rings and satellites) are discussed in Section 10.3.3. Theoretical models suggest that N2 is the most abundant quenched thermochemical species on Saturn and the other giant planets, with a mole fraction of ∼1 ppm (Moses et al. 2010; see also Prinn and Fegley 1981, Lewis and Fegley 1984, Fegley and Prinn 1985, Lodders and Fegley 2002), with the rate-limiting step in the N2 → NH3 conversion being the reaction of H with N2H2 (Moses et al., 2010, 2011). Be- cause it is a homonuclear molecule, N2 is difficult to ob- serve, but there are potential photochemical consequences of a large N2 abundance in the high-altitude stratosphere and ionosphere that may be observable, particularly at UV wavelengths. Other quenched molecules such as CO2, HCN, CH3NH2, H2CO, CH3OH, CH2NH, and HNCO are pre- dicted to have very small mixing ratios (Visscher et al., 2010; Moses et al., 2010) and so far only upper limits on HCN, H2CO and CH3OH have been reported from sub-millimeter spectroscopy (Fletcher et al., 2012a). Figure 10.13 Distribution of PH3 determined by Cassini/CIRS (Fletcher et al., 2009) and Cassini/VIMS (Fletcher et al., 2011) during southern summer, sensing different regions of the troposphere. The CIRS distribution in the upper troposphere (shown as a mole fraction contour plot for p < 1 bar) and VIMS-derived fractional scale height (p < 1 bar) are consistent with one another and show an equatorial enhancement and a north-south asymmetry in southern summertime (the fractional scale height is the scale height of PH3 gas divided by the bulk gas scale height). The deep mole fraction derived from VIMS observations shows opposite trends, suggestive of a different circulation regime beneath the topmost clouds. Solid lines show VIMS night-side calculations without aerosol scattering, dotted lines show VIMS calculations with scattering of thermal photons. The two points with error bars at 60◦S show the typical uncertainties on the VIMS inversions. et al., 1986). As with PH3, these species were not expected in the upper troposphere on the grounds of thermochemi- cal equilibrium. Gaseous GeH4 and AsH3 are expected to be the dominant phases of germanium and arsenic at depth (T > 400 K), to be replaced by condensed Ge and As phases in the upper troposphere (e.g., Fegley and Lodders, 1994), but both species are detected in the upper troposphere, and apparently decrease in abundance with altitude (de Graauw et al., 1997; B´ezard et al., 1989). Cassini/VIMS data are the- oretically sensitive to both species, but only AsH3 could be mapped spatially given the low spectral resolution (Fletcher et al., 2011). As with the deep distribution of PH3, the zonal mean AsH3 distribution showed an equatorial mini- 0.00.10.20.30.40.5Fractional Scale Height1.00.1Pressure (bar)Saturn PH3 Distribution 0.10.10.10.10.10.10.10.10.11.01.01.01.01.02.02.02.02.03.03.03.04.04.05.05.06.06.0 23456Mole Fraction (ppm)-90-75-60-45-30-150153045607590Planetographic Latitude(a) Cassini/CIRS abundances [ppm](b) Cassini/VIMS fractional scale height p<1 bar(c) Cassini/VIMS deep mole fraction [ppm]16 Fletcher, Greathouse, Guerlet, Moses & West Tropospheric photochemistry Asymmetries in the zonal mean distributions of upper tropo- spheric PH3 and NH3 have been identified in Cassini/CIRS, VIMS and RADAR observations, likely related to seasonal contrasts in the efficiency of photolysis and condensation (in the case of NH3). Because NH3 and PH3 are the most abun- dant photochemically-active species in the region of Saturn’s troposphere above the cloud decks, tropospheric photochem- istry revolves around the coupled chemistry of these species (see the reviews of Strobel, 1983, 2005; Atreya et al., 1984; West et al., 1986; Fouchet et al., 2009). Although CH4 is more abundant than either NH3 or PH3, methane is only photolyzed at UV wavelengths shorter than ∼145 nm, and these short-wavelength photons are absorbed by other at- mospheric constituents well above the tropopause, shielding methane from photolysis in the troposphere. Moreover, NH3 and PH3 photolysis products do not readily react with CH4, so methane photochemistry is more prominent in the strato- sphere than the troposphere. Photolysis and photochemical destruction of H2S may occur within and below the NH4SH cloud deck, but little is known about the ultimate fate of the sulfur photochemical products (see Prinn and Owen, 1976; Lewis and Prinn, 1984, for details). The same can be said for AsH3 and GeH4 photochemistry (e.g., Fegley and Prinn, 1985; Nava et al., 1993). Owing to a lack of relevant rate-coefficient information for PH3 photochemistry, in particular, our understanding of tropospheric photochemistry on Saturn (and Jupiter, which is similar) has remained fairly stagnant since the pioneer- ing studies in the late 1970s and 1980s (Strobel, 1975, 1977; Atreya et al., 1977, 1980; Kaye and Strobel, 1983b,a,c, 1984). Advances since that time, which mostly revolve around the relevant kinetics and product quantum yields from PH3 and coupled PH3-NH3 and NH3-C2H2 photochemistry, are re- viewed by Fouchet et al. (2009). Ammonia is photolyzed by photons with wavelengths less than ∼230 nm; the dominant photolysis products are NH2 and H. The NH2 reacts effectively with H2 to reform NH3, releasing another hydrogen atom in the process. Although NH3 condenses in the upper troposphere, its vapor pressure at the tropopause is large enough that NH3 photolysis is an important source of atomic hydrogen in the upper tropo- sphere and lower stratosphere in the models (see Fig. 10.11), affecting the photochemistry of both local PH3 and the com- plex hydrocarbons that are flowing down from their produc- tion regions at higher altitudes. The NH2 and H can also react with PH3 to form PH2. The reaction of NH2 with PH3 helps recycle the ammonia. Two PH2 radicals can re- combine to form P2H4, which condenses in its production region, preventing the effective recycling of PH3. Phosphine itself does not condense under tropospheric conditions on Saturn, but direct photolysis (to produce largely PH2 + H) and reaction of PH3 with NH2 and H serve to drastically de- crease the PH3 abundance above the ammonia clouds (see Fig. 10.11). Fig. 10.14 shows the details of this coupled NH3-PH3 pho- tochemistry, as described by Visscher et al. (2009). The dom- inant product is condensed P2H4, which becomes a major component of the upper-tropospheric haze. Photochemistry of P2H4 can lead to the production of condensed elemental phosphorus as an additional major photochemical product, but the dominant pathways are more speculative (Visscher et al., 2009). Other important but less abundant photoprod- ucts include condensed NH2PH2, condensed N2H4, and gas- phase N2 (e.g., Kaye and Strobel, 1984). The N2 shown in Fig. 10.11 is derived from thermochemical quenching and transport from the deep troposphere, with photochemical production contributing only a tiny fraction in comparison with this deep source. NH3 photochemistry plays only a minor role in shaping its vertical profile at and below the cloud decks. Photochem- istry is responsible for removing NH3 above the clouds (with N2H4 being the dominant product), but dynamics and con- densation control the profile within the cloud region. Hemi- spheric contrasts are more strongly influenced by the ther- mal profile, condensation, regional dynamics (e.g., equato- rial upwelling) and aerosol-microphysical processes. These competing effects have not yet been disentangled to explain the suggested weak enhancement of ammonia in Saturn’s southern summer hemisphere in Section 10.3.2. Further- more, it is not understood whether the NH3 distribution is responding to the seasonal insolation shifts (i.e., following the thermal and aerosol changes), or remaining static with time. We might expect that northern hemisphere warming will permit the release of more NH3 into the vapor phase during northern spring, helping to reverse the asymmetry observed during southern summer. The PH3 vertical profile, on the other hand, is strongly sensitive to chemistry, as well as to vertical transport and to the opacity of upper-tropospheric hazes and clouds, which help shield the PH3 from photolysis and other photochemi- cal losses. The model PH3 vertical profiles drop much more sharply with altitude in the upper troposphere than the pro- files derived from CIRS and VIMS retrievals (e.g., Fletcher et al., 2007b, 2009, 2011), but these differences are likely an artifact of the assumptions in the data retrievals rather than a true model-data mismatch. Indeed, Herschel/SPIRE observations (Fletcher et al., 2012a) suggested that PH3 is not present in the lower stratosphere at the 10-20 mbar level because it is so chemically fragile. The observed hemispheric asymmetry in the upper-tropospheric PH3 abundance (e.g., Fletcher et al., 2009, 2011) may be related to higher haze opacities of the UV-shielding aerosols produced by photo- chemistry in the summer hemisphere, although the global circulation system (i.e., an inter-hemispheric transport from the autumn to the spring hemisphere, see Section 10.2.5) may also play a role. Seasonal tropospheric photochemistry has not yet been investigated theoretically. Summertime in- solation will promote the production of aerosols like P2H4, which will help shield the PH3 and NH3 from photolysis, allowing the PH3 molecules to be carried to higher altitudes during southern summer and autumn, qualitatively consis- tent with Cassini’s observations. Unsaturated hydrocarbons like C2H2 are not particularly abundant in the region in which PH3 and NH3 photochem- istry is active (see Fig. 10.11), limiting the effectiveness of the coupled photochemistry of NH3 and PH3 with C2Hx Saturn’s seasonally changing atmosphere: thermal structure, composition and aerosols 17 and C3Hx species (e.g., Moses et al., 2010), which would otherwise be expected to produce HCN, acetonitrile, methy- lamine, ethylamine, acetaldazine, acetaldehyde hydrazone, vinylphosphine, ethylphosphine, and a whole suite of other interesting organo-nitrogen and organo-phosphorus species Kaye and Strobel (1983b,a); Ferris and Ishikawa (1988); Guillemin et al. (1995); Keane et al. (1996); Moses et al. (2010). The upper-tropospheric C2H2 abundance is not no- tably enhanced in latitude regions known to host thunder- storm activity (Hurley et al., 2012), suggesting that light- ning chemistry does not play much of a role in enhancing upper-tropospheric hydrocarbons beyond their photochem- ically produced abundances. Based on the chemical mecha- nism of Moses et al. (2010), hydrogen cyanide would be the dominant product of the coupled hydrocarbon-NH3 photo- chemistry on Saturn, but these models suggest that pho- tochemically produced HCN is not abundant enough to be observable on Saturn (see Fig. 10.11), and despite a tentative detection of HCN by Weisstein and Serabyn (1996), Herschel analyses have provided upper limits an order of magnitude smaller (mole fractions less than 1.6× 10−11 if the species is well-mixed, Fletcher et al., 2012a). Small amounts of all of the aforementioned organic compounds are produced, how- ever, and will condense, which could potentially contribute to the cloud chromophores on Saturn (see also Carlson et al., 2012). Photo-processing of the condensed P2H4, elemental phosphorus, or NH4SH is also a potential source of the yel- lowish colors on Saturn. The fact that the yellow-brown col- oring is apparently absent in the Cassini images of the high- latitude winter hemisphere of Saturn soon after it emerges back into sunlight — thought to result from a clearing, thin- ning, or reduction in size or depth of the aerosols in the winter hemisphere (e.g., Fletcher et al., 2007a; West et al., 2009) — further points to a photochemical product as the source of the chromophore (see Section 10.4). Figure 10.14 Schematic diagram illustrating the important reaction pathways for coupled PH3 and NH3 in Saturn’s troposphere (based on Visscher et al., 2009; Kaye and Strobel, 1984). The rectangular boxes represent stable molecules, rounded boxes represent radicals or less stable molecules, and hν represents ultraviolet photolysis. The reaction network begins with the photolysis of NH3 and PH3 and the destruction of PH3 from reaction with ammonia photolysis products. The dominant end products are condensed P2H4, Px (elemental phosphorus), N2H4, NH2PH2 and gas-phase N2. Ammonia is also lost via condensation. Para-hydrogen Our final tropospheric species shapes the underlying H2-H2 and H2-He collision-induced continuum throughout the in- frared and sub-millimeter, in addition to readily identifiable quadrupole line features near 0.62-0.64 µm and collision- induced dipole features near 0.826 µm and 2.1-2.2 µm. De- spite the lack of a permanent dipole, collisions between these molecules create instantaneous dipoles that shape the con- tinuum, and the ratio of ortho-H2 (corresponding to the odd rotational spin state of H2 with parallel spins) and para-H2 (the even spin state of H2 with anti-parallel spins) can be deduced from the relative sizes of the broad S(0) (due to para-H2) and S(1) (due to ortho-H2) absorptions near 354 and 587 cm−1, respectively. In “normal” hydrogen, at the high temperatures of Saturn’s lower troposphere, this ratio should be 3:1 in equilibrium (a para-H2 fraction fp of 25%) (e.g., Massie and Hunten, 1982). At the colder temperatures of the upper troposphere the equilibrium para-H2 fraction (feqm) increases beyond 25% as para-H2 has the lowest ro- tational energy state, and we might expect the tropopause para-H2 fraction to be in the 45-50% range, falling as we rise further into the warm stratosphere. However, parcels of air displaced upwards will retain their initial low para- H2 fraction if the vertical mixing is faster than the chemi- cal equilibration timescale, leading to sub-equilibrium con- ditions (fp < feqm) (Conrath et al., 1998; Fletcher et al., 18 Fletcher, Greathouse, Guerlet, Moses & West 2007a). Conversely, downward displacement of cool air with high values of fp can lead to super-equilibrium conditions (fp > feqm). The spatial distribution of the para-H2 fraction, and its deviation from equilibrium, can therefore be used to trace both vertical mixing and the efficiency of chemical equi- libration, both of which may vary with Saturn’s seasons. On Jupiter, where seasonal influences are expected to be negligible, inversions of Voyager/IRIS spectra by Conrath et al. (1998) revealed a distribution of para-H2 that was largely symmetric about the equator, with evidence for sub- equilibrium conditions at the equator and super-equilibrium at higher latitudes. Voyager observations of Saturn in early northern spring (Ls = 8.6 − 18.2◦) revealed local sub- equilibrium conditions near 60◦S, and an asymmetry with a higher fp in the spring hemisphere (super-equilibrium condi- tions from 70◦N to 30◦S, Conrath et al., 1998). These data, most sensitive to the para-H2 fraction between 100-400 mbar (i.e., below the tropopause but above the radiative convec- tive boundary), were interpreted as atmospheric subsidence in the spring hemisphere. Conversely, early measurements by Cassini during south- ern summer (Ls = 297 − 316◦) revealed super-equilibrium over much of the southern summer hemisphere and sub- equilibrium in the winter hemisphere poleward of 30◦N (Fletcher et al., 2007a). Much of this asymmetry was due to the gradient in feqm related to the seasonal tempera- ture contrast, whereas fp itself was found to be largely sym- metric about the equator (e.g., left panels of Fig. 10.15). Fletcher et al. (2010, 2016) demonstrated that the super- equilibrium region was moving slowly northwards through- out the ten-year span of Cassini observations, both due to an increased fp and warmer temperatures in the northern spring hemisphere (i.e., smaller feqm). The para-H2 frac- tion in the southern hemisphere changed very little between 2004 and 2014, although the cooling temperatures as south- ern winter approached caused an increase in the expected feqm, thereby decreasing the discrepancy between fp and equilibrium expectations. In summary, the stark asymmetry in fp − feqm observed during southern summer has reduced Figure 10.15 Seasonal changes in the zonal mean para-H2 fraction measured by Cassini/CIRS using the collision-induced H2-H2 continuum (Fletcher et al., 2007a, 2010, 2016). The upper panels show the subtle changes in fp between 2005 and 2013, including the rising fp at northern high latitudes during spring. The lower panels compares the degree of disequilibrium between 2005 and 2013 (contours have a 1% spacing), showing the reversal of the strong asymmetries found in southern summer. Super-equilibrium conditions (fp > feqm) are shown as solid lines; sub-equilibrium conditions (fp < feqm) are shown as dotted lines and light grey shading. significantly (see Fig. 10.15), such that the southern hemi- sphere is now close to equilibrium, whereas the wintertime sub-equilibrium conditions still prevail at the highest north- ern latitudes. As a result, the zonal mean fp measured by Cassini in 2014 now qualitatively resembles that determined by Voyager in the last northern spring (Fletcher et al., 2016), although quantitative differences remain to be understood, particularly at Saturn’s equator. Given that the timescales for para-H2 equilibration range from decades to centuries in the troposphere (e.g., Con- rath et al., 1998; Fouchet et al., 2003), even in the pres- ence of aerosols whose surfaces could provide paramagnetic sites to catalyze the efficient conversion of uplifted para- H2 back to ortho-H2, the discrepancy from a seasonally- dependent equilibrium (fp − feqm) may not be a good mea- sure of tropospheric circulation. Put another way, the the- oretical feqm varies instantaneously with temperatures on seasonal timescales, whereas fp undergoes much more sub- tle changes with time (showing the largest changes in the northern spring hemisphere). Furthermore, Fletcher et al. (2007a) speculated that the intricate connection between aerosol catalysts and the rate of para-H2 conversion means that equilibrium conditions would be more easily attained when the atmosphere is hazy (e.g., summertime) than when it is aerosol-free (wintertime), and that this was responsible for the asymmetry observed in fp − feqm during southern summer. In either case, it is the tropospheric temperature and/or aerosol availability that determines the magnitude of the disequilibrium, rather than large-scale overturning. -90-75-60-45-30-150153045607590Planetographic Latitude0.500.300.200.100.07Pressure [bar]Para-H2 Fraction: 2005.0.3500.3500.3800.3800.4100.4100.440 -75-60-45-30-1501530456075Planetographic LatitudePara-H2 Fraction: 2009.0.3500.3500.3800.3800.4100.4100.4400.4400.440 -90-75-60-45-30-150153045607590Planetographic Latitude0.500.300.200.100.07Pressure [bar]Para-H2 Fraction: 2013.0.3500.3800.3800.4100.4100.4400.440 Para-H2 Disequilibrium: 2009.-75-60-45-30-1501530456075Planetographic Latitude0.000.000.000.000.000.000.02 Para-H2 Disequilibrium: 2005.-90-75-60-45-30-150153045607590Planetographic Latitude0.500.300.200.100.07Pressure [bar]-0.06-0.04-0.02-0.020.000.000.000.020.02 Para-H2 Disequilibrium: 2013.-90-75-60-45-30-150153045607590Planetographic Latitude0.500.300.200.100.07Pressure [bar]0.000.000.000.000.000.000.00 Saturn’s seasonally changing atmosphere: thermal structure, composition and aerosols 19 Only in regions of strong localized dynamics is fp signifi- cantly altered from equilibrium conditions (e.g., air rising at the equator, storm eruptions, and air sinking at high northern latitudes during spring). For example, the equa- torial minimum in fp (sub-equillibrium conditions) has per- sisted throughout the Cassini mission, consistent with the evidence of powerful equatorial upwelling in the tempera- ture, aerosol, ammonia and phosphine fields (Fletcher et al., 2007a, 2010). The connection between the para-H2 fraction and the tropospheric temperatures, aerosols and circulation remains a topic of active exploration. 10.3.3 Stratospheric composition In this section, we review spatially-resolved observations and modeling of Saturn’s stratospheric hydrocarbons, and their variations with time, to reveal the photochemical and dy- namical processes shaping the middle atmosphere. Strato- spheric photochemistry on Saturn is dominated by Lyman- alpha photolysis of methane at high altitudes, producing a multitude of hydrocarbons, many of which have been ob- served. Ethane (C2H6) was the first photochemical product of methane to be detected in Saturn’s stratosphere from its ν9 emission band at 12 µm by Gillett and Forrest (1974), confirmed shortly after by Tokunaga et al. (1975). Acetylene (C2H2) was detected a few years later by Moos and Clarke (1979) in the UV using data from the International Ultra- violet Explorer (IUE). A column abundance of ethane was obtained using spectra at 3 µm by Bjoraker et al. (1981). The first stratospheric mixing ratios of ethane and acetylene were derived from Voyager/IRIS spectra in the thermal infrared (Courtin et al., 1984; Sada et al., 2005). These measurements confirmed C2H6 and C2H2 as the dominant photochemical products. Constraints from the Infrared Space Observatory (ISO) measurements (de Graauw et al., 1997) were also used by Moses et al. (2000a) to demonstrate that acetylene in- creases with altitude, as expected for a chemical produced in the upper stratosphere and transported downward by eddy diffusion (see also Prang´e et al., 2006). New hydrocarbon species were identified in the 1990s due to the increasing sensitivity of mid-infrared detectors, leading to the determi- nation of the disk-average integrated column abundance of methlyacetylene (CH3C2H), diacetelene (C4H2) (de Graauw et al., 1997; Moses et al., 2000a), benzene (C6H6) (B´ezard et al., 2001a) and the methyl radical (B´ezard et al., 1998), all from ISO. Using ground-based high-resolution spectroscopic data from the NASA/Infrared Telescope Facility (IRTF), ethylene (C2H4) (B´ezard et al., 2001b) and propane (C3H8) (Greathouse et al., 2006) were also detected. Saturn hydrocarbon photochemistry has been recently re- viewed by Fouchet et al. (2009), and there have been few significant theoretical advances since the time of that re- view. A reduced hydrocarbon reaction mechanism suitable for computationally expensive 3D models has been described in Dobrijevic et al. (2011), and recent improvements in pho- tochemical models of Titan and the other giant planets are often directly applicable to Saturn (see Lavvas et al., 2011; Gans et al., 2011; Bell et al., 2011; Vuitton et al., 2012; Westlake et al., 2012; Mandt et al., 2012; Plessis et al., 2012; Moreno et al., 2012; H´ebrard et al., 2013; Lara et al., 2014; Dobrijevic et al., 2014; Krasnopolsky, 2014; Orton et al., 2014; Loison et al., 2015, and references therein). Our knowl- edge of relevant reaction rate coefficients and overall reaction pathways and products is also continually improving due to new laboratory investigations or theoretical calculations (with references too numerous to list). A recent theoretical advance is the new 2D seasonal photochemical modeling of Hue et al. (2015), which takes into account the expected seasonal variation in temperatures. Full details of methane photochemistry on Saturn and the other giant planets can be found in Atreya et al. (1984), Strobel (1983, 2005), Glad- stone et al. (1996), Yung and DeMore (1999), and Moses et al. (2000a, 2005). Methane photolysis leads to the production of CH, CH2 (both the ground-state triplet and the excited singlet) and CH3. The CH radicals tend to favor unsaturated hydrocar- bon production, whereas the CH3 radicals tend to recycle the methane or lead to the production of saturated hydro- carbons. The peak hydrocarbon production region at 10−3– 10−4 mbar occurs just below the methane homopause where the Lyman alpha photons are absorbed by CH4, but there is a secondary peak in the 0.1-10 mbar region due to pho- tosensitized destruction of CH4 resulting from photolysis of C2H2 and other hydrocarbon photochemical products. This secondary production region can particularly affect the rel- ative abundances of the hydrocarbons at the pressures at which the infrared observations are sensitive. As we shall see, the hydrocarbon abundances vary with both altitude and latitude. One obvious cause of this varia- tion is seasonal change: the mean daily solar insolation, at- mospheric temperatures, and perhaps circulation patterns on Saturn change with season, inducing a response in the vertical and meridional distributions of chemical species. Although three-dimensional dynamical models that include chemistry have not yet been developed for Saturn or the other giant planets, Moses and Greathouse (2005) have in- vestigated seasonal stratospheric chemistry on Saturn with a 1D time-variable photochemical model, exploring how the molecular abundances change solely due to seasonally- varying solar insolation. The results of this model are com- pared to the measured distributions of hydrocarbons in Fig. 10.16 at a variety of different pressure levels, showing the ex- pected seasonal variability based on photochemistry alone. This figure will be referred to throughout this section. The Moses and Greenhouse model includes the effects of ring shadowing and solar-cycle variations, but neglects horizontal transport (e.g., Moses et al., 2007) or time-variable temper- atures and vertical winds. The hemispheric asymmetries in the hydrocarbon abundances are more pronounced at higher altitudes where vertical diffusion time scales and chemical lifetimes are short. Our theoretical discussion of the princi- ple production and loss mechanisms for each species will be guided by the results of this 1D model. Note, also, the recent 2D (latitude and altitude) study by Hue et al. (2015) that extends the Moses and Greathouse (2005) study by consid- ering the effects of seasonally varying temperatures. 20 Fletcher, Greathouse, Guerlet, Moses & West C2 hydrocarbons Observations and models both demonstrate that C2H6 is the most abundant hydrocarbon photochemical product in Saturn’s stratosphere. Ethane is produced predominantly by CH3 recombination throughout the stratosphere and by se- quential addition of H to unsaturated C2Hx hydrocarbons in the lower stratosphere; ethane is lost via photolysis and reac- tion with C2H produced from C2H2 photolysis. Production of C2H6 exceeds loss at most altitudes in the Saturn mod- els, and the net C2H6 production is balanced by downward flow in to the deeper troposphere, where the ethane will eventually be thermochemically converted back to methane. The chemical loss time scale for C2H6 exceeds a Saturn year at most altitudes below its peak production region, making C2H6 particularly stable and thus a good tracer for local dynamics and global circulation. Ethylene is produced through the reaction of CH with CH4, reactions of C2H3 with H and H2, and through C2H6 photolysis. It is lost through H-atom addition to form C2H5 Figure 10.16 Comparing zonal mean hydrocarbon distributions measured for five species (ethane, acetylene, propane, methylacetylene and diacetylene) with the photochemical predictions of Moses and Greathouse (2005). Observers have been abbreviated as follows: Gre05 (Greathouse et al., 2005); Pra06 (Prang´e et al., 2006); How07 (Howett et al., 2007); Hes09 (Hesman et al., 2009); Gue09 (Guerlet et al., 2009); and Sin13 (Sinclair et al., 2013). The model output (labelled MG05) is shown for the solstices and equinoxes to show the predicted magnitude of seasonal variability, particularly for p < 0.1 mbar: Ls = 270◦ is dotted (purple), Ls = 0◦ is solid (green), Ls = 90◦ is dashed (orange) and Ls = 180◦ is dot-dashed (red). The colors of the individual measurements are designed to represent the seasonal timing of the observations. The model abundances of C2H6, CH3C2H and C3H8 of Moses and Greathouse (2005) have been approximately scaled to match the data, as described in the figure legends. Note that C3H8 exhibits a similar meridional trend to C2H6, whereas all the other species tend to track C2H2. and through photolysis. Unlike C2H6, ethylene can be pho- tolyzed by UV photons with wavelengths out to ∼200 nm, which makes the C2H4 much less stable in the lower strato- Mole Fraction (ppb)j00.511.5Planetocentric Latitude (deg)-90-60-300306090Guerlet et al. (2010)0.11 mbarC4H2Planetocentric Latitude (deg)-90-60-300306090aMole Fraction (ppm)00.511.51.1 mbarC2H2Gre05Pra06How07Hes09Gue09MG05Sin13, 2005Sin13, 2009Sin13, 201005101520Mole Fraction (ppm)Planetocentric Latitude (deg)-90-60-300306090f2.1 mbarC2H6MG05 model/1.8Gre05Hes09How07 /1.8Gue09Sin13, 2005Sin13, 2009Sin13, 2010Planetocentric Latitude (deg)-90-60-300306090Mole Fraction (ppm)0510150.11 mbarC2H2Greathouse et al. (2005)Guerlet et al. (2009)bMole Fraction (ppm)05101520Planetocentric Latitude (deg)-90-60-300306090g0.11 mbarC2H6MG05 model/1.8Guerlet et al. (2009)0102030Planetocentric Latitude (deg)-90-60-300306090Mole Fraction (ppm)C2H60.012 mbarMG05 model/1.8Guerlet et al. (2009)hdMole Fraction (ppb)00.511.52Planetocentric Latitude (deg)-90-60-3003060901.1 mbarCH3C2HMG05 model/3Guerlet et al. (2010)Guerlet et al. (2010)eMole Fraction (ppb)0246Planetocentric Latitude (deg)-90-60-3003060900.11 mbarCH3C2HiMole Fraction (ppm)00.10.20.3Planetocentric Latitude (deg)-90-60-3003060901.1 mbarC3H8MG05 model x 1.8Guerlet et al. (2009)Planetocentric Latitude (deg)-90-60-300306090cMole Fraction (ppm)05100.012 mbarC2H2Guerlet et al. (2009)Saturn’s seasonally changing atmosphere: thermal structure, composition and aerosols 21 sphere than ethane. Its chemical lifetime is shorter than an Earth year at most altitudes, although recycling can help keep it around longer than its loss time scale would indicate. The very low levels of C2H4 under quiescent conditions on Saturn are consistent with photochemical models; its greatly enhanced abundance in the northern hemispheric storm bea- con region (Hesman et al., 2012, and Chapter 13) is most likely the result of the high dependence of the C2H3 + H2 → C2H4 + H reaction on temperature (e.g., Tautermann et al., 2006; Armstrong et al., 2014; Moses et al., 2014). Acetylene is the second-most abundant hydrocarbon pho- tochemical product on Saturn. Its chemistry is more intri- cate, complicated, and non-linear, and C2H2 is a particularly important parent molecule for other observed hydrocarbons such as CH3C2H and C4H2. Acetylene is produced primar- ily from the photolysis of ethane and ethylene, although “re- cycling” reactions such as C2H + H2 → C2H2 + H, H + C2H3 → C2H2 + H2, and C2H + CH4 → C2H2 + CH3 dominate the total column production rate of C2H2 in the stratosphere. Acetylene is indeed rapidly recycled through- out the stratosphere, although there is a steady leak out of the ongoing recycling reactions into C2H6, C4H2 and higher-order hydrocarbons. Acetylene has a lifetime inter- mediate between that of C2H6 and C2H4: its pure chem- ical loss time scale is of order of a Saturn season through much of the stratosphere (∼0.03-100 mbar, see Moses and Greathouse, 2005), but recycling reactions give it a much longer “effective” lifetime. In the absence of atmospheric circulation, photochemically- generated species at microbar pressures (e.g., Fig. 10.16) tend to have abundances that follow local insolation con- ditions, resulting in higher abundances in the summer hemisphere compared to the winter hemisphere (Moses and Greathouse, 2005). At lower altitudes, however, vertical diffusion time scales and chemical lifetimes are much longer, and the meridional distributions track the yearly average of the mean daily solar insolation rather than the instanta- neous solar fluxes. Because the yearly average insolation is largest at the equator and decreases toward the poles, the photochemical models predict that most of the hydrocarbon photochemical products should have a maximum abundance at the equator, decreasing smoothly toward both poles at pressures of 1 mbar and greater. These predictions can be tested by comparison to latitudinally-resolved hydrocarbon measurements. Ethane and acetylene in southern summer Greathouse et al. (2005) were the first to present latitudinally-resolved distributions of the principle methane photolysis products, ethane and acetylene, in Saturn’s southern (summer) hemisphere (Fig. 10.16a). Using IRTF data acquired in 2002, they showed that acetylene decreased from equator to pole at the 1-mbar pressure level (in agree- ment with photochemical model predictions), whereas the opposite trend was marginally observed for ethane at 2 mbar. These results suggested that dynamical redistribu- tion is effective in Saturn’s stratosphere and operates on timescales less than ethane’s chemical lifetime of ∼700 years and longer than acetylene’s lifetime of ∼100 years (Moses and Greathouse, 2005). Since the work of Greathouse et al. (2005), the distribu- tions of ethane and acetylene have been routinely measured by the Composite Infrared Spectrometer (CIRS) onboard Cassini, using a combination of nadir and limb sounding. Nadir and limb observations provide complementary infor- mation. On the one hand, CIRS nadir data provide an ex- cellent spatio-temporal coverage of Saturn’s atmosphere and are mainly sensitive to the abundance of ethane and acety- lene, with a peak sensitivity near the ∼ 2-mbar pressure level. Under particular conditions, such as the high tempera- tures encountered within the storm beacon region (see Chap- ter 13), other trace species can also be measured (such as ethylene, which has been measured by Hesman et al., 2012). On the other hand, CIRS data acquired in limb-viewing ge- ometry allow the retrieval of the vertical abundance profile of ethane and acetylene over a greater pressure range (typi- cally 5 mbar – 5 µbar), but with a much sparser meridional and temporal coverage. Five Cassini studies have reported on the meridional dis- tribution of ethane and acetylene during southern summer- time based on Cassini/CIRS measurements (Fig. 10.16a): three based on nadir data (Howett et al., 2007; Hesman et al., 2009; Sinclair et al., 2013) and two based on limb observations (Guerlet et al., 2009; Sylvestre et al., 2015). Although these studies do not necessarily agree quantita- tively, they report similar trends. Howett et al. (2007) mea- sured the southern hemisphere hydrocarbon distributions in 2004 between 10 − 70◦S, showing a decrease of acety- lene poleward of 30◦S. They also confirmed that the ethane abundance is relatively uniform with latitude equatorward of 50◦S, but suggested a rather large increase poleward of 50◦S. This increase has not been reproduced by subsequent investigators, and could be ascribed to a bias with emission angle (as the higher latitude data were acquired at higher emission angles) linked to spectroscopic errors. Furthermore, all follow-up studies based on CIRS measurements rely on an updated ethane linelist by Vander Auwera et al. (2007), which has line intensities higher by approximately 30% than the values used by Howett et al. (2007). Hesman et al. (2009) then extended the CIRS retrievals towards the south pole in 2005, and combined them with low-latitude ground- based observations. They confirmed acetylene’s poleward de- crease but found a sharp rise in C2H2 right at the sum- mer pole (87S), within the stratospheric vortex identified by Orton and Yanamandra-Fisher (2005) and Fletcher et al. (2008). This sharp increase has also been observed for ethane (Fletcher et al., 2015), and suggests enhancement by strong subsidence within the summer polar vortex. Using limb ob- servations acquired in 2005-2006 covering both hemispheres (80◦S–45◦N), Guerlet et al. (2009) confirmed that ethane and acetylene follow different meridional trends in the 1– 5 mbar pressure range: while acetylene generally decreases from the equator towards both poles, the ethane distribu- tion is much more homogeneous with latitude. These results are also in agreement with the study of Sinclair et al. (2013) from CIRS nadir observations covering both hemispheres. Figs. 10.16a and 10.16f compare the meridional distribu- tion of ethane and acetylene at the 1-2 mbar pressure level during southern summer, as measured by various authors, 22 Fletcher, Greathouse, Guerlet, Moses & West and as predicted by the photochemical model of Moses and Greathouse (2005). The chemistry of C2H2 and C2H6 is highly linked, such that the C2H2 meridional distribution tends to track the C2H6 distribution in photochemical mod- els (e.g., Moses and Greathouse, 2005; Moses et al., 2007; Hue et al., 2015). The notable differences in the meridional profiles of C2H2 and C2H6 on Saturn is therefore a surprise and suggests the influence of meridional transport on long timescales, as already proposed by Greathouse et al. (2005). Guerlet et al. (2009) also report a sharp and narrow equa- torial maximum in the C2H2 distribution at 1 mbar (and a more moderate local maximum of C2H6), much higher than predicted by the seasonal photochemical model. The authors interpret this strong maximum as the signature of a local subsidence associated with the equatorial oscillation. Indeed, given that C2H2 mixing ratio increases with height, a downwelling wind would carry C2H2 and enrich lower al- titude regions. At sub-millibar pressure levels, Guerlet et al. (2009) took advantage of limb-viewing geometries from Cassini to de- rive the distribution of acetylene and ethane in the upper stratosphere (p < 0.1 mbar, Fig. 10.16b,c,g,h). Contrary to the trends observed in the lower stratosphere, they find that both species follow very similar trends in the upper strato- sphere. These distributions of the two hydrocarbons are asymmetric, with volume mixing ratios 2 to 6 times higher at northern mid-latitudes (with a local maximum centered at 25◦N) than at southern mid-latitudes. This is opposite to what might be expected from hydrocarbon production rates alone, where production rates should have been higher in the summer southern hemisphere than in the winter hemisphere (see Fig. 10.16c,h). Guerlet et al. (2009) hence suggest that a strong meridional transport from the summer to the winter hemisphere, occurring on seasonal timescales, is responsible for the observed enrichment at 25◦N. This hypothesis is con- sistent with the predictions of the global circulation model of Friedson and Moses (2012), in which stratospheric circu- lation cells develop, with a descending branch at 25◦N at this season. Temporal trends in ethane and acetylene The southern summer observations by Cassini revealed asymmetries in ethane and acetylene that would be ex- pected to shift over time, potentially reversing by northern summer solstice in 2017 (Ls = 90◦). Sinclair et al. (2013) were the first to study temporal variability in the distri- butions of ethane and acetylene using nadir sounding at low spectral resolution (15 cm−1) between 2005 and 2010 (Ls = 308− 15◦). They confirmed (i) the absence of a sharp south-poleward rise in ethane as first reported by Howett et al. (2007) from early CIRS nadir data (Ls ≈ 295◦); and (ii) the equatorial peak abundance of acetylene observed in 2005 by Guerlet et al. (2009), although this maximum ap- pears more muted in the nadir data (see Fig. 10.16a). Differ- ences between limb and nadir results in the equatorial region could be ascribed to the temporal evolution of the equatorial oscillation, which strongly perturbs the thermal (and chem- ical) structure of the stratosphere on small vertical scales not fully resolved with nadir, low spectral resolution data. Figure 10.17 Differences in zonal mean retrievals of ethane and acetylene at 2 mbar between 2005 and 2009/2010 from Cassini/CIRS, as presented by Sinclair et al. (2013). Trends are shown in units of the uncertainty (σ) to reveal where the trends are statistically significant. In general, the results show a decline in abundance over the southern hemisphere and an increase over the northern hemisphere, suggestive of large-scale inter hemispheric circulation with subsidence in northern winter/spring. Similar conclusions have been reached using temporal trends in C2H2 and C2H6 distributions at higher latitudes (Fletcher et al., 2015). Away from the equator, the Cassini/CIRS observations are beginning to reveal seasonal shifts in the zonal-mean ethane and acetylene distributions, as shown in Fig. 10.17. At mid-latitudes, Sinclair et al. (2013) report significant en- richments in ethane (and to a lesser extent, in acetylene) at the 2-mbar pressure level in the region 20 − 65◦N and depletions at 15◦S, which they interpret as extended subsi- dence occurring in the northern hemisphere between 2005 and 2010, and upwelling at 15◦S. Sinclair et al. (2013) also showed that the equatorial maximum of C2H2 decreases over time (while the ethane concentration remains fairly con- stant), possibly due to the evolution of the equatorial os- cillation. At higher latitudes, Fletcher et al. (2015) use ten years of CIRS nadir data to demonstrate that the summer- time south polar maximum in both C2H2 and C2H6 has been declining throughout the timespan of the observations (Ls = 293− 54◦) at 1-2 mbar, mirrored by enhancements in both species in the north polar spring (poleward of 75◦N). These changes in the millibar region cannot be explained solely in terms of photochemistry, so that the authors in- voke stratospheric circulation from south to north (broadly speaking, upwelling in the autumnal hemisphere and sub- sidence in the spring hemisphere) to explain the observed changes. In many of these studies (Guerlet et al., 2009; Sinclair et al., 2013; Fletcher et al., 2015), the temporal behavior of C2H2 and C2H6 were found to be different (ethane vari- ations appeared more significant than those of acetylene), despite the highly linked chemistry of the two species, and suggestive of the influence of dynamics. Acetylene is ex- pected to behave differently from ethane in the presence of vertical winds, for example. In the ∼0.1-10 mbar region, downwelling winds could carry more C2H2 from high alti- tudes to lower altitudes, but the increased C2H2 at these lower altitudes results in a non-linear chemical loss rate due to the increased photolysis (and thus C2H and H produc- Saturn’s seasonally changing atmosphere: thermal structure, composition and aerosols 23 tion) in conjunction with the resulting increased rates of the non-recycling loss reactions such as C2H + C2H2 → C4H2 + H and H + C2H2 + M → C2H3 + M. Thus, the subsidence will lead to an increased fraction of the C2H2 photolysis products being permanently converted to other hydrocarbons rather than recycling the acetylene, resulting in less of an increase than one would expect based on sub- sidence alone without chemical coupling. By the same to- ken, upwelling winds will decrease the C2H2 abundance and loss rate non-linearly, favoring recycling over permanent loss. Upwelling and downwelling winds can therefore lead to less significant decreases and increases in the C2H2 mixing ra- tios as compared to C2H6, which might help explain some of the temporal behavior observed by Sinclair et al. (2013); however, the slope of the unperturbed mixing-ratio profile also contributes to the observed magnitude of the increase or decrease during the upwelling/downwelling (e.g., Guerlet et al., 2009), so the resulting response will be complicated. Coupled 2D and 3D photochemical-dynamical models will likely be needed to fully explain the temporal and merid- ional variations of stratospheric hydrocarbons in the Cassini CIRS observations. No such models have been published to date. At higher altitudes, seasonal changes in the hydrocar- bon distributions have also been studied by Sylvestre et al. (2015), who analyzed CIRS limb data acquired in 2010-2012 and compared those results to the 2005-2006 observations by Guerlet et al. (2009). Contrary to Sinclair et al. (2013), Sylvestre et al. (2015) find little change in the ethane and acetylene distribution in the millibar region. However, they find that the local enrichment in ethane and acetylene pre- viously observed at 25◦N in 2005–2006 (especially at p<0.1 mbar) has since disappeared. If subsidence at 25◦N was occurring during the northern winter, then it has stopped by northern spring, qualitatively consistent with the predic- tions of the seasonally-reversing Hadley circulation proposed by Friedson and Moses (2012). Alternatively, the variability at 25◦N could be coupled to secondary meridional circula- tions induced by the quasi-periodic equatorial oscillations whose stacked pattern is descending with time. Despite quantitative differences between spatial distribu- tions and temporal behavior identified by different authors, the general trends for ethane and acetylene are consistent: (i) C2H2 decreasing from equator to pole whereas C2H6 is largely uniform with latitude; (ii) regional enhancements of both species at the highest latitudes within polar vortices, possibly associated with auroral chemistry and dynamic en- trainment; (iii) general spring hemisphere enhancements and autumn hemisphere depletions in the millibar region, sug- gestive of stratospheric circulation from the southern to the northern hemisphere; and (iv) complex temporal behavior in the tropics associated with vertical propagating waves and a seasonally-reversing Hadley-type circulation. As ethane and acetylene are also the principal stratospheric coolants, their spatial and temporal behavior must also be incorporated in radiative climate models to better explain the thermal asymmetries observed by Cassini (see Section 10.2). C3 and higher hydrocarbons In addition to ethane and acetylene, Cassini/CIRS has also revealed hemispheric asymmetries in the higher order hy- drocarbons. Although studies of their temporal evolution are rather limited at the time of writing, we review here the main conclusions from both theory and observations. Propane: Propane is produced in the photochemical models through termolecular reactions, which are more ef- fective at higher pressures. Hence, the production rate for C3H8 peaks near ∼3 mbar, where photosensitized CH4 de- struction occurs, rather than at higher altitudes where CH4 is destroyed by Lyman alpha photolysis. Propane is lost predominantly through photolysis. Because it is shielded to some extent by C2H6 and C2H2, propane is relatively stable in the models. Fig. 10.16i indicates that propane is expected to show a similar latitudinal trend to C2H2, with a decline from equator to pole in both hemispheres. However, the first spatial distribution of propane provided by Greathouse et al. (2006) found that propane at the 5-mbar level was relatively uniform in the southern hemisphere (based on two measurements at 20◦S and 80◦S). Guerlet et al. (2009) derived propane from Cassini/CIRS limb observations be- tween March 2005 and January 2008 (Ls = 300 − 340◦), with a peak sensitivity near 1-mbar. They pointed out that the meridional behavior of C3H8 closely tracks the instan- taneous solar insolation (i.e., more propane in the sum- mer hemisphere), suggesting that the chemical lifetime of C3H8 is shorter than the models indicate. Another possibil- ity is that the C3H8 abundance is sensitive to temperature through reactions that have a higher degree of temperature dependence than is assumed in the models (Dobrijevic et al., 2011). In fact, models tend to underpredict the C3H8 abun- dance on Saturn (Greathouse et al., 2006; Guerlet et al., 2009; Dobrijevic et al., 2011). Most recently, Cassini/CIRS limb observations in 2010 (Sylvestre et al., 2015) suggest little to no change since 2005-2006. Methylacetylene and diacetylene: Guerlet et al. (2010) continued to exploit the same Cassini/CIRS limb data (Ls = 300 − 340◦) to provide the first latitudinally- resolved distributions of methylacetylene (CH3C2H) and di- acetylene (C4H2), finding that mid-southern latitudes are depleted in both hydrocarbons compared to mid-northern latitudes during southern summer. The photochemical mod- els do not predict such a hemispheric asymmetry during this season (see Fig. 10.16), and Guerlet et al. (2010) suggest that the behavior is caused by upwelling at mid-southern latitudes and subsidence at mid-northern latitudes. Methy- lacetylene and diacetylene are important hydrocarbon pho- tochemical products on Saturn that have much shorter chemical lifetimes than C2H6, C2H2, and C3H8. In the up- per atmosphere where CH4 is photolyzed by Lyman α and other short-wavelength radiation, CH insertion into C2H6, C2H2, and C2H4 initiates the production of C3Hx hydro- carbons, which then can be photolyzed or react with hy- drogen or other species to eventually produce CH3C2H (see Moses et al., 2000a, 2005, for details). At pressures greater than ∼0.01 mbar, CH3C2H has additional sources through photolysis of C4Hx species and reactions such as CH3 + 24 Fletcher, Greathouse, Guerlet, Moses & West C2H3, which produce C3Hx species that can again make their way to forming CH3C2H. In this lower-altitude region, acetylene is an important “parent” molecule for CH3C2H. Methylacetylene is lost through photolysis and H-atom ad- dition. The primary (non-recycling) mechanism for C4H2 production is acetylene photolysis followed by C2H + C2H2 → C4H2 + H. Diacetylene formation therefore depends non- linearly on the acetylene abundance. H-atom addition to form C4H3 is the dominant loss reaction for C4H2, but the resulting C4H3 radical can react with H to recycle the C4H2. Photolysis is an effective loss process, although it, too, leads to some C4H2 recycling. Diacetylene is expected to condense in the lower stratosphere to contribute to a substantial frac- tion of the stratospheric haze burden (e.g., Moses et al., 2000b). The meridional distribution of CH3C2H (Fig. 10.16d- e)and C4H2 (Fig. 10.16j) in the 0.05-1 mbar region appears to grossly track that of C2H2 (Guerlet et al., 2010), which makes sense theoretically given the short lifetimes of methy- lacetylene and diacetylene and given that C2H2 is the key “parent” molecule in the middle and lower stratosphere for both these species. The C4H2 abundance is typically well reproduced in 1D global-average models when the predicted C2H2 profiles match observations, suggesting that the C4H2 chemistry is well understood, but global-average photochem- ical models tend to notably overpredict the CH3C2H abun- dance in the 0.1-1 mbar region (e.g., Moses et al., 2000a, 2005; Guerlet et al., 2010; Dobrijevic et al., 2011), suggest- ing that the CH3C2H chemistry is not well described in the models. The meridional variations of CH3C2H and C4H2 ap- pear more extreme than for C2H2, which is likely the result of the non-linear nature of their dependence on C2H2 photo- chemistry; however, these species also exhibit smaller-scale variations and differences between their respective distribu- tions that are not easily explained with the chemical mod- els. Dobrijevic et al. (2011) demonstrate that rate-coefficient uncertainties lead to a large spread in the predicted C3H8, CH3C2H and C4H2 abundances in the photochemical mod- els, which could explain these model-data mismatches. Guer- let et al. (2010) suggest that seasonally variable transport is affecting the C4H2 and CH3C2H distributions. Benzene: Benzene photochemistry is not well under- stood under conditions relevant to Saturn. Neutral C6H6 photochemical production and loss mechanisms on the gi- ant planets are discussed by Moses et al. (2000a, 2005) and Lebonnois (2005). Recent CIRS measurements of the C6H6 column abundance show that benzene is at least as abun- dant in the polar stratosphere (80◦S) compared to equa- torial regions, in contradiction with photochemical model predictions including only neutral chemistry (Guerlet et al., 2015). The latter authors conclude that ion chemistry in the auroral regions plays an important role in the benzene pro- duction rates, as is the case on Jupiter (Wong et al., 2000; Friedson et al., 2002; Wong et al., 2003) and for non-auroral ion chemistry on Titan (Vuitton et al., 2009). The apparent presence of high-altitude hazes on Saturn observed during VIMS stellar occultations (Bellucci et al., 2009; Kim et al., 2012) and by CIRS (Guerlet et al., 2015) further suggests that Titan-like ion chemistry is contributing to the produc- tion of complex organics on Saturn. Benzene, like C4H2, is also expected to condense to form haze particles in Saturn’s lower stratosphere. In summary, Cassini has determined the spatial distribu- tion of Saturn’s stratospheric hydrocarbons (ethane, acety- lene, propane, methylacetylene, diacetylene and benzene), discovering asymmetries during summertime conditions that bear little resemblance to the photochemical model pre- dictions in the absence of circulation. Auroral chemistry is a possible contributor to increased abundances in the high-latitude regions, which should be incorporated into the radiative budget in the polar regions. Coupled 2D and 3D photochemical-dynamical models will likely be needed to fully explain the temporal and meridional variations of stratospheric hydrocarbons in the Cassini CIRS observa- tions, but no such models have been published to date. Oxygen species Finally, the reducing nature of Saturn’s stratospheric com- position is perturbed by a steady influx of oxygenated species from external sources, such as micrometeoroid pre- cipitation, cometary impacts or a connection with Saturn’s rings and satellites. These oxygen compounds can generate new photochemical pathways to form unexpected molecules, attenuate UV flux or provide condensation nuclei (Moses et al., 2000b). The history of ground-based observations of oxygen compounds, using rotational lines in the far-IR and sub-millimeter, was reviewed by Fouchet et al. (2009). Since the time of that review, investigators have continued to study disk-integrated stratospheric CO and H2O from ground-based (Cavali´e et al., 2009, 2010) and space-based (Herschel, Hartogh et al., 2011; Fletcher et al., 2012a) ob- servatories, and Cassini/CIRS has demonstrated an absence of latitudinal CO2 trends for p < 10 mbar (Abbas et al., 2013). However, we discuss oxygen compounds briefly here be- cause, although the chemical and diffusion timescales for CO, H2O and CO2 are too long to generate seasonal asym- metries, the water column abundance itself should change with season due to the altering thermal structure. Because water condenses relatively high in the stratosphere, the far- IR and sub-mm observations will be very sensitive to the altitude at which the water condenses, which is in turn con- trolled by the stratospheric temperatures. If the source is isotropic (interplanetary dust) or favors low latitudes (e.g., from Enceladus, Hartogh et al., 2011), we would expect to see a larger water column in the summer hemisphere and/or wherever mid-to-low stratospheric temperatures are highest. If the source is from the ring atmosphere (O’Donoghue et al., 2013; Tseng and Ip, 2011), then the source itself is seasonal (greater ring atmosphere and thus greater influx when the ring has its highest opening angle at the solstices). The ex- ternal oxygen supply is discussed in Chapter 9. Saturn’s seasonally changing atmosphere: thermal structure, composition and aerosols 25 10.4 Clouds and hazes 10.4.1 Pre-Cassini reflectivity studies Cloud and haze particles provide yet another potential av- enue to probe important atmospheric processes, with both obvious and subtle ties to the wind, temperature, and chem- ical tracer fields. Although observation of clouds and haze is relatively straightforward, retrieval of vertical profiles, par- ticle size, shape and composition and relationships to the other fields is not. Both types of particles can be classified generically as aerosols. We usually refer to ‘clouds’ in the context of condensation/sublimation of volatile constituents (NH3, NH4SH and H2O/NH3) in the upper troposphere, while the term ‘haze’ is used in association with the solid phase of photochemical products: P2H4 is expected to exist in the upper troposphere (although this has yet to be con- firmed by observations); hydrocarbons and condensed ex- ternal water in the stratosphere (Moses et al., 2000a). The two may combine: stratospheric haze sedimenting from high altitudes may serve as condensation nuclei for condensates deeper down. Conversely, ice crystals in tropospheric clouds may acquire a photochemically-produced hydrocarbon coat- ing. The reader is referred to West et al. (2009) for a com- prehensive review of the literature on clouds and hazes, as a full treatment of aerosol composition and derived properties is beyond the scope of this chapter. Common features of the various reflectivity investiga- tions (e.g., Karkoschka and Tomasko, 1992, 1993; Stam et al., 2001; Temma et al., 2005; P´erez-Hoyos et al., 2005; Karkoschka and Tomasko, 2005) include (a) a stratospheric haze (1 < p < 90 mbar) of small radius (r ≈ 0.1 − 0.2 µm) particles, possible originating from photochemical pro- cesses; (b) a tropospheric haze from the tropopause down to the first condensation cloud deck at 1.5-2.0 bar, possi- bly with aerosol-free gaps in the vertical distribution; and (c) a possible thick NH3 cloud, although signatures of fresh NH3 ice have so far only been detected in Saturn’s 2010-11 storm region (Sromovsky et al., 2013). Numerous explana- tions have been presented for the concealment of conden- sate signatures, including large particle sizes, non-spherical particle shapes, coating of the pure ices by photochemical products sedimenting downwards (Atreya and Wong, 2005; Kalogerakis et al., 2008), or by mixing of the condensate phases to mask the spectral signatures. The polar strato- spheric haze appears distinct from all other latitudes, being optically thicker and darker at UV wavelengths, with strong Rayleigh-like polarisation suggestive of the importance of auroral processes in their formation. High spatial resolution limb images have revealed distinct haze layers at some lati- tudes (e.g., Rages and Barth, 2012). The equatorial zone, be- tween ±18◦ latitude, features consistently high clouds with perturbations from major storms (P´erez-Hoyos et al., 2005). The vertical distributions of the tropospheric aerosols have been well-mapped with ground-based and Hubble data, and Cassini is beginning to add to this picture. Here we focus on the seasonal and some non-seasonal behavior of clouds and haze on the scale of the zonal jets. Remote sensing measurements of clouds and haze probe the upper troposphere and stratosphere, and it is in this regime where seasonal effects from atmospheric heating and pho- tochemistry are most likely to have an effect. CCD images in near-infrared methane bands provide a measure of cloud top altitude and haze/cloud density, and it is not surprising that seasonally varying hemispheric asymmetries are appar- ent in ground-based and space-based images in methane ab- sorption bands. Other diagnostics of cloud and haze vertical structure include polarization at phase angles near 90 de- grees and ultraviolet reflectivity. Polarization is sensitive to cloud altitude provided that it is dominated by Rayleigh scattering above a non-polarizing cloud. This is the case for low and middle latitudes but not so for high latitudes on Saturn and Jupiter at blue wavelengths. At high lat- itudes haze particles also provide strong polarization. Ul- traviolet reflectivity is diagnostic of cloud altitude provided that it is also dominated by Rayleigh scattering from gas. Again, this seems to hold reasonably well at low latitude but not at high latitudes where UV-absorbing particles are abundant in the stratosphere. The latitudinal behavior of the highly-polarizing, UV-absorbing stratospheric aerosols is most likely a result of auroral energy deposition at high latitudes (Pryor and Hord, 1991; West and Smith, 1991); polar aerosols are discussed in Chapter 12. The record of reflectivity measurements now span more than an entire seasonal cycle on Saturn. The Pioneer 11 flyby occurred just prior to northern spring equinox (Ls = 354◦); the Voyager 1 and 2 encounters were a little after north- ern spring equinox (Ls = 8.6 − 18.2◦, respectively). Near the northern spring equinox in the early 1980s, methane- band imagery from 1979 (West et al., 1982), polarization imagery from Pioneer 11 in 1979, and Voyager 2 photopo- larimeter scans in 1981 (Lane et al., 1982) all showed hemi- spheric contrasts at mid-latitudes consistent with deeper clouds and/or a smaller column density of haze in the south- ern mid-latitudes during early northern spring. Tomasko and Doose (1984) summarized those results using simple cloud/haze models showing effective cloud top pressures at 40◦N and 40◦S to be near 300 and 460 mbar, respectively. Viewing geometry and spatial resolution limited the north- ern analysis to latitudes less than 50− 60◦N, while at south- ern latitudes the methane-band imagery indicated the effec- tive cloud top pressure rising poleward of 40◦S to 320 mbar at 70◦S. The observed hemispheric difference was initially attributed to higher static stability and suppressed convec- tion at southern latitudes after an extended period of solar heating during southern summer. The equinoctial snapshot provided by Pioneer and Voy- ager was supplemented by ground- and space-based observa- tions prior to Cassini’s observations, most notably using the Hubble Space Telescope. During northern spring and sum- mer, Karkoschka and Tomasko (1992) studied ground-based images from 1986 to 1989; and Karkoschka and Tomasko (1993) presented some of the earliest Hubble imaging in 1991 (Ls = 68− 130◦). Stam et al. (2001) presented ground- based images of Saturn near the northern autumn equinox 26 Fletcher, Greathouse, Guerlet, Moses & West in 1995 (Ls = 176◦), able to view both northern and south- ern hemispheres simultaneously. A seasonal asymmetry was observed, with both the tropospheric and stratospheric opti- cal thicknesses appearing larger over northern mid-latitudes (approaching autumn) than over southern mid-latitudes (approaching spring), interpreted by Stam et al. (2001) as a thicker tropospheric cloud deck in the autumn hemisphere. As the northern hemisphere receded from view and Sat- urn approached southern summer solstice, images from the Hubble Space Telescope (HST) provided a data set of sig- nificant value to the study of seasonal and non-seasonal behavior of Saturn’s haze and clouds. P´erez-Hoyos et al. (2005, 2006) used HST data to study the equator and south- ern hemisphere between 1994-2003 (Ls = 158 − 286◦); and Karkoschka and Tomasko (2005) presented analyses of 134 images of Saturn taken by the Hubble Space Tele- scope between 1991 and 2004 (Ls = 130 − 289◦), and per- formed a principal-component analysis of many latitudes on Saturn. Four statistically-meaningful principal compo- nents emerged. The first principal variation is a strong mid- latitude variation of the aerosol optical depth in the upper troposphere. This structure shifts with Saturn’s seasons, but the structure on small scales of latitude stays constant. This is what is most apparent in a casual comparison of images taken in different seasons. The second principal variation is a variable optical depth of stratospheric aerosols. The optical depth is large at the poles and small at mid- and low lati- tudes, with a steep gradient in between. This structure re- mains essentially constant in time. The third principal vari- ation is a variation in the tropospheric aerosol size, which has only shallow gradients with latitude, but large seasonal variations. Aerosols are largest in the summer and smallest in the winter, broadly consistent with the 1980s-equinox ob- servation of a haze free southern autumn hemisphere. The fourth principal variation is a feature of the tropospheric aerosols with irregular latitudinal structure and fast vari- ability, on the time scale of months. 10.4.2 Cassini’s observations of seasonal aerosol changes The seasonal asymmetry in tropospheric aerosols was there- fore well established prior to Cassini’s arrival just after southern summer solstice. The tropospheric haze optical thickness was expected to be the largest and most ex- tended in the summer hemisphere, and smallest in the winter hemisphere, with the transition occurring at some time near to the equinox as previously observed by Voy- ager (Ls = 8 − 18◦) and Pioneer (Ls = 354◦). The Cassini spacecraft arrived at Saturn at an earlier seasonal phase (Ls ≈ 290◦, southern summer), and has now provided the opportunity to track these changing hemispheric asymme- tries. Indeed, Cassini observations pre-equinox revealed a hemispheric asymmetry that was opposite to those seen by Voyager post-equinox. As shown in Fig. 10.18, the high northern latitudes showed a vibrant blue color in 2004. The interpretation (Edgington et al., 2012) is that Rayleigh scat- tering by gas molecules is responsible, and that the colored haze material is suppressed in the northern (winter) lati- Figure 10.18 This true-color image was obtained in 2004 by the Cassini ISS camera on approach to Saturn. It illustrates the strong hemispheric color difference observed during southern summer. Fig. 10.1 shows how Saturn’s colors evolved with time during the Cassini mission. Courtesy NASA/JPL-Caltech. tudes relative to southern (summer) latitudes. Subsequent Cassini ISS images have shown that the blue color persisted into 2008 but by 2009 (near equinox) the blue color had dis- sipated at northern latitudes (see the images in Fig. 10.1). At the current epoch (2014) the southern high latitudes are be- ginning to show a blue color as they recede into winter con- ditions. These observations are consistent with the idea that the blue color indicates reduced production of haze through- out the winter season. These inferences from the Cassini/ISS visible data are consistent with infrared observations from Cassini/VIMS, particularly at 5 µm where the dearth of hydrogen and methane opacity permits the escape of radiation from rel- atively deep (3-6 bar) levels, and cloud opacity serves to attenuate this 5-µm flux. As first reported by Baines et al. (2006), Fig. 10.19 shows that Saturn’s northern hemisphere was brighter than the southern summer hemi- sphere, with southern hemisphere contrasts muted due to the relatively higher aerosol opacity overlying the contrast- producing clouds. Fletcher et al. (2011) performed a quan- titative analysis of Cassini/VIMS cubes from April 2006 (Ls = 317◦), finding opacity in two regimes: a compact cloud deck centered in the 2.5-2.8 bar region, symmetric between the two hemisphere with small-scale opacity variations re- sponsible for the numerous light/dark axisymmetric lanes; and secondly a hemispherically asymmetric population of aerosols at p < 1.4 bar which was ≈ 2.0× more opaque in the southern summer hemisphere. The upper tropospheric haze asymmetry is shown in Fig. 10.20b, compared to the CIRS-derived 400-mbar temperatures (panel c) and the ob- served contrast in 5-µm brightness temperature (panel a). The vertical structure of this upper-level ‘haze’ could not be Saturn’s seasonally changing atmosphere: thermal structure, composition and aerosols 27 Figure 10.19 Saturn at 5 microns wavelength, showing cloud features in silhouette against the bright background of Saturn’s own thermal emission (Baines et al., 2006; West et al., 2009). In addition to fine-scale banding and numerous dynamic features, note the hemispheric differences in brightness and contrast, caused by a thicker tropospheric haze in the southern hemisphere. Courtesy NASA/JPL-Caltech. constrained by the nightside VIMS observations, but is likely to be the same material responsible for the homogenous haze observed in CCD methane-band imaging. The deep cloud was at higher pressures than the predicted condensation al- titude for NH3 (1.8 bar for a 5× enrichment of heavy ele- ments, Atreya et al., 1999), but at lower pressures than the predicted levels for NH4SH condensation (5.7 bar), so its composition could not be identified unambiguously. Unfor- tunately, there are no studies of the expected reversing of the 5-µm cloud asymmetry available in the literature today. Roman et al. (2013) conducted a quantitative study of Cassini/ISS images of the southern summer hemisphere be- tween 2004-2007 (Ls = 296 − 333◦), reproducing the data with a stratospheric haze merging into a tropospheric haze that sits within the convectively-stable region of the upper troposphere (i.e., above the R-C boundary discussed in Sec- tion 10.2). The tropospheric haze was found to reach the greatest heights (40 ± 20 mbar) at the equator, but to sit deeper (140±20 mbar) at southern mid-latitudes. Fig. 10.20 compares the southern hemisphere optical depths per bar at 619 nm derived by Roman et al. (2013) with the haze opac- ities at 5 µm derived from Fletcher et al. (2011). We have normalized to the maximum haze opacity at the equator to allow intercomparison, highlighting the north-south asym- metry and the comparison with the CIRS-derived tempera- ture structure. The haze location derived by Roman et al. (2013) is consistent with the localized aerosol heating ob- served in Saturn’s thermal field by Fletcher et al. (2007a) (Fig. 10.20d), and Roman et al. (2013) suggest that the tro- pospheric haze is correlated with Saturn’s mid-summer tem- peratures (Fig. 10.20c). At higher pressures, the authors find Figure 10.20 Comparison of the observed 5-µm asymmetry in brightness temperature in southern summer in panel a (2006, Ls = 317◦) with upper tropospheric haze opacities derived from Cassini/ISS for the southern hemisphere (black circles, Roman et al., 2013) and Cassini/VIMS for both hemispheres (solid line with dotted error range, Fletcher et al., 2011). These are found to be largely correlated with the upper tropospheric thermal structure (400 mbar temperatures from Cassini/CIRS) in panel c. Furthermore, the asymmetry correlates with the extent of the temperature ‘knee’ observed by CIRS in the upper troposphere (panel d), and believed to be caused by localized heating in the tropospheric haze layer (Fletcher et al., 2007a). discrete cloud structures in the 1-2 bar range, which may or may not be the same as the 2.5-2.8 bar cloud inferred from VIMS. Quantitative ISS studies have yet to extend into the northern winter hemisphere, and resolving the apparent dis- crepancy in cloud vertical structure with VIMS is a source of ongoing activity. Finally, Sromovsky et al. (2013) analyzed VIMS reflectiv- ity observations in the vicinity of the northern storm, and found that the main haze (of particles 1 µm in radius or 9698100102104Temperature at 400 mbar [K]0.00.20.40.60.81.01.21.4Haze Optical Depth [Normalized]Saturn Clouds in Southern Summer (Ls=317o)170175180185Brightness Temperature [K](a) Cassini/VIMS 5 µm Brightness Temperature (2006)(b) Normalized cloud opacity retrieval from VIMS and ISS(c) Cassini/CIRS 400-mbar temperatures (2006)-90-75-60-45-30-150153045607590Planetographic Latitude30025020015010050Pressure [mbar](d) Extent of Cassini/CIRS temperature knee28 Fletcher, Greathouse, Guerlet, Moses & West smaller) in Saturn’s northern hemisphere in 2011 (away from the storm-perturbed regions) was located between 111-178 mbar (top) and 577-844 mbar (bottom), depending on the latitude, with a deep, compact and opaque cloud near 2.6- 3.2 bar. This is broadly consistent with the 5-µm thermal emission studies. They confirmed that this haze contained no spectroscopically-identifiable features of pure condensates in the VIMS spectral range, and no signatures of hydrazine (N2H4), although diphosphine cannot be definitively ruled out. But, as for ISS, the latitudinal and seasonal dependence of Saturn’s reflectivity has not yet been investigated. In summary, the historical record of Saturn’s aerosol dis- tributions has shown that seasonal insolation changes induce hemispheric asymmetries in the tropospheric (and poten- tially stratospheric) hazes, with higher opacity in the sum- mer hemisphere and low opacity (and blue color) in the winter hemisphere (Fig. 10.20). The haze sits in the region approximately between the tropopause and the radiative- convective boundary, above the main convective region. Cassini images are showing that the asymmetry is reversing, along with the upper tropospheric temperatures, although quantitative studies of VIMS and ISS reflectivity have only been published for single epochs. 10.5 Conclusions and outstanding questions The longevity and broad wavelength coverage of the Cassini mission, coupled with the decades-long record of ground- based observations, have revealed intricate connections be- tween Saturn’s atmospheric temperatures, chemistry and aerosol formation mechanisms. Environmental conditions in the stably-stratified upper troposphere (approximately sit- uated above the radiative-convective boundary at 400-500 mbar) and stratosphere have been observed to vary over time in response to the shifting levels of solar energy depo- sition and the efficiency of radiative cooling to space. At- mospheric temperatures have tracked the seasonal insola- tion changes, albeit with a phase lag that increases with depth into the atmosphere; the upper tropospheric haze has changed in optical thickness, causing differences in the col- oration of reflected sunlight in the summer and winter hemi- spheres; and the zonal mean distributions of both tropo- spheric and stratospheric gaseous constituents exhibit hemi- spheric asymmetries that may be subtly shifting with time. The atmospheric soup of gases and aerosols in turn affects the radiative properties of the atmosphere (i.e., the rates of heating and cooling), which further influences the seasonal temperature shifts that we observe. Atmospheric circulation and localized dynamics can redistribute energy and material from place to place, implying that thermal and chemical per- turbations are superimposed onto the large-scale seasonal asymmetries and sometimes (in the case of equatorial uplift and vertical waves; or within the polar vortices) can domi- nate the observed spatiotemporal trends. Disentangling all of these competing effects is the key chal- lenge for the next generation of modeling activities, towards a complete simulation of the seasonal behavior of Saturn’s cloud-forming weather layer, upper troposphere and strato- sphere. Historically, models have been developed in isolation to explain one subset of the larger problem - for example, radiative climate simulations (with or without convective adjustment and advection of heat via circulation) have been used to understand the magnitude of the seasonal tempera- ture changes; one-dimensional photochemical modeling with parameterized vertical mixing (and an absence of horizon- tal mixing and circulation) demonstrate how stratospheric hydrocarbons vary with time and location; and equilibrium cloud condensation models predict where key condensates should be forming in Saturn’s troposphere. However, each of these models should be intricately linked, as gaseous and aerosol distributions influence the temperatures (and vice versa), which influences chemical and cloud microphysics time scales, as well as dynamic redistribution of heat. The latest generation of numerical simulations are moving in this direction - for example, Friedson and Moses (2012) combined radiative modeling with atmospheric circulation; Hue et al. (2015) connects the photochemically-predicted hydrocarbon distributions with seasonal temperature changes in the con- text of the model of Greathouse et al. (2010); and Spiga et al. (2014) is aiming to incorporate the radiative model of Guerlet et al. (2014) into a full general circulation model. However, some key pieces of the puzzle (the radiative effect of poorly understood and seasonally-variable aerosols; the influence of atmospheric circulation) continue to elude the community and remain the subject of ongoing theoretical development. To date, Cassini has monitored Saturn’s complex atmo- sphere for only a third of a Saturnian year. The upper tro- pospheric and stratospheric temperature fields have been measured from late northern winter through to late north- ern spring, which happens to overlap with Voyager observa- tions just after the previous northern spring equinox, three decades ago. Surprisingly, the atmospheric temperatures measured by Cassini and Voyager at the same point in the seasonal cycle are not identical (Li et al., 2013; Sinclair et al., 2014; Fletcher et al., 2016), suggesting that Saturn might ex- perience non-seasonal variability in its thermal field and cir- culation. Another mystery is why the stratospheric seasonal response at very low pressures (p < 0.1 mbar) appears to be more muted than that at 1 mbar, counter to the expectations of radiative climate models in the absence of stratospheric circulation. Atmospheric temperatures will continue to be monitored in the thermal-infrared using Cassini until north- ern summer solstice, and afterwards with ground-based and space-based observations (e.g., JWST), albeit restricted to the earth-facing summer hemisphere. Long-term consistent datasets are essential to confirm whether Saturn truly does undergo non-seasonal variability. Compared to the study of Saturn’s temperature field, measurements of the distribution of gaseous and aerosol species are less mature. Cassini has identified hemispheric asymmetries in tropospheric species (para-H2, the disequi- librium species PH3 and the condensible volatile NH3, Sec- tion 10.3.2), tropospheric haze opacities and cloud coloration (Section 10.4) and photochemically-produced stratospheric hydrocarbon species (ethane, acetylene, propane, diacety- Saturn’s seasonally changing atmosphere: thermal structure, composition and aerosols 29 lene and methylacetylene, see Section 10.3.3). It remains to be seen whether the timescales for the processes gen- erating these asymmetries match or exceed a Saturnian year. If the timescales are short, then we may expect to see some reversal of the asymmetries as northern summer solstice approaches. If the timescales are long (potentially related to the southern-summer timing of perihelion and the northern-summer timing of aphelion) then we might ex- pect the asymmetries to be quasi-permanent features of Sat- urn’s atmosphere. Temporal studies of these distributions have only been published for the two principle hydrocar- bons (ethane and acetylene) and the tropospheric para-H2 fraction, showing slow and subtle changes to their distri- butions that may be more influenced by atmospheric cir- culation than by seasonally-variable production and loss (e.g., a seasonally-reversing Hadley cell at the equator; inter- hemispheric transport from the autumn to the spring hemi- sphere; and strong subsidence over the north polar region that has recently emerged into spring sunlight). In the near future, new data from Cassini will hopefully determine the magnitude of seasonal shifts in the tropospheric and strato- spheric haze distributions, in addition to the higher-order hydrocarbons and the tropospheric species. By completing Cassini’s observational characterisation of Saturn’s seasonal atmosphere through to northern summer solstice, we hope to inform and guide the development of the next generation of numerical simulations to establish Saturn as the paradigm for seasonal change on a giant planet. Acknowledgements The authors wish to thank A. Laraia, J. Hurley, J. Sinclair, J. Friedson and M. Roman for sharing the results of their studies. The manuscript benefited from thorough reviews by K. Baines, R. Achterberg, F.M. Flasar and G. Bjoraker. Fletcher was supported by a Royal Society Research Fellow- ship at the University of Oxford and University of Leicester. References Abbas, M. M., LeClair, A., Woodard, E., et al. 2013. Distribution of CO2 in Saturn’s Atmosphere from Cassini/cirs Infrared Observations. Astrophys. J., 776(Oct.), 73. Appleby, J. F., and Hogan, J. S. 1984. Radiative-convective equi- Icarus, 59(Sept.), librium models of Jupiter and Saturn. 336–366. Armstrong, E. S., Moses, J. I., Fletcher, L. N., et al. 2014 (Nov.). The chemistry of ethene in the storm beacon region on Sat- urn. Page submitted of: AAS/Division for Planetary Sci- ences Meeting Abstracts. AAS/Division for Planetary Sci- ences Meeting Abstracts, vol. 46. Atreya, S. K., and Wong, A.-S. 2005. Coupled Clouds and Chem- istry of the Giant Planets – A Case for Multiprobes. Space Sci. Rev., 116(Jan.), 121–136. Atreya, S. K., Donahue, T. M., and Kuhn, W. R. 1977. The distribution of ammonia and its photochemical products on Jupiter. Icarus, 31, 348–355. Atreya, S. K., Kuhn, W. R., and Donahue, T. M. 1980. Saturn: Tropospheric ammonia and nitrogen. Geophys. Res. Lett., 7, 474–476. Atreya, S. K., Donahue, T. M., Nagy, A. F., et al. 1984. The- ory, measurements, and models of the upper atmosphere and ionosphere of Saturn. Pages 239–277 of: Gehrels, T., and Matthews, M. S. (eds), Saturn. University of Arizona Press, Tucson. Atreya, S. K., Wong, M. H., Owen, T. C., et al. 1999. A com- parison of the atmospheres of Jupiter and Saturn: deep at- mospheric composition, cloud structure, vertical mixing, and origin. Plan. & Space Sci., 47, 1243–1262. Baines, K. H., Momary, T. W., Roos-Serote, M., et al. 2006. North vs South on Saturn: Discovery of a Pronounced Hemispher- ical Asymmetry in Saturn’s 5-Micron Emission and Associ- ated Deep Cloud Structure by Cassini/VIMS. B.A.A.S., 38, 488. Barnet, C. D., Beebe, R. F., and Conrath, B. J. 1992. A seasonal Icarus, radiative-dynamic model of Saturn’s troposphere. 98, 94–107. Bell, J. M., Westlake, J., and Waite, Jr., J. H. 2011. Simulating the time-dependent response of Titan’s upper atmosphere to periods of magnetospheric forcing. Geophys. Res. Lett., 38, L06202. Bellucci, A., Sicardy, B., Drossart, P., et al. 2009. Titan solar occultation observed by Cassini/VIMS: Gas absorption and constraints on aerosol composition. Icarus, 201, 198–216. B´ezard, B., and Gautier, D. 1985. A seasonal climate model of the atmospheres of the giant planets at the Voyager encounter time. I - Saturn’s stratosphere. Icarus, 61, 296–310. B´ezard, B., Gautier, D., and Conrath, B. 1984. A seasonal model of the Saturnian upper troposphere Comparison with Voy- ager infrared measurements. Icarus, 60, 274–288. B´ezard, B., Drossart, P., Lellouch, E., et al. 1989. Detection of arsine in Saturn. Astrophys. J., 346, 509–513. B´ezard, B., Feuchtgruber, H., Moses, J. I., et al. 1998. Detection of methyl radicals (CH 3) on Saturn. Astron. Astrophys., 334, L41–L44. B´ezard, B., Drossart, P., Encrenaz, T., et al. 2001a. Benzene on the Giant Planets. Icarus, 154(Dec.), 492–500. B´ezard, B., Moses, J. I., Lacy, J., et al. 2001b (Nov.). Detection of Ethylene (C2H4) on Jupiter and Saturn in Non-Auroral Regions. Pages 1079–+ of: Bulletin of the American Astro- nomical Society. B´ezard, B., Lellouch, E., Strobel, D., et al. 2002. Carbon monox- ide on Jupiter: Evidence for both internal and external sources. Icarus, 159, 95–111. Bjoraker, G. L., Larson, H. P., and Fink, U. 1981. A study of ethane on Saturn in the 3 micron region. Astrophys. J., 248, 856–862. Bregman, J. D., Lester, D. F., and Rank, D. M. 1975. Observation of the nu-squared band of PH3 in the atmosphere of Saturn. Astrophys. J., 202. Briggs, F. H., and Sackett, P. D. 1989. Radio observations of Saturn as a probe of its atmosphere and cloud structure. Icarus, 80, 77–103. Burgdorf, M. J., Orton, G. S., Encrenaz, T., et al. 2004. Far- infrared spectroscopy of the giant planets: measurements of ammonia and phosphine at Jupiter and Saturn and the con- tinuum of Neptune. Advances in Space Research, 34, 2247– 2250. Caldwell, J. 1977. The atmosphere of Saturn - an infrared per- spective. Icarus, 30(Mar.), 493–510. 30 Fletcher, Greathouse, Guerlet, Moses & West Caldwell, J., Gillett, F. C., Nolt, I. G., et al. 1978. Spatially resolved infrared observations of Saturn. I - Equatorial limb scans at 20 microns. Icarus, 35(Sept.), 308–312. Carlson, B. E., Caldwell, J., and Cess, R. D. 1980. A model of Saturn’s seasonal stratosphere at the time of the Voyager en- counters. Journal of Atmospheric Sciences, 37(Aug.), 1883– 1885. Carlson, R. W., Baines, K. H., Anderson, M. S., et al. 2012 (Oct.). Chromophores from Photolyzed Ammonia Reacting with Acetylene: Application to Jupiter’s Great Red Spot. Page #205.01 of: AAS/Division for Planetary Sciences Meeting Abstracts. AAS/Division for Planetary Sciences Meeting Ab- stracts, vol. 44. Cavali´e, T., Billebaud, F., Dobrijevic, M., et al. 2009. First obser- vation of CO at 345 GHz in the atmosphere of Saturn with the JCMT: New constraints on its origin. Icarus, 203(Oct.), 531–540. Cavali´e, T., Hartogh, P., Billebaud, F., et al. 2010. A cometary origin for CO in the stratosphere of Saturn? Astron. Astro- phys, 510(Feb.), A88+. Cess, R. D., and Caldwell, J. 1979. A Saturnian stratospheric seasonal climate model. Icarus, 38, 349–357. Conrath, B. J., and Pirraglia, J. A. 1983. Thermal structure of Saturn from Voyager infrared measurements - Implications for atmospheric dynamics. Icarus, 53, 286–292. Conrath, B. J., Gierasch, P. J., and Leroy, S. S. 1990. Tempera- ture and circulation in the stratosphere of the outer planets. Icarus, 83, 255–281. Troposphere of Saturn. Bulletin of the American Astronom- ical Society, 29, 992. Edgington, S. G., Atreya, S. K., Wilson, E. H., et al. 2012. Photo- chemistry in Saturn’s Ring Shadowed Atmosphere: Produc- tion Rates of Key Atmospheric Molecules and Preliminary Analysis of Observations. AGU Fall Meeting Abstracts, Dec., B1946. Encrenaz, T., Owen, T., and Woodman, J. H. 1974. The abun- dance of ammonia on Jupiter, Saturn and Titan. Astron. Astrophys., 37(Dec.), 49–55. Fegley, B., and Prinn, R. G. 1985. Equilibrium and nonequi- librium chemistry of Saturn’s atmosphere - Implications for the observability of PH3, N2, CO, and GeH4. Astrophys. J., 299, 1067–1078. Fegley, Jr., B., and Lodders, K. 1994. Chemical models of the deep atmospheres of Jupiter and Saturn. Icarus, 110(July), 117–154. Ferris, J. P., and Ishikawa, Y. 1988. Formation of HCN and acetylene oligomers by photolysis of ammonia in the pres- ence of acetylene: Applications to the atmospheric chemistry of Jupiter. J. Am. Chem. Soc., 110, 4306–4312. Fink, U., Larson, H. P., Bjoraker, G. L., et al. 1983. The NH3 spectrum in Saturn’s 5 micron window. ApJ, 268(May), 880–888. Flasar, F. M., Kunde, V. G., Abbas, M. M., et al. 2004. Ex- ploring The Saturn System In The Thermal Infrared: The Composite Infrared Spectrometer. Space Science Reviews, 115, 169–297. Conrath, B. J., Gierasch, P. J., and Ustinov, E. A. 1998. Thermal Structure and Para Hydrogen Fraction on the Outer Planets from Voyager IRIS Measurements. Icarus, 135, 501–517. Flasar, F. M., Achterberg, R. K., Conrath, B. J., et al. 2005. Temperatures, Winds, and Composition in the Saturnian System. Science, 307, 1247–1251. Courtin, R., Gautier, D., Marten, A., et al. 1984. The composi- tion of Saturn’s atmosphere at northern temperate latitudes from Voyager IRIS spectra - NH3, PH3, C2H2, C2H6, CH3D, CH4, and the Saturnian D/H isotopic ratio. Astrophys. J., 287, 899–916. Davis, G. R., Griffin, M. J., Naylor, D. A., et al. 1996. ISO LWS measurement of the far-infrared spectrum of Saturn. Astron. Astrophys., 315, L393–L396. de Graauw, T., Feuchtgruber, H., B´ezard, B., et al. 1997. First re- sults of ISO-SWS observations of Saturn: detection of CO2, CH3C2H, C4H2 and tropospheric H2O. Astron. Astrophys., 321, L13–L16. de Pater, I., and Massie, S. T. 1985. Models of the millimeter- centimeter spectra of the giant planets. Icarus, 62, 143–171. Del Genio, A. D., Achterberg, R. K., Baines, K. H., et al. 2009. Saturn Atmospheric Structure and Dynamics, In: Saturn from Cassini-Huygens. Springer. Chap. 6, pages 113–159. Dobrijevic, M., Cavali´e, T., and Billebaud, F. 2011. A methodol- ogy to construct a reduced chemical scheme for 2D-3D pho- tochemical models: Application to Saturn. Icarus, 214, 275– 285. Dobrijevic, M., H´ebrard, E., Loison, J. C., et al. 2014. Coupling of oxygen, nitrogen, and hydrocarbon species in the photo- chemistry of Titan’s atmosphere. Icarus, 228, 324–346. Dowling, T. E., Greathouse, T. K., Sussman, M. G., et al. 2010 (Oct.). New Radiative Transfer Capability in the EPIC At- mospheric Model with Application to Saturn and Uranus. Page 1021 of: AAS/Division for Planetary Sciences Meet- ing Abstracts #42. Bulletin of the American Astronomical Society, vol. 42. Edgington, S. G., Atreya, S. K., Trafton, L. M., et al. 1997. Phos- phine Mixing Ratios and Eddy Mixing Coefficients in the Fletcher, L. N., Irwin, P. G. J., Teanby, N. A., et al. 2007a. Char- acterising Saturn’s Vertical Temperature Structure from Cassini/CIRS. Icarus, 189, 457–478. Fletcher, L. N., Irwin, P. G. J., Teanby, N. A., et al. 2007b. The meridional phosphine distribution in Saturn’s upper tropo- sphere from Cassini/CIRS observations. Icarus, 188(May), 72–88. Fletcher, L. N., Irwin, P. G. J., Orton, G. S., et al. 2008. Tem- perature and Composition of Saturn’s Polar Hot Spots and Hexagon. Science, 319(Jan.), 79–82. Fletcher, L. N., Orton, G. S., Teanby, N. A., et al. 2009. Phos- Icarus, phine on Jupiter and Saturn from Cassini/CIRS. 202(Aug.), 543–564. Fletcher, L. N., Achterberg, R. K., Greathouse, T. K., et al. 2010. Seasonal change on Saturn from Cassini/CIRS observations, 2004-2009. Icarus, 208(July), 337–352. Fletcher, L. N., Baines, K. H., Momary, T. W., et al. 2011. Saturn’s tropospheric composition and clouds from Cassini/VIMS 4.6-5.1 µm nightside spectroscopy. Icarus, 214(Aug.), 510–533. Fletcher, L. N., Swinyard, B., Salji, C., et al. 2012a. Sub- millimetre spectroscopy of Saturn’s trace gases from Her- schel/SPIRE. Astron. Astrophys., 539(Mar.), A44. Fletcher, L. N., Hesman, B. E., Achterberg, R. K., et al. 2012b. The origin and evolution of Saturn’s 2011-2012 stratospheric vortex. Icarus, 221(Nov.), 560–586. Fletcher, L. N., Irwin, P. G. J., Sinclair, J. A., et al. 2015. Seasonal Evolution of Saturn’s Polar Temperatures and Composition. Icarus, 131–153. Fletcher, L. N., Irwin, P. G. J., Achterberg, R. K., et al. 2016. Seasonal Variability of Saturn’s Tropospheric Tem- peratures, Winds and Para-H2 from Cassini Far-IR Spec- troscopy. Icarus, 264, 137–159. Saturn’s seasonally changing atmosphere: thermal structure, composition and aerosols 31 Fouchet, T., Lellouch, E., and Feuchtgruber, H. 2003. The hy- drogen ortho-to-para ratio in the stratospheres of the giant planets. Icarus, 161, 127–143. Guerlet, S., Spiga, A., Sylvestre, M., et al. 2014. Global climate modeling of Saturn’s atmosphere. Part I: Evaluation of the radiative transfer model. Icarus, 238(Aug.), 110–124. Fouchet, T., Guerlet, S., Strobel, DF, et al. 2008. An equa- torial oscillation in Saturn’s middle atmosphere. Nature, 453(7192), 200–202. Fouchet, T., Moses, J. I., and Conrath, B. J. 2009. Saturn: Com- position and Chemistry, In: Saturn from Cassini-Huygens. Chap. 5, pages 83–+. Friedson, A. J., and Moses, J. I. 2012. General circulation and transport in Saturn’s upper troposphere and stratosphere. Icarus, 218(Apr.), 861–875. Friedson, A. J., Wong, A.-S., and Yung, Y. L. 2002. Models for polar haze formation in Jupiter’s stratosphere. Icarus, 158, 389–400. Gans, B., Boy´e-P´eronne, S., Broquier, M., et al. 2011. Photolysis of methane revisited at 121.6 nm and at 118.2 nm: quantum yields of the primary products, measured by mass spectrom- etry. Physical Chemistry Chemical Physics, 13, 8140. Gezari, D. Y., Mumma, M. J., Espenak, F., et al. 1989. New features in Saturn’s atmosphere revealed by high-resolution thermal infrared images. Nature, 342, 777–780. Gillett, F. C., and Forrest, W. J. 1974. The 7.5- to 13.5-MICRON Spectrum of Saturn. Astrophys. J.l, 187(Jan.), L37. Gillett, F. C., and Orton, G. S. 1975. Center-to-limb observations of Saturn in the thermal infrared. Astrophys. J., 195(Jan.), L47–L49. Giver, L. P., and Spinrad, H. 1966. Molecular Hydrogen Features in the Spectra of Saturn and Uranus. Icarus, 5, 586–589. Gladstone, G. R., Allen, M., and Yung, Y. L. 1996. Hydrocarbon photochemistry in the upper atmosphere of Jupiter. Icarus, 119, 1–52. Greathouse, T., Moses, J., Fletcher, L., et al. 2010 (May). Sea- sonal Temperature Variations in Saturn’s Stratosphere: Ra- diative Seasonal Model vs. Observations. Page 4806 of: EGU General Assembly Conference Abstracts. EGU General As- sembly Conference Abstracts, vol. 12. Greathouse, T. K., Lacy, J. H., B´ezard, B., et al. 2005. Meridional variations of temperature, C2H2 and C2H6 abundances in Saturn’s stratosphere at southern summer solstice. Icarus, 177, 18–31. Greathouse, T. K., Lacy, J. H., B´ezard, B., et al. 2006. The first detection of propane on Saturn. Icarus, 181(Mar.), 266–271. Greathouse, T. K., Strong, S., Moses, J., et al. 2008. A Gen- eral Radiative Seasonal Climate Model Applied to Saturn, Uranus, and Neptune. AGU Fall Meeting Abstracts, Dec., P21B–06. Grossman, A. W., Muhleman, D. O., and Berge, G. L. 1989. High- resolution microwave images of Saturn. Science, 245, 1211– 1215. Guerlet, S., Fouchet, T., B´ezard, B., et al. 2009. Vertical and meridional distribution of ethane, acetylene and propane in Saturn’s stratosphere from CIRS/Cassini limb observations. Icarus, 203, 214–232. Guerlet, S., Fouchet, T., B´ezard, B., et al. 2010. Meridional dis- tribution of CH3C2H and C4H2 in Saturn’s stratosphere from CIRS/Cassini limb and nadir observations. Icarus, 209(Oct.), 682–695. Guerlet, S., Fouchet, T., B´ezard, B., et al. 2011. Evolution of the equatorial oscillation in Saturn’s stratosphere between 2005 and 2010 from Cassini/CIRS limb data analysis. Geophysical Research Letters, 38(May), 9201. Guerlet, S., Fouchet, T., Vinatier, S., et al. 2015. Stratospheric benzene and hydrocarbon aerosols detected in Saturn’s au- roral regions. Astronomy and Astrophysics, 580(Aug.), A89. Guillemin, J.-C., Janati, T., and Lassalle, L. 1995. Photolysis of phosphine in the presence of acetylene and propyne, gas mixtures of planetary interest. Advances in Space Research, 16, 85–92. Hanel, R., Conrath, B., Flasar, F. M., et al. 1981. Infrared ob- servations of the Saturnian system from Voyager 1. Science, 212, 192–200. Hanel, R., Conrath, B., Flasar, F. M., et al. 1982. Infrared ob- servations of the Saturnian system from Voyager 2. Science, 215, 544–548. Hanel, R. A., Conrath, B. J., Jennings, D. E., et al. 2003. Ex- ploration of the Solar System by Infrared Remote Sensing: Second Edition. Exploration of the Solar System by In- frared Remote Sensing, by R. A. Hanel and B. J. Con- rath and D. E. Jennings and R. E. Samuelson, ISBN 0521818974. Cambridge, UK: Cambridge University Press, April 2003. Hartogh, P., Lellouch, E., Moreno, R., et al. 2011. Direct detec- tion of the Enceladus water torus with Herschel. Astron. Astrophys, 532(Aug.), L2+. H´ebrard, E., Dobrijevic, M., Loison, J. C., et al. 2013. Photochem- istry of C3Hp hydrocarbons in Titan’s stratosphere revisited. Astron. Astrophys., 552, A132. Hesman, B. E., Jennings, D. E., Sada, P. V., et al. 2009. Saturn’s latitudinal C2H2 and C2H6 abundance profiles from Cassini/CIRS and ground-based observations. Icarus, 202(July), 249–259. Hesman, B. E., Bjoraker, G. L., Sada, P. V., et al. 2012. Elu- sive Ethylene Detected in Saturn’s Northern Storm Region. Astrophys. J., 760(Nov.), 24. Holton, J.R. 2004. An Introduction to Dynamic Meteorology. Academic press. Howett, C. J. A., Irwin, P. G. J., Teanby, N.A., et al. 2007. Meridional Variations in Stratospheric Acetylene and Ethane in the Saturnian Atmosphere as Determined from Cassini/CIRS Measurements. Icarus, 190(2), 556–572. Hue, V., Cavali´e, T., Dobrijevic, M., et al. 2015. 2D photochemi- cal modeling of Saturn’s stratosphere. Part I: Seasonal vari- ation of atmospheric composition without meridional trans- port. Icarus, 257(Sept.), 163–184. Hurley, J., Fletcher, L. N., Irwin, P. G. J., et al. 2012. Lati- tudinal variation of upper tropospheric NH3 on Saturn de- rived from Cassini/CIRS far-infrared measurements. Plan. & Space Sci., 73(Dec.), 347–363. Ingersoll, A. P., Beebe, R. F., Conrath, B. J., et al. 1984. Structure and dynamics of Saturn’s atmosphere. Saturn. Pages 195– 238. Janssen, M. A., Ingersoll, A. P., Allison, M. D., et al. 2013. Sat- urn’s thermal emission at 2.2-cm wavelength as imaged by the Cassini RADAR radiometer. Icarus, 226(Sept.), 522– 535. Kalogerakis, K. S., Marschall, J., Oza, A. U., et al. 2008. The coat- ing hypothesis for ammonia ice particles in Jupiter: Labora- tory experiments and optical modeling. Icarus, 196(July), 202–215. Karkoschka, E., and Tomasko, M. 2005. Saturn’s vertical and latitudinal cloud structure 1991 - 2004 from HST imaging in 30 filters. Icarus, 179, 195–221. 32 Fletcher, Greathouse, Guerlet, Moses & West Karkoschka, E., and Tomasko, M. G. 1992. Saturn’s upper tro- posphere 1986-1989. Icarus, 97, 161–181. Karkoschka, E., and Tomasko, M. G. 1993. Saturn’s upper at- mospheric hazes observed by the Hubble Space Telescope. Icarus, 106, 428–441. Kaye, J. A., and Strobel, D. F. 1983a. Formation and photo- chemistry of methylamine in Jupiter’s atmosphere. Icarus, 55, 399–419. Kaye, J. A., and Strobel, D. F. 1983b. HCN formation on Jupiter: The coupled photochemistry of ammonia and acety- lene. Icarus, 54, 417–433. Kaye, J. A., and Strobel, D. F. 1983c. Phosphine photochemistry in Saturn’s atmosphere. Geophys. Res. Lett., 10, 957–960. Kaye, J. A., and Strobel, D. F. 1984. Phosphine photochemistry in the atmosphere of Saturn. Icarus, 59(Sept.), 314–335. Keane, T. C., Yuan, F., and Ferris, J. P. 1996. Potential Jupiter atmospheric constituents: Candidates for the mass spec- trometer in the Galileo atmospheric probe. Icarus, 122, 205–207. Kerola, D. X., Larson, H. P., and Tomasko, M. G. 1997. Analysis of the Near-IR Spectrum of Saturn: A Comprehensive Radia- tive Transfer Model of Its Middle and Upper Troposphere. Icarus, 127, 190–212. Kim, J. H., Kim, S. J., Geballe, T. R., et al. 2006. High-resolution spectroscopy of Saturn at 3 microns: CH4, CH3D, C2H2, C2H6, PH3, clouds, and haze. Icarus, 185(Dec.), 476–486. Kim, S. J., and Geballe, T. R. 2005. The 2.9-4.2 micron spectrum of Saturn: Clouds and CH4, PH3, and NH3. Icarus, 179, 449–458. Kim, S. J., Sim, C. K., Lee, D. W., et al. 2012. The three-micron spectral feature of the Saturnian haze: Implications for the haze composition and formation process. Planet. Space Sci., 65, 122–129. Krasnopolsky, V. A. 2014. Chemical composition of Titan’s atmo- sphere and ionosphere: Observations and the photochemical model. Icarus, 236, 83–91. Lane, A. L., Hord, C. W., West, R. A., et al. 1982. Photopo- larimetry from Voyager 2 - Preliminary results on Saturn, Titan, and the rings. Science, 215(Jan.), 537–543. Lara, L. M., Lellouch, E., Gonz´alez, M., et al. 2014. A time- dependent photochemical model for Titan’s atmosphere and the origin of H2O. Astron. Astrophys., 566, A143. Laraia, A. L., Ingersoll, A. P., Janssen, M. A., et al. 2013. Analysis of Saturn’s thermal emission at 2.2-cm wavelength: Spatial distribution of ammonia vapor. Icarus, 226(Sept.), 641–654. Infrared spectroscopic observations of the outer planets, their satellites, and the asteroids. Annual Re- view of Astronomy and Astrophysics, 18, 43–75. Larson, H. P. 1980. Lavvas, P., Galand, M., Yelle, R. V., et al. 2011. Energy de- position and primary chemical products in Titan’s upper atmosphere. Icarus, 213, 233–251. Lebonnois, S. 2005. Benzene and aerosol production in Titan and Jupiter’s atmospheres: a sensitivity study. Planet. Space Sci., 53, 486–497. Lellouch, E., B´ezard, B., Fouchet, T., et al. 2001. The deuterium abundance in Jupiter and Saturn from ISO-SWS observa- tions. Astron. Astrophys., 370, 610–622. Lewis, J. S., and Fegley, Jr., M. B. 1984. Vertical distribution of disequilibrium species in Jupiter’s troposphere. Space Sci. Rev., 39(Oct.), 163–192. Lewis, J. S., and Prinn, R. G. 1984. Planets and Their Atmo- spheres: Origin and Evolution. Academic, Orlando. Li, L., Conrath, B. J., Gierasch, P. J., et al. 2010. Saturn’s emitted power. Journal of Geophysical Research (Planets), 115(Nov.), 11002. Li, L., Jiang, X., Ingersoll, A. P., et al. 2011. Equatorial winds on Saturn and the stratospheric oscillation. Nature Geoscience, 4(Nov.), 750–752. Li, L., Achterberg, R. K., Conrath, B. J., et al. 2013. Strong Temporal Variation Over One Saturnian Year: From Voyager to Cassini. Scientific Reports, 3(Aug.). Lindal, G. F., Sweetnam, D. N., and Eshleman, V. R. 1985. The atmosphere of Saturn - an analysis of the Voyager radio oc- cultation measurements. Astron. J., 90, 1136–1146. Lodders, K., and Fegley, B. 2002. Atmospheric Chemistry in Giant Planets, Brown Dwarfs, and Low-Mass Dwarf Stars. I. Carbon, Nitrogen, and Oxygen. Icarus, 155(Feb.), 393– 424. Loison, J. C., H´ebrard, E., Dobrijevic, M., et al. 2015. The neutral photochemistry of nitriles, amines and imines in the atmo- sphere of Titan. Icarus, 247(Feb.), 218–247. Mandt, K. E., Gell, D. A., Perry, M., et al. 2012. Ion densities and composition of Titan’s upper atmosphere derived from the Cassini Ion Neutral Mass Spectrometer: Analysis methods and comparison of measured ion densities to photochemical model simulations. J. Geophys. Res., 117, E10006. Massie, S. T., and Hunten, D. M. 1982. Conversion of para and ortho hydrogen in the Jovian planets. Icarus, 49, 213–226. Moos, H. W., and Clarke, J. T. 1979. Detection of acetylene in the Saturnian atmosphere, using the IUE satellite. Astrophys. J., 229, L107. Moreno, R., Lellouch, E., Lara, L. M., et al. 2012. The abun- dance, vertical distribution and origin of H2O in Titan’s atmosphere: Herschel observations and photochemical mod- elling. Icarus, 221, 753–767. Moses, J. I., and Greathouse, T. K. 2005. Latitudinal and seasonal models of stratospheric photochemistry on Saturn: Compar- ison with infrared data from IRTF/TEXES. Journal of Geo- physical Research (Planets), 110(Sept.), E09007. Moses, J. I., B´ezard, B., Lellouch, E., et al. 2000a. Photochem- istry of Saturn’s Atmosphere. I. Hydrocarbon Chemistry and Comparisons with ISO Observations. Icarus, 143, 244–298. Moses, J. I., Lellouch, E., B´ezard, B., et al. 2000b. Photochem- istry of Saturn’s Atmosphere. II. Effects of an Influx of Ex- ternal Oxygen. Icarus, 145(May), 166–202. Moses, J. I., Fouchet, T., B´ezard, B., et al. 2005. Photochemistry and diffusion in Jupiter’s stratosphere: Constraints from ISO observations and comparisons with other giant plan- ets. Journal of Geophysical Research (Planets), 110(Aug.), E08001. Moses, J. I., Liang, M.-C., Yung, Y. L., et al. 2007 (Mar.). Two-Dimensional Photochemical Modeling of Hydrocarbon Abundances on Saturn. Page 2196 of: Lunar and Planetary Science Conference. Lunar and Planetary Science Confer- ence, vol. 38. Moses, J. I., Visscher, C., Keane, T. C., et al. 2010. On the abundance of non-cometary HCN on Jupiter. Faraday Dis- cussions, 147, 103–136. Moses, J. I., Visscher, C., Fortney, J. J., et al. 2011. Disequilib- rium Carbon, Oxygen, and Nitrogen Chemistry in the At- mospheres of HD 189733b and HD 209458b. Astrophys. J., 737(Aug.), 15. Moses, J. I., Armstrong, E. S., Fletcher, L. N., et al. 2014 (Nov.). Evolution of stratospheric chemistry in the Saturn storm beacon. Page submitted of: AAS/Division for Planetary Saturn’s seasonally changing atmosphere: thermal structure, composition and aerosols 33 Sciences Meeting Abstracts. AAS/Division for Planetary Sciences Meeting Abstracts, vol. 46. Nava, D. F., Payne, W. A., Marston, G., et al. 1993. The re- action of atomic hydrogen with germane: Temperature de- pendence of the rate constant and implications for germane photochemistry in the atmospheres of Jupiter and Saturn. J. Geophys. Res., 98, 5531–5537. Noll, K. S., and Larson, H. P. 1990. The spectrum of Saturn from 1990–2230 cm−1: abundances of AsH3, CH3D, CO, GeH4, and PH3. Icarus, 89, 168–189. Noll, K. S., and Larson, H. P. 1991. The spectrum of Saturn from 1990 to 2230 cm−1 - Abundances of AsH3, CH3D, CO, GeH4, NH3, and PH3. Icarus, 89(Jan.), 168–189. Noll, K. S., Knacke, R. F., Geballe, T. R., et al. 1986. Detection of carbon monoxide in Saturn. ApJ Letters, 309(Oct.), L91– L94. Noll, K. S., Knacke, R. F., Geballe, T. R., et al. 1988. Evidence for germane in Saturn. Icarus, 75, 409–422. Noll, K. S., Geballe, T. R., and Knacke, R. F. 1989. Arsine in Saturn and Jupiter. Astrophys. J., 338, L71–L74. O’Donoghue, J., Stallard, T. S., Melin, H., et al. 2013. The dom- ination of Saturn’s low-latitude ionosphere by ring ‘rain’. Nature, 496(Apr.), 193–195. Ollivier, J. L., Billebaud, F., Drossart, P., et al. 2000. Seasonal effects in the thermal structure of Saturn’s stratosphere from infrared imaging at 10 microns. Astron. Astrophys., 356, 347–356. Orton, G. S., and Ingersoll, A. P. 1980. Saturn’s atmospheric tem- perature structure and heat budget. Journal of Geophysical Research, 85(Nov.), 5871–5881. Orton, G. S., and Yanamandra-Fisher, P. A. 2005. Saturn’s Tem- perature Field from High-Resolution Middle-Infrared Imag- ing. Science, 307, 696–698. Orton, G. S., Serabyn, E., and Lee, Y. T. 2000. Vertical Distri- bution of PH3 in Saturn from Observations of Its 1-0 and 3-2 Rotational Lines. Icarus, 146, 48–59. Orton, G. S., Serabyn, E., and Lee, Y. T. 2001. Erratum, Volume 146, Number 1, pages 48-59 (2000), in the article “Vertical Distribution of PH3 in Saturn from Observations of Its 1-0 and 3-2 Rotational Lines,”. Icarus, 149, 489–490. Orton, G. S., Baines, K. H., Cruikshank, D., et al. 2009. Review of Knowledge Prior to the Cassini-Huygens Mission and Concurrent Research, In: Saturn from Cassini-Huygens. Springer. Chap. 2, pages 9–54. Orton, G. S., Moses, J. I., Fletcher, L. N., et al. 2014. Mid- infrared spectroscopy of Uranus from the Spitzer infrared spectrometer: 2. Determination of the mean composition of the upper troposphere and stratosphere. Icarus, 243(Nov.), 471–493. Orton, G.S., Yanamandra-Fisher, P.A., Fisher, B.M., et al. 2008. Semi-annual oscillations in Saturn’s low-latitude strato- spheric temperatures. Nature, 453(May), 196–198. Palotai, C., Dowling, T. E., and Fletcher, L. N. 2014. 3D Modeling of interactions between Jupiter’s ammonia clouds and large anticyclones. Icarus, 232, 141–156. P´erez-Hoyos, S., S´anchez-Lavega, A., French, R. G., et al. 2005. Saturn’s cloud structure and temporal evolution from ten years of Hubble Space Telescope images (1994-2003). Icarus, 176, 155–174. P´erez-Hoyos, S., S´anchez-Lavega, A., and French, R. G. 2006. Short-term changes in the belt/zone structure of Saturn’s Southern Hemisphere (1996-2004). Astron. Astrophys., 460(Dec.), 641–645. Plessis, S., Carrasco, N., Dobrijevic, M., et al. 2012. Production of neutral species in Titan’s ionosphere through dissociative recombination of ions. Icarus, 219, 254–266. Prang´e, R., Fouchet, T., Courtin, R., et al. 2006. Latitudinal variation of Saturn photochemistry deduced from spatially- resolved ultraviolet spectra. Icarus, 180, 379–392. Prinn, R. G., and Barshay, S. S. 1977. Carbon monoxide on Jupiter and implications for atmospheric convection. Sci- ence, 198, 1031–1034. Prinn, R. G., and Fegley, Jr., B. 1981. Kinetic inhibition of CO and N2 reduction in circumplanetary nebulae - Implications for satellite composition. Astrophys. J, 249(Oct.), 308–317. Prinn, R. G., and Owen, T. 1976. Chemistry and spectroscopy of the Jovian atmosphere. Pages 319–371 of: Gehrels, T. (ed), Jupiter. University of Arizona Press, Tucson. Prinn, R. G., Larson, H. P., Caldwell, J. J., et al. 1984. Compo- sition and chemistry of Saturn’s atmosphere. Saturn. Pages 88–149. Pryor, W. R., and Hord, C. W. 1991. A study of photopolarimeter system UV absorption data on Jupiter, Saturn, Uranus, and Neptune - Implications for auroral haze formation. Icarus, 91, 161–172. Rages, K. A., and Barth, E. L. 2012 (Oct.). Saturn Limb Hazes as Seen from Cassini. Page #500.05 of: AAS/Division for Plan- etary Sciences Meeting Abstracts. AAS/Division for Plane- tary Sciences Meeting Abstracts, vol. 44. Rieke, G. H. 1975. The thermal radiation of Saturn and its rings. Icarus, 26(Sept.), 37–44. Roman, M. T., Banfield, D., and Gierasch, P. J. 2013. Saturn’s cloud structure inferred from Cassini ISS. Icarus, 225(July), 93–110. Sada, P. V., Bjoraker, G. L., Jennings, D. E., et al. 2005. Obser- vations of C 2H 6 and C 2H 2 in the stratosphere of Saturn. Icarus, 173, 499–507. Schinder, P. J., Flasar, F. M., Marouf, E. A., et al. 2011. Sat- urn’s equatorial oscillation: Evidence of descending thermal structure from Cassini radio occultations. Geophys. Res. Lett., 38(Apr.), 8205. Sinclair, J. A., Irwin, P. G. J., Fletcher, L. N., et al. 2013. Seasonal variations of temperature, acetylene and ethane in Saturn’s atmosphere from 2005 to 2010, as observed by Cassini-CIRS. Icarus, 225(July), 257–271. Sinclair, J. A., Irwin, P. G. J., Fletcher, L. N., et al. 2014. From Voyager-IRIS to Cassini-CIRS: Interannual variability in Saturn’s stratosphere? Icarus, 233(May), 281–292. Spiga, A., Guerlet, S., Indurain, M., et al. 2014 (Nov.). An explo- ration of Saturn’s stratospheric dynamics through Global Climate Modeling. Page #508.09 of: AAS/Division for Planetary Sciences Meeting Abstracts. AAS/Division for Planetary Sciences Meeting Abstracts, vol. 46. Sromovsky, L. A., Baines, K. H., and Fry, P. M. 2013. Saturn’s Great Storm of 2010-2011: Evidence for ammonia and water ices from analysis of VIMS spectra. Icarus, 226(Sept.), 402– 418. Stam, D. M., Banfield, D., Gierasch, P. J., et al. 2001. Near- IR Spectrophotometry of Saturnian Aerosols-Meridional and Vertical Distribution. Icarus, 152, 407–422. Strobel, D. F. 1975. Aeronomy of the major planets: Photochem- istry of ammonia and hydrocarbons. Reviews of Geophysics and Space Physics, 13, 372–382. Strobel, D. F. 1977. NH3 and PH3 photochemistry in the Jovian atmosphere. Astrophys. J. Lett., 214, L97–L99. posphere. Page 1201 of: Lunar and Planetary Science Con- ference. Lunar and Planetary Science Conference, vol. 40. Visscher, C., Moses, J. I., and Saslow, S. A. 2010. The deep water abundance on Jupiter: New constraints from thermochemical kinetics and diffusion modeling. Icarus, 209, 602–615. Vuitton, V., Yelle, R. V., and Lavvas, P. 2009. Composition and chemistry of Titan’s thermosphere and ionosphere. Royal Society of London Philosophical Transactions Series A, 367, 729–741. Vuitton, V., Yelle, R. V., Lavvas, P., et al. 2012. Rapid associ- ation reactions at low pressure: Impact on the formation of hydrocarbons on Titan. Astrophys. J., 744, 11. Weidenschilling, S. J., and Lewis, J. S. 1973. Atmospheric and cloud structures of the jovian planets. Icarus, 20, 465–476. Weisstein, E. W., and Serabyn, E. 1994. Detection of the 267 GHz J = 1-0 rotational transition of PH3 in Saturn with a new fourier transfer spectrometer. Icarus, 109, 367–381. Weisstein, E. W., and Serabyn, E. 1996. Submillimeter Line Search in Jupiter and Saturn. Icarus, 123, 23–36. West, R. A., and Smith, P. H. 1991. Evidence for aggregate particles in the atmospheres of Titan and Jupiter. Icarus, 90(Apr.), 330–333. West, R. A., Tomasko, M. G., Smith, B. A., et al. 1982. Spatially resolved methane band photometry of Saturn. I - Absolute reflectivity and center-to-limb variations in the 6190-, 7250-, and 8900-A bands. Icarus, 51, 51–64. West, R. A., Strobel, D. F., and Tomasko, M. G. 1986. Clouds, aerosols, and photochemistry in the Jovian atmosphere. Icarus, 65, 161–217. West, R. A., Baines, K. H., Karkoschka, E., et al. 2009. Clouds and Aerosols in Saturn’s Atmosphere, In: Saturn from Cassini-Huygens. Springer. Chap. 7, pages 161–179. Westlake, J. H., Waite, Jr., J. H., Mandt, K. E., et al. 2012. Ti- tan’s ionospheric composition and structure: Photochemical modeling of Cassini INMS data. J. Geophys. Res., 117, E01003. Winkelstein, P., Caldwell, J., Kim, S. J., et al. 1983. A deter- mination of the composition of the Saturnian stratosphere using the IUE. Icarus, 54, 309–318. Wong, A.-S., Lee, A. Y. T., Yung, Y. L., et al. 2000. Jupiter: Aerosol chemistry in the polar atmosphere. Astrophys. J. Lett., 534, L215–L217. Wong, A.-S., Yung, Y. L., and Friedson, A. J. 2003. Benzene and haze formation in the polar atmosphere of Jupiter. Geophys. Res. Lett., 30, 1447. Yung, Y. L., and DeMore, W. B. 1999. Photochemistry of Plan- etary Atmospheres. Oxford University Press. Yung, Y. L., Drew, W. A., Pinto, J. P., et al. 1988. Estimation of the reaction rate for the formation of CH3O from H + H2CO: Implications for chemistry in the solar system. Icarus, 73, 516–526. 34 Fletcher, Greathouse, Guerlet, Moses & West Strobel, D. F. 1983. Photochemistry of the reducing atmospheres International Reviews in of Jupiter, Saturn, and Titan. Physical Chemistry, 3, 145–176. Strobel, D. F. 2005. Photochemistry in Outer Solar System At- mospheres. Space Science Reviews, 116, 155–170. Sugiyama, K., Nakajima, K., Odaka, M., et al. 2011. Intermit- tent cumulonimbus activity breaking the three-layer cloud structure of Jupiter. Geophys. Res. Lett., 38, L13201. Sugiyama, K., Nakajima, K., Odaka, M., et al. 2014. Numer- ical simulations of Jupiter’s moist convection layer: Struc- ture and dynamics in statistically steady states. Icarus, 229(Feb.), 71–91. Sylvestre, M., Guerlet, S., Fouchet, T., et al. 2015. Seasonal changes in Saturn’s stratosphere inferred from Cassini/CIRS limb observations. Icarus, 258(Sept.), 224–238. Tautermann, C. S., Wellenzohn, B., and Clary, D. C. 2006. Rates of the reaction C2H3 + H2 → C2H4 + H. Molecular Physics, 104, 151–158. Temma, T., Chanover, N. J., Simon-Miller, A. A., et al. 2005. Ver- tical structure modeling of Saturn’s equatorial region using high spectral resolution imaging. Icarus, 175, 464–489. Tokunaga, A., and Cess, R. D. 1977. A model for the temperature inversion within the atmosphere of Saturn. Icarus, 32(Nov.), 321–327. Tokunaga, A., Knacke, R. F., and Owen, T. 1975. The detection of ethane on Saturn. Astrophys. J., 197, L77. Tokunaga, A. T., Caldwell, J., Gillett, F. C., et al. 1978. Spa- tially resolved infrared observations of Saturn. II - The tem- perature enhancement at the South Pole of Saturn. Icarus, 36(Nov.), 216–222. Tokunaga, A. T., Dinerstein, H. L., Lester, D. F., et al. 1980. The phosphine abundance on Saturn derived from new 10- micrometer spectra. Icarus, 42, 79–85. Tomasko, M. G., and Doose, L. R. 1984. Polarimetry and photom- etry of Saturn from Pioneer 11 Observations and constraints on the distribution and properties of cloud and aerosol par- ticles. Icarus, 58, 1–34. Tseng, W.-L., and Ip, W.-H. 2011. An assessment and test of Enceladus as an important source of Saturn’s ring atmo- sphere and ionosphere. Icarus, 212(Mar.), 294–299. van der Tak, F., de Pater, I., Silva, A., et al. 1999. Time Variabil- ity in the Radio Brightness Distribution of Saturn. Icarus, 142(Nov.), 125–147. Vander Auwera, J., Moazzen-Ahmadi, N., and Flaud, J.-M. 2007. Toward an Accurate Database for the 12 µm Region of the Ethane Spectrum. Astrophysical Journal, 662(June), 750– 757. Vasavada, A. R., Horst, S. M., Kennedy, M. R., et al. 2006. Cassini imaging of Saturn: Southern hemisphere winds and vortices. Journal of Geophysical Research (Planets), 111(E10), 5004. Visscher, C., and Fegley, B. J. 2005. Chemical Constraints on the Water and Total Oxygen Abundances in the Deep At- mosphere of Saturn. Astrophys. J., 623, 1221–1227. Visscher, C., and Moses, J. I. 2011. Quenching of carbon monox- ide and methane in the atmospheres of cool brown dwarfs and hot Jupiters. Astrophys. J., 738, 72. Visscher, C., Lodders, K., and Fegley, Jr., B. 2006. Atmospheric Chemistry in Giant Planets, Brown Dwarfs, and Low-Mass Dwarf Stars. II. Sulfur and Phosphorus. Astrophys. J. Let- ters, 648(Sept.), 1181–1195. Visscher, C., Sperier, A. D., Moses, J. I., et al. 2009 (Mar.). Phosphine and ammonia photochemistry in Jupiter’s tro-
1506.07067	1	1506	2015-06-23T16:02:31	The mass of the Mars-sized exoplanet Kepler-138 b from transit timing	[ "astro-ph.EP" ]	Extrasolar planets that pass in front of their host star (transit) cause a temporary decrease in the apparent brightness of the star once per orbit, providing a direct measure of the planet's size and orbital period. In some systems with multiple transiting planets, the times of the transits are measurably affected by the gravitational interactions between neighbouring planets. In favorable cases, the departures from Keplerian orbits implied by the observed transit times permit planetary masses to be measured, which is key to determining bulk densities. Characterizing rocky planets is particularly difficult, since they are generally smaller and less massive than gaseous planets. Thus, few exoplanets near Earth's size have had their masses measured. Here we report the sizes and masses of three planets orbiting Kepler-138, a star much fainter and cooler than the Sun. We measure the mass of the Mars-sized inner planet based on on the transit times of its neighbour and thereby provide the first density measurement for an exoplanet smaller than Earth. The middle and outer planets are both slightly larger than Earth. The middle planet's density is similar to that of Earth, while the outer planet is less than half as dense, implying that it contains a greater portion of low density components such as H2O and/or H2.	astro-ph.EP	astro-ph	Mass of the Mars-sized Exoplanet Kepler-138 b from Transit Timing Daniel Jontof-Hutter1,2, Jason F. Rowe2,3, Jack J. Lissauer2, Daniel C. Fabrycky4,5, and Eric B. Ford1 Extrasolar planets that pass in front of their host star (transit) cause a temporary decrease in the apparent brightness of the star once per orbit, providing a direct measure of the planet's size and orbital period. In some systems with multiple transiting planets, the times of the transits are measurably affected by the gravitational interactions between neighbouring planets1,2. In favorable cases, the departures from Keplerian orbits implied by the observed transit times permit planetary masses to be measured, which is key to determining bulk densities3. Characterizing rocky planets is particularly difficult, since they are generally smaller and less massive than gaseous planets. Thus, few exoplanets near Earth's size have had their masses measured. Here we report the sizes and masses of three planets orbiting Kepler-138, a star much fainter and cooler than the Sun. We measure the mass of the Mars-sized inner planet based on on the transit times of its neighbour and thereby provide the first density measurement for an exoplanet smaller than Earth. The middle and outer planets are both slightly larger than Earth. The middle planet's density is similar to that of Earth, while the outer planet is less than half as dense, implying that it contains a greater portion of low density components such as H2O and/or H2. NASA's Kepler mission has discovered thousands of candidate transiting exoplanets, with a wide range of planetary sizes4,5,6. A small fraction of these planets have had their masses characterized, by either radial velocity spectroscopy (RV) or via transit timing. The latter probes the gravitational perturbations between planets in multi-planet systems by precisely measuring transit times and fitting dynamical models to the observed transit timing variations (TTV)1,2. Both RV and TTV signals are larger for more massive planets, improve with greater planetary masses, although the two techniques sample different populations of exoplanets. The RV technique measures the motion of a host star induced by its planet's gravity, and hence the signal declines with increasing orbital distance. The majority of planets with mass determinations via RV from Kepler's dataset have orbital periods shorter than one week. For Kepler-discovered planets characterized as rocky by this method, the orbital periods are all shortest orbital period- 0.35 days7,8. Characterizing planets by transit timing is quite complementary to RV because transit timing is very sensitive to perturbations between planets that are closely spaced or near orbital resonances3,9. Note that most systems with detected TTVs are not in resonance, but rather near enough to resonance for the perturbations to be coherent for many orbital periods, while also far enough from resonance that the planetary less than one day. Of the RV detections, Kepler-78 b has the lowest mass, (1.7 MEarth), and the """""""""""""""""""""""""""""""""""""""""""""""""""""""" 1"Department"of"Astronomy,"Pennsylvania"State"University,"Davey"Lab,"University"Park," PA"16802,"USA;" 2"NASA"Ames"Research"Center,"Moffett"Field,"CA"94035,"USA;"" 3"SETI"Institute, 189 N Bernardo Ave, Mountain View, CA 94043, USA; 5Alfred P. Sloan Fellow" 4 Department of Astronomy and Astrophysics, University of Chicago, 5640 South Ellis Avenue, Chicago, IL 60637, USA conjunctions cycle around the orbit plane within the four-year Kepler baseline. For near- resonant pairs, the TTVs of neighbouring planets frequently take the form of anti-correlated sinusoids over many orbits10,11,12,13,14 or the sum of sinusoids where a planet is perturbed by two neighbours15. The majority of TTV detections have been found near first-order mean-motion resonances, such that planet pairs have an orbital period ratio near j:j-1 where j is an integer. Near first-order resonances generally cause stronger TTV signals than second-order resonances, although much depends on how close the planet pairs are to resonance and how eccentric (non-circular) their orbits are, as well as the planets' masses. Eccentricity causes the orbital speed to vary during the orbit, and the distance between the planets at conjunction to vary with the position of the conjunction. The bulk of planets with mass characterizations from the Kepler sample using TTVs so far have orbital periods ranging from ~10-100 days. These are generally low density15,16,17,18 and likely possess deep atmospheres, with the exception of the rocky planet Kepler-36 b19. Kepler-138 (formerly known as KOI-314) hosts three validated transiting planets20. The orbital periods of Kepler-138's three planets are given in Table 1. Kepler-138 c and d orbit near a second order mean motion resonance (5:3), whilst 'b' and 'c' orbit near the 4:3 first order resonance. Using transit times up to the fourteenth quarter of the Kepler mission, two of the three known planets orbiting Kepler-13820, have been confirmed and characterized with TTVs21. The derived parameters in that work suggested that the outer planet, Kepler-138 d, has a density so low that it must have a significant hydrogen/helium gaseous envelope21. Using the complete Kepler dataset for Kepler-138, we have detected TTVs for all three planets. We describe our procedure to measure transit times in Methods, and list our transit time measurements in Extended Data Table 1. TTVs are expressed as the difference between the observed transit times and the calculated linear fit to the transit times. Modeling Kepler-138 as a three-planet system, we measure the masses of all three planets, the the size of Mars. We performed dynamical fits by calculating the orbits of the three planets around the star, modeling the orbits as co-planar, since all three planets are transiting, Kepler’s multi-planet systems are known to have small mutual inclinations, and we demonstrate that allowing mutual inclinations has little effect on our results for the planet masses (see Extended Data). Our model parameters for each planet are the orbital period P, the time T0 of the first transit after our chosen epoch, the components of the eccentricity vector (ecosω and esinω, where ω is the angle between the sky-plane and the orbital pericenter of the planet), and Mp/Mstar, the mass of the planet relative to the host star, which we express as (Mp/MEarth)(MSun/Mstar) throughout. We perform Bayesian parameter estimation for these 15 model parameters using Differential Evolution Markov Chain Monte-Carlo. super-Earth sized Kepler-138 c and d, as well as Kepler-138 b, which at 0.52 REarth is roughly We report the properties of the star and planets in Table 1, and details of the parameter estimation algorithm, priors and statistical models in Methods. -0.037MEarth, where uncertainties denote We measure the mass of Kepler-138 b to be 0.066+0.059 the 68.3% confidence limits. The 95.4% interval spans 0.011—0.170 MEarth. The robustness of this result against outlying transit times and mutual inclinations is demonstrated in Methods. The posterior probability for the inner planet having non-zero mass is between 99.82% and 99.91% (depending on the choice of prior for eccentricity), i.e., equivalent to a “3-sigma” detection. This calculation is based on the Savage-Dickey density ratio for calculating the Bayes factor, which fully accounts for posterior width and shape, including asymmetries and non-Gaussianity as described in Methods. Kepler-138 b is by far the smallest exoplanet, both by radius and mass, to have a density measurement. Thus it opens a new regime to physical study. It is likely to become the prototype for a class of small close-in planets that could be common. The prospect of further constraints on this planet are excellent: NASA’s Transiting Exoplanet Survey Satellite should be able to measure transit times for the two largest planets, improving constraints on the dynamical model for all 3 planets. ESA’s Plato mission will continue this process and ground- based measurements are also possible. These future observations, plus more accurate stellar classification using the distance to the star measured by the Gaia mission, will further improve the characterization of this system, especially the inner planet. Our measurements of the mass and density of the small inner planet Kepler-138 b are consistent with various compositions and formation locations. If future observations imply that the planet is less dense than rock, then the only physically and cosmogonically plausible low-density constituents are H2O and/or other astrophysical ices, which could only have condensed far from the star. This would be the first definitive evidence for substantial inward orbital migration of a small planet. For the two outer planets, Kepler-138 c and d, we find a lower mass ratio between these two planets than previous work21. This is not surprising, since planet 'b's perturbations explain part of the TTVs observed in planet 'c', which were previously attributed solely to perturbations by planet 'd'. Nevertheless, the mass ratios between each of these planets and their host star remains consistent with published results21. We find higher densities for both of these planets than previous work, due to our improved stellar properties, particularly the higher stellar density and consequent smaller stellar radius. Previous estimates of the size and mass of the outer planet Kepler-138 d implied that the planet possessed a hydrogen-rich atmosphere21, which is difficult to explain with our current understanding of the accretion and retention of light gases from low-mass planets orbiting close to their star28,29,30. Our new measurements could be explained by a composition of rock and H2O. A rock and H2O planet would be more stable against mass loss, and would imply that the planet formed at a greater distance from the star and migrated. 1. Agol, E., Steffen, J., Sari, R. & Clarkson, W. On Detecting Terrestrial Planets with Timing of Giant Planet Transits. Mon. Not. R Astron. Soc. 359, 567-579 (2005). 2. Holman, M.J. & Murray, N.W. The Use of Transit Timing to Detect Terrestrial-Mass Extrasolar Planets. Science 307, 1288-1291 (2005). 3. Holman, M. J. et al. Kepler-9: A System of Multiple Planets Transiting a Sun-Like Star, Confirmed By Timing Variations. Science 330, 51-54 (2010). 4. Borucki, W.J., et al. Characteristics of Planetary Candidates Observed by Kepler. II. Analysis of the First Four Months of Data. Astrophys. J. 736, 19-40 (2011). 5. Batalha, N.M. et al. Planetary Candidates Observed by Kepler. III. Analysis of the First 16 Months of Data. Astrophys. J. Supp. 204, 24-44 (2013). 6. Burke, C.J. et al. Planetary Candidates Observed by Kepler IV: Planet Sample from Q1-Q8 (22 Months). Astrophys. J. Supp. 210, 19-30 (2014). 7. Pepe, F. et al. An Earth-sized Planet with an Earth-like Density. Nature 503, 377-380 (2013). 8. Howard, A.W. et al. A Rocky Composition for an Earth-sized Exoplanet. Nature 503, 381- 384 (2013). 9. Cochran, W. D. et al. Kepler-18b, c and d: A System of Three Planets Confirmed by Transit Timing Variations, Light Curve Validation, Warm Spitzer Photometry, and Radial Velocity Measurements. Astrophys. J. Supp. 197, 7-25 (2011). 10. Steffen. J.H. et al. Transit Timing Observations from Kepler - III. Confirmation of Four Multiple Planet Systems by a Fourier-domain Study of Anticorrelated Transit Timing Variations. Mon. Not. R. Astron. Soc. 421, 2342-2354 (2012). 11. Ford, E. B. et al. Transit Timing Observations from Kepler. V. Transit Timing Variation Candidates in the First Sixteen Months from Polynomial Models. Astrophys. J. 756, 185-191 (2012). 12. Steffen. J.H. et al. Transit Timing Observsations from Kepler. VI. Potentially Interesting Candidate Systems from Fourier-based Statistical Tests. Astrophys. J. 756, 186-190 (2012). 13. Lithwick. Y., Xie, J. & Wu, Y. Extracting Planet Mass and Eccentricity from TTV Data. Astrophys. J. 761, 122-132 (2012). 14. Wu. Y. & Lithwick, Y., Density and Eccentricity of Kepler Planets. Astrophys. J. 772, 74- 86 (2013). 15. Jontof-Hutter. D., Lissauer, J.J., Rowe, J.F. & Fabrycky, D.C. Kepler-79's Low Density Planets. Astrophys. J. 785, 15-18 (2014). 16. Lissauer. J.J. et al. All Six Planets Known to Orbit Kepler-11 Have Low Densities. Astrophys. J. 770, 131-145 (2013). 17. Masuda, K. Very Low Density Planets around Kepler-51 Revealed With Transit Timing Variations and an Anomoly Similar to a Planet-Planet Eclipse Event. Astrophys. J. 783, 53-60 (2014). 18. Ofir, A., Dreizler, S., Zechmeister, M. & Husser, T.-O. An Independent Planet Search in the Kepler Dataset. II. An Extremely Low-Density Super-Earth Mass Planet Around Kepler- 87. Astron. Astrophysics 561, 103-109 (2014). 19. Carter, J.A. et al. Kepler-36: A Pair of Planets With Neighboring Orbits and Dissimilar Densities. Science 337, 556-559 (2012). 20. Rowe, J.F. et al. Validation of Kepler's Multiple Planet Candidates. III. Light Curve Analysis and Announcement of Hundreds of New Multi-Planet Systems. Astrophys. J. 784, 45-64 (2014). 21. Kipping, D.M. et al. The Hunt for Exomoons with Kepler (HEK). IV. A Search for Moons around Eight M Dwarfs. Astrophys. J. 784, 28-42 (2014). 22. Ferraz-Mello, S., Tadeu Dos Santos, M., Beaugé, C., Michtchenko, T. A. & Rodríguez, A. On the Mass Determination of Super-Earths Orbiting Active Stars: the CoRoT-7 System. Astron. Astrophysics 531, 161-171 (2011). 23. Batalha, N.M. et al. Kepler's First Rocky Planet: Kepler-10b. Astrophys. J. 729, 27-47 (2011). 24. Fressin, F.G. et al. Two Earth-sized Planets Orbiting Kepler-20. Nature 482, 195-198 (2012). 25. Gautier, T.N. III. et al. Kepler-20: A Sun-like Star with Three Sub-Neptune Exoplanets and Two Earth-size Candidates. Astrophys. J. 749, 15-33 (2012). 26. Weiss, L.M. et al. The Mass of KOI-94d and a Relation for Planet Radius, Mass, and Incident Flux. Astrophys. J. 768, 14-32 (2013). 27. Marcy, G.W. et al. Masses, Radii, and Orbits of Small Kepler Planets: The Transition from Gaseous to Rocky Planets. Astrophys. J. Supp. 210, 20-89 (2014). 28. Bodenheimer, P. & Lissauer, J.J. Accretion and Evolution of ~2.5 Earth-mass Planets with Voluminous H/He Envelopes. Astrophys. J. 791, 103-134 (2014). 29. Lopez, E. D. & Fortney, J.J. The Role of Core Mass in Controlling Evaporation: The Kepler Radius Distribution and the Kepler-36 Density Dichotomy. Astrophys. J. 776, 2-12 (2013). 30. Owen, J.E. & Wu, Y. Kepler Planets: A Tale of Evaporation. Astrophys. J. 775, 105-116 (2013). Acknowledgements D.J. acknowledges support through the NASA Postdoctorial Program and funding from the Center for Exoplanets and Habitable Worlds. J.F.R. acknowledges NASA grant NNX14AB82G issued through the Kepler Participating Scientist Program. D.C.F. was supported by the NASA Kepler Participating Scientist Program award NNX14AB87G. E.B.F. was supported in part by NASA Kepler Participating Scientist Program award NNX14AN76G and NASA Exoplanet Research Program award NNX15AE21G. The Center for Exoplanets and Habitable Worlds is supported by the Pennsylvania State University, the Eberly College of Science, and the Pennsylvania Space Grant Consortium. Author Contributions D. Jontof-Hutter led the research effort to model the transit timing variations, constrain planetary masses, and wrote the manuscript. J. Rowe measured transit times from the Kepler dataset, characterized the host star using spectral follow-up of the target and constraints from the transits and edited the manuscript. J. Lissauer led the interpretation effort, assisted in the dynamical study and writing the manuscript. D. Fabrycky wrote software to simulate planetary transits, assisted in interpreting results and edited the manuscript. E. Ford assisted in the development of statistical methodologies and robustness tests for the TTV modeling and edited the manuscript. Author Information Reprints and permissions information is available at www.nature.com/reprints. Correspondence and requests for materials should be addressed to D. Jontof-Hutter [email protected]. Planet Period (days) T0 ecosω esinω -0.011 −0.140 +0.096 -0.015 −0.126 +0.086 -0.037 −0.092 +0.060 Rp (REarth) 0.522 +/-0.032 1.197 +/- 0.070 1.212 +/- 0.075 -0.024 −0.135 +0.075 -0.020 −0.117 +0.064 -0.057 −0.387 +0.674 Density (g cm-3) 2.6 −1.5 +2.4 6.2 −3.4 +5.8 2.1 −1.2 +2.2 M p MSun MEarth Mstar 0.13 −0.08 +0.12 3.85 −2.30 +3.77 1.28 −0.78 +1.36 Incident Flux (rel. to Earth) 6.81 +/- 0.84 4.63 +/- 0.57 2.32 +/- 0.29 Stellar Parameters Mstar: 0.521 +/- 0.055 MSun Rstar: 0.442 +/- 0.024 RSun Teff: 3841 +/- 49 K Density: 9.5 +/- 2.2 g cm-3 [Fe/H]: -0.280 +/- 0.099 log(g) (cm s-2): 4.886 +/0.055 b c d Planet b c d 10.3126 −0.0006 +0.0004 13.7813 −0.0001 +0.0001 23.0881 −0.0008 +0.0009 Rp Rstar 0.0108 +/- 0.0003 0.0247 +/- 0.0005 0.0251 +/- 0.0007 (BJD-2,454,900) 788.4142 −0.0027 +0.0027 786.1289 −0.0005 +0.0005 796.6689 −0.0013 +0.0013 Mp (MEarth) 0.066 −0.037 +0.059 1.970 −1.120 +1.912 0.640 −0.387 +0.674 Table 1: Stellar and planetary parameters for the Kepler-138 system. The left column lists our adopted stellar parameters. The upper right panel lists the solutions for the parameters we have explored with dynamical modeling. The lower right panel shows our adopted physical characteristics for the three planets. Figure 1: Transit Timing Variations of the three planets orbiting Kepler-138. In black are the differences between measured transit times and a calculated linear fit to the transit times, with 1σ uncertainties shown as error-bars. Grey points mark the difference between the simulated transit times based on the best-fit dynamical model and a linear fit to the transit times. Panels ‘a’, ‘b’ and ‘c’ display the TTVs of Kepler-138 b, c and d respectively. Figure 2: Mass-Radius diagram of well-characterized planets smaller than 2.1 REarth. Prior exoplanet characterizations are shown as grey points6,7,9,16,19,22,23,24,25,26,27. Black points from left to right are Mercury, Mars, Venus and Earth. Red data points are our results for Kepler-138. Open circles mark previously measured masses for Kepler-138 c and d21. Error bars mark published 1σ uncertainties for the planets of Kepler-138 and published masses and radii of all other characterized exoplanets within this size range. The curves mark bulk densities of 1, 3 and 10 g cm-3. Methods We have used all available short cadence Kepler data and long cadence data wherever short cadence are unavailable to complete the dataset for 17 quarters. We list the transit times for each planet in Extended Data (ED) Table 1. Throughout, we express times since Barycentric Julian Day (BJD)- 2,454,900. Photometric Transit and Stellar Models From the light curve, we filtered instrumental and astrophysical effects that are independent of planetary transits. To each segment of the photometric time series, we fitted a cubic polynomial of width 2 days, centered on the time of each measurement20. We excluded measurements taken within 1 transit-duration (defined as the time from first to last contact) of the measured center of the transit and extrapolate the polynomial to estimate corrections during transits. This process strongly filters astrophysical signals with timescales of approximately 2 days, which could affect the shape of a planetary transit. We also excluded measurements for which the associated segment has gaps longer than 2.5 hours. We fitted the detrended Kepler light curve using a transit model for quadratic limb- darkening31 and non-circular Keplerian orbits. We stacked transits of each planet with corrections for the measured TTVs20. To account for Kepler’s observation cadence, we averaged our transit model with 11 equal spacings within the 1 minute or 30 minute integration window. We evaluated the photometric noise for each quarter of data to fit transit models, adopting published stellar parameters for Kepler-13832. We adopted a two-parameter quadratic model for limb-darkening with fixed coefficients (0.3576, 0.3487) appropriate for Kepler’s bandpass and Kepler-138’s effective temperature (Teff), log(g) and metallicity [Fe/H]33. The light curve model parameters consist of the mean stellar density34 (ρstar), a photometric zero point for each light curve segment, and for each planet the orbital period, time of transit, planet-to-star radius ratio, impact parameter, and eccentricity parameterized as ecosω and esinω. We determined posterior distributions of our model parameters using MCMC techniques20. Our best-fit transit models, shown in ED Figure 1, resulted in consistent estimates for ρstar from each planet. We determined the mass and radius of Kepler-138 by fitting the spectroscopic parameters (Teff , [Fe/H])32 and our light curve contraints of ρstar to Dartmouth Stellar Evolution models35, assuming a Gaussian probability density for each parameter32. For the Dartmouth models, we varied initial conditions of Mass, Age, and [Fe/H] and interpolated over a grid to evaluate Teff, ρstar, and [Fe/H] for any set of initial conditions. We computed posteriors using MCMC to obtain stellar model-dependent posteriors on Mstar and Rstar. Table 1 lists our adopted stellar parameters. We tested the effects of eccentricity priors on our measurement of ρstar, adopting a uniform prior in eccentricity as our nominal results. We compare results with eccentricity fixed at zero in ED Figure 2. Although a uniform prior on eccentricity results in a slightly wider range of inferred radii for the star, both of these models are consistent with the spectroscopic study of Kepler-13832. Our solution for the impact parameter of the middle planet is 0.3 +/- 0.2, significantly lower than the previous estimate of 0.92 +/- 0.0221. The apparent U-shaped transit for planet ‘c’ (ED Figure 1) is consistent with a low impact parameter. Our measured impact parameter for planet ‘d’ is 0.810 +/- 0.057, consistent with the previous measurement21. Our revised impact parameters imply ρstar = 9.0 +/- 1.9 g cm-3, agrees with our transit models for each planet, and with the spectroscopic study of Kepler-13832. We find that Kepler-138 has a smaller mass and radius than previous estimates based on the absolute K-band magnitude (MK) of the star from high-resolution Keck spectra36. These relied on mass-luminosity relations37 and a mass-radius relation from interferometry38. However, the calibration stars used to correlate the MK to spectral index excluded cool stars that are active. In the case of Kepler-138, the photometric time-series exhibits large (1%) variations due to starspots. These increase the risk of systematic errors in the measurement of stellar luminosity, and therefore the stellar properties derived from the mass-luminosity relation. The time-scale for star-spot modulation, ~20 days, was much longer than the transit duration, and was likely dominated by two spots. We found no evidence of star-spot crossings, nor did we find any TTV periodicities related to the rotation period of the star. Hence, the stellar activity is unlikely to effect our transit model or transit times. Analytical Constraints from Transit Timing Variations Orbital period ratios determine how close planets are to mean motion resonance, and over what period TTVs are expected to cycle. The inner pair of planets of Kepler-138 orbit near the 4:3 resonance with an expected TTV cycle of 1570 days, slightly longer than the 1454-day observational baseline of Kepler-138 b's transits. We fitted a sinusoid at this periodicity to the TTVs of the inner planet, and detected a TTV amplitude of 34 +/- 4 minutes. This permits a rough estimate for the mass13 of the middle planet of 6.8 +/- 0.9 MEarth (MSun/Mstar), which is close to our final measure of the mass of the middle planet. The detection of TTVs at planet ‘b’ imply that the TTVs in ‘c’ have a component caused by planet ‘b’. However, the middle planet's TTVs are the combined effect of perturbations from its two neighbours ‘b’ and ‘d’, and the outer pair orbit near a second order mean motion resonance for which there is no known analytical model. Hence, the masses of the inner and outermost planets cannot be estimated by fitting such a simplified sinusoidal model to the TTVs. Detailed TTV Modelling For each set of initial conditions, we calculate the transit times of all three planets based on Newtonian gravity, using an eighth order Dormand-Prince Runge-Kutta integrator15,16,39. We compare the simulated and observed transit times for each planet assuming each observed transit time has an independent Gaussian measurement uncertainty. We found an excess of outlying transit times to our model, where either instrumental effects or stellar activity led to a few unlikely transit times with underestimated uncertainties. ED Figure 3 shows the distribution of residuals to our best-fit TTV model compared to a Gaussian distribution, revealing these outliers. Of the 257 measured transit times, 5 are outliers where for simulated transit times S, and measurement uncertainties σTT, (O-S)/σTT > 3. We removed these outliers, and used the 252 remaining measurements as our nominal dataset for dynamical models. Later, we tested our results for robustness against outliers. Performing extensive grid searches and Levenberg-Marquardt, we found a single region of high posterior probability including multiple closely related local minima. To characterize the masses and orbital parameters of the three planets, we performed Bayesian parameter estimation using a Differential Evolution Markov Chain Monte Carlo (DEMCMC) algorithm40,41. We used Metropolis-Hasting acceptance rules on 45 "walkers" exploring parameter space in parallel, where for each walker, a proposal is a scaled vector between two other walkers chosen at random. Using differential evolution for the proposal steps increases the probability that proposals will be accepted, particularly for target distributions with significant correlations between model parameters. The walkers were launched near the best-fit model found by Levenberg-Marquardt. We updated the vector scale length factor every 15 generations to keep the acceptance rate near the optimum value of 0.2540,41,42 and thinned the data by recording every 20th generation in the Markov Chain. We discarded the first 50,000 generations and continued the Markov Chain for 1,250,000 additional generations. The mean correlation between parameter values in the same MCMC chain for a given separation in the chain begins near unity for consecutive generations, and declines for greater separations43. We used this property to assess how well-mixed our MCMC chains were. The autocorrelation length, defined as the lag between generations required for the correlation to fall below 0.5, varied between the walkers, averaging 9000 generations for the planet-star mass ratio of Kepler-138 b. Hence, the mean length of the DEMCMC chains were ~139 autocorrelation lengths. For each planet, we adopted a uniform prior in orbital period, T0, e, and ω. For Mp/Mstar, we adopted a uniform prior that allowed negative masses to enable a simple estimate for the significance of a positive mass for Kepler-138 b from the posterior samples27. For Kepler-138 b, the mass is greater than zero in 99.84% of posterior samples. As a further test, we considered an alternative more realistic uniform prior where planetary masses are positive definite and limited to the mass of a pure iron planet given their sizes44,45. Posteriors for the mass ratios of each planet to the host star, and each eccentricity vector component are shown in ED Figure 4. We list 68.3%, 95.4%, and 99.7% credible intervals for all parameters in ED Table 2. We also include the mass ratios between the planets and relative eccentricity vector components, which are more constrained by the data than absolute masses and eccentricities. Orbital Eccentricity Joint posteriors for the eccentricity vector components are displayed in ED Figure 5. These show extreme correlations due to a broad class of orbital models satisfying the data, in which there is a precise apsidal alignment of orbits. We performed long-term integrations on a subset of our posterior planet masses and orbital parameters including a wide range of eccentricities using the HNBODY code46. We investigated whether long-term stability could further constrain the eccentricities. A sample of the solutions were integrated for 10 million orbits and all were found to be stable. Integrating one of the best-fit solutions with high eccentricities for b, c and d at 0.23, 0.20 and 0.18 respectively, we confirmed that the orbits were stable for over 1 Gyr; the apsidal lock was maintained with periapses of all three planets closely aligned. Although rare in the Solar System, apsidal alignment has been observed and studied in the Uranian ring system47,48 and it has been detected in exoplanetary systems, including υ Andromedae49, GJ 87650 and possibly 55 Cancri51. Significance of Mass Detection for Kepler-138 b We establish the significance of a non-zero mass for Kepler-138 b by computing the Bayes factor, i.e., the ratio of the marginalized posterior probability for a model with the mass of planet Kepler-138 b fixed at zero relative to the marginalized posterior probability for a model where all three planets have a non-zero mass. Intuitively, the Bayes factor quantifies how much the transit timing data has increased our confidence that planet b has a non-zero mass. When performing Bayesian model selection, it is essential to choose proper (i.e., normalized) priors for any parameters not occurring in both models. For the mass of Kepler- 138 b we adopt a uniform prior ranging between zero mass and the mass of an iron sphere the size of Kepler-138 b. We tested two models of iron planets as upper limits on our mass priors44,45. We compute the Bayes factor using the generalized Savage-Dickey density ratio based on the posterior samples from our nominal model52. We find that the three massive planet model is strongly favored, with the posterior probability for the three massive planet model equal to 99.82%44 (99.80%45) for our nominal model (i.e., three massive planets, each with a uniform eccentricity prior) or 99.91%44 (99.90%45) for a model with three massive planets, each with a Rayleigh distribution (sigma = 0.0253) for an eccentricity prior. A more restrictive prior for the mass of Kepler-138 b would further increase the posterior probability for the three massive planet model. The generalized Savage-Dickey density ratio (SDDR) is superior to more commonly used substitutes (e.g., Akaike information criterion, AIC, or Bayesian Information Criterion, BIC), since the SDDR provides a practical means for calculating the Bayes factor, the actual quantity of interest for rigorous Bayesian model comparison. The AIC and BIC use only the likelihood of the two best-fit models and do not account for the width or shape of the posterior probability distributions. ED Figure 4a shows that the marginal posterior for the mass of planet ‘b’ is asymmetric and non-Gaussian, so the AIC or BIC would be a particularly poor choice for our problem. Therefore, we have computed the rigorously correct Bayes factor using the SDDR, which provides an efficient way of calculating the Bayes factor when comparing two nested models, meaning that the simpler model is equivalent to the more general model when the additional parameters (θ) take on a particular value (θ=θ0). In our case, this occurs when Mb = 0. One advantage of the SDDR over computing fully marginalized likelihoods is that the SDDR can be computed from the posterior for the more general model. This is computationally practical when comparing models that differ by one to a few dimensions, since then the posterior at θ0 can be computed from a posterior sample using a kernel density estimator. We estimate the posterior density using a Gaussian kernel density estimator with bandwidth 0.001 MEarth(Mstar/MSun), which was found to be optimal when analyzing synthetic posterior samples data. We verified that the results were insensitive to varying the choice of bandwidth by an order of magnitude. Sensitivity Analyses We performed several tests to assess the robustness of our results to: 1) the choice of prior for eccentricity, 2) the treatment of transit time outliers, 3) the assumption of coplanarity, and 4) our algorithm. For these sensitivity analyses, we adopt our nominal prior for planet masses to allow for negative planet masses. While any such models are clearly unphysical, allowing for such model offers an efficient and intuitive means for evaluating whether the lower limit on the planet masses is robust to the above assumptions. Choice of Eccentricity Prior As noted above, our nominal model has a uniform prior in eccentricity. The joint posterior for the mass ratio of Kepler-138 b to the host star and its orbital eccentricity is shown in ED Figure 5. We show this plot with two alternative priors in eccentricity; a Rayleigh distribution with a scale length 0.154, and a more constrained one, consistent with Kepler’s multi-planet systems, with a scale length of 0.0253. Since the data constrain the eccentricity so weakly, the choice of priors strongly affects the posteriors of eccentricity. However, the planet-star mass ratios were much more weakly affected by the choice of eccentricity prior, as shown in ED Figure 7a. Because the apsidally-locked solutions are long-term stable, we adopt the uniform prior as our nominal solution, and note that a more constraining prior on eccentricity results in a marginally wider posterior for mass ratio of Kepler-138 b to the star. Although, as noted above, the orbital eccentricities in the system are only weakly constrained, the relative eccentricities and mass ratios between the planets are tightly constrained by the TTVs, as shown in ED Figure 6 and ED Table 2. Transit Time Observation Outliers To assess the effect of residual outlying transit times, we repeated the analysis with two other sets of measured transit times, differing only in that we removed transit time residual outliers beyond i) 4σ (leaving 254 transit times) or ii) 2.5σ (leaving 244 transit times), as opposed to our nominal dataset which excluded 3σ outliers. ED Figure 3 shows the relative frequency of residuals compared to a Gaussian. The majority of outliers are between 2.5 and 3σ. We excluded outliers beyond 4σ from all models. The posterior for the mass of Kepler-138 b for each of these three datasets is displayed for comparison in ED Figure 7b. Overall, the outliers have a modest effect on the measured mass ratio for Kepler-138 b to the host star. Since most of the outliers are in the transit times of Kepler-138 b, these have a small effect on our mass measurement of planet ‘b’. Furthermore, the mass of planet ‘c’ is constrained by its effect on the transit times of both planets ‘b’ and ‘d’. Nevertheless, we note that the inclusion of more outliers increases the skewness of the posterior for planetary mass and causes the mode of the distribution to shift to a slightly lower mass. Assumption of Co-planarity The known low inclination dispersion amongst Kepler’s multiplanet systems makes coplanarity a reasonable assumption for TTV modeling55,56,13. Furthermore, geometric considerations make multi-planet transiting less likely to be observed for systems with large mutual inclinations. Nevertheless, we tested the effect of mutual inclinations on our solutions. We performed two additional sets of simulations: i) with the longitude of the ascending node of ‘b’ as a free parameter, leaving the other two planets as coplanar, and ii) with a free ascending node for ‘d’. In each case, we adopted a uniform prior for the ascending node. In ED Figure 7c, we compare the posteriors for the mass ratio of Kepler-138 b to the host star with mutual inclinations to our nominal result. The nominal result gives a consistent, but slightly wider posterior than with free ascending nodes. Tests with Synthetic Transit Times We evaluated our method by generating synthetic datasets of transit times with known planetary masses and orbital parameters to test how well the input parameters were recovered. The results are shown in ED Figure 8. We generated synthetic transit times based on the median values of each parameter from the marginal posteriors of our nominal model and added Gaussian noise to each observation with a standard deviation equal to the measured timing uncertainties. Our analysis of the simulated data results in a posterior for the mass of Kepler-138 b consistent with the true values. While the resulting posterior for the mass of Kepler-138 b shows a slightly higher mode and is less skewed than the posterior for the nominal model, the differences are comparable to the minor effects of transit timing outliers or choice of eccentricity prior. This result validates our both our transit timing method and our TTV analysis. Additionally, we generated eight independent synthetic datasets with zero mass for Kepler- 138 b and other parameters based on our nominal model (ED Figure 8). In all eight cases, the posterior probability for planet b's mass was insignificant, with only 16% to 86% of posterior samples having a mass greater than zero, consistent with expectations for non-detections. This is in sharp contrast to our analysis of the actual data, which results in 99.84% of the posterior yielding a positive mass for Kepler-138 b. Planetary Characteristics Our adopted credible intervals in planetary mass and density were calculated by repeatedly multiplying samples from the posteriors of planet-star mass ratios, and Mstar. Uncertainties in planetary radii were calculated with the fractional uncertainty in the stellar radius and the uncertainty in planet-star radius ratio added in quadrature. Time-averaged incident flux for each planet compared to the Earth was calculated in the low eccentricity limit, although we note that if the orbits were highly eccentric, the fluxes would be marginally higher. 31. Mandel, K. & Agol, E. Analytic Light Curves for Planetary Transit Searches. Astrophys. J. 580, 171-175 (2002). 32. Muirhead, P.S. et al. Characterizing the Cool Kepler Objects of Interest. New Effective Temperatures, Metallicities, Masses and Radii of Low-mass Kepler Planet-candidate Host Stars. Astrophys. J. 750, 37-42 (2012). 33. Claret, A. & Bloeman, S. Gravity and Limb-Darkening Coefficients the Kepler, CoRoT, Spitzer, uvby, UBVRIJHK, and Sloan photometric systems. Astron. Astrophysics 529, 75-79 (2011). 34. Seager, S. & Mallen-Ornelas, G. A Unique Solution of Planet and Star Parameters from an Extrasolar Planet Transit Light Curve. Astrophys J. 585, 1038-1055 (2003). 35. Dotter, A. The Dartmouth Stellar Evolution Database. Astrophys. J. Supp. 178, 89-101 (2008). 36. Pineda, J.S., Bottom, M. & Johnson, J.A. Using High Resolution Optical Spectra to Measure Intrinsic Properties of Low-Mass Stars: New Properties for KOI-314 and GJ 3470. Astrophys. J. 767, 28-38 (2013). 37. Delfosse, X. et al. Accurate Masses of Very Low Mass Stars. IV. Improved Mass- Luminosity Relations. Astron. Astrophysics 364, 217-224 (2000). 38. Boyajian, T.S. et al. Stellar Diameters and Temperatures. II. Main Sequence K- and M- Stars. Astrophys. J. 757, 112-142 (2012). 39. Lissauer J.J. et al. A Closely Packed System of Low-mass, Low-density Planets Transiting Kepler-11. Nature 470, 53-58 (2011). 40. Ter Braak, C. A Markov Chain Monte Carlo Version of the Genetic Algorithm. Differential Evolution: Easy Bayesian Computer for Real Parameter Spaces. Statistics and Computing 16, 239-249 (2006). 41. Nelson, B., Ford, E.B. & Payne, M.J. RUN DMC: An Efficient, Parallel Code for Analyzing Radial Velocity Observations Using N-body Integrations and Differential Evolution Markov Chain Monte Carlo. Astrophys. J. Supp. 210, 11-23 (2014). 42. Ford, E. Quantifying the Uncertainty in the Orbits of Extrasolar Planets, Astron. J. 129, 1706-1717 (2005). 43. Tegmark, M. et al. Cosmological Parameters from SDSS and WMAP. Physical Review D 69, 103501 (2004). 44. Fortney, J. J. Marley, M. S. & Barnes, J. W. Planetary Radii Across Five Orders of Magnitude in Mass and Steller Insolation: Application to Transits. Astrophys. J. 659, 1661- 1672 (2007). 45. Zeng, L. & Sasselov, D. A Detailed Model Grid for Solid Planets from 0.1 through 100 Earth Masses, Pub. Astron. Soc. Pacific 125, 925, 227-239 (2013). 46. Rauch, K.P. & Hamilton, D.P. HNBody: Hierarchical N-Body Symplectic Integration Package. Astrophysics Source Code Library (2012). 47. Goldreich, P. & Tremaine, S. Precession of the Epsilon Ring of Uranus. Astron. J. 84, 1638-1641 (1979). 48. Chiang, E. I. & Goldreich, P. Apse Alignment of Narrow Eccentric Planetary Rings. Astrophys. J. 540, 1084-1090 (2000) 49. Chiang. E. I., Tabachnik, S. & Tremaine, S. Apsidal Alignment in Upsilon Andromedae. Astron. J. 122, 1607-1615 (2001). 50. Rivera, E.J., Laughlin, G., Butler, R.P., Vogt, S.S. & Haghighipour, N. The Lick-Carnegie Exoplanet Survey. A Uranus-Mass Fourth Planet for GJ 876 in an Extrasolar Laplace Configuration. Astrophys. J. 719, 890-899 (2010). 51. Nelson, B.E. et al. The 55 Cancri Planetary System: Fully Self-Consistent N-body Constraints and a Dynamical Analysis. Mon. Not. R. Astron. Soc. 441, 442-451 (2014). 52. Verdinelli, I. & Wasserman, L., Computing Bayes Factors Using a Generalization of the Savage-Dickey Density Ratio. Journal of the American Statistical Association. 90, 430, 614- 618 (1995). 53. Hadden, S. & Lithwick, Y., Densities and Eccentricities of 139 Kepler Planets from Transit Timing Variations. Astrophys. J. 787, 80-87 (2014). 54. Moorhead, A.V. et al. The Distribution of Transit Durations for Kepler Planet Candidates and Implications for their Orbital Eccentricities. Astrophys. J. Supp. 197, 1-15 (2011). 55. Fabrycky, D.C. et al. Architecture of Kepler’s Multi-transiting Systems. II. New Investigations with Twice as Many Candidates. Astrophys. J. 790, 146-157 (2014). 56. Lissauer, J.J. et al. Architecture and Dynamics of Kepler’s Candidate Multiple Transiting Planet Systems. Astrophys. J. 197, 8-33 (2011). Extended Data Table 1: Transit times of Kepler-138. All times are expressed in days since Barycentric Julian Day: 2,454,900. Estimated uncertainties give 68% confidence limits. Outliers beyond 3σ in the residuals of dynamical fits are marked with an asterisk. Extended Data Table 2: Confidence intervals from distributions found with DEMCMC TTV analysis. We include the parameters of our dynamical fits, as well as the mass ratios and relative eccentricity vector components between the planets, which have tighter constraints than the absolute masses or eccentricity vector components. Extended Data Figure 1: Folded light curves with corrections for observed transit timing variations for Kepler-138. The scattered points are photometric relative fluxes and the curves are analytical models of the transit shape described in the text. Panels ‘a’, ‘b’ and ‘c’ correspond to Kepler-138 b, c and d respectively. Extended Data Figure 2: Stellar mass and radius models using constraints on the stellar mean density inferred from the light curve. In cyan are models that adopted a uniform prior in eccentricity, and in magenta constraints found with orbital eccentricities fixed at zero. Grey points with error bars mark stellar parameters found in the literature whilst the black error bars mark our adopted solution for stellar mass and radius. Extended Data Figure 3: The distribution of residual normalized deviations from our best fit dynamical model to the raw transit times. The histogram marks deviations (O-S)/σTT where O is the observed transit time, S is the simulated transit time and σTT is the measurement uncertainty. The curve marks a Gaussian distribution. Extended Data Figure 4: Posterior distributions for TTV model parameters. The planet-star mass ratios (Mp/Mstar) for each planet of Kepler-138 b, c, and d are shown in panels a, b, and c, ecosω (in panels d,e and f), and esinω (in panels g,h, and i) respectively for our nominal model. The relative frequency for each histogram is scaled to the mode. Extended Data Figure 5: Joint posteriors of model parameters and the effects of eccentricity priors. The dark (light) grey marks the 68.3% (95.4%) credible intervals for each joint posterior. Panels ‘a’, ‘b’ and ‘c’ plot Mp/Mstar and eccentricity vector components for the inner and middle planets, whilst panels ‘d’, ‘e’, and ‘f’ plot the same for the middle and outer planets. Panels g, h and i compare Mp/Mstar for Kepler-138 b only and its orbital eccentricity, for three eccentricity priors, a uniform prior on eccentricity (g), and models with a Rayleigh distribution of scale factor 0.1 (h), and 0.02 (i). Extended Data Figure 6: Posterior distributions for mass ratios and relative eccentricities between planets. The mass ratio of the inner and middle planets is shown in panel ‘a’, and relative eccentricity vector components (the difference in ecosω (esinω) in the inner pair in panel b (c)). The mass ration of the middle and outer planets are plotted in panels ‘d’, relative eccentricity vector components (the difference in ecosω (esinω) in the outer pair in panel e (f)). Extended Data Figure 7: Sensitivity tests for the effects of eccentricity prior, outlying transit times and free inclinations on the mass of Kepler-138 b relative to the host star. Panel ‘a’ compares a uniform prior (black curve, our nominal posterior for all comparisons) and a Rayleigh Distribution with scale factors 0.1 (navy) and 0.02 (cyan). Panel ‘b’ compares posteriors with 3σ outliers excluded (black), with two alternatives; 4σ outliers (blue) and 2.5σ outliers removed (light green). Panel ‘c’ compares our nominal model with one with a free ascending node for the inner (purple) or outer (red) planet. Extended Data Figure 8: Validation of our method with synthetic datasets. The green curve marks the posterior for a synthetic dataset generated with the same parameters as the medians of our nominal posteriors (in Table 1). The agreement between the green and black curves validates our method and our claim for a positive mass detection for Kepler-138 b. The magenta and purple shades are posteriors for models using data generated with zero mass for Kepler-138 b. These zero-mass synthetic models all reproduced non-detections. – 1 – 150 150 a a Kepler-138 b Kepler-138 b 15 15 b b Kepler-138 c Kepler-138 c 30 30 c c Kepler-138 d Kepler-138 d 100 100 ) ) s s e e t t u u n n i i m m ( ( V V T T T T 50 50 0 0 -50 -50 -100 -100 10 10 5 5 0 0 -5 -5 -10 -10 -15 -15 -20 -20 20 20 10 10 0 0 -10 -10 -20 -20 -30 -30 0 0 400 400 800 800 1200 1200 1600 1600 0 0 400 400 800 800 1200 1200 1600 1600 0 0 400 400 800 800 1200 1200 1600 1600 BJD-2,454,900 BJD-2,454,900 BJD-2,454,900 BJD-2,454,900 BJD-2,454,900 BJD-2,454,900 Fig. 1.— Main paper, Figure 1 – 2 – -3 -3 1 g cm 3 g c m 0 g 1 3 - c m ⊕ R p R / 2 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0.01 0.1 1 10 Mp/M⊕ Fig. 2.— Main paper, Figure 2 – 3 – – 4 – Fig. 4.— ED Table 2 – 5 – Fig. 5.— ED Figure 1 Fig. 6.— ED Figure 2 – 6 – Fig. 7.— ED Figure 3 – 7 – Fig. 8.— ED Figure 4 – 8 – Fig. 9.— ED Figure 5 – 9 – Fig. 10.— ED Figure 6 Fig. 11.— ED Figure 7 – 10 – Fig. 12.— ED Figure 8
1505.06576	1	1505	2015-05-25T09:26:33	Modelling the local and global cloud formation on HD 189733b	[ "astro-ph.EP" ]	Context. Observations suggest that exoplanets such as HD 189733b form clouds in their atmospheres which have a strong feedback onto their thermodynamical and chemical structure, and overall appearance. Aims. Inspired by mineral cloud modelling efforts for Brown Dwarf atmospheres, we present the first spatially varying kinetic cloud model structures for HD 189733b. Methods. We apply a 2-model approach using results from a 3D global radiation-hydrodynamic simulation of the atmosphere as input for a detailed, kinetic cloud formation model. Sampling the 3D global atmosphere structure with 1D trajectories allows us to model the spatially varying cloud structure on HD 189733b. The resulting cloud properties enable the calculation of the scattering and absorption properties of the clouds. Results. We present local and global cloud structure and property maps for HD 189733b. The calculated cloud properties show variations in composition, size and number density of cloud particles which are strongest between the dayside and nightside. Cloud particles are mainly composed of a mix of materials with silicates being the main component. Cloud properties, and hence the local gas composition, change dramatically where temperature inversions occur locally. The cloud opacity is dominated by absorption in the upper atmosphere and scattering at higher pressures in the model. The calculated 8{\mu}m single scattering Albedo of the cloud particles are consistent with Spitzer bright regions. The cloud particles scattering properties suggest that they would sparkle/reflect a midnight blue colour at optical wavelengths.	astro-ph.EP	astro-ph	Astronomy & Astrophysics manuscript no. Modelling_the_local_and_global_cloud_formation_on_HD_189733b c(cid:13)ESO 2021 June 4, 2021 Modelling the local and global cloud formation on HD 189733b E. K. H. Lee1, Ch. Helling1, I. Dobbs-Dixon2, and D. Juncher3 1 SUPA, School of Physics and Astronomy, University of St Andrews, North Haugh, St Andrews, Fife KY16 9SS, 2 NYU Abu Dhabi, PO Box 129188, Abu Dhabi, UAE 3 Niels Bohr Institute & Centre for Star and Planet Formation, University of Copenhagen, DK-2100 Copenhagen, UK Denmark Received: 27/02/2015 / Accepted: 21/05/2015 ABSTRACT Context. Observations suggest that exoplanets such as HD 189733b form clouds in their atmospheres which have a strong feedback onto their thermodynamical and chemical structure, and overall appearance. Aims. Inspired by mineral cloud modelling efforts for Brown Dwarf atmospheres, we present the first spatially varying kinetic cloud model structures for HD 189733b. Methods. We apply a 2-model approach using results from a 3D global radiation-hydrodynamic simulation of the atmosphere as input for a detailed, kinetic cloud formation model. Sampling the 3D global atmosphere structure with 1D trajectories allows us to model the spatially varying cloud structure on HD 189733b. The resulting cloud properties enable the calculation of the scattering and absorption properties of the clouds. Results. We present local and global cloud structure and property maps for HD 189733b. The calculated cloud properties show variations in composition, size and number density of cloud particles which are strongest between the dayside and nightside. Cloud particles are mainly composed of a mix of materials with silicates being the main component. Cloud properties, and hence the local gas composition, change dramatically where temperature inversions occur locally. The cloud opacity is dominated by absorption in the upper atmosphere and scattering at higher pressures in the model. The calculated 8µm single scattering Albedo of the cloud particles are consistent with Spitzer bright regions. The cloud particles scattering properties suggest that they would sparkle/reflect a midnight blue colour at optical wavelengths. Key words. planets and satellites: individual: HD 189733b -- planets and satellites: atmospheres -- methods: numerical 1. Introduction The atmospheres of exoplanets are beginning to be char- acterised in great detail. Spectroscopic measurements from primary and secondary transits of hot Jupiters have opened up the first examinations into the composition of their atmospheres. HD 189733b is one of the most stud- ied hot Jupiters due to its bright parent star and large planet-to-star radius ratio. Observational data of this ob- ject has been collected by various groups from X-ray to radio wavelengths (Knutson et al. 2007; Tinetti et al. 2007; Charbonneau et al. 2008; Grillmair et al. 2008; Swain et al. 2008, 2009; Danielski et al. 2014). Tran- sit spectra from the Hubble space telescope (Lecave- lier Des Etangs et al. 2008; Sing et al. 2011; Gibson et al. 2012) between 0.3µm and 1.6µm show a fea- tureless spectrum consistent with Rayleigh scattering from the atmosphere. These observations sample the outer limbs (dayside-nightside terminator regions) of the planet's atmosphere during primary eclipse. Lecave- lier Des Etangs et al. (2008) suggests haze/clouds to be responsible for the Rayleigh slope and estimate cloud particle radii of 10−2. . .10−1 µm at ∼10−6. . .10−3 bar local gas pressures. They suggest MgSiO3[s] as the most likely cloud particle material due to its strong scattering properties. Pont et al. (2013) who reanal- ysed and combined the Hubble and Spitzer (Agol et al. 2010; Désert et al. 2011; Knutson et al. 2012) obser- vations found that the spectrum was 'Dominated by Rayleigh scattering over the whole visible and near- infrared range. . .' and also suggest cloud/haze layers as a likely scenario. Subsequent analysis by Wakeford & Sing (2015) showed that single composition cloud condensates could fit the Rayleigh scattering slope with grain sizes of 0.025µm. Evans et al. (2013) measured the geometric albedo of HD 189733b over 0.29. . .0.57 µm and infer the planet is likely to be a deep blue colour at visible wavelengths, which they attribute to scattering mid-altitude clouds. These albedo observations sample the dayside face of the planet by comparing measure- ments before, during and after the secondary transit (oc- culation). Star spots (McCullough et al. 2014), low at- mospheric metallicity (Huitson et al. 2012) and photo- chemical, upper atmosphere haze layers (Pont et al. Article number, page 1 of 23 5 1 0 2 y a M 5 2 . ] P E h p - o r t s a [ 1 v 6 7 5 6 0 . 5 0 5 1 : v i X r a A&A proofs: manuscript no. Modelling_the_local_and_global_cloud_formation_on_HD_189733b 2013) have been proposed as alternative explanations to these observations. Clouds present in optically thin atmospheric regions are thought to significantly impact the observed spectra of exoplanets by flattening the ultra-violet and visible spectrum through scattering by small cloud particles, by depletion of elements and by providing an additional opacity source (e.g GJ 1214b; Kreidberg et al. (2014)). The strong radiative cooling (or heating) resulting from the high opacity of the cloud particles also affect the local pressure which influences the local velocity field (Helling et al. 2004). These effects have been observed and modelled in Brown Dwarf atmospheres which are the inspiration for the current study. Clouds are also im- portant for ionisation of the atmosphere by dust-dust collisions (Helling et al. 2011) and cosmic ray ionisa- tion (Rimmer & Helling 2013). This leaves open the possibility of lightning discharge events in exo-clouds (Bailey et al. 2014). Furthermore, Stark et al. (2014) suggest that charged dust grains could aid the synthesis of prebiotic molecules by enabling the required ener- gies of formation on their surfaces. Multi-dimensional atmosphere simulations have been used to gain a first insight into the atmospheric dynamics of HD 189733b (e.g. Showman et al. (2009); Fortney et al. (2010); Knut- son et al. (2012); Dobbs-Dixon & Agol (2013)). Out- put from Showman et al. (2009) has been used for ki- netic, non-equilibrium photo-chemical models of HD 189733b (Moses et al. 2011; Venot et al. 2012; Agúndez et al. 2014). In this work, we present cloud structure calculations for HD 189733b based on 3D radiation-hydrodynamic [3D RHD] atmosphere results of Dobbs-Dixon & Agol (2013). With this 2-model approach, we investigate how the specific cloud properties vary locally and globally throughout HD 189733b's atmosphere. In Sect. 2 we outline our 2-model approach. We summarise our cloud formation model and discuss our approach to atmo- spheric mixing, cloud particle opacities and model lim- itations. In Sect. 3 the spatially varying cloud properties (particle sizes, number density, material composition, nucleation rate, growth rate and opacity) are presented. In Sect. 4 we use the results of the locally and spatially varying cloud properties to calculate the wavelength de- pendent opacity and contribution of the clouds to the reflected light and transit spectra for HD 189733b. Sec- tion 5 discusses our results with respect to observational findings for HD 189733b. Our conclusions are outlined in Sect. 6. 2. 2-model approach We apply our kinetic dust cloud formation model to re- sults from 3D radiation-hydrodynamic [3D RHD] sim- ulations of HD 189733b's atmosphere (Dobbs-Dixon & Agol 2013) and present a first study of spatially vary- ing cloud formation on a hot Jupiter. We briefly sum- marise the main features of the kinetic cloud formation model (see: Woitke & Helling (2003, 2004); Helling et al. (2008b); Helling & Fomins (2013)). This cloud model has been successfully combined with 1D atmo- sphere models (DRIFT-PHOENIX; Dehn et al. (2007); Article number, page 2 of 23 Table 1. Input quantities for the cloud formation model. Lo- cal Tgas, pgas, ρgas, vz and z are taken from the 3D radiation- hydrodynamic model. Input Tgas (r) pgas (r) ρgas (r) vz (r) z ε0 x /εH(r) g (r) Definition local gas temperature local gas pressure local gas density local vertical gas velocity vertical atmospheric height element abundance surface gravity Units K dyn cm−2 g cm−3 cm s−1 cm - cm s−2 Helling et al. (2008a); Witte et al. (2009, 2011)) to pro- duce synthetic spectra of M Dwarfs, Brown Dwarfs and non-irradiated hot Jupiter exoplanets. We then sum- marise the multi-dimensional radiative-hydrodynamical model from Dobbs-Dixon & Agol (2013) used as in- put for the kinetic cloud formation model. Finally, we outline our approach in calculating the absorption and scattering properties of multi-material, multi-disperse, mixed cloud particles. 2.1. Cloud formation modelling Cloud formation proceeds via a sequence of processes that are described kinetically in our cloud formation model: 1. Formation of seed particles by homomolecular ho- mogeneous nucleation (Jeong et al. 2003; Lee et al. 2015). 2. Growth of various solid materials by gas-grain chemical surface reactions on the existing seeds or grains (Gail & Sedlmayr 1986; Helling & Woitke 2006; Helling et al. 2008b). 3. Evaporation of grains when the materials that they are composed of become thermally unstable (Helling & Woitke 2006; Helling et al. 2008b). 4. Gravitational settling allowing a continuation of grain growth through transport of grains out of under-saturated regions (Woitke & Helling 2003, 2004). 5. Element depletion in regions of cloud formation which can stop new grains from forming (Helling & Woitke 2006). 6. Convective/turbulent mixing from deeper to higher atmospheric regions to provide element replenish- ment (Woitke & Helling 2003, 2004). Without the nucleation step 1., grains do not form and a stationary atmosphere remains dust-free. We refer the reader to Helling & Fomins (2013) for a summary of the applied theoretical approach to model cloud forma- tion in an oxygen-rich gas. The nucleation rate (forma- tion of seed particles) J∗ [cm−3 s−1] for homomolecular homogeneous nucleation (Helling & Woitke 2006) is J∗(t,r) = f (1,t) Z(N∗) (cid:18) τgr(1,N∗,t) (N∗ − 1)lnS(T )− exp (cid:19) , (1) ∆G(N∗) RT E. K. H. Lee et al.: Modelling the local and global cloud formation on HD 189733b Fig. 1. Illustration of the sample trajectories (black points) taken from the 3D radiative-hydrodynamic model atmosphere of HD 189733b (Dobbs-Dixon & Agol 2013); longitudes φ = 0◦. . .315◦in steps of ∆φ = +45◦, latitudes θ = 0◦, +45◦. The sub-stellar point is at φ = 0◦, θ = 0◦. We assume that the sample trajectories are latitudinal (north-south) symmetric. respectively; where ρgas(r) [g cm−3] is the local gas volume density and L0(r), L1(r) are derived by solv- ing for L j(r) [cm jg−1] in the dust moment equations (Woitke & Helling 2003). 2.2. 3D radiative-hydrodynamical model The 3D radiative-hydrodynamical [3D RHD] model solves the fully compressible Navier-Stokes equations coupled to a wavelength dependent two-stream radia- tive transfer scheme. The model assumes a tidally- locked planet with φ = 0◦, θ = 0◦denoting the sub- stellar point (the closest point to the host star). There are three components to the wavelength dependent opaci- ties: molecular opacities consistent with solar composi- tion gas, a gray component representing a strongly ab- sorbing cloud deck, and a strong Rayleigh scattering component. Equations are solved on a spherical grid with pressures ranging from ∼10−4.5 to ∼103 bar. In- put parameters were chosen to represent the bulk ob- served properties of HD 189733b, but (with the ex- ception of the opacity) were not tuned to match spec- troscopic observations. The dominate dynamical fea- ture is the formation of a super-rotating, circumplane- tary jet (Tsai et al. 2014) that efficiently advects energy from day to night near the equatorial regions. This jet forms from the planetary rotation (Rossby waves) cou- pled with eddies which pump positive angular momen- tum toward the equator (Showman & Polvani 2011). A counter-rotating jet is present at higher latitudes. Signif- icant vertical mixing, discussed later, is seen through- out the atmosphere. Calculated transit spectra, emission spectra, and light curves agree quite well with current observations from 0.3µm to 8.0µm. Further details can be found in Dobbs-Dixon & Agol (2013). (cid:18) 1− (cid:19) , where f(1,t) is the number density of the seed form- ing gas species, τgr the growth timescale of the critical cluster size N∗, Z(N∗) the Zeldovich factor, S(T) the su- persaturation ratio and ∆G(N∗) the Gibbs energy of the critical cluster size. We consider the homogeneous nu- cleation of TiO2 seed particles. A new value for the sur- face tension σ∞ = 480 [erg cm−2] is used based on up- dated TiO2 cluster calculations from Lee et al. (2015). The net growth/evaporation velocity χnet [cm s−1] of a grain (after nucleation) due to chemical surface re- actions (Helling & Woitke 2006) is χnet(r) = 3√36π ∑ s R ∑ r=1 r αr ∆Vrnrvrel νkey r 1 Sr 1 bs surf r (2) where 'r' is the index for the chemical surface reaction, ∆Vr the volume increment of the solid 's' by reaction r, nr the particle density of the reactant in the gas phase, vrel the relative thermal velocity of r the gas species taking part in reaction r, αr the stick- ing coefficient of reaction r and νkey the stoichiomet- ric factor of the key reactant in reaction r (Helling & Woitke 2006). Sr is the reaction supersaturation ra- tio and 1/bs surf = Vs/Vtot the volume ratio of solid s. Our cloud formation method allows the formation of mixed heterogeneous grain mantles. We simultaneously consider 12 solid growth species (TiO2[s], Al2O3[s], CaTiO3[s], Fe2O3[s], FeS[s], FeO[s], Fe[s], SiO[s], SiO2[s], MgO[s], MgSiO3[s], Mg2SiO4[s]) with 60 sur- face chemical reactions (Helling et al. 2008b). The local dust number density nd [cm−3] and mean grain radius (cid:104)a(cid:105) [cm] are given by nd(r) = ρgas(r)L0(r), (3) (cid:114) 3 3 4π (cid:104)a(cid:105)(r) = L1(r) L0(r) , 2.3. Model set-up and input quantities (4) We sampled vertical trajectories of the 3D RHD model at longitudes of φ = 0◦. . .315◦in steps of ∆φ = +45◦and Article number, page 3 of 23 NE(0◦,0◦)(45◦,0◦)(90◦,0◦)(270◦,0◦)(315◦,0◦)(0◦,45◦)(45◦,45◦)(90◦,45◦)(270◦,45◦)(315◦,45◦)DaysideSWNE(180◦,0◦)(225◦,0◦)(270◦,0◦)(90◦,0◦)(135◦,0◦)(180◦,45◦)(225◦,45◦)(270◦,45◦)(90◦,45◦)(135◦,45◦)NightsideSWA&A proofs: manuscript no. Modelling_the_local_and_global_cloud_formation_on_HD_189733b Fig. 2. 1D trajectory results from the 3D RHD HD 189733b atmosphere simulation used as input for the cloud formation model. Top Row: (Tgas, pgas) profiles in steps of longitude ∆φ = +45◦at equatorial, θ = 0◦and latitude, θ = 45◦profiles. Bottom Row: Smoothed vertical gas velocities vz [km s−1] in steps of longitude ∆φ = +45◦at equatorial, θ = 0◦and latitude, θ = 45◦profiles. Both latitudes show a temperature inversion on the dayside. The sub-stellar point is at φ = 0◦, θ = 0◦. Solid, dotted and dashed lines indicate dayside, day-night terminator and nightside profiles respectively. latitudes of θ = 0◦, +45◦(Fig. 1). The horizontal wind velocity moves in the positive φ direction. Figure 2 shows the (Tgas, pgas) and vertical velocity profiles used to derive the cloud structure at the equator θ = 0◦and lat- itude θ = 45◦. Local temperature inversions are present in the dayside (Tgas, pgas) profiles which are exposed to the irradiation of the host star without the need for an additional opacity source. The local temperature max- imum migrates eastward with increasing atmospheric depth. This is due to hydrodynamical flows funnelling gas towards the equator causing viscous, compressive and shock heating. The (Tgas, pgas) profile at latitude θ = 45◦has steeper temperature inversions on the dayside. Sample trajectories converge to equal temperatures at pgas > 1 bar for all longitudes and latitudes. We ap- ply a 3-point boxcar smoothing to the vertical veloc- ity profiles in order to reduce the effects of unresolved turbulence. Further required input quantities are a con- stant surface gravity of log g = 3.32 and initial solar ele- ment abundances ε0 x /εH (element 'x' to Hydrogen ratio) with C/O = 0.427 (Anders & Grevesse 1989) for all at- mospheric layers. However, we note the element abun- dance of the gas phase will increase or decrease due to cloud formation or evaporation (Fig. 8). We assume lo- cal thermal equilibrium (LTE) for all gas-gas and dust- gas chemical reactions. The required input quantities for the kinetic cloud formation model are summarised in Table 1. 2.4. Atmospheric vertical mixing Vertical mixing is important to resupply the upper at- mosphere with elements which have been depleted by cloud particle formation and their subsequent gravita- tional settling into deeper atmospheric layers (Woitke & Helling 2003). Without mixing, the cloud particles in the atmosphere would rain out and remove heavy ele- ments from the upper atmosphere. This leaves a metal poor gas phase where no future cloud particles could form (Woitke & Helling 2004; Appendix A). 2.4.1. Previous Brown Dwarf approach The main energy transport in the core of a brown dwarf is convection. The atmosphere is convectively unstable in the inner, hotter parts. This atmospheric Article number, page 4 of 23 -5-4-3-2-10123log10pgas[bar]40060080010001200140016001800200022002400Tgas[K]θ=0◦Longitudeφ0◦45◦90◦135◦180◦225◦270◦315◦-5-4-3-2-10123log10pgas[bar]40060080010001200140016001800200022002400Tgas[K]θ=45◦Longitudeφ0◦45◦90◦135◦180◦225◦270◦315◦-5-4-3-2-10123log10pgas[bar]-8-7-6-5-4-3-2-10log10vz[kms−1]θ=0◦Longitudeφ0◦45◦90◦135◦180◦225◦270◦315◦-5-4-3-2-10123log10pgas[bar]-8-7-6-5-4-3-2-10log10vz[kms−1]θ=45◦Longitudeφ0◦45◦90◦135◦180◦225◦270◦315◦E. K. H. Lee et al.: Modelling the local and global cloud formation on HD 189733b convection causes substantial overshooting into even higher, and radiation dominated parts (e.g. Ludwig et al. (2002)). Woitke & Helling (2003, 2004) define the mix- ing timescale in low mass stellar atmospheres as the time for convective motions vconv to travel a fraction of the pressure scale height, Hp(r) τmix,conv(r) = const· Hp(r) vconv(r) . (5) Helling et al. (2008b); Witte et al. (2009, 2011) represent values for vconv(r), in their 1D Drift-Phoenix model atmospheres, above the convective zone (defined by the Schwarzschild criterion) from inertially driven overshooting of ascending gas bubbles. 2.4.2. Previous Hot Jupiter approach On hot Jupiters, in contrast to Earth, Jupiter and Brown Dwarf atmospheres, the intensity of stellar irradiation can suppress convective motions down to very large pressures (Barman et al. 2001). Therefore turbulent dif- fusion is likely the dominant vertical transport mecha- nism. The use of the 3D RHD model results allows us to apply the local vertical velocity component, vz(r), in our cloud formation model. No additional assumptions are required. Hence, vz(r) is known for each atmospheric trajectory chosen, as visualised in Fig. 2. Some chemical models of hot Jupiters (e.g. Moses et al. (2011)) use an approximation of the vertical diffu- sion coefficient Kzz(r) [cm2 s−1] of the gaseous state. Kzz(r) = Hp(r)· vz(r). The diffusion timescale is then τmix,diff(r) = const· H2 p (r) Kzz(r) . (6) (7) Moses et al. (2011) apply the root-mean-square (rms) vertical velocities derived from global horizon- tal averages at each atmospheric layer to their chemi- cal models. This results in a horizontally homogenous Kzz(r) mixing parameter. Parmentier et al. (2013) de- rive a global mean value for Kzz(r) as a function of lo- cal pressure using a passive tracer in their global circu- lation model [GCM] of HD 189733b resulting in Kzz(r) = 107 p−0.65 (p [bar]). Using a global mean smooths away all horizontally dependent vertical velocity varia- tions. Any parameterisation of the vertical mixing will depend on the details of the underlying hydrodynamical structure. Parmentier et al. (2013) use a 3D primitive equation model where vertical hydrostatic equilibrium is assumed. The radiation hydrodynamic simulations performed by Dobbs-Dixon & Agol (2013), applied in this paper, solve the full Navier-Stokes equations and will produce larger vertical velocities compared to the primitive equations. In summary, vertical velocity may be substantially damped in models using the primitive equations. Both Eq. (6) and Parmentier et al. (2013) def- initions for Kzz(r) assume that the dominant mixing oc- curs in the vertical direction, possible mixing from hor- izontal flows are neglected. The time-scale comparison in Sect. 2.4.4 supports this assumption for cloud forma- tion processes. It is worthwhile noting that the idea of diffusive mixing originates from shallow water approx- imations as applied in solar system research where a 2D velocity field produces shear which creates a turbulent velocity component. Hartogh et al. (2005) outline how local wind shear and the hydrostatic gas pressure are used to represent a vertical mixing. Some arguments reinforce why vertical mixing is important: -- Regions with low vertical velocity can be replen- ished of elements by the horizontal winds from high vertical velocity regions in a 3D situation. -- Large Hadley cell circulation is present and here vertical velocities can be significant and element re- plenishment to the upper atmosphere may be more efficient. -- Vertical transport is a key mixing process on Earth which has been successfully applied to hot Jupiter chemical models (Moses et al. 2011; Venot et al. 2012; Agúndez et al. 2014). We use Eq. (7) (const = 1) as the 1D mixing timescale input for our kinetic dust formation model, with Eq. (6) as the definition of the diffusion coefficient. We adopt the local vertical velocities that result from the Dobbs-Dixon & Agol (2013) 3D RHD atmosphere sim- ulations for HD 189733b as the values for vz(r) (Fig. 2). A 3-point boxcar smoothing was applied to these ve- locities to reduce the effects of unresolved turbulence. The longitude dependent Kzz (φ, r) for latitudes θ = 0◦, 45◦is shown in Fig. 3 (first row) and the resulting verti- cal mixing timescales (second row). We include the Kzz relation from Parmentier et al. (2013) in Fig. 3 (dash- dotted line) for comparison. Their linear fit is approxi- mately one order of magnitude lower which is similar to the difference found in HD 189733b and HD 209458b chemical models (Agúndez et al. 2014) who compared the two Kzz(r) expressions for Showman et al. (2009) GCM simulations. The current approach aims to capture the unique vertical mixing and thermodynamic condi- tions at each trajectory, while also accounting for prac- tical modelling of atmospheric mixing. 2.4.3. Advective timescales An important timescale to consider is the charcteristic advective timescale which is a representative timescale for heat to redistribute over the circumference of the globe. The advective timescale is given by τadv(r) = πr vhoriz(r) , (8) Where r(z) is the radial height of the planet and vhoriz(r) the local gas velocity in the horizontal direc- tion (φ). This timescale gives an idea of how fast ther- modynamic conditions can change in the longitudinal direction at a particular height z in the atmosphere. Figure. 3 compares the advective timescale to the ver- tical diffusion timescale (Eq. (7)) and find that both time scales are of the same order of magnitude. This Article number, page 5 of 23 A&A proofs: manuscript no. Modelling_the_local_and_global_cloud_formation_on_HD_189733b Fig. 3. Top Row: Vertical diffusion coefficient Kzz(r) = Hp · vz(r) [cm2 s−1] applying the smoothed vertical velocities of the radiative-hydrodynamic HD 189733b model at latitudes θ = 0◦and θ = 45◦, as a function of pressure at ∆φ = +45◦longitude intervals. The Kzz(r) = 107p−0.65 (p [bar]) expression from Parmentier et al. (2013) is shown as dash-dotted lines. Middle Row: The vertical mixing timescales τmix,diff [s] derived from the HD 189733b radiative-hydrodynamic model results at latitudes θ = 0◦, θ = 45◦, as a function of pressure at ∆φ = +45◦longitude intervals. Bottom Row: Ratio of the mixing and advective timescales at θ = 0◦, 45◦respectively. The ratio of the timescales stays approximately equal at all pressures. Solid, dotted and dashed lines indicate dayside, day-night terminator and nightside profiles respectively. suggests that the convective/turbulent diffusion of up- ward gaseous material as described in Sect. 2.4.1 occurs approximately at the same timescales as gas advected across the globe. This suggests that, to a first approx- imation, the 1D (Tgas, pgas) trajectories used as input for the cloud formation model do not fluctuate rapidly enough to substantially influence the cloud properties. Article number, page 6 of 23 -5-4-3-2-10123log10pgas[bar]56789101112log10Kzz[cm2s−1]θ=0◦Kzz=107p−0.65Longitudeφ0◦45◦90◦135◦180◦225◦270◦315◦-5-4-3-2-10123log10pgas[bar]56789101112log10Kzz[cm2s−1]θ=45◦Kzz=107p−0.65Longitudeφ0◦45◦90◦135◦180◦225◦270◦315◦-5-4-3-2-10123log10pgas[bar]2345678910log10τmix,diff[s]θ=0◦Longitudeφ0◦45◦90◦135◦180◦225◦270◦315◦-5-4-3-2-10123log10pgas[bar]2345678910log10τmix,diff[s]θ=45◦Longitudeφ0◦45◦90◦135◦180◦225◦270◦315◦-5-4-3-2-10123log10pgas[bar]0.00.20.40.60.81.01.21.41.61.82.02.22.42.62.83.0τmix/τadvθ=0◦Longitudeφ0◦45◦90◦135◦180◦225◦270◦315◦-5-4-3-2-10123log10pgas[bar]0.00.20.40.60.81.01.21.41.61.82.02.22.42.62.83.0τmix/τadvθ=45◦Longitudeφ0◦45◦90◦135◦180◦225◦270◦315◦E. K. H. Lee et al.: Modelling the local and global cloud formation on HD 189733b Fig. 4. The vertical mixing, horizontal advection, cloud settling, cloud growth and cloud nucleation timescales at the sub-stellar point (Left) and φ = 180◦, θ = 0◦(Right). 2.4.4. Cloud formation timescales We compare the cloud particle settling, growth and nu- cleation timescales that result from our cloud model (Sect. 3) to the mixing and advection timescales that are derived from the hydrodynamic fluid field. The nucle- ation timescale τnuc is defined as τnuc = nd J(cid:63) , the growth timescale τgr by τgr = 3√36π(cid:104)a(cid:105) 3χnet , and the cloud particle settling timescale τsetl by τsetl = Hp (cid:104)vdr(cid:105) , (9) (10) (11) where (cid:104)vdr(cid:105) is the large Knudsen number frictional regime (Kn >> 1) mean drift velocity (Woitke & Helling 2003, Eq. 63). Figure 4 shows the timescales at the sub-stellar and anti-stellar points. Our results agree with earlier timescale analysis in Woitke & Helling (2003) who point out an hierarchical dominance of the cloud formation processes through the atmosphere. In the upper atmosphere nucleation is the most efficient process: Table 2. References for the (n, k) optical constants of the 12 different solids. Solid species Source TiO2[s] Al2O3[s] CaTiO3[s] Fe2O3[s] FeS[s] FeO[s] Fe[s] SiO[s] SiO2[s] MgO[s] MgSiO3[s] Mg2SiO4[s] Ribarsky in Palik (1985) Zeidler et al. (2013) Posch et al. (2003) Unpublished1 Henning et al. (1995) Henning et al. (1995) Posch et al. (2003) Philipp in Palik (1985) Posch et al. (2003) Palik (1985) Dorschner et al. (1995) Jäger et al. (2003) 2.5. Dust opacity Based on the results of the spatially varying cloud prop- erties we calculate the opacity of the cloud particles. We determine the cloud particle absorption and scat- tering coefficients in each atmospheric layer for wave- lengths 0.3µm, 0.6µm, 1.1µm, 1.6µm, 3.6µm, 4.5µm, 5.8µm, 8.0µm, 24.0µm, corresponding to the Hubble and Spitzer average band passes. Since the cloud parti- cles are made of mixed solids, the effective (n, k) optical constants is calculated for each particle using effective medium theory. We follow the approach of Bruggeman (1935) where τnuc (cid:46) τgr << τmix ∼ τadv << τsetl. Deeper in the atmosphere nucleation eventually be- comes inefficient and all timescales become comparable ∑ s (cid:18) Vs Vtot (cid:19) εs − εav εs + 2εav = 0, (12) τgr (cid:46) τmix ∼ τadv (cid:46) τsetl. The chemical processes that determine the cloud particle formation occur on shorter timescales than the large scale hydrodynamical timescales. This emphasises that the cloud particle formation is a local process. where Vs/Vtot is the volume fraction of solid species s, εs the dielectric function of solid species s and εav the average dielectric function over the total cloud particle volume. The average dielectric function is then found 1 http://www.astro.uni-jena.de/ Laboratory/OCDB/oxsul.html Article number, page 7 of 23 -5-4-3-2-10123log10pgas[bar]23456789log10τ[s]φ=0◦,θ=0◦τVertMixHorizAdvecSettlingGrowthNucl-5-4-3-2-10123log10pgas[bar]23456789log10τ[s]φ=180◦,θ=0◦τVertMixHorizAdvecSettlingGrowthNuclA&A proofs: manuscript no. Modelling_the_local_and_global_cloud_formation_on_HD_189733b by Newton-Raphson minimisation of Eq. (12). The scat- tering and extinction cross sections are then calculated using Mie theory for spherical particles (Mie 1908). We follow the approach of Bohren & Huffman (1983) where the scattering and extinction cross sections are defined as Csca(λ ,a) = Cext(λ ,a) = 2πa2 x2 2πa2 x2 ∞ ∑ n=1 (2n + 1)(an2 +bn2), ∞ ∑ n=1 (2n + 1)Re(an + bn), (13) (14) respectively; where x = 2πa/λ is the wavelength dependent size parameter. The scattering coefficients an and bn are calculated from the material optical k constant (Bohren & Huffman 1983). The wavelength- dependent absorption and scattering efficiency of a cloud particle is then Qsca(λ ,a) = Csca(λ ,a) πa2 , Qabs(λ ,a) = Cext(λ ,a) πa2 − Qsca, (15) (16) respectively. The total absorption and scattering ef- ficiency κ [cm2 g−1] can then be derived by multiplying the corresponding efficiencies with the area and occur- rence rate of each cloud particle. κsca(λ ,a) = Qsca(λ ,a)πa2nd/ρgas, κabs(λ ,a) = Qabs(λ ,a)πa2nd/ρgas. (17) (18) We include all 12 solid dust species in the opacity calculations. References for their optical constants are presented in Table 2. We assume that for all Mie calcu- lations the grain size is given by the mean grain radius (cid:104)a(cid:105) (Eq. 4) with number density nd (Eq. 3) for each at- mospheric layer. 3. Mineral clouds in the atmosphere of HD 189733b We apply the dust formation theory developed by Woitke & Helling (2003, 2004); Helling & Woitke (2006) and Helling et al. (2008b) in its 1D stationary form to 1D output trajectories of a 3D HD 189733b radiation-hydrodynamical [3D RHD] model, as de- scribed in Dobbs-Dixon & Agol (2013). In the follow- ing, we present local and global cloud structures for HD 189733b and discuss detailed results on cloud proper- ties such as grain sizes, material composition, element abundances, dust-to-gas ratio and C/O ratio. We inves- tigate general trends of the cloud structure as it varies throughout the atmosphere and make first attempts to study the cloud properties across the planetary globe. Article number, page 8 of 23 3.1. The cloud structure of HD 189733b at the sub-stellar point the local The substellar point (φ = 0◦, θ = 0◦) is the most directly irradiated point in atmospheres of hot Jupiters such as HD 189733b which is measured by observing before, during and after secondary transit (occulatation). We use this well defined location to provide a detailed description of the vertical cloud structure. We compare this atmospheric trajectory to other longitudes in Sec. 3.2 and to other latitudes in Sec. 3.3. Figure 5 shows that the cloud structure starts with the formation of seed particles (J∗ [cm−3 s−1] - nucleation rate) occurring at the upper pressure boundary of ∼10−4 bar. After the first surface growth processes occur on the seed particles, the cloud particles then gravitationally settle into the atmospheric regions below (toward higher density/pressure). Primary nucleation is efficient down to ∼10−2.5 bar where it drops off significantly, indicating that the local temperature is too high for further nucleation and that the seed forming species has been substantially depleted. The gas-grain surface chemical reactions that form the grain mantle (Eq. (2)) increase in rate as the cloud particles fall inward. This is due to the cloud particles encountering increasing local gas density, and therefore more condensible material is available to react with cloud particles. This surface growth becomes more efficient from ∼10−3. . .∼10−2 bar until temperature is so hot that the materials become thermally unstable and evaporate. The evaporation region results in a half magnitude decrease of the cloud particle sizes (negative χ) in the center region of the cloud. The relative volume fractions of the solid 's', Vs/Vtot, represents the material composition of the cloud particles. The cloud particle composition is dominated by silicates and oxides such as MgSiO3[s](∼27%), the upper Mg2SiO4[s](∼20%), SiO2[s](∼21%) at regions (cid:46)10−2.5 bar. Fe[s] contributes (cid:46)20% to the volume of the cloud particle in this region. The other 8 growth species (TiO2[s](∼0.03%), Al2O3[s](∼2%), CaTiO3[s](∼0.15%), Fe2O3[s](∼0.001%), FeS[s](∼1.6%), FeO[s](∼0.35%), SiO[s](∼0.05%), MgO[s](∼7%)) constitute the remaining grain volume. An evaporation region at ∼10−2.5 bar before the temperature maximum at ∼10−1.5 bar alters the grain composition dramatically. Al2O3[s] and Fe[s] dominate the grain composition in this region as the less stable silicates and oxides have evaporated from the grain surface. At the temperature maximum the grain is composed of Al2O3[s](∼80%) and Fe[s](∼15%). This suggests the presence of a cloud section more transparent than surrounding layers (Fig. 12). We address the issue of transparency in Sect. 5. Deeper in the atmosphere, a temperature inversion occurs (Fig. 2) starting from ∼10−1.5 bar. The temperature drops by ∼400 K from ∼10−1.5. . .∼10−0.5 bar. This temperature decrease allows a secondary nucleation region from ∼10−1. . .1 bar. The number density of grains jumps by ∼2 orders of magnitude at the secondary nucleation region, as a result of formation of many new cloud par- ticles. This causes a dip in the mean grain size at ∼10−1 E. K. H. Lee et al.: Modelling the local and global cloud formation on HD 189733b Fig. 5. HD 189733b's calculated cloud structure at the sub-stellar point (θ = 0◦, φ = 0◦). Left: 1st panel: Local gas temperature Tgas [K] (solid, left) and mixing timescale τmix [s] (dashed, right). 2nd panel: Nucleation rate J∗ [cm−3s−1] (solid, left) and dust number density nd [cm−3] (dashed, right). 3rd panel: Growth velocity of material s χ [cm s−1]. 4th panel: Volume fraction Vs/Vtot of solid s. 5th panel: Effective supersaturation ratio Seff of individual solids. 6th panel: Cloud particle mean radius (cid:104)a(cid:105) [µm] (solid, left) and mean drift velocity (cid:104)vdr(cid:105) [cm s−1] (dashed, right). Right: Key showing line-style and colour of our considered dust species. bar. Such secondary nucleation has not been seen in our cloud model results in Brown Dwarf nor non-irradiated giant gas planet atmospheres. Silicates and oxides are now thermally stable and once again form the grain mantle. The grain composition is approximately a 70-20-10 % mix of silicates and oxides, iron and other material respectively in this region, similar to the composition before the temperature maximum. At the cloud base, Fe[s] dominates the composition (∼35%) with MgO[s] and Mg2SiO4[s] making up ∼16% and ∼23% respectively. Table. 3 shows the percentage volume fraction Vs/Vtot of the 12 dust species at the sub-stellar point and the φ = 180◦, θ = 0◦ trajectory at 10−4, 10−3, 10−2, 10−1, 1 and 10 bar. Our results show that the entire vertical atmospheric range considered here (∼10−4.5. . .103 bar) is filled with dust. The exact properties of this dust such as size, com- position and number density change depending on the local thermodynamical state. The 3D RHD model does not expand to such low pressures and densities that the cloud formation processes becomes inefficient (Sect. 5). This suggests that clouds can be present at a higher and lower pressure than the present 3D RHD model bound- ary conditions allow. Article number, page 9 of 23 800100012001400160018002000220024002600T[K]−100−80−60−40−20020log10J∗[cm−3s−1]−300−300−200−200−100−10000100100200200300300χ[10−10cms−1]0.00.00.20.20.40.40.60.60.80.81.01.0Vs/Vtot0020020040040060060080080010001000Seff-5-4-3-2-10123log10pgas[bar]−3−2−10123log10hai[µm]3456789log10τmix[s]−8−6−4−20246log10nd[cm−3]−0.50.00.51.01.52.02.53.0log10hvdri[cms−1]Mg2SiO4MgSiO3MgOSiO2SiOFeFeOFeSFe2O3CaTiO3Al2O3TiO2A&A proofs: manuscript no. Modelling_the_local_and_global_cloud_formation_on_HD_189733b Fig. 6. Dust properties as a function of gas pressure for ∆φ = +45◦longitude intervals for latitudes θ = 0◦(Left column) and 45◦(Right column): Top Row: Nucleation rate J∗ [cm−3 s−1]. Second Row: Dust number density nd [cm−3]. Third Row: Net dust growth velocity χnet [cm s−1]. Bottom Row: Mean grain size (cid:104)a(cid:105) [µm]. Solid, dotted and dashed lines indicate dayside, day-night terminator and nightside profiles respectively. Article number, page 10 of 23 -5-4-3-2-10123log10pgas[bar]-30-25-20-15-10-505log10J∗[cm−3s−1]θ=0◦Longitudeφ0◦45◦90◦135◦180◦225◦270◦315◦-5-4-3-2-10123log10pgas[bar]-30-25-20-15-10-505log10J∗[cm−3s−1]θ=45◦Longitudeφ0◦45◦90◦135◦180◦225◦270◦315◦-5-4-3-2-10123log10pgas[bar]-9-8-7-6-5-4-3-2-101234567log10nd[cm−3]θ=0◦Longitudeφ0◦45◦90◦135◦180◦225◦270◦315◦-5-4-3-2-10123log10pgas[bar]-9-8-7-6-5-4-3-2-101234567log10nd[cm−3]θ=45◦Longitudeφ0◦45◦90◦135◦180◦225◦270◦315◦-5-4-3-2-10123log10pgas[bar]−0.50.00.51.01.52.02.53.0χnet[10−6cms−1]θ=0◦Longitudeφ0◦45◦90◦135◦180◦225◦270◦315◦-5-4-3-2-10123log10pgas[bar]−0.50.00.51.01.52.02.53.0χnet[10−7cms−1]θ=45◦Longitudeφ0◦45◦90◦135◦180◦225◦270◦315◦-5-4-3-2-10123log10pgas[bar]-3-2-101234log10hai[µm]θ=0◦Longitudeφ0◦45◦90◦135◦180◦225◦270◦315◦-5-4-3-2-10123log10pgas[bar]-3-2-101234log10hai[µm]θ=45◦Longitudeφ0◦45◦90◦135◦180◦225◦270◦315◦E. K. H. Lee et al.: Modelling the local and global cloud formation on HD 189733b Fig. 7. Top: Dust composition volume fraction Vs/Vtot as a function of pressure at θ = 0◦(Left) and θ = 45◦(Right) in ∆φ = +45◦longitude intervals. Grain composition is generally dominated by silicate-oxides including MgSiO3[s] apart from localised regions such as 10−2. . .10−1 bar at φ = 0◦, 45◦, θ = 0◦. Fe[s] dominates at 10−2. . .10−1 bar at φ = 0◦, θ = 45◦. Fe[s] grains remain at the cloud base as the most thermally stable element. Bottom: Key showing line-style and colour of our considered dust species. 3.2. Cloud structure changes with longitude (East-West) One of the main features of the radiation-hydrodynamic simulation is the equatorial jet structure which trans- ports heat over the entire 360◦longitude. The presence of this jet changes the thermodynamic structure of the atmosphere which affects the resulting global cloud structure. We sampled the 3D RHD results in longitude steps of ∆φ = +45◦at the equatorial region θ = 0◦to in- vestigate if cloud properties change significantly from dayside to nightside. Figure 2 shows that the nightside (Tgas, pgas) profiles can be ∼200 K cooler than the day- side in the upper atmospheric regions (∼10−4.5 . . .10−1 bar). Figure 6 shows the nucleation rate J∗ [cm−3 s−1] for dayside (solid lines), day-night terminator (dotted lines) and nightside (dashed lines) sample trajectories. The dayside nucleation rates fall off quicker with at- mospheric depth compared to the terminator and night- side profiles, particularly relevant for interpreting transit spectra. At longitude φ = 315◦ a secondary nucleation region emerges, similar to the sub-stellar point struc- ture (Fig. 5). The terminator and nightside nucleation regions extend further into the atmosphere to ∼10−1 bar. As a consequence of the different nucleation rates between dayside and nightside, the number density of cloud particles nd [cm−3] is greater on the nightside down to pressures of ∼10−1 bar. At this pressure, the secondary nucleation regions on the dayside, φ = 0◦, 315◦profiles spike up the number density comparable to values on the nightside. From Fig. 6 the net growth ve- locity χnet [cm s−1] shows that the most efficient growth Article number, page 11 of 23 0.00.20.40.60.80◦0.00.20.40.60.845◦0.00.20.40.60.890◦0.00.20.40.60.8135◦0.00.20.40.60.8180◦0.00.20.40.60.8225◦0.00.20.40.60.8270◦-5-4-3-2-101230.00.20.40.60.8315◦log10pgas[bar]Vs/Vtotφ(θ=0◦)0.00.20.40.60.80◦0.00.20.40.60.845◦0.00.20.40.60.890◦0.00.20.40.60.8135◦0.00.20.40.60.8180◦0.00.20.40.60.8225◦0.00.20.40.60.8270◦-5-4-3-2-101230.00.20.40.60.8315◦log10pgas[bar]Vs/Vtotφ(θ=45◦)Mg2SiO4MgSiO3MgOSiO2SiOFeFeOFeSFe2O3CaTiO3Al2O3TiO2A&A proofs: manuscript no. Modelling_the_local_and_global_cloud_formation_on_HD_189733b Table 3. Volume fraction Vs/Vtot[%] for the 12 solid species included in kinetic the cloud model. The first row of each species corresponds to the sub-stellar trajectory (φ = 0◦, θ = 0◦) cloud structure. The second row corresponds to the nightside trajectory φ = 180◦, θ = 0◦cloud structure. Note: The pressure at the cloud base is different for the two profiles. Pressure [bar] (cid:104)a(cid:105) [µm] Solid species TiO2[s] Al2O3[s] CaTiO3[s] Fe2O3[s] FeS[s] FeO[s] Fe[s] SiO[s] SiO2[s] MgO[s] MgSiO3[s] Mg2SiO4[s] 10−4 0.025 0.018 10−3 0.23 0.035 10−2 19.9 0.15 10−1 307 12.3 1 146 174 10 275 338 Cloud base 605 1088 0.03 0.04 2.43 2.06 0.16 0.10 0.26 0.03 13.99 2.42 0.87 0.17 0.05 0.02 3.68 2.44 0.24 0.17 0.02 0.03 0.01 0.06 2.84 2.11 2.49 2.06 0.22 0.13 0.20 0.07 0.10 >0.01 >0.01 >0.01 >0.01 >0.01 >0.01 >0.01 >0.01 9.69 0.02 0.09 14.44 0.02 0.10 12.12 8.87 0.07 0.02 0.02 7.63 0.03 22.49 45.00 7.87 21.75 21.14 4.52 2.42 0.70 3.82 5.20 0.11 8.84 20.60 17.90 7.88 20.33 21.26 9.92 8.87 11.10 7.73 7.21 5.40 9.35 15.79 22.47 8.96 15.25 22.02 21.74 25.31 11.09 15.93 16.15 27.31 25.39 9.68 0.22 12.12 0.06 7.63 21.08 4.52 0.06 9.03 20.08 10.13 5.13 7.28 24.11 20.71 26.64 16.70 0.05 0.05 0.02 0.02 24.29 21.29 0.48 0.44 19.94 22.18 8.05 6.51 20.77 21.82 21.41 26.59 1.22 0.03 60.98 2.13 3.56 0.14 0.07 0.03 14.45 0.03 7.93 30.93 8.50 0.08 1.12 0.75 19.35 1.97 7.61 0.14 22.41 0.30 16.23 0.22 0.24 10.03 10.03 0.64 0.77 >0.01 >0.01 0.03 0.03 0.09 0.09 87.15 86.80 0.92 0.99 0.12 0.13 0.79 0.89 >0.01 >0.01 0.02 0.04 regions for the grains is approximately 10−3. . .1 bar for the dayside and 10−2. . .1 bar for the nightside. Al- though the temperature of the local gas phase plays a role in increasing the growth rate, it is the increasing local density of material (as the particle falls thought the atmosphere) that is the dominating factor in deter- mining growth rates (Eq. (2); Fig. 8). Consequently, the mean grain size (cid:104)a(cid:105) [µm] shows a stronger increase from ∼10−4 . . .10−1 bar on the dayside than the night- side. The mean grain size dips at ∼10−1 bar for lon- gitudes φ = 0◦and 315◦due to the increase of grain number density as now the same number of gaseous growth species have to be distributed over a larger sur- face area. Figure 7 shows the volume fraction Vs/Vtot of the solid species 's'. The dust composition is gen- erally dominated by silicates and oxides (∼60 %) such as MgSiO3[s], Mg2SiO4[s] and SiO2[s] with (∼20 %) Fe[s] and FeO[s] content. At the cloud base the grain is primarily composed of Fe[s] (∼85%) and Al2O3[s] (∼10%). The φ = 45◦, θ = 0◦cloud structure contains an Al2O3[s] (∼60 %) and Fe[s] (∼30 %) dominant re- gion from ∼10−2. . .10−1 bar similar to the sub-stellar point structure. Table 3 summarises the percentage vol- ume fraction Vs/Vtot and the mean cloud particle size (cid:104)a(cid:105) of the 12 dust species at the sub-stellar point and the φ = 180◦, θ = 0◦trajectory at 10−4, 10−3, 10−2, 10−1, 1 and 10 bar pressures. 3.3. Cloud structure changes with latitude (North-South) At higher latitudes θ (cid:38)40◦the 3D RHD model produces a jet structure flowing in the opposite direction to the equatorial jet at dayside longitudes (Tsai et al. 2014). This significantly alters the (Tgas, pgas) and vertical ve- locity profiles (Fig. 2). These latitudes also contain the coldest regions of the nightside where vortexes easily form and dominate the atmosphere dynamics (Dobbs- Dixon & Agol 2013, Fig. 1). To investigate the cloud structure at these latitudes we repeated our trajectory sampling for latitudes of θ = 45◦in longitude steps of ∆φ = +45◦. Figure 6 shows the nucleation rate J∗ [cm−3 s−1] for dayside (solid lines), day-night terminator (dot- ted lines) and nightside (dashed lines) sample trajec- tories. The profiles are similar to the equatorial θ = 0◦regions with double nucleation peaks at φ = 0◦and 315◦. Again, there is an increase in number density nd [cm−3] (Fig. 6) from 10−1. . .1 bar due to second nu- cleation regions at φ = 0◦and 315◦. The growth ve- locity χnet is generally an order of magnitude lower on the dayside than the equatorial regions. This results in higher latitude clouds containing smaller grain sizes compared to their equatorial counterparts. Article number, page 12 of 23 E. K. H. Lee et al.: Modelling the local and global cloud formation on HD 189733b Fig. 8. Gas phase element abundances εx as a function of pressure at θ = 0◦(Left) and θ = 45◦(Right) in ∆φ = +45◦longitude intervals. We consider 8 elements: Mg(orange), Si(maroon), Ti(blue), O(red), Fe(green), Al(cyan), Ca(purple), S(magenta) that constitute the growth species. Horizontal black lines indicate solar abundance ε0 x . A decrease in element abundance indicates condensation of growth species onto cloud particle surface. An increase indicates evaporation of molecules constituted of that element from the cloud particle surface. Dayside profiles (solid) are φ = 0◦, 45◦, 315◦, θ = 0◦, 45◦. Day-night terminator profiles (dashed) are φ = 90◦, 270◦, θ = 0◦, 45◦. Nightside profiles (dotted) are φ = 135◦, 180◦, 225◦, θ = 0◦, 45◦. 3.4. Element depletion, C/O ratio and dust-to-gas ratio The cloud formation process strongly depletes the local gas phase of elements, primarily through extremely effi- cient solid surface growth processes. We consider the 8 elements that constitute the solid materials of the cloud particles, Mg, Si, Ti, O, Fe, Al, Ca and S assuming an initial solar element abundance (ε0 x ) for all layers. Fig- ure 8 shows the elemental abundance εx (ratio to Hydro- gen abundance) of each element as a function of pres- sure at each of the sample trajectories. Depletion occurs due to the formation of solids made of the these ele- ment onto the cloud particle surface and by nucleation of new cloud particles. Increase in element abundances correspond to regions of solid material evaporation. Ti is depleted at the upper boundary due to immediate effi- cient nucleation. For dayside profiles φ = 0◦, 45◦, 315◦, θ = 0◦, 45◦, from 10−4.5. . .10−3 bar, Mg, Ti, Si, Al and Fe are depleted by ∼1 order of magnitude while O, S and Ca are depleted by ∼ 10%. These profiles return to initial solar abundance values at ∼10−2 bar where the solid material from the grain surface evaporates, re- turning elements to the gas phase. O, Fe, Si S and Mg abundance can slightly overshoot solar abundance val- ues at their respective maximums. Elements are most heavily depleted in these profiles at ∼10−1 bar where the most efficient surface growth occurs. Mg, Si, and Fe are depleted by ∼3 orders of magnitude, Ti by 8 orders of magnitude and Al by 5 orders of magnitude. O, S and Ca are again depleted by ∼ 10%. Fe, Al, S and Ti return to solar abundance or slightly sub-solar abundance at the cloud base, where all materials have evaporated. The cloud base is enriched in O, Ca, Mg and Si which are ∼50% above solar abundance values. For nightside and day-night terminator profiles φ = 90◦, 135◦, 180◦, 225◦, 270◦θ = 0◦, 45◦, from 10−4.5. . .10−2 bar, Mg, Ti, Si, Al and Fe are depleted by 4 to 8 or- ders of magnitude while O and S are depleted by ∼1 order of magnitude and Ca by ∼10%. The φ = 90◦and 135◦, θ = 0◦, 45◦show a return to near initial abun- dance at ∼10−0.5 bar from material evaporation. Other nightside/terminator profiles gradually return to initial abundance from ∼1 bar to their respective cloud bases. Again, O, Ca, Mg and Si are slightly above solar abun- dance at the cloud base. We calculate the dust-to-gas ratio and the C/O ra- tio of our cloud structures. Figure. 9 shows the local dust-to-gas ratio of the sample trajectories at latitudes θ = 0◦and 45◦respectively. Dayside profiles show in- creases and decreases in dust-to-gas ratio correspond- ing to regions of nucleation/growth and evaporation. Nightside profiles show less cloud particle evapora- tion throughout the upper atmosphere, with only small changes in the dust-to-gas ratio which starts to drop off from ∼10−2 bar. Figure. 9 shows the local gaseous Article number, page 13 of 23 −14−13−12−11−10−9−8−7Ti−3.20−3.15−3.10−3.05−3.00O−5.66−5.65−5.64−5.63−5.62Ca−5.8−5.6−5.4−5.2−5.0−4.8−4.6S−11−10−9−8−7−6Al−9−8−7−6−5Fe−10−9−8−7−6−5Si-5-4-3-2-10123−10−9−8−7−6−5−4Mglog10pgas[bar]log10xθ=0◦−16−14−12−10−8Ti−3.20−3.15−3.10−3.05−3.00−2.95O−5.67−5.66−5.65−5.64−5.63−5.62Ca−5.8−5.6−5.4−5.2−5.0−4.8S−13−12−11−10−9−8−7−6Al−11−10−9−8−7−6−5Fe−12−11−10−9−8−7−6−5Si-5-4-3-2-10123−10−9−8−7−6−5−4Mglog10pgas[bar]log10xθ=45◦A&A proofs: manuscript no. Modelling_the_local_and_global_cloud_formation_on_HD_189733b Fig. 9. Gas properties as a function of gas pressure for ∆φ = +45◦longitude intervals from the cloud formation process. Top Row: Dust-to-gas ratio ρd/ρgas at latitudes θ = 0◦and 45◦. Bottom Row: C/O ratio at latitudes θ = 0◦and 45◦. Solid, dotted and dashed lines indicate dayside, day-night terminator and nightside profiles respectively. The horizontal black line indicates solar C/O ratio. Regions of decreasing ρd/ρgas and C/O indicate cloud particle evaporation. C/O ratio of our sample trajectories at latitudes θ = 0◦and 45◦respectively. These follow similar trends to the dust-to-gas ratio. The C/O ratio lowers where evap- oration of cloud particles releases their oxygen baring materials, replenishing the local gas phase, The abun- C = 10−3.45 (solar abun- dance of C is kept constant at ε0 dance) and is not affected by the formation of cloud particles in our model. Dayside equatorial profiles show C/O ratio dips by ∼5% below solar values at pressures of 10−2.5. . .10−1 bar. The φ = 0◦, θ = 45◦profile also shows a dip below solar values at similar pressure lev- els. Apart from these localised regions of oxygen re- plenishment, the C/O ratio remains above solar values for the majority of the atmospheric profiles; except from the cloud base, which is enriched with oxygen by 10%- 20% for all profiles. 3.5. Cloud property maps of HD 189733b Global cloud property maps of the atmosphere of HD 189733b enable the comparison between different at- mospheric regions as a whole. This has implications for interpreting cloudy observations which sample different atmospheric regions. To produce a global cloud map of the atmosphere of HD 189733b we bi-cubically inter- polate our cloud structure results across longitudes φ = 0◦. . .360◦and latitudes θ = 0◦. . .80◦. We include sample trajectories at latitude θ = 80◦to interpolate to higher latitudes. We assume latitudinal (north-south) symme- try as thermodynamic conditions do not vary signifi- cantly from positive to negative latitudes (Dobbs-Dixon & Agol 2013). Figures 10 and 11 show maps of the mean grain radius (cid:104)a(cid:105) [µm] and grain number density nd [cm−3] at 10−4 bar, 10−3, 10−2, 10−1 bar, 1 bar and 10 bar. The mean grain radius and number den- sity (along with material composition) are key values in calculating the wavelength dependent opacity of the clouds. At all pressures there is a contrast between the dayside and nightside mean grain radii. In the upper at- mosphere (< 10−1 bar) the globally largest cloud parti- cles occur on the dayside face. In the deeper atmosphere (> 10−1 bar) global maximum of the cloud particle size may also occur in nightside regions. The difference be- tween the largest and smallest grain radii at each pres- sure can range from 1 to 2 orders of magnitude. The highest difference in cloud particle sizes occurs at 10−2 bar where grains on the dayside are 2 orders of mag- nitude bigger than the nightside. This is due to dayside Article number, page 14 of 23 -5-4-3-2-10123log10pgas[bar]0.0000.0020.0040.0060.0080.010ρd/ρgasθ=0◦Longitudeφ0◦45◦90◦135◦180◦225◦270◦315◦-5-4-3-2-10123log10pgas[bar]0.0000.0020.0040.0060.0080.010ρd/ρgasθ=45◦Longitudeφ0◦45◦90◦135◦180◦225◦270◦315◦-5-4-3-2-10123log10pgas[bar]0.30.40.50.60.7C/Oratioθ=0◦Longitudeφ0◦45◦90◦135◦180◦225◦270◦315◦-5-4-3-2-10123log10pgas[bar]0.30.40.50.60.7C/Oratioθ=45◦Longitudeφ0◦45◦90◦135◦180◦225◦270◦315◦E. K. H. Lee et al.: Modelling the local and global cloud formation on HD 189733b Fig. 10. Mean grain radius (cid:104)a(cid:105) [µm] interpolated across the globe (assuming latitudinal (north-south) symmetry) of HD 189733b. Top Row: 10−4 bar, 10−3 bar. Middle Row: 10−2 bar, 10−1 bar. Bottom Row: 1 bar, 10 bar, respectively. The sub-stellar point is at φ = 0◦, θ = 0◦. The largest mean grain radii of cloud particles are generally found on the dayside at all pressure levels. Consequently, cloud particles on the dayside are larger than their pressure level counterparts on the nightside. Note: each plot contains a different colour scale. profiles undergoing very efficient cloud particle surface growth at ∼10−2 bar, while nightside cloud particles do not grow as efficiently (Fig. 6). There are also steep gra- dients between the maximum and minimum mean grain radius (indicating very rapid cloud formation processes) which occur at the terminator regions (φ = 90◦, 270◦) at pressure profiles (cid:46)10−1 bar. The number density maps show similar differences between dayside and nightside profiles but with maximum values of number density generally occurring on the nightside of the planet at each pressure level. Steep gradients in number density are also present at the terminator regions. These results, taken as a whole, suggest that the wavelength depen- dent dust opacity significantly varies between the day- side and nightside. The mixed local composition of the cloud particles will also have an effect on the dust opac- ity. 4. Cloud opacities We calculated the scattering and absorption properties of our cloud particles and produce global cloud opac- ity maps of the HD 189733b atmosphere. For an il- lustration of the results, we present cloud opacity cal- Article number, page 15 of 23 A&A proofs: manuscript no. Modelling_the_local_and_global_cloud_formation_on_HD_189733b Fig. 11. Number density of cloud particles nd [cm−3] interpolated across the globe (assuming latitudinal (north-south) symme- try) of HD 189733b. Top Row: 10−4 bar, 10−3 bar. Middle Row: 10−2 bar, 10−1 bar. Bottom Row: 1 bar, 10 bar, respectively. The sub-stellar point is at φ = 0◦, θ = 0◦. The maximum number of cloud particles occurs on the nightside of the planet. Note: each plot contains a different colour scale. culations at longitude φ = 0◦and 180◦at the equato- rial region θ = 0◦. Figure 12 demonstrates that the efficiency of light extinction depends on wavelength. Bluer light is more heavily scattered/absorbed in the upper atmosphere compared to the infrared. For the φ = 0◦trajectory the region of highest extinction ex- tends from ∼10−4. . .10−2 bar. The φ = 180◦trajectory extends from ∼10−4. . .1 bar. The fraction of scatter- ing to absorption is shown in Fig. 12. Absorption of the stellar light is the most efficient extinction mecha- nism in the upper atmosphere (∼10−3 bar) while scat- tering dominates the extinction in the deeper atmo- sphere (cid:38)10−2.5 bar. There is a large increase in the scattering component corresponding to the Fe[s] rich (∼50%) grain regions at 10−2 and 10−1 bar in the φ = 0◦, θ = 0◦trajectory. Between these Fe[s] regions a ∼80% Al2O3[s] grain composition region occurs which is more transparent in our wavelength range than the Fe[s] surrounding regions. For a discussion of these re- sults, we refer to Sect. 5. From the upper atmosphere down to ∼10−3 bar the extinction profile follows an ab- sorption profile of κext ∝ λ−2. A transition region from ∼10−2 to 10−1 bar occurs where profiles gradually flat- ten from optical to infrared wavelengths. This is due to the wavelength dependent extinction efficiency of cloud particles with blue light more absorbed/scattered by Article number, page 16 of 23 E. K. H. Lee et al.: Modelling the local and global cloud formation on HD 189733b Fig. 12. Dust opacities κ [cm2 g−1] for the cloud structure at φ = 0◦, 180◦, θ = 0◦as a function of pressure. Top Row: Total extinction κext. Second Row: Absorption κabs. Third Row: Scattering κsca. Bottom Row: Scattering to absorption ratio κsca/κabs. Bluer wavelengths are absorbed/scattered more efficiently than redder wavelengths in the upper atmosphere. The absorption dominates the total extinction in the upper regions and scattering in the deeper atmosphere. The spikes in scattering ratio at 10−2 and 10−1 bar correspond to Fe rich grain composition. Article number, page 17 of 23 -5-4-3-2-10123log10pgas[bar]051015202530354045505560κext[cm2g−1]φ=0◦,θ=0◦λ0.3µm0.6µm1.1µm1.6µm3.6µm4.5µm5.8µm8.0µm24.0µm-5-4-3-2-10123log10pgas[bar]051015202530354045505560657075808590κext[cm2g−1]φ=180◦,θ=0◦λ0.3µm0.6µm1.1µm1.6µm3.6µm4.5µm5.8µm8.0µm24.0µm-5-4-3-2-10123log10pgas[bar]0510152025303540κabs[cm2g−1]φ=0◦,θ=0◦λ0.3µm0.6µm1.1µm1.6µm3.6µm4.5µm5.8µm8.0µm24.0µm-5-4-3-2-10123log10pgas[bar]051015202530354045κabs[cm2g−1]φ=180◦,θ=0◦λ0.3µm0.6µm1.1µm1.6µm3.6µm4.5µm5.8µm8.0µm24.0µm-5-4-3-2-10123log10pgas[bar]0510152025κsca[cm2g−1]φ=0◦,θ=0◦λ0.3µm0.6µm1.1µm1.6µm3.6µm4.5µm5.8µm8.0µm24.0µm-5-4-3-2-10123log10pgas[bar]051015202530354045κsca[cm2g−1]φ=180◦,θ=0◦λ0.3µm0.6µm1.1µm1.6µm3.6µm4.5µm5.8µm8.0µm24.0µm-5-4-3-2-10123log10pgas[bar]05101520κsca/κabsφ=0◦,θ=0◦λ0.3µm0.6µm1.1µm1.6µm3.6µm4.5µm5.8µm8.0µm24.0µm-5-4-3-2-10123log10pgas[bar]05101520κsca/κabsφ=180◦,θ=0◦λ0.3µm0.6µm1.1µm1.6µm3.6µm4.5µm5.8µm8.0µm24.0µmA&A proofs: manuscript no. Modelling_the_local_and_global_cloud_formation_on_HD_189733b Fig. 13. Dust extinction κext [cm2 g−1] pressure isobars at φ = 90◦(Left), 270◦(Right), θ = 0◦(equatorial dayside-nightside terminator regions). Triangles denote the values from Dobbs-Dixon & Agol (2013) assumed additional Rayleigh slope opacity κext ∝ λ−4. Isobars follow an absorption profile κext ∝ λ−2 at the upper atmosphere (cid:46)10−1 bar. Isobars flatten deeper than (cid:38)10−1 bar. small, upper atmosphere grains and red light by larger, deeper atmosphere grains. In addition, the grain com- position fraction of highly opaque solid species such as Fe[s] increases from ∼10% to ∼20% deeper in the at- mosphere (>10−1.5 bar). A wavelength dependent 'cloud opacity' global map can be produced by interpolation assuming latitudinal (north-south) symmetry of the cloud properties sample trajectories. We choose the 0.6µm and 5.8µm wave- lengths as representative of optical and infrared wave- length extinction respectively. Figures 14 and 15 show the bi-cubic interpolated cloud particle extinction ef- ficiency at 0.6µm and 5.8µm wavelengths across the globe from 10−4. . .10 bar. The optical 0.6µm cloud map shows the maximum extinction regions migrate from dayside to nightside regions with increasing pressure. The 5.8µm infrared cloud map shows the maximum ex- tinction region also undergoes a shift, from nightside at 10−4 bar, to dayside at 10−3. . .10−2 bar, to nightside for <10−1 bar. The most efficient extinction region deeper in the atmosphere (10−1. . .10 bar) for both maps occurs at longitudes φ ∼225◦. . .315◦which correspond to the coldest parts of the atmosphere (Fig. 2). 4.1. Reflection and sparkling of cloud particles The visible appearance of our cloud particles can be estimated from their scattering properties. By estimat- ing the relative fraction of scattered light in red, blue and green colour wavelengths, a rough RGB scale can be constructed to visualise a sparkling/reflection colour. We use the κsca results from Sect. 4 for cloud layers at upper regions in the atmosphere ∼10−4.5. . .10−2 bar at the sub-stellar point φ = 0◦, θ = 0◦. This profile was used as being best comparable to observations from sec- ondary transit (occultation) observations in which the albedo (or colour) of the atmosphere can be determined (Evans et al. (2013)). We linearly interpolated the κsca to proxy RGB wavelengths and calculated their relative red, blue and green scattering fractions. This results in a deep midnight blue colour. Deeper in the atmosphere, >10−3 bar, cloud particles scatter red, blue and green light in more equal fractions which results in redder and grayer cloud particle appearance. Helling & Rietmei- jer (2009) suggest that amorphous cloud particles can re-arrange themselves into crystalline lattice structures as they gravitationally settle. This would allow light to refract and reflect inside the cloud particle volume pro- ducing a sparkle. The sparkling colour is likely to be a similar colour to the reflected light. 4.2. Reflectance of cloud particles Cloud particles have a large effect on the observable properties by reflecting incident light on the atmosphere back into space. We estimate the reflectance of the cloud particles by calculating the pressure dependent single scattering Albedo ω0 (Bohren & Clothiaux 2006) of the cloud particles defined as ω0 = κsca κabs + κsca (19) This ratio indicates where the cloud particles extinc- tion is dominated by scattering (∼1) or absorption (∼0). Figure 16 shows maps of the calculated single scatter- ing Albedo for 8µm at 10−1 bar and 1 bar. The maxi- mum of the reflectance occurs in the approximate longi- tude range 0◦. . .135◦at the equatorial region. From these maps we expect the peak of the 8µm albedo to occur from 20◦. . .40◦East of the sub-stellar point. 5. Discussion Our results strongly support the idea that the Hubble and Spitzer observations of HD 189733b described in Lecavelier Des Etangs et al. (2008); Sing et al. (2011); Gibson et al. (2012) and Pont et al. (2013) can be plau- sibly explained by the presence of cloud particles in the atmosphere. The cloud composition is dominated Article number, page 18 of 23 10−1100101102λ[µm]10−210−1100101102κext[cm2g−1]φ=90◦,θ=0◦xpgas[bar]<10−310−3<x<10−110−1<x<1>110−1100101102λ[µm]10−210−1100101102κext[cm2g−1]φ=270◦,θ=0◦xpgas[bar]<10−310−3<x<10−110−1<x<1>1E. K. H. Lee et al.: Modelling the local and global cloud formation on HD 189733b Fig. 14. Dust κext [cm2 g−1] opacity for 0.6 µm interpolated across the globe of HD 189733b assuming latitudinal (north- south) symmetry. Top Row: 10−4 bar, 10−3 bar. Middle Row: 10−2 bar, 10−1 bar. Bottom Row: 1 bar, 10 bar, respectively. We take 0.6 µm as representative of the optical wavelength extinction due to cloud. The sub-stellar point is at φ = 0◦, θ = 0◦. The maximum extinction efficiency shifts from the dayside of the planet to the nightside with increasing depth. Deep in the atmosphere the most opaque region remains at ∼225◦. . .315◦longitudes. Note: The colour scale for each plot is different by a silicate-oxide-iron mix (Fig. 7) which supports the suggestion of a MgSiO3[s] dominated cloudy at- mosphere by Lecavelier Des Etangs et al. (2008). They suggest a MgSiO3[s] dominated grain composition with sizes ∼10−2. . .10−1 µm at pressures of 10−6. . .10−3 bar. Their grain size estimate is based on finding that Rayleigh scattering fits their observed feature-free slop in the optical spectral range. This analysis is consis- tent with our detailed model results which produce grains of ∼20% MgSiO3[s] composition with sizes ∼10−2. . .10−1 µm at ∼10−4.5. . .10−3 bar, for all atmo- spheric profiles in this region. However, Fig. 13 shows that the isobars of the extinction κext at the day-night terminator regions follow an absorption profile slope in the upper atmosphere which transitions into a flat profile deeper in the atmosphere. The absorption dom- inated profile from 10−4. . .10−3 bar does not support the Rayleigh scattering parameterisations for the obser- vations of HD 189733b. This does not rule out cloud particles as the source of the Rayleigh slope. A possi- bility is that many smaller grains form at lower pressure regions than the boundary applied. This would increase the scattering component of the cloud opacity. Further- more, in a 3D atmosphere, small grains may be more Article number, page 19 of 23 A&A proofs: manuscript no. Modelling_the_local_and_global_cloud_formation_on_HD_189733b Fig. 15. Dust κext [cm2 g−1] opacity for 5.8 µm interpolated across the globe of HD 189733b assuming latitudinal (north-south) symmetry. Top Row: 10−4 bar, 10−3 bar. Middle Row: 10−2 bar, 10−1 bar. Bottom Row: 1 bar, 10 bar, respectively. We take 5.8 µm as representative of the infrared extinction due to cloud. The sub-stellar point is at φ = 0◦, θ = 0◦. The maximum extinction efficiency shifts from the dayside of the planet to the nightside with increasing depth. Deep in the atmosphere the most opaque region remains at ∼225◦. . .315◦longitudes. Note: The colour scale for each plot is different. easily lofted than larger grains in regions of upward gas flow. Our static 1D model does not include the 3D ef- fects of a dynamic, changing atmosphere. Section 3.4 shows the effect that cloud particles have on local element abundances. Depletion of elements oc- cur when they are consumed by cloud particle formation and replenishment occurs when these cloud particles evaporate. In our cloud model the elements that consti- tute the main cloud particle composition were depleted by 3 orders of magnitude or more from ∼10−4.5. . .10−2 bar. These results suggest that the presence of cloud particles would generally flatten the spectral signatures from these elements and associated molecules in the atmosphere. The evaporation of molecules from cloud particles transports elements from upper to lower atmo- spheric height. This suggests that elements thought to contribute to upper atmosphere thermal inversion layers such as Ti (more specifically TiO; Fortney et al. (2008); Spiegel et al. (2009); Showman et al. (2009)) are trans- ported to the deeper atmosphere by cloud particles. The depletion and movement of TiO from the upper to lower atmosphere could reduce the intensity of a stratospheric inversion and/or move it deeper into the atmosphere or destroy it completely. Figure 9 shows that regions of the Article number, page 20 of 23 E. K. H. Lee et al.: Modelling the local and global cloud formation on HD 189733b Fig. 16. Interpolated single scattering albedo ω0 (Eq. (19)) at a wavelength of 8µm at 10−1 bar (Left), 1 bar (Right) for our cloud properties. The maximum of the reflectance occurs approximately from φ = 20◦. . .135◦longitude. These maximum reflectance regions are consistent with the 8µm Spitzer flux maps of HD 189733b from Knutson et al. (2007). Note: the sub-stellar point in the diagrams is located at φ = 0◦, θ = 0◦, at the left side of the figures. Fig. 17. (Tgas, pgas) profiles for Drift-Phoenix atmosphere models (coloured, solid) for log g = 3, Teff = 1000, 1500, 2000 K and the Dobbs-Dixon & Agol (2013) RHD model at φ = 0◦, 90◦, 180◦, 270◦, θ = 0◦(black, styled). Downward and upward facing triangles denote the cloud deck and base respectively. The Drift-Phoenix profiles show that cloud formation begins at ∼10−11 bar. atmosphere can be oxygen poor or rich depending on the cloud processes. This significantly alters the C/O ra- tio, where it is increased by the growth of oxygen bar- ing solid materials onto cloud particle surfaces and de- creased when these materials evaporate back into the gas phase. This could have an impact on interpreting observations of over/under abundant C/O ratios in exo- planet atmospheres as well as altering the local chem- istry in oxygen depleted regions. The opacity of the clouds will strongly influence the spectral signature from the atmosphere by absorb- ing or scattering various wavelengths at different effi- ciency. From Fig. 12 optical wavelengths are preferen- tially absorbed and scattered in the upper atmosphere. We therefore suggest that mineral clouds are responsi- ble for the planets observed albedo and deep blue colour suggested by Berdyugina et al. (2011) and Evans et al. (2013). The migration of the maximum efficiency of the cloud extinction from dayside to nightside (Fig. 14, 15) at 0.6µm and 5.8µm shows a strong dayside-nightside opacity contrast. The terminator regions at longitude φ = 90◦and 270◦also show differences in dust extinc- tion efficiency, especially deeper in the atmosphere. Dif- ferent thermodynamic conditions result in mean grain Article number, page 21 of 23 -12-11-10-9-8-7-6-5-4-3-2-10123log10pgas[bar]02004006008001000120014001600180020002200240026002800300032003400Tgas[K]CloudbaseClouddeckHD189733bRHD-θ=0◦φ=0◦φ=90◦φ=180◦φ=270◦Drift-Phoenix-logg=3Teff=1000KTeff=1500KTeff=2000KA&A proofs: manuscript no. Modelling_the_local_and_global_cloud_formation_on_HD_189733b sizes, compositions and opacity that is different at each terminator region. This has consequences for interpret- ing transmission spectroscopy measurements since ob- servations of opposite limbs of the planet would have different extinction properties. In addition, the cloud maps also show steep gradients of cloud properties at the terminator regions. Transit spectroscopy observa- tions that measure these regions would sample a vari- ety of cloud particle number density, sizes and distri- butions dependent on wavelength. In Fig. 7 the profiles at φ = 0◦and 45◦show a region compositionally domi- nated by Al2O3[s] cloud particles from ∼10−2. . .10−1 bar with Fe[s]-rich grains on above and below this re- gion. The effect of this Al2O3[s] region is to produce a locally lower cloud opacity layer flanked by high opac- ity regions (Fig. 12). This would have a significant ef- fect on radiation propagation through the atmosphere as the Fe[s] rich grains could shield photons reaching the Al2O3[s] layer from above or below. Our qualitative calculation of the cloud contribution to the albedo provides yet another insight into the obser- vational consequences of our cloud modelling. Our cal- culated single scattering albedo for the 8µm band (Fig. 16) show that the maximum reflectance occurs approxi- mately in the longitude range 0◦. . .135◦at the equator at 10−1 and 1 bar. With the peak occurring from 20◦. . .40◦. These maximum reflectance regions reproduce the ar- eas of maximum 8µm flux map from Knutson et al. (2007). This suggests that cloud particles influence ob- served phase curves of hot Jupiters depending on the position and reflectivity of cloud particles in the atmo- sphere. These effects have been observed for hot Jupiter Kepler 7-b (Demory et al. 2013), where Kepler obser- vations showed a westward shift in optical phase curves (GCM/RHD models predict an eastward shift). Spitzer phase curves showed that the shift was non-thermal in origin. This shift was attributed to non-uniform, longi- tude dependent optical scattering from clouds. Our 8µm reflectance result, albeit qualitatively, show that clouds can contribute to the infrared flux observed from the exoplanet atmosphere. It may not be simple to disen- tangle contributions by thermal emission and scatter- ing/reflection clouds from infrared observational prop- erties. Figure. 17 shows a (Tgas, pgas) structure comparison between Drift-Phoenix (log g = 3, Teff = 1000, 1500, 2000 K) and the Dobbs-Dixon & Agol (2013) 3D RHD model. Comparing trajectories from the 3D RHD model and the 1D Drift-Phoenix models suggests that clouds could be more extended into the lower-pressure regions than our present results show. Previous studies (e.g. Drift-Phoenix; Witte et al. (2011), Woitke & Helling (2004)) involving cloud formation, the lower pressure boundary can be up to ∼10−12 bar with cloud forma- tion occurring from ∼10−11 bar, these studies contain a smoother non-cloud to cloudy upper atmospheric re- gion than our present results. The cloud deck in the Drift-Phoenix models starts from ∼10−11 bar, 7 orders of magnitude lower that the RHD upper boundary pres- sure. Therefore, our results do not capture cloud forma- tion outside the RHD model boundary conditions yet. We also suggest that cloud formation can continue fur- ther inward than the results presented here. Due to the increasing density and high pressure, cloud particle ma- terial remains thermally stable until considerably higher temperatures and survives deeper into the atmosphere. We probed a 3D hot Jupiter atmospheric structure, irradiated by a host star, through selecting 16 1D at- mospheres trajectories along the equator and latitude θ = 45◦. We aimed to capture the main features of a dynamic atmosphere like east-west jet streams, latitu- dinal differences and dayside-nightside differences. We used these thermodynamic and velocity profiles to study cloud formation in order to present consistently calcu- lated cloud properties for HD 189733b. Our ansatz does not yet include a self-consistent feedback onto the (Tgas, pgas, vgas) structure due to radiative transfer processes. We note that due to an additional wavelength depen- dent dust opacity term in the radiative transfer scheme of Dobbs-Dixon & Agol (2013), this feedback is al- ready accounted for to a significant degree in the cur- rent (Tgas, pgas) profiles. Furthermore, we do not in- clude non-LTE kinetic gas-phase chemistry such as pho- tochemistry. Departures from LTE have been detailed in non-equilibrium models of HD 189733b's atmosphere performed by Moses et al. (2011); Venot et al. (2012); Agúndez et al. (2014) using thermodynamic input from the Showman et al. (2009) global circulation model. 6. Conclusion and summary We have presented the first spatially varying kinetic cloud simulation of the irradiated hot Jupiter exoplanet HD 189733b. We applied a 2-model approach with our cloud formation model using 1D thermodynamic input from a 3D radiation-hydrodynamic simulation of HD 189733b. Our results suggest that HD 189733b has a significant cloud component in the atmosphere, span- ning a large pressure scale throughout the entire globe. Cloud particles remain thermally stable deep in the at- mosphere up to ∼10 bar pressures and reach ∼mm sizes at the cloud base. We suggest that cloud particles form at a lower pressure boundary than considered here, and also survive to deeper depths. Cloud properties change significantly from dayside to nightside and from equa- tor to mid latitudes with variations in grain size, number density and wavelength dependent opacity. The cloud property maps show that steep gradients between day- side and nightside cloud proprieties occur at dayside- nightside transition regions. These cloud property dif- ferences have implications on interpreting observations of HD 189733b, depending what region and depth of the planet is probed. The single scattering albedo calcu- lations showed that cloud particles could play a signif- icant role in planetary phase curves. The reflectance of the clouds at 8µm showed that a significant fraction of infrared flux could originate from scattering/reflecting cloud particles. Since the most reflective clouds in the infrared correspond to regions of highest temperature it may be difficult to distinguish between cloud reflec- tions and thermal emission. The extinction properties of the cloud particles exhibit absorption signatures in the upper atmosphere ((cid:38)10−1 bar) which flatten deeper in the atmosphere ((cid:46)10−2 bar). However, this does not Article number, page 22 of 23 E. K. H. Lee et al.: Modelling the local and global cloud formation on HD 189733b Lecavelier Des Etangs, A., Pont, F., Vidal-Madjar, A., & Sing, D. Lee, G., Helling, C., Giles, H., & Bromley, S. T. 2015, A&A, 575, Ludwig, H.-G., Allard, F., & Hauschildt, P. H. 2002, A&A, 395, 99 McCullough, P. R., Crouzet, N., Deming, D., & Madhusudhan, N. 2008, A&A, 481, L83 A11 2014, ApJ, 791, 55 Mie, G. 1908, Annalen der Physik, 330, 377 Moses, J. I., Visscher, C., Fortney, J. J., et al. 2011, ApJ, 737, 15 Palik, E. D. 1985, Handbook of optical constants of solids Parmentier, V., Showman, A. P., & Lian, Y. 2013, A&A, 558, A91 Pont, F., Sing, D. K., Gibson, N. P., et al. 2013, MNRAS, 432, 2917 Posch, T., Kerschbaum, F., Fabian, D., et al. 2003, ApJS, 149, 437 Rimmer, P. B. & Helling, C. 2013, ApJ, 774, 108 Showman, A. P., Fortney, J. J., Lian, Y., et al. 2009, ApJ, 699, 564 Showman, A. P. & Polvani, L. M. 2011, ApJ, 738, 71 Sing, D. K., Pont, F., Aigrain, S., et al. 2011, MNRAS, 416, 1443 Spiegel, D. S., Silverio, K., & Burrows, A. 2009, ApJ, 699, 1487 Stark, C. R., Helling, C., Diver, D. A., & Rimmer, P. B. 2014, Inter- national Journal of Astrobiology, 13, 165 Swain, M. R., Vasisht, G., & Tinetti, G. 2008, Nature, 452, 329 Swain, M. R., Vasisht, G., Tinetti, G., et al. 2009, ApJL, 690, L114 Tinetti, G., Vidal-Madjar, A., Liang, M.-C., et al. 2007, Nature, 448, Tsai, S.-M., Dobbs-Dixon, I., & Gu, P.-G. 2014, ApJ, 793, 141 Venot, O., Hébrard, E., Agúndez, M., et al. 2012, A&A, 546, A43 Wakeford, H. R. & Sing, D. K. 2015, A&A, 573, A122 Witte, S., Helling, C., Barman, T., Heidrich, N., & Hauschildt, P. H. Witte, S., Helling, C., & Hauschildt, P. H. 2009, A&A, 506, 1367 Woitke, P. & Helling, C. 2003, A&A, 399, 297 Woitke, P. & Helling, C. 2004, A&A, 414, 335 Zeidler, S., Posch, T., & Mutschke, H. 2013, A&A, 553, A81 A&A, 564, A73 53, 197 2014, ApJ, 784, 43 ApJL, 728, L6 Radiation Bailey, R. L., Helling, C., Hodosán, G., Bilger, C., & Stark, C. R. Barman, T. S., Hauschildt, P. H., & Allard, F. 2001, ApJ, 556, 885 Berdyugina, S. V., Berdyugin, A. V., Fluri, D. M., & Piirola, V. 2011, 169 Bohren, C. F. & Clothiaux, E. E. 2006, Fundamentals of Atmospheric Bohren, C. F. & Huffman, D. R. 1983, Absorption and scattering of 2011, A&A, 529, A44 rule out a cloud particle origin for the observations of Rayleigh scattering. The scattering properties also sug- gest that the cloud particles would sparkle/reflect a mid- night blue colour over the optical wavelength regime. Acknowledgements. EL and ChH highlight the financial support of the European community under the FP7 ERC starting grant 257431. Our respective local computer support is highly acknowledged. We thank P. Woitke, C.R. Stark and P. Rimmer for helpful discussions and feedback. We thank L. Neary for insightful discussion on atmospheric mixing. References Agol, E., Cowan, N. B., Knutson, H. A., et al. 2010, ApJ, 721, 1861 Agúndez, M., Parmentier, V., Venot, O., Hersant, F., & Selsis, F. 2014, Anders, E. & Grevesse, N. 1989, Geochimica et Cosmochimica Acta, light by small particles Bruggeman, D. A. G. 1935, Annalen der Physik, 416, 636 Charbonneau, D., Knutson, H. A., Barman, T., et al. 2008, ApJ, 686, 1341 Danielski, C., Deroo, P., Waldmann, I. P., et al. 2014, ApJ, 785, 35 Dehn, M., Helling, C., Woitke, P., & Hauschildt, P. 2007, in IAU Sym- posium, Vol. 239, IAU Symposium, ed. F. Kupka, I. Roxburgh, & K. L. Chan, 227 -- 229 Demory, B.-O., de Wit, J., Lewis, N., et al. 2013, ApJL, 776, L25 Désert, J.-M., Sing, D., Vidal-Madjar, A., et al. 2011, A&A, 526, A12 Dobbs-Dixon, I. & Agol, E. 2013, MNRAS, 435, 3159 Dorschner, J., Begemann, B., Henning, T., Jaeger, C., & Mutschke, H. 1995, A&A, 300, 503 Evans, T. M., Pont, F., Sing, D. K., et al. 2013, ApJL, 772, L16 Fortney, J. J., Lodders, K., Marley, M. S., & Freedman, R. S. 2008, ApJ, 678, 1419 1396 Fortney, J. J., Shabram, M., Showman, A. P., et al. 2010, ApJ, 709, Gail, H.-P. & Sedlmayr, E. 1986, A&A, 166, 225 Gibson, N. P., Aigrain, S., Pont, F., et al. 2012, MNRAS, 422, 753 Grillmair, C. J., Burrows, A., Charbonneau, D., et al. 2008, Nature, 456, 767 677, L157 Hartogh, P., Medvedev, A. S., Kuroda, T., et al. 2005, Journal of Geo- physical Research (Planets), 110, 11008 Helling, C., Dehn, M., Woitke, P., & Hauschildt, P. H. 2008a, ApJL, Helling, C. & Fomins, A. 2013, Royal Society of London Philosophi- cal Transactions Series A, 371, 10581 Helling, C., Jardine, M., & Mokler, F. 2011, ApJ, 737, 38 Helling, C., Klein, R., Woitke, P., Nowak, U., & Sedlmayr, E. 2004, Helling, C. & Rietmeijer, F. J. M. 2009, International Journal of As- Helling, C. & Woitke, P. 2006, A&A, 455, 325 Helling, C., Woitke, P., & Thi, W.-F. 2008b, A&A, 485, 547 Henning, T., Begemann, B., Mutschke, H., & Dorschner, J. 1995, Huitson, C. M., Sing, D. K., Vidal-Madjar, A., et al. 2012, MNRAS, Jäger, C., Dorschner, J., Mutschke, H., Posch, T., & Henning, T. 2003, Jeong, K. S., Winters, J. M., Le Bertre, T., & Sedlmayr, E. 2003, A&A, Knutson, H. A., Charbonneau, D., Allen, L. E., et al. 2007, Nature, A&A, 423, 657 trobiology, 8, 3 A&AS, 112, 143 422, 2477 A&A, 408, 193 407, 191 447, 183 Knutson, H. A., Lewis, N., Fortney, J. J., et al. 2012, ApJ, 754, 22 Kreidberg, L., Bean, J. L., Désert, J.-M., et al. 2014, Nature, 505, 69 Article number, page 23 of 23
1603.06596	3	1603	2016-10-06T00:54:13	Evolutionary Analysis of Gaseous Sub-Neptune-Mass Planets with MESA	[ "astro-ph.EP" ]	Sub-Neptune-sized exoplanets represent one of the most common types of planets in the Milky Way, yet many of their properties are unknown. Here, we present a prescription to adapt the capabilities of the stellar evolution toolkit Modules for Experiments in Stellar Astrophysics (MESA) to model sub-Neptune mass planets with H/He envelopes. With the addition of routines treating the planet core luminosity, heavy element enrichment, atmospheric boundary condition, and mass loss due to hydrodynamic winds, the evolutionary pathways of planets with diverse starting conditions are more accurately constrained. Using these dynamical models, we construct mass-composition relationships of planets from 1 to 400 $M_{\oplus}$ and investigate how mass-loss impacts their composition and evolution history. We demonstrate that planet radii are typically insensitive to the evolution pathway that brought the planet to its instantaneous mass, composition and age, with variations from hysteresis. We find that planet envelope mass loss timescales, $\tau_{\rm env}$, vary non-monotonically with H/He envelope mass fractions (at fixed planet mass). In our simulations of young (100~Myr) low-mass ($M_p\lesssim10~M_\oplus$) planets with rocky cores, $\tau_{\rm env}$ is maximized at $M_{\rm env}/M_p=1\%$ to $3\%$. The resulting convergent mass loss evolution could potentially imprint itself on the close-in planet population as a preferred H/He mass fraction of ${\sim}1\%$. Looking ahead, we anticipate that this numerical code will see widespread applications complementing both 3-D models and observational exoplanet surveys.	astro-ph.EP	astro-ph	DRAFT VERSION OCTOBER 7, 2016 Preprint typeset using LATEX style emulateapj v. 12/16/11 EVOLUTIONARY ANALYSIS OF GASEOUS SUB-NEPTUNE-MASS PLANETS WITH MESA 1Department of Physics, Boston University, 590 Commonwealth Ave., Boston, MA 02215, USA HOWARD CHEN1 & LESLIE A. ROGERS2,3,4,5,6 2Department of Astronomy, California Institute of Technology, MC249-17, 1200 East California Boulevard, Pasadena, CA 91125, USA 3Division of Geological and Planetary Sciences, California Institute of Technology, MC249-17, 1200 East California Boulevard, Pasadena, CA 91125, USA and 4Department of Earth & Planetary Sciences, University of California, 307 McCone Hall, Berkeley, CA, 94720-4767, USA 6 1 0 2 t c O 6 . ] P E h p - o r t s a [ 3 v 6 9 5 6 0 . 3 0 6 1 : v i X r a Draft version October 7, 2016 ABSTRACT Sub-Neptune-sized exoplanets represent the most common types of planets in the Milky Way, yet many of their properties are unknown. Here, we present a prescription to adapt the capabilities of the stellar evolution toolkit Modules for Experiments in Stellar Astrophysics (MESA) to model sub-Neptune mass planets with H/He envelopes. With the addition of routines treating the planet core luminosity, heavy element enrichment, atmospheric boundary condition, and mass loss due to hydrodynamic winds, the evolutionary pathways of planets with diverse starting conditions are more accurately constrained. Using these dynamical models, we construct mass-composition relationships of planets from 1 to 400 M⊕ and investigate how mass-loss impacts their composition and evolution history. We demonstrate that planet radii are typically insensitive to the evo- lution pathway that brought the planet to its instantaneous mass, composition and age, with variations from hysteresis . 2%. We find that planet envelope mass loss timescales, τenv, vary non-monotonically with H/He envelope mass fractions (at fixed planet mass). In our simulations of young (100 Myr) low-mass (Mp . 10 M⊕) planets with rocky cores, τenv is maximized at Menv/Mp = 1% to 3%. The resulting convergent mass loss evo- lution could potentially imprint itself on the close-in planet population as a preferred H/He mass fraction of ∼1%. Looking ahead, we anticipate that this numerical code will see widespread applications complementing both 3-D models and observational exoplanet surveys. Keywords: methods: numerical, planets and satellites: atmospheres - interiors - physical evolution 1. INTRODUCTION A striking revelation from NASA's Kepler mission is the profusion of sub-Neptune sized planets discovered with short orbital periods (Porb . 50 days) (Borucki et al. 2011; Batalha et al. 2013; Burke et al. 2014; Rowe et al. 2014; Han et al. 2014; Mullally et al. 2015; Burke et al. 2015). De- spite the absence of sub-Neptune, super-Earth sized planets in our Solar System, they are quite ubiquitous in the Milky Way, comprising the majority of planets found by Kepler (Fressin et al. 2013; Petigura et al. 2013b). For inner orbital distances . 0.25 AU, the number of sub-Neptune mass plan- ets exceeds that of Jovian planets by more than a factor of 30 (Howard et al. 2012). Among the transiting sub-Neptune-sized planets that have measured masses, many of them have mean densities so low that they must have significant complement of light gasses (hydrogen and helium) contributing to the planet volume (e.g., Kepler-11c,d,e,f,g Lissauer et al. 2011, 2013). Wolfgang & Lopez (2015) inferred from the Kepler radius distribution that most sub-Neptunes should have present day composition of ∼1% H/He envelope, under the assumption that all close-in planets consist of rocky cores surrounded by H/He envelopes. Considering the sample of Kepler tran- siting planets with Keck-HIRES radial velocity follow-up (Marcy et al. 2014), Rogers (2015) showed that at planet radii of 1.6 R⊕ (and larger) most close-in planets of that size have sufficiently low mean densities that they require a volatile en- velope (consisting of H/He and/or water). Notably, a large scattering of mass-radius measurements is found in the sub- [email protected] 5 Hubble Fellow 6 NASA Sagan Fellow Neptune, super-Earth size regime (Marcy et al. 2014), partic- ularly between 1.6 and 4 R⊕ (Weiss & Marcy 2014). Noise in the mass-radius measurements (which often have large er- ror bars) does not account for all of the apparent scatter; there is evidence for intrinsic dispersion in the masses of planets of specified size (Wolfgang et al. 2015). The intrinsic scatter of small planet mass-radius measurements is an indicator of compositional diversity (e.g., Rogers & Seager 2010). The composition distribution of planets observed today re- flects both the initial outcomes from planet formation, and subsequent post-formation evolution processes. To study the latter, a useful avenue is numerical simulations of planets' thermo-physical evolution. Stellar flux delays the cooling and contraction of close-in planets. At the same time, higher lev- els of incident flux also mean greater susceptibility to atmo- spheric escape. By "backtracking" the thermal and mass loss evolution history, inferences about the present-day composi- tion and past history of a planet can be made. Such anal- yses have been carried out for GJ 1214b, CoRoT-7b, and the Kepler 11 and Kepler 36 systems (Valencia et al. 2010; Nettelmann et al. 2011; Lopez et al. 2012; Lopez & Fortney 2013; Howe & Burrows 2015). For example, The effect of atmospheric escape has also been stud- ied for exoplanets in the broader population-level con- text. the planet mass-loss simulations of Lopez & Fortney (2013), Owen & Wu (2013), and Jin et al. (2014) predict a "radius occurrence valley" dividing the sub-populations of close-in planets that have lost/retained their volatile envelopes. Focusing on higher-mass planets, Kurokawa & Nakamoto (2014) investigated whether mass- loss from hot-Jupiters could reproduce the observed "desert of sub-Jupiter size exoplanets." More recently, Luger et al. 2 CHEN & ROGERS (2015) assessed migration, habitability, and transformation of Neptune-like to Earth-like worlds orbiting M-dwarfs. In this work, we make two contributions. First, we im- plement several physical formulations related to low mass planets with hydrogen/helium envelopes into the state-of-the- art MESA code (Modules for Experiments in Stellar As- trophysics) (Paxton et al. 2011, 2013, 2015). MESA is an open-source, 1-D stellar evolution code that has seen wide use to address problems in stellar astrophysics such as low mass and high mass stars, white dwarfs, young neutron stars, and pre-supernova outbursts or supernova core-collapse (e.g., Wolf et al. 2013; Perna et al. 2014; Mcley & Soker 2014). Only a handful of planetary studies using MESA exist in the literature (Batygin & Stevenson 2013; Owen & Wu 2013; Valsecchi et al. 2014, 2015; Jackson et al. 2016), and there is still much potential to push the numerical code to even lower planet masses (1 - 10 M⊕). It is our aim that this pa- per will serve as a platform for future exoplanetary studies with MESA. The open source nature of MESA allows the as- tronomical community to readily access the numerical exten- sions introduced here. Second, we generate suites of planet evolution simula- tions to compute mass-radius-composition-age relations for low mass planets, to quantify how those relations depend on evolution history, and, finally, to explore whether photo- evaporation can produce a "favored" planet composition (en- velope mass fraction). The numerical models and results ob- tained are available for public access. This paper is structured as follows. In Section 2, we de- scribe methods for modeling low-mass planets with MESA. We benchmark are simulations against previously published planet evolution calculations in Section 3. We present our nu- merical results for the coupled thermal-mass-loss evolution of planets in Section 4, and discuss and conclude in Sections 5 and 6. 2. MODEL & APPROACH We employ the MESA toolkit (Paxton et al. 2011, 2013, 2015) (version 7623) to construct and evolve thousands of planet models. We consider spherically symmetric planets consisting of a heavy-element interior (comprised of rocky material, or a mixture of rock and ice) surrounded by a hydrogen-helium dominated envelope. The one-dimensional stellar evolution module, MESA star, is adapted to evolve planetary H/He envelopes. In simulating planetary H/He envelopes, we employ the default MESA input options, unless otherwise stated. For equation of state in planetary conditions, we adopted the hy- drogen/helium equation of state (EOS) from Saumon et al. (1995). For the sake of simplicity, we restrict ourselves to solar values of metallicity Z = 0.03 and helium frac- tion Y = 0.25, unless otherwise stated. We used the stan- dard low temperature Rosseland tables (Freedman et al. 2008) and (Freedman et al. 2014) for visible and infrared opacities. The MESA EOS and opacity tables are further described in Paxton et al. (2011, 2013). MESA already has the study of planets several useful built-in functions designed for (e.g., Paxton et al. 2013, Wu & Lithwick 2013, Becker & Batygin 2013, and Valsecchi et al. 2014). Some of these build-in capabilities however, run into issues when trying to create and evolve small planets. The following sections (Section 2.1 to 2.4) detail our modifications to the MESA code, and recipes for constructing low-mass planets. 2.1. Initial Starting Models Creating an initial starting model in MESA is a multi-step process. First, we make an initial model (using the cre- ate_initial_model option) with a fixed pressure and temper- ature boundary. We specify these values to be comparable to those in the subsequent evolutionary stages. This technique avoids large discrepancies between the created model and pa- rameters imposed in the evolution phase. Next, we insert an inert core (using the relax_core option) at the bottommost zone, to represent the heavy-element inte- rior of the planet. We use the models of Rogers et al. (2011) to determine the core radius and bulk density. We consider both a "rocky composition" (70% silicates and 30% Fe) and an "ice-rock mixture composition" (67% H2O, 23% silicates and 10% Fe) as options for the heavy-element interior. The addition of an inert core at this stage allows the envelope to be more strongly gravitationally bound in the steps that fol- low. it In the next step, we rescale the entire planet envelope to the desired total planet mass (using the relax_mass_scale option). Another option to reduce the planet mass is via a mass-loss wind (relax_mass Batygin & Stevenson 2013). Finally, after having constructed a model with the desired core mass and envelope mass, we still must quantitatively standardize the initial entropy starting conditions before en- tering the evolution phase. In each simulation, we take the general approach of "re-inflating" planets, to reset planet evo- lution. For high mass planets (cid:0)Mp ≥ 17 M⊕(cid:1), we use an ar- tificial core luminosity to re-inflate the planet envelope reaches a specified interior entropy threshold until test (which depends on mass). In a discrete set of cases, we determined the maximum entropy (smax) to which the planets could be inflated before they reach runaway inflation. We fit a linear function to the maximum entropy data as a function of planet mass, (cid:0)smax/(cid:0)kB baryon- 1(cid:1) = 2.1662 log(cid:0)Mp/M⊕(cid:1) + 10.6855(cid:1), and use this interpolating function to specify the initial entropy of our high-mass simulated planets. For planets with Mp ≥ 17 M⊕, the artificial core luminosity deposited at the base of the envelope during the reinflation phase is taken to be three times the surface luminosity (total intrinsic luminosity) of the planet after it is left to passively evolve for 1000 years at the end of the mass-reduction phase. For low mass models . 17 M⊕, special care must be taken when reinflating the planet envelopes. In particular, the pro- cedure used to set the artificial reinflation luminosity in the high mass regime (three times the planet luminosity at the end of the mass-reduction phase) can blow the envelope of these low-mass loosely bound planets apart. In the low-mass regime, we took a discrete set of test cases (spanning a grid of Mp and fenv) and determined in each case the maximum reinflation luminosity factor for which the planet envelope would remain bound (within a precision of 0.02). In practice this involved finding the highest artificial core luminosity for which the planet radius plateaued to a new equilibrium within 100 Myr, instead of expanding to infinity. Then, given any ar- bitrary planet mass and envelope mass fraction as inputs, we use 2D interpolation to resolve the specific luminosity. Once the planet is reinflated, we turn off the artificial re- inflation luminosity, and reset the planet age to zero. The zero-age planet model that we have created so far, does not yet account for radiation from the host star. We then use the EVOLVING GASEOUS SUB-NEPTUNES WITH MESA 3 MESA relax_irradiation option to gradually irradiate our non- irradiated model to the desired level. Convergence issues may arise if a high irradiation flux is abruptly introduced on an unirradiated model; not only does the irradiation have to dif- fuse its way in from the surface of the envelope, but the heat flux escaping from the core also has to adjust to new surface boundary conditions. The relax_irradiation option solves this issue by adding a phase in which the flux is turned on slowly. After the initial starting conditions have been set, we al- low the planet to undergo pure thermal evolution (without mass-loss) for 10 Myr before turning on mass loss (follow- ing Lopez et al. 2012). These initial cooling periods avoid the runaway mass loss scenario where the size of planet increase without bound as the simulations enter the mass-loss evolu- tion phases. By the onset of evaporation at 10 Myr, models in general should be insensitive to the details of their initial entropy choice. This is because the cooling timescale is typi- cally much less than 10 Myr. 2.2. Atmospheric Boundary Conditions When simulating highly irradiated low-mass planets (cid:0)Mp . 20 M⊕(cid:1), direct application of MESA's default irradi- ated atmosphere boundary conditions can lead to problems. In particular, care must be taken to ensure that the atmospheric boundary conditions account for the large changes in surface gravity the low-mass low-density planets may experience over their evolution as they cool and contract. In MESA's current grey_irradiated planetary atmosphere option, the surface pressure (at the base of the atmosphere) is resolved at a fixed value. This can lead to issues for simula- tions of low-mass planets (cid:0). 20 M⊕(cid:1) contracting from very puffy initial states, as the optical depth and opacities at a spec- ified pressure level vary significantly over time. We follow the approach of Owen & Wu (2013) to overcome this issue in im- plementing the T (τ) relation of Guillot (2010) specifying a fixed optical depth τ (instead of fixed pressure) at the base of the atmosphere. This allows the surface pressure boundary to vary over the course of the planets' evolution. Unless other- wise stated, the planet radii quoted through out this work, are defined at optical depth τ = 2/3 (for outgoing thermal radia- tion). To relate the irradiation flux absorbed by a planet to the planet's orbital separation and host star properties, a model for the fraction of the irradiation that is absorbed versus reflected by the planet's atmosphere is needed. In general, the Bond albedo of a planet depends on the precise atmospheric com- position and the scattering properties of clouds in the planet's atmosphere. Constructing the planet atmosphere directly is beyond the scope of this article, instead we assume albedo values are taken from Fortney et al. (2007). 2.3. Hydrodynamic Evaporative Mass-Loss Extreme ultraviolet and X-ray radiation (EUV; 200 . λ . 911Å) from a planet's host star heat the outer reaches of a planet's H/He envelope. For close-orbiting planets, this energy imparted on the at- mosphere generates a hydrodynamic wind and causes some gas to escape the planets' gravitational potential well. Similar processes are also proposed to explain the escape of atomic hydrogen in Early Venus and Early Earth (Kasting & Pollack 1983; Watson et al. 1981). For this study, we implement irradiation-driven mass loss in MESA using the prescrip- tions of Murray-Clay et al. (2009). Note that in this work we only focus on implementing hydrodynamic mass loss; we do not directly treat Jeans escape, direct blow-off, or "photon- limited" escape (Owen & Alvarez 2016) which each may be relevant in other regimes. At low levels of irradiation, the mass loss is assumed to be energy limited. We follow the energy limited escape for- mulation first described by Watson et al. (1981) then studied by Lammer et al. (2003) Erkaev et al. (2007), Valencia et al. (2010), Lopez et al. (2012), and Luger et al. (2015). The energy-limited mass loss rate is given by: dMp dt = - ǫEUVπFEUVRpR2 GMpKtidal EUV , (1) where ǫEUV is the mass loss efficiency (i.e. the fraction of inci- dent EUV energy that contributes to unbinding the outer lay- ers of the planet), which depends on atmospheric composition and the EUV flux. Here we adopt a mean efficiency value of 0.1 (Jackson et al. 2012; Lopez et al. 2012) unless otherwise stated. FEUV is the extreme ultraviolet energy flux from the host star impinging on the planet atmosphere. Rp and Mp are planet radius at optical depth τvisible = 1 (in the visible) and the total mass of the planet respectively. G is the gravitational constant. RHill ≈ a(cid:0)Mp/3M∗(cid:1)1/3 represents the distances out to which the planet's gravitational influence dominates over the gravitational influence of the star. Note that all our models we assume RHill to be located well within the exobase where the particle mean free path and at- mospheric scale height are comparable. Ktidal (a factor that de- pends on the ratio of RHill and REUV) corrects for tidal forces, which modify the geometry of the potential energy well and decrease the energy deposition needed to escape the planet's gravity (Erkaev et al. 2007). Finally, REUV is distance from the center of the planet to the point where the atmosphere is optically thick to EUV pho- tons. To calculate REUV, which changes with time, we first approximate the difference between τvisible = 1 and τEUV = 1 (the photo-ionization base) with REUV ≈ Rp + H ln(cid:18) Pphoto PEUV (cid:19) (2) where H = (kBTphoto)/(2mHg) is the atmospheric scale height at the photosphere (the factor of 2 in the scale height equa- tion denotes the molecular form of hydrogen in this regime). Pphoto and Tphoto are the pressure and temperature at the visible photosphere. Following Murray-Clay et al. 20 eV instead of integrating over the host star spectrum. (2009), we estimate the pressure at τEUV = 1 from the photo-ionization of hy- - 3 cm2 as PEUV ≈ drogen, σν0 = 6 × 10- 18(cid:0)hν0/13.6 eV(cid:1) (cid:0)mHGMp(cid:1) /(cid:0)σν0 R2 p(cid:1), adopting a typical EUV energy of hν0 = At high EUV fluxes (cid:0)& 104 erg s- 1 cm- 3(cid:1), radiative losses from Lyα cooling become important, mass loss ceases to be energy limited (Murray-Clay et al. 2009) and a constant mass loss efficiency parameter assumption no longer holds. In this regime, photo-ionizations are balanced by radiative recombi- nations and radiative losses maintain the temperature of the wind at Twind ∼ 104 K. This radiation-recombination limited mass loss rate is approximated by, 4 CHEN & ROGERS GMp s REUV (cid:19), (3) c2 c2 dMp e(cid:18)2- s (cid:19)1/2 s (cid:19)2 = - π(cid:18) GMp csmH(cid:18) FEUVGMp hν0αrecR2 EUVc2 dt (cid:12)(cid:12)(cid:12)(cid:12)rr- lim where cs = (cid:0)2kBTwind/mH(cid:1)1/2 is the isothermal sound speed tion coefficient at 104 K (cid:0)2.7 × 10- 13 cm3 s- 1(cid:1). Equation 3 is identically Equation 20 of Murray-Clay et al. (2009), but with the dependencies on planet mass and surface gravity explicitly preserved (see also Kurokawa & Nakamoto 2014). of the fully ionized wind, and αrec is the radiative recombina- For the EUV luminosity of the planet host star (assumed sun-like), we adopt XUV evolution model of Ribas et al. (2005). At each time step, we evaluate both the energy- limited or radiation-recombination limited mass loss rates and impose the lesser of the two on the MESA planet model. Above, χ is the mass fraction of "chrondritic" material in the planet heavy element interior; this factor is 1 for the Earth-like core composition and 0.33 for the ice-rock core composition. λi is the decay rate constant, and Hinitial,i is the initial rate of energy released (per unit mass of rocky material) at t = 0 by the decay of the ith nuclide. The important long lived radioac- tive nuclides are 232Th, 238U, 40K, and 235U and their half- lives are respectively 1.405 × 1010, 4.468 × 109, 1.26 × 109, and 7.04 × 108 years. We use the chrondritic abundances and initial energy production rates from Hartmann (2004). 3. MODEL BENCHMARKING Due to the extensive nature of our adaptations to MESA, both in the default parameters and the subroutine modules, there is a need to ensure the that our calculations are con- sistent with those in current literature. Thus our numerical results are first presented with a series of benchmarking exer- cises. 3.1. Thermal Evolution Benchmarking We first simulate the thermal evolution of Jovian and sub- Neptune planets without the effects of atmospheric escape. We run simulations of high mass (Mp ≥ 17 M⊕) gas giant planets to compare to the models of Fortney et al. (2007) (Ta- ble 1). These simulated planet orbit Sun-twins, have heavy element interiors consisting of 50% rock and 50% ice by mass, and have no heating from the core (Lcore = 0) (as in Fortney et al. 2007). By ages of 1 Gyr, the planet radii predicted by MESA typi- cally agree with the tabulated radii from Fortney et al. (2007) to within 3% and to better than 7% for all planet masses and orbital separations. Radius offsets exist at younger ages (100- 300 Myr), as would be expected due to the differences in how we set the initial conditions. A crucial difference be- tween our approaches is how we model the absorption of stellar flux in the atmosphere. Fortney et al. (2007) use a grid of self-consistent radiative-convective equilibrium atmo- spheric structure models computed following a correlated-K approach, whereas we use a semi-grey atmospheric bound- ary condition based on the modified T (τ) relation of Guillot (2010) (Section 2.2). Nonetheless, we find encouraging agreement between MESA and the model planet radii of Fortney et al. (2007). For planet masses below 20 M⊕, we bench- mark our results against Lopez & Fortney (2014) and Bodenheimer & Lissauer (2014) (Table 2). These simulations have Earth-like rocky cores (70% silicate and 30% Fe) with heat capacity of 1.0 J K- 1 g- 1 and do not experience the effects of photo-evaporation. At planet masses near 20 M⊕ and at low levels of inci- dent irradiation, we find good agreement (within 5%) be- tween the planet radii predicted by MESA and the results of Lopez & Fortney (2014). The differences between the two sets of model radii are most extreme at low planet masses and high fenv, though all simulations agree to within 20%. The modeled MESA planet radii further show a stronger depen- dence on the irradiation flux, predicting low-mass planet radii that are larger than Lopez & Fortney (2014) at 1000 F⊕, but smaller than those of Lopez & Fortney (2014) at 10 F⊕. As for the comparisons with Bodenheimer & Lissauer (2014), we look at their Table 2 runs with the accretion cut- off times at 2 Myr (runs 2H, 1, and 0.5). At Teq = 500 K 2.4. Heavy Element Interior We incorporate two main updates to MESA to more realisti- cally model planet heavy element interiors. First, as described in Section 2.1, we use the models of Rogers et al. (2011) to set the core radius and bulk density for a specified heavy element interior composition and mass. In this work, we do not model the variations in the core radius as a function of time or pressure over-burden. This assumption is appropriate for the low-mass planets that are the focus of this paper. However, the compression of the planet heavy-element interior due to the pressure of the surrounding envelope starts to be significant for cases in which the pressure at the boundary of the core and envelope exceeds ∼ 1010 Pa (e.g., Mordasini et al. 2012). Second, we incorporate into MESA a time-varying core lu- minosity to account for the heavy-element interior's contri- bution to the envelope energy budget. This contribution may be negligible in hot Jupiters, but is more significant in cases where the core represents a substantial fraction of the total planet mass. For this reason, the energy budget for planets Mp . 20 M⊕ is more significantly influenced by the presence of a core. The core luminosity, Lcore, (i.e., the energy input to the base of the envelope from the heavy element interior) is commonly modeled as, Lcore = - cvMcore dTcore dt + Lradio. (4) The first term on the right hand side of Equation 4 accounts for the thermal inertia of the core. Therein, cv is the effective con- stant volume heat capacity of the core (in ergsK- 1g- 1), Mcore is the mass of the planet's core, and dT core/dt is time deriva- tive of the effective (mass-weighted) core temperature. We take the Lagrangian time derivative of the temperature at the base of the planet envelope as an approximation to dT core/dt. The most uncertain factor is cv, which could vary depending on the composition of the core and the presence of spatial composition gradients or thermal boundary layers. For the ice-rock core, we adopt a value of 1.2 J K- 1 g- 1. And the rocky core 1.0 J K- 1 g- 1 (Guillot et al. 1995). The second term on the right hand side of Equation 4, Lradio, represents the contribution of the decay of radio nuclei to the core lumi- nosity. Lradio = χMcoreXi Hinitial,ie- λit. EVOLVING GASEOUS SUB-NEPTUNES WITH MESA 5 and 200 K and ages of 4 Gyr, the final planet radii computed by MESA and by Bodenheimer & Lissauer (2014) typically agree within 3% and in all cases agree to better than 10%. In addition to the Earth-composition heavy element interi- ors that are used for benchmarking purposes, we also simu- late the same grid of planet masses, envelope mass fractions, and orbital separations with heavy element interiors that are 70% ice and 30% rock by mass (Table 2). Notice the deci- sive role that changing the core composition has on the over- all planetary radii. As expected, keeping all other parame- ters identical, the simulated planets with ice-rock cores have larger radii than those with rocky cores. Ice-rock cores have a smaller radio-luminosity than rocky cores due to their smaller mass fractions of radioactive nuclides. However, this factor is greatly compensated by the fact that ice-rock cores have a lower densities and surface gravities than rocky cores. We shall see that this increase of planet radii with the replacement of ice-rock heavy-element interiors have important implica- tions for their H/He envelope survival rates. 3.2. Mass-Loss Evolution Benchmarking In Section 3.1, we modeled planet thermal evolution in the absence of mass loss and compared the MESA-simulated planet radii to Fortney et al. (2007) and Lopez & Fortney (2014) (Tables 1 and 2). Here, we turn to benchmarking our evolution calculations including the full evaporation prescrip- tion described in Section 2.3. We follow Murray-Clay et al. (2009) in calculating the mass loss rates (for hot Jupiters). For a 0.7 MJ planet with radius 1.4 RJ located at 0.05 AU around a solar-mass (G-type) star, they computed a present mass-loss rate of ∼4 ×1010 g s- 1 and a maximum mass loss rate of ∼6 × 1012 g s- 1. Our evo- lutionary calculations yield a close result of 5.5 × 1010 g s- 1 at 4.5 Gyr, although our modeled planet radius at this time is only 1.2 RJ since we do not include any hot Jupiter inflation mechanisms in our simulations. Additionally, these mass loss rates agree qualitatively well with more detailed calculations including magneto-hydrodynamic effects (e.g. Vidotto et al. 2015 Chadney et al. 2015, Tripathi et al. 2015 ). Our mass- loss rate compared to these studies do not differ by more than an order of magnitude at old ages for planets ∼1 MJ. We also compared our coupled thermal-mass loss evolu- tion calculations against previous planet evolution studies. Owen & Wu (2013) found that a 318 M⊕ planet with a 15 M⊕ core at 0.025 AU lost ∼0.51% of its total mass over 10 Gyr. In our simulations, a planet with the same mass, composi- tion, and orbital separation lost 5 M⊕ (or ∼1.81% of its total mass) over the same time span of 10 Gyr. This discrepancy may be attributed to the fact that we use a fixed mass-loss efficiency factor of ǫEUV = 0.1, whereas Owen & Wu (2013) perform a more detailed calculation of the mass loss rate. In performing a full calculation of the hydrodynamics of X-ray driven escape, Owen & Jackson (2012) concluded that Jovian mass planets drive the least efficient winds, and that the effec- tive approximate escape efficiency for this particular planet scenario is ∼0.01. Assuming ǫEUV = 0.01 in our simulation, the cumulative mass lost agrees very well with the results of Owen & Jackson (2012), with our simulation losing ∼0.57% of its total mass. Lopez & Fortney (2013) simulated mass loss from a 320 M⊕ Jupiter-mass planet with a 64 M⊕ core at 0.033 AU (roughly 1000 F⊕). They found that the planet lost less than 2% of its total mass. When we model an identical planet we find a total of 1.1% of mass lost over 10 Gyr. The fact that we find a lower cumulative mass loss is to be expected be- cause Lopez & Fortney (2013) assume energy limited escape throughout the entire planet evolution, which would lead to an overestimation of the total mass lost. For simulations of mass loss from planets in the sub- Neptune/Super-Earth regime, we compare with figure 3 in Owen & Wu (2013). This is a (initially) 20 M⊕ planet with a 12 M⊕ core. With their X-ray and EUV driven evaporation scheme, Owen & Wu (2013) found that the planet ended up with about 13 M⊕. We find a slightly higher final mass value of 13.45 M⊕. This difference may be partly due to the fact that Owen & Jackson (2012) more carefully treats the differ- ence in EUV and X-ray driven wind scenarios. We also com- pute the mass loss evolution of the Kepler-36 system to repro- duce Figure 1 from Lopez & Fortney (2013) (Figure 1). These simulations start with the initial compositions and masses (Kepler-36c: fenv = 22%, and Mp = 9.41M⊕, Kepler-36b: fenv = 22%, and Mp = 4.45M⊕) that Lopez & Fortney (2013) used it their evolution calculations to match the current prop- erties of the planets inferred by (Carter et al. 2012). Our cal- culated evolution tracks of both planets agree qualitatively well with those from Lopez & Fortney (2013) (Figure 1); we also find that 36b ends with a density higher than 36c by a factor of 8 despite the two beginning with the same compo- sition. Note however, that the predicted planet radii in our models are slightly above the measure radii (for instance, we predicted 3.75 R⊕ as opposed to the measured 3.67 R⊕ for 36c). / e H H % ] i R [ s u d a r t e n a p l ⊕ 14 12 25 20 15 10 5 0 Kepler 36b Kepler 36c Kepler 36c w/o atm esc 10 8 6 4 2 0 8.0 8.5 9.0 9.5 10.0 age since disk clearing [log10 years] Figure 1. Plot of the thermal/mass-loss evolution of Kepler-36 b and c, with radius and envelope mass fraction as a function of time. This plot is de- signed to be compared with Figure 1 from Lopez & Fortney (2013). The dashed curves have initial compositions and masses (Kepler-36c: fenv = 22%, and Mp = 9.41M⊕, Kepler-36b: fenv = 22%, and Mp = 4.45M⊕) that Lopez & Fortney (2013) used it their evolution calculations to match the cur- rent properties of the planets inferred by (Carter et al. 2012). The solid curve is a simulation that begins with the inferred current composition and evolved in the absence of photo-evaporation. Note the relatively small final radii dif- ference (∼1.35%) between solid and dashed curves for Kepler-36c. To summarize Sections 3.1 and 3.2, we find that our planet evolution results are in general agreement with those in pub- lished literature. We also find that the majority of the dis- crepancies can be attributed to differences in our evaporation schemes and in the way the initial starting conditions are spec- ified. 6 14 12 ] ⊕ 10 i R [ s u d a r t e n a p l ] ⊕ i R [ s u d a r t e n a p l 8 6 4 2 0 14 12 10 8 6 4 2 0 14 12 10 8 6 4 2 0 0 a = 0.1 AU age = 4.5 Gyr 100 200 300 planet mass [M ] ⊕ 400 500 500 Myr Fe-Si Core 4.5 Gyr 6 5 4 3 2 1 0 500 Myr 6 4.5 Gyr 5 Fe-Si Core 4 3 2 1 Ice-Fe Core 5 10 0 20 15 planet mass [M 0 Ice-Fe Core 10 15 20 5 ] ⊕ Figure 2. Planetary mass-radius relationship varying core mass or envelope mass fraction in the absence of mass-loss at 0.1 AU. For the top panel, the core mass values from top to bottom are 0, 5, 10, 25, 50, 100 M⊕. Notice an "ultra-low" density region where Neptune-mass planets attain sizes well above that of Saturn. For the bottom panel, compare particularly between the radii of rocky interior versus ice-rock interior simulations. The fenv = Menv/Mtot values from bottom to top is 0.05, 0.1, 0.5 ,1.0, 2, 5, 10, 15%. A more exhaustive table is included in Appendix A. 4. CALCULATIONS & RESULTS We now turn to applying MESA, with the extensions dis- cussed in Section 2, to simulate the evolution of planets with H/He envelopes. We explore some of the important attributes of H/He-laden planets including ultra-low density configura- tions, the (in)dependence on evolution history of the current planetary radii and compositions, the role of evaporation in determining planet survival lifetimes as a function of com- position, and a synthetic planet population generated by our models. 4.1. Creating Mass-Radius-Composition Relations with MESA Theoretical planetary mass-radius relationships have long been a helpful tool in aiding interpretation and characteriza- tion efforts. To produced mass-radius-composition relations (Figure 2), we evolve planets spanning a mass range of 2 M⊕ to 20 M⊕, initial envelope mass fractions of 0.05, 0.1, 0.5 ,1, 2, 5, 10, 15, 20, 25%, and orbital separations of 0.05, 0.075, 0.1, and 1.0 AU. We evolve planets with mass loss for 1 Gyr, CHEN & ROGERS and 10 Gyr time spans and record both the radii and final com- positions of the planets at those ages. With our adaptations, MESA can simulate H/He envelopes surrounding planets down to masses of 1 M⊕ and H/He mass fractions down to fenv = 1 × 10- 6, in the absence of H/He mass loss. Below fenv = 1 × 10- 6 the approximation of an optically thick envelope start to break down. Typically, it is harder to evolve low mass (Mp . 15 M⊕) planets as fenv values become greater than ∼40%. With mass loss turned on in the simulations, there are cer- tain regimes of planet parameter space in which MESA runs into issues, specifically at low planet masses, high fenv, and high levels of irradiation. Some of the numerical difficul- ties can be attributed to the fact that highly-irradiated, low- mass, low-density planets are more unstable to mass loss. For example, a 8 M⊕ planet within 0.045 AU orbital dis- tance might be unable to hold on to its atmosphere even for a short (∼5,000 yr) timescale. The time steps in the evolution- ary phases are generally much larger than the nominal 1,000 years. For this reason, MESA would unable to resolve the boundary conditions between the two time steps and outputs convergence error. For these regimes, instead of losing its H/He over ∼100 Myr these simulations crash at the start. To study low mass gaseous planets in extreme equilibrium tem- peratures, one would need to develop a more suitable bound- ary condition and initialization parameters. However, these limitations largely do not impede our effort to create mass- radius isochrons in our regimes of interest. From our grid of planet simulations, we derive fitting for- mulae that may be used to estimate the radius of a planet at specified mass, composition, irradiation flux, and age. Though we advocate interpolating within Tables 2 and 2 (or directly simulating the desired planet using MESA) to derive planet radii for most applications, we provide these fitting for- mula for situations where a quick analytic approximation may come in handy. We fit our simulated planet radii to a model in which the logarithm of the contribution of the planet's H/He layer to the planet radius(cid:0)Rp - Rcore(cid:1) varies quadratically with the logarithms of Mp, fenv, Fp, and age. log10(cid:18) Rp - Rcore R⊕ (cid:19) = c0 + 4 4 4 Xi=1 ci jxix j (5) Xj≥i cixi + Xi=1 x1 = log10(cid:18) Mp M⊕(cid:19) x2 = log10(cid:18) fenv 0.05(cid:19) x3 = log10(cid:18) Fp F⊕(cid:19) x4 = log10(cid:18) t 5 Gyr(cid:19) We provide the best fitting coefficients in Table 2, both for planets with rocky-composition cores and planets with ice- rock cores. These expressions are valid for 10- 4 ≤ fenv ≤ 0.2, 1 ≤ Mp/M⊕ ≤ 20, 4 ≤ Fp/F⊕ ≤ 400, and 100 Myr ≤ t ≤ 10 Gyr (and should not be extrapolated beyond these ranges). Though we initially endeavored to apply a simpler linear power-law model (as in Lopez & Fortney 2014), we found non-negligible curvature in the constant envelope radius hy- perplanes in log10 Mp, log10 fenv, log10 Fp, and log10 t space, EVOLVING GASEOUS SUB-NEPTUNES WITH MESA 7 motivating the addition of quadratic terms to the model. Ul- timately, these quadratic fits to the rocky core and ice-rock core simulations have adjusted R2 of 0.993 and 0.987 (com- pared to 0.914 and 0.816 for the pure powerlaw) and have root mean squared residuals in the envelope radii of 0.0347 dex and 0.0465 dex (compared to 0.122 dex and 0.133 dex for the pure powerlaw). In the above equations, the radius of the planet H/He enve- lope has a steep dependence on change in envelope fraction and a much shallower dependence on age. This is because non-opacity-enhanced models cool at much higher rates dur- ing the first 100 Myr of evolution. 4.2. The Case of Ultra Low-Density Planets As shown in Figure 2, there exist situations in which sim- ulated sub-Saturn mass planets have larger sizes than Jupiter. These ultra low-density "puff-ball" planets (ρ ∼ 0.2g/cm3) are an interesting new class that challenges planet formation theories. Rogers et al. (2011) showed that moderately irra- diated planets low mass-planets (3 - 8 M⊕) with extended H/He envelopes can plausibly have transit radii comparable to Jupiter. Several measurements by Kepler provide evi- dence for the existence of these extremely low density planets, for example, Kepler-30d, Kepler-79 and Kepler-51 systems (Sanchis-Ojeda et al. 2012; Jontof-Hutter et al. 2014; Masuda 2014). Batygin & Stevenson (2013) considered these planets over a wide range of parameter space and calculated their mass- radius isochrons. However, their models assumed a con- stant core density (ρcore = 5 g/cm3) independent of core mass, a luminosity independent of time, and relied on mass- loss timescale arguments to assess the survival of the planet envelope. Self-consistent calculations including a refined thermal-physical model for the planet core and coupled thermal-evaporative evolution are needed, and are now pos- sible with the updates that we have made to MESA. Us- ing these updates, we reproduce the mass-radius figures from Batygin & Stevenson (2013). With our coupled thermal-mass loss simulations (shown in Figure 3), we still find ultra low-density planets (with radii above 10 R⊕ and masses below 30 M⊕), increasing the confi- dence in the survivability of these planet configurations. With the exception of the simulations with 1 M⊕ cores, our sim- ulated planetary radii show good agreement with those from Batygin & Stevenson (2013). The main differences between our mass-radius isochrons and those from Batygin & Stevenson (2013) are threefold. First, our predicted planet radii are generally lower than those in Batygin & Stevenson (2013). This can be attributed to the different ways in which we set our initial starting con- ditions. Second, we find that planets over a slightly larger range of planet mass-core mass-incident flux parameter space are susceptible to loosing their entire envelope, with the dif- ference being most pronounced for planets with low-mass cores (∼ 1 M⊕). In particular, while a 13 M⊕ planet with a 1 M⊕ core is expected to survive at 500 K based on the Batygin & Stevenson (2013) timescale criterion, such a planet is unstable to mass loss in our simulations. Third, among the planets that manage to retain their envelopes, we find fewer in- stances where planet radius increases with decreasing planet mass (at constant core mass). We do still find a negative ra- dius versus mass slope in our Teq = 700 - 300 K simulations with Mcore = 1 M⊕, however, this rise is radius toward smaller 0 5 10 15 20 25 30 35 planet mass [M ] ⊕ 700 K 500 K 300 K no irradiation Mcore = 1 M ⊕ 700 K 500 K 300 K no irradiation Mcore = 3 M ⊕ 12 ] ⊕ 10 8 6 4 i R [ s u d a r t e n a p l ] 12 ⊕ 10 i R [ s u d a r t e n a p l 8 6 4 0 5 10 15 20 25 30 35 planet mass [M ] ⊕ 12 ] ⊕ 10 i R [ s u d a r t e n a p l 8 6 4 700 K 500 K 300 K no irradiation Mcore = 5 M ⊕ 0 5 10 15 20 25 30 35 planet mass [M ] ⊕ Figure 3. Plot of planet radii versus mass at 5 Gyr, for planets suffering evaporation. The top panel is for planets with 1 M⊕ cores, the middle panel is for 3 M⊕ cores, and the bottom panel is for 5 M⊕ cores. Note that in contrast to the M-R relations from the previous sections, these are not for constant composition, but for constant core mass. This is designed to be compared with Figure 3 from Batygin & Stevenson (2013). planet masses is less steep than that in Batygin & Stevenson (2013). Planets with dRp/dMp < 0 at constant core mass can be susceptible to runaway mass-loss (e.g., Baraffe et al. 2008). 8 CHEN & ROGERS In these configurations, removal of mass from the planet leads to an expansion of the planet radius, and further in- crease in the mass loss rate. This scenario is not treated in Batygin & Stevenson (2013)'s instantaneous mass-loss timescale criterion. This accounts for why our simulations find a larger area of parameter space excluded by envelope mass loss in the mass-radius diagram (Figure 3). 4.3. Assessing Radius as a Proxy for Composition Lopez & Fortney (2014) recasted the mass-radius relations of low mass planets as more convenient radius-H/He mass fraction relations. However, they were focused on models that did not incorporate mass-loss in their evolutionary cal- culations. Such neglect is appropriate if the primary concern is to assess the effect of stellar irradiation on planetary evolu- tion tracks. With the inclusion of hydrodynamic escape, we reassess some of their main conclusions. Figure 2 shows that the radius of a planet with a given H/He mass fraction is quite independent of planet mass. This is because at old ages (& 500 Myr), the upturn in the radii of low mass planets at early times is largely offset by the higher cooling rate. Such striking behavior was previously noted by Lopez & Fortney (2014), who found that the scatter in the % H/He of their planet models at specifed radius was very low (∼0.3 dex). This behavior suggests that planet sizes could act as direct proxy for the bulk H/He content. However, the inclu- sion of mass-loss may complicate matters because the planet composition does not stay constant but varies over time. With the inclusion of mass-loss in our evolution calcula- tions, we find that radius is a proxy for the current planet composition (H/He envelope mass fraction). Figure 4 shows mass radius contours both at constant initial composition and at constant final compositions for planets at 0.1 AU that evolved for 300 Myr and 1 Gyr. Despite the ad- dition of mass-loss, the contour lines of constant final compo- sition are flat, with radius largely independent of planet mass (Mp . 30 M⊕) for planets older than ∼ 800 Myr. For exam- ple, in the mass range of 5 to 20 M⊕, the radii of our sim- ulated planets with 1% H/He vary by no more than 0.5 R⊕, while the radii of planets having 15% H/He vary by no more than 0.1 R⊕ over the same mass range. Interestingly, these results hold even at a closer orbital separations; at 0.05 AU; simulations that survive over 1 Gyr (typically & 8 M⊕) still have flat mass-radius curves. For young planets . 500 Myr there is an upturn in the mass-radius relations (at constant fi- nal compositions) at low-masses. Radiative cooling over time decreases these "inflation" of low mass planet sizes to which planets from 1 to 20 M⊕ have comparable radii. 4.4. (In)Dependence on Evolution History Planet evolution calculations, in principle, provide a useful mapping from planet mass, composition, age, and irradiation to planet radius. Often, mass-radius isochrones calculated ne- glecting mass loss from H/He envelopes (Fortney et al. 2007; Lopez & Fortney 2014; Howe et al. 2014), are applied even in scenarios in which planetary evaporation processes may be significant. Lopez et al. (2012) previously noted that evap- oration only strongly affects thermal evolution in cases of extreme mass loss when the evaporation timescale becomes comparable to the thermal cooling timescale. However, it is still crucial to more quantitatively calculate the error intro- duced by not accounting for the full mass-loss history of a planet. In this section we explore the regimes of parameter 7 6 ] ⊕ 5 i R [ s u d a r t e n a p l 4 3 2 1 0 7 6 5 10 15 25 planet mass [M 20 ⊕ ⊕ 5 ] i R [ s u d a r t e n a p l 4 3 2 1 0 5 10 15 25 planet mass [M 20 ⊕ 0.1 AU 300 Myr 30 35 40 0.1 AU 1 Gyr 30 35 40 ] ] Figure 4. Planetary mass-radius relationships showing contours of constant initial (square markers) and final (solid curves) compositions for models ex- periencing evaporation. The top panel presents planets that have evolved for 300 Myr, while the bottom panel shows planets at 1 Gyr. All simulations are presented at an orbital separation of 0.1 AU. The composition sequence from bottom to top is as follows: . Note the decisive "flattening" of the radii- composition curves for constant final composition. space in which planet evolution may exhibit hysteresis, or in other words, in which the radii of planets with identical com- positions, masses, and ages depend on their earlier evolution history. We first use the specific case of Kepler-36c to provide an illustrative example of moderate hysteresis. Figure 1 shows two distinct evolution tracks for Kepler-36c, which both end up at 10 Gyr, with identical final compositions and masses ( fenv = 8.2%, Mp = 8.01 M⊕). The first evolu- tion track (dashed line) includes atmospheric mass loss over the cumulative history of the planet, and began with an ini- tial composition of 22% H/He (9.41 M⊕ total mass). In con- trast, the second evolution track (solid-line) does not include mass loss, staying at constant mass and composition through- out its lifetime. The final planet radii of the two simulations at 10 Gyr show a ∼1.35% difference (3.65 vs 3.51 R⊕), with the non-evaporating model having a smaller radius. While this difference in the model radii appears small, it approaches the measurement uncertainty on Kepler-36c's radius (. 2%). This example motivates further investigation to map out the scenarios in which planet radii are even more strongly depen- dent on the planet evolution history. EVOLVING GASEOUS SUB-NEPTUNES WITH MESA 9 6.0 5.5 5.0 ] a = 0.10 AU age = 1 Gyr 6.0 5.5 5.0 ] a = 0.05 AU age = 1 Gyr 8 10 12 14 16 18 20 planet mass [M ] ⊕ 25.00 20.00 15.00 10.00 5.00 1.00 0.05 18 20 22 ] / ) e H H % t n e r r u c ( ⊕ 4.5 i R [ s u d a r t e n a p l ] i R [ s u d a r t e n a p l ⊕ 4.5 4.0 3.5 3.0 2.5 2.0 1.5 4 6.0 5.5 5.0 6 8 10 14 planet mass [M 12 ⊕ a = 0.10 AU age = 10 Gyr 4.0 3.5 3.0 2.5 2.0 1.5 4 6 8 10 14 planet mass [M 12 ⊕ ] ] ⊕ 4.5 i R [ s u d a r t e n a p l 16 18 20 i l ] s u d a r e p o e v n e [ h t p e d e n o z e v i t a d a r i 4.0 3.5 3.0 2.5 2.0 1.5 6 1.1 1.0 0.9 0.8 0.7 0.6 0.5 16 18 20 4 6 8 12 10 16 planet mass [M 14 ⊕ Figure 5. Quantifying the degree of hysteresis with planet evolution tracks experiencing mass-loss. The top panels show plot of planet mass-radius relations with varying initial envelope mass fractions, at an age of 1 Gyr, and at orbital distances of 0.10 AU (top left) and 0.05 AU (top right). Dotted lines represent simulations of planets experiencing mass-loss, while the solid curves represent planets (of identical instantaneous mass and composition) that evolved at constant mass. The initial envelope mass fraction values, from top to bottom, are: 25%, 20%, 15%, 10%, 5.0%, 2.0%, 1.0%, 0.05%, 0.01%. Accompanying higher values of % H/He in the planet composition is the increase of the discrepancy between the dotted and solid lines. This discrepancy is weakly dependent of planet mass, with lower mass simulations (Mp . 10 M⊕) having more significant discrepancy (up to 3% in planet radius). Conversely, in the regime of low H/He mass fractions, there is little difference between the radii of planets following the two different evolution pathways. The bottom left panel shows the same relations at 0.1 AU but for models with the same instantaneous H/He fractions at 10 Gyr. The bottom right panel present the fractional radiative zone depth (in units of the total radial thickness of the planet H/He envelope, Renv) as function of planet mass, for 1 Gyr old planets at 0.1 AU. Notice how the outer radiative zone represents a larger fraction of the envelope radius at lower H/He mass fractions and smaller planet masses, explaining the relative small display of hysteresis in this region of parameter space. We simulate a large grid of planet evolution models to ex- tend the hysteresis experiment to a wider range of planet masses, envelope fractions, and orbital separations. We use the same masses as Section 4.1, initial fenv = 25, 20, 15, 10, 5.0, 2.0, 1.0, 0.05, 0.01%, simulating a grid of evolution cal- culations with mass loss turned on. We then performed a second suite of planet evolution simulations with mass loss turned off. These simulations have compositions identical to those of the mass-losing simulations at 1 and 10 Gyr. The re- sults for 1 Gyr old planets at 0.10 AU and 0.05 AU are shown in Figure 5. In contrast to Section 4.3, where we presented simulations with the same initial compositions, here we com- pare suites of simulations that begin with different initial com- positions but end up with identical final compositions. With the inclusion of mass-loss, planets tend to be larger at a specified mass and instantaneous composition. For example, at 0.05 AU and 1 Gyr ages, the radius of a planet with a spec- ified instantaneous mass and composition can vary by up to 2.5% (at 15 M⊕, and fenv = 25%). At ∼0.1 AU, where lower mass planets (cid:0)Mp ∼ 5 M⊕(cid:1) may retain their atmospheres (1- 10% H/He), we find overall planet radius differences of ∼1%. At a given composition, higher mass-loss rates lead to greater differences in the final sizes. As a corollary, younger simu- lations receiving higher irradiation with higher envelope mass fractions are marked by a greater difference in planet radii. At older ages beyond 10 Gyr, hysteresis is almost non-existence for planets with fenv . 15%. At lower initial fenv below 10%, the difference between evo- lution computations including and disregarding mass-loss is never more than 0.5% regardless of flux received or planet mass. At further out orbital semi-major axes of 0.4 AU, there is minimal discrepancy (. 0.01%) between simulations in- cluding and excluding mass-loss in their evolutionary compu- tations. Nonetheless, with the exception of strongly irradiated orbital separations below ∼0.06 AU, the difference is always within the 1% mark. This suggests that, in most cases the er- 10 CHEN & ROGERS ror introduced by using mass-radius isochrons calculated for planets neglecting mass loss is negligible. Nonetheless, the mass loss history of a planet can introduce a systematic shift in planet mass-radius relations that can be comparable to the observations radius uncertainties for planets with fenv & 10% at 1 Gyr. This difference typically decreases to between 0.2 and 0.5% at even older ages of 10 Gyr. The degree of hystere- sis is only weakly dependent on planet mass. To get insight for the reason behind these trends, we can ex- amine the structure of the planet H/He envelope. Since more massive planets start with higher interior entropy, removing mass from a planet's envelope leads to an interior structure that is effectively "younger" (higher entropy) than a planet that evolved at a constant lower mass. This explains the fact that the simulated planets that experienced mass loss have systematically higher radii (at specified mass, age and com- position) than the simulated planets that evolved at constant mass. Planets for which the convective H/He envelope con- tributes a significant fraction of the planet radius, should be more susceptible to hysteresis than planets with deep radia- tive envelopes. The temperature of the outer radiative zone in a planet's H/He envelope is set by the radiation incident on the planet from its host star (independent of the cooling his- tory of the planet). It is the location of the radiative-convective boundary and the entropy of the convective zone that may de- pend on the planet's earlier evolution. The radiative zone depth in our simulated planets (at speci- fied age and orbital separation) depends only weakly on mass but more heavily on fenv (Figure 5). The fraction of the en- velope radius that is radiative decreases with increasing H/He mass fraction. Planets with low H/He mass fractions (below ∼1%) have envelopes that are almost entirely radiative (& 80 to 100% of the envelope radius lying in the radiative zone). Planets with very low total mass (Mp . 7 M⊕) also have en- velopes that are almost entirely radiative. At the other ex- treme, comparison of the top and bottom panels in Figure 5 indicates that in regimes where the simulated planets do ex- hibit hysteresis (at fenv & 10%, where the eventual planet radii differ by ∼0.5 - 1.5%) planets generally have more substan- tial convective regions, accounting for at least 30% of the total radial extent of the planet envelope. It is possible however, that the cause of the apparent hys- teresis in our planet evolution simulations is numerical in na- ture. This is because our modified T (τ) atmospheric boundary condition does not conserve energy globally to machine pre- cision. Since mass-loss has some pdV work associated with it in the upper atmosphere, a fixed temperature boundary con- dition does not treat this correctly, and in some cases will lead to more energy input into the planet's atmosphere. In reality however, the advection due to evaporation carries some extra thermal energy out of the planet, thereby enhancing cooling. For the "normal" EUV evaporation, the expectation is that this is a small effect, hence using a fixed T (τ) relation would not introduce significant errors. However, once the escape enters the "boil-off" regime (e.g. Owen & Wu 2016), then our cur- rent boundary condition would not be suitable. Regardless of the root cause, (numerical or physical in na- ture), our main conclusion is unchanged. The error intro- duced by using mass-composition-radius isochrons calculated for planets neglecting mass loss is typically small (. 1% in planet radius). 4.5. A Favored % H/He mass fraction? it Based on the analysis of the radius distribution of a subset of Kepler planet candidates, Wolfgang & Lopez (2015) found a typical H/He mass fraction of 0.7% (with a standard devi- ation of ∼0.6 dex). Ultimately the compositions of planets observed today are consequences of both the initial formation and subsequent evolution. What role might evaporative planet mass loss play in producing this 0.7% typical H/He envelope mass fraction? Our simulations of the coupled thermal and mass-loss evolution of low-mass planets show evidence for a non- monotonic relation between planet composition and envelope survival rate. Similar behavior has also been noted in previous planet evaporation parameter studies (Lopez & Fortney 2013; Jin et al. 2014). In Figure 4, we see that, by ages of 1 Gyr, low mass planets (. 6 M⊕) with high (∼20%) or very low (∼0.05 - 0.1%) envelope mass fractions tend to have lower survival rates. Envelope mass fraction values in the "inter- mediate" range seem to have a greater probability of survival. This is seen in Figure 4 by the fact that the mass-radius curves for fenv = 0.05 - 2% extend to lower masses than the mass ra- dius curves of both higher and lower envelope fractions. This could provide a mechanism to imprint a preferred envelope mass on the planet population. To more quantitatively examine the situation, is il- luminating to consider the planet envelope mass loss timescale, defined as τenv = Menv/ Mp (e.g., Rogers et al. 2011; Batygin & Stevenson 2013). This quantity provides an instan- taneous measure of how long a planet envelope could "sur- vive" at its current mass loss rate. For this experiment, we calculate τenv at early ages of 100 Myr. We present in Fig- ure 6, envelope mass loss timescales for planets at 0.1 AU. In our simulations of planets with rocky cores, the mass loss timescale (for specified planet mass) is maximized at an intermediate value of fenv = 1- 2%. At smaller envelope mass fractions ( fenv = 0.05 and 0.1%), lifetimes are shorter because there is less envelope to lose. At higher envelope mass frac- tions above 5%, planet radii increases the mass loss rate as the energy limited escape is sensitive to the cross-sectional radius (energy-limited ∝ R3 p). Furthermore, planets that loses a sig- nificant fraction of their total mass early on would likely com- pletely evaporate before reaching ages of 1 Gyr (Lopez et al. 2012). This may lead to convergent evolution behavior, where planets end up with similar final envelope masses fractions (near where the mass loss timescale is maximized) for a wide range of initial envelope mass fractions. This trend in our re- sults also indicates that planets with ∼ 1% H/He may "linger" longer at the composition, compared to other values of fenv. Once planets with high initial H/He fractions reaches this ∼1% value, they typically remain with similar values of en- velope fraction for & 5 Gyr. Upon closer examination of the top panel in Figure 6, the planet envelope mass fraction at which the mass loss timescale is highest increases with planet mass. Below 12 M⊕, the simulated planets with 1% H/He mass fraction maximize τenv. Above 12 M⊕, the value of fenv that max- imizes τenv switches to 5-6%. This hints that "fixed point" compositions to which planets converge may depend on the planet mass, potentially providing an observational diagnos- tic. Recall, though, that these τenv values correspond to the stellar fluxes at 0.1 AU. At further out orbital distances (& 0.1 AU), the entire distribution shifts upward (toward higher τenv). Conversely, at a closer-in distance, the envelope mass EVOLVING GASEOUS SUB-NEPTUNES WITH MESA 11 ] s r a e y [ e a c s e m l i i t l a v v r u s e H H / ] s r a e y [ e a c s e m l i i t l a v v r u s e H H / 1010 109 108 107 0 1010 109 108 107 106 0 a = 0.1 AU age = 100 Myr rocky interior 5 10 15 20 25 H/He mass fraction [%] a = 0.1 AU age = 100 Myr ice-rock interior 5 10 15 20 25 H/He mass fraction [%] Figure 6. Plot of instantaneous envelope mass-loss timescales, τenv, as a function of H/He fraction and planet mass at 0.1 AU. These simulations have Earth-like rocky (top panel) or ice-rock (bottom panel) heavy element interi- ors, and solar metallicity envelopes. The initial planet masses for the scatter points from bottom to top are 4.0 (blue), 6.0 (light blue), 8.0 (teal), 13 (green), 17 (red), 20 (white) M⊕ respectively. The τenv values presented are instanta- neous mass-loss rate at early times (100 Myr) when the stellar EUV is high. We see evidence of non-monotonic behavior of τenv with fenv. At this in- stant, the most "survivable" envelope fraction (at which τenv is maximized) is between 1 and 5%. loss timescales decreases across all values of planet masses, largely preserving the general shape of the distribution. For planets with lower density ice-rich cores (lower panel, Figure 6), we see a distinct upward shift in the value of fenv that maximizes τenv compared to planets with higher-density rocky cores. The icy interior models do not generate the same 1% signature but instead favor a higher set of "most sur- vivable" envelope mass fraction values (namely in the &5% range for planets & 5 M⊕). This shift is a consequence of how a lower density heavy element interior distorts the planet Rp - fenv relation. At masses below ∼13 M⊕, the ice-rock in- terior planets with 5 - 15% by mass H/He envelopes all have similar trends and near maximal values of τenv. At yet higher planet masses (above ∼13 M⊕), the ∼15% H/He composition takes over as the composition maximizing τenv. Compared with the rocky-core models, this take-over point occurs at a much lower planet mass. A major caveat in this study of τenv is that even if the H/He envelopes mass fractions of 1-5% surrounding rocky cores maximize the mass loss timescale at 100 Myr, we have not yet established whether this could have a substantial effect on the planet population, specifically whether this effect could plausibly reproduce the H/He mass fraction distribution noted by Wolfgang & Lopez (2015). We calculated τenv by taking instantaneous mass-loss rates to predict planet survival life- times. Are the differences in τenv sufficiently large relative to the duration of active mass loss to leave an observable im- print on the planet population? To what extent would the planet composition distribution be altered over the course of Gyr timespans? In the next section, we turn to evaluating the evolution in the H/He inventory of a larger synthetic popula- tion of planets. 4.6. Sampling a Simulated Planet Population In the previous section, we illustrated the effect of planet mass loss on the fenv distribution using instantaneous diagnos- tics at early times. We now synthesize all the implementations done so far to test the cumulative ramifications on planets in the present day. The planet population observed today (e.g., by Kepler) is a combination of both the initial outcomes from planet formation, and the subsequent effect of planet evolu- tion. In this section, focus solely on the latter process, iso- lating how planet evolution would affect an arbitrary initial planet composition distribution. We borrow the language of observational cosmology to ex- press the effects of planet mass-loss evolution on the distri- bution, g (Mcore, fenv,a,t) of planet masses and compositions, at orbital separation, a, and time, t. The transfer function T(cid:0)Mcore, fenv, fenv,0,a,t(cid:1) encapsulates how the initial planet mass-composition distribution output by the planet formation process, g (Mcore, fenv,a,t = 0), will be modified over time g(cid:0)Mp, fenv,a,t(cid:1) = Z T(cid:0)Mcore, fenv, fenv,0,a,t(cid:1)g(cid:0)Mcore, fenv,0,a,t = 0(cid:1)d fenv,0. Planet mass loss causes planets to evolve towards lower enve- lope mass fractions (and lower total masses) over time. We present in Figure 7 a snapshot in time (at 4.5 Gyr age) and orbital separations at 0.1, 0.07 and 0.25 AU of the planet evolution transfer function, marginalized over fenv,0. The dis- tribution of initial planet properties, g(cid:0)Mcore, fenv,0,a,t = 0(cid:1), were chosen to be uniformly distributed in log Mp-log fenv space; any peaks and deserts in Figure 7 are due solely to the subsequent consequences of mass-loss evolution. We consid- ered initial masses ranging from 1 M⊕ to 100 M⊕, and ini- tial fenv,0 from 0.001 to 0.90. For each orbital separation, we evolved 1200 planets with randomly generated conditions for 4.5 Gyr. Figure 7 presents final mass-composition distribution of the simulated planets, which effectively represent the dis- tortion of the initial logMp - log fenv distribution due to planet evolution. We analyze some observations of the 0.1 AU diagram (large panel in Figure 7). There is an over-density in the planet mass- composition distribution at envelope fractions of ≈ 10- 2.0 to 10- 1.5 and planet masses . 20 M⊕. Planets residing in this regime corresponds to those with compositions close to the most "optimal" longest-lived envelope fractions (with highest τenv). This is the manifestation over the cumulative history of the planets of the non-monotonic behavior of τenv with fenv highlighted in Section 4.5). 12 CHEN & ROGERS ] ⊕ M [ s s a m t e n a p l 0 1 g o l ] ⊕ M [ s s a m t e n a p l 0 1 g o l ] ⊕ M [ s s a m t e n a p l 0 1 g o l 0.07 AU 2.0 1.8 1.6 1.4 1.2 1.0 0.8 0.6 0.0 5 0 . 0 8 0 . 3 4 0 . 2 0 1 . 1 2 6 0.2 0.0 5 −4.5 −4.0 −3.5 −3.0 −2.5 −2.0 −1.5 −1.0 −0.5 0.0 2.0 1.8 1.6 1.4 1.2 1.0 0.8 0.6 0.1 AU 5 0 . 0 2 1 . 0 0 . 0 8 0.0 8 6 0 . 2 0.34 0.21 0.08 0.05 −4.5 −4.0 −3.5 −3.0 −2.5 −2.0 −1.5 −1.0 −0.5 0.0 0.25 AU 2 1 . 0 6 2 . 0 5 0 . 0 8 0 . 0 2.0 1.8 1.6 1.4 1.2 1.0 0.8 0.6 −4.5 −4.0 −3.5 −3.0 −2.5 −2.0 −1.5 log10 H/He mass fraction 0 . 1 2 0.34 0.12 1 0.2 0.05 0.08 −1.0 −0.5 0.0 Figure 7. Final mass-composition distribution of simulated mass-losing planets at 4.5 Gyr (top panel: 0.07 AU, middle panel: 0.1 AU, bottom panel 0.25 AU). The distribution of initial planet properties were chosen to be uniformly distributed in log Mp-log fenv space, with initial masses ranging from 1 M⊕ to 100 M⊕, and initial fenv,0 from 0.001 to 0.95. We carried out calculations for 1200 simulated planets with randomly generated initial conditions for 4.5 Gyr. The final mass-composition distribution obtained from Gaussian kernel density estimation, is indicated by color-shading and contours, and represents a measure of the transfer function (at 4.5 Gyr), marginalized over fenv,0. EVOLVING GASEOUS SUB-NEPTUNES WITH MESA 13 = 56.1 % = 33.1 % 1-10 M ⊕ 10-30 M ⊕ 30-100 M ⊕ = 10.93 % 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 e c n e r u c c o f o y t i l i b a b o r p 0.0 −5.0 −4.5 −4.0 −3.5 −2.0 log10 H/He mass fraction −3.0 −2.5 −1.5 −1.0 −0.5 Figure 8. Synthetic planet probability-composition distribution at 4.5 Gyr and 0.1 AU. The initial fenv,0 distribution assumed is flat in log fenv,0 (between fenv,0 of 0.001 to 0.90). The planet mass distribution is sampled from the RV mass distribution of Howard et al. (2010); the proportion of simulated planets within each mass bin is listed in the upper left corner of the plot. This 1D histogram is effectively Figure 7, marginalized over planet mass, weighted by the California Planet Search radial velocity mass distribution. The 0.1 AU diagram in Figure 7 also displays a paucity of planets with masses . 17 M⊕ surviving at 4.5 Gyr with fenv . 10- 3%. This is due to the short envelope mass loss timescales in this regime (because there is little H/He to lose). For planets with masses above ∼16 M⊕ (101.2), the "distor- tion" of the original distribution g(cid:0)Mp, fenv,0,a,t = 0(cid:1) is less significant compared to the low mass regime and even less so for further orbital separations (compare the degree of distor- tion for the 0.07 AU and the 0.25 AU panels). Though there is an apparent dearth of planets with fenv . 0.001, this desert arises as a pure artifact due to the way in which we set the lower bound initial compositions ( fenv ≥ 0.001). These re- sults reflect the fact that planets in this regime are less likely to lose mass at this specific orbital distance, Fp ≈ 100 F⊕. Planet interior structure and evolution may also lead to low mass (Mp . 10 M⊕) and high envelope mass fractions ( fenv & 10%) being rare. This is visible in Figure 7 as an underdensity in the lower right-hand corner. This desert arises from the combined consequence of the massive mass-loss rates due to large planet cross-sectional area as well as the instability of MESA computations at even greater sizes (or fenv). In the latter cases, planets in this regimes can be created but unable to be evolved past ages of ∼5 - 10 Myr. Simulations with extremely low mean densities such as these cause the H/He envelopes to be "unbounded" over a very short period of time (even without mass loss). The low-Mp-high- fenv boundary in the lower right corner of Figure 7 corresponds to the same boundary in the upper left corner of Figure 2. Finally, we recover the convergent behavior of evolution tracks, in which points (if imagined as vector fields) move toward the lower left, along a constant core mass curve (not shown in figure). Particularly, a range of initial % H/He values from 1 to 10% ended up with a similar composition at 4.5 Gyr ( fenv∼0.8%). Ultimately, the transfer function (Figure 7) is convolved with the initial distribution of planet properties (immediately after formation) to yield the current planet mass-composition distribution observed today. As a proof of concept, we sample from the mass distribution of radial velocity planets from the California Planet Search reported by Howard et al. (2010) to define a planet mass distribution (still assuming a flat distribu- tion of initial log fenv). The resulting composition distribution of planets at 4.5 Gyr is shown in Figure 8. With this more real- istic initial planet mass distribution, we can see a distinct peak of simulated planet occurrence at about 1% H/He mass frac- tion. This result suggests that evolution via evaporation may partly explain the compositional distribution of sub-Neptune sized planets found by Wolfgang & Lopez (2015). In the past two sections, we have shown that photo- evaporation effects may, in certain regimes, lead to a fa- vored envelope mass fraction where the envelope mass-loss timescale is optimized. However, we have focused only on evolutionary processes, encapsulated in the transfer function. The eventuality of planet evolution is also strongly shaped by the choices of initial masses and compositions. While the ini- tial planet distribution we assumed here is an oversimplifi- cation, it nonetheless provides an encouraging demonstration that evolutionary processes play a large role in sculpting the planets observed today. 5. DISCUSSION 5.1. Insights into the Kepler Planet Population Planet interior structure and evolution calculations can pro- vide insights into the observed distribution of Kepler planet properties. Several planet modelers have already noted fea- tures that photo-evaporation could produce in the radius-flux distribution of close-in planets, specifically a declining occur- rence of sub-Neptune-size planets with increasing irradiation as vulnerable low-density planets lose their envelopes (e.g., Lecavelier Des Etangs 2007; Lopez et al. 2012), and an "oc- currence valley" in the planet distribution between ∼ 1.5 R⊕ and ∼ 2.5 R⊕ between the populations of planets that have re- tained their volatile envelopes and the population of remnant evaporated cores (e.g., Owen & Wu 2013; Lopez & Fortney 2013; Mordasini et al. 2012; Jin et al. 2014). Our planets evo- lution simulations with MESA show good agreement (as ex- pected) with these previously reported trends in the dividing line between complete and incomplete evaporation. We have proposed yet another potential observable signa- ture of evaporation in the close-in planet population, in Sec- tions 4.5 and 4.6. Our simulations have hinted that the typical ∼ 1% H/He mass fraction inferred from the Kepler radius dis- tribution by Wolfgang & Lopez (2015) could potentially be due to convergent evolution produced by evaporative planet mass loss. This feature appears among the population of plan- ets that suffer significant but incomplete evaporation. With a toy model for the initial planet mass-composition distribution, we have provided an illustrative proof of concept that this ef- fect can have an observable influence sculpting the eventual fenv distribution at Gyr ages. Further work is needed to more fully quantify the effect of convergent photo-evaporative evo- lution on the observed Kepler population, including models incorporating a more sophisticated treatment of planet mass- loss, a fully de-biased joint radius-period distribution of Ke- pler planet candidates, and robust statistical methods to com- pare models and observations. Ultimately, Kepler measured planet radii (transit depths) and it is to the distribution of planet radii that any model should be compared. In Figure 9 we present the distribution of planet radii associated with the synthetic planet popula- tion at 0.1 AU in Figure 8 along with synthetic planet radius 14 CHEN & ROGERS l ] s t e n a p d e t a u m i s f o # l [ e c n e r u c c o 200 150 100 50 0 1 2 3 4 = 10.93% = 33.1% = 56.1% 1-10 M ⊕ 10-30 M ⊕ 30-100 M ⊕ Red = 0.07 AU Blue = 0.1 AU Green = 0.25 AU 600 500 400 300 200 100 Black histrogram = Petigura et al. (2013) Figure 3 l ] s t e n a p d e t a u m i s f o # l [ e c n e r u c c o 9 10 11 12 0 1.0 8 ⊕ ] 2.0 4.0 planet radius [R ] ⊕ 8.0 16.0 5 6 planet radius [R 7 Figure 9. Synthetic planet radius distributions at 4.5 Gyr on a linear radius scale (left) and logarithmic radius scale (right). The histogram colors indicate different orbital separations: 0.07 AU (red), 0.10 AU (blue) and 0.25 AU (green). A total of 1200 planets were simulated at each orbital separation. The solid lines represent planet models that have retained some H/He in their envelopes, while dotted lines correspond to evaporated planet cores. The blue histogram at 0.10 AU corresponds to the same suite of simulations presented in Figure 8. The bottom panel is plotted to ease comparison with the Kepler radius occurrence rates from Petigura et al. (2013a) (black step-curve). distributions at 0.07 AU and 0.25 AU, generated following an identical approach. Our synthetic population shows the occurrence valley between evaporated cores and volatile rich planets (between 1 and 2 R⊕) that was found in previous the- oretical works. The planet radius corresponding to the valley decreases with increasing orbital separation. Like in the ob- served Kepler radius distribution from Petigura et al. (2013a), our synthetic radius distributions display a occurrence peak in the 2 to 2.83 R⊕ bin (when logarithmic radius bins are employed). On the other hand, Petigura et al. (2013a) see a much steeper decline in planet occurrence between the 2- 2.8 R⊕ and 2.8-4 R⊕ bins than is present in our synthetic population. We emphasize however, that we assumed one of the simplest and broadest initial Mp - fenv distributions possible (flat in log fenv, and no correlation between initial Mp and fenv); we made no attempt to tune the initial mass- composition distribution to fit the observed Kepler radius dis- tribution. The steeper observed decline around 3 R⊕ could be a consequence of a narrower distribution of initial composi- tions output from formation, evidence for a sub-population of water-rich planets, and/or correlations between planet initial planet mass and fenv. Indeed, planet formation models predict that correlations should exist between envelope mass and core mass (Bodenheimer & Lissauer 2014; Mordasini et al. 2014; Lee & Chiang 2015). Further statistical studies linking Kepler data to models are warranted to better disentangle the effects of evolution and the outcomes of formation. 5.2. Model Extension Opportunities We extended the MESA stellar evolution code to simu- late H/He envelopes surrounding low-mass planets in 1D. The physics that we incorporated into MESA is not neces- sarily new; rather, we followed common assumptions and ap- proaches employed in closed-source (proprietary) 1D planet evolution codes that are in use in the field (e.g. Valencia et al. 2010, Lopez & Fortney 2014, Kurokawa & Nakamoto 2014, and Howe & Burrows 2015). With our adaptations to MESA, there now exists a publicly available open-source code to sim- ulate H/He planets down to a few Earth masses. True planets are surely more complicated than the vanilla spherically symmetric, homogeneously layered scenarios simulated in this paper (and in prior works Valencia et al. 2010; Lopez & Fortney 2014; Kurokawa & Nakamoto 2014; Howe & Burrows 2015). These 1D simulations are, nonethe- less, the current workhorses of exoplanet evolution studies, and help to provide context to more complex and computa- tionally intensive models. There are a number of physical processes that could be added to MESA in future work, to improve upon its capa- bilities to model planets. These include, a treatment of composition gradients within planet interiors (e.g. double diffusive convection Nettelmann et al. 2015), using self-consistent model planet atmosphere grids to set the outer boundary conditions and planet cooling rates, (as in, e.g., Fortney et al. 2007), a more sophisticated treatment of atmospheric escape (includ- ing energy-limited scaling laws (Salz et al. 2015) and MHD effects), and finally the addition of a water EOS to facilitate the simulation of high mean molecular weight planetary en- velopes. It is our hope that the adaptations to MESA presented in this paper will help to provide the baseline groundwork for future applications of MESA to low-mass planets. 6. SUMMARY & CONCLUSIONS We summarize briefly below, the main outcomes and con- clusions of this work. • We implemented extensions to the MESA stellar evo- lution code that now permit MESA to simulate H/He envelopes surrounding planets down to a few Earth masses. These extensions include a thermo-physical model for planet heavy-element interiors, energy- limited and radiation-recombination limited mass loss, and an improved atmospheric boundary condition. • Coupled thermal and mass-loss evolution confirm that ultra low-density planets (with radii above 10 R⊕ and masses below 30 M⊕) can plausibly survive for mul- tiple Gyr timescales, though over a narrower range of parameter space that predicted by instantaneous mass loss timescale criteria (Batygin & Stevenson 2013). EVOLVING GASEOUS SUB-NEPTUNES WITH MESA 15 and Frank Timmes- without whom this project would not have been a possibility. • For mass-losing planets even at close orbital distances . 0.1 AU, radius is a proxy for the planet's current composition. • Planet radii typically show very little hysteresis. The systematic error introduced by applying planet isochrons calculated neglecting mass loss to define a mapping from planet mass, composition, and age to ra- dius for evaporating planets is typically small (. 1% in planet radius), with the exception of models with very high envelope fractions. • Planet envelope mass loss timescales, τenv vary non- monotonically with fenv (at fixed planet mass). In our simulations of young (100 Myr) low-mass (Mp . 10 M⊕) planets with rocky cores, τenv is maximized at fenv = 1% to 3%. The resulting convergent evo- lution could potentially imprint itself on the close-in planet population as a preferred H/He mass fraction of ∼ 1% (as inferred from the Kepler radius distribution by Wolfgang & Lopez 2015) With a succession of space-based exoplanet transit surveys on the horizon (K2, TESS, CHEOPS and PLATO) combined with improving resolution and stability of ground-based spec- trographs (e.g., SPIRou, Keck SHREK, EXPRES, Carmenes, HPF, ESPRESSO, G-CLEF, HiJaK), our purview of exoplan- etary systems is bound to expand vastly in the years to come. For this reason, there is a need for a fast yet robust series of modeling and computational schemes to complement obser- vational measurements. Recent years has seen a near simul- taneous rise in the use of the stellar evolution code MESA and planetary mass loss evolution studies. In this article, we provide a suite of basic planet models from which more com- plicated studies can be built; this is again eased by the open- source nature of MESA. Looking ahead, we see this wonder- ful evolution code acting as a complement to more sophis- ticated 3-D models to interpret the measurements of future space missions and exoplanetary surveys. We thank Dr. James Owen for assisting us with the atmo- spheric boundary conditions and Professor Phil Arras for pro- viding us with helpful templates that enhanced the efficiency of simulating MESA planets. We also thank Dr. Eric Lopez for great suggestions to our manuscript. H.C. acknowledges the Undergraduate Research Opportunities Program (UROP) at Boston University for funding this research while in res- idence at Caltech during the summer of 2014 and through- out the academic year. The authors thank Professors Philip Muirhead and Heather Knutson for facilitating the summer re- search. L.A.R. gratefully acknowledges support provided by NASA through Hubble Fellowship grant #HF-51313 awarded by the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., for NASA, under contract NAS 5-26555. This work was performed in part under contract with the Jet Propulsion Lab- oratory (JPL) funded by NASA through the Sagan Fellow- ship Program executed by the NASA Exoplanet Science In- stitute. Most of the calculations have made use of H.C.'s MSI GE70 APACHE laptop, which was supported by his parents to pursue his interest in computational astrophysics and plan- etary science. Lastly, we express gratitude toward the MESA code creators- Dr. Bill Paxton, and Professors Lars Bildsten REFERENCES Baraffe, I., Chabrier, G., & Barman, T. 2008, A&A, 482, 315 Batalha, N. M., Rowe, J. F., Bryson, S. T., et al. 2013, ApJS, 204, 24 Batygin, K., & Stevenson, D. J. 2013, ApJ, 769, L9 Becker, J. C., & Batygin, K. 2013, ApJ, 778, 100 Bodenheimer, P., & Lissauer, J. J. 2014, ApJ, 791, 103 Borucki, W. J., Koch, D. G., Basri, G., et al. 2011, ApJ, 728, 117 Burke, C. J., Bryson, S. T., Mullally, F., et al. 2014, ApJS, 210, 19 Burke, C. J., Christiansen, J. L., Mullally, F., et al. 2015, ApJ, 809, 8 Carter, J. A., Agol, E., Chaplin, W. J., et al. 2012, Science, 337, 556 Chadney, J. M., Galand, M., Unruh, Y. C., Koskinen, T. T., & Sanz-Forcada, J. 2015, Icarus, 250, 357 Erkaev, N. V., Kulikov, Y. N., Lammer, H., et al. 2007, A&A, 472, 329 Fortney, J. J., Marley, M. S., & Barnes, J. W. 2007, ApJ, 659, 1661 Freedman, R. S., Lustig-Yaeger, J., Fortney, J. J., et al. 2014, ApJS, 214, 25 Freedman, R. S., Marley, M. S., & Lodders, K. 2008, ApJS, 174, 504 Fressin, F., Torres, G., Charbonneau, D., et al. 2013, ApJ, 766, 81 Guillot, T. 2010, A&A, 520, A27 Guillot, T., Chabrier, G., Gautier, D., & Morel, P. 1995, ApJ, 450, 463 Han, E., Wang, S. X., Wright, J. T., et al. 2014, PASP, 126, 827 Hartmann, W. 2004, Brooks Cole Howard, A. W., Marcy, G. W., Johnson, J. A., et al. 2010, Science, 330, 653 Howard, A. W., Marcy, G. W., Bryson, S. T., et al. 2012, ApJS, 201, 15 Howe, A. R., & Burrows, A. 2015, ApJ, 808, 150 Howe, A. R., Burrows, A., & Verne, W. 2014, ApJ, 787, 173 Jackson, A. P., Davis, T. A., & Wheatley, P. J. 2012, MNRAS, 422, 2024 Jackson, B., Jensen, E., Peacock, S., Arras, P., & Penev, K. 2016, ArXiv e-prints, arXiv:1603.00392 Jin, S., Mordasini, C., Parmentier, V., et al. 2014, ApJ, 795, 65 Jontof-Hutter, D., Lissauer, J. J., Rowe, J. F., & Fabrycky, D. C. 2014, ApJ, 785, 15 Kasting, J. F., & Pollack, J. B. 1983, Icarus, 53, 479 Kurokawa, H., & Nakamoto, T. 2014, ApJ, 783, 54 Lammer, H., Selsis, F., Ribas, I., et al. 2003, ApJ, 598, L121 Lecavelier Des Etangs, A. 2007, A&A, 461, 1185 Lee, E. J., & Chiang, E. 2015, ApJ, 811, 41 Lissauer, J. J., Fabrycky, D. C., Ford, E. B., et al. 2011, Nature, 470, 53 Lissauer, J. J., Jontof-Hutter, D., Rowe, J. F., et al. 2013, ApJ, 770, 131 Lopez, E. D., & Fortney, J. J. 2013, ApJ, 776, 2 -. 2014, ApJ, 792, 1 Lopez, E. D., Fortney, J. J., & Miller, N. 2012, ApJ, 761, 59 Luger, R., Barnes, R., Lopez, E., et al. 2015, Astrobiology, 15, 57 Marcy, G., Isaacson, H., Howard, A. W., et al. 2014, ApJS, 210, 20 Masuda, K. 2014, ApJ, 783, 53 Mcley, L., & Soker, N. 2014, MNRAS, 445, 2492 Mordasini, C., Alibert, Y., Georgy, C., et al. 2012, A&A, 547, A112 Mordasini, C., Klahr, H., Alibert, Y., Miller, N., & Henning, T. 2014, A&A, Mullally, F., Coughlin, J. L., Thompson, S. E., et al. 2015, ApJS, 217, 31 Murray-Clay, R. A., Chiang, E. I., & Murray, N. 2009, ApJ, 693, 23 Nettelmann, N., Fortney, J. J., Kramm, U., & Redmer, R. 2011, ApJ, 733, 2 Nettelmann, N., Fortney, J. J., Moore, K., & Mankovich, C. 2015, MNRAS, 566, A141 447, 3422 Owen, J. E., & Alvarez, M. A. 2016, ApJ, 816, 34 Owen, J. E., & Jackson, A. P. 2012, MNRAS, 425, 2931 Owen, J. E., & Wu, Y. 2013, ApJ, 775, 105 -. 2016, ApJ, 817, 107 Paxton, B., Bildsten, L., Dotter, A., et al. 2011, ApJS, 192, 3 Paxton, B., Cantiello, M., Arras, P., et al. 2013, ApJS, 208, 4 Paxton, B., Marchant, P., Schwab, J., et al. 2015, ApJS, 220, 15 Perna, R., Duffell, P., Cantiello, M., & MacFadyen, A. I. 2014, ApJ, 781, Petigura, E. A., Howard, A. W., & Marcy, G. W. 2013a, Proceedings of the National Academy of Science, 110, 19273 Petigura, E. A., Marcy, G. W., & Howard, A. W. 2013b, ApJ, 770, 69 Ribas, I., Guinan, E. F., Güdel, M., & Audard, M. 2005, ApJ, 622, 680 Rogers, L. A. 2015, ApJ, 801, 41 Rogers, L. A., Bodenheimer, P., Lissauer, J. J., & Seager, S. 2011, ApJ, 738, Rogers, L. A., & Seager, S. 2010, ApJ, 712, 974 Rowe, J. F., Bryson, S. T., Marcy, G. W., et al. 2014, ApJ, 784, 45 Salz, M., Schneider, P. C., Czesla, S., & Schmitt, J. H. M. M. 2015, ArXiv e-prints, arXiv:1511.09348 Sanchis-Ojeda, R., Fabrycky, D. C., Winn, J. N., et al. 2012, Nature, 487, Saumon, D., Chabrier, G., & van Horn, H. M. 1995, ApJS, 99, 713 Tripathi, A., Kratter, K. M., Murray-Clay, R. A., & Krumholz, M. R. 2015, ApJ, 808, 173 Valencia, D., Ikoma, M., Guillot, T., & Nettelmann, N. 2010, A&A, 516, Valsecchi, F., Rappaport, S., Rasio, F. A., Marchant, P., & Rogers, L. A. 2015, ApJ, 813, 101 119 59 449 A20 16 CHEN & ROGERS Valsecchi, F., Rasio, F. A., & Steffen, J. H. 2014, ApJ, 793, L3 Vidotto, A. A., Fares, R., Jardine, M., Moutou, C., & Donati, J.-F. 2015, MNRAS, 449, 4117 Watson, A. J., Donahue, T. M., & Walker, J. C. G. 1981, Icarus, 48, 150 Weiss, L. M., & Marcy, G. W. 2014, ApJ, 783, L6 Wolf, W. M., Bildsten, L., Brooks, J., & Paxton, B. 2013, ApJ, 777, 136 Wolfgang, A., & Lopez, E. 2015, ApJ, 806, 183 Wolfgang, A., Rogers, L. A., & Ford, E. B. 2015, ArXiv e-prints, arXiv:1504.07557 Wu, Y., & Lithwick, Y. 2013, ApJ, 763, 13 EVOLVING GASEOUS SUB-NEPTUNES WITH MESA 17 Planetary Radii Table of Jovian and Sub-Jovian Gas Giants with 50% Ice - 50%Rock Cores Table 1 Separation Core Mass Age (Gyr) 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 4.5 4.5 4.5 4.5 4.5 4.5 4.5 4.5 4.5 4.5 4.5 4.5 4.5 4.5 4.5 4.5 4.5 4.5 4.5 4.5 (AU) 0.02 0.02 0.02 0.02 0.02 0.045 0.045 0.045 0.045 0.045 0.1 0.1 0.1 0.1 0.1 1.0 1.0 1.0 1.0 1.0 0.02 0.02 0.02 0.02 0.02 0.045 0.045 0.045 0.045 0.045 0.1 0.1 0.1 0.1 0.1 1.0 1.0 1.0 1.0 1.0 0.02 0.02 0.02 0.02 0.02 0.045 0.045 0.045 0.045 0.045 0.1 0.1 0.1 0.1 0.1 1.0 1.0 1.0 1.0 1.0 (M⊕) 0.0 10.0 25.0 50.0 100.0 0.0 10.0 25.0 50.0 100.0 0.0 10.0 25.0 50.0 100.0 0.0 10.0 25.0 50.0 100.0 0.0 10.0 25.0 50.0 100.0 0.0 10.0 25.0 50.0 100.0 0.0 10.0 25.0 50.0 100.0 0.0 10.0 25.0 50.0 100.0 0.0 10.0 25.0 50.0 100.0 0.0 10.0 25.0 50.0 100.0 0.0 10.0 25.0 50.0 100.0 0.0 10.0 25.0 50.0 100.0 0.0535 0.0881 0.115 0.242 0.406 0.676 1.0 1.46 2.44 4.07 6.78 11.31 nan 13.17 nan nan nan nan 9.23 nan nan nan nan 8.25 nan nan nan nan 6.98 nan nan nan nan 11.24 nan nan nan nan 8.2 nan nan nan nan 7.39 nan nan nan nan 6.41 nan nan nan nan 9.43 nan nan nan nan 7.24 nan nan nan nan 6.56 nan nan nan nan 5.86 nan nan nan nan 15.37 5.12 nan nan nan 11.34 4.62 nan nan nan 10.41 4.47 nan nan nan 8.99 4.23 nan nan nan 13.1 4.83 nan nan nan 10.14 4.42 nan nan nan 9.37 4.27 nan nan nan 8.27 4.09 nan nan nan 11.03 4.6 nan nan nan 8.98 4.23 nan nan nan 8.34 4.09 nan nan nan 7.6 3.93 nan nan nan 15.26 8.28 nan nan nan 11.8 7.17 nan nan nan 11.02 6.88 nan nan nan 9.69 6.35 nan nan nan 13.22 7.59 nan nan nan 10.67 6.7 nan nan nan 10.0 6.44 nan nan nan 8.97 6.03 nan nan nan 11.32 6.91 nan nan nan 9.54 6.24 nan nan nan 8.98 6.0 nan nan nan 8.29 5.72 nan nan nan 14.57 11.95 8.18 nan nan 12.45 10.51 7.5 nan nan 12.02 10.2 7.34 nan nan 11.05 9.51 6.99 nan nan 13.19 11.0 7.72 nan nan 11.58 9.87 7.16 nan nan 11.18 9.59 7.01 nan nan 10.39 9.01 6.72 nan nan 11.79 10.0 7.2 nan nan 10.62 9.16 6.77 nan nan 10.29 8.91 6.64 nan nan 9.77 8.53 6.45 nan 15.32 14.22 12.79 10.7 6.73 13.56 12.71 11.57 9.84 6.4 13.24 12.42 11.33 9.67 6.33 12.44 11.7 10.73 9.23 6.15 14.1 13.18 11.95 10.1 6.48 12.76 12.01 10.99 9.41 6.21 12.45 11.74 10.76 9.24 6.14 11.76 11.11 10.23 8.86 5.99 12.86 12.1 11.05 9.44 6.2 11.83 11.2 10.3 8.89 5.98 11.57 10.96 10.11 8.75 5.92 11.1 10.53 9.75 8.49 5.83 14.47 13.87 13.15 11.98 9.8 13.28 12.82 12.16 11.14 9.22 13.08 12.63 12.0 11.0 9.12 12.53 12.1 11.52 10.6 8.84 13.61 13.12 12.44 11.37 9.37 12.68 12.26 11.66 10.71 8.91 12.49 12.08 11.49 10.57 8.81 11.98 11.6 11.06 10.2 8.56 12.72 12.3 11.69 10.73 8.9 11.98 11.61 11.07 10.2 8.54 11.81 11.45 10.92 10.08 8.45 11.44 11.1 10.61 9.82 8.28 14.09 13.79 13.33 12.6 11.14 13.14 12.87 12.47 11.81 10.51 13.0 12.73 12.34 11.68 10.41 12.58 12.31 11.94 11.33 10.13 13.37 13.11 12.69 12.0 10.66 12.62 12.38 12.0 11.38 10.16 12.48 12.24 11.87 11.26 10.07 12.08 11.85 11.5 10.93 9.79 12.64 12.4 12.02 11.39 10.16 12.04 11.83 11.48 10.9 9.76 11.91 11.7 11.36 10.79 9.68 11.6 11.4 11.08 10.53 9.47 14.0 13.83 13.54 13.04 12.05 13.14 12.98 12.73 12.28 11.39 13.02 12.87 12.62 12.18 11.3 12.68 12.53 12.29 11.88 11.05 13.33 13.18 12.91 12.45 11.54 12.65 12.52 12.28 11.86 11.02 12.54 12.41 12.17 11.76 10.93 12.19 12.06 11.84 11.45 10.67 12.65 12.52 12.28 11.86 11.02 12.12 12.01 11.79 11.4 10.61 12.02 11.91 11.69 11.31 10.53 11.73 11.62 11.42 11.06 10.31 14.02 13.95 13.8 13.51 12.91 13.19 13.14 13.01 12.76 12.22 13.1 13.05 12.92 12.67 12.14 12.86 12.81 12.69 12.45 11.94 13.36 13.31 13.18 12.91 12.36 12.72 12.68 12.57 12.33 11.82 12.63 12.59 12.48 12.25 11.74 12.34 12.3 12.19 11.97 11.5 12.71 12.68 12.56 12.33 11.82 12.22 12.2 12.09 11.88 11.4 12.14 12.11 12.01 11.8 11.33 11.87 11.84 11.75 11.55 11.1 14.0 14.04 13.98 13.82 13.47 13.16 13.2 13.15 13.02 12.71 13.08 13.11 13.07 12.94 12.63 12.94 12.97 12.93 12.81 12.51 13.34 13.36 13.31 13.18 12.86 12.68 12.72 12.69 12.57 12.28 12.6 12.64 12.61 12.5 12.21 12.38 12.41 12.38 12.28 12.01 12.68 12.71 12.68 12.57 12.28 12.19 12.23 12.21 12.12 11.85 12.11 12.16 12.14 12.05 11.78 11.87 11.91 11.9 11.81 11.56 13.87 13.98 13.98 13.92 13.73 12.94 13.05 13.07 13.03 12.88 12.86 12.96 12.98 12.95 12.8 12.82 12.91 12.94 12.9 12.76 13.14 13.22 13.24 13.2 13.04 12.44 12.54 12.58 12.55 12.42 12.37 12.47 12.5 12.48 12.35 12.24 12.33 12.37 12.35 12.22 12.44 12.54 12.57 12.55 12.41 11.94 12.05 12.1 12.09 11.97 11.88 11.98 12.03 12.02 11.91 11.67 11.77 11.82 11.82 11.71 13.55 13.69 13.74 13.75 13.68 12.53 12.69 12.76 12.79 12.75 12.43 12.58 12.66 12.69 12.66 12.41 12.56 12.64 12.68 12.64 12.73 12.86 12.93 12.96 12.92 11.99 12.14 12.23 12.27 12.25 11.91 12.06 12.15 12.2 12.18 11.87 12.02 12.11 12.15 12.13 11.99 12.13 12.22 12.27 12.24 11.47 11.64 11.74 11.8 11.79 11.41 11.57 11.68 11.74 11.73 11.28 11.44 11.54 11.6 11.6 Note. - Top row are total mass values in MJup. The planet radii values in R⊕. The specific mass and flux values were chosen to ease comparison with Fortney et al. (2007). APPENDIX 18 CHEN & ROGERS Planetary Radii Table of Sub-Neptune Mass Planets with Earth-like Rocky Interiors at 10 Gyr Table 2 Stellar Flux (F⊕) 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 10.0 10.0 10.0 10.0 10.0 10.0 10.0 10.0 10.0 1000.0 1000.0 1000.0 1000.0 1000.0 1000.0 1000.0 1000.0 1000.0 Planet Mass (M⊕) 1.0 1.5 2.0 2.4 3.6 5.5 8.5 13.0 20.0 1.0 1.5 2.0 2.4 3.6 5.5 8.5 13.0 20.0 1.0 1.5 2.0 2.4 3.6 5.5 8.5 13.0 20.0 0.01% 0.02% 0.05% 0.1% 0.2% 0.5% 1.0% 2.0% 5.0% 10% 20% 1.03 1.15 1.24 1.3 1.45 1.62 1.8 2.0 2.2 1.17 1.26 1.34 1.39 1.52 1.67 1.85 2.03 2.23 2.18 1.86 1.78 1.76 1.77 1.83 1.94 2.08 2.21 1.05 1.17 1.26 1.32 1.46 1.63 1.81 2.01 2.21 1.19 1.28 1.36 1.41 1.54 1.69 1.86 2.04 2.24 2.6 2.05 1.9 1.85 1.83 1.88 1.98 2.11 2.23 1.08 1.19 1.28 1.34 1.49 1.65 1.84 2.03 2.24 1.24 1.32 1.39 1.44 1.57 1.71 1.89 2.07 2.26 3.45 2.34 2.1 2.01 1.93 1.95 2.03 2.14 2.27 1.11 1.22 1.31 1.37 1.51 1.68 1.86 2.06 2.26 1.29 1.36 1.43 1.47 1.6 1.74 1.91 2.09 2.29 4.66 2.65 2.25 2.14 2.02 2.01 2.08 2.17 2.31 1.15 1.26 1.35 1.41 1.55 1.72 1.9 2.09 2.3 1.35 1.41 1.48 1.52 1.64 1.79 1.95 2.14 2.33 nan 3.06 2.49 2.3 2.13 2.1 2.14 2.22 2.36 1.23 1.34 1.43 1.49 1.63 1.8 1.98 2.18 2.37 1.47 1.52 1.57 1.62 1.73 1.87 2.04 2.22 2.42 nan 4.01 2.95 2.64 2.32 2.23 2.25 2.34 2.47 1.34 1.44 1.53 1.59 1.73 1.9 2.08 2.24 2.43 1.62 1.65 1.7 1.73 1.84 1.98 2.15 2.31 2.47 nan 5.56 3.54 3.06 2.59 2.42 2.41 2.48 2.6 1.5 1.6 1.68 1.74 1.88 2.05 2.2 2.51 nan 1.87 1.86 1.89 1.92 2.02 2.15 2.28 2.4 2.64 nan nan 4.7 3.8 2.99 2.7 2.65 2.68 2.79 1.87 1.96 2.03 2.09 2.23 2.32 2.59 2.83 nan nan nan 2.34 2.35 2.42 2.56 2.71 2.91 3.13 nan nan nan nan 4.05 3.4 3.18 3.17 3.25 2.38 2.43 2.5 2.55 2.69 2.87 3.08 3.31 nan nan nan nan 2.97 2.99 3.09 3.24 3.43 3.71 nan nan nan nan nan 4.48 3.99 3.84 3.87 3.27 3.25 3.29 3.33 3.45 3.62 3.83 4.07 nan nan nan nan nan 4.0 4.0 4.11 4.28 4.51 nan nan nan nan nan 6.8 5.45 5.0 4.89 Note. - The planet radii values in R⊕. Format and specific mass and flux values are taken to ease comparison with those from Lopez & Fortney (2014). Planetary Radii Table of Sub-Neptune Mass Planets with Ice-Rock Interiors at 10 Gyr Table 3 Stellar Flux Planet Mass 0.01% 0.02% 0.05% 0.1% 0.2% 0.5% 1.0% 2.0% 5.0% 10% 20% 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 10.0 10.0 10.0 10.0 10.0 10.0 10.0 10.0 10.0 1000.0 1000.0 1000.0 1000.0 1000.0 1000.0 1000.0 1000.0 1000.0 1.0 1.5 2.0 2.4 3.6 5.5 8.5 13.0 20.0 1.0 1.5 2.0 2.4 3.6 5.5 8.5 13.0 20.0 1.0 1.5 2.0 2.4 3.6 5.5 8.5 13.0 20.0 1.36 1.51 1.62 1.69 1.87 2.08 2.31 2.56 2.82 1.58 1.68 1.77 1.83 1.99 2.17 2.38 2.62 2.87 3.65 2.8 2.58 2.49 2.42 2.46 2.57 2.7 2.9 1.38 1.52 1.63 1.71 1.89 2.09 2.32 2.57 2.83 1.63 1.72 1.8 1.86 2.01 2.19 2.4 2.63 2.88 2.6 3.21 2.81 2.67 2.55 2.54 2.62 2.74 2.93 1.41 1.55 1.66 1.74 1.91 2.12 2.35 2.59 2.86 1.7 1.78 1.85 1.9 2.04 2.22 2.43 2.66 2.9 3.45 2.34 2.1 2.01 2.7 2.65 2.69 2.81 2.98 1.45 1.58 1.69 1.77 1.94 2.14 2.37 2.62 2.88 1.77 1.83 1.89 1.94 2.08 2.25 2.46 2.68 2.93 4.66 2.65 3.62 3.24 2.86 2.73 2.75 2.86 3.02 1.49 1.63 1.73 1.8 1.98 2.18 2.41 2.65 2.92 1.85 1.89 1.95 2.0 2.13 2.29 2.5 2.72 2.97 nan 3.06 4.16 3.6 3.05 2.85 2.84 2.93 3.07 1.58 1.71 1.81 1.88 2.05 2.26 2.49 2.73 3.0 2.0 2.01 2.06 2.1 2.22 2.38 2.58 2.8 3.05 nan 4.01 5.25 4.25 3.35 3.06 2.99 3.05 3.17 1.68 1.81 1.91 1.98 2.15 2.35 2.58 2.82 3.09 2.17 2.15 2.18 2.22 2.33 2.48 2.68 2.9 3.15 nan 5.56 6.79 5.04 3.69 3.26 3.14 3.18 3.29 1.83 1.95 2.05 2.12 2.29 2.49 2.72 2.97 3.2 2.44 2.37 2.37 2.4 2.49 2.64 2.83 3.06 3.27 nan nan 4.7 3.8 4.24 3.6 3.39 3.38 3.49 2.19 2.29 2.37 2.44 2.6 2.8 3.03 3.23 3.6 nan nan 2.82 2.82 2.87 3.0 3.18 3.42 3.68 nan nan nan nan 4.05 4.35 3.94 3.84 3.91 2.62 2.73 2.8 2.86 3.01 3.22 3.47 3.74 4.04 nan nan nan 2.97 3.39 3.49 3.66 3.89 4.15 nan nan nan nan nan 5.49 4.7 4.48 4.48 3.51 3.49 3.53 3.57 3.71 3.91 4.15 4.43 4.75 nan nan nan nan 4.0 4.35 4.47 4.66 4.93 nan nan nan nan nan 6.8 5.45 5.58 5.42 Note. - The planet radii values in R⊕. Format and specific mass and flux values are identical to those from Table 2. EVOLVING GASEOUS SUB-NEPTUNES WITH MESA 19 Best fit coefficients to analytically estimate the radius of a planet at specified mass, composition, irradiation flux, and age Table 4 using Equation 5. Coefficient Rocky-Core Planets Ice-Rock Core Planets c0 c1 c2 c3 c4 c12 c13 c14 c23 c24 c34 c11 c22 c33 c44 adjusted R2 rms error Rcore 0.131 ± 0.006 - 0.348 ± 0.008 0.631 ± 0.003 0.104 ± 0.006 - 0.179 ± 0.005 0.028 ± 0.002 - 0.168 ± 0.002 0.008 ± 0.003 - 0.045 ± 0.001 - 0.036 ± 0.001 0.031 ± 0.002 0.209 ± 0.005 0.086 ± 0.001 0.052 ± 0.002 - 0.009 ± 0.002 0.993 0.035 ≈ 0.97M0.28 core 0.169 ± 0.008 - 0.436 ± 0.013 0.572 ± 0.005 0.154 ± 0.009 - 0.173 ± 0.007 0.014 ± 0.003 - 0.210 ± 0.003 0.006 ± 0.005 - 0.048 ± 0.001 - 0.040 ± 0.002 0.031 ± 0.002 0.246 ± 0.007 0.074 ± 0.001 0.059 ± 0.003 - 0.006 ± 0.003 0.987 0.047 ≈ 1.27M0.27 core i=1P4 Note. - The radius contribution of the planet envelope is fit with a quadratic model, log10 (cid:16) Rp- Rcore i=1 cixi + R⊕ (cid:17) = c0 +P4 j≥i ci jxix j, where x1 = log10 (cid:16) Mp 0.05(cid:17), M⊕ (cid:17), x2 = log10 (cid:16) fenv P4 x3 = log10 (cid:16) Fp 5 Gyr(cid:17). These fitting for- mulae are valid for 10- 4 ≤ fenv ≤ 0.2, 1 ≤ Mp/M⊕ ≤ 20, 4 ≤ Fp/F⊕ ≤ 400, and 100 Myr ≤ t ≤ 10 Gyr. F⊕ (cid:17), and x4 = log10 (cid:16) t
1302.3312	1	1302	2013-02-14T05:14:59	Proton Cyclotron Waves Upstream from Mars: Observations from Mars Global Surveyor	[ "astro-ph.EP", "physics.space-ph" ]	We present a study on the properties of electromagnetic plasma waves in the region upstream of the Martian bow shock, detected by the magnetometer and electron reflectometer (MAG / ER) onboard the Mars Global Surveyor (MGS) spacecraft during the period known as Science Phasing Orbits (SPO). The frequency of these waves, measured in the MGS reference frame (SC), is close to the local proton cyclotron frequency. Minimum variance analysis (MVA) shows that these 'proton cyclotron frequency' waves (PCWs) are characterized - in the SC frame - by a left-hand, elliptical polarization and propagate almost parallel to the background magnetic field. They also have a small degree of compressibility and an amplitude that decreases with the increase of the interplanetary magnetic field (IMF) cone angle and radial distance from the planet. The latter result supports the idea that the source of these waves is Mars. In addition, we find that these waves are not associated with the foreshock . Empirical evidence and theoretical approaches suggest that most of these observations correspond to the ion-ion right hand (RH) mode originating from the pick-up of ionized exospheric hydrogen. The left-hand (LH) mode might be present in cases where the IMF cone angle is high. PCWs occur in 62% of the time during SPO1 subphase, whereas occurrence drops to 8% during SPO2. Also, SPO1 PCWs preserve their characteristics for longer time periods and have greater degree of polarization and coherence than those in SPO2. We discuss these results in the context of possible changes in the pick-up conditions from SPO1 to SPO2, or steady, spatial inhomogeneities in the wave distribution. The lack of influence from the Solar Wind's convective electric field upon the location of PCWs indicates that there is no clear relation between the spatial distribution of PCWs and that of pick-up ions.	astro-ph.EP	astro-ph	Proton Cyclotron Waves Upstream from Mars: Observations from Mars Global Surveyor. N. Romanellia, C. Bertuccia, D. G´omeza, C. Mazelleb, M. Delvac aAstrophysical Plasmas, IAFE, Buenos Aires, Argentina. bG´eophysique Plan´etaire et Plasmas Spatiaux (GPPS), IRAP, Toulouse, France. cSpace Research Institute, Graz, Austria. Abstract We present a study on the properties of electromagnetic plasma waves in the region upstream of the Martian bow shock, detected by the magnetometer and electron reflectometer (MAG / ER) onboard the Mars Global Surveyor (MGS) spacecraft during the period known as Science Phasing Orbits (SPO). The frequency of these waves, measured in the MGS reference frame (SC), is close to the local proton cyclotron frequency. Minimum variance analy- sis (MVA) shows that these 'proton cyclotron frequency' waves (PCWs) are characterized - in the SC frame - by a left-hand, elliptical polarization and propagate almost parallel to the background magnetic field. They also have a small degree of compressibility and an amplitude that decreases with the increase of the interplanetary magnetic field (IMF) cone angle and radial dis- tance from the planet. The latter result supports the idea that the source of these waves is Mars. In addition, we find that these waves are not associated with the foreshock and their properties (ellipticity, degree of polarization, ✩ Corresponding author: Tel.: (5411) 4789-0179 int. 134. Fax: (5411) 4786-8114. Email address: [email protected] (N. Romanelli) Preprint submitted to Planetary and Space Science October 27, 2018 direction of propagation) do not depend on the IMF cone angle. Empirical evidence and theoretical approaches suggest that most of these observations correspond to the ion-ion right hand (RH) mode originating from the pick-up of ionized exospheric hydrogen. The left-hand (LH) mode might be present in cases where the IMF cone angle is high. PCWs occur in 62 % of the time during SPO1 subphase, whereas occurrence drops to 8% during SPO2. Also, SPO1 PCWs preserve their characteristics for longer time periods and have greater degree of polarization and coherence than those in SPO2. We discuss these results in the context of possible changes in the pick-up conditions from SPO1 to SPO2, or steady, spatial inhomogeneities in the wave distribution. The lack of influence from the Solar Wind's convective electric field upon the location of PCWs indicates that, as suggested by recent theoretical results, there is no clear relation between the spatial distribution of PCWs and that of pick-up ions. Keywords: Mars, Upstream, Cyclotron, Waves, Pick-up, Exosphere 1. Introduction Our Solar System is embedded in a plasma flow emanating from the Sun, known as the Solar Wind. The Solar Wind expands into the Solar System reaching supersonic speeds at a few solar radii from the Sun. The obstacles in its path can be classified into three types: absorbers (for example, un- magnetized asteroids or the Moon) where the Solar Wind interacts directly with their surface, obstacles with intrinsic magnetic fields (such as the Earth and Jupiter), and objects with no intrinsic magnetic field and whose atmo- spheres directly interact with the Solar Wind. In this group we find planets 2 like Venus and Mars, Saturn's satellite Titan, and also active comets. The interaction between the atmosphere of a non-magnetic obstacle and the Solar Wind can be described essentially as a non-collisional interaction between a magnetic plasma wind and a neutral cloud being ionized by solar radiation and charge-exchange with the Solar Wind plasma. An atmo- spheric obstacle such as Mars's atmosphere generates an induced magneto- sphere [Acuna et al., 1998; Bertucci et al., 2005], which has its origin in the exchange of energy and momentum between the Solar Wind and the plane- tary ions. The induced magnetosphere is preceded by a bow shock because of the supersonic nature of the Solar Wind. However, the interaction can start far beyond the bow shock because particles from the exosphere (mainly hydrogen) [Chaufray et al., 2008] are ionized several planetary radii away from the object. These particles are ionized mostly by photoionization and charge exchange [Modolo et al., 2005], which add a small amount of energy to the ions with respect to their parent neutrals. As the latter are approximately at rest with respect to the planet, the ions' planetocentric velocities are also considered to be negligible. These newborn ions start gyrating around the interplanetary magnetic field (IMF) while preserving their parents neutral's parallel velocity. In the planet's frame the gyrating ions also drift perpendicular to ~B and ~Ec( ~Ec = −~v × ~B). The physics of planetary ion pick-up is exactly the same as the one seen at comets and about which there is a vast literature [e.g., Mazelle and Neubauer [1993]; Tsurutani et al. [1989]; Tsurutani [1991]; Gary [1991], and references therein]. These newborn ions represent a non thermal component of the total ion distribution function which is unstable to the generation of electromag- 3 netic waves [Wu and Davidson, 1972; Wu and Hartle, 1974]. The instability and wave polarization resulting from the pick up process depends on the IMF cone angle αV,B, which is the angle between the solar wind velocity (VSW ) and the IMF at the time of pick-up [Tsurutani and Smith, 1986; Tsurutani et al., 1987]. If Vsw is parallel to the IMF, the newborn ions will form a beam in the solar wind frame and the electromagnetic ion-ion right-hand (RH) resonant instability will be predominant [Gary, 1993]. On the other hand, if Vsw is perpendicular to the IMF, the newborn ion distribution function will drive the electromagnetic ion-ion left-hand (LH) mode unstable. Both instabilities have maximum growth rates at ~k × ~B = 0, where ~k is the propagation wave- number. At moderate angles of αV,B the RH instability is still predominant. Brinca and Tsurutani [1989] found that the maximum growth rate of the LH instability is larger than that of the RH instability for αV,B > 75°, whereas, according to Convery and Gary [1997]; Gary and Madland [1988], this cutoff cone angle is αV,B = 90°. The ion-ion RH instability satisfies that the expression ω − ~k · ~vion // + Ωi is approximately zero for moderate αV,B [Gary et al., 1989; Brinca, 1991]. In this expression Ωi is the newborn ion gyrofrequency, ~vion is the ion drift // velocity along the magnetic field ~B, and ω and ~k are respectively the wave frequency and wave vector in the Solar Wind frame in the case of propaga- tion parallel to ~B. As the spacecraft has a negligible planetocentric velocity compared to that of the Solar Wind ~Vsw, we obtain the following expression: ωsc = ω − ~k · ~vsc // ; // = −[~Vsw · k] k ~vsc (1) where ωsc is the frequency of the plasma wave in the SC frame and k = ~k/~k. Since the frequency of the electromagnetic plasma wave in the newborn 4 ion reference frame is ωion = ω − ~k · ~vion // , and the velocity of the planetary particle before ionization is negligible with respect to ~Vsw, we obtain the useful expression: ωsc = −Ωi (2) where Ωi = (qiB/mi), with B = ~B , and q and m being the charge and mass of the newborn ion respectively. This means that the waves generated by the RH instability are characterized by a frequency which, in the SC frame, is close to the local ion cyclotron frequency with a left-handed polarization. When the parallel propagating ion-ion LH instability prevails, the Doppler correction is very small and therefore the waves will be observed with a left- hand polarization in both the plasma and SC frames at the local ion cyclotron frequency. This suggests that the occurrence of waves at the local ion cyclotron frequency of a particular ion species in the SC frame can be associated with the occurrence of the pick up of such ions. In this sense, the presence of plasma waves is, a priori, an important diagnostic tool for the evidence of ionized exospheric particles. The first observation of PCWs upstream from Mars' bow shock was made by Phobos-2 [Russell et al., 1990]. These waves had small amplitudes (∼ 0.15 nT), they were left-hand elliptically polarized in the SC frame, and propa- gated at a small angle to the mean magnetic field. PCWs have also been observed by Mars Global Surveyor (MGS) [Brain et al., 2002; Mazelle et al., 2004]. The frequency, polarization and propagation angle of the waves de- tected by MGS were similar to those determined from Phobos-2 observations, except their amplitude (2-3 times greater). Waves at the local proton cy- 5 clotron frequency have also been observed upstream from Venus [Delva et al., 2009] and active comets [Tsurutani [1991]; Mazelle and Neubauer [1993] and references therein]. The eccentric orbits of MGS during the mission's pre-mapping phase [Albee et al., 2001] allowed observations of PCWs up to 15 RM (1RM =3390 km: Mars radius). Brain et al. [2002] performed a statistical analysis of the properties of these waves during the first aerobraking (AB1) and science phas- ing orbit (SPO) phases. A few years later, Wei and Russell [2006] analyzed 85 events during the AB1 phase and discussed the generation mechanisms at large distances, as well as the possible distribution of waves depending on the direction of ~Ec = −~v × ~B. Similar analyses were presented by Delva et al. [2009], based on measurements provided by the magnetometer onboard the Venus Express spacecraft during 450 orbits. In this study we carry out an analysis of the PCWs detected by the magnetometer and the electron reflectometer (MAG/ER) onboard MGS in the region upstream from the Martian bow shock during the mission's SPO phase. We analyze the frequency, propagation and polarization properties of these waves and discuss the generation mechanisms and their relationship to the neutral densities at the exosphere of Mars. We also analyze the spatial distribution of these waves in a magneto-electric coordinate system centered on Mars (MBE). Finally, we study the implications of our results and compare them with recent studies around Mars [Wei and Russell, 2006] and Venus [Delva et al., 2011]. The article is structured as follows. In section (2) we describe the capabilities and limitations of MAG/ER in characterizing the properties of the PCW's, along with a description of the various methods of 6 analysis that we applied to the measurements. In section (3) we show typical examples of PCW's as detected by MAG/ER and obtain their properties based on the methods described in section (2). Next, we show statistical analyses of the amplitude and the spatial distribution of these waves, as well as the Solar Wind IMF cone angle associated with them. In section (4) we present a discussion of the results, and the theoretical approaches that might explain the generation of these waves. Finally in section (5) we summarize our conclusions. 2. Upstream waves: Analysis Methods MGS entered into orbit around Mars on 11 September 1997 [Albee et al., 2001]. During the pre-mapping AB and SPO phases, MGS provided measure- ments of the Martian environment from the unperturbed Solar Wind down to the neutral atmosphere from 1683 elliptical orbits. After these orbital phases, MGS reached a final circular mapping orbit at 400 km altitude. MGS carried a combination of a twin-triaxial fluxgate magnetometer system (MAG) and an electron spectrometer used as a reflectometer (ER) [Acuna et al., 1992]. MGS did not carry any instruments dedicated to the measurements of ion properties. The magnetometers (MAG) provided fast measurements (32 vectors/s) over a wide dynamic range (from ±4 nT to ±65536 nT), and the electron spectrometer (ER) measured the electron fluxes in 30 logarithmically spaced energy channels ranging from 10 eV to 20 keV with a maximum integration resolution of 2 s [Mitchell et al., 2001]. In this study we analyze the averaged fluxes over all directions with energies higher than 30 eV (below this energy value, the electron distribution function is 7 affected by spacecraft photoelectrons). The complete removal of spacecraft fields from the upstream MAG mea- surements was difficult, as the sensors were not mounted on a boom, but on the outer edge of each of the two solar panels. However, spacecraft field modeling allowed the reduction of their influence to ±1 nT [Acuna et al., 2001]. We used the MAG/ER measurements from 27 March 1998 through 24 September 1998. During this period, MGS performed 372 elliptical orbits grouped into two subphases: • From P202 (27/03/98) to P327 (27/05/98). [SPO1 sub-phase] • From P328 (27/05/98) to P573 (24/09/98). [SPO2 sub-phase] The P prefix indicates the periapsis number. The SPO orbits had a constant period of 11.6 hours with apoapses above the south pole at distances of roughly 6 RM . They also displayed a mono- tonic change in their local time from noon to 4AM. During the same period, the Martian season varied from early southern hemisphere summer to mid northern hemisphere spring. The SPO1 and SPO2 sub-phases arise as a re- sult of the presence of a hiatus due to a solar conjunction around July, 1998. As a result, SPO1 orbits had local times between noon and 10 AM, and SPO2 orbits had local times between 10 AM and 4AM. Also, SPO1 orbits are characterized by low to moderate solar zenith angles (SZA), whereas up- stream observations during SPO2 were closer to the terminator plane (see figure 1, Brain et al. [2002]). 8 The sampling frequency of MAG measurements analyzed here is fsamp = 1.33Hz, and due to the influence of spacecraft fields, the study is focused on periods where the strength of the IMF is equal to or larger than 4 nT. Analysis The analyses performed on MAG/ER data consisted in a characterization of their spectral properties, their polarization, their degree of coherence, their magnetic connection to Mars's bow shock, and their spatial distribution in an electromagnetic coordinate system. A more detailed description of them is given below. 2.1. Dynamic spectra and correlation We generate dynamic Fourier spectra of the magnetic field components in the Mars Solar Orbital (MSO) coordinate system. This coordinate system is centered at Mars with its ~XM SO axis pointing toward the Sun, ~ZM SO being perpendicular to Mars's orbital plane and positive to the ecliptic north, and while ~YM SO completing the right-hand system. We also calculate the cross correlation between the fluctuations present in the electron flux measurements for specific energy channels, and in the component of the magnetic field along its mean value. 2.2. Polarization - MVA The wave-vector and polarization of PCW's were obtained from minimum variance analysis (MVA) [Smith and Tsurutani, 1976; Khrabrov and Sonnerup, 1998]. This method provides an estimate of the direction of propagation for a plane wave by calculating the eigenvalues of the covariance matrix of the 9 magnetic field within each interval ( the maximum, intermediate and mini- mum eigenvalues are λ1, λ2 and λ3, respectively). Then, the wave-vector ~k is associated with the minimum variance eigenvector. Additionally, the angle between ~k and the mean magnetic field were calculated with an error based on the number of measurements and eigenvalues ratios [Sonnerup and Scheible, 1998]. 2.3. Polarization and coherence - Coherence Matrix An additional tool of analysis consisted in calculating the coherence ma- trix of the magnetic field for a 0.015 Hz interval around the local proton cy- clotron frequency (MAG uncertainty) [Fowler et al., 1967; McPherron et al., 1972]. The elements of this matrix provide information about the coherence [Tsurutani et al., 2009], polarization and ellipticity within the selected fre- quency range. These parameters are explicitly indicated in equations (21), (20) and (23), respectively [Rankin and Kurtz, 1970]. 2.4. Magnetic connectivity with the Martian bow shock We also determine the spatial location of wave observations with respect to the Martian foreshock. For this purpose, we use a static model of the bow shock [Vignes et al., 2000] and look for an intersection point with the field line sampled by the spacecraft, assuming a uniform IMF. 2.5. Spatial distribution Several theoretical studies [Modolo et al., 2005] suggest a dependence of the spatial distribution of pick-up ions upon convective electric field ~Ec. We investigate if a similar pattern is found on the distribution of waves. As- suming ~VSW = −400 km/s XM SO, we study the spatial distribution of waves 10 (including their amplitude) , with respect to ~Ec, by introducing a "electro- magnetic"coordinate system (MBE) which is centered at Mars and where the ~ZM BE axis is parallel to ~Ec, ~XM BE is antiparallel to ~VSW , and ~YM BE completes the right hand triad. The IMF cone angle is also included in this analysis. These methods have been applied on selected time intervals to illustrate the main properties of these waves and then on 10-minute overlapping inter- vals in order to study the changes in the waves' properties along the MGS trajectory. The duration of these segments was chosen so as to contain a significant number of proton cyclotron periods. 3. Results Figures 1 and 2 show typical examples of wave activity present in MAG/ER measurements. Figure 1 illustrates an example of PCW's seen by MAG up- stream from the Martian bow shock. The top three panels show the magnetic field components in the MSO reference frame. At first glance, the oscilla- tions show a well-defined frequency, large amplitudes (up to 4 nT) and a high degree of coherence. In this interval, the average magnetic field ( ~Bo) is quite steady: its magnitude is slightly above 8 nT and its orientation makes an angle of 34°with respect to the Mars(cid:21)Sun direction. Figure 2 shows the electron fluxes measurements seen by ER for energies 116, 191 and 314 eV (with maximum time resolution of 2s). Oscillations with features similar to those measured by MAG can be observed, which we assume also corresponds to fluctuations in the total electron density. 11 ) T n ( x B ) T n ( y B ) T n ( z B 10 8 6 4 −1 −3 −5 −7 6 4 2 0 10 9 8 7 ) T n ( B 10:37:00 Orbit P216 10:38:00 10:37:30 Universal Time (hours) 10:38:30 Figure 1: Magnetic field measurements during part of the orbit P216 in MSO coordinates, (April 3, 1998). 3.1. Temporal spectra When simultaneous MAG and ER measurements are compared, we find that the component of the magnetic field parallel to ~Bo and the omnidi- rectional electron flux are in phase. Figure 3 shows the fluctuations in the parallel component of the field (upper panel), and in the electron flux at 116 eV (lower panel) during part of the orbit P232. The cross-correlation of the two time series shown in the enclosed panel displays a peak at zero displace- ment with a correlation value of 0.7, falling down to 0.44 at ±2s. We repeat this analysis for other events and for different energy channels with similar results, which is consistent with both time series being in phase, considering an error of ±2s associated with the cadence of the ER instrument. We also find that most of the waves observed by MGS MAG/ ER have frequencies in the SC frame which coincide with the local proton cyclotron 12 Orbit P233 314 eV 191 eV 116 eV 3.10^4 ) V e . r e t s . s . 10^4 2 m c ( / t r a p ( x u l F 10^3 17:33:00 17:34:30 17:36:00 17:37:30 17:39:00 Universal Time (hours) Figure 2: Electron flux measurements during part of the orbit P233, (April 11, 1998). frequency (Ωp = B/mp), within a ±1 nT uncertainty. Figure 4 shows a dynamic Fourier spectrum of the ~YM SO component of the magnetic field (BY M SO) in the orbit P216. The peak in the frequency of the oscillations is systematically close to the calculated proton cyclotron frequency for several hours. 3.2. Polarization - MVA MVA shows that the waves are planar, propagate almost parallel to the background magnetic field and have a left-hand circular polarization (in the SC frame) with respect to ~Bo. Figure 5 shows the results of MVA applied on 10 cyclotron periods of data during orbit P216. In this case, MVA yields a high λ2/λ3 ratio (50.3) and a small angle between ~k and ~Bo (θkB = 8°). The hodogram describing the maximum and intermediate variance components clearly shows a left hand polarization, as the mean magnetic field for this interval points out of the page. 13 Orbit P232. δB and δflux (E=116 eV) par 0.5 0.25 0 −0,25 −0,5 ) T n ( r a p B δ ) V e . r e t s . s . 2 m c / t r a p ( x u l f δ 3000 1500 0 −1500 −3.000 6:40 6:41 6:42 Cross−correlation function 1 0.5 0 −0.5 −18 −12 −6 0 6 12 18 Time displacement (seg) 6:43 6:44 6:45 6:46 6:47 Universal Time (hours) Figure 3: Fluctuations in the parallel component of the magnetic field and in the electron flux measurements (E=116 eV) during part of the orbit P232, (April 11, 1998). In the case of orbit P216, the propagation properties of the waves remain unchanged for several hours. As a result, an estimation of the θkB from 10 proton cyclotron periods windows for which λ2/λ3 > 10 [Knetter et al., 2004] yields a value of 9.9°. We also find that these waves typically have more power perpendicular to the mean-field direction than along it, which again indicates that ~k makes a small or moderate angle with the mean-field direction. 14 Figure 4: Fourier dynamic spectrum of BY M SO, Orbit P216, (April 3, 1998). The calculated local proton cyclotron frequency from MAG data is plotted in black for reference (line in the middle). The black lines at the sides correspond to the error bars associated with the MAG data. P is the power spectral density. 3.3. Polarization and coherence - Coherence Matrix The apparent coherence of the waves shown in Figure 1 is confirmed with an estimate of the coherence coefficient described in section (2). Figure 6 illustrates an example of PCW's (orbit P204) with a high degree of coher- ence (≥ 0.7 ) and polarization (≥ 70%) for over an hour. Their ellipticity coefficient indicates a left-hand elliptical polarization (< -0.6). Similar anal- yses applied to other SPO1 orbits suggest that these properties also remain unchanged for long periods of time (typically 1 hour). We also see that the wave amplitudes decrease with radial distance from the planet, supporting the idea that Mars is the source of these waves. Figure 7 illustrates what is observed in orbit P216. A similar behavior is observed for other orbits. 15 Orbit P216 April 3, 1998. 10:37:11 − 10:38:42 2 1.5 1 0.5 0 −0.5 −1 −1.5 −2 ) T n ( B 2 −2.5 −3 −2.5 −2 −1.5 −1 −0.5 0 0.5 1 1.5 B (nT) 1 Figure 5: MVA hodogram corresponding to a 10 cyclotron period interval during orbit P216. The mean magnetic field points out of the page ( ~Bo = [−1.02, −0.18, 8.27] nT). The cross indicates the start of the time series analyzed. 3.4. IMF dependence: Connection to the Martian bow shock The properties of PCW's are the same inside and outside the nominal Martian foreshock. The results are summarized in Table 1 which shows the average properties of PCWs for selected orbits that were chosen based on their particularly high values of degree of polarization and coherence. For example, orbits P204 and P207 have similar wave properties: they both are characterized by a high degree of polarization (≥ 90%), a large coherence (0.9) and a left-hand ellipticity (-0.72). θkB is less than 15°and their IMF cone angles differ in approximately 8°. Furthermore, in both cases the waves are observed in the same region (approximately between (2.09,-1.27,-2.22) RM and (1.7,-1.38,-4.21) RM in MSO coordinates). However, whereas in P204 16 f o e e r g e D n o i t a z i r a l o p 100 90 80 70 y t i c i t p i l l E e c n e r e h o C −0.5 −0.6 −0.7 −0.8 1 0.9 0.8 0.7 Orbit P204 15:30 15:45 16:00 16:15 16:30 15:30 15:45 16:00 16:15 16:30 15:30 15:45 16:00 16:15 16:30 Universal Time (hours) Figure 6: Properties of the ultra low frequency waves observed by the MGS spacecraft during a part of the orbit P204, (March 28, 1998). the magnetic field lines are connected to the bow shock, this does not occur in P207. The difference arises because in the first case, ~Bo is approximately (3,-1.5,-3) nT and in the second one (3,-5,0.7) nT. This strongly suggests that these are not associated with Solar Wind backstreaming ions within the ion Martian foreshock. 3.5. Overall occurrence A study of the overall occurrence demanded the definition of a few criteria of the type of oscillation that would be considered to be a PCW. These criteria are: • Frequency close to the calculated proton cyclotron frequency (±1nT). • MVA: events with λ2/λ3 ≥ 10. 17 o B / B δ 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 1 Orbit P216 2 3 4 Altitude (R ) M 5 6 Figure 7: The relative wave amplitudes decrease with radial distance from the planet. Orbit P216, (April 3, 1998). An exponential fit y = A exp(− r a ) yields a scale height a = (4.15 ± 1.5) RM . • Events with coherence≥ 0.7 , ellipticity ≤ −0.5 , degree of polarization ≥ 70%. As we explained in section (2), we study the temporal variation of the different waves properties breaking time series into 10-minute overlapping segments. Figure 8 shows the wave amplitude versus planetocentric distance for all SPO orbits. The amplitude is estimated from the power spectral density in an interval centered at the proton cyclotron frequency with a width of 0.015 Hz. Cyan crosses correspond to SPO1, while brown open circles indicate wave occurrence for SPO2. The black curve displays the average amplitudes for 0.5 RM bins, showing a general decrease of amplitude with increasing 18 distance. Assuming that the source for these waves are exospheric pick-up protons, the obtained result is to be expected for at least two reasons: the ion density decreases with distance, and for smaller MGS altitudes, waves have more time to grow as they are convected by the Solar Wind. 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 ) T n ( e d u t i l p m A 0 2 2.5 3 3.5 4 4.5 Altitude (RM) 5 5.5 6 6.5 Figure 8: Amplitude of the PCW's as a function of the altitude for all SPO orbits. (Criteria stated at the beginning of section 3.5). Figure 9 shows the amplitude of the waves (estimated in the same way as in figure 8) as a function of the observed IMF cone angle for all SPO orbits. The cyan crosses correspond to SPO1 and the brown open circles to SPO2, while the black curve displays the average amplitudes for 10° bins. This figure does not show any statistical shift between these two populations. The 10° bin curve shows a slight shift toward small amplitudes as the IMF cone angle increases. This trend is similar to the behavior of the saturation amplitude 19 of the instablity shown in Cowee et al. [2012]. However, Cowee et al. [2012] suggested that the saturation level is likely not reached in the upstream region of Mars, which is not necessarily the case for the waves measured by MGS. The observed differences could be due to the non homoscedasticity of the αV,B distribution and the αV,B estimation errors, which make direct comparison with their results not so straightforward. 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 ) T n ( e d u t i l p m A 0 10 20 30 40 50 Cone angle α V,B 60 70 80 90 Figure 9: Amplitude of the PCW's as a function of the IMF cone angle for all SPO orbits. (Criteria stated at the beginning of section 3.5). We also analyze the spatial distribution of the orbit segments where waves were found, making use of the MBE coordinate system. Figure 10.a shows the wave amplitudes (color coded) for the projection of the trajectory on the (~YM BE, ~ZM BE) plane for all SPO1 and SPO2 orbits. Figure 10.b shows the αV,B value instead. In the absence of velocity vector measurements, αV,B is obtained assuming that ~VSW is parallel to the Sun-Mars line and therefore: 20 αV,B = acos( BXM SO B ). In both figures, closed and open circles correspond to SPO1 and SPO2 respectively. The analyses do not show any clear difference in the spatial distribution of the waves between these two sub-phases: 60% of the waves occur above the ZM BE = 0 plane and 40% are below it, while the spacecraft is 48% and 52% of the time in each hemisphere. As a result, the spatial distribution of PCWs does not seem to be affected by the Solar Wind's convective electric field. In fact, the spatial occurrence of the amplitude of the waves and the cone angle αV,B associated with them are not affected by the electric field either. Moreover, these waves are observed even when ~Ec is relatively weak. Another interesting aspect of this study is the clear difference in the occurrence rate of the waves between SPO1 and SPO2: whereas PCWs are present in 62% of the SPO1 upstream observations, they only appear in 8% of the time spent by MGS in the upstream region during SPO2. The waves during SPO1 last for longer and have a higher degree of polarization and coherence than those observed in SPO2. This difference in the occurrence of PCW's is in principle not due to changes in the αV,B as the distributions for both subphases peak at around 60°(Parker's spiral angle at Mars 55°) with a standard deviation of 19°. 4. Discussion - Planetary Ion Pick-up and Waves at the Cyclotron Frequency One of the main results of the present study is that the waves in the region upstream of the Martian bow shock are observed, in the SC frame, at the local proton cyclotron frequency and are left-hand elliptically polarized. 21 M ) R ( E B M − Z ) M R ( E B M − Z 6 4 2 0 −2 −4 −6 −6 6 4 2 0 −2 −4 −6 −6 Amplitude (nT) 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 −4 −2 0 Y−MBE (R ) M 2 4 6 α V,B 80 70 60 50 40 30 20 −4 −2 0 2 4 6 Y−MBE (R ) M Figure 10: Trajectory of the spacecraft SPO orbits with PCW's (MBE coordinate frame). Upper panel (a): Color coded by the amplitude of the waves. Lower panel (b): Coded by color by the cone angle associated with the MAG measurements where it is observed PCW's. Closed and open circles correspond to SPO1 and SPO2, respectively. 22 As described in the introduction section, the Vlasov-Maxwell theory pro- vides a mechanism capable of generating this type of waves as a result of the interaction between the magnetic field and the distribution function of "exospheric " pick-up protons in the Solar Wind [Russell et al., 1990]. In particular, numerical studies [Gary, 1993] suggest two possible modes LH and RH whose predominance depends on the initial distribution function of the pick-up protons, which in turn, depends on the IMF cone angle. The dispersion obtained in the values of the IMF cone angle (where ~VSW was assumed to be along the Sun-Mars direction) indicates that in principle both modes could be present in MAG observations. On the other hand, all wave observations in the SC frame displays a LH polarization. In the absence of ion measurements it is impossible to obtain a value for the phase velocity of these waves. In particular, the angle between ~k and ~VSW (that is neces- sary to compute the Doppler correction) at which the polarization changes remains unknown. However, following theoretical results [Cowee et al., 2012] one could assume that the phase speed is of the order of the 10% of the VSW , in which case the cutoff angle would be around 95°. In that case, most of the waves observed would correspond to the RH mode. LH waves have been ob- served at comet Grigg-Skjellerup [Neubauer et al., 2003] for IMF cone angles near 90°[Cao et al. [1998] and references therein]. MVA analysis shows that the observed PCW's are planar and propagate almost parallel (but not exactly) to ~B. This could be the result of nonlinear effects not addressed in this paper. This non zero θkB is responsible for a small compressibility (typically, δBpar/B ≃ 0.25). The compressibility of these waves was studied from the cross-correlation 23 between the fluctuating electron flux (with the electron fluxes taken as a proxy of the electron density) and the parallel magnetic field component. The maximum values for the cross-correlation coefficient between both time series are associated with phase shifts between 0 and 2s. The quality of this result is limited by the following two factors: the influence of the magnetic fields generated by the spacecraft and the time resolution of the ER instrument(2s). These two sources of error impose a tradeoff in our aim to obtain the best possible cross-correlation between the two time series. With this in mind we calculated the cross-correlation for several intervals with mean magnetic fields between 4 and 8 nT. For all the cases considered, the results indicate a maximum in the correlation coefficient located within 2s from zero time lag. This means that there is a systematic correlation between the two data sets within the errors caused by the spacecraft fields and the ER time resolution. But also, these results indicate that anticorrelation is not possible (errors are not as high as half a proton cyclotron period). The analysis of MAG measurements also shows that the frequency, polar- ization and coherence of these waves do not seem to depend on the direction of the IMF (see Table 1). It is then not surprising that the observed waves are not associated with the foreshock. In addition, the amplitude of the observed waves shows a decrease as the IMF cone angle increases. The lack of influence of ~Ec on the spatial distribution of PCWs as well as their amplitudes suggests that the link between the spatial distribution of the pick up ions and the waves is not obvious [Cowee et al., 2012]. On the other hand, the decrease of the observed wave amplitude with radial distance from the planet, supports the idea that Mars is indeed the source of these 24 waves. We also analyzed the PCWs overall occurrence using very strict criteria. We found a very clear difference in the percentage and the properties of PCWs between SPO1 and SPO2. However, a statistical study of the IMF cone angle did not show any significant difference between both periods, suggesting that the pick-up geometry might have not changed significantly from SPO1 to SPO2. The difference in the occurrence of the waves between SPO1 and SPO2 might be related to temporal or spatial changes in the properties of the Martian hydrogen exosphere. Wei and Russell [2006] studied the wave occurrence in AB1, as seen from a reference frame that takes into account the convective electric field. They observed that the PCWs at large distances (85 events present in 9 orbits with B > 5.6 nT) occur intermittently and predominantly in the hemi- sphere where ~Ec points out from the planet (+E hemisphere). In order to explain this behavior the authors proposed a mechanism where an exospheric hydrogen atom is ionized and then accelerated by the Solar Wind electric field. A charge exchange collision would then produce fast neutrals able to reach distant regions where they are re-ionized and generate cyclotron waves downstream of the planet and on the + E hemisphere. During SPO1 and SPO2 the wave spatial distribution for distances smaller than 6 RM does not seem to depend on the orientation of the Solar Wind convective electric field (even after more restrictive criteria have been ap- plied). Furthermore, waves are found even when ~Ec is relatively weak. We also observe PCWs far away into the -E hemisphere, although there is no known mechanism to move ions against the electric field. One explanation 25 to this is that the wave distribution does not necessarily follow the one of pick-up ions. It is important to note that pick-up ions generate waves whose wavelengths are of planetary scale. Ion density is not likely to be homoge- neous over such large spatial distances. The difference between the results obtained in Wei and Russell [2006] and the ones presented here could be due to the fact that MGS sampled different regions during AB1 and SPO. On the other hand it is also important to point out that the lack of correlation between the spatial distribution of the PCWs and the convective electric field has also been observed at Venus [Delva et al., 2011]. 5. Conclusions In this study we analyzed upstream MGS MAG/ER measurements during the pre-mapping SPO phase. Analyses of these measurements show waves whose frequencies in the reference system of the satellite are near the local cyclotron frequency of protons. These ultra low frequency plasma waves are also characterized (in the spacecraft frame) by a left-hand elliptical polar- ization and by an amplitude that decreases with the increase of the IMF cone angle and the radial distance from the planet. They propagate almost parallel to the background magnetic field and show a small degree of com- pressibility. The observed waves are not associated to the foreshock and their properties do not depend on the angle αvB between the solar wind velocity and the magnetic field direction. The most plausible explanation for the existence of these waves is that they are generated from the pick-up of newborn protons from the Martian exosphere whose distribution function kinetically interacts with the Solar 26 Wind plasma and the interplanetary magnetic field [Russell et al., 1990]. From a theoretical perspective, there are two possible modes that can be generated in the plasma frame which depends on the pick-up proton distri- bution function. For relative orientations between ~VSW and ~B ranging from parallel to almost perpendicular, the RH mode will predominate, while for almost perpendicular orientations the LH mode will dominate. The reason why we always observe left-hand polarized waves in the SC frame is because of their corresponding Doppler corrections. The spatial distribution of these waves does not seem to depend on the orientation of the Solar Wind convective electric field and they are also ob- served even when it is relatively weak. PCWs are also found far into the negative convective electric field region. This supports the idea from recent simulation results [Cowee et al., 2012] that, even though the origin of these waves are the pick-up ions, there is no clear relation between their corre- sponding spatial distributions. We also found a clear difference in the PCWs occurrence between SPO1 and SPO2. Moreover, it is observed that the waves observed during SPO1 last longer, and have higher degree of polarization and coherence than those observed in SPO2. Since there is no significant difference in the IMF cone angle distribution between both periods, these differences might be due to changes in the properties of the exosphere and/or ionization rates. A more accurate determination of the wave modes present at Mars will require information about the pick up ion distribution function as well as that of the Solar Wind. We hope MAVEN spacecraft (Mars Atmosphere and Volatile Evolution Mission) will provide such measurements contributing to a 27 better understanding of the microscopic processes arising from the interaction between the Solar Wind and Mars. Acknowledgments The authors also wish to thank Misa Cowee for helpful and interest- ing suggestions and comments. This work was done in conjunction with the International Space Science Institute (ISSI) Working Group on Induced Magnetospheres. N.R. is supported by a National Science and Technology Research Council (CONICET) Phd grant. References Acuna, M., et al., 1992. The Mars Observer magnetic fields investigations. J. Geophys. Res. 97, E5, 7799-7814. Acuna, M., et al., 1998. Magnetic field and plasma observations at Mars: initial results of the Mars Global Surveyor Mission. Science 279, 1676-1680. Acuna, M., et al., 2001. Magnetic field of Mars: summary of results from the aerobraking and mapping orbits. J. Geophys. Res. 106, E10, 23403-23417. Albee, A., et al., 2001. Overview of the Mars Global Surveyor mission. J. Geophys. Res. 106, E10, 23291-23316. Bertucci, C., et al., 2005. Interaction of the Solar Wind with Mars from Mars Global Surveyor MAG/ER observations. J. Atmospheric and Solar- Terrestrial Physics 67, 1797-1808. 28 Brain, A., et al., 2002. Observations of low-frecuency electromagnetic plasma waves upstream from the Martian shock. J. Geophys. Res. 107, A6, 1076. Brinca, A., Tsurutani, B.T., 1989. Influence of multiple ion species on low- frequency electromagnetic wave instabilities, J. Geophys. Res., 94, 13565- 13569. Brinca, A., 1991. Cometary linear instabilities: from profusion to perspertive in Cometary Plasma Processes. Geophys. Monograph 61, 211-221. Cao, J.B., Mazelle, C., Belmont, G., Reme, H.: J. Geophys. Res. 103,2055 (1998) Chaufray, J.Y., Bertaux, J.L., Leblanc, F., Qumerais, E., 2008. Observation of the hydrogen corona with SPICAM on Mars Express. Icarus, 195, Issue 2, 598-613. Convery, P., and Gary, P., 1997. Electromagnetic proton cyclotron ring in- stability: Threshold and saturation. J. Geophys. Res. 102, A2, 2351-2358. Cowee, M. M., S. P. Gary, and H. Y. Wei (2012), Pickup ions and ion cyclotron wave amplitudes upstream of Mars: First results from the 1D hybrid simulation, Geophys. Res. Lett., 39, L08104, doi:10.1029/2012GL051313. Delva, M., et al., 2009. Hydrogen in the extended Venus exosphere. J. Geo- phys. Res. 36, L01203. 29 Delva, M., Mazelle, C., Bertucci, C., Volwerk,M., Voros, Z., Zhang, T.L., 2011. Proton cyclotron wave generation mechanisms upstream of Venus. J. Geophys. Res. 116, A02318, doi:10.1029/2010JA015826. Fowler R. A., Kotick B.J., Elliot D., 1967. Polarization Analysis of Natural and Artificially Induced GeomagneticMicropulsations. J. Geophys. Res. 72, No. 11. Gary, S.P., 1991. Electromagnetic ion/ion instabilities and their consequences in space plasmas: A review, Space Sci. Rev., 56, 373415. Gary, S.P., 1993. Theory of Space Plasma Microinstabilities. In: Cambridge Atmospheric and Space Series. Cambridge University Press, Cambridge. Gary, S.P., and Madland, C., 1988. Electromagnetic ion instabilities in a cometary environment. J. Geophys. Res. 93, A1, 235-241. Gary, S.P., Akimoto, K., Winske D., 1989. Computer simulations of cometary ion/ion instabilities and wave growth. J. Geophys. Res. 94, A4, 3513-3525. Khrabrov, A., and Sonnerup, B., 1998. Error estimates for minimum variance analysis. J. Geophys. Res. 103, A4, 6641-6651. Knetter, T., et al., 2004. Four-point discontinuity observations using Cluster magnetic field data: A statical survey. J. Geophys. Res. 109, A06102. Mazelle, C., and Neubauer, F.M., 1993. Discrete wave packets at the proton cyclotron frequency at Comet P/Halley.Geophys. Res. Lett. 20, 153. Mazelle, C., et al., 2004. Bow shock and upstream phenomena at Mars, Space Science Review 111, 115-181. 30 McPherron, R.L., Russell,C.T., Coleman, P.J., 1972. Fluctuating magnetic fields in the magnetosphere, II, ULF waves. Space Sci. Rev. 13, 411-454. Mitchell, D., et al., 2001. Probing Mars' crustal magnetic field and ionosphere with the MGS Electron Reflectometer. J. Geophys. Res. 106, E10, 23419- 23427. Modolo, R., Chanteur, G. M., Dubinin, E., Matthews, A.P., 2005. Influ- ence of the solar EUV flux on the Martian plasma environment. Annales Geophysicae 23, 433444. Neubauer, F. M., K. .-H. Glassmeier, A. J. Coates, and A. D. Johnstone (1993), Low-Frequency Electromagnetic Plasma Waves at Comet P/Grigg- Skjellerup: Analysis and Interpretation, J. Geophys. Res., 98(A12), 20,93720,953, doi:10.1029/93JA02532. Rankin, D., and Kurtz, R., 1970. Statistical study of micropulsation polar- izations. J. Geophys. Res. 75, 5444-5458. Russell, C.T., et al., 1990. Upstream waves at Mars - PHOBOS observations. Geophysical Research Letters 17, 6, 897-900. Smith, E.J., and Tsurutani, B.T., 1976. Magnetosheath Lion Roars, J. Geo- phys. Res. 81, 2261-2266. Sonnerup, B. U. O., Scheible, M., 1998. Minimum and Maximum Variance Analysis, in: Paschmann, G., Daly, P. W. (Eds), Analysis Methods for Multi-Spacecraft Data. Int. Space Sci. Inst., Bern, Switzerland. 31 Tsurutani, B.T., and Smith, E.J., 1986. Strong hydromagnetic turbulence associated with comet Giacobini-Zinner, Geophys. Res. Lett., 13, 259-262. Tsurutani, B.T., et al., 1987. Steepened magnetosonic waves at comet Giacobini-Zinner, J. Geophys. Res. 92, 11074-11082. Tsurutani, B.T., et al., 1989. Magnetic Pulses with Durations near the Local Proton Cyclotron Period: Comet Giacobini-Zinner, J. Geophys. Res., 94, 29-35. Tsurutani, B. T., 1991. Comets: A laboratory for plasma waves and instabil- ities, in Cometary Plasma Processes, Geophys. Monogr. Ser., 61, 189209, AGU, Washington, D. C. Tsurutani, B. T., O. P. Verkhoglyadova, G. S. Lakhina, and S. Yagi- tani (2009), Properties of dayside outer zone chorus during HILDCAA events: Loss of energetic electrons, J. Geophys. Res., 114, A03207, doi:10.1029/2008JA013353. Vignes, D., et al., 2000. The Solar Wind interaction with Mars: locations and shapes of the Bow Shock and the Magnetic Pile-up Boundary from the observations of the MAG/ER experiment onboard Mars Global Surveyor. Geophysical Research Letters 27, 1, 49-52. Wu, C. S., and R. C. Davidson (1972), Electromagnetic Instabilities Pro- duced by Neutral-Particle Ionization in Interplanetary Space, J. Geophys. Res., 77(28), 53995406, doi:10.1029/JA077i028p05399. Wu, C.S., and Hartle, R.E. 1974. Further remarks on plasma instabilities produced by ions born in the solar wind. J. Geophys. Res. 79, 283-285. 32 Wei, H., and Russell, C.T., 2006. Proton cyclotron waves at Mars: Exosphere structure and evidence for a fast neutral disk. J. Geophys. Res. 33, L23103. 33 Orbit D. Pol Ell Coh λ2/λ3 θkB (°) B αV,B (°) Connection 204 207 208 209 211 215 216 217 239 241 242 249 257 258 260 267 268 94.7 95.4 89.6 88.6 91.7 85.4 94.5 93.3 93.3 91.5 91.5 87.1 86.4 90.8 90.6 91.7 89.4 -0.72 0.90 31.4 -0.72 0.90 32.8 -0.67 0.80 17.4 -0.67 0.78 13.9 -0.68 0.83 15.3 -0.65 0.73 18.1 -0.69 0.89 20.2 -0.70 0.84 19.6 -0.70 0.87 37.6 -0.68 0.83 32.1 -0.69 0.81 19.7 -0.63 0.76 10.0 -0.64 0.75 16.1 -0.68 0.82 30.4 -0.67 0.82 33.3 -0.70 0.83 42.6 -0.67 0.80 23.1 8.9 14.2 10.0 19.3 15.6 38.4 17.6 24.3 6.14 17.8 20.6 29.6 15.2 21.9 14.1 19.6 12.7 4.86 5.94 5.04 9.04 4.65 13.5 7.31 4.62 4.10 7.20 15.0 4.87 4.44 4.63 5.27 5.02 5.06 51.9 59.6 54.1 49.8 56.6 40.2 33.6 31.5 65.2 76.9 71.7 32.6 61.7 54.1 42.8 68.4 24.4 Yes No Yes Yes Both No Both No No No No Both Yes No Yes No Yes Table 1: List of some selected orbits. Orbit number and wave properties. From left to right: degree of polarization, ellipticity, coherence, λ2/λ3, θkB, B, Cone Angle αV,B, Connection to the Martian Bow Shock (Yes, No, Both cases). 34
1206.0749	1	1206	2012-06-04T20:04:47	Back to the Moon: The Scientific Rationale for Resuming Lunar Surface Exploration	[ "astro-ph.EP", "astro-ph.IM" ]	The lunar geological record has much to tell us about the earliest history of the Solar System, the origin and evolution of the Earth-Moon system, the geological evolution of rocky planets, and the near-Earth cosmic environment throughout Solar System history. In addition, the lunar surface offers outstanding opportunities for research in astronomy, astrobiology, fundamental physics, life sciences and human physiology and medicine. This paper provides an interdisciplinary review of outstanding lunar science objectives in all of these different areas. It is concluded that addressing them satisfactorily will require an end to the 40-year hiatus of lunar surface exploration, and the placing of new scientific instruments on, and the return of additional samples from, the surface of the Moon. Some of these objectives can be achieved robotically (e.g. through targeted sample return, the deployment of geophysical networks, and the placing of antennas on the lunar surface to form radio telescopes). However, in the longer term, most of these scientific objectives would benefit significantly from renewed human operations on the lunar surface. For these reasons it is highly desirable that current plans for renewed robotic surface exploration of the Moon are developed in the context of a future human lunar exploration programme, such as that proposed by the recently formulated Global Exploration Roadmap.	astro-ph.EP	astro-ph	(Accepted for publication in Planetary and Space Science) Back to the Moon: The Scientific Rationale for Resuming Lunar Surface Exploration I. A. Crawforda,b,, M. Anandc,d, C. S. Cockelle, H. Falckef,g,h, D. A. Greeni, R. Jaumannj, M. A. Wieczorekk a Department of Earth and Planetary Sciences, Birkbeck College, Malet Street, London, WC1E 7HX, UK. b Centre for Planetary Sciences at UCL/Birkbeck, UK. c Planetary and Space Sciences, Department of Physical Sciences, The Open University, Milton Keynes, MK7 6AA, UK. d Department of Mineralogy, The Natural History Museum, London, SW7 5BD, UK. e School of Physics and Astronomy, University of Edinburgh, UK. f Department of Astrophysics, Radboud University, Nijmegen, The Netherlands. g Netherlands Institute for Radio Astronomy (ASTRON), Postbus 2, 7990 AA Dwingeloo, The Netherlands. h Max-Planck-Institut für Radioastronomie, Auf dem Hügel 69, 53121 Bonn, Germany. i Centre of Human and Aerospace Physiological Sciences, King's College London, London, SE1 1UL, UK. j DLR, Institute of Planetary Research, Berlin, Germany. k Institut de Physique du Globe de Paris, Univ Paris Diderot, France. Corresponding author: Tel : +44 203 073 8026 Email address: [email protected] Abstract The lunar geological record has much to tell us about the earliest history of the Solar System, the origin and evolution of the Earth-Moon system, the geological evolution of rocky planets, and the near-Earth cosmic environment throughout Solar System history. In addition, the lunar surface offers outstanding opportunities for research in astronomy, astrobiology, fundamental physics, life sciences and human physiology and medicine. This paper provides an interdisciplinary review of outstanding lunar science objectives in all of these different areas. It is concluded that addressing them satisfactorily will require an end to the 40-year hiatus of lunar surface exploration, and the placing of new scientific instruments on, and the return of additional samples from, the surface of the Moon. Some of these objectives can be achieved robotically (e.g. through targeted sample return, the deployment of geophysical networks, and the placing of antennas on the lunar surface to form radio telescopes). However, in the longer term, most of these scientific objectives would benefit significantly from renewed human operations on the lunar surface. For these reasons it is highly desirable that current plans for renewed robotic surface exploration of the Moon are developed in the context of a future human lunar exploration programme, such as that proposed by the recently formulated Global Exploration Roadmap. Keywords: Moon; Lunar science; Lunar geology; Lunar geophysics; Lunar astronomy; Space exploration; Astrobiology; Space life sciences; Space medicine 1. Introduction Over the last decade there has been a renaissance in lunar exploration conducted from orbit about the Moon, with the following countries all sending remote-sensing spacecraft to lunar orbit during this period: European Space Agency: SMART-1 (2004); Japan: Kaguya (2007); China: Chang’e-1 (2007), Chang’e-2 (2010); India: Chandrayaan-1 (2008); and the United States: Lunar Reconnaissance Orbiter (LRO; 2009), GRAIL (2012). However, none of these spacecraft were designed to land on the Moon’s surface in a controlled manner, although the US Lunar CRater Observation and Sensing Satellite (LCROSS; co-launched with LRO) and the Chandrayaan-1 Moon Impact Probe (MIP) did deliberately impact the lunar surface in an effort to detect polar volatiles. Indeed, it is sobering to realise that no spacecraft of any kind has successfully landed on the Moon in a controlled manner since the Russian robotic sample return mission Luna 24 in August 1976, and that no human being has set foot on the Moon since Apollo 17 in December 1972. In this paper we argue that this long hiatus in lunar surface exploration has been to the detriment of lunar and planetary science, and indeed of other sciences also, and that the time has come to resume the robotic and human exploration of the surface of the Moon. The strong scientific case for renewed lunar exploration was already recognized in a comprehensive study conducted by the European Space Agency in 1992 on “Europe’s Priorities for the Scientific Exploration and Utilization of the Moon” (ESA 1992). Many of the conclusions of this earlier study remain valid today, albeit with some need of updating in light of more recent discoveries, and we will follow its basic framework in what follows. In particular, the ESA study found that lunar science objectives can logically be divided into three categories: (i) Science of the Moon (i.e. studies of the Moon itself); (ii) Science on the Moon (i.e. studies using the lunar surface as a platform for scientific investigations not directly related to the Moon itself); and (iii) Science from the Moon (i.e. studies utilising the lunar surface as a platform for astronomical observations). 2. Science of the Moon From a planetary science perspective, the primary importance of the Moon arises from the fact that, owing to a lack of recent geological activity and active erosional processes, it has an extremely ancient surface, mostly older than 3 billion years with some areas extending almost all the way back to the origin of the Moon 4.5 billion years ago (e.g. Hiesinger and Head 2006; Jaumann et al., 2012). Its relatively accessible near-surface environment therefore preserves a record of the early geological evolution of a terrestrial planet, which more complex planets such as Earth, Venus and Mars, have long lost, and of the Earth-Moon system in particular. Moreover, the lunar surface provides a platform for geophysical instruments (e.g. seismometers and heat-flow probes) to probe the structure and composition of the deep interior, which is also required if we are to use the Moon as a model for terrestrial planet evolution. Last, but not least, the Moon’s outer layers also preserve a record of the environment in the inner Solar System (e.g. meteorite flux, interplanetary dust density, solar wind flux and composition, galactic cosmic ray flux) throughout Solar System history, much of which is relevant to understanding the past habitability of our own planet (e.g. Crawford, 2006; Cockell, 2010). Gaining access to this information lies primarily in the domain of the geosciences (i.e. geology, geophysics, and geochemistry), and will require a return to the lunar surface with robotic and/or human missions. The most recent study of ‘Science of the Moon’ objectives is the US National Research Council’s Report on the Scientific Context for Exploration of the Moon (NRC 2007; hereinafter the ‘SCEM Report’). This study identified, and prioritized, eight top-level scientific ‘concepts’ (each of which can be broken down into multiple individual science goals), and identified the capabilities that would be required of space missions designed to address them. We note that several of these objectives (i.e. those which could be addressed by a South Pole-Aitken Basin sample return mission and a lunar geophysics network) were also strongly endorsed by the recent US Planetary Science Decadal Survey (NRC, 2011). As a basis for discussion, we summarise these here (as given in Table 4.1 of NRC 2007), and in each case highlight how surface (human and/or robotic) exploration will aid in meeting these scientific objectives. 2.1 The Bombardment History of the Inner Solar System The vast majority of lunar terrains have never been directly sampled, and their ages are based on the observed density of impact craters calibrated against the ages of Apollo and Luna samples (e.g. Neukum et al., 2001; Stöffler et al., 2006). However, the current calibration of the cratering rate, used to convert crater densities to absolute ages, is neither as complete nor as reliable as it is often made out to be. For example, there are no calibration points that are older than about 3.85 Ga, and crater ages younger than about 3 Ga are also uncertain (e.g. Hiesinger et al., 2012). Furthermore, longitudinal variations in the cratering rate on the surface of the Moon are predicted by orbital dynamics models, but these are difficult to confirm with existing data (Le Feuvre and Wieczorek, 2011). Improving the calibration of the cratering rate would be of great value for planetary science for the following three reasons: (i) It would provide better estimates for the ages of unsampled regions of the lunar surface; (ii) It would provide us with a more reliable estimate of the impact history of the inner Solar System, especially that of our own planet; and (iii) The lunar impact rate is used, with various other assumptions, to date the surfaces of other planets for which samples have not been obtained (including key events and stratigraphic boundaries on Mars), and to the extent that the lunar rate remains unreliable so do the age estimates of surfaces on the other terrestrial planets. Moreover, there is still uncertainty over whether the lunar cratering rate has declined monotonically since the formation of the Moon, or whether there was a bombardment ‘cataclysm’ between about 3.8 and 4.0 billion years ago characterised by an unusually high rate of impacts (Hartmann et al., 2000; Stöffler et al., 2006). Clarifying this issue is especially important from an astrobiology perspective because it defines the impact regime under which life on Earth became established (e.g. Maher and Stevenson, 1988; Sleep et al., 1989; Ryder, 2003; for an illustration of this unresolved issue see Fig. 2.3 of NRC 2007). In principle, obtaining an improved cratering chronology is straightforward: it ‘merely’ requires the sampling, and radiometric dating, of surfaces having a wide range of crater densities, supplemented where possible by dating of impact melt deposits from individual craters and basins (Stöffler, et al., 2006). However, in practice this is likely to require the implementation of multiple missions to many different sites (see the detailed discussion by Stöffler, et al., 2006). These might be robotic missions, ideally involving rover-facilitated mobility and either in situ radiometric dating or, preferably, a sample return capability (e.g. Shearer and Borg, 2006). The possibilities of robotic in situ radiometric dating techniques have been investigated by a number of workers (e.g. Talboys et al., 2009; Trieloff et al., 2010), and such in situ measurements may be able to address some key questions in lunar science (e.g. verifying the approximate ages of the youngest lava flows estimated by crater counting). However, the practicality of performing in situ radiometric dating with sufficient accuracy to obtain definitive results for chronology has been questioned (e.g. Taylor et al., 2006). Additional robotic sample return missions to key localities, such as the South Pole-Aitken basin (as envisaged by the MoonRise proposal; Jolliff et al., 2010) would undoubtedly improve our knowledge in key respects (Cohen and Coker, 2010). However it is likely to prove impractical to target dedicated robotic sample return missions to sufficient locations to fully characterise lunar chronology. Given the enhanced mobility, sample collection efficiency and sample return capacity actually demonstrated by the Apollo missions (e.g. Vaniman et al., 1991; Crawford, 2012), it seems clear that obtaining anything approaching a complete lunar impact chronology will, at the very least, be greatly facilitated by new human missions to the lunar surface. 2.2 The structure and composition of the lunar interior As noted in the SCEM Report, the structure of the lunar interior provides fundamental information on the evolution of differentiated planetary bodies. The Moon is especially important in this respect because, lacking plate tectonics, its crust and mantle have remained almost completely isolated from each other for more than 4 billion years. Given the limited volcanic activity on the Moon since its formation, and given the sluggish convection in the mantle of this single-plate planet, the interior of the Moon should retain a record of early planetary differentiation processes that more evolved planetary bodies have since lost. Despite data from a rudimentary geophysical network that was set up during the Apollo missions (ALSEP; Apollo Lunar Surface Experiment Package), keys aspects of the Moon’s interior structure, composition, and evolution are left unanswered (for a summary, see Wieczorek et al. 2006). The thickness of the crust, which is critical for estimating the bulk composition of the Moon, is debated, with recent estimates being almost half of those made during the Apollo era. The existence of a seismic discontinuity in the mantle 500 km below the surface, which could either be a reflection of primordial differentiation or later magmatic processes, is also debated. Four heat flow measurements (at two sites each) were made during Apollo (Langseth et al. 1976) in order constrain the Moon’s bulk composition and thermal evolution, but in retrospect, these were found to be made near the edge of an atypical province enriched in heat-producing elements (Jolliff et al., 2000). It is not clear if these unique measurements are representative of either the ancient feldspathic highlands terrane, or of the more volcanically active Procellarum KREEP terrane. Though most data indicate that the Moon has a small molten core (e.g. Weber et al., 2011), neither its size nor composition are well constrained, and the existence of a solid inner core has remained elusive. Resolving these questions will require making further geophysical measurements. While a few such measurements can be made from orbit (such as the measurement of the Moon’s gravity and magnetic fields; Purucker and Nicholas 2010, Zuber et al., 2011), most require geophysical instruments to be placed on, or below, the lunar surface. Key instruments in this respect are seismometers, to probe the structure of the deep interior (e.g. Lognonné, 2005; Yamada et al., 2011), heat-flow probes to measure the heat loss from the lunar interior and its spatial variations (Langseth, et al., 1976), and magnetometers to measure the thermal conductivity profile. Such instruments can in principle be placed and operated on the lunar surface robotically without direct human intervention, such as envisaged for the proposed Farside Explorer (Mimoun et al., 2012) and Lunette (Neal et al., 2011) missions, the International Lunar Network (Cohen et al., 2010; also http://iln.arc.nasa.gov/), and the LunarNet penetrator concept (Smith et al., 2012). Nevertheless, as discussed below, human deployment of such instruments would be extremely beneficial. The thermal perturbation caused by an autonomous lander, which would bias the heat flow measurements, could be minimized by placing the heat flow experiment far from the landing site (Kiefer, 2012). The emplacement of thermal probes several meters below the surface would be greatly facilitated by a human operated drill, as was aptly demonstrated during the deployment of the heat flow package by the Apollo astronauts. Moreover, thanks to the human operated portable magnetometers used during the Apollo 14 and 16 missions, it is known that the surface magnetic fields can vary dramatically, and even change direction, over kilometer scale distances (e.g., Fuller and Cisowski 1987); human-led magnetic surveys over large distances would be invaluable in deciphering the origins of these enigmatic fields. Strong magnetic fields can help in deflecting harmful solar and cosmic rays, but it is currently difficult to quantify this potential benefit to human exploration activities in the absence of detailed surface magnetic field map (e.g., Halekas et al., 2010). A human return to the lunar surface would therefore benefit geophysical investigations by the additional payload capacity, operational flexibility, and long traverses that are inherent in human missions (e.g. Spudis, 2001; Garvin, 2004; Crawford, 2004). 2.3 The diversity of lunar crustal rocks The SCEM Report highlighted the fact that key planetary processes are manifested in the diversity of lunar crustal rocks. Quantifying and understanding this diversity will require detailed chemical and mineralogical analysis of rocks and soils from as yet unsampled regions of the lunar surface. In particular, no samples have yet been returned from the polar regions or the far-side, thus greatly limiting our knowledge of lunar geological processes. Although, statistically, many of the 160 or so known lunar meteorites (http://meteorites.wustl.edu/lunar/moon_meteorites_list_alpha.htm) must originate from these areas, and provide an excellent statistical sampling of the lunar crust (e.g. Korotev, 2005), the value of these materials is limited by a lack of knowledge of their source regions and thus geological context. The diversity of lunar crustal materials has been demonstrated most recently by new results from orbital remote sensing instruments on Kaguya, Chandrayaan-1 and LRO. These include outcrops of pure anorthosite which may represent pristine magma ocean flotation cumulates (Ohtake et al., 2009); olivine rich outcrops which may sample mantle material (Yamamoto et al., 2010); and spinel-rich (Sunshine et al., 2010) and silica-rich (Greenhagen et al. 2010; Glotch et al. 2010; Jolliff et al., 2011) lithologies not represented in the existing sample collection. It is important to confirm the interpretation of the remote-sensing data as no ‘ground-truth’ has yet been obtained for any of these localities. It is also important to obtain measurements of minor and trace elements in these materials which cannot be detected by orbital remote-sensing instruments, but which are essential to discriminate between different suggested origins and formation mechanisms. Sample return missions to currently unsampled regions would be the preferred means of furthering our knowledge of lunar geological diversity. Although an alternative would be to make robotic in situ geochemical measurements, as proposed for both the Farside Explorer (Mimoun et al., 2012) and LunarNet (Smith et al., 2012) mission concepts, a full understanding of geological processes requires the ability to measure the isotopic composition and concentrations of minor and trace elements at levels that may be beyond the capabilities of in situ robotic instruments. In addition, mobility is crucially important if the local geological diversity in the vicinity of a landing site is to be properly characterized. As was the case for determining surface ages (Section 2.1), and for essentially the same reasons, characterizing lunar geological diversity would greatly benefit from a renewed human presence on the lunar surface provided that such missions were targeted at sites carefully chosen with this objective in mind. 2.4 Volatiles at the lunar poles As noted in the SCEM Report, the lunar poles potentially bear witness to the flux of volatiles present in the inner Solar System throughout much of Solar System history. In 1998 the Lunar Prospector neutron spectrometer found evidence of enhanced concentrations of hydrogen at the lunar poles (Feldman et al., 1998), which was widely interpreted as indicating the presence of water ice in the floors of permanently shadowed polar craters. Radar observations have also been used to infer the presence of ice in such regions (e.g. Spudis et al., 2010; but see Fa et al. 2011 for an alternative interpretation). The presence of water ice in permanently shadowed craters was supported by the LCROSS impact experiment, which found a water ice concentration of 5.6 ± 2.9 % by weight in the target regolith at the Cabeus crater (Colaprete et al., 2010). It seems likely that this water is ultimately derived from the impacts of comets (and/or water-rich asteroids) with the lunar surface, although solar wind implantation and endogenic sources might also contribute (for a brief review see Anand, 2010). However, the inferred quantity of water is sensitive to the calibration of the spectrometers on the LCROSS probe and a number of other assumptions (Colaprete et al. 2010). Ideally, therefore, this result needs to be confirmed by in situ measurements. In addition to ice in permanently shadowed craters, infra-red remote-sensing observations have found evidence for hydrated minerals, and/or adsorbed water or hydroxyl molecules, over large areas of the high latitude (but not permanently shadowed) lunar surface (Pieters et al. 2009; Sunshine et al., 2009; Clark, 2009). It is hypothesised this OH/H2O, which cannot exist as ice, is produced by the reduction of iron oxides in the regolith by solar wind-implanted hydrogen, with OH/H2O being retained in the relatively cold high-latitude regolith. It is possible that, over time, this high-latitude OH/H2O may migrate to polar cold traps and contribute to ice deposits there (Crider and Vondrak 2002), and this scenario is supported by observations of water molecules in the lunar exosphere made by the Moon Impact Probe (MIP) released by Chandrayaan-1 (Sridharan et al., 2010). As discussed by Anand (2010) and Smith et al. (2012), obtaining improved knowledge of the presence, composition, and abundance of water (and other volatiles) at the lunar poles is important for several reasons: • It is probable that the ice in permanently shadowed regions is ultimately derived from comet and/or asteroid impacts. Even though the original volatiles will have been considerably reworked, it is likely that some information concerning the composition of the original sources will remain. Among other things, this may yield astrobiologically important knowledge on the role of comets and meteorites in delivering volatiles and pre-biotic organic materials to the terrestrial planets (Chyba and Sagan 1992; Pierazzo and Chyba 1999). • The processes involved in the creation, retention, migration, and destruction of OH and H2O across the surface of the Moon are likely to be common on other air-less bodies, and quantifying them on the Moon will give us better insight into the volatile history and potential availability of water elsewhere in the inner solar system. • Lunar polar ice deposits are of considerable astrobiological interest, even if they do not retain vestigial information concerning their ultimate sources. This is because any such ices will have been subject to irradiation by galactic cosmic rays and, as such, may be expected to undergo organic synthesis reactions (e.g. Lucey, 2000). Analogous reactions may be important for producing organic molecules in the icy mantles of interstellar dust grains, and on the surfaces of outer Solar System satellites and comets (e.g. Bernstein et al., 2002; Elsila et al., 2007), but the lunar poles are much more accessible than any of these other locations. • The presence of water ice at the lunar poles, and even hydrated materials at high-latitude but non-shadowed localities, could potentially provide a very valuable resource in the context of future human exploration of the Moon (e.g. Anand et al., 2012; this issue). Confirming the interpretation of the remote sensing measurements, and obtaining accurate values for the concentration of polar ice and high latitude surficial OH/H2O will require in situ measurements by suitably instrumented and landed spacecraft. The requirements for permanently shadowed environments and high latitude, but non- permanently shadowed, environments are rather different. The conditions within permanently shadowed areas, where temperatures can be below 40 K (Paige et al., 2010), are not readily amenable to human exploration or even robotic vehicles (unless these are equipped with a nuclear power source or long-lived batteries capable of operating at very low temperatures). For these environments a penetrator-based system, such as envisaged in the MoonLITE and LunarNet mission concepts (Gao et al., 2008; Smith et al., 2012), may be a possibility. On the other hand, non- permanently shadowed polar localities are amenable to solar-powered robotic exploration, and proposals such as ESA’s Lunar Lander (Carpenter et al., 2010; 2012), and ‘Lunar Beagle’ (Gibson et al., 2010), would provide valuable initial measurements of volatiles in these environments. In the longer term, a full characterization of polar volatiles, as for other aspects of lunar geology, would benefit from the increased mobility and flexibility that would be provided by human exploration (which would of course be facilitated at the poles if exploitable quantities of volatiles prove to be present). 2.5 Lunar volcanism The SCEM Report identified the charaterization of lunar volcanism as a high lunar science priority because of the window it provides into the thermal and compositional evolution of the lunar mantle. Since the SCEM Report was published new remote sensing observations have indicated that lunar volcanism is even more diverse than previously thought, with the identification of probable areas of non-mare silicic volcanism on both the near- and far-sides (Glotch et al., 2010; Jolliff et al., 2011). There is no doubt that these areas would benefit from in situ investigation. However, from a lunar exploration perspective, this is just one aspect of the wider requirement to sample a diverse set of lunar rocks (Section 2.3), and the implications for exploration capabilities are essentially the same. 2.6 Impact processes Impact cratering is a fundamental planetary process, an understanding of which is essential for our knowledge of planetary evolution. Yet our knowledge of impact processes is based on a combination of theoretical modelling, small-scale laboratory hyper-velocity impact experiments, and field geological studies of generally poorly- preserved terrestrial impact craters (Melosh, 1989). The Moon provides a unique record of essentially pristine impact craters of all sizes (from micron-sized pits up to 1000-km impact basins). Field studies, combining sample collection (including drill cores) and in situ geophysical studies (e.g. active seismic profiling), of the ejecta blankets and sub-floor structures of pristine lunar craters of a range of sizes would greatly aid in our understanding of the impact cratering process. As discussed by Crawford (2004), the implied requirements for mobility, deployment of complex geophysical instruments, sub-surface drilling, and sample return capacity are likely to outstrip the capabilities of purely robotic exploration and would be greatly facilitated by a human exploration programme. 2.7 Regolith processes The SCEM Report noted that the lunar surface is a natural laboratory for understanding regolith processes and space weathering on airless bodies throughout the Solar System. In addition to their scientific interest, better characterisation of the composition, volatile content and mechanical properties of lunar regolith will also be important for planning In-Situ Resource Utilisation (ISRU) applications in support of future lunar exploration activities (e.g. Anand et al., 2012; Schwandt et al., 2012). As noted in Section 2.4, the nature of relatively cold high-latitude regoliths, which have never been sampled or studied in situ, and which may contain a volatile component, are of particular interest. Many of the properties of both high- and low-latitude regoliths could be investigated by appropriate instruments on robotic soft-landers such as ESA’s proposed Lunar Lander (Carpenter et al., 2012) or penetrators such as the LunarNet concept (Smith et al., 2012). Another important aspect of the lunar regolith is the record it contains of early solar and Solar System history. Studies of Apollo samples have revealed that solar wind particles are efficiently implanted in the lunar regolith (McKay et al., 1991; Lucey et al., 2006), which therefore contains a record of the composition and evolution of the Sun throughout Solar System history (e.g. Wieler et al., 1996). Recently, samples of the Earth’s early atmosphere appear to have been retrieved from lunar regolith samples (Ozima et al., 2005; 2008), and it has been suggested that samples of Earth’s early crust may also be preserved there in the form of terrestrial meteorites (Gutiérrez,, 2002; Armstrong et al., 2002; Crawford et al., 2008; Armstrong, 2010). Meteorites derived from elsewhere in the Solar System will likely also be found on the Moon, preserving a record of the dynamical evolution of small bodies throughout Solar System history (Joy et al., 2011; 2012). Last but not least, the lunar regolith may contain a record of galactic events, by preserving the signatures of ancient galactic cosmic ray (GCR) fluxes, and the possible accumulation of interstellar dust particles during passages of the Sun through dense interstellar clouds (Crozaz et al., 1977; McKay et al., 1991; Crawford et al., 2010). Collectively, these lunar geological records would provide a window into the early evolution of the Sun and Earth, and of the changing galactic environment of the Solar System, that is unlikely to be obtained in any other way. Much of this record has clear astrobiological implications, as it relates to the conditions under which life first arose and evolved on Earth. From the point of view of accessing ancient Solar System history it will be desirable to find layers of ancient regoliths (palaeoregoliths) that were formed and buried billions of years ago, and thus protected from more recent geological processes, (e.g. Spudis, 1996; Crawford et al., 2010; Fagents et al., 2010; see Figure 1 of Crawford et al., 2010 for a pictorial representation of the process). Locating and sampling such deposits will likely be an important objective of future lunar exploration activities, but they will not be easy to access. Although robotic sampling missions might in principle be able to access palaeoregoliths at a limited number of favorable sites (for example where buried layers outcrop in the walls of craters or rilles), fully sampling this potentially rich geological archive of Solar System history will probably require the mobility and sample return capabilities of human exploration (Spudis, 2001; Garvin, 2004; Crawford, 2004). The specific requirements of a human exploration architecture capable of locating and sampling palaeoregolith deposits have been described elsewhere (Crawford et al., 2010). 2.8 Atmospheric and dust environment The final broad lunar science area discussed by the SCEM Report related to studies of the lunar atmosphere and near-surface dust environment. Of particular interest is the need to characterise the composition of the tenuous lunar exosphere (which has a variable density in the range 105 – 107 atoms or molecules cm-3; Lucey et al., 2006), and its interaction with the surface regolith. This in turn will help constrain models of the lunar volatile budget discussed in Section 2.4. The extent to which transient releases of gasses into the atmosphere may occur is also of interest, as this may correlate with on-going low-level geological activity (Crotts, 2008). The surface dust environment, and especially the extent to which dust grains may become electrostatically charged and transported, is another important research topic (e.g. Grün et al., 2011; Pines et al., 2011; Horanyi and Stern, 2011), not least because of the potential hazards mobile dust may pose to scientific instruments and human operations on the surface (NRC, 2007; Loftus et al., 2010; Linnarsson et al., 2012). Some of the processes involved are likely to be common on other air-less bodies, and quantifying them on the relatively accessible lunar surface will therefore give us better insight into regolith/exosphere interactions throughout the Solar System. Although some aspects of these investigations can be performed from lunar orbit (for example relevant observations will be performed by the Lunar Atmosphere and Dust Environment Explorer, LADEE, mission due for launch in 2013), more detailed studies of the lunar dust and plasma environment will require in situ surface measurements. These may be provided by suitably instrumented robotic landers, such as ESA’s proposed Lunar Lander mission (Carpenter et al., 2012). However, it is important to note that the landed missions have the potential to significantly disturb the tenuous lunar atmospheric environment, and that this will be especially true of human operations (NRC, 2007). It is therefore scientifically highly desirable that the lunar atmosphere/exosphere be properly characterized by minimally invasive robotic probes before human operations are resumed on the lunar surface. 3. Science on the Moon 3.1 Life sciences and astrobiology The Moon is a potentially valuable site to address a range of life science and astrobiology questions (e.g. Crawford, 2006; Gronstal et al., 2007; Cockell 2010; Crawford and Cockell, 2010). These questions broadly fall into three categories: 3.1.1 Research that enhances our understanding of the habitability of the Earth through time As the Earth's closest celestial neighbour the Moon retains a unique record of the inner Solar System environment under which life evolved on our planet. The metamorphism and alteration of terrestrial Archaean (i.e. >2.5 Gyr old) rocks and their organic microfossils limits the quantity of material that can be used to understand the nature of early life on the Earth. The possibility that rocks ejected by asteroid and comet impacts on the early Earth may have landed on the Moon provides a tantalising possibility for a lunar surface source of early Earth material (Gutiérrez,, 2002; Armstrong et al. 2002; Crawford et al., 2008; Armstrong, 2010). The quantity of this material is predicted to be as much as 200 kg/km2. The Moon may also have collected material ejected from other planetary bodies in the Solar System. Early rocks from Mars and Venus, both of which have early histories of enormous biological interest, might also exist on the Moon (Gladman et al. 1996; Armstrong et al. 2002), although the quantity and distribution and condition of this material is more uncertain. In addition, the lunar regolith, and especially buried palaeoregoliths (Spudis 1996), likely contains a record of solar wind flux (and thus solar luminosity) and galactic cosmic rays (and thus the galactic environment of the solar system) throughout solar system history (Crawford et al., 2010; Fagents et al., 2010). Much of this record will be directly relevant to understanding the past habitability of our own planet. 3.1.2 Research that enhances our understanding of the possibility of life elsewhere in the Universe Although the Moon has, almost certainly, never supported any life of its own, lunar exploration will nevertheless inform our searches for life elsewhere. This includes a record of volatile fluxes in the inner solar system (NRC, 2007; Anand, 2010), and information on the survival of both microorganisms and organic matter in extreme planetary conditions. The impacts that occurred into the Moon in its early history included material in addition to terrestrial rocks, such as cometary material and chondritic and carbonaceous meteorites (e.g. Joy et al., 2012). Protected either in the subsurface or in permanently shadowed craters (Seife, 2004; Vasavada et al., 1999), these might provide insights into the inventory of volatile and/or organic material that penetrated the early inner Solar System, and what quantity may still do so today. Insofar as these organics might have provided an exogenous source of prebiotic organics necessary to kick-start life on the Earth, investigating organics on the Moon has important contributions to make to understanding the origin of life on the Earth. Indigenous organic processing on the Moon may also yield insights into the chemical pathways of alteration, and fate, of organics in interplanetary space (e.g., Lucey, 2000). In addition, the lunar environment contains the crashed remains of unsterilized spacecraft. The bacterial spores that these craft may contain could be collected and examined for DNA and other biochemical damage, as well as examined for their viability. Human or robotic explorers could be developed to collect these organisms for study on the Moon or return to the Earth. The organisms returned from these craft would answer many questions about the longevity of microorganisms in the space environment that will inform fields as diverse as planetary protection, for example allowing for an assessment of how long contaminant organisms survive on other planetary surfaces (Rummel, 2004; Glavin et al., 2004, 2010) and biogeography, for example, showing whether organisms can survive the conditions in interplanetary space and the impact conditions of landing on another planetary surface after being transferred from one planet to another (Clark, 2001; Horneck et al., 2001; Mastrapa et al., 2001; Burchell, 2004; Cockell et al., 2007; Nicholson et al., 2005). 3.1.3 Research that advances the human exploration and settlement of space The space environment is hostile to life and includes hard vacuum, high radiation (both UV and ionizing radiation), altered gravity regimes, the presence of biologically and mechanically aggravating dust, and difficulties in acquiring liquid water and gases to breathe (e.g. Horneck, 1996; Horneck et al., 2003). The Moon is therefore a testing ground for technological principles and approaches for dealing with the major environmental parameters that affect life in outer space. For example, the Moon can be used to investigate whether the effects of gravity are linear or whether there are critical threshold in effect (e.g. Cockell, 2010), the biological effects of the radiation environment beyond the Earth’s magnetosphere, and the toxicity of lunar dust (Carpenter et al., 2010; Loftus et al., 2010; Linnarsson et al., 2012). Organisms could also be taken to the lunar surface and used to carry out investigations in situ using surface laboratories (as proposed for ESA’s Lunar Lander; Carpenter et al., 2010). Microorganisms, plants and animals could be used to investigate a variety of questions, including the cumulative effects of space conditions (Mileikowsky et al., 2000; Blakely, 2000; Clark, 2001; Brenner et al., 2003; Giusti et al., 1998; Horneck et al., 2003; Stein and Leskiw ,2000; Zayzafoon et al., 2005; Carpenter et al., 2010) and their use in life support systems (Tamponnet 1996; Sadeh and Sadeh, 1997; Bluem and Paris, 2001; Henrickx et al., 2006). These experiments would yield new insights into the evolution of organisms in the space environment, the possibility of microbial, plant, and cultivated crop production, and the potential for healthy human and animal reproduction in space. 3.2 Human physiology and medicine Acceleration produced by the force of gravity is an omnipresent factor that modulates many biological processes (Clément and Slenzka, 2006). Historically, much of the space biomedical science focus has been upon whole-body physiology, reflecting the immediate challenge of maintaining astronaut (and thus mission) functionality (Garshnek, 1989; Williams, 2003). Most physiological systems are known to progressively ‘adapt’ to life in microgravity (Nicogossian et al., 1994), however many encounter problems upon re-exposure to a gravitation vector. Alterations in spatial orientation (Lipshits et al., 2005) and sensory-motor function (e.g. Kalb and Solomon, 2007; Souvestre et al., 2008) are acutely challenging but re-adapt rapidly, whereas cardiovascular (Hargens and Richardson, 2009; Hughson, 2009) and musculoskeletal system (Narici and de Boer 2011) de-conditioning can precipitate chronic health issues as well as imparing astronaut operations. A raft of countermeasures including various exercise regimes, nutrition and behavioural support are employed in space missions (Convertino, 2002; Cavanagh et al., 2005), but these fail to entirely ameliorate microgravity-induced de-conditioning. As a consequence, the maximum ‘recommended’ stay on the International Space Station (ISS) is currently 6 months (Williams et al., 2009). Partly, this failure relates to the fact that just how gravity affects biological processes is largely unknown, at the level of an organism, a system, or of an individual cell. In fact, whilst life appears well adapted to gravity today, it presented a major challenge to emergence from the prehistoric aquatic environment. Gravitropism (a plant growth’s sensitivity to gravity) has been well documented since Darwin, and it is highly likely that gravity also has an important role in the regulation of animal cells. In fact, intra-cellular force sensors (sensitive to physical pressure and structural strain) have been proposed as part of standard cellular architecture (Wayne et al, 1992). Even minor alterations in the physical force environment (Klaus, 1998) may have significant downstream effects. An abnormal gravity environment is likely to inhibit sedimentation, equalising molecular (and thus affecting electro-chemical gradients) and organelle distribution, and modify cytoskeletal activity and gene transcription (Cogoli and Cogoli-Greuter, 1997). Such processes may in part explain observed gravity-dependent changes in cellular growth, proliferation and regeneration (Sonnenfeld and Shearer, 2002; Borchers, et al., 2002), organisation (Vico et al., 2000), and healing (Davidson et al., 1999; Radek et al., 2008). For instance, reduced lymphocyte proliferation, delayed bone cell differentiation (Hughes-Fulford and Lewis, 1996), and retardation of pre-natal (e.g. Bruce, 2003) and post-natal development (e.g. Ronca and Alberts, 1997) and locomotion (Walton, 1998) have all been observed in a number of animal models in microgravity. Monitoring human adaptation to prolonged exposure to partial gravity, such as exists on the Moon, may offer significant insights into vestibular disorders (Clément et al., 2005) and a range of processes beyond associated in ageing (Vernikos and Schneider, 2010), disuse pathology (Edgerton et al., 2000; Elmann-Larsen and Schmitt, 2003) and lifestyle conditions such as the metabolic syndrome and cardiovascular disease. The lunar partial gravity model would supplement the knowledge accrued by ‘long duration’ ISS microgravity exposure, by adding a point at 1/6th g from which to investigate gravity dose-dependence (e.g. Cockell, 2010). In addition, by providing a more tightly controlled environment in terms of the nutritional, medical (including remote monitoring) and exercise countermeasure support, the Moon would offer an opportunity to test the efficacy of countermeasure and/or therapeutic approaches within otherwise ‘healthy’ and well motivated individuals (e.g. Green, 2010). The lunar partial gravity is likely to offer some protection to de-conditioning that will likely not only facilitate longer operational durations than the ISS, but also has greater relevance and applicability to life on Earth than the microgravity environment of the ISS. There would therefore be the opportunity to understand the physiological effects of gravity that are not possible within either a 1-g or a microgravity environment. Such studies were beyond the scope of the limited duration Apollo missions, but would be facilitated by longer-term human operations on the Moon. Finally, there would be much to learn about life support (e.g. bio-regenerative food, breathable air, and water closed-loops; Ekhart, 1996), and medical support provision, from human operations in a lunar base beyond research into partial gravity effects. Examples include individualised medicine (Kalow 2002), carcinogenesis (Rykova et al., 2008), viral virulence (Wilson et al., 2007), robotic surgery (Cermack, 2006), telemetric medical monitoring (Grigoriev and Egorov, 1997), and even basic health care delivery due to the severe resource and technical assistance limitations (Mortimer et al., 2004). In contrast to the popular myth, space biomedical devices must be simple (to operate and repair), robust, compact, light, low consumers, non-invasive and multi-faceted (e.g. diagnostic and interventional ultrasound; Ma et al., 2007) and ideally preventative (Thirsk et al., 2009). Such characteristics possess value well beyond space applications, e.g. in remote societies and disaster zones, and could also facilitate de-centralised care provision within the developed world offering efficacy and cost-effectiveness benefits (Williams, 2002). It follows that while incorporating humans in future lunar exploration may add to the associated risks and challenges, it would radically enhance both its scientific capabilities and the resulting benefits (including potential biomedical benefits) to society. 3.3 Fundamental physics Although not a major driver for lunar exploration, it is recognized that a number of research fields in the area of fundamental physics may benefit from the ability to place scientific instruments on the lunar surface. These include tests of General Relativity through improved lunar laser ranging measurements (e.g. Livio, 2006; Burns et al., 2009; Currie et al., 2010), tests of quantum entanglement over large baselines (Schneider, 2010), and searches for strange quark matter (Banerdt, et al., 2007; Han et al., 2009). 4. Science from the Moon – Astronomy 4.1 Why Astronomy? A natural area to use the Moon as a platform for performing scientific experiments is astronomy (for summaries see, e.g., Burns et al., 1990; Livio, 2006; Crawford and Zarnecki, 2008; Jester and Falcke, 2009). Almost the entire electromagnetic spectrum is currently being used to study the universe from radio to high-energy gamma ray emission. Different frequencies typically relate to different physical processes, and consequently the universe looks markedly different in optical, infrared, or radio wavelengths. Hence, during the last century modern telescopes have diversified and evolved enormously, fundamentally changing our view of the universe and our place therein. Due to their ever increasing sensitivity, which allows one to peer deeper and deeper into the earliest phases of the cosmos, the requirements for telescope sites have become more and more extreme: one simply needs the best possible observing conditions. The most important factors here are light pollution (at the relevant frequencies) and distortions due to the atmosphere. Light pollution is generally caused by any form of civilization, thereby pushing observatories to more and more remote locations. Detrimental effects of the atmosphere include: • temporary effects such as clouds and water vapour, which temporarily absorb and disturb optical or high-frequency radio radiation, • turbulence in the ionosphere or troposphere, which distorts radio or optical wave fronts, thereby severely degrading the image quality, • air glow, which can overpower sensitive infrared observations, • total absorption of radiation, e.g., of very low-frequency radio, infrared, X-ray, and gamma-ray radiation. The best – and in many cases only – remedy is to observe from dry deserts, high mountains, or from space. Two of the most remote, but also most exquisite, astronomical sites on Earth are the Atacama desert and Antarctica. The former currently hosts some of the world’s largest telescopes, including ESO’s 8m-class Very Large Telescopes (VLT), the ALMA sub-mm-wave radio telescope, and in the future probably also the ~40 m diameter European Extremely Large Telescope (E-ELT; see http:// www.eso.org). A century after its initial exploration, Antarctica now also hosts a number of somewhat smaller telescopes (e.g., the South Pole Telescope, Carlstrom et al., 2011) as well as the giant IceCube detector. IceCube is the world’s largest neutrino observatory, using the ice itself as detector material (e.g., Abbasi et al., 2011). The Moon would be a logical next step in the quest for the most suitable sites to be used for astronomy. An important secondary important factor in selecting a site, however, is the available infrastructure: How accessible is the site for people and material? How does one obtain power and how good is the data connection? Already for Antarctica this poses serious constraints, and it took a long time until this continent became useful for scientific exploitation. It is needless to say that the Moon is even more difficult to reach. Hence, like Antarctica, any significant exploitation of the Moon requires a developed infrastructure – something that would likely become available only in conjunction with human exploration of the Moon. Even then one has to assess how unique and useful the Moon is for astronomy in the first place. After all, the International Space Station (ISS), while having a well-developed infrastructure available, is not used for telescopes; its small, relatively unstable platform in low- Earth orbit (LEO) is simply too poor a telescope site to be competitive. Hence, the vast majority of space-based telescopes have been associated with free-flying satellites. Of course, some of these satellites, most notably the Hubble Space Telescope (HST), benefited from the heavy lift capabilities of the Space Shuttle and the servicing possibilities the human space flight program offered (NRC, 2005). Indeed, it is interesting to note that the one human-serviced space telescope, HST, is in fact the most productive of all astronomy space missions even many years after its launch (see Tables 4 and 6 in Trimble and Ceja, 2008; HST produced 1063 papers in the time frame 2001-2003, compared to 724 for Chandra, the next most productive). So, the question to ask is: Which type of telescopes would uniquely benefit from a lunar surface location? This question has been addressed in a couple of workshops and scientific roadmaps in recent years (Falcke et al., 2006; Livio, 2006; NRC, 2007; Crawford and Zarnecki, 2008; Worms et al., 2009). In the following section we try to synthesize these findings. 4.2 Which astronomy? There is a wide consensus that a low-frequency radio telescope (i.e. a radio telescope operating at frequencies below 30-100 MHz) would be the highest priority (e.g., Jester and Falcke, 2009; Burns et al., 2009). Radio waves at these frequencies are seriously distorted by the Earth’s ionosphere and completely absorbed or reflected at frequencies below 10-30 MHz. Hence, the low-frequency universe is the last uncharted part of the electromagnetic spectrum, and a lunar infrastructure would greatly benefit its exploration. Of particular relevance for science here is the investigation of the “dark ages” of the universe. This is the epoch several hundred million years after the big bang, but before the formation of the first stars and black holes, when the cosmos was mainly filled with dark matter and neutral hydrogen. This epoch contains still pristine information of the state of the big bang and can essentially only be observed through radio emission from atomic hydrogen red-shifted to several tens of MHz. The best location to study this treasure trove of cosmology (Loeb and Zaldariaga 2004) would indeed be on the lunar far-side. Other science topics of interest for a lunar radio telescope include: • A high-resolution map of the universe and a general inventory of low- frequency radio sources – ground-based maps around 10 MHz are very poor compared to any other wavelength due to the effects of the ionosphere (Cane and Whitham 1977); • Search for and study of radio emission from planets and exoplanets (Griessmeier et al. 2007); • Investigation of the local plasma bubble around our solar system; and • Radio-detection of ultra-high energy cosmic rays. In fact the Moon itself is already now being used as a detector for ultra-high energy neutrinos (Gorham et al. 2004, Buitink et al. 2010, Jaeger et al. 2011) and cosmic rays (ter Veen et al. 2010) with the help of radio telescopes. More details on the various science cases, an extensive review of past studies, and a summary of relevant observing constraints can be found in Jester and Falcke (2009). Low-frequency telescopes are currently being built using a network of many, relatively simple dipole-like antennas (e.g., the LOFAR telescope; van Haarlem et al. 2012). The individual antennas are digitally connected to form a large interferometer (a “phased array”), which acts as one large, steerable telescope dish, but without any moving parts. These radio antennas can be made robust and lightweight to transport many of them into space, while the Moon’s gravity and surface would keep them fixed relative to each other. Moreover, if placed at the far-side of the Moon, the Moon itself would shield the telescope from any radio pollution (also called radio frequency interference, RFI) originating from Earth. This RFI can be either of man-made origin or due to natural Auroral Kilometric Radiation (AKR) in the Earth’s magnetosphere. In fact – due to this shielding and lack of its own magnetic activity – the farside of the Moon probably belongs to the most radio-quiet places in our solar system. This has been long recognized and the International Telecommunication Union (ITU) has accordingly designated the lunar far-side as a radio-quiet zone for radio astronomy (ITU Radio Regulations, Article 22, Section V). The installation of such a low- frequency radio array could proceed in several steps, starting with a few antennas and then growing to a larger and larger network (ideally out to tens to hundreds or kilometres with thousands of antennas). While the first steps can certainly be done robotically, a large-scale installation over such diverse terrain and large distances, will ultimately be much more efficient with the presence of humans on the surface. In addition to radio telescopes, a number of other types of lunar observatories have been discussed (for example, see the discussion reported by Crawford and Zarnecki, 2008). The huge cost overruns of the HST-successor, JWST, have shown how difficult it is to fold-up and bring even medium-sized optical telescopes into space. For certain applications liquid-mirror telescopes have been proposed, whereby the combination of a rotating fluid and the Moon’s gravity would form a perfect (non- steerable) mirror that could detect the very first stars in the universe (Angel et al. 2008), although this technique currently has a rather low technical readiness. Perhaps a simpler, and scientifically topical and important, implementation of lunar-based telescopes would be to obtain disc-integrated spectral, polarimetric, and albedo measurements of the Earth. Such observations could provide important insights into both the Earths radiation budget and climate (e.g. Pallé and Goode, 2009), and the interpretation of observations of Earth-like exoplanets and the interpretation of planetary biosignatures (e.g. Sparks et al., 2010; Karalidi et al., 2012). With regard to larger instruments, the lunar surface and gravity may facilitate the construction of interferometers by combining a set of smaller optical, IR, and sub- mm-wave telescopes – a concept similar to the radio interferometer array. However, here shielding is not such an issue and one needs to compare these Moon-based interferometers with free-flying and formation-flying satellites. The latter concept has been the preferred option for small-number-of-elements interferometers (e.g., Cockell et al. 2009). On the other hand, formation flight to the required precision for astronomical interferometers has not been realized so far, while ground-based interferometers are well understood and tested. Moreover, the Moon as a large airless body could host detectors for cosmic ray particles. Those particles are typically absorbed in the Earth’s atmosphere but would reach the lunar surface unimpeded. The detectors could be distributed over the Moon to study solar-wind induced structural variations of the cosmic ray flux at low energies. At energies above 1015 eV the composition of cosmic rays could be directly measured, if a large detector can be built using, e.g., lunar water resources. Here, however, one needs to critically assess whether the science return justifies the undoubtedly high costs for these concepts. Finally, the simple presence of a lunar infrastructure for human activities, may lend itself to the relatively simple installation of smaller telescopes and experiments. Indeed, Apollo 16 deployed a UV telescope on the Moon (Carruthers, 1973) – a class of telescopes that was later flown on satellites (e.g. the International Ultraviolet Explorer, IUE). This would qualify as “opportunistic” science, i.e., missions that simply catch a ride, given it is there. At the very least, an existing ‘science-park’ on the Moon would take the burden of taking along one’s own attitude control, power- supply, or data handling capability, which now has to be carried by every astronomy satellite. Whether by the time such an infrastructure is likely to be available there will still a sufficient science case left for such small telescopes remains to be seen. On the other hand, it is worth noting that, perhaps surprisingly, small robotic telescopes have revolutionized some areas of astronomy (e.g., planet searches, Gamma-Ray-Burst afterglows) in recent years and may continue to do so. To summarize: a low-frequency lunar radio telescope to map the last unexplored frequency window to the universe, and to explore the dark ages of the universe, should be feasible and would be uniquely suited for the Moon’s far-side. Small optical/IR telescopes to observe the Earth from the nearside may also be scientifically valuable. Large (10 metre plus) optical telescopes or interferometers might at some point benefit from the lunar surface as a platform, but the technical case for this still needs to be developed, and the Moon as a site for such telescopes needs to be better studied. Smaller telescopes could be installed, using lunar activities as opportunity. All options would certainly benefit from a sustainable human return to the Moon, but certainly will not drive it. This situation is, however, not very much different from the usage of Antarctica for astronomy, where astrophysicists simply made use of the fact that Antarctic exploration programs and infrastructure already existed. Perhaps the same will happen with the Moon in the coming decades. As a final note: a potentially disturbing factor for many telescopes at wavelengths shorter than radio is the unknown dust mobility on the Moon, which might cover mirrors and joints over a period of time. Also, for radio telescopes the dielectric properties of the lunar surface, the existence of an ionosphere, and the temporal (e.g., through cosmic rays) and spectral radio background variations should be studied to verify the feasibility of the lunar surface for future astronomy applications. Hence, any lunar precursor missions, such as ESA’s proposed Lunar Lander (Carpenter et al., 2008; 2012) should investigate these effects in more detail in order to optimize future astronomical observations from the Moon.. 5 Conclusions Summarising the above, we see that the lunar geological record still has much to tell us about the earliest history of the Solar System, the origin and evolution of the Earth- Moon system, the geological evolution of rocky planets, and the near-Earth cosmic environment throughout Solar System history. These lunar science objectives were strongly endorsed by the US National Research Council Report on the Scientific Context for Exploration of the Moon (NRC 2007), and a number of them (i.e. those which could be addressed by sample return from the South Pole-Aitken Basin and by the creation of a lunar geophysics network) received high priority in the recent US Planetary Science Decadal Survey (NRC, 2011). Addressing these objectives requires an end to the 40-year hiatus of lunar surface exploration, with the placing of scientific instruments on, and the return of samples from, the surface of the Moon, with a particular emphasis on regions not previously visited. It is also clear that the lunar surface offers outstanding opportunities for research in astronomy, astrobiology, fundamental physics, life sciences and human physiology and medicine. Many of these objectives, can be addressed robotically, as reflected by the large number of proposals for lunar surface robotic exploration that have been put forward in recent years (e.g. MoonLITE and MoonRaker, Gao et al., 2008; MoonRise, Jolliff et al., 2010; Lunar Beagle, Gibson et al., 2010; SELENE-2, Hashimoto et al., 2011; Luna-Glob, Mitrofanov et al., 2011; Lunette, Neal et al., 2011; LunarNet, Smith et al., 2012; Farside Explorer, Mimoun et al., 2012; and ESA’s Lunar Lander, Carpenter et al., 2012). However, in the longer term, it is also clear that most of these scientific objectives would benefit from the scientific infrastructure, on the spot decision making, enhanced surface mobility, and sample return capacity that would be provided by renewed human operations on the lunar surface (e.g. Spudis, 2001; Garvin, 2004; Cockell, 2004; Crawford, 2004; 2012). Indeed, some of these scientific objectives will be impossible to conduct robotically (i.e. those related to human physiology and medicine where humans will form the experimental subjects), and others (e.g. deep drilling into the lunar crust to extract undisturbed palaeoregolith deposits with their potentially rich record of Solar System history) may be wholly impractical without a human presence. That said, it is also true that a human return to the Moon will benefit from robotic precursor missions, for example to assess possible seismic and impact hazards, regolith properties (including possible dust toxicity; Linnarsson et al. 2012, this issue), the radiation environment, and in situ resource availability (Anand et al. 2012, this issue). For all these reasons it is highly desirable that current plans for robotic exploration of the lunar surface are developed in the context of a future human exploration programme. Fortunately, such a programme is under active international discussion. In 2007 the World’s space agencies came together to develop the Global Exploration Strategy (GES), which lays the foundations for a global human and robotic space exploration programme (GES 2007). One of the first fruits of the GES has been the development of a Global Exploration Roadmap (GER 2011), which outlines possible international contributions to human and robotic missions to the Moon, near-Earth asteroids and, eventually, Mars. Implementation of this roadmap would provide many opportunities for pursuing the science objectives outlined in this paper, and lunar science would therefore be a major beneficiary of its implementation. Acknowledgements We wish to thank Dr James Carpenter for his enthusiastic support of lunar science in Europe, for organising this Special Issue of Planetary and Space Science, and for his invitation to contribute to it. We thank the two referees (Dr. Wim van Westrenen, the other anonymous), for comments which have greatly improved the quality of the manuscript. References Abbasi, R. et al., 2011 Limits on Neutrino Emission from Gamma-Ray Bursts with the 40 String IceCube Detector, Physical Review Letters, 106, 141101. Anand, M., 2010. Lunar water. Earth Moon Planets, 107, 65-73. Anand, M, Crawford, I.A., Balat-Pichelin, M., Abanades, S., van Westrenen, W., Péraudeau, G., Jaumann, R., Seboldt, W., 2012 A brief review of chemical and mineralogical resources on the Moon and their potential utilization. Planet. Space Sci., (this volume). Angel, R., et al., 2008. A Cryogenic Liquid-Mirror Telescope on the Moon to Study the Early Universe. Astrophys. Journal, 680, 1582-1594. Armstrong, J.C., 2010. Distribution of impact locations and velocities of Earth meteorites on the Moon. Earth Moon Planets, 107, 43-54. Armstrong, J.C., Wells, L.E. and Gonzales, G., 2002. Rummaging through Earth’s attic for remains of ancient life. Icarus, 160:183-196. Banerdt, W. B., Chui, T., Griggs, C.E., Herrin, E.T., Nakamura, Y.,; Paik, H.J., Penanen, K., Rosenbaum, D., Teplitz, V. L., Young, J., 2007. Using the Moon as a low-noise seismic detector for strange quark nuggets, Nuclear Physics B Proceedings Supplements, 166, 203-208. Bernstein, M.P., Dworkin, J.P., Sandford, S.A., Cooper, G.W., and Allamandola, L.J. 2002. The formation of racemic amino acids by ultraviolet photolysis of interstellar ice analogs. Nature, 416:401-403. Blakely, E.A., 2000. Biological effects of cosmic radiation: deterministic and stochastic: setting the framework. Health Phys. 79, 495–506. Bluem, V., Paris, F., 2001. Aquatic modules for bioregenerative life support systems based on the CEBAS biotechnology. Acta Astronaut 48, 287–297. Borchers, A.T., Keen, C.L., Gershwin, M.E., 2002. Microgravity and immune responsiveness: implications for space travel. Nutrition 18(10), 889-898. Brenner, D.J., Doll, R., Goodhead, D.T., Halla, E.J., Land, C.E., Little, J.B., Lubin, J.H., Preston, D.L., Preston, R.J., Puskin, J.S., Ron, E., Sachs, R.K., Samet, J.M., Bruce, L.L., 2003. Adaptations of the vestibular system to short and long-term exposures to altered gravity. Adv Space Res. 32(8), 1533-9. Bruce, L.L., 2003. Adaptations of the vestibular system to short and long-term exposures to altered gravity. Adv Space Res., 32(8), 1533-9. Buitink, S., Scholten, O., Bacelar, J., Braun, R., de Bruyn, A.~G., Falcke, H., Singh, K., Stappers, B., Strom, R.G., Yahyaoui, R.A., 2010. Constraints on the flux of ultra- high energy neutrinos from Westerbork Synthesis Radio Telescope observations. Astron. Astrophys., 521, A47. Burchell, M.J., 2004. Panspermia today. Int. J. Astrobiology 3, 73–80. Burns, J.O., Duric, N., Taylor, G.J., Johnson, S.W., 1990. Observatories on the Moon. Scientific American, 262(3), 18-25. Burns, J.O., et al., 2009. Science from the Moon: The NASA/NLSI Lunar University Network for Astrophysics Research (LUNAR). Submitted as a white paper to the planetary sciences decadal review, ArXiv e-prints arXiv:0909.1509. Cane, H.V., Whitham, P.S., 1977. Observations of the southern sky at five frequencies in the range 2-20 MHz. Mon. Not. Roy. Astron. Soc., 179, 21-29. Carlstrom, J.E., et al., 2011. The 10 Meter South Pole Telescope. Pub. Astron. Soc. Pacific, 123, 568-581. Carpenter, J.D., Houdou, B., Koschny, D., Crawford, I.A., Falcke, H., Kempf, S., Lognonne, P., Ricci, C., Pradier, A., 2008. The MoonNEXT Mission: A European Lander at the Lunar South Pole. Joint Annual Meeting of LEAG-ICEUM-SRR, October 28-31, 2008, Cape Canaveral, Florida; LPI Contributions 1446, 33. Carpenter, J.D., Angerer, O., Durante, M., Linnarson, D., Pike, W.T., 2010. Life siciences investigations for ESA’s first lunar lander. Earth Moon Planets, 107, 11-23. Carpenter, J.D., Fisackerly, R., Pradier, A., Houdou, B., De Rosa, D., Gardini., B., 2012. Science and payload activities in support of the ESA lunar lander. Lunar and Planetary Science Conference, 43, 1990. Carruthers, G.R., 1973. Apollo 16 far-ultraviolet camera/spectrograph: Instrument and operations. Applied Optics,12, 2501-2508. Cavanagh, P.R., Licata, A.A., Rice, A.J., 2005. Exercise and pharmacological countermeasures for bone loss during long-duration space flight. Gravit. Space Biol. Bull. 18(2), 39-58. Cermack, M., 2006. Monitoring and telemedicine support in remote environments and in human space flight. Brit. J. Anaesth. 97(1), 107-114. Chyba, C.F., Sagan, C.,1992. Endogenous production, exogenous delivery and impact-shock synthesis of organic molecules: an inventory for the origins of life. Nature, 355, 125-132. Clark, B.C., 2001. Planetary interchange of bioactive material: probability factors and implications. Orig. Life Evol. Biosph. 31, 185–197. Clark, R.N., 2009. Detection of adsorbed water and hydroxyl on the Moon. Science, 236, 562-564. Clément, G., Reschke, M.S., Wood, S., 2005. Neurovestibular and sensorimotor studies in space and Earth benefits. Curr. Pharm. Biotechnol. 6(4), 267-283. Clément, G., Slenzka, K., (eds.), 2006. Fundamentals of Space Biology: Research on Cells, Animals and Plants in Space (Springer Science Business Media, New York). Cockell, C.S., 2004. The value of humans in the biological exploration of space. Earth Moon Planets 94 233-243. Cockell, C.S., 2010. Astrobiology – what can we do on the Moon? Earth, Moon Planets 107, 3-10. Cockell, C.S, Brack, A., Wynn-Williams, D.D., Baglioni, P. Demets, R., Edwards, H., Gronstal, A., Kurat, G., Lee, P., Osinski, G.R., Pearce, D., Pillinger, J., Roten, C.A., Sancisi-Frey, S., 2007. Interplanetary transfer of photosynthesis: an experimental demonstration of a selective dispersal filter in planetary island biogeography. Astrobiology 7, 1–9. Cockell, C.S. et al., 2009. Darwin-a mission to detect and search for life on extrasolar planets. Astrobiology 9, 1-22. Cogoli, A., Cogoli-Greuter, M., 1997. Activation and proliferation of lymphocytes and other mammalian cells in microgravity. Adv. Space Biol. Med. 6, 33–79. Cohen, B.A., et al., 2010. NASA's International Lunar Network Anchor Nodes and Robotic Lunar Lander Project Update, Lunar Exploration Analysis Group 2010, 14. Cohen, B. A., Coker, R. F., 2010. Pulling Marbles from a Bag: Deducing the Regional Impact History of the SPA Basin from Impact-Melt Rocks, Lunar and Planetary Science Conference, 41, 2475. Colaprete, A., Schultz,P., Heldmann, J., Wooden,D., Shirley, M., Ennico, K., Hermalyn, B., Marshall, W., Ricco, A., Elphic, R.C., Goldstein, D., Summy, D., Bart, G.D., Asphaug, E., Korycansky D., Landis D., Sollitt L., 2010. Detection of water in the LCROSS ejecta plume. Science, 330, 463-468. Convertino, V.A., 2002. Planning strategies for development of effective exercise and nutrition countermeasures for long-duration space flight. Nutrition 18(10), 880–888. Crawford, I.A., 2004. The scientific case for renewed human activities on the Moon. Space Policy, 20, 91-97. Crawford, I.A., 2006. The Astrobiological Case for Renewed Robotic and Human Exploration of the Moon. Internat. J. Astrobiology, 5, 191-197. Crawford, I.A., 2012. Dispelling the myth of robotic efficiency: why human space exploration will tell us more about the Solar System than relying on robotic exploration alone. Astronomy and Geophysics, 53, 2.22-2.26. Crawford, I.A., Cockell, C.S., 2010. Astrobiology on the Moon, Astronomy and Geophysics, 51, 4.11-4.14. Crawford, I.A., Zarnecki, J., 2008., Astronomy from the Moon, Astronomy and Geophysics, 49, 2.17-2.19. Crawford, I.A., Baldwin, E.C., Taylor, E.A., Bailey, J. and Tsembelis, K., 2008. On the survivability and detectability of terrestrial meteorites on the Moon. Astrobiology, 8, 242-252. Crawford, I.A., Fagents, S.A., Joy, K.H., Rumpf, M.E., 2010. Lunar palaeoregolith deposits as recorders of the galactic environment of the Solar System and implications for astrobiology. Earth Moon Planets, 107, 75-85. Crider, D. H. and Vondrak, R. R., 2002. Hydrogen migration to the lunar poles by solar wind bombardment of the Moon. Adv. Space Res., 30, 1869-1874. Crotts, A.P.S., 2008. Lunar Outgassing, Transient Phenomena, and the Return to the Moon. Astrophysical Journal, 687, 692-705. Crozaz, G., Poupeau, G., Walker, R.M., Zinner, E., Morrison, D.A., 1977. The record of solar and galactic radiations in the ancient lunar regolith and their implications for the early history of the Sun and Moon. Phil. Trans. Royal Soc., A285, 587-592. Currie, D. G., Delle Monache, G., Dell'Agnello, S., 2010. A lunar laser retroreflector for the for the 21st century (LLRRA-21): selenodesy, science and status. American Geophysical Union, Fall Meeting 2010, abstract #P51C-1467. Davidson, J.M., Aquino, A.M., Woodward, S.C., Wilfinger, W.W., 1999. Sustained microgravity reduces intrinsic wound healing and growth factor responses in the rat. FASEB J. 13, 325–329. Eckart P., 1996. Spaceflight Life Support and Biospherics, Microcosm Press, Torrance, CA. Edgerton, V.R., Roy, R.R.., Hodgson, JA., Day, M.K., Weiss, J., Harkema, S.J., Dobkin, B., Garfinkel, A., Konigsberg, E., Koslovskaya, I., 2000. How the science and engineering of spaceflight contribute to understanding the plasticity of spinal cord injury. Acta Astronaut. 47(1), 51-62. Elsila, J.E., Dworkin, J.P., Bernstein, M.P., Martin, M.P., Sandford, S.A., 2007. Mechanisms of Amino Acid Formation in Interstellar Ice Analogs. Astrophys. J., 660, 911-918. Elmann-Larsen, B., Schmitt, B,D., 2003. Staying in bed to benefit ESA's astronauts and Europe's citizens. ESA Bull. 113, 34-39. ESA, 1992. Mission to the Moon: Europe’s priorities for the scientific exploration and utilization of the Moon, ESA SP-1150. Fa, W., Wieczorek, M.A., Heggy, E., 2011. Modeling polarimetric radar scattering from the lunar surface: Study on the effect of physical properties of the regolith layer. Journal Geophys. Res., 116, CiteID E03005. Fagents, S.A., Rumpf, M.E., Crawford, I.A. and Joy, K.H., 2010. Preservation potential of implanted solar wind volatiles in lunar paleoregolith deposits buried by lava flows. Icarus, 207, 595-604. Falcke, H. et al., 2006. Towards a European Infrastructure for Lunar Observatories – Joint Statement, Workshop Bremen 2006 (http://www.astron.nl/moon/). Feldman, W.C., Maurice, S., Binder, A.B., Barraclough, B.L., Elphic, R.C., Lawrence, D.J., 1998. Fluxes of fast and epithermal neutrons from Lunar Prospector: evidence for water ice at the lunar poles. Science, 281, 1496-1500. Fuller M, and S.M. Cisowski, 1987. Lunar paleomagnetism. In: Geomagnetism, Jacobs J.A. (ed), Academic Press, p 307-455. Gao, Y., Phipps, A., Taylor, M., Crawford, I.A., Ball, A.J., Wilson, L., Parker, D., Sweeting, M., da Silva Curie, A., Davies, P., Baker, A., Pike, W.T., Smith, A., and Gowen, R., 2008. Lunar science with affordable small spacecraft technologies: MoonLITE and Moonraker. Planet. Space Sci., 56, 368-377. Garshnek, V., 1989. Soviet space flight: the human element. Aviation Space Environ. Med. 60(7), 695-705. Garvin, J., 2004. The science behind the vision for US space exploration: the value of a human-robotic partnership. Earth Moon Planets 94 221-232 GES, 2007. The Global Exploration Strategy: Framework for Coordination (http://esamultimedia.esa.int/docs/GES_Framework_final.pdf). GER, 2011. The Global Exploration Roadmap (http://www.nasa.gov/pdf/591067main_GER_2011_small_single.pdf). Gibson, E.K., Pillinger, C.T., Waugh, L.J., 2010. Lunar Beagle and lunar astrobiology. Earth Moon Planets, 107, 25-42. Giusti, A.M., Raimondi, M., Ravagnan, G., Sapora, O., Parasassi, T., 1998. Human cell membrane oxidative damage induced by single and fractionated doses of ionizing radiation: a fluorescence spectroscopy study. Int. J. Radiat. Biol. 74, 595–605. Gladman, B.J., Burns, J.A., Duncan, M., Lee, P., Levison, H.F., 1996. The exchange of impact ejecta between terrestrial planets. Science 271, 1387–1392. Glavin, D.P., Dworkin, J.P., Lupisella, M., Kminek, G., Rummel, J.D., 2004. Biological contamination studies of lunar landing sites: implications for future planetary protection and life detection on the Moon and Mars. Int. J. Astrobiology, 3, 265–271. Glavin, D.P., Dworkin, J.P., Lupisella, M., Williams, D.R., Kminek, G., Rummel, J.D., 2010. In situ biological contamination studies of the Moon: implications for planetary protection and life detection missions. Earth, Moon, Planets, 107, 87-93. Glotch, T.D., Lucey, P.G., Bandfield, J.L., Greenhagen, B.T., Thomas, I.R., Elphic, R.C., Bowles, N., Wyatt, M.B., Allen, C.C., Hanna, K.D., Paige, D.A., Highly silicic compositions on the Moon. Science 329, 1510-1513, (2010). Gorham, P.W., Hebert, C.L., Liewer, K.M., Naudet, C.J., Saltzberg, D., Williams, D., 2004. Experimental limit on the cosmic diffuse ultrahigh energy neutrino flux. Physical Rev. Lett., 93, 041101. Green, D.A., 2010. How the UK can lead the terrestrial translation of biomedical advances arising from lunar exploration activities. Earth, Moon Planets, 107, 127- 146. Greenhagen, B.T., et al., Global silicate mineralogy of the Moon from the Diviner lunar radiometer. Science, 329, 1507-1509, (2010). Griessmeier, J.-M., Zarka, P., Spreeuw, H.,2007. Predicting low-frequency radio fluxes of known extrasolar planets. Astron. Astrophys., 475, 359-368. Grigoriev, A.I., Egorov. A.D., 1997. Medical monitoring in long-term space missions. Adv Space Biol. Med. 6, 167-191. Gronstal, A.,Cockell, C.S., Perino, M.A., Bittner, T., Clacey, E., Clark, O., Ingold, O., Alves de Oliveira, C., and Wathiong, S., 2007. Lunar Astrobiology: A Review and Suggested Laboratory Equipment. Astrobiology, 7:767-782. Grün, E., Horanyi, M., Sternovsky, Z., 2011. The lunar dust environment. Planet. Space Sci., 59, 1672-1680. Gutiérrez, J.L., 2002. Terrene meteorites in the moon: its relevance for the study of the origin of life in the Earth. ESA SP-518, 187 – 191. Halekas, J.S., Lillis, R.J., Lin, R.P., Manga, M., Purucker, M.E., Carley, R.A., 2010. How strong are lunar crustal magnetic fields at the surface? Considerations from a re- examination of the electron reflectometry technique. J. Geophys. Res., 115, E03006. Han, K., and the LSSS Collaboration, 2009. Search for stable strange quark matter in lunar soil using the mass spectrometry technique. Journal of Physics G: Nuclear and Particle Physics, 36, 064048. Hargens A.R., Richardson, S., 2009. Cardiovascular adaptations, fluid shifts, and countermeasures related to space flight. Respir. Physiol. Neurobiol. 169, S30-S33. Hartmann, W.K., Ryder, G., Dones, L., Grinspoon, D., 2000. The time dependent intense bombardment of the primordial Earth-Moon system. In Origin of the Earth and Moon, eds. Canup, R.M., Righter, K., University of Arizona Press, 493-512. Hashimoto, T., Hoshino, T., Tanaka, S., Otsuki, M., Otake, H., Morimoto, H., 2011. Japanese moon lander SELENE-2—Present status. Acta Astronautica., 68, 1386- 1391. Hendrickx, L., De Wever, H., Hermans, V., Mastroleo, F. , Morin, N., Wilmotte, A., Janssen, P., Mergeay, M., 2006. Microbial ecology of the closed artificial ecosystem MELiSSA (Micro-Ecological Life Support System Alternative): reinventing and compartmentalizing the Earth’s food and oxygen regeneration system for long-haul space exploration missions. Res. Microbiol., 157, 77–86. Hiesinger, H., Head, J.W., 2006. New views of lunar geoscience: an introduction and overview. Rev. Min. Geochem., 60, 1-81. Hiesinger, H., van der Bogert, C.H., Pasckert, J.H., Funcke, L., Giacomini, L., Ostrach, L.R., Robinson, M.S., 2012. How old are young lunar craters? J. Geophys. Res., 117, E00H10. Horanyi, M., Stern, A., 2011. Lunar dust, atmosphere and plasma: The next steps. Planet. Space Sci. 59, 1671). Horneck, G., 1996. Life sciences on the Moon. Adv. Space Res. 18, 95–101. Horneck, G., Rettberg, P., Reitz, G., Wehner, J., Eschweiler, U., Strauch, K., Panitz, C., Starke, V., Baumstark-Kahn, C., 2001. Protection of bacterial spores in space, a contribution to the discussion on panspermia. Orig. Life Evol. Biosph. 31, 527–547. Horneck, G., Facius, R., Reichert, M., Rettberg, P., Seboldt, W., Manzey, D., Comet, B., Maillet, A., Preiss, H., Schauer, L., Dussap, C.G., Poughon, L., Belyavin, A., Reitz, G., Baumstark-Khan, C., Gerzer, R.., 2003. Humex, a study on the survivability and adaptation of humans to long-duration exploratory missions, part I: lunar missions. Adv. Space Res. 31, 2389–2401. Hughes-Fulford M, Lewis M.L., 1996. Effects of microgravity on osteoblast growth activation. Exp. Cell. Res. 224(1),103-9. Hughson, R.L., 2009. Recent findings in cardiovascular physiology with space travel. Respir. Physiol. Neurobiol. 169, S38-S41. Jaeger, T.R., Mutel, R.L., Gayley, K.G., 2010. Project RESUN, a Radio EVLA Search for UHE Neutrinos. Astroparticle Physics, 34, 293-303. Jaumann, R., et al., 2012. Lunar Exploration and Scientific Results: Current Status of Lunar Research. Planet. Space Sci. (in press; this volume). Jester, S., Falcke, H., 2009. Science with a lunar low-frequency array: from the dark ages of the Universe to nearby exoplanets. New Astron. Rev., 53, 1-26. Jolliff, B.L., Gillis, J.J., Haskin, L.A., Korotev, R.L., Wieczorek, M.A., 2000. Major lunar crustal terranes: surface expressions and crust-mantle origins. J. Geophys. Res., 105, 4197-4216. Jolliff, B. L., Shearer, C., Gaddis, L. R., Pieters, C. M., Head, J. W., Haruyama, J., Jaumann, R., Ohtake, M., Osinski, G., Papanastassiou, D. A., Petro, N. E., 2010. MoonRise: Sampling South Pole-Aitken Basin as a Recorder of Solar System Events. American Geophysical Union, Fall Meeting 2010, abstract #P43A-01. Jolliff, B.L.,et al., 2011. Non-mare silicic volcanism on the lunar farside at Compton- Belkovich. Nature Geoscience, 4, 566-571. Joy, K.H., Kring, D.A., Bogard, D.D., McKay, D.S., Zolensky, M.E.. 2011. Re- examination of the formation ages of the Apollo 16 regolith breccias. Geochimica et Cosmochimica Acta, 75, 7208-7225. Joy, K.H., Zolensky, M.E., Nagashima, K., Huss, G.R., Kent, Ross, D.K., McKay, D.S., Kring, D.A., 2012. Direct detection of projectile relics from the end of the lunar basin–forming epoch. Science, (in press; DOI: 10.1126/science.1219633). Kalb, R., Solomon, D., 2007. Space exploration: Mars, and the nervous system. Arch. Neurol. 64, 485–490. Kalow, W., 2002. Pharmacogenetics and personalised medicine. Fundam. Clin. Pharmacol. 16(5), 337-342. Karalidi, T., Stam, D.M., Snik, F., Bagnulo, S., Sparks, W.B., Keller, C.U., 2012. Observing the Earth as an exoplanet with LOUPE, the Lunar Observatory for Unresolved Polarimetry of Earth. Planet. Space Sci., (this issue). Kiefer, W.S., 2012. Lunar heat flow experiments: science objectives and a strategy for minimising the effects of lander-induced perturbations. Planet Space Sci., 60, 155- 165. Klaus., D.M., 1998. Microgravity and its implication for fermentation biotechnology. Trends Biotechnol. 16(9), 369-73. Korotev, R.L., 2005. Lunar geochemistry as told by lunar meteorites, Chemie der Erde, 65, 297-346. Langseth, M.G., Keihm, S.J., Peters, K.., 1976. Revised lunar heat-flow values, Lunar Planet. Sci. Conf., 7, 3143-3171. Le Feuvre, M., Wieczorek, M.A., 2011. Nonuniform cratering of the Moon and a revised crater chronology of the inner Solar System. Icarus, 214, 1-20. Linnarsson, D., Carpenter, J., Fubini, B., Gerde, P., Karlsson, L.L., Loftus, D.J., Prisk, G.K., Staufer, U., Tranfield, E.M., van Westrenen, W., 2012. Toxicity of lunar dust. Planet. Space. Sci., (in press; this volume) Lipshits, M., Bengoetxea, A., Cheron, G., McIntyre, J., 2005. Two reference frames for visual perception in two gravity conditions. Perception, 34(5), 545-555. Livio, M., 2006. Astrophysics enabled by the return to the Moon: a brief summary of highlights based on the workshop held at Space Telescope Science Institute November 28–30, 2006 (available on line at: http://www.star.ucl.ac.uk/~iac/STScI_Report.pdf). Loeb, A., Zaldarriaga, M., 2004. Measuring the small-scale power spectrum of cosmic density fluctuations through 21cm tomography prior to the epoch of structure formation. Phys. Rev. Lett., 92, 211301. Loftus, D.J., Rask, J.C., McCrossin, C.G., Tranfield, E.M., 2010. The chemical reactivity of lunar dust: from toxicity to astrobiology. Earth Moon Planets, 107, 95- 105. Lognonné, P., 2005. Planetary Seismology, Ann. Rev. Earth. Planet. Sci., 33, 571- 604. Lucey, P.G., 2000. Potential for prebiotic chemistry at the poles of the Moon. Proc. SPIE, 4137, 84-88. Lucey, P.G., Korotev, R.L., Gillis, J.J., Taylor, L.A., Lawrence, D., Campbell, B.A., Elphic, R., Feldman, B., Hood, L.L., Hunten, D., Mendillo, M., Noble, S., Papike, J.J., Reedy, R.C., Lawson, S., Prettyman, T., Gasnault, O., Maurice, S., 2006. Understanding the lunar surface and space-Moon interaction. Rev. Min. Geochem. 60, 82–219. Ma, O.J., Norvell, J.G., Subramanian, S., 2007. Ultrasound applications in mass casualties and extreme environments. Crit Care Med. 35(5 Suppl), S275-S279. Maher, K.A., Stevenson, D. 1988. Impact frustration of the origin of life, Nature, 331, 612-614. Mastrapa, R.M.E., Glanzberg, H.,.Head, J.N., Melosh, H.J., Nicholson, W.L., 2001. Survival of bacteria exposed to extreme acceleration: implications for panspermia. Earth Planet. Sci. Lett. 189, 1–8. McKay, D.S., Heiken, G.H., Basu, A., Blanford, G., Simon, S., Reedy, R., French, B.M. and Papike, J., 1991. The lunar regolith. In: Heiken, G.H., Vaniman, D. and French, B.M. (Eds.), The Lunar sourcebook: A user's guide to the Moon, Cambridge University Press, pp. 285-356. Melosh, H.J., 1989. Impact Cratering: A Geologic Process, Oxford University Press, Oxford. Mileikowsky, C., Cucinotta, F.A., Wilson, J.W., Gladman, B., Horneck, G., Lindegren, L., Melosh, J., Rickman, H., Valtonen, M., Zheng, J.Q., 2000. Risks threatening viable transfer of microbes between bodies in our solar system. Planet. Space Sci. 48, 1107–1115. Mimoun, D., et al., 2012. Farside explorer: unique science from a mission to the farside of the Moon, Experimental Astronomy, (in press; DOI: 10.1007/s10686-011- 9252-3). Mitrofanov, I. G., Zelenyi, L. M., Tret'yakov, V. I., Dolgopolov, V. P., 2011. Science program of lunar landers of "Luna-Glob" and "Luna-Resource" missions. Lunar and Planetary Science Conference, 42, 1798. Mortimer, A.J., DeBakey, M.E., Gerzer, R., Hansen, R., Sutton, J., Neiman, S.N., 2004. Life science research in space brings health on Earth. Acta Astronaut. 54, 805- 812. Narici, M.V., de Boer, M.D., 2011. Disuse of the musculo-skeletal system in space and on Earth. Eur. J. Appl. Physiol., 111(3), 403-20. Neal, C. R., Banerdt, W. B., Alkalai, L., 2011. Lunette: a two-lander Discovery-class geophysics mission to the Moon. Lunar and Planetary Science Conference, 42, 2832. Neukum, G., Ivanov, B. A., Hartmann, W. K., 2001. Cratering Records in the Inner Solar System in Relation to the Lunar Reference System. Space Science Reviews, 96, 55-86. Nicholson, W.L., Schuerger, A.C., 2005. Bacillus subtilis spore survival and expression of germination-induced bioluminescence after prolonged incubation under simulated Mars atmospheric pressure and composition: implications for planetary protection and lithopanspermia. Astrobiology 5, 536–544. Nicogossian, A.E, Huntoon, C.L., Pool, S.L. (eds.), 1994. Space physiology and medicine. (3rd ed., Lea and Febiger, Philadelphia, USA, 1994) pp. 167–193. NRC, 2005. Assessment of Options for Extending the Life of the Hubble Space Telescope. National Research Council, National Academies Press, Washington DC. NRC, 2007. The Scientific Context for Exploration of the Moon (the ‘SCEM Report’). National Research Council, National Academies Press, Washington DC. NRC, 2011. Vision and voyages for planetary science in the decade 2013-2022. National Research Council, National Academies Press, Washington DC. Ohtake, M., et al., 2009. The global distribution of pure anorthosite on the Moon. Nature, 461, 236-240. Ozima, M., Seki, K., Terada, N., Miura, Y.N., Podosek, F.A. and Shinagawa, M., 2005. Terrestrial nitrogen and noble gases in lunar soils. Nature 436:655-659. Ozima, M., Yin, Q.-Z., Podosek, F.A and Miura, Y.N., 2008. Toward prescription for approach from terrestrial noble gas and light element records in lunar soils understanding early Earth evolution. Proc. Nat. Acad. Sci. 105:17654–17658. Pallé, E., Goode, P.R., 2009. The Lunar Terrestrial Observatory: Observing the Earth using photometers on the Moon’s surface. Adv. Space. Res., 43, 1083-1089. Paige, D.A., et al., 2010. Diviner Lunar Radiometer observations of cold traps in the Moon’s south polar region. Science. 330, 479-482. Pierazzo, E., Chyba, C.F., 1999. Amino acid survival in large cometary impacts. Meteorit. Planet. Sci., 34, 909-918. Pieters, C.M. et al., 2009. Character and spatial distribution of OH/H2O on the surface of the Moon seen by M3 on Chandrayaan-1. Science, 326, 568-572. Pines, V., Zlatkowski, M., Chait, A., 2011. Lofted charged dust distribution above the Moon surface. Planet. Space Sci., 59, 1795-1803. Purucker, M.E., Nicholas, J.B., 2011. Global spherical harmonic models of the internal magnetic field of the Moon based on sequential and coestimation approaches. J. Geophys. Res., 115, E12007. Radek, K.A., Baer L.A., Eckhardt, J., DiPietro, L.A., Wade, C.E., 2008. Mechanical unloading impairs keratinocyte migration and angiogenesis during cutaneous wound healing. J. Appl. Physiol. 104, 1295–1303. Ronca, A.E, Alberts, J.R, 1997. Altered vestibular function in fetal and newborn rats gestated in space. J. Gravi.t Physiol. 4(2), P63-6. Rummel, J.D., 2004. Strep, lies, and 16mm Film: did S. mitis survive on the Moon? Should humans be allowed on Mars? Int. J. Astrobiology 3 (Suppl.), S7–S8. Ryder, G., 2003. Bombardment of the Hadean Earth: wholesome or deleterious? Astrobiology, 3, 3-6. Rykova, M., Antropova, E., Larina, I.M., Morukov, B.V., 2008. Humoral and cellular immunity in cosmonauts after the ISS missions. Acta Astronaut. 63, 697–705. Sadeh, W.Z., Sadeh, E., 1997. An integrated engineered closed/controlled ecosystem for a lunar base. Adv. Space Res. 20, 2001–2008. Seife, C., 2004. Moon’s ‘abundant resources’ largely an unknown quantity. Science 303, 1603. Schneider, J., 2010. Test of Quantum Physics from Earth-Moon correlations. European Planetary Science Congress 2010, Abstract #202. Schwandt, C., Hamilton J.A., Fray, D.J., Crawford, I.A., 2012. The production of oxygen and metal from lunar regolith. Planet. Space Sci., (this issue). Shearer, C. K., Borg, L.E., 2006. Big returns on small samples: Lessons learned from the analysis of small lunar samples and implications for the future scientific exploration of the Moon. Chemie der Erde, 66, 163-185. Sleep N.H., Zahnle K.J., Kasting K.F. and Morowitz H.J., 1989. Annihilation of ecosystems by large asteroid impacts on the early Earth, Nature, 342, 139-142. Smith, A., et al., 2012. Lunar Net—a proposal in response to an ESA M3 call in 2010 for a medium sized mission, Experimental Astronomy, (in press; DOI: 10.1007/s10686-011-9250-5). Sonnenfeld, G., Shearer, W.T., 2002. Immune function during space flight. Nutrition, 18, 899–903. Souvestre, P. Blaber A., Landrock, C., 2008. Space motion sickness: the sensory motor controls and cardiovascular correlation. Acta Astronaut. 63, 745–757. Sparks, W. B., Meadows, V., McCullough, P., Postman, M., Bussey, B., Christian, C., 2010. Lunar Based Observations of the Earth as a Planet. Astrobiology Science Conference 2010, Abstract No. 5397. Spudis, P.D., 1996. The Once and Future Moon. Smithsonian Institution Press, Washington D.C. Spudis. P.D., 2001. The case for renewed human exploration of the Moon. Earth Moon Planets 87 159-171. Spudis, P.D. et al., 2010. Initial results for the north pole of the Moon from Mini- SAR, Chandrayaan-1 mission. Geophysical Research Letters, 37, CiteID L06204. Sridharan, R., Ahmed, S. M., Pratim Das, T., Sreelatha, P., Pradeepkumar, P., Naik, N., Supriya, G., 2010. The sunlit lunar atmosphere: a comprehensive study by CHACE on the Moon Impact Probe of Chandrayaan-1. Planet. Space Sci., 58, 1567- 1577. Stein, T.P., Leskiw, M.J., 2000. Oxidant damage during and after spaceflight. Am. J. Physiol. Endocrinol. Metab. 278, E375–E382. Stöffler, D., Ryder, G., Ivanov, B.A.,, Artemieva, N.A., Cintala, M.J., Grieve, R.A.F., 2006. Cratering History and Lunar Chronology, Rev. Min. Geochem., 60, 519 – 596. Sunshine, J.M., Farnham, T.L., Feaga, L.M., Groussin, O., Merlin, F., Milliken, R.E., A'Hearn, M.F., 2009. Temporal and Spatial Variability of Lunar Hydration As Observed by the Deep Impact Spacecraft. Science, 326, 565-568. Sunshine, J., Besse, S., Petro, N., 2010. Hidden from plain sight: spinel-rich deposits on the nearside of the Moon as revealed by Moon Mineralogy Mapper (M3). Lunar and Planetary Science Conference, 41, 1508. Talboys, D. L., Barber, S., Bridges, J. C., Kelley, S. P., Pullan, D., Verchovsky, A. B., Butcher, G., Fazel, A., Fraser, G. W., Pillinger, C. T., Sims, M. R., Wright, I. P., 2009. In situ radiometric dating on Mars: Investigation of the feasibility of K-Ar dating using flight-type mass and X-ray spectrometers, Planet. Space Sci., 57, 1237- 1245. Tamponnet, C., 1996. Life support systems for lunar missions. Adv. Space Res. 18, 103–110. Taylor, S.R., Pieters, C.M., MacPherson, G.J., 2006. Earth-Moon system, planetary science, and lessons learned, Rev. Min. Geochem., 60, 657 – 704. Ter Veen, S., Buitink, S., Falcke, H., James, C.W., Mevius, M., Scholten, O., Singh, K., Stappers, B., de Vries, K.D., 2010. Limit on the ultrahigh-energy cosmic-ray flux with the Westerbork synthesis radio telescope. Phys. Rev.D, 82, 103014. Thirsk, R., Kuipers, A., Mukai, C., Williams, D., 2009. Spinoffs from space. CMAJ, 180, 1324-1325. Trieloff, M., Jessberger, E. K., Hiesinger, H., Schwarz, W. H., Hopp, J., Burfeindt, J., Bernhard, H.-G., Hofmann, P., Li, X., Breitkreutz, H., 2010. ISAGE - in situ dating of planetary surfaces, EPSC 2010, 754. Trimble, V., Ceja, J. A., 2008. Productivity and impact of astronomical facilities: three years of publications and citation rates. Astronomische Nachrichten, 329,.632- 647. van Haarlem, M. et al., 2012. LOFAR: The Low Frequency Array. Astron. Astrophys., (to be submitted). Vaniman, D., Dietrich, J., Taylor, G.J. and Heiken, G., 1991. Exploration, samples, and recent concepts of the Moon, 1991. In: Heiken, G.H., Vaniman, D. and French, B.M. (Eds.), The Lunar sourcebook: A user's guide to the Moon, Cambridge University Press. Vasavada, A.R., Paige, D.A., Wood, S.E., 1999. Near surface temperatures on Mercury and the Moon and the stability of polar ice deposits. Icarus 141, 179–193. Vernikos, J., Schneider, V.S., 2010. Space, gravity and the physiology of aging: parallel or convergent disciplines? Gerontology 56(2), 157-166. Vico, L., Collet, P., Guignandon, A., Lafage-Proust, M..H., Thomas, T., Rehaillia, M.. Alexandre, C., 2000. Effects of longterm microgravity exposure on cancellous and cortical weightbearing bones of cosmonauts. Lancet 355, 1607–1611. Walton, K., 1998. Postnatal development under conditions of simulated weightlessness and space flight. Brain Res. Rev. 28, 25-34. Wayne, R, Staves, M.P, Leopold, A.C, 1992. The contribution of the extracellular matrix to gravisensing in characean cells. J. Cell Sci.101, 611-23. Weber, R.C., Lin, P.-Y., Garnero, E.J.; Williams, Q., Lognonné, P., 2011. Seismic Detection of the Lunar Core. Science, 331, 309-312. Wieczorek, M.A., et al., 2006. The constitution and structure of the lunar interior, Rev. Min. Geochem., 60, 221 – 364. Wieler, R., Kehm, K., Meshik, A.P., Hohenberg, C.M.,1996. Secular changes in the xenon and krypton abundances in the solar wind recorded in single lunar grains. Nature, 384:46-49. Williams, D.R., 2002. Bioastronautics: optimizing human performance through research and medical innovations. Nutrition, 18(10), 794-796. Williams, D.R., 2003. The biomedical challenges of space flight. Ann. Rev Med. 54, 245-256. Williams, D., Kuipers, A., Mukai, C., Thirsk, R., 2009. Acclimation during space flight: effects on human physiology. CMAJ, 180, 1317-1323. Wilson, J.W. et al., 2007. Spaceflight alters bacterial gene expression and virulence and reveals a role for global regulator Hfq. Proc Natl. Acad. Sci. USA, 104(41), 16299-16304. Worms, J.-C., et al., 2009. ESSC-ESF Position Paper-Science-Driven Scenario for Space Exploration: Report from the European Space Sciences Committee (ESSC). Astrobiology, 9, 23-41. Yamada, R., Garcia, R. F.;, Lognonné, P., Le Feuvre, M., Calvet, M., Gagnepain- Beyneix, J., 2011. Optimisation of seismic network design: application to a geophysical international lunar network. Planet. Space Sci., 59, 343-354. Yamamoto, S., Nakamura, R., Matsunaga, T., Ogawa, Y., Ishihara, Y., Morota, T., Hirata, N., Ohtake, M., Hiroi, T., Yokota, Y., Haruyama, J., 2010. Possible mantle origin of olivine around lunar impact basins detected by SELENE. Nature Geoscience, 3, 533-536. Zayzafoon, M., Meyers, V.E., McDonald, J.M., 2005. Microgravity: the immune response and bone. Immunol. Rev. 208, 267–280. Zuber, M.T., et al., 2011. Mission status and future prospects for improving understanding of the internal structure and thermal evolution of the Moon from the Gravity Recovery and Interior Laboratory (GRAIL) Mission, Lunar and Planetary Science Conference, 42, 1967.
1802.06116	1	1802	2018-02-16T20:47:07	Evolution of H$_2$O Production in Comet C/2012 S1 (ISON) as Inferred from Forbidden Oxygen and OH Emission	[ "astro-ph.EP" ]	We present H$_2$O production rates for comet C/2012 S1 (ISON) derived from observations of [OI] and OH emission during its inbound leg, covering a heliocentric distance range of 1.8-0.44 AU. Our production rates are in agreement with previous measurements using a variety of instruments and techniques and with data from the various observatories greatly differing in their projected fields of view. The consistent results across all data suggest the absence of an extended source of H$_2$O production, for example sublimation of icy grains in the coma, or a source with spatial extent confined to the dimensions of the smallest projected field of view (in this case $<$ 1,000 km). We find that ISON had an active area of around 10 km$^2$ for heliocentric distances R$_h$ > 1.2 AU, which then decreased to about half this value from R$_h$=1.2-0.9 AU. This was followed by a rapid increase in active area at about R$_h$=0.6 AU, corresponding to the first of three major outbursts ISON experienced inside of 1 AU. The combination of a detected outburst in the light curve and rapid increase in active area likely indicates a major nucleus fragmentation event. The 5-10 km$^2$ active area observed outside of R$_h$=0.6 AU is consistent with a 50-100% active fraction for the nucleus, larger than typically observed for cometary nuclei. Although the absolute value of the active area is somewhat dependent on the thermal model employed, the changes in active area observed are consistent among models. The conclusion of a 50-100+% active fraction is robust for realistic thermal models of the nucleus. However the possibility of a contribution of a spatially unresolved distribution of icy grains cannot be discounted.	astro-ph.EP	astro-ph	Evolution of H2O Production in Comet C/2012 S1 (ISON) as Inferred from Forbidden Oxygen and OH Emission ⋆ Adam J. McKay a, Anita L. Cochran b, Michael A. DiSanti c,d, Neil Dello Russo e, Harold Weaver e, Ronald J. Vervack Jr. e, Walter M. Harris f, Hideyo Kawakita g aNASA GSFC/USRA, 8800 Greenbelt Rd, Greenbelt, MD 20771 (U.S.A.); [email protected] bUniversity of Texas Austin/McDonald Observatory, 2512 Speedway, Stop C1402 , Austin, TX 78712, (U.S.A); [email protected] cNASA Goddard Center for Astrobiology, NASA GSFC, Mail Stop 690, Greenbelt, MD 20771 (U.S.A.); [email protected] dSolar System Exploration Division, Mail Stop 690, Greenbelt, MD 20771 (U.S.A) eJohns Hopkins University Applied Physics Laboratory, 11100 Johns Hopkins Rd., Laurel, MD, 20723 (U.S.A.); [email protected], [email protected], [email protected] f Lunar and Planetary Laboratory, University of Arizona, 1629 E University Blvd., Tucson, AZ, 85721 (U.S.A);[email protected] gKoyama Astronomical Observatory, Kyoto Sangyo University, Motoyama, Kamigamo, Kita-ku, Kyoto 603-8555, Japan; [email protected] Copyright c(cid:13) 2018 Adam J. McKay, Anita L. Cochran, Michael A. DiSanti, Neil Dello Russo, Harold Weaver, Ronald J. Vervack, Walter M. Harris, Hideyo Kawakita Preprint submitted to Icarus 20 February 2018 Number of pages: 26 Number of tables: 5 Number of figures: 5 ⋆ This paper includes data taken at The McDonald Observatory of The Univer- sity of Texas at Austin, as well as data collected at the W.M. Keck Observatory, Maunakea, HI, USA, operated as a scientific partnership among Caltech, UCLA, and NASA, and made possible by the generous financial support of the W.M. Keck Foundation. 2 Proposed Running Head: H2O Production in Comet C/2012 S1 (ISON) Please send Editorial Correspondence to: Adam J. McKay NASA GSFC/USRA 8800 Greenbelt Rd Greenbelt, MD 20771, USA. Email: [email protected] Phone: 301-614-5701 3 ABSTRACT We present H2O production rates for comet C/2012 S1 (ISON) derived from observations of [O I] and OH emission during its inbound leg, covering a helio- centric distance range of 1.8-0.44 AU. Our production rates are in agreement with previous measurements using a variety of instruments and techniques and with data from the various observatories greatly differing in their projected fields of view. The consistent results across all data suggest the absence of an extended source of H2O production, for example sublimation of icy grains in the coma, or a source with spatial extent confined to the dimensions of the smallest projected field of view (in this case < 1,000 km). We find that ISON had an active area of around 10 km2 for heliocentric distances Rh > 1.2 AU, which then decreased to about half this value from Rh=1.2-0.9 AU. This was followed by a rapid increase in active area at about Rh=0.6 AU, corresponding to the first of three major outbursts ISON experienced inside of 1 AU. The combination of a detected outburst in the light curve and rapid increase in active area likely indicates a major nucleus fragmentation event. The 5-10 km2 active area observed outside of Rh=0.6 AU is consistent with a 50-100% active fraction for the nucleus, larger than typically observed for cometary nuclei. Al- though the absolute value of the active area is somewhat dependent on the thermal model employed, the changes in active area observed are consistent among models. The conclusion of a 50-100+% active fraction is robust for re- alistic thermal models of the nucleus. However the possibility of a contribution of a spatially unresolved distribution of icy grains cannot be discounted. As our [OI]-derived H2O production rates are consistent with values derived using other methods, we conclude that the contribution of O2 photodissociation to the observed [O I] emission is at most 5-10% that of the contribution of H2O 4 for ISON. This is consistent with the expected contribution of O2 photodis- sociation if O2/H2O ∼ 4%, meaning [O I] emission can still be utilized as a reliable proxy for H2O production in comets as long as O2/H2O . 4%, similar to the abundance measured by the ROSINA instrument on Rosetta at comet 67P/Churyumov-Gerasimenko. Keywords: Comets; Comets, Coma; Comets, Composition 5 1 Introduction Comet C/2012 S1 (ISON) was a dynamically new comet from the Oort Cloud, meaning it was making its first passage through the inner Solar System (Novski et al., 2012). It passed within 3 solar radii of the Sun in late November 2013, mak- ing ISON a sungrazing comet. Many sungrazing comets have been discovered, most belonging to a dynamical family called the Kreutz group. However, ISON was unique in two respects: 1) it was dynamically new, whereas the Kreutz group comets are not, and 2) more importantly, ISON was discovered over a year before its perihelion passage, meaning observations could be planned well in advance to follow its plunge toward the Sun. This meant that ISON could be observed over a very large range of heliocentric distances, which was not possible for previous sun-grazing comets, potentially providing new insights into cometary composition and activity. Because of this unique opportunity, a world-wide observing campaign was organized, providing an unprecedented data set. Comet ISON experienced very irregular activity levels as it approached the Sun, experiencing several outbursts that may or may not have been correlated with fragmentation or disruption events (e.g. Sekanina and Kracht, 2014). Measuring the H2O production rate throughout its inbound trajectory pro- vides a window into the sublimation processes driving the outbursts. While direct observations of H2O are possible at IR wavelengths (e.g. Dello Russo et al., 2011), H2O production can also be measured indirectly through its dissociation products H, O, and OH. The main emission mechanism for the [O I]6300 A line 6 is prompt emission after photodissociation of an H2O molecule (Festou and Feldman, 1981). Therefore observations of this line have been used as a reliable proxy for H2O production in the past (e.g. Morgenthaler et al., 2001, 2007; Fink, 2009; McKay et al., 2015). Also, OH in the coma results from photodissociation of H2O, and OH is often used as a proxy for H2O at NUV (e.g. A'Hearn et al., 1995; Opitom et al., 2015), IR (e.g. Bonev et al., 2006; Dello Russo et al., 2011), and radio wavelengths (e.g. Biver et al., 2002). Lyman-α emission has also been used to derive H2O production rates, particularly with the SOHO satellite (e.g. Combi et al., 2013). We present analysis of observations of [O I] and OH emission in comet ISON as a proxy for H2O production. We obtained multiple observations spanning the time period from September-November 2013, covering a heliocentric distance range of 1.8-0.44 AU. Our observations of [O I] in particular, which have excellent sensitivity due to the abilities of optical detectors and the prompt nature of the [O I] emission mechanism, cover larger heliocentric distances than other studies, providing a characterization of ISON's activity earlier in the apparition. This paper is organized as follows: in section 2 we describe our observations and reduction and analysis procedures. Section 3 presents our results. In section 4 we discuss our results in the context of the wider observing campaign, including comparison to other measurements of the H2O production rate obtained with different methods. Section 5 presents a summary of our conclusions. 7 2 Observations and Data Analysis 2.1 Observations We obtained spectra of C/2012 S1 (ISON) using three facilities. The ARCES instrument is mounted on the Astrophysical Research Consortium 3.5-m tele- scope at Apache Point Observatory (APO) in Sunspot, New Mexico. ARCES provides a spectral resolution of R ≡ λ ∆λ = 31,500 and a spectral range of 3500-10,000 A with no interorder gaps. More specifics for this instrument are discussed elsewhere (Wang et al., 2003). In mid-October we obtained spectra using the Tull Coude Spectrograph mounted on the 2.7-m Harlan J. Smith tele- scope at McDonald Observatory. The spectral range is the same as ARCES, but the Tull Coude provides twice the spectral resolution, with R ∼ 60,000. However, unlike ARCES, the Tull Coude spectrograph has interorder gaps redward of 5700 A. To account for this, we used a spectrograph setup such that the interorder gaps did not fall on emission features of interest, which for the work presented here constitutes the [O I]6300 A line. We obtained ob- servations with the HIRES instrument on Keck I in late October. We utilized the HIRESb mode, which covers wavelengths from 3000-5700 A at R ∼ 48,000 with small gaps in coverage at the edges of the CCD detectors. Although this mode does not cover the redder wavelengths sampled by the ARCES and Tull Coude data (and therefore not the [O I]6300 A line), the additional extension blueward into the UV provides observations of the OH A-X (0-0) band at 3080 A that is not sampled in the ARCES or Tull Coude spectra. The observation dates and geometries, as well as standard stars used and ob- 8 serving conditions are described in Table 1. For all observations we centered the slit on the optocenter of the comet. All the instruments employed have rela- tively narrow and short slits, specifically 1.6′′×3.2′′(ARCES), 0.86′′×7′′(HIRES), and 1.2′′×8.2′′(Tull Coude). We used an ephemeris generated from JPL Hori- zons for non-sidereal tracking of the optocenter. For short time-scale tracking, the guiding software uses a boresight technique, which utilizes optocenter flux that falls outside the slit to keep the slit on the optocenter. For the ARCES and HIRES observations, we observed a G2V star in order to remove the un- derlying solar continuum and Fraunhofer absorption lines. For the McDonald observations, we obtained observations through a solar port which feeds sun- light directly into the spectrograph. We obtained spectra of a fast rotating (vsin(i) > 150 km s−1), O, B, or A star to account for telluric features and spectra of a flux standard to establish absolute intensities of cometary emis- sion lines. The calibration stars used for each observation date are given in Table 1. We obtained spectra of a quartz lamp for flat fielding and acquired spectra of a ThAr lamp for wavelength calibration. 2.2 Data Reduction Spectra were extracted and calibrated using IRAF scripts that perform bias subtraction, cosmic ray removal, flat fielding, and wavelength calibration. We removed telluric absorption features, removed the reflected solar continuum from the dust coma, and flux calibrated the spectra employing our stan- dard star observations. We assumed an exponential extinction law and extinc- tion coefficients for the observatory site when flux calibrating the cometary 9 spectra (e.g. Hogg et al., 2001). More details of this procedure can be found in McKay et al. (2012) and Cochran and Cochran (2002). We determined slit losses for the flux standard star observations by performing aperture photom- etry on the slit viewer images as described in McKay et al. (2014). Slit losses introduce a systematic error in the flux calibration of ∼ 10%. Slit viewer im- ages are not available for the McDonald and Keck data sets, so for these data sets we adopted a slit loss value based on the measured seeing and assumption of a Gaussian PSF. Accounting for possible variability in seeing and the ideal nature of this assumption, we estimate the uncertainties in slit losses are ∼ 20% in these cases. For UT September 23 and UT November 20, flux stan- dards taken from UT October 3 and UT November 15, respectively, were used to flux calibrate the spectra since a flux standard star was not obtained those nights. This may introduce additional uncertainty that is not accounted for in our quoted measurements. With the high airmass (> 2-3) of some of our observations and the slits em- ployed (widths of 0.9-1.6′′, lengths of 3.2-8.2′′), the effect of differential refrac- tion on fluxes obtained at different wavelengths could be significant. This could be particularly important for OH, whose primary emission band is in the near UV (3100 A) and therefore suffers from more significant differential refraction than redder spectral features. To evaluate the effect of differential refraction, we employ the computed differential refraction as a function of wavelength and airmass from Filippenko (1982), and interpolate to the relevant wavelengths and airmasses for our observations. We calculate the amount of differential re- fraction and use a Haser model profile convolved with a Gaussian point spread function to estimate the slit loss due to the differential refraction. For all our 10 observations we estimate the effect of differential refraction is less than 10% and is insignificant compared to the uncertainties in flux calibration noted above (10-20%). The uncertainties quoted in this paper are dominated by the uncertainty in flux calibration and not the signal-to-noise ratio of the data nor differential refraction effects. 2.3 Analysis of [O I] Emission The [O I]6300 A line is among the emission features present in the ARCES and Tull Coude bandpasses and can be employed to derive the H2O production rate. The [O I]6300 A line is also present as a telluric emission feature, and so a combination of high spectral resolution and adequate geocentric velocity (and therefore Doppler shift) is needed to resolve the cometary line from the tel- luric feature. For all observation dates except November 20 the cometary and telluric lines are fully separated. On November 20 the comet was so bright and active that the cometary line is expected to dwarf the telluric feature, there- fore any influence of the telluric feature on the measured flux is negligible. An example spectrum from November 15 is shown in the top panel of Fig. 1, with a more detailed view of the [O I]6300 A and [O I]6364 A lines shown in Fig. 2. To measure the line flux we fit a Gaussian function to the line profile. More details for our line fitting procedure can be found in McKay et al. (2012). For all of our observations, to check our calibration we measured the flux ratio of the [O I]6300 A line to the [O I]6364 A line (hereafter the red line ratio). As these lines both originate from the 1D state, the red line ratio is determined 11 directly by the branching ratio into the different transitions and is not depen- dent on coma chemistry or physics. This value is well established at a value of 3.0 (Cochran and Cochran, 2001; Sharpee and Slanger, 2006). However, our observations in September and October yield significantly lower values for the red line ratio, in the range 1.5-2.5. Examination of the spectra revealed that at the very large Doppler shift that ISON had at the time of these observations (-50 km s−1), the [O I]6300 A line falls precisely on an O2 telluric absorption feature, as shown in Fig. 3. Although we remove telluric signatures during our reduction process, the removal may not be perfect. On nights for which we had multiple observations, [O I]6300 A line intensities have a large amount of scatter, evidence that our [O I]6300 A line intensities were affected by im- perfect removal of this telluric absorption feature from spectrum to spectrum. There are no telluric or solar absorption features near the [O I]6364 A line, suggesting this line should not be affected by the presence of these features (see Fig. 3). The [O I]6364 A line intensities exhibit significantly less scatter, consistent with the stochastic uncertainties and supporting this conclusion. Therefore for the September and October data we use the [O I]6364 A line flux and multiply it by the branching ratio of 3.0 to obtain the line flux that is used to calculate H2O production rates. This was not required for our Novem- ber observations, as the Doppler shift had decreased (in absolute value) so that the [O I]6300 A line was no longer located on an O2 telluric absorption (see Fig. 3), and spectra obtained on these dates have red line ratios consistent with the known branching ratio of 3.0. 12 2.4 Spectral Fitting Model for OH Unlike [O I] emission, the OH emission exhibits ro-vibrational structure su- perimposed on the electronic transition, which results in an intricate band structure for the emission features (see Fig. 1). To fit the OH emission we employ the same modeling technique used by McKay et al. (2016), which we briefly describe here. The program references a line list for the species of in- terest (in this case the OH A-X (0-0) band). Once this line list is compiled, we fit each spectral feature in the list to a Gaussian profile. We add all the line profile fits together to create an empirical fit to the spectrum and then integrate over this model fit to obtain a flux for the OH band. While in general the program handles multiple species simultaneously, the OH A-X (0-0) band does not suffer from contamination of spectral features due to other species, therefore their inclusion is not necessary. More details on the general program can be found in McKay et al. (2014), with the specific application to OH de- tailed in McKay et al. (2016). 2.5 Conversion of Observed Flux to Production Rate We derived H2O production rates based on two different spectral features: [O I]6300, 6364 A emission and the OH A-X (0-0) band at 3080 A. For [O I], we employ the same Haser model for the [O I] emission used in McKay et al. (2012) to infer H2O production rates from our [O I] observations. For OH, we employ the same Haser model as in McKay et al. (2016), which has modifica- tions that emulate the vectorial model (see McKay et al. (2014) for more de- 13 tails), and adopt fluorescence efficiencies from Schleicher and A'Hearn (1988) that account for the Swings effect (the dependence of the fluorescence efficiency on the heliocentric velocity of the comet). The OH production rates are con- verted to H2O production rates using the relation from Cochran and Schleicher (1993): QH2O = 1.361R−0.5 h QOH (1) where Rh is heliocentric distance in AU. The relation depends on heliocentric distance because the expansion velocity of H2O in the coma is dependent on heliocentric distance while the ejection velocity of OH into the coma is not. Parameters for the Haser models are given in Table 2. 3 Results The resulting average H2O production rates for each observation date are given in Table 3 and plotted in Fig. 4, along with other measured H2O production rates for ISON from the literature. In Table 3 we also present the surface area needed to provide the observed H2O production (i.e. the active area). We em- ploy the sublimation model of Cowan and A'Hearn (1979) with a visual albedo of 0.05 and thermal emissivity of unity and the slow-rotator approximation. The slow-rotator approximation is appropriate for objects with slow rota- tion rates or low thermal inertia, and is typically the model used to describe cometary nuclei (e.g. Bodewits et al., 2014). Table 3 also displays the active fraction of the nucleus based on our active area calculations and assuming a spherical nucleus with a radius of 0.75 ± 0.15 km as derived by Lamy et al. 14 (2014) from HST observations (although they quote a more precise value of 0.68 ± 0.02 km, Lamy et al. (2014) also note that this precise number is model dependent, so for this work we will employ their more conservative value of 0.75 ± 0.15 km). Size estimates for Oort Cloud comets like ISON are very rare, so the size estimate for ISON provides a unique opportunity to constrain the active fraction of an Oort Cloud comet. As both active area and active fraction are model dependent, Table 4 shows active areas and active fractions derived for isothermal, subsolar, and fast-rotator models in addition to the slow-rotator model to examine the effect of model assumptions on our results. Table 5 is the same as Table 4, except for a visual albedo of 0.5 instead of 0.05. The active area results are also plotted in Fig. 5. The derived active areas are as high as 100% and in some cases even higher. Implications for this will be discussed in the next section. 4 Discussion 4.1 Evolution of H2O Production Our H2O production rates show a very slow increase from 1.8 to 0.9 AU, with a dramatic rise in H2O production at around 0.6 AU, in agreement with previous measurements as shown in Fig. 4. Our observation at a heliocentric distance of 0.6 AU was only ∼ 12 hours after the first of several outbursts that were observed at heliocentric distances less than 0.6 AU (see Sekanina and Kracht, 2014, for an analysis of comet ISON's lightcurve and outbursts). The differ- ence in H2O production before and after the outburst is about a factor of 15 15, which matches well the observed increase in brightness during the out- burst (Sekanina and Kracht, 2014), suggesting that the outburst was indeed correlated to a massive increase in H2O production. We will discuss what led to this strong increase in gas production in the following sections. The H2O production rate seems to have leveled out between 0.6 and 0.44 AU, how- ever we note that clouds may have influenced our derived fluxes on this date. Although our derived production rate is consistent with other measurements around this time, the uncertainty introduced by clouds means the uncertainty in this measurement could be a factor of two or maybe even higher, mak- ing it difficult to derive any strong conclusions on the behavior of the H2O production between 0.6 and 0.44 AU. 4.2 Extended Sources of H2O Past observations of comets such as C/2009 P1 (Garradd) have shown trends in derived H2O production rate with the projected field of view at the comet, with larger fields of view showing larger production rates. This has been cited as evidence for an icy grain halo with an extent of 104-105 km that serves as an extended source of H2O production (Combi et al., 2013; Bodewits et al., 2014; McKay et al., 2015). The Lyman-α observations of Combi et al. (2014a) have very large fields of view (∼ 106-107 km), while slit-based spectroscopy like the observations presented here and IR observations such as Dello Russo et al. (2016) and DiSanti et al. (2016) have much smaller projected fields (a few × 103 km), and narrowband OH imaging (Knight and Schleicher, 2015; Opitom et al., 2014) has intermediate fields of view on the order of tens of thousands of km. However, throughout the apparition for ISON there appears to be no correla- 16 tion between projected field of view and derived H2O production, with the pos- sible exception of our [O I] observations at Rh=1.8 AU and 1.6 AU and nearly concurrent measurements of OH from Knight and Schleicher (2015), which employ a field of view on the order of 5×104 km (see Fig. 4). Knight and Schleicher (2015) argue from analysis of the dust spatial profiles that before late October (i.e. at the time of our observations at Rh=1.8 AU and 1.6 AU) the coma did contain icy grains, so it is possible that there was an extended source of H2O production at larger heliocentric distance that dissipated with the disappear- ance of the icy grains in late October. Combi et al. (2014a) reached a similar conclusion about the presence of an extended source of H2O production in comet ISON from late October onward. It is also possible that the extended source did not disappear, but rather became much smaller in spatial extent and was no longer spatially resolved by observations. With derived H2O pro- duction rates being consistent across platforms, any extended source present at Rh < 1.3 AU must have had a spatial extent smaller than the projected field of view of our observations (< 1000 km). 4.3 Analysis of the Active Area: Evidence for Fragmentation 4.3.1 Evolution of the Active Area Our derived active areas (Table 3) show a relatively constant value of ∼10 km2 outside of Rh=1.2 AU, which then decreased to about half this value over the Rh=0.9-1.2 AU range, in agreement with calculations by Combi et al. (2014a), although Combi et al. (2014a) have a slightly later onset (∼1 AU) for their tentative detection of a decrease in active area. For the nucleus radius of 0.75 ± 0.15 km determined by Lamy et al. (2014), the active areas derived for Rh 17 > 0.9 AU are consistent with a very high active fraction for the nucleus of 50% or even a completely active surface, though the error bars are quite large due to uncertainty in the actual size of the nucleus coupled with uncertainties in the H2O production rate. In addition, our calculations of the active fraction for ISON assume a spherical nucleus, which is not necessarily a valid assumption. Although a precise measurement of the active fraction for ISON is difficult, both our results and those of Combi et al. (2014a) suggest that it is 50-100% before the first outburst at Rh ∼ 0.6 AU. However, at Rh=0.6 AU, both we and Combi et al. (2014a) find an active area of ∼ 30 km2, much larger than the total surface area of a spherical nucleus with radius ∼ 0.75 km. This implies that at this time there was a large fragmentation event, or at least shedding of a large amount of ice-rich material, that resulted in more surface area being exposed directly to sunlight. There is also the possibility that the active area decreased to half this value by 0.44 AU, but as stated in Section 4.1 the water production rate derived on this date may be affected by clouds, making it difficult to arrive at any definitive conclusions about the active area at this time. Comparison to other measurements of the water production rates in Fig. 2 and the active areas they imply suggest that it seems unlikely that the active area was higher at 0.44 AU than at 0.6 AU. 4.3.2 Alternate Models To test the robustness of this conclusion, we examine not only active areas from the slow-rotator model, but also isothermal, subsolar, and fast-rotator models. The slow-rotator model is valid for slow rotation periods and/or low thermal inertia of the surface material. These conditions mean that every point on the cometary surface is in thermal equilibrium with the solar radiation in- 18 cident upon it. Comet rotation periods vary anywhere from hours to days, with a possible value of 10.4 hours determined for ISON (Lamy et al., 2014). This does not fulfill the slow-rotator approximation. However, the thermal inertias for comets where such measurements are available are quite low (e.g. Groussin et al., 2013; Choukroun et al., 2015), hence the tendency in the lit- erature is to assume cometary nuclei follow the slow-rotator approximation. Therefore the slow-rotator is our preferred model. The isothermal model as- sumes the whole nucleus surface is the same temperature, while the subsolar model assumes the whole nucleus surface has the temperature of the subsolar point. Although these models are idealizations that likely do not occur in re- ality, they provide bounds on the sublimation rate coming off the surface. The fast-rotator model assumes that lines of latitude on the cometary nucleus are isotherms, which occurs for fast rotation periods and/or high thermal inertias of the surface material. The active areas and active fractions derived for the different models with a visual albedo of 0.05 are shown in Table 4 and Fig. 5. At the heliocentric dis- tances of our observations (< 2 AU, well within the H2O sublimation line, such that H2O can be assumed to be fully activated), the fast-rotator and isother- mal approximations give very similar active areas, which are slightly higher than those derived from the slow-rotator model. This discrepancy is maximum at the largest heliocentric distances, and decreases as the comet moves towards the Sun. The subsolar point model produces significantly smaller values for the active area than the other models, and is the only model with active frac- tions less than 50%. Because all of these models suggest ISON's active fraction was much higher 19 than the typical 5% seen in other comets (e.g. A'Hearn et al., 1995), this ei- ther suggests that a significant fraction of the comet's surface was active, or there was an extended source of H2O in the coma such as sublimation from icy grains. We discussed the possibility of an extended source in the previous section. For the following we will explore the possibility of a surface that was 20-100% active (encompassing values from all the thermal models). This high an active fraction could suggest that a large fraction of the cometary surface was exposed H2O ice, meaning assuming a visual albedo of 0.05 may not be a reasonable assumption. Therefore we explored thermal models for the nucleus where both the visual albedo and thermal albedo of the nucleus have a much higher value of 0.50. The results of these models are presented in Table 5 and Fig. 5. Assuming a higher albedo shifts the derived active areas upward by about a factor of two compared to the lower albedo models. This only strengthens the conclusion that ISON's active fraction was quite large and perhaps an icy grain source of H2O is needed to account for the comet being hyperactive, though this source would have to have a small spatial extent (see section 4.2). We will now examine which of our thermal models is most likely to describe ISON's activity. The subsolar model is the only model with active fractions less than 50%, meaning this is the only model that does not require a nearly uniformly active surface or an extended source. The subsolar model is only valid if all the activity is dominated by the subsolar point, which is gener- ally not observed in comets. A special case that would fit this description is where a polar source region is the dominant source of activity and the rotation pole is pointed directly at the Sun (i.e. the rotation pole is at the subsolar 20 point). Early in the apparition (Spring 2013, Rh=4.0 AU) there was evidence for a polar source region and a rotation pole pointed at the Sun based on HST (Li et al., 2013) and ground-based observations (Knight and Schleicher, 2015) of a dust jet pointed in the solar direction, the orientation of which did not change over time. However, this feature dissipated as ISON moved towards the Sun and the CN jet morphology during October-November did not match well with a polar jet (Knight and Schleicher, 2015), making it likely that during the time of our observations the activity was not driven by a po- lar source region located at the subsolar point. Therefore we do not find the subsolar model a likely description of ISON's activity. We cannot rule out the plausibility of the other models with current results. 4.3.3 Implications for ISON's Activity The arguments of the previous section mean that ISON's surface was at least 50% active or there was an extended source of H2O production in the coma, most likely due to the presence of icy grains. The high albedo (A=0.5) cases all favor an active area of more than 100%, implying an extended source, in which case there is a degeneracy between the active area of the extended source and the active area of the nucleus. The low albedo cases (A=0.05) are consistent with both a highly active nucleus or an extended source. If an extended source is present it must have a small spatial extent (less than ∼1000 km, the projected field of view of our slit spectroscopy) as argued in section 4.2. This is plausible, considering that 103P/Hartley 2 was shown to have a hyperactive nucleus and that this extra active area came from an ex- tended source of icy grains in the coma (A'Hearn et al., 2011; Kelley et al., 21 2013). However, H2O production rates derived from remote sensing observa- tions agree over drastically different fields of view, from ∼500 km for slit spec- troscopy (Dello Russo et al., 2011; Mumma et al., 2011; McKay et al., 2014) to 104-105 km for imaging of OH (Knight and Schleicher, 2013) and 106 km for Lyman-α (Combi et al., 2011), suggesting that the spatial extent of the icy grain source was not much larger than the smallest fields of view used (in this case a few hundered km). This is consistent with the expected lifetimes of micron-sized icy grains at the heliocentric distance of 103P/Hartley 2 at the time of observation (Rh ∼1.1 AU) of only 100-1000 seconds (Beer et al., 2006), which is comparable to the crossing time for icy grains moving at ∼1 km s−1. Therefore for ISON to exhibit a similar phenomenon at similar and smaller heliocentric distances to 103P/Hartley 2 is certainly plausible, although un- certainties in the active area and nucleus size preclude this from being demon- strated conclusively. Using observations of the spatial distribution of H2O and rotational temperature in the inner coma (nucleocentric distance < 1000 km), Bonev et al. (2014) argued that ISON was releasing icy grains at he- liocentric distances less than 0.6 AU, but as they obtained no observations at larger heliocentric distances (in particular before the first outburst in mid- November), their observations do not place a constraint on an icy grain source before the outburst. All models for ISON on November 15 display active fractions greater than unity (Fig. 5, Tables 4 and 5), implying a large increase in the active area of the comet/extended coma source. This could be due to either a catastrophic fragmentation of the nucleus or a large outburst ejecting icy grains into the coma. As discussed above, Bonev et al. (2014) suggested that after the first outburst at Rh ∼ 0.6 AU sublimation from icy grains was a significant source of 22 water production. However, Steckloff et al. (2015) argued that the coma mor- phology during and after the outburst in mid-November is inconsistent with a simple explosive event such as amorphous-crystalline water ice transition or trapped hypervolatiles triggering the outbursts and ejecting icy grains into the coma. They argued instead that ISON split into a swarm of fragments with a characteristic size of 100 m. As they are all the same size, drifting between fragments would not show up in Earth-based observations of the coma mor- phology until days after the event, consistent with observations. However, it is likely that both the fragmentation event and release of small (micron-sized) icy grains contributed to the increased water production. Although the absolute values of the active area and active fraction are model dependent, the evolution of the active area over the course of the appari- tion is much less sensitive to the specific thermal model employed, as shown in Fig. 5. All models support a relatively constant active area during Octo- ber (Rh > 1.2 AU), a decrease in late October and early November, and a massive increase in active area in mid-November. This is very similar to the results presented in Combi et al. (2014a). The massive increase in active area in mid-November was likely associated with the first of several fragmentation events (Steckloff et al., 2015), which increased the sublimating surface area in the form of smaller fragments and may also have released micron-sized icy grains into the coma, further increasing the effective active area. It is also possible in mid-November that both thermal emission from sub- micron grains in the coma and reflected solar radiation from the dust coma could provide significant heating to the surface, meaning our simple thermo- dynamical models for the nucleus temperature and sublimation would not be accurate. If this is the case, this would mean less active area is required to 23 explain the observed H2O production rates. However, quantifying this effect requires detailed calculations that are beyond the scope of this work. 4.4 Influence of O2 Recently, the Rosetta mission discovered the presence of O2 in the coma of 67P/Churyumov-Gerasimenko at an abundance of 4% relative to H2O (Bieler et al., 2015). This discovery was unexpected and means that O2 photodissociation could be an unaccounted source of O I in cometary comae. Specifically of rele- vance to this work is that O2 has a high branching ratio for releasing O I in the 1D state upon photodissociation (0.85, Huebner et al., 1992), which will then radiatively decay and release [O I]6300 A line emission. This could potentially influence attempts to use [O I]6300 A line emission as a proxy for H2O pro- duction. To test the potential influence of O2 on [O I]6300 A line intensities we modified our Haser model for [O I] to have O2 be the parent of O I instead of H2O. For this purpose we employed a branching ratio of 0.854 for photodisso- ciation of O2 → O(3P) + O(1D) and a photodestruction timescale of 2.08×105 s at 1 AU (Huebner et al., 1992). We find that at the abundance of 4% relative to H2O observed by Rosetta at 67P by the ROSINA instrument (Bieler et al., 2015), O2 can only account for 5-10% of the observed [O I]6300 line flux. As our uncertainties are 10-20%, O2 at an abundance of 4% relative to H2O does not significantly influence our derived H2O production rates. Additionally, the consistency between our production rates derived from [O I] and OH, as well as agreement with other values already in the literature from methods other than [O I], support the conclusion that any O2 present in the coma of ISON did not influence our derived H2O production rates. Although this suggests 24 [O I] studies cannot be used to infer O2 abundances from the ground (which is very challenging to do directly due to severe telluric absorption), it does show that [O I]6300 A line emission can still be used as a proxy for H2O in comets for O2/H2O ratios similar to that measured by ROSINA. This con- clusion is valid for the small FOV employed for our observations, with the potential contribution of O2 becoming more important for larger FOV. Using Rosetta Alice UV stellar occultation data Keeney et al. (2017) found higher O2/H2O ratios than those measured by ROSINA (some as high as 50%). For O2/H2O of 50% we expect that O2 photodissociation would be a significant contributor to the [O I] emission, being responsible for approximately half of the emission. It is possible that such a high O2 abundance could artificially raise our derived H2O production rates so that they agree with observations such as Combi et al. (2014b) with larger FOV's, and could hide the presence of an extended source. However, the consistency between our OH and [O I] derived H2O production rates (which have similar FOV) argues against this scenario and therefore against such a high O2 abundance for comet ISON. 5 Conclusions In this work we presented H2O production rates for comet C/2012 S1 (ISON) derived from observations of [O I] and OH emission during its inbound leg, cov- ering a heliocentric distance range of 1.8-0.44 AU. Our production rates are in agreement with previous measurements using a variety of different techniques. As our [O I]-derived H2O production rates are consistent with other meth- ods, we conclude that the contribution of O2 photodissociation to the [O I] emission is negligible compared to the contribution of H2O (for abundances 25 observed by the ROSINA instrument aboard Rosetta), meaning [O I] emission can still be utilized as a reliable proxy for H2O production in comets (though for the much higher abundances observed by Rosetta Alice the presence of O2 could introduce systematic error). The lack of any apparent dependence of derived production rate with instrument field of view suggests that any ex- tended source of H2O production must have had a spatial extent smaller than the smallest projected field of view observation (in this case < 1,000 km). We find that ISON had an evolving active area throughout the apparition, with a rapid increase at about Rh=0.6 AU corresponding to the first of three ma- jor outbursts ISON experienced inside of 1 AU, consistent with other studies. This rapid increase in active area likely indicates a major fragmentation event of the nucleus. The 5-10 km2 active area observed outside of Rh=0.6 AU is consistent with a 50-100% active fraction for the nucleus, much higher than typically observed for cometary nuclei. Although the absolute value of the ac- tive area is somewhat dependent on the thermal model employed, the changes in observed active area are consistent among models, and the conclusion of a 50-100+% active fraction is robust for realistic thermal models of the nucleus. However, the possibility of a spatially unresolved icy grain source cannot be discounted. Acknowledgements We are grateful to Lori Feaga and an anonymous reviewer for helpful comments that improved the quality of this manuscript. We thank John Barentine, Jurek Krzesinski, Chris Churchill, Pey Lian Lim, Paul Strycker, and Doug Hoffman for developing and optimizing the ARCES IRAF reduction script used to 26 reduce these data. We thank the APO and Keck observing staffs for their invaluable help in conducting the observations. We are extremely grateful to Dr. Zlatan Tsvetanov for yielding some of his observing time so we could ob- tain the UT November 20 observations presented in this paper. We thank Dr. Cyrielle Opitom for sharing some of her unpublished water production rates for inclusion in Fig. 4. We would also like to acknowledge the JPL Horizons System, which was used to generate ephemerides for nonsidereal tracking of the comets during the observations, and the SIMBAD database, which was used for selection of reference stars. This work was supported by the NASA Postdoctoral Program, administered by USRA, as well as the NASA Plane- tary Atmospheres Program, the NASA Planetary Astronomy Program, and the NASA Emerging Worlds Program. The authors wish to recognize and ac- knowledge the very significant cultural role and reverence that the summit of Maunakea has always had within the indigenous Hawaiian community. We are most fortunate to have the opportunity to conduct observations from this mountain. References M. F. A'Hearn, R. L. Millis, D. G. Schleicher, D. J. Osip, and P. V. Birch. The ensemble properties of comets: Results from narrowband photometry of 85 comets, 1976-1992. Icarus, 118:223–270, December 1995. doi: 10.1006/icar. 1995.1190. M. F. A'Hearn, M. J. S. Belton, W. A. Delamere, L. M. Feaga, D. Hampton, J. Kissel, K. P. Klaasen, L. A. McFadden, K. J. Meech, H. J. Melosh, P. H. Schultz, J. M. Sunshine, P. C. Thomas, J. Veverka, D. D. Wellnitz, D. K. Yeomans, S. Besse, D. Bodewits, T. J. Bowling, B. T. Carcich, S. M. Collins, 27 T. L. Farnham, O. Groussin, B. Hermalyn, M. S. Kelley, M. S. Kelley, J.-Y. Li, D. J. Lindler, C. M. Lisse, S. A. McLaughlin, F. Merlin, S. Protopapa, J. E. Richardson, and J. L. Williams. EPOXI at Comet Hartley 2. Science, 332:1396–, June 2011. doi: 10.1126/science.1204054. E. H. Beer, M. Podolak, and D. Prialnik. The contribution of icy grains to the activity of comets. I. Grain lifetime and distribution. Icarus, 180:473–486, February 2006. doi: 10.1016/j.icarus.2005.10.018. A. Bhardwaj and S. Raghuram. A Coupled Chemistry-emission Model for Atomic Oxygen Green and Red-doublet Emissions in the Comet C/1996 B2 Hyakutake. Astrophysical Journal, 748:13, March 2012. doi: 10.1088/ 0004-637X/748/1/13. A. Bieler, K. Altwegg, H. Balsiger, A. Bar-Nun, J.-J. Berthelier, P. Bochsler, C. Briois, U. Calmonte, M. Combi, J. de Keyser, E. F. van Dishoeck, B. Fi- ethe, S. A. Fuselier, S. Gasc, T. I. Gombosi, K. C. Hansen, M. Hassig, A. Jackel, E. Kopp, A. Korth, L. Le Roy, U. Mall, R. Maggiolo, B. Marty, O. Mousis, T. Owen, H. R`eme, M. Rubin, T. S´emon, C.-Y. Tzou, J. H. Waite, C. Walsh, and P. Wurz. Abundant molecular oxygen in the coma of comet 67P/Churyumov-Gerasimenko. Nature, 526:678–681, October 2015. doi: 10.1038/nature15707. N. Biver, D. Bockel´ee-Morvan, P. Colom, J. Crovisier, F. Henry, E. Lel- louch, A. Winnberg, L.E.B. Johansson, M. Gunnarsson, H. Rickman, F. Rantakyro, Davies. J.K., W.R.F. Dent, G. Paubert, R. Moreno, J. Wink, D. Despois, D. Benford, M. Gardner, D.C. Lis, D. Mehringer, T.G. Phillips, and H. Rauer. The 1995-2002 long-term monitoring of comet C/1995 O1 (Hale-Bopp) at radio wavelength. Earth, Moon, and Planets, 90(1):5–14, 2002. D. Bodewits, T. L. Farnham, M. F. A'Hearn, L. M. Feaga, A. McKay, D. G. 28 Schleicher, and J. M. Sunshine. The Evolving Activity of the Dynamically Young Comet C/2009 P1 (Garradd). Astrophysical Journal, 786:48, May 2014. doi: 10.1088/0004-637X/786/1/48. B. P. Bonev, M. J. Mumma, M. A. DiSanti, N. Dello Russo, K. Magee-Sauer, R. S. Ellis, and D. P. Stark. A Comprehensive Study of Infrared OH Prompt Emission in Two Comets. I. Observations and Effective g-Factors. Astrophysical Journal, 653:774–787, December 2006. doi: 10.1086/508452. B. P. Bonev, M. A. DiSanti, G. L. Villanueva, E. L. Gibb, L. Paganini, and M. J. Mumma. The Inner Coma of Comet C/2012 S1 (ISON) at 0.53 AU and 0.35 AU from the Sun. Astrophysical Journal Letters, 796:L6, November 2014. doi: 10.1088/2041-8205/796/1/L6. M. Choukroun, S. Keihm, F. P. Schloerb, S. Gulkis, E. Lellouch, C. Leyrat, P. von Allmen, N. Biver, D. Bockel´ee-Morvan, J. Crovisier, P. Encre- naz, P. Hartogh, M. Hofstadter, W.-H. Ip, C. Jarchow, M. Janssen, S. Lee, L. Rezac, G. Beaudin, B. Gaskell, L. Jorda, H. U. Keller, and H. Sierks. Dark side of comet 67P/Churyumov-Gerasimenko in Aug.-Oct. 2014. MIRO/Rosetta continuum observations of polar night in the south- ern regions. Astronomy and Astrophysics, 583:A28, November 2015. doi: 10.1051/0004-6361/201526181. A. L. Cochran and D. G. Schleicher. Observational Constraints on the Lifetime of Cometary H 2O. Icarus, 105:235–253, September 1993. doi: 10.1006/icar. 1993.1121. A.L. Cochran and W.D. Cochran. Observations of O (1S) and O (1D) in spectra of C/1999 S4 (LINEAR). Icarus, 154(2):381–390, 2001. A.L. Cochran and W.D. Cochran. A high spectral resolution atlas of comet 122P/de Vico. Icarus, 157(2):297–308, 2002. M. R. Combi, J.-L. Bertaux, E. Qu´emerais, S. Ferron, and J. T. T. Makinen. 29 Water Production by Comet 103P/Hartley 2 Observed with the SWAN Instrument on the SOHO Spacecraft. Astrophysical Journal Letters, 734: L6, June 2011. doi: 10.1088/2041-8205/734/1/L6. M. R. Combi, J. T. T. Makinen, J.-L. Bertaux, E. Qu´emerais, S. Ferron, and N. Fougere. Water production rate of Comet C/2009 P1 (Garradd) throughout the 2011-2012 apparition: Evidence for an icy grain halo. Icarus, 225:740–748, July 2013. doi: 10.1016/j.icarus.2013.04.030. M. R. Combi, N. Fougere, J. T. T. Makinen, J.-L. Bertaux, E. Qu´emerais, and S. Ferron. Unusual Water Production Activity of Comet C/2012 S1 (ISON): Outbursts and Continuous Fragmentation. Astrophysical Journal Letters, 788:L7, June 2014a. doi: 10.1088/2041-8205/788/1/L7. M. R. Combi, J. T. Makinen, J. L. Bertaux, E. Qu´emerais, and S. Ferron. The Water Production Rate of Recent Comets (2013-2014) by SOHO/SWAN: 2P/Encke (2013), C/2013 R1 (Lovejoy), and C/2013 A1 (Siding Spring). In AAS/Division for Planetary Sciences Meeting Abstracts, volume 46 of AAS/Division for Planetary Sciences Meeting Abstracts, page 110.09, November 2014b. J. J. Cowan and M. F. A'Hearn. Vaporization of comet nuclei - Light curves and life times. Moon and Planets, 21:155–171, October 1979. doi: 10.1007/ BF00897085. N. Dello Russo, R. J. Vervack, Jr., C. M. Lisse, H. A. Weaver, H. Kawakita, H. Kobayashi, A. L. Cochran, W. M. Harris, A. J. McKay, N. Biver, D. Bockel´ee-Morvan, and J. Crovisier. The Volatile Composition and Activity of Comet 103P/Hartley 2 During the EPOXI Closest Ap- proach. Astrophysical Journal Letters, 734:L8+, June 2011. doi: 10.1088/ 2041-8205/734/1/L8. N. Dello Russo, R. J. Vervack, H. Kawakita, A. Cochran, A. J. McKay, W. M. 30 Harris, H. A. Weaver, C. M. Lisse, M. A. DiSanti, H. Kobayashi, N. Biver, D. Bockel´ee-Morvan, J. Crovisier, C. Opitom, and E. Jehin. The compo- sitional evolution of C/2012 S1 (ISON) from ground-based high-resolution infrared spectroscopy as part of a worldwide observing campaign. Icarus, 266:152–172, March 2016. doi: 10.1016/j.icarus.2015.11.030. M. A. DiSanti, B. P. Bonev, E. L. Gibb, L. Paganini, G. L. Villanueva, M. J. Mumma, J. V. Keane, G. A. Blake, N. Dello Russo, K. J. Meech, R. J. Vervack, Jr., and A. J. McKay. En Route to Destruction: The Evolution in Composition of Ices in Comet D/2012 S1 (ISON) between 1.2 and 0.34 AU from the Sun as Revealed at Infrared Wavelengths. Astrophysical Journal, 820:34, March 2016. doi: 10.3847/0004-637X/820/1/34. M.C. Festou and P.D. Feldman. The forbidden oxygen lines in comets. Astronomy and Astrophysics, 103(1):154–159, 1981. A. V. Filippenko. The importance of atmospheric differential refraction in spectrophotometry. Publications of the Astronomical Society of the Pacific, 94:715–721, August 1982. doi: 10.1086/131052. U. Fink. A taxonomic survey of comet composition 1985-2004 using CCD spectroscopy. Icarus, 201:311–334, May 2009. doi: 10.1016/j.icarus.2008.12. 044. O. Groussin, J. M. Sunshine, L. M. Feaga, L. Jorda, P. C. Thomas, J.-Y. Li, M. F. A'Hearn, M. J. S. Belton, S. Besse, B. Carcich, T. L. Farn- ham, D. Hampton, K. Klaasen, C. Lisse, F. Merlin, and S. Protopapa. The temperature, thermal inertia, roughness and color of the nuclei of Comets 103P/Hartley 2 and 9P/Tempel 1. Icarus, 222:580–594, February 2013. doi: 10.1016/j.icarus.2012.10.003. D. W. Hogg, D. P. Finkbeiner, D. J. Schlegel, and J. E. Gunn. A Photometric- ity and Extinction Monitor at the Apache Point Observatory. Astronomical 31 Journal, 122:2129–2138, October 2001. doi: 10.1086/323103. W.F. Huebner, J.J. Keady, and S.P. Lyon. Solar photo rates for planetary atmospheres and atmospheric pollutants. Astrophysics and Space Science, 195(1):1–289, 291–294, 1992. B. A. Keeney, S. A. Stern, M. F. A'Hearn, J.-L. Bertaux, L. M. Feaga, P. D. Feldman, R. A. Medina, J. W. Parker, J. P. Pineau, E. Schindhelm, A. J. Steffl, M. Versteeg, and H. A. Weaver. H2O and O2 absorption in the coma of comet 67P/Churyumov-Gerasimenko measured by the Alice far-ultraviolet spectrograph on Rosetta. Monthly Notices of the Royal Astronomical Society, 469:S158–S177, July 2017. doi: 10.1093/mnras/stx1426. M. S. Kelley, D. J. Lindler, D. Bodewits, M. F. A'Hearn, C. M. Lisse, L. Kolokolova, J. Kissel, and B. Hermalyn. A distribution of large par- ticles in the coma of Comet 103P/Hartley 2. Icarus, 222:634–652, February 2013. doi: 10.1016/j.icarus.2012.09.037. M. M. Knight and D. G. Schleicher. The highly unusual outgassing of Comet 103P/Hartley 2 from narrowband photometry and imaging of the coma. Icarus, 222:691–706, February 2013. doi: 10.1016/j.icarus.2012.06.004. M. M. Knight and D. G. Schleicher. Observations of Comet ISON (C/2012 S1) from Lowell Observatory. Astronomical Journal, 149:19, January 2015. doi: 10.1088/0004-6256/149/1/19. P. L. Lamy, I. Toth, and H. A. Weaver. Hubble Space Telescope Observations of the Nucleus of Comet C/2012 S1 (ISON). Astrophysical Journal Letters, 794:L9, October 2014. doi: 10.1088/2041-8205/794/1/L9. J.-Y. Li, M. S. P. Kelley, M. M. Knight, T. L. Farnham, H. A. Weaver, M. F. A'Hearn, M. J. Mutchler, L. Kolokolova, P. Lamy, and I. Toth. Characterizing the Dust Coma of Comet C/2012 S1 (ISON) at 4.15 AU from the Sun. Astrophysical Journal Letters, 779:L3, December 2013. doi: 32 10.1088/2041-8205/779/1/L3. A. J. McKay, N. J. Chanover, J. P. Morgenthaler, A. L. Cochran, W. M. Harris, and N. D. Russo. Forbidden oxygen lines in Comets C/2006 W3 Christensen and C/2007 Q3 Siding Spring at large heliocentric distance: Implications for the sublimation of volatile ices. Icarus, 220:277–285, July 2012. doi: 10.1016/j.icarus.2012.04.030. A. J. McKay, N. J. Chanover, M. A. DiSanti, J. P. Morgenthaler, A. L. Cochran, W. M. Harris, and N. D. Russo. Rotational variation of daugh- ter species production rates in Comet 103P/Hartley: Implications for the progeny of daughter species and the degree of chemical heterogeneity. Icarus, 231:193–205, March 2014. doi: 10.1016/j.icarus.2013.11.029. A. J. McKay, A. L. Cochran, M. A. DiSanti, G. Villanueva, N. D. Russo, R. J. Vervack, J. P. Morgenthaler, W. M. Harris, and N. J. Chanover. Evolution of H2O, CO, and CO2 production in Comet C/2009 P1 Garradd during the 2011-2012 apparition. Icarus, 250:504–515, April 2015. doi: 10.1016/j. icarus.2014.12.023. A. J. McKay, M. S. P. Kelley, A. L. Cochran, D. Bodewits, M. A. DiSanti, N. D. Russo, and C. M. Lisse. The CO2 abundance in Comets C/2012 K1 (PanSTARRS), C/2012 K5 (LINEAR), and 290P/Jager as measured with Spitzer. Icarus, 266:249–260, March 2016. doi: 10.1016/j.icarus.2015.11.004. J.P. Morgenthaler, W.M. Harris, F. Scherb, C. Anderson, R.J. Oliversen, N.E. Doane, M.R. Combi, M.L. Marconi, and W.H. Smyth. Large aperture [O I] 6300 A photometry of comet Hale-Bopp: Implications for the photochem- istry of OH. Astrophysical Journal, 563(1):451–461, 2001. J.P. Morgenthaler, W.M. Harris, and M.R. Combi. Large Aperture O I 6300 A Observations of Comet Hyakutake: Implications for the Photochemistry of OH and O I Production in Comet Hale-Bopp. Astrophysical Journal, 33 657:1162–1171, March 2007. doi: 10.1086/511062. M. J. Mumma, B. P. Bonev, G. L. Villanueva, L. Paganini, M. A. DiSanti, E. L. Gibb, J. V. Keane, K. J. Meech, G. A. Blake, R. S. Ellis, M. Lippi, H. Boehn- hardt, and K. Magee-Sauer. Temporal and Spatial Aspects of Gas Re- lease During the 2010 Apparition of Comet 103P/Hartley 2. Astrophysical Journal Letters, 734:L7, June 2011. doi: 10.1088/2041-8205/734/1/L7. V. Novski, A. Novichonok, O. Burhonov, W. H. Ryan, E. V. Ryan, H. Sato, E. Guido, G. Sostero, N. Howes, and G. V. Williams. Comet C/2012 S1 (Ison). Central Bureau Electronic Telegrams, 3238:1, September 2012. C. Opitom, E. Jehin, J. Manfroid, D. Hutsem´ekers, and M. Gillon. TRAPPIST monitoring of comets C/2012 S1 (ISON) and C/2013 R1 (Lovejoy). In K. Muinonen, A. Penttila, M. Granvik, A. Virkki, G. Fedorets, O. Wilkman, and T. Kohout, editors, Asteroids, Comets, Meteors 2014, July 2014. C. Opitom, E. Jehin, J. Manfroid, D. Hutsem´ekers, M. Gillon, and P. Ma- gain. TRAPPIST photometry and imaging monitoring of comet C/2013 R1 (Lovejoy): Implications for the origin of daughter species. Astronomy and Astrophysics, 584:A121, December 2015. doi: 10.1051/0004-6361/ 201526427. D. G. Schleicher and M. F. A'Hearn. The fluorescence of cometary OH. Astrophysical Journal, 331:1058–1077, August 1988. doi: 10.1086/166622. Z. Sekanina and R. Kracht. Disintegration of Comet C/2012 S1 (ISON) Shortly Before Perihelion: Evidence from Independent Data Sets. ArXiv e-prints, August 2014. BD Sharpee and TG Slanger. O (1d2-3p2, 1, 0) 630.0, 636.4, and 639.2 nm forbidden emission line intensity ratios measured in the terrestrial nightglow. The Journal of Physical Chemistry A, 110(21):6707–6710, 2006. J. K. Steckloff, B. C. Johnson, T. Bowling, H. Jay Melosh, D. Minton, C. M. 34 Lisse, and K. Battams. Dynamic sublimation pressure and the catastrophic breakup of Comet ISON. Icarus, 258:430–437, September 2015. doi: 10. 1016/j.icarus.2015.06.032. S.-i. Wang, R. H. Hildebrand, L. M. Hobbs, S. J. Heimsath, G. Kelderhouse, R. F. Loewenstein, S. Lucero, C. M. Rockosi, D. Sandford, J. L. Sund- wall, J. A. Thorburn, and D. G. York. ARCES: an echelle spectrograph for the Astrophysical Research Consortium (ARC) 3.5m telescope. In M. Iye and A. F. M. Moorwood, editors, Instrument Design and Performance for Optical/Infrared Ground-based Telescopes, volume 4841 of Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, pages 1145–1156, March 2003. doi: 10.1117/12.461447. 35 Table 1 Observation Log Date (UT) Instrument r (AU) r (km s−1) ∆ (AU) ∆ (km s−1) FOV (km) G2V A0V Flux Cal Conditions 9/23/2013 ARCES 1.80 -31.3 2.38 -48.9 2760 × 5540 HD30854 HR3224 HR 3454a Clear 10/3/2013 ARCES 1.61 -33.0 2.09 -50.9 2430 × 4860 HR 79078 HR 3711 HR 3454 Clear 10/18/2013 Tull Coude 1.31 -36.6 1.64 -51.9 1430 × 9780 Solar Port γ Gem γ Gem Clear 3 6 10/20/2013 Tull Coude 1.27 -37.2 1.58 -51.8 1380 × 9420 Solar Port γ Gem γ Gem Clear 10/21/2013 Tull Coude 1.25 -37.5 1.55 -51.7 1350 × 9240 Solar Port γ Gem γ Gem Clear 10/25/2013 HIRESb 1.15 -39.1 1.42 -50.8 890 × 7230 Hyades 64 HR 2207 Hilt 600 Clear 10/28/2013 HIRESb 1.08 -40.3 1.33 -49.8 830 × 6770 Hyades 64 HR 3134 Hilt 600 Clear 11/6/2013 ARCES 0.87 -44.8 1.10 -43.5 1280 × 2560 HD 95868 HR 4464 HR 4468 Clear 11/15/2013 ARCES 0.60 -53.5 0.90 -25.7 1050 × 2100 HD 110747 HR 4722 HR 4963 Clear 11/20/2013 ARCES 0.44 -62.8 0.86 -4.1 1000 × 2000 HD 110747 HR 4722 HR 4963b Passing Clouds a Flux Standard taken from UT October 3 observation b Flux Standard taken from UT November 15 observation Table 2 Parameter Values Used in the Haser Models Molecule τp (s)a τd (s)a Vej (km s−1) g-factor (ergs s−1 molecule−1)a OH O Ib O Ic 1.3 × 104 2.1 × 105 1.02 3.5 × 10−15 8.3 × 104 1.3 × 105 - - - - - - a Given for r=1 AU. The listed g-factor also accounts for the Swings effect. b For [O I] from dissociation of H2O into H2 and O; branching ratio employed is 0.07 (Bhardwaj and Raghuram, 2012) c For [O I] from dissociation of OH; branching ratio for H2O to OH + H employed is 0.855 (Huebner et al., 1992) and the branching ratio for OH to O + H is 0.094 (Bhardwaj and Raghuram, 2012). 37 Table 3 Production Rates UT Date Rh (AU) QH2O (1027 mol s−1) Active Area (km2)a Active Fractionb (%) 9/23/2013 1.80 7.65 ± 1.03 9.0 ± 1.2 127 ± 54 10/3/2013 1.61 9.10 ± 1.15 7.9 ± 1.0 112 ± 47 10/18/2013 1.31 14.5 ± 3.5 7.5 ± 1.8 106 ± 49 10/20/2013 1.27 21.8 ± 4.9 10.5 ± 2.4 149 ± 68 10/21/2013 1.25 17.3 ± 3.9 8.0 ± 1.8 113 ± 52 10/25/2013 1.15 12.7 ± 2.4 4.8 ± 0.9 68 ± 30 10/28/2013 1.08 14.1 ± 2.4 4.7 ± 0.8 66 ± 29 11/6/2013 0.87 21.3 ± 4.3 4.3 ± 1.0 61 ± 28 11/15/2013 0.60 347 ± 29.1 31.8 ± 2.7 450 ± 180 11/20/2013 0.44 400 ± 29.0 19.2 ± 1.4 270 ± 110 a Assuming an albedo of 0.05, thermal emissivity of unity, and a slow-rotator ther- mal model (Cowan and A'Hearn, 1979). See Tables 4 and 5 for different parameter values. b Assuming a spherical nucleus with radius of 0.75 ± 0.15 km (Lamy et al., 2014). 38 Table 4 Active Area (km2) and Active Fraction (%) for 0.05 Albedo Model Isothermal Fast-Rotator Slow-Rotator Subsolar UT Date Active Area Active Fraction Active Area Active Fraction Active Area Active Fraction Active Area Active Fraction 9/22/2013 16.2 ± 2.2 230 ± 97 15.3 ± 2.1 217 ± 92 9.0 ± 1.2 127 ± 54 1.9 ± 0.3 27 ± 11 10/3/2013 12.2 ± 1.5 173 ± 73 11.8 ± 1.5 168 ± 70 7.9 ± 1.0 112 ± 47 1.7 ± 0.2 24 ± 10 3 9 10/18/2013 9.8 ± 2.4 139 ± 65 9.7 ± 2.3 137 ± 64 7.5 ± 1.8 106 ± 50 1.7 ± 0.4 24 ± 11 10/20/2013 13.5 ± 3.0 190 ± 87 13.3 ± 3.0 188 ± 86 10.5 ± 2.4 148 ± 68 2.4 ± 0.5 34 ± 16 10/21/2013 10.2 ± 2.3 144 ± 66 10.1 ± 2.3 143 ± 65 8.0 ± 1.8 113 ± 52 1.8 ± 0.4 26 ± 12 10/25/2013 5.9 ± 1.1 84 ± 37 5.9 ± 1.1 83 ± 37 4.8 ± 0.9 69 ± 30 1.1 ± 0.2 16 ± 7 10/28/2013 5.5 ± 0.9 78 ± 34 5.5 ± 0.9 78 ± 34 4.7 ± 0.8 66 ± 29 1.1 ± 0.2 15 ± 7 11/6/2013 4.8 ± 1.0 68 ± 31 4.8 ± 1.0 68 ± 31 4.3 ± 0.9 61 ± 27 1.0 ± 0.2 15 ± 7 11/15/2013 33.7 ± 2.8 480 ± 200 33.7 ± 2.8 480 ± 200 31.8 ± 2.7 450 ± 180 7.7 ± 60 110 ± 40 11/20/2013 19.8 ± 1.4 280 ± 110 19.9 ± 1.4 280 ± 110 19.2 ± 1.4 270 ± 110 4.7 ± 30 67 ± 27 Table 5 Active Area (km2) and Active Fraction (%) for 0.50 Albedo Model Isothermal Fast-Rotator Slow-Rotator Subsolar UT Date Active Area Active Fraction Active Area Active Fraction Active Area Active Fraction Active Area Active Fraction 9/22/2013 26.3 ± 3.5 370 ± 160 25.2 ± 3.4 360 ± 150 16.2 ± 2.2 230 ± 100 3.5 ± 0.5 49 ± 21 10/3/2013 20.7 ± 2.6 290 ± 120 20.3 ± 2.6 290 ± 120 14.4 ± 1.8 200 ± 90 3.2 ± 0.4 45 ± 19 4 0 10/18/2013 17.4 ± 4.2 250 ± 120 17.2 ± 4.2 240 ± 110 13.9 ± 3.4 200 ± 90 3.2 ± 0.8 45 ± 21 10/20/2013 23.9 ± 5.4 340 ± 160 23.7 ± 5.3 340 ± 150 19.5 ± 4.4 280 ± 130 4.5 ± 1.0 63 ± 29 10/21/2013 18.2 ± 4.1 260 ± 120 18.0 ± 4.1 260 ± 120 14.9 ± 3.4 210 ± 100 3.4 ± 0.8 49 ± 22 10/25/2013 10.7 ± 2.0 150 ± 70 10.6 ± 2.0 150 ± 70 9.0 ± 1.7 130 ± 60 2.1 ± 0.4 30 ± 13 10/28/2013 10.1 ± 1.7 140 ± 60 10.0 ± 1.7 140 ± 60 8.7 ± 1.5 120 ± 50 2.0 ± 0.3 29 ± 13 11/6/2013 8.9 ± 1.8 130 ± 60 8.9 ± 1.8 130 ± 60 8.1 ± 1.6 120 ± 50 2.0 ± 0.4 28 ± 12 11/15/2013 63.1 ± 5.3 890 ± 370 63.1 ± 5.3 890 ± 370 60.0 ± 5.0 850 ± 350 14.7 ± 1.2 210 ± 90 11/20/2013 37.4 ± 2.7 530 ± 210 37.4 ± 2.7 530 ± 210 36.4 ± 2.6 510 ± 210 9.0 ± 0.7 130 ± 50 Figure Captions Fig 1: Top: Spectrum showing the [O I] line region on November 15. The [O I]6300 A and [O I]6364 A lines are labeled. Most other emission features present are due to NH2. Bottom: Spectrum showing the OH A-X band on October 25. Error bars have been omitted from both plots for clarity. Fig 2: Top: Spectrum showing the [O I]6300 A line on November 15. The telluric line is the weaker feature redward of the cometary line. Bottom: Spec- trum showing the [O I]6364 A line on November 15. The telluric line is the weaker feature redward of the cometary line. Fig 3: Raw extracted spectra (i.e. no telluric removal, solar subtraction, or flux calibration) showing the [O I]6300 A line (top row) and the [O I]6364 A line (bottom row) on UT October 3 (left column) and UT November 15 (right col- umn), along with raw extracted spectra of the telluric standard on each date. The telluric standard has been shifted down and for October 3 the counts were scaled by a factor of 10 to facilitate comparison with the comet spectrum. On October 3 the [O I]6300 A line is directly coincident with an O2 telluric absorp- tion feature, while by November 15 the cometary line is no longer coincident with any O2 telluric absorption. The placement of the cometary line relative to the O2 absorption is similar for all our September and October observa- tions. There are no strong telluric features underneath the [O I]6364 A line on either date. Therefore for our September and October observations we calcu- late the H2O production rate using the the [O I]6364 A line rather than the [O I]6300 A line. Fig 4: Plot of H2O production in comet ISON as a function of heliocentric distance. Production rates from this work are plotted as filled symbols, with 41 circles denoting production rates based on [O I] observations and squares denoting production rates based on OH. Other results from the literature are plotted for comparison as empty symbols. Our values are in agreement with other measurements, and extend coverage of ISON's H2O production rate to larger heliocentric distances than many previously published results. Fig 5: Derived active area as a function of heliocentric distance for different thermal models and albedos. The top panel shows results for an albedo of 0.05, while the bottom panel shows results for a much larger albedo of 0.50. For both albedos the isothermal and fast-rotator models give nearly identical values, so for clarity the isothermal model has not been plotted. The dashed line is the total surface area of a spherical nucleus with a radius of 0.75 km (Lamy et al., 2014) and the shaded area depicts the uncertainty in this measurement. There- fore active areas that fall in the shaded region are consistent with an active fraction of 100%, while those above the shaded region indicate ISON is hyper- active, and additional surface area besides that of the nucleus is required to explain the observed H2O production rates. Changing the albedo and employ- ing different thermal models changes the absolute value of the derived active area, but trends over the apparition are independent of the model employed. The fast-rotator and slow-rotator models suggest an active fraction of at least 50%, which could be indicative of a largely active surface or the presence of icy grains in the coma. The subsolar model gives much lower active areas, but we believe this model does not provide a realistic description of ISON's activity (see Section 4.3.2). 42 [OI]6300 [OI]6364 Fig. 1. 43 Fig. 2. 44 200 0 -200 -400 200 100 0 [OI]6300 Oct 3 [OI]6300 Nov 15 0 6298.5 6299 6299.5 6300 6299 6299.5 6300 6300.5 Wavelength (Angstroms) Wavelength (Angstroms) [OI]6364 Oct 3 [OI]6364 Nov 15 6362.5 6363 6363.5 6364 Wavelength (Angstroms) 6362 6362.5 6363 6363.5 Wavelength (Angstroms) 5000 0 Fig. 3. 45 This Work (OI) This Work (OH) DiSanti et al. 2016 Dello Russo et al. 2016 Combi et al. 2014 Knight and Schleicher 2015 Bonev et al. 2014 Opitom et al. 2014 Fig. 4. 46 Fast-Rotator Model Slow-Rotator Model Subsolar Model Fast-Rotator Model Slow-Rotator Model Subsolar Model Fig. 5. 47
1308.4968	2	1308	2013-12-23T22:30:28	Viscoelastic Tidal Dissipation in Giant Planets and Formation of Hot Jupiters Through High-Eccentricity Migration	[ "astro-ph.EP" ]	We study the possibility of tidal dissipation in the solid cores of giant planets and its implication for the formation of hot Jupiters through high-eccentricity migration. We present a general framework by which the tidal evolution of planetary systems can be computed for any form of tidal dissipation, characterized by the imaginary part of the complex tidal Love number, ${\rm Im}[{\tilde k}_2(\omega)]$, as a function of the forcing frequency $\omega$. Using the simplest viscoelastic dissipation model (the Maxwell model) for the rocky core and including the effect of a nondissipative fluid envelope, we show that with reasonable (but uncertain) physical parameters for the core (size, viscosity and shear modulus), tidal dissipation in the core can accommodate the tidal-Q constraint of the Solar system gas giants and at the same time allows exoplanetary hot Jupiters to form via tidal circularization in the high-e migration scenario. By contrast, the often-used weak friction theory of equilibrium tide would lead to a discrepancy between the Solar system constraint and the amount of dissipation necessary for high-e migration. We also show that tidal heating in the rocky core can lead to modest radius inflation of the planets, particularly when the planets are in the high-eccentricity phase ($e\sim 0.6$) during their high-e migration. Finally, as an interesting by-product of our study, we note that for a generic tidal response function ${\rm Im}[{\tilde k}_2(\omega)]$, it is possible that spin equilibrium (zero torque) can be achieved for multiple spin frequencies (at a given $e$), and the actual pseudo-synchronized spin rate depends on the evolutionary history of the system.	astro-ph.EP	astro-ph	Mon. Not. R. Astron. Soc. 000, 1 -- ?? (2013) Printed 9 June 2021 (MN LATEX style file v2.2) Viscoelastic Tidal Dissipation in Giant Planets and Formation of Hot Jupiters Through High-Eccentricity Migration Natalia I Storch⋆ and Dong Lai Center for Space Research, Department of Astronomy, Cornell University, Ithaca, NY 14853, USA 9 June 2021 ABSTRACT We study the possibility of tidal dissipation in the solid cores of giant planets and its implication for the formation of hot Jupiters through high-eccentricity migration. We present a general framework by which the tidal evolution of planetary systems can be computed for any form of tidal dissipation, characterized by the imaginary part of the complex tidal Love number, Im[k2(ω)], as a function of the forcing frequency ω. Using the simplest viscoelastic dissipation model (the Maxwell model) for the rocky core and including the effect of a nondissipative fluid envelope, we show that with reasonable (but uncertain) physical parameters for the core (size, viscosity and shear modulus), tidal dissipation in the core can accommodate the tidal-Q constraint of the Solar system gas giants and at the same time allows exoplanetary hot Jupiters to form via tidal circularization in the high-e migration scenario. By contrast, the often-used weak friction theory of equilibrium tide would lead to a discrepancy between the Solar system constraint and the amount of dissipation necessary for high-e migration. We also show that tidal heating in the rocky core can lead to modest radius inflation of the planets, particularly when the planets are in the high-eccentricity phase (e ∼ 0.6) during their high-e migration. Finally, as an interesting by-product of our study, we note that for a generic tidal response function Im[k2(ω)], it is possible that spin equilibrium (zero torque) can be achieved for multiple spin frequencies (at a given e), and the actual pseudosynchronized spin rate depends on the evolutionary history of the system. Key words: hydrodynamics -- planets and satellites: general -- planets and satellites: interiors -- planet-star interactions -- binaries: close 1 INTRODUCTION In recent years, high-eccentricity migration has emerged as one of the dominant mechanisms responsible for the forma- tion of hot Jupiters. In this mechanism, a gas giant which is formed beyond the snow line is first excited into a state of very high eccentricity (e >∼ 0.9) by few-body interac- tions, either via dynamical planet-planet scatterings (Ra- sio & Ford 1996; Weidenschilling & Marzari 1996; Zhou, Lin & Sun 2007; Chatterjee et al. 2008; Juric & Tremaine 2008) or/and secular interactions between multiple planets, or the Kozai effect induced by a distant companion (Wu & Murray 2003; Fabrycky & Tremaine 2007; Wu, Murray & Ramshhai 2007; Nagasawa, Ida & Bessho 2008; Katz, Dong & Malhotra 2011; Naoz et al. 2011,2013; Wu & Lithwick ⋆ Email: [email protected], [email protected] c(cid:13) 2013 RAS 2011; Naoz, Farr & Rasio 2012; see also Dawson & Murray- Clay 2013). Due to the high eccentricity, the planet passes quite close to its host star at periastron, and tidal dissipa- tion in the planet extracts energy from the orbit, leading to inward migration and circularization of the planet's orbit. Tidal effects on the orbital evolution of binaries are of- ten discussed using the weak friction theory of equilibrium tides (Darwin 1879; Alexander 1973; Hut 1981; Eggleton et al. 1998), according to which the rate of decay of the semi-major axis (a) for a pseudosynchronized planet can be written as (cid:12)(cid:12)(cid:12)(cid:12) a a(cid:12)(cid:12)(cid:12)(cid:12) = 6k2τ(cid:18) GM⋆ F (cid:19)(cid:18) Rp a3 aF (cid:19)5 M⋆ Mpr aF a F (e). (1) Here, Mp and Rp are the mass and radius of the planet, M⋆ is the mass of the central star, aF ≡ a(1 − e2) is the final circularization radius (assuming orbital angular momentum 2 Natalia I Storch and Dong Lai conservation), k2 the tidal Love number, τ is the tidal lag time (assumed constant in the weak friction theory), and F (e) is a function of eccentricity of order (1 − 10) and given by F (e) = f1(e) − f 2 2 (e)/f5(e), where f1, f2 and f5 are given by Eq. (11) of Hut (1981). By requiring that the high-e mi- gration happens on a timescale less than 10 Gyr we can place a constraint on τ : (cid:18) GM⋆ a3 τ ∼ F (cid:19)1/2 ×(cid:18) M⋆ > 3 × 10−5(cid:16) a MJ (cid:18) Rp 5AU(cid:17)1/2(cid:16) aF 0.06AU(cid:17)6 RJ(cid:19)−5(cid:18) k2 0.38(cid:19)−1 M⊙(cid:19)−3/2 Mp . (2) Note that instead of τ , tidal dissipation is often parametrized by the tidal quality factor Q ≡ (τ ω)−1, where ω is the tidal forcing frequency. Thus, the above constraint on τ trans- lates to Q <∼ 3 × 104 at ω ∼ (GM⋆/a3 F )1/2 ∼ 2π/(5 d) [for the canonical parameters adopted in Eq. (2)]. A similar constraint can be obtained by integration over the planets' orbital evolution (e.g., Fabrycky & Tremaine 2007; Leconte et al. 2010; Matsumura, Peale & Rasio 2010; Hansen 2012; Naoz et al. 2012; Socrates, Katz & Dong 2012b). The tidal Q for Solar system giant planets can be mea- sured or constrained by the tidal evolution of their satellites (Goldreich & Soter 1966). For Jupiter, Yoder & Peale (1981) derived a bound 2 × 10−7 < k2/QJ < 6 × 10−6 based on Io's long-term orbital evolution (particularly the eccentricity equilibrium), with the upper limit following from the limited expansion of satellite orbits. Recent analysis of the astromet- ric data of Galilean moons gave k2/QJ = (1.1 ± 0.2) × 10−5 for the current Jupiter-Io system (Lainey et al. 2009), cor- responding to QJ ≃ 3.5 × 104 for the conventional value of the Love number k2 = 0.38 (Gavrilov & Zharkov 1977). With Jupiter's spin period 9.9 hrs and Io's orbital pe- riod 42.5 hrs, the tidal forcing frequency on Jupiter from Io is ω = 2π/(6.5 hr), and the tidal lag time is then τJ = (QJ ω)−1 ≃ 0.1 s. For Saturn, theoretical considerations based on the long-term evolution of Mimas and other main moons (with the assumption that they formed above the synchronous or- bit 4.5 Gyr ago) lead to the constraint 3 × 10−6 < k2/QS < 2 × 10−5 (Sinclair 1983; Peale 1999). However, using as- trometric data spanning more than a century, Lainey et al. (2012) found a much larger k2/QS = (2.3 ± 0.7) × 10−4, corresponding to QS = (1 − 2) × 103 for k2 = 0.34; they also found that QS depends weakly on the tidal period in the range between 2π/ω = 5.8 hrs (Rhea) and 7.8 hrs (Ence- ladus). Assuming that extra-solar giant planets are close analogs of our own gas giants, we can ask whether the afore- mentioned empirical constraints on k2/Q for Jupiter and Saturn are compatible with the extra-solar constraint [see Eq. (2)]. The difference between the two sets of constraints is the tidal forcing frequencies: For example, the Jupiter-Io constraint involves a single frequency (P = 6.5 hrs), while high-e migration involves tidal potentials of many harmon- ics, all of them with periods longer than a few days. Socrates et al. (2012b) showed that the two sets of constraints are in- compatible with the weak friction theory (see also Naoz et al. 2012): In order for hot Jupiters to undergo high-e migra- tion within the age of their host stars, their required tidal lag times must be more than an order of magnitude larger than the Jupiter-Io constraint. Tidal dissipation in giant planets is complex, and de- pends strongly on the internal structure of the planet, such as the stratification of the liquid envelope and the presence and properties of a solid core. There have been some at- tempts to understand the physics of tidal Q in giant planets (see Ogilvie & Lin 2004 for a review). It has long been known (Goldreich & Nicholson 1977) that simple turbulent viscos- ity in the fluid envelopes of giant planets is many orders of magnitude lower than required by observations. Ioannou & Lindzen (1993a,b) considered a prescribed model of Jupiter where the envelope is not fully convective (contrary to the conventional model where the envelope is neutrally buoy- ant to a high degree; see Guillot 2005) and showed that the excitation and radiative damping of gravity waves in the envelope provide efficient tidal dissipation only at specific "resonant" frequencies. Lubow et al. (1997) examined simi- lar gravity wave excitations in the radiative layer above the convective envelope of hot Jupiters. So far the most sophis- ticated study of dynamical tides in giant planets is that by Ogilvie & Lin (2004) (see also Goodman & Lackner 2009; Ogilvie 2009,2013), who focused on the tidal forcing of iner- tial waves (short-wavelength disturbances restored primar- ily by Coriolis force) in the convective envelope of a rotating planet [see Ivanov & Papaloizou (2007) and Papaloizou & Ivanov (2010) for highly eccentric orbits, and Wu (2005) for a different approach]. They showed that because of the rocky core, the excited inertial waves are concentrated on a web of "rays", leading to tidal dissipation which depends on the forcing frequency in a highly erratic way. The tidal Q obtained is typically of order 106−7. It remains unclear whether this mechanism can provide sufficient tidal dissipa- tion compared to the observational constraints. The possibility of core dissipation in giant planets was first considered by Dermott (1979) but has not received much attention since. Recently, Remus et al. (2012a) showed that dissipation in the solid core could in principle satisfy the constraints on tidal Q obtained by Lainey et al. (2009) for Jupiter and by Lainey et al. (2012) for Saturn. In this paper, we continue the study of tidal dissipation in the solid core of giant planets and examine its conse- quences for the high-e migration scenario and for the ther- mal evolution of hot Jupiters. In section 2 we present the general tidal theory which may be used with any tidal re- sponse model. In section 3 we discuss a simple viscoelastic tidal response model and its range of applicability. In sec- tion 4 we use the general theory of section 2 in conjunction with the model of section 3 to compute high-e migration timescales and compare with the weak friction theory. We also examine the effect of tidal heating in the core for the radius evolution of the planets. We summarize our findings and conclude in section 5. 2 EVOLUTION OF ECCENTRIC SYSTEMS WITH GENERAL TIDAL RESPONSES Here we formulate the tidal evolution equations for eccen- tric binary systems. These equations can be applied to any tidal response model, where the complex Love number (de- fined below) is an arbitrary function of the tidal forcing fre- c(cid:13) 2013 RAS, MNRAS 000, 1 -- ?? quency [see also Efroimsky & Makarov (2013), Mathis & Le Poncin-Lafitte (2009) and Remus et al. (2012a) for sim- ilar formalisms]. This formulation is valid as long as the responses of the body to different tidal components are in- dependent of each other. We consider a planet of mass Mp, radius Rp and rota- tion rate Ωs (assumed to be aligned with the orbital angu- lar momentum axis), moving around a star (mass M⋆) in an eccentric orbit with semi-major axis a and mean motion frequency Ω. The tidal potential exerted on the planet by the star is given by U (r, t) = −GM⋆Xm W2mr2 D(t)3 e−imΦ(t)Y2m(θ, φ), (3) where (r, θ, φ) is the position vector (in spherical coordi- nates) relative to the center of mass of the planet, D(t) and Φ(t) are the time-dependent separation and phase of the orbit, and m = 0, ±2, with W20 = −(π/5)1/2 and W2±2 = (3π/10)1/2. The potential U (r, t) can be decom- posed into an infinite series of circular harmonics: U (r, t) = −Xm,N where N ∈ (−∞, ∞) and UmN r2Y2m(θ, φ)e−iNΩt, (4) UmN ≡ GM⋆ a3 W2mFmN (e), (5) with FmN (e) being the Hansen coefficient (e.g., called X N 2m in Murray & Dermott 2000), given by FmN (e) = with 1 π Z π 0 cos [N (Ψ − e sin Ψ) − m Φ(t)] (1 − e cos Ψ)2 dΨ, (6) cos Φ(t) = cos Ψ − e 1 − e cos Ψ . (7) Each harmonic of the tidal potential produces a per- turbative response in the planet, expressible in terms of the Lagrangian displacement ξmN and the Eulerian density per- turbation δρmN . These responses are proportional to the di- 0 = (M⋆/Mp)(Rp/a)3W2mFmN , mensionless ratio, UmN /ω2 where ω0 ≡ (GMp/R3 p)1/2 is the dynamical frequency of the planet. Without loss of generality, we can write the tidal responses as (see Lai 2012) ξmN (r, t) = δρmN (r, t) = with UmN ω2 0 UmN ω2 0 ¯ξmN (r, θ)eimφ−iNΩt, δ ¯ρmN (r, θ)eimφ−iNΩt, (8) (9) δρmN = −∇ · (ρξmN ). (10) Note that δ ¯ρmN and ¯ξmN are in general complex functions (implying that the tidal response is phased-shifted relative to the tidal potential), and they depend on the forcing fre- quency ωmN of each harmonic in the rotating frame of the primary, ωmN ≡ N Ω − mΩs. (11) Given the Eulerian density perturbation, we can obtain the perturbation to the gravitational potential of the c(cid:13) 2013 RAS, MNRAS 000, 1 -- ?? Tidal Dissipation in Giant Planets 3 planet, δΦmN , by solving the Poisson equation, ∇2δΦmN = 4πGδρmN . We define the dimensionless Love number kmN as the ratio of δΦmN and the (mN )-component of the tidal po- tential [U (r, t)]mN = −r2UmN Y2m(θ, φ) exp(−iN Ωt), evalu- ated at the planet's surface: 2 (12) kmN 2 ≡ . δΦmN [U (r, t)]mN(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)r=Rp p Z δ ¯ρmN (r, θ)eimφr2Y ∗ Note that, just as δρmN is complex, so in general is kmN We find that 2 . kmN 2 = 4π 5 1 MpR2 2md3x. (13) We now have all the information necessary to calculate the time-averaged torque and energy transfer rate (from the orbit to the planet): T = ReZ d3x δρ(r, t) r × [−∇U ∗(r, t)]+, E = ReZ d3x ρ(r) · [−∇U ∗(r, t)]+, ∂ξ(r, t) ∂t (14) (15) where h i denotes time averaging. After plugging in the ansatz for ξ and δρ [Eqs. (8)-(9)] and the expression for kmN 2 [Eq. (13)], we find Tz = E = 5 4π 5 4π T0 Xm,N T0ΩXm,N m [W2mFmN (e)]2 Im(kmN 2 ), (16) N [W2mFmN (e)]2 Im(kmN 2 ), (17) where T0 ≡ G(cid:0)M⋆/a3(cid:1)2 R5 p. The tidal evolution equations for the planet's spin Ωs, the orbital semi-major axis a and the eccentricity e are Ωs = Tz I , a a = − 2a E GM⋆Mp , e e 1 − e2 = − a E GM⋆Mp + Tz L , (18) (19) (20) where I is the moment of inertia of the planet and L = momentum. As noted before, kmN M⋆Mp(cid:2)Ga(1 − e2)/(M⋆ + Mp)(cid:3)1/2 is the orbital angular depends on the forcing frequency ωmN = N Ω − mΩs and physical properties of the planet. We can write kmN 2 = k2(ωmN ). In general, given a model for k2(ω), the sum over (mN ) must be computed numeri- cally. Note that Im(kmN ) is related to the often-defined tidal quality factor Q by 2 2 Im(kmN 2 ) ≡(cid:18) k2 Q(cid:19)mN , (21) with k2 the usual (real) Love number, except that in our general case (k2/Q)mN is for a specific (mN )-tidal compo- nent. In the special case of the weak friction theory of equilib- rium tide 1, one assumes Im[k2(ω)] = k2τ ω, with k2 and the 1 Note that for equilibrium tides in general, the tidal response 4 Natalia I Storch and Dong Lai lag time τ being independent of the frequency ω. In this case, the sum over (mN ) can be carried out analytically, giving the usual expressions (see Alexander 1973, Hut 1981): Tz = E = 3 T0 Ω k2 τ (1 − e2)6 (cid:20)f2 − (1 − e2)3/2f5 (1 − e2)15/2 (cid:20)f1 − (1 − e2)3/2f2 3 T0 Ω2 k2 τ Ωs Ω (cid:21) , Ω (cid:21) , Ωs (22) (23) where f1, f2, and f5 are functions of eccentricity given by (Hut 1981) f1(e) = 1 + f2(e) = 1 + 31 2 15 2 e2 + e2 + f5(e) = 1 + 3e2 + 3 8 8 45 8 e4. 255 e4 + e4 + 5 16 25 64 e6 + 185 16 e6, and e8, (24) (25) (26) 3 VISCOELASTIC DISSIPATION IN GIANT PLANETS WITH ROCKY CORES We now discuss a theoretical model of k2(ω) for giant planets based on viscoelastic dissipation in rocky cores. We consider first a homogeneous solid core, and subsequently introduce a homogeneous non-dissipative liquid envelope. 3.1 Viscoelastic Solid Core The rocky/icy core of a giant planet can possess the charac- teristics of both elastic solid and viscous fluid, depending on the frequency of the imposed periodic shear stress or strain. Dissipation in rocks arises from thermally activated creep processes associated with the diffusion of atoms or the mo- tion of dislocations when the rocks are subjected to stress. We use the simplest phenomenological model, the Maxwell model, to describe such viscoelastic materials (Turcotte & Schubert 2002). The model contains two free parameters, the shear modulus (rigidity) µ and viscosity η. Other rheolo- gies are possible (see Henning, O'Connell & Sasselov 2009), but contain more free parameters and are not warranted at present given the large uncertainties associated with the solid cores of giant planets. The incompressible constitutive relation of a Maxwell solid core takes the form εij = 1 2µ σij + 1 2η σij, (27) where εij and σij are strain and stress tensors, respec- tively, and a dot denotes time derivative. For periodic forcing εij, σij ∝ e−iωt, the complex shear modulus, µ ≡ σij/(2εij ), is given by µ = ωµη ωη + iµ = µ 1 + i(ωM /ω) , (28) where the Maxwell frequency is ωM ≡ µ/η. (29) Im[k2(ω)] does not have to be a linear function of ω (i.e., con- stant lag time). For example, Remus et al. (2012b) showed that for convective stars/planets, Im[k2(ω)] is independent of ω (i.e., constant lag angle) when ω exceeds the convective turnover rate. Clearly, the core behaves as an elastic solid (with µ ≃ µ) for ω ≫ ωM , and as a viscous fluid (with µ ≃ −iωη) for ω ≪ ωM . Consider a homogeneous rocky core (mass Mc, radius Rc and density ρc) with a constant µ. When the tidal forcing frequency ω is much less than the dynamical frequency of the body, i.e., when ω ≪ (GMc/R3 c)1/2, the tidal Love number in the purely elastic case (Im[µ] = 0) can be obtained analytically (Love 1927). Following Remus et al. (2012a) we invoke the correspondence principle (Biot 1954), which allows us to simply replace the real shear mod- ulus in the elastic solution by the full complex shear modulus in order to obtain the viscoelastic solution, yielding c)1/2 and ω ≪ (µ/ρcR2 k2c = 3 2 1 1 + ¯µ , where ¯µ is the body's (dimensionless) effective rigidity ¯µ ≡ ¯µ1 + i¯µ2 ≡ 19µ 2β , with β ≡ ρcgcRc and gc = GMc/R2 c. Thus we have Im(k2c) = 57ωη 4β "1 +(cid:18) ωη µ (cid:19)2(cid:18)1 + 19µ 2β (cid:19)2#−1 . (30) (31) (32) Note that Im(k2c) is a non-monotonic function of ω (see Fig. 1, top panel). For ω ≪ ωM , we have Im(k2c) ≃ 57ωη/(4β); for ω ≫ ωM , we have Im(k2c) ∝ ω−1. For a given core model, the maximum Im(k2c)max = 3µ 4(1 + µ) (33) is attained at ω = ωM /(1 + µ), where µ ≡ 19µ/(2β). 3.2 Application to a giant planet with a rocky core In order to apply the results of section 3.1 to a gas giant, we introduce a non-dissipative fluid envelope on top of the rocky body. While the fluid envelope does not, itself, dissipate en- ergy, it is deformed by the tidal potential and interacts with the central solid body by exerting variable pressure on its surface, thus creating additional stress. We consider a core of radius Rc and density ρc within a planet of radius Rp, with a fluid envelope of density ρF . We then use the analyt- ical expression of Remus et al. (2012a), who used Dermott's 1979 solution for the effect of a liquid envelope on the defor- mation of an elastic core, together with the correspondence principle (Biot et al. 1954), to calculate the resulting mod- ified Love number of the core, defined as the ratio of the potential generated by the deformed core and the tidal po- tential, evaluated at the core radius (Rc): k2c = 1 (B + ¯µ1)2 + ¯µ2 2("(B + ¯µ1)(cid:18)C + + 3 2α ¯µ2 2# − iAD ¯µ2), 3 2α ¯µ1(cid:19) (34) c(cid:13) 2013 RAS, MNRAS 000, 1 -- ?? where (Remus et al. 2012a) Tidal Dissipation in Giant Planets 5 α = 1 + 5 2 A = (cid:18)1 − B = 1 − ρF ρc C = D = 3 2 (cid:18)1 − 2 (cid:18)1 − 3 ρF ρc (cid:19) , ρF 3 + ρc ρF 3 2 ρF (cid:18) Rc ρc (cid:19)(cid:18)1 + Rp(cid:19)3 (cid:18)1 − 2α(cid:19) , ρc (cid:18)1 − ρc (cid:19) (cid:18)1 − ρc (cid:19) "1 + ρF ρc ρF ρF ρF ρc (cid:19) − + 5 9 4α (cid:18) Rc 2α(cid:19) + Rp(cid:19)5# . 3 2α (cid:18) Rc 0.100 0.050 0.010 0.005 c 2 kŽ m I , ρF Rp(cid:19)5(cid:18)1 − 4α (cid:18) Rc ρc (cid:19)2 Rp(cid:19)5 (cid:18)1 − 9 ρF ρc (cid:19)2 Since in our model, all the dissipation happens in the core, we then have, from section 2, E = 5 4π (cid:18) GM 2 ⋆ R5 c a6 (cid:19) ΩXm,N N [W2mFmN (e)]2 Im[kmN 2c ], (35) 2c = k2c(N Ω − mΩs). However, rather than keep where kmN the explicit dependence on Rc, we prefer to re-cast the equa- tion such that all core parameters appear in k2 only. We write, 5 4π (cid:18) GM 2 ⋆ R5 p a6 (cid:19) ΩXm,N E = where N [W2mFmN (e)]2 Im[kmN 2 ], (36) (37) Rp(cid:19)5 k2(ω) ≡(cid:18) Rc k2c(ω). This (complex) Love number is now, effectively, the Love number for the entire planet rather than for the core only. 3.3 The specific case of Jupiter The size of the rocky/icy core of Jupiter is uncertain, with estimates in the range of ∼ (0 − 10)M⊕ (Guillot 2005) and ∼ (14 − 18)M⊕ (Militzer et al. 2008). The viscous and elas- tic properties of materials at the high pressure (∼ 40 Mbar) found at the center of giant planets are also poorly known. We mention here values of η and µ for several materials to give the reader an idea for the range of parameter space in- volved. The inner core of the Earth has a measured viscosity of η ∼ 108±3 bar · s (Jeanloz 1990) and a shear modulus of µ ∼ 1500 kbar, while the central pressure is ∼ 3600 kbar (Montagner & Kennett 1996). In contrast, the Earth's man- tle has η ∼ 1015 − 1018 bar · s, depending on depth (Mitro- vica & Forte 2004), and shear modulus similar to the core. Icy materials have η ∼ 106 − 109 bar · s, and µ ∼ 50 kbar (Poirier, Sotin & Peyronneau 1981, Goldsby & Kohlstedt 2001). Evidently, η in particular has a very large dynamical range, and since very little is known about the interior of Jupiter, all of this range is hypothetically accessible. In ad- dition to varying η and µ, we may also vary the size of the core Rc and the core density ρc. Figure 1 presents three models for the tidal response Im(k2c) of Jupiter's rocky core (upper panel), and the corre- sponding effective tidal response of the entire planet Im(k2) (lower panel). For each curve, different values of η and Rc c(cid:13) 2013 RAS, MNRAS 000, 1 -- ?? , with fluid envelope no fluid envelope 0.001 0.001 0.01 0.1 1 10 100 Weak Friction Model 2 Model 1 2 kŽ m I 10-4 10-5 10-6 10-7 Model 3 Jupiter-Io 5 day orbit 10-8 0.01 0.1 1 Ω 2Πday 10 100 Figure 1. Theoretical curves for the tidal Love number of a Jupiter-mass planet as a function of the tidal forcing frequency, for several values of Rc/Rp and η, each calibrated to satisfy the Jupiter-Io constraint. Top: The intrinsic Love number of the rocky core with (blue lines) and without (red lines) the presence of liq- uid envelope. Bottom: The effective Love number for the entire planet with fluid envelope. The density ratio of the core and en- velope is ρc/ρF = 5 and the core rigidity is µ = 485 kbar for all models. The other model parameters are as follows. Model 1 (blue solid line): Rc/Rp = 0.13, η = 4.4 × 109 bar · s; Model 2 (blue long-dashed line): Rc/Rp = 0.19, η = 2 × 1010 bar · s; Model 3 (blue short-dashed line): Rc/Rp = 0.13, η = 3.3 × 108 bar · s. Green solid line: weak friction theory with τ = 0.06 s (the lag time obtained using the value of k2/Q from Lainey et al. (2009) and assuming k2 = 0.38). were chosen such that the Jupiter-Io tidal dissipation con- straint is satisfied (see also Fig. 10 of Remus et al. 2012a). Also plotted is the weak friction theory, similarly calibrated. For all the theoretical curves of Figure 1, we choose to fix µ and ρc, due to their smaller dynamical ranges. We note that of the remaining parameters, changing η acts primar- ily to alter the transition frequency ωM ∼ µη−1, effectively moving the curve horizontally left-right, while changing Rc effectively moves the curve up-down due to the strong de- pendence of k2 on Rc/Rp. From Figure 1, it is evident that the use of weak fric- tion theory, which due to having only one parameter needs only one data point to be completely constrained, can lead to strong over- or under- estimation of tidal dissipation at different frequencies, as compared to more realistic models. 6 Natalia I Storch and Dong Lai i r e p W s p W 1.15 1.10 1.05 1.00 0.95 0.90 0.85 Model 3 Model 1 Model 2 100.0 50.0 10.0 5.0 1.0 0.5 k a e w a(cid:160) a(cid:160) Model 2 Model 1 Model 3 0.2 0.4 0.6 e 0.8 1.0 Figure 2. Ratio of the pseudosynchronized spin frequency Ωps to the pericenter frequency Ωperi for each of the viscoelastic models of Figure 1, as well as for the (analytical) weak friction model. Each blue curve corresponds to one of the Maxwell model curves depicted in Fig. 1. The green curve shows the result of the weak friction theory. 0.02 0.05 0.10 0.20 0.50 1-e Figure 3. Ratio of the orbital decay rate a for different viscoelas- tic tidal dissipation models and aweak for the weak friction theory, as a function of eccentricity, for a fixed aF = a(1−e2) correspond- ing to final mean motion period of 5 days. Each curve corresponds to one of the blue Maxwell model curves of Fig. 1. In all cases, the weak friction theory is that of the green curve in Fig. 1. 4 HIGH-ECCENTRICITY MIGRATION OF A GIANT PLANET WITH A ROCKY CORE 4.1 Orbital Evolution We now compute the rates of high-e migration for a giant planet with a rocky core for different viscoelastic dissipation models depicted in Fig. 1, and compare the results with weak friction theory. We numerically carry out the sums in Eqs. (16)-(17) for different values of orbital eccentricity and a fixed final semi-major axis, i.e., the semi-major axis a and eccentricity e always satisfy a(1 − e2) ≡ aF =constant, corresponding to a final circular orbital period of 5 days. Since the timescale for changing the planet's spin is much shorter than the orbital evolution time, we assume that the planet is in the equilibrium spin state (Tz = 0) at all times. For the weak friction theory, the result is [see Eq. (22)] Ωps/Ωperi = (1 + e)−3/2f2/f5, where Ωperi = Ω/(1 − e)3/2 is the orbital frequency at the pericenter. For general vis- coelastic models, we set the right-hand-side of Eq. (16) to 0 and numerically solve for the equilibrium spin rate Ωps. The results are shown in Fig. 2. We note that while for the model parameters considered in Figs. 1-2, there exists a single Ωps for a given e (for a given model), as in the weak friction theory, for other model parameters where the torque is created by a primarily elastic rather than viscous response > ωM ), it is possible to find multiple spin frequencies for (ω ∼ which Tz = 0, some of which are resonant in nature, for a given e. We discuss this interesting phenomenon in the Appendix. Figures 3 and 4 present the results of the orbital evolu- tion for different viscoelastic tidal dissipation models. While all these models satisfy the same Jupiter-Io tidal Q con- straint as the weak friction theory, the predicted high-e mi- gration rate can be easily larger, by a factor of 10 or more, than that predicted by the weak friction theory. For exam- ple, while it takes ∼ 100 Gyrs to complete the orbital cir- cularization in the weak friction theory, only 10 Gyrs is needed in Model 1 and only ∼ 2 Gyrs is needed in Model 2. e 1.0 0.8 0.6 0.4 0.2 0.0 Weak friction Model 2 Model 1 Model 3 0.5 1.0 5.0 time Gyrs 10.0 50.0 100.0 Figure 4. Eccentricity as a function of time, for an initial eccen- tricity of 0.9945 and a final mean motion period of 5 days. Each blue curve corresponds to one of the blue theoretical Maxwell model curves of Figure 1. The green curve corresponds to the weak friction theory of Fig. 1. 4.2 Tidal heating of giant planets during migration Many hot Jupiters are found to have much larger radii than predictions based on "standard" gas giant theory (e.g., Baraffe, Chabrier & Barman 2010). A number of possi- ble explanations for the "radius inflation" have been sug- gested, including tidal heating (e.g., Bodenheimer, Lin & Mardling 2001, Bodenheimer, Laughlin & Lin 2003; Miller, Fortney & Jackson 2009; Ibgui et al. 2010; Leconte et al. 2010), the effect of thermal tides (Arras & Socrates 2010; Socrates 2013), enhanced envelope opacity (Burrows et al. 2007), double-diffusive envelope convection (Chabrier & Baraffe 2007; Leconte & Chabrier 2012) and Ohmic dis- sipation of planetary magnetic fields (Batygin & Steven- son 2010; Batygin, Stevenson & Bodenheimer 2011; but see Perna, Menou & Rauscher 2010; Huang & Cumming 2012; Menou 2012; Wu & Lithwick 2013; Rauscher & Menou 2013). It is possible that more than one mechanism is c(cid:13) 2013 RAS, MNRAS 000, 1 -- ?? needed to explain all of the observed radius anomalies of hot Jupiters (see Fortney & Nettelmann 2010; Spiegel & Burrows 2013). Several papers have already pointed out the potential importance of tidal heating in solving the radius anomaly puzzle (see above for references). In particular, Leconte et al. (2010) studied the combined evolutions of the planet's orbit (starting from high eccentricity) and thermal structure including tidal heating, and showed that tidal dissipation in the planet provides a substantial contribution to the planet's heat budget and can explain some of the moderately bloated hot Jupiters but not the most inflated objects (see also Miller et al. 2009; Ibgui et al. 2010). However, all these studies were based on equilibrium tide theory with a parametrized tidal quality factor Q or lag time, and assume that the heating is distributed uniformly across the planet. Here we study the heating of proto-hot-Jupiters via tidal dissipation in the core. To model this effect, we use the MESA code (Paxton et al. 2011, 2013) to evolve the internal structure of giant planets in conjunction with the orbital evolution starting from high eccentricity. We create a zero-age Jupiter-mass giant planet (initially hot and in- flated) with an inert rocky core, for which we can prescribe a time-varying luminosity. Assuming the core is in thermal equilibrium with its surroundings, we consider the core lumi- nosity to be equal to E as given by Eq. (17) (with Ωs = Ωps such that Tz = 0). We assume the planet starts at a high eccentricity of e = 0.9945 and circularizes to a 5-day or- bit, while conserving orbital angular momentum (so that aF = a(1 − e2) at all times). These assumptions enable us to calculate E(t) and observe its effect on the radius of the planet. Figure 5 presents the planet heating rate and radius vs age curves. Evidently, it is possible to inflate a proto-hot- Jupiter by up to 40% via tidal heating in the core. How- ever, this happens early in the planet's evolution, around eccentricities of 0.6, when the heating rate is largest. By the time the planet's orbit has circularized (e <∼ 0.05), its ra- dius is only ∼ 10% larger than the zero-temperature planet and continues to decline over time. Therefore, regardless of the details of the tidal models, it appears that tidal heating cannot fully explain the population of observed hot Jupiters with significant radius inflation. Nevertheless, tidal effects can significantly delay the radius contraction of gas giants. By keeping the planet somewhat inflated until (possibly) an- other effect due to proximity to the host star takes over, tidal dissipation may still play an important role in the creation of inflated hot Jupiters. Interestingly, these cooling curves suggest that if tidal dissipation in the core is indeed strong enough to play a sig- nificant role in circularizing the planet's orbit, as we have shown to be possible in this paper, we may expect to ob- serve a population of gas giants (proto-hot-Jupiters) in wide, eccentric orbits, which are nevertheless inflated more than expected (see Socrates et al. 2012a; Dawson & Murray-Clay 2013). 5 CONCLUSION The physical mechanisms for tidal dissipations in giant plan- ets are uncertain. Recent works have focused on mechanisms c(cid:13) 2013 RAS, MNRAS 000, 1 -- ?? Tidal Dissipation in Giant Planets 7 8 ´ 1027 6 ´ 1027 4 ´ 1027 2 ´ 1027 1 - s g r e t d E d J R s u i d a R 0 0 1.6 1.5 1.4 1.3 1.2 1.1 1.0 0 Model 2 Model 1 9 9 9 9 10 0.5 ae 0.75 ae e=0.9 ae ae 0.25 0 ae 0.75 ae 0.5 ae e=0.9 ae 0.25 ae Model 1 Model 2 0 ae 2 ´ 109 4 ´ 109 6 ´ 109 8 ´ 109 1 ´ 1010 age yrs Figure 5. Top: Core luminosity due to viscoelastic tidal dissipa- tion in a Jupiter-mass gas giant. The blue solid curve corresponds to Model 1 of Fig. 1, and the blue long-dashed curve corresponds to Model 2 of Fig. 1. Bottom: Evolution of radius vs time for each of the models (top), assuming an initial eccentricity of 0.9945 and a final circularized orbital period of 5 days. The red solid curve has no tidal heating and asymptotes to the zero-temperature ra- dius at later times. Black dots and labels on each curve denote when the particular planet model passes through that value of eccentricity in its orbital evolution. Note that since Model 2 is more dissipative than Model 1, the maximum heating rate and radius inflation (around e = 0.6) occur earlier in time than Model 1. of dissipation in the planet's fluid envelope, but it is not clear whether they are adequate to satisfy the constraints from the Solar system gas giants and the formation of close-in exoplanetary systems via high-e migration. In this paper, we have studied the possibility of tidal dissipation in the solid cores of giant planets. We have pre- sented a general framework by which the effects of tidal dis- sipation on the spin and orbital evolution of planetary sys- tems can be computed. This requires only one input - the imaginary part of the complex tidal Love number, Im[k2(ω)], as a function of the forcing frequency ω. We discussed the simplest model of tidal response in solids - the Maxwell vis- coelastic model, which is characterized by a transition fre- quency ωM , above which the solid responds elastically, below - viscously. Using the Maxwell model for the rocky/icy core, and including the effect of a non-dissipative fluid envelope, we have demonstrated that with a modest-sized rocky core and reasonable (but uncertain) physical core parameters, tidal dissipation in the core can account for the Jupiter-Io tidal-Q constraint (Remus et al. 2012) and at the same time allows exoplanetary hot Jupiters to form via tidal circular- ization in the high-e migration scenario. By contrast, in the often-used weak friction theory of equilibrium tide, when the tidal lag is calibrated with the Jupiter-Io constraint, hot 8 Natalia I Storch and Dong Lai Jupiters would not be able to go through high-e migration within the lifetime of their host stars. We have also examined the consequence of tidal heating in the rocky cores of giant planets. Such heating can lead to modest radius inflation of the planets, particularly when the planets are in the high-eccentricity phase (e ∼ 0.6) during their high-e migration. As an interesting by-product of our study, we have shown that when Im(k2) exhibits nontrivial dependence on ω (as opposed to the linear dependence in the weak friction theory), there may exist multiple spin frequencies at which the torque on the planet vanishes (see Appendix A). We emphasize that there remain large uncertainties in the physical properties of solid cores inside giant planets, in- cluding the size, density, composition, viscosity and elastic shear modulus. These uncertainties make it difficult to draw any definitive conclusion about the importance of core dis- sipation. Nevertheless, our study in this paper suggests that within the range of uncertainties, viscoelastic dissipation in the core is a possible mechanism of tidal dissipation in giant planets and has several desirable features when confronting the current observational constraints. Thus, core dissipation should be kept in mind as observations in the coming years provide more data on tidal dissipations in giant planets. APPENDIX A: SPIN EQUILIBRIUM/PSEUDOSYNCHRONIZATION IN VISCOELASTIC TIDAL MODELS In the weak friction theory, the tidal Love number Im(k2) is a linear function of the tidal frequency ω, and thus spin equilibrium (Tz = 0) occurs at a unique value of Ωs, termed the pseudosynchronous frequency, for a given orbital eccen- tricity e. When Im(k2) depends on ω in a more general way, as in the case of viscoelastic tidal models of giant planets, it is possible that multiple solutions for the equilibrium spin frequency Ωps exist at a given e. The reason for the existence of multiple pseudo- synchronized spins can be understood in simple algebraic terms. For clarity here we demonstrate how multiple roots arise naturally even at low eccentricities. Consider Eq. (16), which we rewrite here to make the dependence on spin fre- quency explicit: Tz = 5 4π T0 Xm,N m [W2mFmN (e)]2 Im[k2(N Ω−mΩs)]. (A1) For very low eccentricities e ≪ 1, the Hansen coefficients FmN are negligible for all except the following combinations of (m, N ): (0, 0), (0, ±1),(±2, ±2), and (±2, ±3). We can then rewrite Tz as Tz = Im[Ak2(2Ω − 2Ωs) + Bk2(3Ω − 2Ωs)], (A2) with A and B real constants. Plugging in for k2 using the Maxwell model (Eq. 32) (neglecting fluid envelope for sim- plicity), we have: Tz = ¯A (2Ω − 2Ωs) 1 + C(2Ω − 2Ωs)2 + ¯B (3Ω − 2Ωs) 1 + C(3Ω − 2Ωs)2 , (A3) where ¯A, ¯B, and C are constants. Thus, when solving for Ωs from Tz(Ωs) = 0, it is obvious that upon finding the least common denominator, we end up solving a cubic equation 100 10 1 0.1 0.01 2 kŽ m I 5 0 1 Model 4 Model 2 Weak Friction 0.001 0.001 0.01 0.1 Ω 2Πday 1 10 6 4 2 0 -2 -4 2 kŽ m I 5 0 1 Model 2 Model 4 Weak Friction -6 -1.0 -0.5 0.0 Ω 2Πday 0.5 1.0 Figure A1. Theoretical curves for the tidal Love number of a Jupiter-mass planet as a function of the tidal forcing frequency. The blue dashed curve is the same as Model 2 of Fig. 1, and the green solid curve corresponds to the weak friction model of Fig. 1. The blue solid curve is a viscoelastic Maxwell model (model 4), with Rc/Rp = 0.2 and η = 4 × 1011 bar · s. In the top panel, the models are plotted on a log-log scale, as in Fig. 1, while in the bottom panel we plot the models on a linear scale to clarify how the shape of the tidal response curve leads to resonant equilibrium spin states. for Ωs. The above discussion can be generalized to higher ec- centricities: the pseudosynchronized spin Ωps is determined by solving equations of increasingly higher (always odd) de- gree in Ωs. The top panel of Figure A2 shows the two terms on the RHS of Eq. (A3), as well as their sum, for the viscoelas- tic Model 4 of Figure A1 at an eccentricity of 0.13. This demonstrates the way in which multiple solutions for Ωps arise. Furthermore, we see that two (out of three) of the so- lutions are resonant in nature: that is, they occur, roughly, at multiples of Ω/2, where Ω is the orbital frequency. This can be understood by considering that the viscoelastic re- sponse (Figure A1) is quite sharply peaked and localized. Each term on the RHS of Eq. (A3) vanishes when Ωs = Ω and 1.5Ω, respectively. Due to the sharply peaked nature of the viscoelastic response, to which Tz is proportional, the sum of the two terms then shows resonant crossings at both of these values. This generalizes easily to the case of arbitrary eccentric- ity, where each (m, N ) harmonic of the sum for Tz (Eq. A1) vanishes when N Ω − mΩs = 0. The number and location of the resonant crossings then depends on the relative im- c(cid:13) 2013 RAS, MNRAS 000, 1 -- ?? 5 C RR z T 0 T 0.2 0.1 0.0 -0.1 -0.2 0.0 0.5 1.0 WSW 1.5 2.0 6 4 2 0 -2 5 C RR z T 0 T 0 2 4 8 10 12 6 WSW Figure A2. Tidal torque on the planet as a function of the spin frequency for different values of eccentricity and different tidal dissipation models. Top: Model 4 (solid blue) of Figure A1, for e = 0.13. The black dot-dashed and dashed curves show the (m, N ) = (±2, ±2) and (±2, ±3) terms of Eq. A3, respectively. The red curve shows their sum. The resonant features in each of the harmonics combine into three different zero-crossings in the sum. Bottom: Model 4 (solid) and Model 2 (dashed) of Figure A1. The red curves have e = 0.8, while the blue curves have e = 0.5. In order to fit all the curves on the same plot, we show 10Tz for the blue solid curve, and 0.1Tz for the red dashed curve. Equi- librium spins are determined by Tz = 0. Evidently, in the case of Model 2, the viscoelastic response is not localized enough to permit more than one resonant solution. portance of each of the harmonics; the strongest crossing is expected to occur Ωs ∼ Ωperi. This is demonstrated in Figure A2 (bottom) and Figure A3. Thus, the nonlinearity of the function Im[k2(ω)] of the viscoelastic Maxwell model is responsible for the existence of multiple pseudosynchronized spins. As shown in Figure A2 (bottom panel), there will not be multiple solutions if the viscoelastic response is not localized enough, compared with the mean motion frequency Ω (that is, the width of > Ω), and all important the resonant transition ∆ ∼ ωM ∼ harmonics of Tz are solidly on the viscous (linear) side of the Maxwell curve. On the other hand, there may exist multiple solutions when ∆ ∼ ωM ≪ Ω and the relevant tidal forcing frequencies lie on the elastic side of the Maxwell curve. Finally, we note that all the resonant zero crossings of Tz are stable (negative slope), while the non-resonant crossings are necessarily unstable (positive slope). The in- nermost and outermost crossings are always resonant. This phenomenon is analogous to that discussed by Makarov & Efroimsky (2013), who used a different model for viscoelastic c(cid:13) 2013 RAS, MNRAS 000, 1 -- ?? Tidal Dissipation in Giant Planets 9 5 C R R z T 0 T 6 4 2 0 -2 -4 0.6 0.8 1.0 WSWperi 1.2 1.4 Figure A3. Tidal torque on the planet as a function of the ratio of spin frequency to pericenter frequency for e = 0.8 and differ- ent tidal dissipation models. Red solid: Model 4 of Figure A1; red dashed: 0.1Tz for Model 2 of Figure A1; green solid: Weak friction model of Figure A1. Note that the equilibrium spin frequencies of the viscoelastic models can differ from the weak friction pseu- dosynchronous spin by as much as ∼ 10 − 20%. dissipation in solid bodies to analyze the pseudosynchroniza- tion of telluric planets. They demonstrated the presence of multiple equilibrium spin solutions, and showed that only the resonant solutions are stable equilibria, thus conclud- ing that the telluric planets possess no true (non-resonant) pseudosynchronous state. The implications of our finding may be of practical in- terest when it becomes possible to measure the spin of exo- planets on eccentric orbits. We may then look for evidence of the existence of multiple stable spin equilibria in rocky planets or gas giants with rocky cores. ACKNOWLEDGEMENTS We thank M. Efroimsky, M.-H. Lee, J. Lunine, P. Nichol- son and Y. Wu for discussions and information. This work has been supported in part by NSF grants AST-1008245, 1211061 and NASA grant NNX12AF85G. REFERENCES Alexander M.E., 1973, Astrophys. Space Sci., 23, 459 Arras P., Socrates A., 2010, ApJ, 714, 1 Baraffe I., Chabrier G., Barman T., 2010, Rep. Prog. Phys., 73, 016901 Batygin K., Stevenson D. J., 2010, ApJ, 714, L238 Batygin K., Stevenson D.J., Bodenheimer P.H., 2011, ApJ, 738, 1 Biot M.A., 1954, J. Appl. Phys., 25, 1385 Bodenheimer P., Lin D. N. C., Mardling R. A., 2001, ApJ, 548, 466 Bodenheimer P., Laughlin G., Lin D. N. C., 2003, ApJ, 592, 555 Burrows A., Hubeny I., Budaj J., Hubbard W.B., 2007, ApJ, 661, 502 Chabrier G., Baraffe I., 2007, ApJ, 661, L81 Chatterjee S., Ford E. B., Matsumura S., Rasio F. A., 2008, ApJ, 686, 580 10 Natalia I Storch and Dong Lai Darwin G.H., 1879, Phil. Trans. R. Soc., 170, 1 Dawson R.I., Murray-Clay R.A., 2013, ApJ, 767, L24 Dermott S.F., 1979, Icarus, 37, 310 Eggleton P.P., Kiseleva L.G., Hut P., 1998, ApJ, 499, 853 Efroimsky M., Makarov V.V., 2013, ApJ, 764, 26 Fabrycky D., Tremaine S., 2007, ApJ, 699, 1298 Fortney J.J., Nettelmann N., 2010, Space Sci. Rev., 152, 423 Gavrilov S.V., Zharkov V.N., 1977, Icarus, 32, 443 Goldreich P., Nicholson P.D., 1977, Icarus, 30, 301 Goldreich P., Soter S., 1966, Icarus, 5, 375 Goldsby D. L., Kohlstedt D. L., 2001, J. Geophys. Res.,106, 11017 Goodman J., Lackner C., 2009, ApJ, 696, 2054 Guillot T., 2005, Ann. Rev. Earth Planet. Sci., 33, 493 Hansen B., 2012, ApJ, 757, 6 Henning W.G., O'Connell R.J., Sasselov D.D., 2009, ApJ, 707, 1000 Huang X., Cumming A., 2012, ApJ, 757, 47 Hut P., 1981, A&A, 99, 126 Ibgui L., Burrows A., Spiegel D.S., 2010, ApJ, 713, 751 Ioannou P.J., Lindzen R.S., 1993a, ApJ, 406, 252 Ioannou P.J., Lindzen R.S.,1993b, ApJ, 406, 266 Ivanov P.B., Papaloizou J.C.B., 2007, MNRAS, 376, 682 Jeanloz R., 1990, Annu. Rev. Earth Planet. Sci. 18, 357 Juric M., Tremaine S., 2008, ApJ, 686, 603 Katz B., Dong S., Malhotra R., 2011, Phys. Rev. Lett., 107, 181101 Lai D., 2012, MNRAS, 423, 486 Lainey V., Arlot J.-E., Karatekin O., van Hoolst T., 2009, Nature, 459, 957 Lainey V. et al., 2012, ApJ, 752, 14 Leconte J., Chabrier G., Baraffe I., Levrard B., 2010, A&A, 516, A64 Leconte J., Chabrier G., 2012, A&A, 540, A20 Love A., 1927, A Treatise on the Mathematical Theory of Elasticity (Dover, NY), p. 259 Lubow S.H., Tout C.A., Livio M., 1997, ApJ, 484, 866 Makarov V.V., Efroimsky M., 2013, ApJ, 764, 27 Mathis S., Le Poncin-Lafitte C., 2009, A&A, 497, 889 Matsumura S., Peale S. J., Rasio F. A., 2010, ApJ, 725, 1995 Menou K., 2012, ApJ, 745, 138 Militzer B., Hubbard W. B., Vorberger J., Tamblyn I., Bonev S. A., 2008, ApJ, 688, L45 Miller N., Fortney J.J., Jackson B., 2009, ApJ, 702, 1413 Mirouh G.M., Garaud P., Stellmach S., Traxler A. L., Wood T. S., 2012, ApJ, 750, 61 Mitrovica J.X., Forte, A.M., 2004, Earth Planet. Sci. Lett., 225, 177 Montagner J.P., Kennett, B.L.N., 1996, Geophys. J. Int. 125, 229 Murray C.D., Dermott, S.F., 2000, Solar System Dynamics. Cambridge Univ. Press, Cambridge. Nagasawa M., Ida S., Bessho,T., 2008, ApJ, 678, 498 Naoz S., Farr W. M., Lithwick Y., Rasio F. A., Teyssandier J., 2011, Nature, 473, 187 Naoz S., Farr W.M., Rasio F.A., 2012, ApJ, 754, 36 Naoz S., Farr W. M., Lithwick Y., Rasio F. A., Teyssandier J., 2013, MNRAS, 431, 2155 Ogilvie G.I., 2009, MNRAS, 396, 794 Ogilvie G.I., 2013, MNRAS, 429, 613 Ogilvie G.I., Lin D.N.C., 2004, ApJ, 610, 477 Papaloizou J.C.B., Ivanov P.B., 2010, MNRAS, 407, 1631 Paxton B., et al., 2011, ApJS, 192, 3 Paxton B., et al., 2013, ApJS, 208, 4 Peale S.J., 1999, ARAA, 37, 533 Perna R., Menou K., Rauscher E., 2010, ApJ, 724, 313 Poirier J. P., Sotin C., Peyronneau J., 1981, Nature, 292, 225 Rasio F.A., Ford E.B., 1996, Science, 274, 954 Rauscher E., Menou K., 2013, ApJ, 764, 103 Remus F., Mathis S., Zahn J.-P., Lainey V., 2012a, A&A, 541, 165 Remus F., Mathis S., Zahn J.-P., 2012b, A&A, 544, 132 Sinclair A.T., 1983, in Dynamical Trapping and Evolution in the Solar System, ed. V.V. Markellos & Y. Kozai (D Reidel), p. 19 Socrates A., 2013, arXiv:1304.4121 Socrates A., Katz B., Dong S., Tremaine S., 2012a, ApJ, 750, 106 Socrates A., Katz B., Dong S., (arXiv:1209.5724) Spiegel D.S., Burrows A., 2013, ApJ, 772, 76 Turcotte D.L., Schubert G., 2002, Geodynamics. Cam- bridge Univ. Press, Cambridge Weidenschilling S.J., Marzari F., 1996, Nature, 384, 619 Wu Y., 2005, ApJ, 635, 688 Wu Y., Lithwick Y., 2011, ApJ, 735, 109 Wu Y., Lithwick Y., 2013, ApJ, 763, 13 Wu Y., Murray N.W., 2003, ApJ, 589, 605 Wu Y., Murray N.W., Ramshhai J.M., 2007, ApJ, 670, 820 Yoder C.F., Peale S.J., 1981, Icarus, 47, 1 Zhou J.-L., Lin D.N.C., Sun Y.-S., 2007, ApJ, 666, 423 2012b, preprint c(cid:13) 2013 RAS, MNRAS 000, 1 -- ??
1805.08774	1	1805	2018-05-22T17:59:21	Expected spectral characteristics of (101955) Bennu and (162173) Ryugu, targets of the OSIRIS-REx and Hayabusa2 missions	[ "astro-ph.EP" ]	NASA's OSIRIS-REx and JAXA's Hayabusa2 sample-return missions are currently on their way to encounter primitive near-Earth asteroids (101955) Bennu and (162173) Ryugu, respectively. Spectral and dynamical evidence indicates that these near-Earth asteroids originated in the inner part of the main belt. There are several primitive collisional families in this region, and both these asteroids are most likely to have originated in the Polana-Eulalia family complex. We present the expected spectral characteristics of both targets based on our studies of our primitive collisional families in the inner belt: Polana-Eulalia, Erigone, Sulamitis, and Clarissa. Observations were obtained in the framework of our PRIMitive Asteroids Spectroscopic Survey (PRIMASS). Our results are especially relevant to the planning and interpretation of in-situ images and spectra to be obtained by the two spacecraft during the encounters with their targets.	astro-ph.EP	astro-ph	Pre-print submitted to Icarus - Accepted Expected spectral characteristics of (101955) Bennu and (162173) Ryugu, targets of the OSIRIS-REx and Hayabusa2 missions J. de León(1,2)1, H. Campins(3), D. Morate(1,2), M. De Prá(4), V. Alí-Lagoa(5), J. Licandro(1,2), J. L. Rizos(1,2), N. Pinilla-Alonso(6), D. N. DellaGiustina(7), D. S. Lauretta(7), M. Popescu(1,2), V. Lorenzi(9,1) (1) Instituto de Astrofísica de Canarias, C/Vía Láctea s/n, E-38205 La Laguna, Tenerife, Spain (2) Departamento de Astrofísica, Universidad de La Laguna, E-38206 La Laguna, Tenerife, Spain (3) Physics Department, University of Central Florida, P.P. box 162385, Orlando, FL 32816-2385, USA (4) Departamento de Astrofísica, Observatorio Nacional, Rio de Janeiro, 20921-400, Brazil (5) Max-Planck-Institut für extraterrestrische Physik, Giessenbachstrasse 1, 85748 Garching, Germany (6) Florida Space Institute, University of Central Florida, Orlando, FL 32816, USA (7) Lunar and Planetary Laboratory, University of Arizona, Tucson, AZ, USA (8) Fundación Galileo Galilei – INAF, Rambla José Ana Fernández Pérez, 7, E-38712 Breña Baja, La Palma, Spain Abstract NASA's OSIRIS-REx and JAXA's Hayabusa2 sample-return missions are currently on their way to encounter primitive near-Earth asteroids (101955) Bennu and (162173) Ryugu, respectively. Spectral and dynamical evidence indicates that these near-Earth asteroids originated in the inner part of the main belt. There are several primitive collisional families in this region, and both these asteroids are most likely to have originated in the Polana-Eulalia family complex. We present the expected spectral characteristics of both targets based on our studies of four primitive collisional families in the inner belt: Polana-Eulalia, Erigone, Sulamitis, and Clarissa. Observations were obtained framework of our PRIMitive Asteroids Spectroscopic Survey (PRIMASS). Our results are especially relevant to the planning and interpretation of in situ images and spectra to be obtained by the two spacecraft during the encounters with their targets. Keywords: Near-Earth objects; Asteroids, surfaces; Spectroscopy 1. Introduction Near-Earth asteroids (NEAs) are among the most interesting populations of minor bodies of the Solar System: their proximity to the Earth makes them impact hazards and also accessible to spacecraft. Consequently, they can be studied in detail to mitigate potential impacts and also sample their surface material for analysis in terrestrial laboratories. Also, NEAs are considered ideal targets for In-Situ Resource the in 1 e-mail: [email protected] 1 Pre-print submitted to Icarus - Accepted Utilization (ISRU) and could become a source of materials for space activities in the near future. Among NEAs, those with a primitive composition are of particular interest, as they might contain water and organic compounds, being the remnants of the formative stages of our Solar System and thereby providing information on the early conditions of the solar nebula. Polana-Eulalia! Klio! Erigone! Svea! Sulamitis! Clarissa! Chaldaea! Chimaera! 0.30! 0.25! 0.20! 0.15! 0.10! ! y t i c i r t n e c c E 0.05! 2.00! 2.10! 2.20! 2.40! Semi-major axis (au)! 2.30! 2.50! 2.60! Polana-Eulalia! Klio! Erigone! Svea! Sulamitis! Clarissa! Chaldaea! Chimaera! 0.30! 0.25! 0.20! 0.15! 0.10! 0.05! 0.00! 2.00! 2.10! 2.20! 2.40! Semi-major axis (au)! 2.30! 2.50! 2.60! n o i t a n i l c n ! i f o e n S i Figure 1. Proper semi-major axis vs. proper eccentricity (upper panel) and sine of proper inclination (bottom panel) for the currently identified eight primitive collisional families located in the inner belt, according to Nesvorný et al. (2015). For the Polana-Eulalia family complex we have used the definition from Walsh et al. (2013). NEA lifetimes are short compared to the age of the Solar System (Morbidelli et al. 2002), and there must exist a replenishment mechanism to sustain the population. Dynamical models indicate that most of the NEAs come from the main asteroid belt (Bottke et al. 2002), in particular from the region enclosed by the ν6 secular resonance with Jupiter and Saturn, located at 2.1 au, and the 3:1 mean motion resonance with Jupiter, at 2.5 au. We refer to this region as the inner main belt. The process that best describes how NEAs reach near-Earth space can be separated into two steps. First, collisions in the main belt generate small fragments in the size range of NEAs, i.e., meters to a few kilometers. Second, the action of the Yarkovsky effect, which is more 2 Pre-print submitted to Icarus - Accepted efficient for small-diameter objects, modifies the semi-major axes of the orbits of these fragments until they reach one of the transport routes above mentioned (Morbidelli & Vokrouhlický 2003; Bottke et al. 2006). Therefore, collisional families in the inner main belt are considered the most likely source of NEAs. This rationale also applies to primitive NEAs that are targets of space missions. As of early 2018, there are eight primitive collisional families identified in this region (Nesvorný et al. 2015). These are, the Polana-Eulalia complex, and the Erigone, Sulamitis, Clarissa, Klio, Chaldaea, Svea, and Chimaera collisional families (Fig. 1). There are currently two sample-return missions on their way to two primitive NEAs: NASA's OSIRIS-REx mission to asteroid (101955) Bennu (Lauretta et al. 2017) and JAXA's Hayabusa2 mission to asteroid (162173) Ryugu (Tsuda et al. 2013). Asteroid Bennu is classified as a B-type asteroid from its visible spectrum (Clark et al. 2011), and its most likely origin is the Polana-Eulalia family complex (Campins et al. 2010; Walsh et al. 2013; Bottke et al. 2015). Likewise, Ryugu, a C-type asteroid, most likely originated in the Polana-Eulalia complex or in the population of low-albedo and low- inclination background asteroids (Campins et al. 2013), identified for the first time by Gayon-Markt et al. (2012). With the aim of enhancing the science return of both missions, we started our PRIMitive Asteroids Spectroscopic Survey (PRIMASS) in 2010. As part of the PRIMASS project, we obtain visible and near-infrared spectra of the members of the collisional families and dynamical groups of the main asteroid belt, as well as other populations of primitive objects (De Prá et al. 2018). As of 2018, we have characterized four primitive families in the inner belt: Polana-Eulalia, Erigone, Sulamitis, and Clarissa. We are in the process of data reduction and analysis of the visible spectra of members of the rest of the primitive families in the inner belt (Klio, Chaldaea, Svea, and Chimaera). Our results on the spectral characterization of the members of the Polana-Eulalia family complex both in the visible (de León et al. 2016) and the near-infrared (Pinilla- Alonso et al. 2016) provide essential context for this study. Despite the dynamical complexity of this region (Walsh et al. 2013; Milani et al. 2014; Dykhuis & Greenberg 2015), asteroids belonging to this complex show spectral homogeneity, with a continuum of spectral slopes from blue to moderately red, typical of B- and C-type objects. In contrast, Morate et al. (2016) showed that the Erigone collisional family shows a more extensive spectral diversity, with B-, C-, X-, and T-type objects. Also, the majority of the Erigone family members have an absorption band at 0.7 μm, as does the parent body (163) Erigone; this absorption is associated with aqueously altered silicates, i.e., phyllosilicates (Vilas 1994). Neither (142) Polana nor (495) Eulalia or their family members show this band. Finally, our most recent results on the Sulamitis and the Clarissa families suggest that there might be a connection between Polana-Eulalia and Clarissa (both families with no hydration signatures and presenting mainly B- and C-type asteroids). Our work revealed a similar association between Erigone and Sulamitis (both families showing the 0.7 μm feature and a similar fraction of B-, C-, X-, and T-type asteroids). This tentative connection has been presented in Morate et al. (2018) and will be further explored in this work. Visible spectra of the parent bodies of the collisional families, as well as their members are presented in Section 2, together with a summary of the results obtained for the spectral analysis of each family. In Section 3 we compare the available visible spectra of both Bennu and Ryugu with the spectra of the B- and C-type asteroids found in the Polana-Eulalia complex and also in the four collisional families of primitive asteroids of the inner belt studied in this work. In the case of Bennu, we also compute (b'-v) and (v- x) colors using the response curves of the MapCam filters (one of the three scientific 3 Pre-print submitted to Icarus - Accepted cameras on-board OSIRIS-REx: Rizk et al. 2018) and the mean spectra of the different taxonomies mentioned before. Section 4 presents a discussion of the obtained results, and conclusions are presented in Section 5. 2. Collisional families of primitive asteroids in the inner belt The primary aim of this work is to make a comparative study between the ground- based visible spectra of the primitive NEAs (101955) Bennu and (163172) Ryugu, and the visible spectra of the primitive collisional families in the inner belt, which are considered as the most likely source region of these NEAs. By studying the spectral properties of the members of these families, we can constrain the expected global spectral properties of the surface of the targets before the arrival of the spacecraft. In this section, we study the spectral characteristics of the parent bodies and the family members. 2.1. Parent bodies In Fig. 2 we present the available visible spectra of the parent bodies of the primitive collisional families studied in this work: asteroids (142) Polana, (495) Eulalia, (163) Erigone, (752) Sulamitis, and (306) Clarissa. For those asteroids having more than one 1.35! 1.25! (762) Sulamitis! 1.15! (163) Erigone! (302) Clarissa! (495) Eulalia! (142) Polana! 1.05! 0.95! 0.85! 0.75! ! e c n a t c e fl e R e v i t a l e R 0.65! 0.45! 0.50! 0.55! 0.60! 0.70! 0.65! 0.75! Wavelength (µm)! 0.80! 0.85! 0.90! 0.95! Figure 2. Visible spectra of the parent bodies of the primitive collisional families studied in this work. The spectra are normalized to unity at 0.55 μm and offset vertically for clarity. Asteroids have been ordered according to their spectral slope, from the reddest object at the top (Sulamitis) to the bluest asteroid at the bottom (Polana). See the main text for details on individual spectra. published visible spectrum we compute the average spectrum and the corresponding ±1σ deviation of the mean (shown as error bars). Table 1 provides a summary of the observational circumstances for each parent body, including the observation date, the phase angle (α) at the time of observation, the telescope used, and the corresponding bibliographic reference. In a separate table (Table 2) we show some physical properties of these asteroids, including size and visible geometric albedo (pV). The table also includes the computed values for the spectral slope S'. This slope has been computed in the range from 0.55 to 0.88 μm (this is the common wavelength interval to 4 Pre-print submitted to Icarus - Accepted (142) Polana (495) Eulalia Parent body (752) Sulamitis Phase angle (°) (302) Clarissa (163) Erigone 8.1 14.5 23.1 27.9 18.3 9.8 13.7 22.6 10.4 9.1 17.4 23.9 19.6 Telescope 1.3m McG-H 1.5m CTIO 1.52m ESO 3.6m NTT 1.5m CTIO Obs. Date 25/04/1996 19/04/1992 16/04/1999 20/07/2011 06/07/1987 10/09/1992 05/05/2014 31/08/2016 21/04/1996 29/11/1992 26/01/2001 21/07/2015 31/08/2015 Ref. B02 V92 F14 dL16 V92 X95 2.4m H dL16 3.6m TNG 10.4m GTC M17 B02 1.3m McG-H X95 1.52m ESO L04 2.5m INT M17 10.4m GTC M17 Telescopes: McG-H – McGraw-Hill (Kitt Peak, Arizona); CTIO – Cerro Tololo Inter- american Observatory (La Serena, Chile); ESO – European Southern Observatory (La Silla, Chile); H – Hiltner (Kitt Peak, Arizona); INT – Isaac Newton Telescope (El Roque de Los Muchachos, Spain); TNG – Telescopio Nazionale Galileo (El Roque de los Muchachos, Spain); NTT – New Technology Telescope (La Silla, Chile); GTC – Gran Telescopio Canarias (El Roque de Los Muchachos, Spain). Ref: B02 – Bus & Binzel (2002); dL16 – de León et al. (2016); F14 – Fornasier et al. (2014); L04 – Lazzaro et al. (2004); M17 – Morate et al. (2018); V92 – Vilas & McFadden (1992); X95 – Xu et al. (1995) the spectra of the five asteroids), using the expression as defined by Luu & Jewitt (1990), S' = (dS/dλ)/S0.55, where dS/dλ is the variation of the reflectance in the selected wavelength range, and S0.55 is the reflectance at 0.55 μm. For those asteroids having more than one published spectrum we see no correlation between phase angle and spectral slope (these are 142 Polana, 495 Eulalia, and 752 Sulamitis). Table 1. Summary of the observational circumstances of the parent bodies of the primitive families studied in this work. The slope is computed by a simple linear fit in the 0.55-0.88 μm range, in units of %/1000 Å. Typically, the most significant contribution to the error in the spectral slope comes from the process of dividing the spectrum of the asteroid by the spectra of solar analog stars. This error is usually lower that 1.0 %/1000 Å and is typically of the order of 0.5 %/1000 Å. As we do not have access to this information for all the targets, we show in Table 2 the error associated with the slope computation process: we perform 100 iterations, randomly removing 10% of the data points and doing a linear fitting on each iteration. The resulting slope is the mean of these 100 iterations, and the error is the standard deviation of this mean. Table 2. Physical properties of the parent bodies of the primitive families studied in this work. 2.4m H S' Parent body Size1 (km) pV 1 (%/1000Å) -1.17 ± 0.01 (142) Polana -0.11 ± 0.02 (495) Eulalia 1.00 ± 0.02 (302) Clarissa 1.25 ± 0.02 (163) Erigone (752) Sulamitis 2.17 ± 0.02 1 Average diameter computed using the NEATM model from Alí-Lagoa et al. (2018) to fit NEOWISE and AKARI data. The corresponding visible geometric albedos used the H-G12 values from Oszkiewicz et al. (2012). 58 ± 6 38 ± 4 39 ± 4 76 ± 7 61 ± 6 0.044 0.054 0.046 0.041 0.040 5 Pre-print submitted to Icarus - Accepted We see from the spectra in Fig. 2 and the slopes in Table 2 that there is a significant variation in the spectral properties of these parent bodies, even if they are all considered primitive asteroids. Their visible spectra range from featureless and blue sloped (Polana) to neutral colors (Clarissa) and red sloped (Sulamitis and Erigone) with a clear absorption band at 0.7 μm associated with phyllosilicates and indicative of aqueous alteration processes. Interestingly, we also observe a wide range in the estimated ages of their corresponding collisional families (Table 3). Also, the observed spectral diversity among the parent bodies is even more extensive in that of the members of these families; we discuss this point in detail in the next section. 2.2. Family members We have spectrally characterized four out of the eight primitive collisional families in the inner belt: Polana-Eulalia, Erigone, Sulamitis, and Clarissa. We have computed the spectral slopes not only for the parent bodies (Table 2) but also for all the members observed within the families. The results can be found in de León et al. (2016) and Morate et al. (2016, 2018). In this section we summarize the main results obtained from this spectral characterization, describing each family separately. 1. Polana-Eulalia. Visible spectra of a total of 65 members of this complex (about 2 % of the family) were obtained using the 10.4m Gran Telescopio Canarias (GTC) and the 3.56m Telescopio Nazionale Galileo (TNG), both located at the El Roque de Los Muchachos Observatory, in the island of La Palma (Spain), and the 3.6m New Technology Telescope (NTT), located at La Silla Observatory (Chile). The results obtained from the spectral analysis of these asteroids are summarized in Fig. 3a and published by de León et al. (2016). Despite the apparent dynamical complexity of this particular region of the asteroid belt, we found no spectral differences between the members of the so-called New Polana and the Eulalia families (Walsh et al. 2013), neither in the visible nor the near infrared (Pinilla-Alonso et al. 2016). An almost equal proportion of B- and C-type asteroids populates this complex, i.e., the family is characterized by asteroids presenting featureless spectra, with slopes ranging in a continuum from blue (-2.9 %/1000Å) to moderately red (2.7 %/1000Å), and with no signs of aqueous alteration. This group is by far the largest primitive collisional family in the inner belt and is also the oldest. 2. Erigone. A total of 101 members of this collisional family were observed using the 10.4m Gran Telescopio Canarias, and the results were published by Morate et al. (2016). This set accounts for about 6% of the family, which has 1776 members. Contrary to what is observed in the Polana-Eulalia complex, the Erigone collisional family shows a broader diversity of primitive spectral classes, as is shown in Fig. 3b. The obtained spectral slopes range from blue (-2.8 %/1000Å) to significantly red (7.2 %/1000Å). Perhaps the most remarkable result is the high abundance of asteroids (more than 50%) showing the 0.7 μm absorption band produced by aqueous alteration of silicates, in contrast to the lack of such band found in the Polana-Eulalia complex. The parent body of this family, asteroid (163) Erigone, also presents this hydration feature in its visible spectrum (see Fig. 2). Although the majority (75%) of the asteroids with this absorption band are C-types, as expected, we also find signs of hydration among X-types (15%), B-types (8%), and T-types (2%). 3. Sulamitis. We obtained visible spectra of a total of 64 members of the Sulamitis family, accounting for 21% of the total number, using the 10.4m GTC (Morate et al. 2018). As in the case of Erigone, the members of the family showed a significant 6 Pre-print submitted to Icarus - Accepted diversity of primitive spectral classes (see Fig. 3d), distributed in a similar proportion, except B-type asteroids (there is only one object in the sample). Similarly, (752) Sulamitis, like Erigone, presents the 0.7 μm absorption feature in its visible spectrum, and about 60% of the members of the family show this hydration band. Spectral slopes in this family range from moderately blue (-1.6 %/1000Å) to considerably red (8.3 %/1000Å), again similar to what we see in the Erigone family. Also, the two families have comparable ages (see Table 2). 4. Clarissa. This group is the smallest family of the ones we present in this work. It has 179 members, and we have obtained visible spectra of a total of 33 (18% of the family) using the 10.4m GTC (Morate et al. 2018). The Clarissa family shares some of the spectral properties of the Polana-Eulalia complex: it is mostly composed of featureless B- and C-type asteroids, although it presents a more significant proportion of redder X- types. Spectral slopes range from blue (-4.4 %/1000Å) to moderately red (4.7 %/1000Å). Interestingly, as in the case of the Polana-Eulalia complex, we find almost no asteroids (3 out of 33) showing the 0.7 μm absorption band (Fig.3c). Figure 3. Summary of the results obtained from the spectral characterization of the inner belt primitive families studied so far. Each panel contains: the distribution of the taxonomical classes found within the family (percentage); the mean spectrum obtained from the averaging of all the spectra from each class in the family; the corresponding ±1σ deviation from the mean for each class, plotted as error bars at the bottom of each panel. For all the families blue is for B-types, red is for C-types, orange is for X-types and green is for T-types. Spectra are all normalized to unity at 0.55 μm. In the case of the Sulamitis family, there is only one B-type object, asteroid (122109). Figs. 4 and 5 summarize the spectral characteristics of the primitive families we have analyzed so far. Fig. 4 shows the distributions of the spectral slopes S' computed for each family. The vertical lines correspond to the mean slope of each family, shown in Table 3. A simple visual inspection reveals the similarity between the Polana-Eulalia complex and the Clarissa family on one side, and between the Erigone and the Sulamitis families on the other. These similarities were quantified and confirmed by Morate et al. (2018) using a two-sample Kolmogorov-Smirnov test over the two pairs of distributions. In Fig. 5 we have plotted together the distribution (pie charts) of the taxonomical classes of each family. 7 Pre-print submitted to Icarus - Accepted Polana-Eulalia Clarissa Erigone Sulamitis 30 25 20 15 10 5 i s d o r e t s A f o r e b m u N 0 -6 -4 -2 0 2 Spectral Slope (%/1000 A) 4 6 8 10 Figure 4. Spectral slope distributions of the four inner primitive families studied in this work. Details for each family can be found in de León et al. (2016) for the Polana-Eulalia complex, Morate et al. (2016) for the Erigone family, and Morate et al. (2018) for the Sulamitis and the Clarissa families. Polana-Eulalia! 4%! Clarissa! 16%! Erigone! 8%! 12%! 44%! 52%! 36%! 48%! 48%! 32%! 56%! Sulamitis! 2%! 12%! 30%! X-types (4%)! C-types (52%)! B-types (44%)! X-types (16%)! C-types (48%)! B-types (36%)! T-types (8%)! X-types (32%)! C-types (48%)! B-types (12%)! T-types (12%)! X-types (30%)! C-types (56%)! B-types (2%)! No 0.7 μm absorption band 0.7 μm absorption band! Figure 5. Comparison of the distributions of taxonomic classes found for each family. We have grouped the Polana-Eulalia complex and the Clarissa family (left), and the Erigone and Sulamitis families (right) to enhance the similarities between each pair. We have also computed the average geometric albedo for each family, using WISE/NEOWISE data (Wright et al. 2010; Mainzer et al. 2011), the NEATM model implemented by Alí-Lagoa et al. (2017), and values for H and G12 from Oszkiewicz et al. (2012). These average albedos and their associated errors are shown in Table 3, together with the number of asteroids used to compute the average for each family (sample size). We also include in the last column of Table 3 a flag for the quality criteria (Q) as defined by Alí-Lagoa et al. (2016), indicating whether the sample includes only those fits having at least 10 data points in both W3 and W4 filters2 (labeled "Yes" or "No"). The first requirement ensures a reasonable rotational sampling and averaging out of the irregularities in both thermally dominated bands is a requisite 2 The NASA Wide-field Infrared Survey Explorer (WISE) mission is equipped with four filters, W1, W2, W3, and W4 centered at 3.4, 4.6, 12, and 22 µm, respectively. 8 Pre-print submitted to Icarus - Accepted to fit the beaming parameter and improve the inferred diameters (Harris 1998). As expected we find average albedo values for the families in good agreement with the albedo values of the parent bodies. Table 3. Properties of the four primitive collisional families studied in this work. Family #1 Age2 (Myr) <S'> (%/1000Å) <pV>3 Sample size Q4 Polana-Eulalia 3783* Clarissa Erigone Sulamitis 179 1776 303 1400 ± 150 +370 830-100 ~60 130 ± 30 200 ± 40 +0,010 -0.43 ± 1.17 0.052-0,009 +0,020 0.14 ± 2.09 0.042-0,012 +0,020 1.81 ± 2.02 0.047-0,013 +0,010 2.12 ± 1.83 0.053-0,009 615 Yes 19 212 50 No Yes Yes 1 From Nesvorny et al. (2015). 2 From Bottke et al. (2015). The two ages correspond to the so-called "new Polana" (top) and Eulalia (bottom) families, as described in the paper. 3 Albedo values computed from the NEATM diameters of Alí-Lagoa et al. (2017) to fit WISE/NEOWISE thermal infrared data and the H and G12 values from Oszkiewicz et al. (2012). 4 Q: applied quality criteria based on the number of available data per band defined in Alí-Lagoa et al. (2016). * K. Walsh, personal communication. From the overall spectral characteristics shown in Figs. 3-5, we can roughly differentiate two groups in the inner belt primitive families studied so far: the Polana- like group and the Erigone-like group. The Polana-like family members present featureless (no 0.7 μm absorption band), homogeneous spectra ranging from slightly blue to moderately red. The Erigone-like group contains members showing a more extensive spectral diversity and, in their majority, the 0.7 μm band associated with phyllosilicates. While Erigone and Sulamitis families have similar ages and mean albedo values, the Clarissa family is much younger than the Polana-Eulalia complex (see Table 3). Also, we find a more substantial fraction of red, X-type asteroids in the Clarissa family (Fig. 5). Recent laboratory experiments to simulate space weathering effects on low-albedo, primitive materials suggest that their visible spectra tend to get bluer and their albedo tend to get higher as exposure age increases (Lantz et al. 2015, 2017). This result can explain the fact that the younger Clarissa family presents a lower fraction of B-types (blue) and a larger fraction of X-types (red) than the much older Polana-Eulalia complex (Campins et al. 2018). Regarding the albedo, we cannot establish any conclusion, as the number of asteroids with good-quality albedo values in the Clarissa family is too low. 3. Comparison with the spectra of Bennu and Ryugu Primitive near-Earth asteroids (101955) Bennu and (163172) Ryugu are the targets of NASA's OSIRIS-REx and JAXA's Hayabusa2 sample-return missions, respectively. These two missions are currently in space and are expected to arrive at their targets during the second half of 2018. In both cases, our results are especially relevant to the planning and interpretation of in situ images and spectra to be obtained by the two spacecraft. Spectra and color images will not only provide information on the visible spectral characteristics of the surface of the asteroid but will be used to identify and select the site for the sample collection. Therefore, it is essential to have as much information as possible on what to expect regarding colors in the visible wavelength 9 Pre-print submitted to Icarus - Accepted region of the surface of the two asteroids before the arrival. This is only possible from the examination of ground-based spectra of the two targets, and also from the visible spectra of the asteroids located in the inner belt, their most likely source region. 3.1. (101955) Bennu: target of the NASA OSIRIS-REx mission Primitive NEA (101955) Bennu (previously known by its provisional designation 1999 RQ36) is the target of the NASA's OSIRIS-REx sample-return mission (Lauretta et al. 2015; Lauretta et al. 2017). This asteroid has a diameter of about 500 m, a surface with an overall blue color (B-type asteroid) and a low albedo (pV = 0.04). It has been observed only two times in the visible wavelength region using ground-based Range of B-type spectra from inner belt primitive families 1.10 1.05 1.00 0.95 0.90 e c n a t c e l f e R e v i t l a e R Clark et al. (2011) Hergenrother et al. (2013) (101955) Bennu +1σ -1σ 0.9 0.4 0.5 0.6 0.7 Wavelength (µm) 0.8 Figure 6. Visible spectrum and colors for primitive NEA Bennu, target of the NASA OSIRIS- REx mission. We show here the mean spectrum of all the B-type asteroids found among the inner belt primitive families, as well as the corresponding ±1σ deviation of the mean. telescopes. The spectrum shown in Fig. 6 corresponds to the average of the visible spectra obtained over five consecutive nights between 15 and 20 September 1999, during one of Bennu's close approaches to the Earth. Observations span a range of phase angles from 35.9° to 65.0° and visual magnitudes from 16.6 to 14.4. Visible colors (black squares in Fig. 6) were obtained using ubvwxp ECAS filters on September 14-17 2005 (Hergenrother et al. 2013). Both B-type and F-type asteroids show a turnover in reflectance between 0.4 and 0.5 μm, with the B-types presenting a larger decrease in reflectance short ward 0.5 μm. This is due to absorption in the ultra- violet (UV) wavelength region associated with the presence of aqueously altered minerals. The ECAS colors of Bennu do not show this UV drop-off in reflectance, and the only published visible spectrum of Bennu starts at 0.47 μm, so we cannot detect this drop-off. Interestingly, neither the spectrum nor the colors show the 0.7 μm absorption band associated with hydrated silicates (phyllosilicates). This was first noted by Hergenrother et al. (2013) and later studied by de León et al. (2016). There are several published spectra of Bennu in the near-infrared wavelength region (0.8-2.4 μm) that range from negative, blue spectral slope (-0.30 %/1000Å) to positive, red slope (1.40 %/1000Å), i.e., from a B-type to a C-type according to the DeMeo et al. (2009) taxonomical classification. This variation is compatible with having a negative, B-type slope in the visible, as shown by de León et al. (2012), and it was thoroughly studied by Binzel et al. (2015). The authors found no correlation between near-infrared slope and any systematic observational effect (including phase angle), and so, they propose 10 Pre-print submitted to Icarus - Accepted an alternative explanation related to the accumulation of finer grained material in an equatorial ridge created by regolith migration during episodes of rapid rotation. We showed in Fig. 8 from de León et al. (2016) that the visible spectrum of Bennu (S' = -1.28 ± 0.03 %/1000Å) is almost identical to the visible spectrum of (142) Polana. Bennu also lies within the boundaries defined by the ±1σ of the mean spectrum of the Polana-Eulalia family members (S'= -0.43 ± 1.17 %/1000Å). Here we have computed the mean spectrum of all the asteroids classified as B-types in the four primitive families studied in this work. The resulting mean spectrum and the corresponding ±1σ deviation from the mean are shown in Fig. 6 (orange), together with the spectrum and the colors of Bennu. As expected, the agreement is excellent. Regarding the origin of Bennu, spectroscopic and dynamical arguments suggest that it most likely originated in the primitive collisional families of the inner belt, in particular in the Polana-Eulalia complex (Campins et al. 2010; Bottke et al. 2015). The absence of a 0.7 μm absorption band in the existing spectra of Bennu is additional evidence in favor of an origin in the Polana-Eulalia complex (Campins et al. 2018). Nevertheless one should also consider the possibility that Bennu has lost the 0.7 μm band due to its proximity to the Sun as a near-Earth asteroid or that space weathering might have as well removed the signs of such band. Regarding this last point, a work by Matsuoka et al. (2015) shows that space weathering effects on C-type asteroids include a diminishing in the depth of the 0.7 μm band; on the other hand, Lantz et al. (2018) state that the effects of space weathering on the 0.7 μm absorption band have not been deciphered yet through laboratory experiments. 3.2. (163172) Ryugu: target of the JAXA Hayabusa2 mission Primitive NEA (162173) Ryugu (previously known by its provisional designation 1999 JU3) is the main target of the Japanese sample-return mission Hayabusa2 (Tsuda et al. 2013). It is a small (about 800 m), low-albedo (pV = 0.05 – 0.07; Campins et al. 2009, Müller et al. 2017) asteroid classified as a C-type and thoroughly observed from ground-based telescopes, in particular in the visible wavelength range. Fig. 7 shows all the available visible spectra of Ryugu. When more than one spectrum have been obtained, we show the mean spectrum and its corresponding standard deviation (error bars). This is the case for spectra labeled as L12 (Lazzaro et al. 2013), S12 (Sugita et al. 2013), M12 (Moskovitz et al. 2013), P16F (Perna et al. 2017, using FORS2), and P16X (Perna et al. 2017, using XShooter). Table 4 summarizes the observational circumstances for all the spectra, including the observation date and the apparent visual magnitude (mV) and phase angle (α) at the time of observation. We have also included in Table 4 the taxonomical classification of each spectrum, as well as our computed spectral slope S' in the 0.55-0.90 μm wavelength range. We do not find any clear correlation between the spectral slope and the phase angle nor the aspect angle, and so the apparent differences between the available spectra might be explained by the use of different instruments/telescopes, observing and calibration issues, use of various solar analogs, etc. Nevertheless, the variation falls within the error bars of the data. Several attempts have been made to search for surface heterogeneity in Ryugu (Lazzaro et al. 2013; Moskovitz et al. 2013; Perna et al. 2017), with no conclusive results. In the near-infrared wavelength range, published spectra by Abe et al. (2008), Pinilla-Alonso et al. (2013) and Perna et al. (2017) are similar, showing a neutral to slightly red spectral slope, while the most recent near-infrared spectra by LeCorre et al. (2018) presents a significantly redder slope. These authors fail to find a plausible explanation for such discrepancy. 11 Pre-print submitted to Icarus - Accepted (162173) Ryugu e c n a t c e l f e R e v i t l a e R B99 L12 S12 M12 V07 P16F P16X 1.5 1.0 0.5 0.3 0.4 0.5 0.6 0.7 Wavelength (µm) 0.8 0.9 Figure 7. Visible spectra of primitive NEA (162173) Ryugu, target of the JAXA Hayabusa2 mission. This asteroid has been thoroughly observed from the ground (see main text for data sources). Some of the visible spectra of Ryugu led to a taxonomical classification as Cg. Asteroids belonging to this spectral class show a drop in reflectance short ward of 0.55 μm, due to absorption in the ultra-violet wavelength region associated with the presence of aqueously altered minerals. Only one visible spectrum of Ryugu, obtained by Vilas (2008) presented an absorption band at 0.7 μm produced by hydrated silicates (this spectrum had poor signal-to-noise). The rest of the available spectra, shown in Fig. 7, do not illustrate this absorption. However, it is important to emphasize that the absence of the 0.7 μm band does not imply the lack of hydrated minerals. In roughly half of the studied cases, asteroids showing an absorption band in the 3-μm region due to hydrated silicates do not show the corresponding band at 0.7 μm (Vilas 1994; Rivkin et al. 2002, 2015). A recent study made by Busarev et al. (2018) based on the shape of visible spectra of Ryugu from Vilas (2008) and Sugita et al. (2013) obtained a month after aphelion passage, suggests the existence of sublimation/degassing activity of Ryugu and the presence of an ice reservoir at small depths, indicative of a relatively short residence time in the near-Earth space. Table 4. Summary of the observational circumstances of the available visible spectra of Ryugu. α (°) mV 17.7 17.9 19.9 18.9 19.3 Obs. Date Tax. S' (%/1000 Å) 17/05/1999 10/09/2007 1-3/06/2012 24-26/06/2012 9-10/07/2012 12/07/2016 11/08/2016 Data ID -0.75 ± 0.20 B99 0.93 ± 0.07 V07 -0.02 ± 0.02 M12 1.38 ± 0.11 S12 -0.06 ± 0.23 L12 1.52 ± 0.09 P16F P16X -0.37 ± 0.06 B99 – Binzel et al. (2001); V07 – Vilas (2008); M12 – Moskovitz et al. (2013); S12 – Sugita et al. (2013); L12 – Lazzaro et al. (2013); P16F – Perna et al. (2017), using FORS2; P16X – Perna et al. (2017), using XShooter Cg 6.1 Cb 22.5 0.2-2.0 C 22.7-30.3 C Cg Cb C 17.7-17.9 19.1-19.6 33.0 13.9 24.5 12 Pre-print submitted to Icarus - Accepted As in the case of Bennu, different studies indicate that Ryugu's most likely origin is in the primitive collisional families of the inner belt or the low-albedo and low-inclination population of background asteroids in the same region (Campins et al. 2013; Bottke et al. 2015). To further support this idea, we have taken the most recently obtained visible spectra of Ryugu from Perna et al. (2017) and compared them with the mean spectra of all the B-types and all the C-types found among the four primitive families studied in this work (Polana-Eulalia, Erigone, Sulamitis, and Clarissa). The two selected spectra of Ryugu represent very well the range of variation in spectral slope that is observed among all the available spectra (Table 4). Fig. 8 shows a comparison between these two spectra of Ryugu and the mean spectra obtained for all the B-types (orange) and C-types (red) find among the four primitive families presented in this work. The spectrum labeled as P16X corresponds to a visible-to-near-infrared (VNIR) spectrum obtained using X-Shooter (Perna et al. 2017), i.e., although the visible part presents a negative spectral slope (B-type), the complete VNIR spectrum corresponds to a C-type according to the DeMeo et al. (2009) taxonomy. As in the case of Bennu, the absence of a 0.7 μm absorption band in the spectra of Ryugu obtained to date is additional evidence in favor of an origin in the Polana-Eulalia complex (Campins et al. 2018). (162173) Ryugu 1.15 1.10 1.05 1.00 0.95 0.90 e c n a t c e l f e R e v i t l a e R Range of B-type spectra from inner-belt primitive families Range of C-type spectra from inner-belt primitive families P16F P16X +1σ +1σ -1σ 0.4 0.5 0.6 0.7 Wavelength (µm) 0.8 Figure 8. Visible spectra of primitive NEA Ryugu, target of the JAXA Hayabusa2 mission. We show here the mean spectra of all the B-type (orange) and C-type (red) asteroids found among the inner belt primitive families, as well as the corresponding ±1σ deviation of the mean. Visible spectra of Ryugu are from Perna et al. (2017) and have been smoothed to a factor of 50 for a better visualization. P16F corresponds to the spectra obtained using FORS2 instrument, while P16X corresponds to those obtained using XShooter. 4. Discussion (101955) Bennu. The only published visible spectrum of asteroid Bennu presents a blue, negative spectral slope, similar to the visible spectrum of B-type asteroid (142) Polana. This spectrum is compatible with the mean of the members of the Polana- Eulalia complex and, in general, consistent with the mean spectrum of all the B-type asteroids found within the four collisional families of primitive asteroids studied so far: Polana-Eulalia, Erigone, Sulamitis, and Clarissa. Dynamical simulations indicate that -1σ 0.9 13 Pre-print submitted to Icarus - Accepted the most likely origin of Bennu is the Polana-Eulalia complex (Bottke et al. 2015). Therefore, assuming this origin, and considering the visible spectral slopes computed for this family in de León et al. (2016), we can expect a variation in the slope of S' = [- 1.64, 0.70] %/1000Å. When considering only members of the Polana-Eulalia complex classified as B-types, the expected variation would be S' = [-1.83, -0.84] %/1000Å. Being even more conservative and including also B-type asteroids from the other studied inner-belt primitive families as potential sources of Bennu, we can expect a variation of S' = [-2,28, -0.78] %/1000Å. These ranges have been computed as [<S'>+1σ, <S'>-1σ] using the mean values <S'> and their corresponding 1σ errors shown in Table 5. Note that we have computed the average of the spectral slopes and not the spectral slope of the average spectra. ! e d u t i n g a m e v i t a e R l OSIRIS-REx MapCam filters! b'! v! w! x! 1! 0.9! 0.8! 0.7! 0.6! 0.5! 0.4! 0.3! 0.2! 0.1! 0! 0.30! 0.40! 0.50! 0.60! 0.70! μ! Wavelength ( m)! 0.80! 0.90! 1.00! Figure 9. OCAMS MapCam filter transmission curves. Figure modified from Rizk et al. (2018). In August 2018 the OSIRIS-REx mission will begin the approach phase to asteroid Bennu and start obtaining images as a point source using its suite of optical cameras. Resolved images will be obtained starting in late September 2018. The OSIRIS-REx Camera Suite (OCAMS) includes three cameras: SamCam, MapCam, and PolyCam. MapCam is equipped with a filter wheel and a set of 6 filters: 2 clear (panchromatic) filters and four color filters, aligned with the ECAS wavelengths (b'vwx) covering a wavelength range between 0.4 and 0.9 μm, approximately (see Fig. 9). Filter b' has been shifted toward longer wavelengths from the ECAS blue filter to improve optical and radiometric performance and minimize aging effects due to radiation (Rizk et al. 2018). Color images are fundamental for assessing surface composition of the target, and to subsequently select the best sample-collection site. Also, MapCam colors will be used to assess the extent of space weathering on Bennu and to help in the determination of surface "freshness", which will have higher scientific value for the sampling decision. The calculation of the b'/v and v/x color ratio maps is stipulated in two of the mission requirements documents in order to characterize the UV slope, and the visible slope, respectively, as well as the presence and depth of any 0.7 μm feature. 14 Pre-print submitted to Icarus - Accepted Figure 10. Example of the obtained reflectances and their corresponding errors for the OCAMS MapCamp filters (b'vwx) after the convolution of their transmission curves and the represented spectra. These are the mean spectra of all the B- (left panel) and C-type (right panel) asteroids in the Polana-Eulalia complex (black) and the shadowed grey region corresponds to the standard deviation of the mean (de León et al. 2016). The red part of these spectra correspond to the mean spectrum of those B-type (left) and C-type (right) asteroids in the Polana-Eulalia complex observed down to 0.35 µm, and used to compute the b' reflectance (see main text for details). Reflectances and mean spectra have been normalized to unity at the central wavelength of v filter for comparison. Colored regions correspond to the width of the different filters. Using the OCAMS MapCam transmission curves of the b'vwx filters, we have computed the range of (b'-v) and (v-x) colors expected from the visible spectra of the different populations considered as representative of the surface composition of Bennu and listed in Table 5. These include the B-type and C-type asteroids in the Polana- Eulalia complex, and B-type and C-type asteroids from the four collisional families of primitive asteroids presented in this work: Polana-Eulalia, Erigone, Sulamitis, and Clarissa. To obtain these colors we have computed the integral of the average spectra of these populations through the OCAMS MapCam filters using the relation , ri= λ2 fi λ r(λ)λdλ λ=λ1 λ2 fi(λ)λdλ λ=λ1 where ri is the reflectance in the i-th filter, λ is the wavelength, fi(λ) is the filter transmission function, r(λ) represents the asteroid reflectance spectra, and λ1 and λ2 are the lower and upper bounds of the filter transmittance, respectively. Note that, as we are working with reflectance spectra (relative to the Sun) the instrumental terms in the above relation, like the detector QE or the camera optics reflectivity, are cancelled. We apply this relation to the mean spectrum of each population, considering the ±1σ deviation associated with this mean spectrum to compute the errors in reflectances (see Fig. 10). The filter reflectance and its associated error are given by the average and the variance of these values. Finally, we obtain the colors using the relation mi – mj = -2.5 x log10(ri/rj), where ri and rj are the reflectances at the i-th and j-th filters and the color i-j is the difference in the magnitudes mi and mj. These colors are directly comparable to ECAS colors since they use a solar spectrum as the reference for the magnitude system. Fig. 15 Pre-print submitted to Icarus - Accepted 10 shows an example of the obtained reflectances through the OCAMS filters for the mean spectrum of all the B-type asteroids in the Polana-Eulalia complex (black line). Note that in this case and also for the rest of the populations, we do not have data below 0.5 μm. To compute the reflectance through the b' filter (and so the b'-v color) in this particular case, we used spectra of a sample of B-type asteroids in the Polana- Eulalia complex obtained using the 3.6m Telescopio Nazionale Galileo (TNG) and New Technology Telescope (NTT). These spectra go down to 0.35 μm (see de León et al. 2016). We computed the mean spectrum of a total of 16 B-types and used this mean spectrum to obtain the b' reflectance (red line in Fig. 10). The same approach has been followed with a sample of 9 C-type asteroids in the Polana-Eulalia complex. For the rest of the populations listed in Table 5 we used visible spectra of B- and C-type asteroids from the SMASS database (0.43 – 0.92 μm), all of them located in the same region of the belt as that of the 4 primitive families studied here (Bus & Binzel 2002). The obtained b'-v and v-x colors are shown in Table 5 and the regions covered by the different populations in the color-color plot considering the errors in the computed colors are shown in Fig. 11. We have also plotted the b'-v and v-x colors computed for Bennu by Hergenrother et al. (2013). Table 5. Summary of the average slopes and their corresponding 1σ errors for the different populations presented in this work. We also include the b'-v and v-x colors computed using the OCAMS MapCam filters transmission curves (see main text for details). The visible spectral slope of Bennu has been taken from de León et al. (2016) and colors from Hergenrother et al. (2013). Population Polana-Eulalia B-types Polana-Eulalia C-types Primitive families B-types Primitive families C-types Bennu <S'> (%/1000Å) -1.34 ± 0.49 -0.10 ± 0.87 -1.53 ± 0.75 0.61 ±0.76 -1.28 ± 0.03 b'-v v-x -0.0124 ± 0.0225 0.0043 ± 0.0184 0.0053 ± 0.0256 0.0379 ± 0.0300 -0.03 ± 0.02 -0.0348 ± 0.0626 0.0113 ± 0.0573 -0.0540 ± 0.0194 0.0247 ± 0.0132 -0.01 ± 0.03 0.15 0.10 0.05 v - ' b 0.00 -0.05 -0.15 B-types in Polana-Eulalia C-types in Polana-Eulalia B-types in inner-belt primitive families C-types in inner-belt primitive families Bennu (Hergenrother et al. 2013) 0.7-µm band -0.10 -0.05 v-x -0.00 0.05 0.10 Figure 11. Expected color variation computed for the MapCam filters on OSIRIS-REx and the mean spectra of B- and C-type asteroids from the Polana-Eulalia complex only and from the four collisional families of primitive asteroids in the inner belt: Polana-Eulalia, Erigone, Sulamitis, and Clarissa. See main text for details on the color computation. The dashed-line region corresponds to the color variation computed from the asteroids in the primitive families showing the 0.7 µm absorption band. 16 Pre-print submitted to Icarus - Accepted As we can see in Fig. 11, the color regions associated with the B- and C-type asteroids in the Polana-Eulalia complex present a larger dispersion in the (v-x) color than in the (b'-v) color. This is expected, as the (v-x) represents the visible slope and what we found in the Polana-Eulalia complex was a continuum of spectral slopes, from blue to moderately red (de León et al. 2016). Also, the data obtained by these authors presented higher levels of noise as the spectra approached 0.9 μm (x filter, see Fig. 10). If one considers the colors obtained from the mean B- and C-type spectra of the four collisional families of primitive asteroids in the inner belt, we can easily separate between the two taxonomies. The colors of Bennu fall in the region outlined by the B- type asteroids in the Polana-Eulalia complex, as expected from their visible spectra. Although neither the spectrum of Bennu nor the spectra of any of the Polana-Eulalia complex present the 0.7 μm absorption band, we do not rule out the possibility of finding such band when observing the resolved surface of the asteroid. Therefore, we have computed the (b'-v) and (v-x) colors on the mean spectra of all the asteroids in the inner belt families showing the 0.7 μm absorption band, mainly members of the Erigone and Sulamitis families. As we did for the other populations, we computed the mean spectra of Ch-type asteroids from the SMASS database and located in the inner part of the main belt, to obtain the b' reflectance value and so the b'-v color. The region covered by the two colors and their corresponding errors is depicted with a dashed-line in Fig. 11. We have also computed the band depth on both the mean spectrum and the spectrum of the asteroid showing the deepest absorption band. Band depth has been calculated following the procedure described in De Prá et al. (2018). We fit a straight line (continuum) to the reflectances at the v and x filters; divide v, w, and x reflectances by this line (continuum removal); compute the band depth as 1 – Rw, where Rw is the new reflectance at the w filter after continuum removal. According to the obtained results, if the 0.7 μm band is detected, we can expect it to have depths of the order of 2.1 ± 3.5 % on average, with the deepest absorption being 7.6 ± 2.0 %. (162173) Ryugu. Hayabusa2 will arrive at Ryugu about one month before OSIRIS-REx encounters Bennu. The spacecraft is provided with the Optical Navigation Cameras (ONC). The ONC consists of a narrow-angle camera (ONC-T, Kameda et al. 2017), equipped with a filter wheel and a set of 7 filters, covering the wavelength range from 0.4 μm to 0.95 μm, and two wide-angle cameras (ONC-W1 and ONC-W2). In the case of asteroid Ryugu, there are several published visible spectra, as detailed in the previous section. Although the majority of them are in agreement with a taxonomical classification of a C-type, there is one recently obtained spectrum that presents a negative spectral slope in the visible and shows the concave-up shape typical of C-types in the near-infrared according to DeMeo et al. (2009). Again, the most likely origin of Ryugu from dynamical simulations is the Polana-Eulalia complex, although the low-albedo, low-inclination background population has also been considered (Campins et al. 2010). Assuming an origin in the Polana-Eulalia complex one can expect the same variation in spectral slope as that for Bennu and indicated above. Considering only those members of the family complex classified as C-type asteroids, then S' = [-0.97, 0.77] %/1000Å. If one also considers the possibility of Ryugu coming from any of the four primitive families in the inner belt studied so far, and only from the C-type asteroids, we can expect a variation in slope of S' = [-0.15, 1.37] %/1000Å. This is in good agreement with the observed variation in the spectral slope of Ryugu, S' = [-0.75,1.52] %/1000Å. 17 Pre-print submitted to Icarus - Accepted For both Ryugu and Bennu, the possibility remains of finding more spectral variation in the visible spectra as surface resolved measurements are acquired. For the two targets of the space missions, the best spectral variation range to consider is the one computed from all B-types and C-types studied so far in the primitive collisional families of the inner belt. Table 5 summarizes these results. 5. Conclusions The latest results obtained within the frame of our PRIMitive Asteroids Spectroscopic Survey (PRIMASS) have allowed us to characterize the spectral behavior in the visible wavelengths of 4 collisional families of primitive asteroids located in the inner belt: the Polana-Eulalia complex and the Erigone, Sulamitis, and Clarissa families. This region, and in particular the Polana-Eulalia complex, is considered the most likely origin of the two targets of the current sample-return missions: NASA's OSIRIS-REx (Bennu) and JAXA's Hayabusa2 (Ryugu). In this paper, we have performed a global analysis on the spectral characteristics of the members of these families and have compared them to the available visible spectra of Bennu and Ryugu. The results can be summarized as follows: • According to the visible spectra obtained from their members, we can differentiate two groups among the families of primitive asteroids studied so far. The Polana-like group presents homogeneous, featureless spectra in a continuum of slopes from blue to moderately red, and no 0.7 μm band. The Erigone-like group, shows a spectral diversity among primitive taxonomic classes and a majority of spectra with the 0.7 μm band associated with phyllosilicates. • The obtained spectral slopes of the inner-belt families of primitive asteroids indicate that we should expect variations in slope between -2.28 and -0.78 %/1000 Å for Bennu and between -0.15 and 1.37 %/1000 Å for Ryugu. • While the Erigone and the Sulamitis families show very similar properties regarding spectra, collisional ages, and visible albedo, we see that the Clarissa family has on average a redder spectral slope and is significantly younger than the Polana-Eulalia complex. A tentative explanation for these differences is space weathering. Laboratory experiments with carbonaceous chondritic material show that lowest-albedo material gets bluer with space exposure age (Lantz et al. 2015, 2017). This has testable implications for Bennu and Ryugu, where older terrains would be expected to be bluer than younger surfaces (Campins et al. 2018). • The mean spectra of B- and C-type asteroids in the studied inner-belt families through the OCAMS filters show that we will be able to differentiate between B- and C-type surfaces (blue and red) using the b'-v and v-x colors. In the event of detecting the presence of the 0.7 μm absorption band, we expect to find band depths of 2.1 ± 3.5 % on average, with maximum absorption of 7.6 ± 2.0 %. Acknowledgements We want to especially thank Dr. Davide Perna for sharing his spectral data from Ryugu and Dr. Sonia Fornasier for her constructive and extremely useful review. JdL acknowledges the Spanish Ministry of Economy and Competitiveness (MINECO) under the 2015 Severo Ochoa Program MINECO SEV- 2015-0548. HC acknowledges support from NASA's Near-Earth Object Observations financial support from 18 Pre-print submitted to Icarus - Accepted program and the Center for Lunar and Asteroid Surface Science funded by NASA's SSERVI program at the University of Central Florida. DM gratefully acknowledges MINECO for the financial support received in the form of a Severo Ochoa Ph.D. Fellowship. JdL, JL, MP, and JLR acknowledge support from the project AYA2015- 67772-R (MINECO). The research leading to these results has received funding from the European Union's Horizon 2020 Research and Innovation Programme, under Grant Agreement no 687378. This material is based partly upon work supported by NASA under Contract NNM10AA11C issued through the New Frontiers Program. References - Abe, M., Kawakami, K., Hasegawa, S., et al. 2008. Ground-based Observational Campaign for Asteroid 162173 1999 JU3. 3th Lunar and Planetary Science Conference, No. 1391, p.1594 - Alí-Lagoa, V., Licandro, J., Gil-Hutton, R., et al. 2016. Differences between the Pallas collisional family and similarly sized B-type asteroids. Astron. Astroph. 591, 14-25. doi:10.1051/0004-6361/201527660 - Alí-Lagoa, V., Müller, T. G., Usui, F., et al. 2017. The AKARI IRC asteroid flux catalogue: updated diameters and albedos. Astron. Astroph. Available online 15 December 2017. doi:10.1051/0004-6361/201731806 - Binzel, R. P., Harris, A. W., Bus, S. J., et al. 2001. Spectral Properties of Near-Earth Objects: Palomar and IRTF Results for 48 Objects Including Spacecraft Targets (9969) Braille and (10302) 1989 ML. Icarus 151, 139-149. doi:10.1006/icar.2001.6613 - Binzel, R. P., DeMeo, F. E., Burt, B. J., et al. 2015. Spectral slope variations for OSIRIS-REx target Asteroid (101955) Bennu: Possible evidence for a fine-grained regolith equatorial ridge. Icarus 256, 22-29. doi:10.1016/j.icarus.2015.04.011 - Bottke, W. F., Morbidelli, A., Jedicke, R., et al., 2002. Debiased orbital and absolute magnitude distribution of Icarus 156, 399-433. doi:10.1006/icar.2001.6788 - Bottke, W. F., Vokrouhlický, D., Rubincam, D. P., et al., 2006. The Yarkovsky and Yorp effects: Implications for asteroid dynamics. Ann. Rev. Earth Planet. Sci. 34, 157- 191. doi:10.1146/annurev.earth.34.031405.125154 - Bottke, W. F., Vokrouhlický, D., Walsh, K. J., et al., 2015. In search of the source of asteroid (101955) Bennu: Applications of the stochastic YORP model. Icarus 247, 191- 217. doi:10.1016/j.icarus.2014.09.046 - Bus, S. J. & Binzel, R. P. 2002. Phase II of the Small Main-Belt Asteroid Survey. Spectroscopic 106-145. doi:10.1006/icar.2002.6857 - Busarev, V. V., Makalkin, A. B., Vilas, F., et al. 2018. New candidates for active asteroids: Main-belt (145) Adeona, (704) Interamnia, (779) Nina, (1474) Beira, and near-Earth (162173) Ryugu. Icarus, 304, 83-94. doi:10.1016/j.icarus.2017.06.032. - Campins, H., Emery, J. P., Kelley, M., et al. 2009. Spitzer observations of spacecraft target 162173 (1999 JU3). Astron. Astroph. 503, L17-L20. doi:10.1051/0004- 6361/200912374. - Campins, H., Morbidelli, A., Tsiganis, K., et al., 2010. The origin of asteroid 101955 (1999 RQ36). Astrophys. J. Lett. 721, L53-L57. doi:10.1088/2041-8205/721/1/L53. - Campins, H., de León, J., Morbidelli, A., et al., 2013. The origin of Asteroid 162173 (1999 JU3). Astron. J. 146, 26-32. doi:10.1088/0004-6256/146/2/26. - Campins, H., de León, J., Licandro, J. et al. 2018. Compositional Diversity Among Primitive Asteroids. In Primitive Meteorites and Asteroids: Physical, Chemical and the near-Earth objects. Observations. The Icarus 158, 19 Pre-print submitted to Icarus - Accepted Icarus 202, into the taxonomy groups. Icarus, Available online 13 November near-infrared. family complex: Spectral homogeneity. Spectroscopic Observations Paving the Way for Exploration. Elsevier. Accepted for publication. - Clark, B. E., Binzel, R. P., Howell, E. S., et al., 2011. Asteroid (101955) 1999 RQ36: Spectroscopy from 0.4 to 2.4 μm and meteorite analogs. Icarus 216, 462-475. doi:10.1016/j.icarus.2011.08.021. - de León, J., Pinilla-Alonso, N., Campins, H., et al. 2012. Near-infrared spectroscopic survey of B-type asteroids: Compositional analysis. Icarus 218, 196-206. doi:10.1016/j.icarus.2011.11.024 - de León, J., Pinilla-Alonso, N., Delbo, M., et al., 2016. Visible spectroscopy of the Polana-Eulalia Icarus 266, 57-75. doi:10.1016/j.icarus.2015.11.014. - DeMeo, F. E., Binzel, R. P., Slivan, S. M., et al. 2009. An extension of the Bus asteroid 160-180. doi:10.1016/j.icarus.2009.02.005 - De Prá, M. N., Pinilla-Alonso, N., Carvano, J. M., et al., 2018. PRIMASS visits Hilda and Cybele 2017. doi:10.1016/j.icarus.2017.11.012 - Dykhuis, M. J. & Greenberg, R. 2015. Collisional family structure within the Nysa- Polana complex. Icarus 252, 199-211. doi:10.1016/j.icarus.2015.01.012 - Fornasier, S., Lantz, C., Barucci, M. A., Lazzarin, M. 2014. Aqueous alteration on main belt primitive asteroids: Results from visible spectroscopy. Icarus, 233, 163-178. doi:10.1016/j.icarus.2014.01.040 - Gayon-Markt, J., Delbo, M., Morbidelli, A., et al., 2012. On the origin of the Almahata Sitta meteorite and 2008 TC3 asteroid. Mon. Not. R. Astron. Soc. 424, 508-518. doi:10.1111/j.1365-2966.2012.21220.x - Harris, A. W. 1998. A Thermal Model for Near-Earth Asteroids. Icarus 131, 291-301. doi:10.1006/icar.1997.5865. - Hergenrother, C. W., Nolan, M. C., Binzel, R. P., et al. 2013. Lightcurve, Color and Phase Function Photometry of the OSIRIS-REx Target Asteroid (101955) Bennu. Icarus 226, 663-670. doi:10.106/j.icarus.2013.05.044 - Kameda, S., Suzuki, H., Takamatsu, T., et al. 2017. Preflight Calibration Test Results for Optical Navigation Camera Telescope (ONC-T) Onboard the Hayabusa2 Spacecraft. Space Science Reviews 208, 17-31. doi:10.1007/s11214-015-0227-y - Lantz, C., Brunetto, R., Barucci, M. A., et al. 2015. Ion irratiation of the Murchison meteorite: Visible to mid-indrared spectroscopic results. Astron. Astroph. 577, 41-50. doi:10.1051/0004-6361/201425398 - Lantz, C., Brunetto, R., Barucci, M. A., et al. 2017. Ion irradiation of carbonaceous chondrites: A new view of space weathering on primitive asteroids. Icarus 285, 43-57. doi:10.1016/j.icarus.2016.12.019 - Lantz, C., Binzel, R. P., DeMeo, F. E. 2018. Space weathering trends on carbonaceous asteroids: A possible explanation for Bennu's blue slope. Icarus 302, 10- 17. doi:10.1016/j.icarus.2017.11.010. - Lauretta, D. S., Bartels, A. E., Barucci, M. A., et al. 2015. The OSIRIS-REx target asteroid (101955) Bennu: Constraints on its physical, geological, and dynamical nature from 834-849. doi:10.1111/maps.12353 - Lauretta, D. S., Balram-Knutson, S. S., Beshore, E., et al. 2017. OSIRIS-REx: Sample Return from Asteroid (101955) Bennu. Space Science Reviews 212, 925-984. doi:10.1007/s11214-017-0405-1 - Lazzaro, D., Angeli, C. A., Carvano, J. M., et al. 2004. S3OS2: the visible spectroscopic 179-220. of doi:10.1016/j.icarus.2004.06.006 observations. Met. astronomical asteroids. & Plan. Sci. 50, Icarus 172, survey 820 20 Pre-print submitted to Icarus - Accepted - Lazzaro, D., Barucci, M. A., Perna, D., et al. 2013. Rotational spectra of (162173) 1999 JU3, the target of the Hayabusa 2 mission. Astron. Astroph. 549, L2-L6. doi:10.1051/0004-6361/201220629 - LeCorre, L., Sanchez, J. A., Reddy, V., et al. 2018. Ground-based characterization of Hayabusa2 mission target asteroid 162173 Ryugu: constraining mineralogical composition in preparation for spacecraft operations. Month. Not. Roy. Astron. Soc. 475, 614-623. doi: 10.1093/mnras/stx3236 - Luu, J. X. & Jewitt, D. C. 1990. Charge-coupled device spectra of asteroids. I – Near- earth and 3:1 resonance asteroids. Astron. Journal 99, 1985-2011. doi:10.1086/115481 - Mainzer, A., Bauer, J., Grav, T., et al. 2011. Preliminary Results from NEOWISE: An Enhancement to the Wide-field Infrared Survey Explorer for Solar System Science. Astrophys. J. 731, 53-66. doi:10.1088/0004-637X/731/1/53. - Matsuoka, M., Nakamura, T., Kimura, Y., et al. 2015. Pulse-laser irradiation experiments of Murchison CM2 chondrite for reproducing space weathering on C-type asteroids. Icarus 254, 135-143. doi:10.1016/j.icarus.2015.02.029. - Milani, A., Cellino, A., Knežević, Z., et al., 2014. Asteroid families classification: Exploiting very large databases. Icarus 239, 46-73. doi:10.1016/j.icarus.2014.05.039 - Morate, D., de León, J., De Prá, M., et al., 2016. Compositional study of asteroids in the Erigone collisional family using visible spectroscopy at the 10.4m GTC. Astron. Astrophys. 586, A129. doi:10.1051/0004-6361/201527453 - Morate, D., de León, J., De Prá, M., et al, 2018. Visible spectroscopy of the Sulamitis and Clarissa primitive families: a possible link to Erigone and Polana. Astron. Astrophys. Available online 5 October 2017. doi:10.1051/0004-6361/201731407. - Morbidelli, A., Bottke, Jr. W. W., Froeschlé, C., et al., 2002. Origin and evolution of near-Earth objects. In Asteroids III. University of Arizona Press, Tucson, p. 409-422. - Morbidelli, A. & Vokrouhlicky, 2003. The Yarkovsky-driven origin of near-Earth asteroids. Icarus 163, 120-134. doi:10.1016/S0019-1035(03)00047-2 - Moskovitz, N. A., Abe, S., Pan, K-S., et al. 2013. Rotational characterization of Hayabusa Icarus 224, 24-31. target Asteroid doi:10.1016/j.icarus.2013.02.009 - Müller, T. G., Ďurech, J., Ishiguro, M., et al. 2017. Hayabusa-2 mission target asteroid 162173 Ryugu (1999 JU3): Searching for the object's spin-axis orientation. Astron. Astroph. 599, A103. doi:10.1051/0004-6361/201629134. - Nesvorny, D., Brož, M., Carruba, V. 2015. Identification and Dynamical Properties of Asteroid Families. In Asteroids IV. University of Arizona Press, Tucson, p.297-321 - Oszkiewicz, D. A., Bowell, E., Wasserman, L. H., et al. 2012. Asteroid taxonomic signatures 283-296. doi:10.1016/j.icarus.2012.02.028 - Perna, D., Barucci, M. A., Ishiguro, M., et al. 2017. Spectral and rotational properties of near-Earth asteroid (162173) Ryugu, target of the Hayabusa 2 sample return mission. Astron. Astroph. 599, L1-L5. doi:10.1051/0004-6361/201630346 - Pinilla-Alonso, N., Lorenzi, V., Campins, H., et al. 2013. Near-infrared spectroscopy of 1999 JU3, the target of the Hayabusa 2 mission. Astron. Astroph. 552, 79-82. doi:10.1051/0004-6361/201221015 - Pinilla-Alonso, N., de León, J., Walsh, K., et al. 2016. Portrait of the Polana-Eulalia family complex: Surface homogeneity revealed from near-infrared spectroscopy. Icarus 274, 231-248. doi:10.1016/j.icarus.2016.03.022 - Rivkin, A., S., Howell, E. S., Vilas, F., et al. 2002. Hydrated Minerals on Asteroids: The Astronomical Record. In Asteroids III. University of Arizona Press, Tucson, p.235- 253. - Rivkin, A. S., Campins, H., Emery, J. P., et al. 2015. Astronomical Observations of Volatiles on Asteroids. In Asteroids IV, University of Arizona Press, Tucson, p.65-87. doi:10.2458/azu_uapress_9780816532131-ch004 (162173) 1999 photometric JU3. from phase curves. Icarus 219, II 21 Pre-print submitted to Icarus - Accepted - Rizk, B., Drouet d'Aubigny, C., Golish, D. et al. 2018. OCAMS: The OSIRIS-REx Camera Suite. Space Sience Reviews 214, 26-81. doi:10.1007/s11214-017-0460-7 - Sugita, S., Kuroda, D., Kameda, S., et al. 2013. Visible Spectroscopic Observations of Asteroid 162173 (1999 JU3) with the Gemini-S Telescope. 44th Lunar and Planetary Science Conferences. No. 1719, p.2591 - Tsuda, Y., Yoshikawa, M., Abe, M., et al., 2013. System design of the Hayabusa 2 – Asteroid sample return mission to 1999 JU3. Acta Astronaut. 91, 356-362. doi: 10.1016/j.actaastro.2013.06.028 - Vilas, F. & McFadden, L. A. 1992. CCD reflectance spectra of selected asteroids. I – Presentation and data analysis considerations. Icarus 100, 85-94. doi:10.1016/0019- 1035(92)90020-8 - Vilas, F. 1994. A cheaper, faster, better way to detect water of hydration on Solar System bodies. Icarus 111, 456-467. doi:10.1006/icar.1994.1156 - Vilas, F. 2008. Spectral Characteristics of Hayabusa 2 Near-Earth Asteroid Targets 162173 1999 JU3 and 2001 QC34. Astron. Journal 135, 1101-1105. doi:10.1088/0004- 6256/135/4/1101 - Walsh, K. J., Delbo, M., Bottke, W. F., et al., 2013. Introducing the Eulalia and new Polana asteroid families: Re-assessing primitive asteroid families in the inner main belt. Icarus 225, 283-297. doi:10.1016/j.icarus.2013.03.005 - Wright, E. L., Eisenhardt, P. R. M., Mainzer, A. K., et al. 2010. The Wide-field Infrared Explorer (WISE): Mission Description and Initial On-orbit Performance. Astron. J. 140, 1868-1881. doi:10.1088/0004-6256/140/6/1868. - Xu, S., Binzel, R. P., Burbine, T. H., et al. 1995. Small main-belt asteroid spectroscopic survey: Initial results. Icarus 115, 1-35. doi:10.1006/icar.1995.1075 22
1007.0579	1	1007	2010-07-04T19:36:45	From planetesimals to terrestrial planets: N-body simulations including the effects of nebular gas and giant planets	[ "astro-ph.EP" ]	We present results from a suite of N-body simulations that follow the accretion history of the terrestrial planets using a new parallel treecode that we have developed. We initially place 2000 equal size planetesimals between 0.5--4.0 AU and the collisional growth is followed until the completion of planetary accretion (> 100 Myr). All the important effect of gas in laminar disks are taken into account: the aerodynamic gas drag, the disk-planet interaction including Type I migration, and the global disk potential which causes inward migration of secular resonances as the gas dissipates. We vary the initial total mass and spatial distribution of the planetesimals, the time scale of dissipation of nebular gas, and orbits of Jupiter and Saturn. We end up with one to five planets in the terrestrial region. In order to maintain sufficient mass in this region in the presence of Type I migration, the time scale of gas dissipation needs to be 1-2 Myr. The final configurations and collisional histories strongly depend on the orbital eccentricity of Jupiter. If today's eccentricity of Jupiter is used, then most of bodies in the asteroidal region are swept up within the terrestrial region owing to the inward migration of the secular resonance, and giant impacts between protoplanets occur most commonly around 10 Myr. If the orbital eccentricity of Jupiter is close to zero, as suggested in the Nice model, the effect of the secular resonance is negligible and a large amount of mass stays for a long period of time in the asteroidal region. With a circular orbit for Jupiter, giant impacts usually occur around 100 Myr, consistent with the accretion time scale indicated from isotope records. However, we inevitably have an Earth size planet at around 2 AU in this case. It is very difficult to obtain spatially concentrated terrestrial planets together with very late giant impacts.	astro-ph.EP	astro-ph	– 1 – From planetesimals to terrestrial planets: N-body simulations including the effects of nebular gas and giant planets Ryuji Morishima1,2, Joachim Stadel3, Ben Moore3 [email protected] 1: Jet Propulsion Laboratory/California Institute of Technology, Pasadena, CA, USA 2: Institute of Geophysics and Planetary Physics/UCLA, Los Angels, CA, USA 3: Institute for Theoretical Physics, University of Zurich, Switzerland Accepted for publication in Icarus Total manuscript page : 51 Table: 2 Figures: 19 0 1 0 2 l u J 4 . ] P E h p - o r t s a [ 1 v 9 7 5 0 . 7 0 0 1 : v i X r a – 2 – Proposed Running Head: Formation of terrestrial planets with gas and giant planets Editorial correspondence to: Ryuji Morishima Jet Propulsion Laboratory, M/S 230-205, 4800 Oak Grove Drive, Pasadena, CA 91109, USA Phone:+1 818 393 1014; Fax:+1 818 393 4495 e-mail: [email protected] – 3 – ABSTRACT We present results from a suite of N-body simulations that follow the formation and accretion history of the terrestrial planets using a new parallel treecode that we have developed. We initially place 2000 equal size planetesimals between 0.5–4.0 AU and the collisional growth is followed until the completion of planetary accre- tion (> 100 Myr). A total of 64 simulations were carried out to explore sensitivity to the key parameters and initial conditions. All the important effect of gas in laminar disks are taken into account: the aerodynamic gas drag, the disk-planet interaction including Type I migration, and the global disk potential which causes inward mi- gration of secular resonances as the gas dissipates. We vary the initial total mass and spatial distribution of the planetesimals, the time scale of dissipation of nebular gas (which dissipates uniformly in space and exponentially in time), and orbits of Jupiter and Saturn. We end up with one to five planets in the terrestrial region. In order to maintain sufficient mass in this region in the presence of Type I migration, the time scale of gas dissipation needs to be 1-2 Myr. The final configurations and collisional histories strongly depend on the orbital eccentricity of Jupiter. If today's eccentricity of Jupiter is used, then most of bodies in the asteroidal region are swept up within the terrestrial region owing to the inward migration of the secular reso- nance, and giant impacts between protoplanets occur most commonly around 10 Myr. If the orbital eccentricity of Jupiter is close to zero, as suggested in the Nice model, the effect of the secular resonance is negligible and a large amount of mass stays for a long period of time in the asteroidal region. With a circular orbit for Jupiter, giant impacts usually occur around 100 Myr, consistent with the accretion time scale indicated from isotope records. However, we inevitably have an Earth size planet at around 2 AU in this case. It is very difficult to obtain spatially concen- trated terrestrial planets together with very late giant impacts, as long as we include all the above effects of gas and assume initial disks similar to the minimum mass solar nebular. Key words: Accretion– Origin, Solar system– Planetary formation–Terrestrial planets 1. Introduction Accretion models for the formation of the terrestrial planets need to reproduce the follow- ing properties of our solar system. (1) The spatial mass distribution of the terrestrial planets is radially concentrated. Namely, the Earth and Venus contain most of the mass inside of Jupiter's orbit, whereas the masses of Mars and Mercury are an order of magnitude smaller than the Earth's mass (M . The orbital separation between Earth and Venus is only 0.3 AU. (2) The orbits of the Earth and Venus are nearly cir- cular and coplanar even though orbital excitements are expected in the late stage of accretion ) and the total mass of asteroids is only 5 × 10−4 M ⊕ ⊕ – 4 – where large protoplanets experience mutual collisions. This probably indicates that there were some damping mechanisms to reduce their orbital eccentricities and inclinations near the end of their formation processes. (3) Isotope records such as W/Hf and U/Pb for the Earth and Moon indicate that the Moon-forming giant impact occurred ∼ 100 Myr after the beginning of the solar system (Touboul et al., 2007; All´egre et al., 2008). Some accretion models can reproduce two of the three properties but there is no accretion model that satisfies all three properties. N-body simulations that follow only the gravitational interactions of protoplanets (or with the giant planets) usually produce too large orbital eccentricities of the final planets (Chambers, 1998; Agnor et al., 1999; Raymond et al., 2004; Kokubo et al., 2006). To resolve this issue two damping mechanisms have been proposed: one is by remnant planetesimals (Chambers, 2001; O'Brien et al., 2006; Raymond et al., 2006, 2009; Morishima et al., 2008) and another is by remnant nebular gas (Agnor and Ward, 2002; Kominami et al., 2002, 2004; Nagasawa et al., 2005; Ogihara et al., 2007; Thommes et al., 2008). In particular, the dynamical shake-up model of Nagasawa et al. (2005) and Thommes et al. (2008) is the most successful model so far in reproducing both the small orbital eccentric- ities and the radial mass concentration. Their model initially places a few tens of protoplanets embedded in a gas disk and takes into account the effect of the gas giant planets with orbital eccentricities comparable to or slightly larger than the present values (the present eccentrici- ties of both Jupiter and Saturn are ∼ 0.05). The secular resonance with Jupiter migrates from ∼4.0 to 0.6 AU with dissipation of the gas disk and sweeps protoplanets toward the current location of the Earth. Mutual impacts between protoplanets occur even when a large amount of gas remains, and the remnant gas disk reduces the orbital eccentricities of formed Earth-like planets. If the formation of the Moon occurred very late, as indicated by recent isotope measure- ments (Touboul et al., 2007), significant damping due to remnant gas may be problematic for the following reasons. Significant depletion of gas already at ∼ 10 Myr is suggested from di- rect measurements of line emissions of circumstellar disks (Pascucci et al. 2006) and from Hα emission profiles of central stars due to accretion shock (Muzerolle et al., 2000), although we cannot rule out the possibility that our protosolar nebular was exceptionally long-lived. Another potential problem for long-lived gas disks is depletion of the solid component due to Type I mi- gration, which is neglected in the model of Nagasawa et al. (2005) and Thommes et al. (2008). Once a protoplanet becomes as massive as Mars in a gas disk, the gravitational torque from spi- ral density waves launched by the protoplanet causes rapid inward migration of the protoplanet, referred to as Type I migration (Goldreich and Tremaine, 1980; Ward, 1986, 1997; Tanaka et al., 2002). McNeil et al. (2005) and Daisaka et al. (2006) conducted N-body simulations taking Type I migration into account and found that the decay time scale for a gas disk needs to be ∼ 1 Myr in order to retain sufficient solid mass for the formation of Earth-size planets. Some remnant planetesimals are likely to have existed even at 100 Myr particularly in the – 5 – asteroidal region. 1 Several authors have conducted N-body simulations starting with proto- planets and remnant planetesimals (Chambers, 2001; Chambers and Cassen, 2002; O'Brien et al. 2006; Raymond et al. 2006, 2009). Their results show that final eccentricities of planets tend to be lower with increasing total mass of remnant planetesimals. Even with the same total mass of planetesimals, the damping is more effective when the size of an individual planetes- imal is smaller and planetesimals are placed uniformly over both the terrestrial and asteroidal regions rather than in the asteroidal region only (O'Brien et al., 2006; Raymond et al., 2006). An eccentricity of Jupiter comparable to or slightly larger than the present value is preferred in regard to the radial mass concentration (Chambers and Cassen, 2002; O'Brien et al., 2006; Ray- mond et al., 2009), although none of simulations have succeeded in reproducing the very high radial mass concentration of the current terrestrial planets. The spatial and size distributions of remnant planetesimals are arbitrarily assumed in these simulations. In order to clarify these distributions one needs to follow the growth of planetesimals and protoplanets in the runaway growth stage. In addition, the above simulations neglect the effects of gas. The protosolar neb- ular must have existed at least until the formation of Jupiter, and probably survived somewhat longer. Even a rapid dissipation of gas of the order of 0.1 Myr in the presence of Jupiter strongly affects the orbital evolution of asteroidal bodies (Nagasawa et al., 2000). In the present study we perform N-body simulations starting with a wide planetesimal disk and continue until the end of the accretion phase. Since we directly calculate the runaway growth of the protoplanets, we can obtain reasonable spatial and size distributions of remnant planetesimals in the late stage. We take into account the gas giant planets and all the possible effects of gas in laminar disks: aerodynamic gas drag, disk-planet interaction including Type I migration, and global disk potential. In a previous study, Morishima et al. (2008) also con- ducted N-body simulations that followed the evolution of a protoplanetary disk starting with planetesimals only until the end of accretion. However, their simulations started from radially narrow annuli (widths of 0.3-0.5 AU) and neglected any external forces. In Sec. 2, we introduce our new N-body code, review the effects of nebular gas, and outline our choice of simulation parameters. In Sec. 3, we show time evolution of two contrasting examples, present the results of all simulations and compare them with the solar system. In principle, none of our simula- tions succeed in satisfying all the three constraints listed in the first paragraph of this section. In Sec. 4, we discuss possible solutions for this problem. In Sec. 5, we give our conclusions. 2. Method 2.1. N-body code Any N-body code consists of a gravity solver to calculate the mutual gravity force and an integrator to update positions and velocities for a next time step. Existing N-body codes applied 1 In the present paper, we loosely refer to the outside and inside of 2 AU as the asteroidal and terrestrial regions, respectively. – 6 – for the runaway growth stage have good gravity solvers, such as tree methods (Richardson et al., 2000; Stadel, 2001) or special hardware such as HARP or GRAPE (Kokubo and Ida, 1996). On the other hand, N-body codes applied for the late stage have good integrators, such as SyMBA (Duncan et al., 1998) and Mercury (Chambers, 1999), so that one can take a large time step size, while solving the mutual gravity by direct N2 calculation. The new N-body code we have developed has both a good gravity solver and a good integrator, allowing us to follow accretion stages of planetary systems with large N and over a large number of orbits. The code is still being upgraded and details of the complete version will be reported elsewhere. Here we briefly describe the version of the code used for simulations in this paper. The gravity solver we use is a parallel-tree method, PKDGRAV (Stadel, 2001). The orig- inal version of PKDGRAV uses a leap-frog integration scheme, where individual particles are updated in hierarchical time step blocks, each block with a factor of two smaller time-step according to the local dynamical time of its particles. Richardson et al. (2000) added some functions for planetary accretion to the code and conducted simulations for the runaway growth stage. In the tree method, all particles are divided into hierarchal tree-structured groups. Then, the gravity forces from distant particle groups are calculated using their multipole expansions. The zeroth-order multipole expansion corresponds to the gravity from the mass center of a par- ticle group. In order to keep good accuracy, we include multipole expansions up to the fourth order. The computer time of gravity calculation using this gravity solver is only proportional to N log N. We find a significant speed up in the gravity calculation with our tree method as compared with the direct N2 calculation when N > 200. In addition, the tree method is suitable for parallel computing, and this results in further speed up when N > 1000. Since the number of particles decreases with growth of planets, we switch calculation methods from parallel-tree to serial-tree, and to serial-direct N2 calculations. The error in the total energy due to the tree method appears as a random walk, and thus is proportional to the square root of the number of time steps. Therefore, with some short additional simulations, we can estimate the accumulated error after a long term evolution of a system. The accuracy of the method is controlled by the opening angle for tree cells and we set it to be 0.5 in the present paper. We confirm that the expected error due to the gravity calculation is sufficiently smaller than the error due to the integrator's, which rises most notably during close encounters. The integrator implemented in the code is a mixed-variable symplectic (MVS) integrator, SyMBA (Duncan et al., 1998). In MVS integrators such as SyMBA and Mercury, the Hamilto- nian is split into a part which describes the Keplerian motion around the central star and another part which describes the perturbation from other bodies. The former part corresponds to up- dates of positions and velocities along Kepler orbits. This can be done nearly analytically. The latter part corresponds to velocity changes due to the perturbation from other bodies, which we calculate using the tree method described above. These types of integrators are very accurate and allow us to take very large time steps for long term simulations. We use a time step size, ∆t, of 6 days. This is comparable to those used in previous high-resolution N-body simulations using MVS integrators (O'Brien et al., 2006; Raymond et al., 2006). Practically, the time step size needs to be chosen so that a fraction of particles in close encounters defined below is small enough. – 7 – In close encounter search, we first search for 10–20 nearest neighbors of each particle using the same algorithm in the tree SPH code, Gasoline (Wadsley et al., 2004). With the relative velocities and positions at the beginning and end of a time step, the closest distance of a pair during the step is calculated using a third order interpolation (Eq. (11) of Chambers, 1999). In order to judge whether a particle is in a close encounter or not, we first define the critical distance for each particle. The critical distance is defined as the larger of (1) n1RH or (2) n2vep∆t, where n1 and n2 are numerical factor (we use n1 = 4.5 and n2 = 1.0), RH is the Hill radius, and vep = vkep √e2 + i2 is the epicyclic velocity of the particle with vkep, e, and i being the Kepler velocity, eccentricity, and inclination, respectively. If the closest distance of a pair is smaller than the sum of the critical distances, we regard this pair to be in a close encounter. In the original SyMBA code (Duncan et al., 1998), only the first condition is used. However, the second condition is necessary if the relative velocity is much larger than the mutual escape velocity; such encounters usually occur near resonances of the giant planets or between small planetesimals stirred by large protoplanets. In the Mercury code (Chambers, 1999), the second condition uses the largest vkep in all particles instead of each body's vep which we apply. Although this criterion of the Mercury code gives better energy conservation (for a same n2), it also results in too many pairs being identified as involved in close encounters during the runaway growth stage, where the number density of particles is large. If the number density is not large, as is the case in the late stage, use of the Mercury's criterion is recommended (although we do not use it even in the late stage in this paper). All the above computations are performed in parallel (if N > 1000), while particles in close encounters are collected on a single cpu during the domain decomposition; the procedure which distributes the workload, and hence the remaining particles, amongst processors. Then, multistep time stepping is done until all particles proceed to the end of the original time step, following the procedure described in Duncan et al. (1998). The closest mutual distance in a substep is always estimated by interpolation using the information of both the beginning and the end of the step. This guarantees time symmetry of the integrator which is essential to avoid secular errors. The computational time for the substep procedure is small enough as it is simply proportional to the number of particles in close encounters. We set the ratio of the cut-off radius to that of the next order to be 2.08 and the number of next order substeps in the current substep to be 3 after Duncan et al. (1998). We assume that two bodies merge when they partially overlap. We assume the density of particles to be 2 g cm−3. Since we do not apply the artificial enhancement of radii (Kokubo and Ida, 2002, Morishima et al., 2008), the accretion time scale we will show is realistic. A particle is assumed to collide with the Sun when either the heliocentric distance is less than 0.1 AU or the semimajor axis is less than 0.25 AU. A particle is assumed to escape from the system when the heliocentric distance is larger than 100 AU. The energy error after completion of accretion is typically ∆E/E = 10−5–10−3. Note that energy changes due to merging or gas forces can be explicitly calculated and are subtracted in the above error. The largest portion of the error usually accumulates after the runaway growth stage where a large number of excited small particles experience mutual encounters. Therefore, the energy error primarily depends on how long small particles survive. – 8 – 2.2. Effects of gas In this section, we review the forces due to the nebular gas. The acceleration due to the gas is incorporated in the integrator by placing the half kicks before and after the Kepler drift oper- ation. This symmetric placement once again avoids secular errors coming from the inclusion of gas forces. 2.2.1. Aerodynamic drag The aerodynamic drag force per unit mass (or acceleration) is given by (Adachi et al., 1976) f drag = − 1 2m cDπr2 pρgasurelurel, (1) where m and rp are the mass and the radius of a particle, cD is the numerical coefficient (assumed to be 2 in the present paper), ρgas is the gas density, and urel = u − ugas is the velocity of a particle relative to the gas velocity. In the cylindrical coordinate (r, θ, z), the gas velocity in the θ component is obtained by the force balance in the r component as v2 gas(r, z) r = c2 r ∂ ln ρgas ∂ ln r + ∂ ln T ∂ ln r! + GM ⊙ (r2 + z2)3/2 r − fglob,r(r, z), (2) ⊙ where c is the isothermal sound velocity, T is the gas temperature, G is the gravitational con- stant, and M is the mass of the central star (assumed to be the solar mass). The left hand side of Eq. (2) represents the centrifugal force while the first, second, and third terms of the right hand side represent the pressure gradient, the gravity of the central star, and the self-gravity of the gas disk (Eq. (16)), respectively. Assuming the gas is isothermal in the z direction and neglecting the self gravity of the gas, the gas density is approximately given by ρgas(r, z) = Σgas √2πh! exp − 2h2!, where Σgas is the surface density of the gas and h = c/Ωkep is the scale height, with the Keplerian frequency Ωkep. If we assume Σgas ∝ r−p, T ∝ r−q, we obtain, 2! 1 −(cid:18) z = −p + q 2 − ∂ ln ρgas ∂ ln r h(cid:19)2! . (3) (4) z2 3 We adopt the temperature profile of an optically thin disk, T = 280 (r/1 AU)−1/2K (q = 1/2), after Hayashi et al. (1981). We will explain how Σgas and p are given in Sec. 2.3. The growth rate of a protoplanet is proportional to e−2, where e is the orbital eccentricity of planetesimals, and the balance between viscous stirring by protoplanets and damping due to gas drag gives e ∝ m1/18 (Eq. (25) of Kokubo and Ida, 2000). The initial planetesimal mass m0 in our simulations is ∼ 1025 g. On the other hand, the size of planetesimals mGI formed by – 9 – gravitational instability is estimated to be several order of magnitude smaller (Goldreich and Ward, 1973). Therefore, the growth time scale of protoplanets will be significantly overesti- mated in our model. In order to reproduce the growth rate of protoplanets surrounded by small planetesimals (∼ mGI), we artificially enhance the gas drag force on planetesimals by a factor of (m/m′)2/3 in N-body simulations if the particle mass m is smaller than m1 (> m0). Here the mass m′ is given as log m′ − log mGI log m1 − log mGI = log m − log m0 log m1 − log m0 . (for m < m1) (5) With this scheme, a bunch of small planetesimals of mass m′ are represented by a single super planetesimal of mass m (e.g., McNeil et al., 2005). The mass m1 should be as small as possible in order to avoid artificial clean up of large planetesimals by protoplanets. We use m1 = 0.01 M ⊕ as a compromise, and mGI = 1019 g. The migration and damping time scales with the artificial . enhancement of the gas drag are shown in Fig. 1 for the case of e = 0.05 and m0 = 0.0025 M ⊕ [Fig. 1] 2.2.2. Tidal interaction between a protoplanet and a nebular disk We use the formulation of Tanaka et al. (2002) for Type I migration rate. For the damping rates of orbital eccentricity e and inclination i, we use the formulation given in Tanaka and Ward (2004). We make corrections in these formulations for the case of large e and i after Papaloizou and Larwood (2000) and Cresswell et al. (2007). Tanaka et al. (2002) derived the torque due to tidal interaction between a protoplanet and a three-dimensional baroclinic disk, for the case of e = i = 0. The decay time scale of the semimajor axis a is given as = (2.7 + 1.1p)−1 M ⊙ m where ftid1,θ is the acceleration in the azimuthal direction. τtid1 = − vkep ftid1,θ = − a a 1 2 ⊙ M vkep!2 Σgasa2 c Ω−1 kep, (6) Tanaka and Ward (2004) extended their work to the case of nonzero e and i (≪ h/a) and obtained the rate for damping of e and i due to the tidal interaction. The r, θ, and z components of the damping force per unit mass are given as (a typo was corrected; see Ogihara et al., 2007) ftid,r = ftid,θ = ftid,z = 1 1 τwave h0.104[vθ − v′kep(r)] + 0.176vri , τwave h−1.736[vθ − v′kep(r)] + 0.325vri , τwave h−1.088vz + 0.871zΩkep(r)i , 1 where characteristic time of the orbital evolution, τwave, is given by τwave = m M ⊙ !−1 ⊙!−1 Σgasa2 M vkep!4 c Ω−1 kep. (7) (8) (9) (10) – 10 – In Eqs. (7)–(9), the velocity components of a particle are given as u = (vr, vθ, vz) and v′kep(r) = (v2 kep − r fglob,r(r, 0))1/2 is the Keplerian velocity of solid bodies including the gravity of the gas disk. Note that all the components of the damping force become zero when e = i = 0. The formulation of Tanaka and Ward (2004) is limited to small e and i (≪ h/a), or that the planet motion relative to the gas remains subsonic. From a two dimensional linear theory, Papaloizou and Larwood (2000) found that the time scales τtid1 and τwave (Eqs. (6) and (10)) increase with increasing e by factors of gtid1 and gwave, respectively: gtid1 = 1 + (x/1.3)5 1 − (x/1.1)4 , x3. gwave = 1 + 1 4 (11) (12) Here the original x used in Papaloizou and Larwood (2000) is xPL = ae/h. On the other hand, we use a modified value as x = ap(e2 + i2) 2h , (13) in order to reproduce similar results of simulations of Cresswell et al. (2007) (see the next paragraph). The evolution of the orbital eccentricity with the corrections for large e is given as 1 e de dt!tid = − 0.78 τwavegwave . (14) This equation corresponds to Eq. (45) of Tanaka and Ward (2004), but the correction factor is multiplied by τwave. From Eqs. (6), (11), and (14), the evolution of the semimajor axis due to the tidal interaction is given by, 1 a da dt!tid 1 τtid1gtid1 − = − 1 e de dt!tid 2e2 √1 − e2! , (15) where the first term in the right hand side comes from exchange of orbital angular momenta while the second term represents the migration rate due to energy dissipation. Cresswell et al. (2007) conducted hydrodynamic simulations for migration of a 20 Earth mass planet with nonzero e and i embedded in a gas disk. They found good agreement with Tanaka et al. (2002) and Tanaka and Ward (2004) for small e and i (< 0.1), while the damping time scale e/e is proportional to e3 for large e as predicted by Papaloizou and Larwood (2000) (Eq. (12)). They also showed the same dependence of the inclination decay rate on i, so we first simply replace e by (e2 + i2)1/2 in Eq. (13). With the original xPL, Eq. (15) gives outward migration for ae/h > 1.1. Cresswell et al. (2007) also found that the sign of the torque changes to positive (gtid1 < 0) when ae/h > 1.6, but did not find any indication of outward migration. This means that the absolute value of the second term in the right hand side of Eq. (15) is always larger than the first term for large e. It is unclear whether outward migration really occurs or not and we need to wait for future studies. In the present paper, we use the modified x (Eq. (13)) to reproduce similar dependences of e and a on e and i to those obtained in Cresswell – 11 – et al. (2007). The migration and damping time scales (e/e and a/a) with various e are shown in Fig. 1. Although there are some differences between the time scales shown in Fig. 1 and those from Cresswell et al. (2007), these do not result in significant differences in our N-body simulations unless the time scales become too large or negative. Practically, large protoplanets obtain large e when they are trapped in secular resonances. If outward migration really occurs, protoplanets can be removed from secular resonances which migrate inward. 2.2.3. Global nebular force and secular resonances We calculate the gravity of the gas disk after Nagasawa et al. (2000). The r and z compo- nents of the nebular gravitational force are given as r′ r ρgas(r′, z′) fglob,r(r, z) = −2GZ ∞ −∞Zr′ p(r + r′)2 + (z − z′)2 ×" r2 − r′2 − (z − z′)2 (r − r′)2 − (z − z′)2 E(k) + K(k)# dr′dz′ ρgas(r′, z′)r′(z − z′) fglob,z(r, z) = −4GZ ∞ p(r + r′)2 + (z − z′)2 " −∞Zr′ k = s (r + r′)2 + (z − z′)2 . 4rr′ E(k) (r − r′)2 + (z − z′)2# dr′dz′, (16) (17) where K(k) and E(k) are elliptic integrals of the first and second kind, and k is given by These components of the force are initially tabulated in a (r, z) grid, and interpolated in sim- ulations. A secular resonance occurs when precession rates of two orbiting bodies coincide (Heppenheimer, 1980; Ward, 1981; Nagasawa et al., 2000). We analytically calculate preces- sion rates of planetesimals and the gas giant planets for e, i ≪ 1, following Appendix A of Nagasawa et al. (2000). Then, we obtain the locations of the secular resonances with Jupiter (ν5) and Saturn (ν6) (e.g., Fig. 2). 2.3. Input parameters We adopt four different initial spatial distributions of planetesimals, four different dissipa- tion time scales of the nebular gas, and four different types of orbits of Jovian planets. In total, therefore, we made 64 runs with different combinations of parameters. Each run takes one to three cpu months until completion of accretion. – 12 – 2.3.1. Initial spatial distribution of planetesimals We initially place 2000 equal-mass planetesimals between 0.5 and 4.0 AU. The initial surface density distribution of planetesimals is given by Σsolid = Σsolid,0(cid:18) r 1 AU(cid:19)−p , (for 0.5 AU ≤ r ≤ 4.0 AU) (18) where Σsolid,0 is Σsolid at 1 AU. We adopt p = 1 and 2 in simulations. The initial total solid mass is given as MT =Z 4.0 AU 0.5 AU 2πrΣsoliddr. (19) , Σsolid,0 = 6.1 and 10.2 gcm−2 for p = 1 We adopt MT = 5 and 10 M and 2, respectively. For comparison, the minimum mass solar nebula (MMSN) by Hayashi et al. (1981) has Σsolid,0 = 7.1 gcm−2, p = 1.5, and MT = 4.3 M . In the case of MT = 5 M ⊕ ⊕ . ⊕ 2.3.2. Dissipation time scale of nebular gas In the present paper, we limit ourselves to the case where the surface density of gas dissi- pates exponentially in time and uniformly in space: Σgas(r, t) = Σgas,0(cid:18) r 1 AU(cid:19)−p exp − t τdecay!, (for 0.1 AU ≤ r ≤ 36.0 AU) (20) where Σgas,0 is Σgas at 1AU and t = 0 and τdecay is the time scale for gas decay. We adopt four different decay times: τdecay = 1, 2, 3, and 5 Myr. We assume that Σgas,0 = 2000 and 3366 gcm−2 for p = 1 and 2, respectively. The gas-to-solid ratio is constant between 0.5 AU and 4.0 AU in the beginning of simulations and is 329 and 165 for MT = 5 and 10 M , respectively. For comparison, MMSN has Σgas,0 = 1700 gcm−2 (Hayashi et al., 1981). We do not consider a gap around Jupiter's orbit in the present paper. The formation of a gap does not significantly alter the time of the sweeping secular resonance's passage in the gaseous region (Nagasawa et al., 2005). However, orbital evolutions of asteroidal bodies in a gas-free gap will be different, in particular, if a gap is very wide. ⊕ 2.3.3. Orbits of Jovian planets We introduce Jupiter and Saturn after τdecay from the beginning of a simulation, assuming that they form via core accretion (Bodenheimer and Polack, 1986; Ikoma et al., 2000; Lissauer et al., 2009). Rapid formation of Jovian planets via disk instability is unlikely for our solar system because efficient cooling for clump formation is only possible at r > 100 AU (Boley, 2009). – 13 – We define aJ and eJ to be the semimajor axis and the orbital eccentricity of Jupiter, and aS and eS to be the corresponding orbital elements for Saturn. We adopt four types of orbits for Jupiter and Saturn at their formation and use the same notation introduced in O'Brien et al. (2006) and Raymond et al. (2009), although we have slightly different orbits: • EJS ("Eccentric Jupiter and Saturn"). Jupiter and Saturn are placed on their current orbits: aJ = 5.20 AU, eJ = 0.048, aS = 9.55 AU, and eS = 0.056. • CJS ("Circular Jupiter and Saturn"). These are the initial conditions used in the Nice model (Tsiganis et al., 2005). Jupiter and Saturn are radially more confined with circular orbits: aJ = 5.45 AU, eJ = 0.0, aS = 8.18 AU, and eS = 0.0. The mutual inclination is 0.5 degrees. • EEJS ("Extra Eccentric Jupiter and Saturn"). The same as EJS except a higher eccentric- ity of Jupiter: eJ = 0.1. • CJSECC. Jupiter and Saturn have the CJS semimajor axes with higher eccentricities: eJ = 0.05 and eS = 0.05. The EJS and EEJS simulations assume that Jupiter and Saturn formed at their current locations. On the other hand, the Nice model suggests that Jupiter migrated inward and Sat- urn migrated outward from their original CJS orbits via interactions with Kuiper Belt objects leading to a sudden change to the EJS orbits when Jupiter and Saturn crossed their 1:2 orbital resonance (Tsiganis et al., 2005). Asteroidal and cometary bodies strongly perturbed by this event are considered to have caused the Late Heavy Bombardment at ∼ 3.85 billion years ago; well after formation of the terrestrial planets (Gomes et al., 2005; Strom et al., 2005). Initially higher eJ and eS in the EEJS and CJSECC simulations than those in the EJS and CJS simula- tions, respectively, are considered because eJ and eS decrease via interaction with planetesimals (Sec. 3). We only take into account the gravity of other bodies and the global nebular force on motion of Jupiter and Saturn. We neglect any further interaction between a gas disk and the giant planets. Details of orbital evolution of a giant planet, which opens a gap in a gas disk, have a complicated dependence on disk properties, such as shape of a gap and disk mass, for which explicit and reliable formulae are not yet available. Nevertheless, some basic trends, which should be included in future studies, started to be clarified as follows. Analytic (Goldreich and Sari, 2003; Ogilvie and Lubow, 2003) and numerical (D'Angelo et al., 2006) studies indicate that eccentricity growth for a Jupiter-mass planet occurs together with the saturation of the corotation resonance in a low-viscosity disk. The migration time scale of a Jovian planet is much longer than the viscous diffusion time scale and even outward migration occurs, either when the gap is clean and the orbit is eccentric (D'Angelo et al., 2006) or when the gap is not deep enough (Crida and Morbidelli, 2007). A long migration time scale of Jupiter due to disk-sculpturing by Saturn is also pointed out by Morbidelli and Crida (2007). – 14 – 3. Results 3.1. Time evolution 3.1.1. Case with EJS orbits [Figs. 2- 4] Figures 2-4 show an example of the time evolution for the case with the present orbits of Jupiter and Saturn: snapshots on the plane of the semimajor axis versus the orbital eccentricity (Fig. 2), evolution of the total masses and the mean eccentricities of small and big particles (Fig. 3), and evolution of the masses and orbital excursions of surviving planets (Fig. 4). The gas decay time is tdecay = 1 Myr, the initial total mass is 5 M , the power-law index for the surface density is p = 2. In the present paper, we distinguish between a big particle, or (proto)planet, defined as a body with m > 2 × 1026 g (∼ 0.03 M ), and smaller bodies or planetesimals. This border mass is slightly smaller than Mercury's mass (3.3 × 1026 g), and one order of magnitude smaller than the isolation mass at 1AU (Kokubo and Ida, 2002). ⊕ ⊕ ⊕ We follow growth of protoplanets without Jupiter and Saturn, until tdecay from the begin- at 105 yr; see Fig. 3) by the ning of the simulation. We first observe the mass loss (∼ 0.2 M migration of the smallest planetesimals due to gas drag at the inner most region. Protoplanets grow rapidly in the inner regions while many smaller bodies remain unaccreted in the outer regions. Protoplanets first tend to collide with smaller planetesimals because their orbital ec- centricities are smaller. At 1 Myr, the smallest particles significantly deplete at r < 2 AU, while some larger planetesimals still survive there (Fig. 2). Once a planet reaches a certain size, it starts rapid inward migration (Type I migration). Since the growth time is shorter for smaller r, significant mass depletion first occurs in the inner disk. With increasing time, the mass deple- tion propagates outward (see Ida and Lin (2008) for more details). The mass loss due to Type I migration is easily recognized in the upper panel of Fig. 3, as we see stepwise depletion in the total mass. More than half of the total mass is lost by this mechanism until 3 Myr. The half of the total mass corresponds to the initial total mass inside of ∼1.5 AU. The three planets, which survive till the end, are only lunar-size at 1 Myr and located outside of 1.2 AU (Fig. 4). There are other larger bodies located inside of 1.2 AU at 1 Myr, but they will not survive Type I migration. At 1 Myr, we introduce Jupiter and Saturn with their present orbital elements. Their for- mation causes a sudden increase of orbital eccentricities of nearby particles, particularly those in mean and secular resonances, while such a jump does not appear for inner planets (see the lower panel of Fig. 3). The locations of the ν5 and ν6 resonances at t = 1 Myr are 3.10 and 3.34 AU, respectively. These resonances, in particular ν5, effectively enhance orbital eccentricities of planetesimals and protoplanets, and these bodies migrate inward as the gas drag and the tidal interaction damp their eccentricities. At 3-5 Myr, the mean eccentricities of big and small bod- ies are comparable (Fig. 3) as many protoplanets are also involved in the ν5 resonance. Near the resonance, the surface density of solid material is highly enhanced and relative velocities of solid bodies are small due to orbital alignments (Nagasawa et al., 2005). Consequently, pro- toplanets grow very quickly (Fig. 4). Collisions between protoplanets also commonly occur at – 15 – this stage. When the gas density becomes so low that inward migration of planetesimals and proto- planets due to the gas drag or the tidal interaction is slower than the migration rate of the secular resonance, these bodies are no longer swept by the resonance (Nagasawa et al., 2005). After the passage of the ν5 resonance, orbital eccentricities of planets are reduced first by tidal interaction with a remnant gas disk. The damping due to gas is effective as long as the damping time is shorter than the gas decay time (τdamp < τdecay). For an Earth-size planet at 1 AU, this condition is satisfied until ∼ 7–8 τdecay. Indeed, we see a rapid decrease of the mean eccentricity of big bodies at 5-7 Myr in Fig. 3. Later, the damping due to dynamical friction of planetesimals fur- ther reduces the eccentricities of planets. Its effect is determined by the total mass and the mass distribution of planetesimals (Morishima et al., 2008). Small bodies contain ∼ 20% of the total mass at 10 Myr and this amount is likely to be sufficient to achieve energy equipartition between big and small bodies as long as the planet-planet interaction is negligible. The typical size of remnant planetesimals is the lunar size or slightly smaller. There are only a few remnant plan- etesimals with the initial size because the smallest planetesimals had small orbital eccentricities due to strong gas drag and efficiently merged with larger bodies. The equilibrium eccentricity of Earth-size bodies due to dynamical friction of lunar-size planetesimals is estimated to be ∼ 0.03 (Morishima et al., 2008), and we obtain a quite consistent value after most of the remnant planetesimals merge with planets. while both of other two have 0.3 M Through the above processes, three planets form near 1 AU at the end of this simulation. The largest planet has 1.4 M . The locations of these three planets resemble those of the current terrestrial planets (except we do not have a Mercury analog), although the masses are somewhat different. Orbital eccentricities are as small as those for the current planets. The most serious issue for this simulation is that the last of the potentially Moon-forming impacts occurs too early (at 13.5 Myr; see Sec. 3.2 for the definition) when compared with that suggested from isotope records (Touboul et al., 2007). ⊕ ⊕ 3.1.2. Case with CJS orbits Figures 5-7 show the time evolution for the case with circular orbits of Jupiter and Saturn, which are given the initial configuration of the Nice model. Except for the orbits of the gas giant planets, all other parameters are the same as those used in the EJS simulation in Sec. 3.1.1 (Figs. 2-4). If the orbits of the gas giant planets are nearly circular, the effects of the secular resonances are much weaker than the case of the EJS orbits. Consequently, a large fraction of particles stay in the asteroidal region for a long period of time. Nevertheless, the orbital eccentricities of the gas giant planets can not be completely zero when there are two (or more) gas giant planets. Therefore, resonances gradually sculpture orbits of particles in the asteroidal region. Some enhancements at the 3:1 mean motion resonance (2.6 AU) and the ν6 secular resonance (3.2 AU) are seen at 3 Myr (Fig. 5). These enhancements gradually propagate to non-resonant locations via mutual gravitational interactions (Raymond et al., 2006). The mass [Figs. 5- 7] – 16 – transfer from the asteroidal region to the terrestrial region slowly occurs as excited asteroidal bodies encounter and collide with inner planets. At 3 Myr, there are about 15 Mars-size protoplanets in the terrestrial region. The fraction of small particles is very small in the terrestrial region whereas small particles are still dominant in mass in the asteroidal region. This situation is similar to some of the initial conditions adopted in previous N-body simulations (e.g., Raymond et al., 2006). In our simulation, however, the mean orbital separation between protoplanets is ∼ 20 Hill radius at 3 Myr while previous late stage simulations usually assume the orbital separation to be ∼ 10 Hill radius, based on the oligarchic growth model of Kokubo and Ida (1998). The larger separation in our simulation is due to Type I migration, which makes inner protoplanets start to migrate inward earlier than outer planets (McNeil et al., 2005). With the large orbital separation and the orbital stabilization due to the nebular gas, mutual collisions between protoplanets are rare until ∼ 10 Myr. After 10 Myr, mutual collisions between protoplanets start to occur, and the orbital separa- tion increases as the number of protoplanets decreases. Strictly speaking, the orbital separation normalized by the mutual Hill radius, and averaged over pairs of planets, increases nearly log- arithmically with time (from ∼ 20 at 3 Myr to ∼ 60 at 800 Myr), and this is probably because the stable time scale of a planetary system exponentially increases with the normalized separa- tion (Chambers et al., 1996; Ito and Tanikawa, 1999; Yoshinaga et al., 1999). Through giant impacts, an Earth-size planet forms slightly inside of 2 AU. This planet highly enhances orbital eccentricities of asteroidal bodies and their number gradually decreases by either hitting this planet or escaping from the system after encounters with Jupiter. There are still a large number of remnant planetesimals at 100 Myr, but not at 300 Myr (Figs. 5 and 6). Therefore, damping due to planetesimals is expected until giant impacts com- plete at around 100 Myr. In this simulation, the last potentially Moon-forming impact occurs at 139.5 Myr on the outermost planet (Fig. 7). Nevertheless, the mean orbital eccentricity in the end of the simulation (∼ 800 Myr) is not small, as the two middle planets are still strongly inter- acting (Fig. 7). These two planets are stirred by the innermost and outermost planets which have larger masses than the middle planets. This simulation succeeded in reproducing the very late giant impact, but failed to reproduce the nearly circular orbits and the radial mass concentration of the current system. 3.2. Comparison with constraints In this section, we examine the properties of planetary systems formed in all runs and compare them with those for our own system. We consider the following three properties of systems (see also Raymond et al., 2009). 1. The spatial mass distribution. The radial mass concentration statistic, S c, is defined as – 17 – (Chambers, 2001) S c = max P m j P m j log10(a/a j)! , (21) where m j and a j are the mass and the semimajor axis of each planet, and the subscript j is given in the order of a j. The function in the parenthesis is calculated for a throughout the terrestrial planet region, with S c being the maximum value. This quantity increases with radial mass concentration. We use two additional parameters as constraints. The first one is the minimum orbital separation between neighboring planets bmin normalized by their mutual Hill radius rH, j = 0.5(a j + a j+1)[(m j + m j+1)/(3M )]1/3: ⊙ bmin = min a j+1 − a j rH, j ! . The second being the mass weighted mean semimajor axis of planets am = P a jm j MT,final , (22) (23) where MT,final = P m j is the final total mass of planets. For the terrestrial planets, S c = 89.9, bmin = 26.3 (the normalized orbital separation between the Earth and Venus), and am =0.90 AU. 2. The deviation from circular and coplanar orbits. The orbital excitation of terrestrial plan- ets is characterized by the angular momentum deficit, S d (Laskar, 1997): S d = P m j √a j(cid:18)1 − cos(i j)q1 − e2 j(cid:19) P m j √a j  , (24) where e j and i j are the orbital eccentricity and inclination of each planet with respect to the fiducial plane of an initial planetesimal disk. We also measure the mass weighted mean eccentricity: em = P e jm j MT,final . (25) We use the orbital elements averaged over a few Myr. For the terrestrial planets, S d = 0.0018 and em = 0.038 (Quinn et al., 1991). 3. The timing of the Moon-forming impact. Based on simulations for giant impacts and accretion of moons (Cameron and Benz, 1991; Ida et al., 1997; Canup et al., 2001; Canup, 2004; 2008), we roughly regard an impact as potentially Moon-forming, if a) the total mass of the impactor and the target is > 0.5 M , b) the impactor's mass (< the target's mass) is > 0.05 M , and c) the impact angular momentum is > 0.5 LEM, where LEM is the current total angular momentum of the Earth-Moon system (i.e., the Earth's spin angular momentum and the Moon's orbital angular momentum). We define the time of the last ⊕ ⊕ – 18 – potentially Moon-forming impact in a run to be timp. The time of the actual Moon-forming impact is estimated as timp = 50-150 Myr (Touboul et al., 2007; All´egre et al., 2008). 2 The small content of iron in the Moon is explained when the impact velocity is as slow as the mutual escape velocity (Canup, 2004, 2008); this constraint will be also discussed (Figs. 13 and 16). ⊕ Raymond et al. (2009) considered two additional constraints. The first one is the absence of planetary bodies (> 0.05 M ) in the asteroidal region (> 2 AU). They showed that bodies which are too massive destroy the observed radial variation of the asteroidal taxonomic types (Gradie and Tedesco, 1982). If massive bodies exist in the asteroidal region, this system usually has too large am or too small S c, because we do not exclude a massive body in the asteroidal region to be defined as a planet. Therefore, this constraint is automatically included in our first constraint. Their second additional constraint is the Earth's water content. They showed that the mass supply from asteroidal bodies to terrestrial planets decreases with increasing eccentricity of Jupiter. As a result, the water content of an Earth-analog is insufficient in their EJS and EEJS simulations. On the other hand, in our EJS and EEJS simulations with the nebular gas, a large fraction of mass of an Earth-analog comes from the asteroidal region, while protoplanets which formed from planetesimals originally in the terrestrial region spiral into the Sun by Type I migration (Sec. 3.1.1). Probably, the problem here is rather that we need to consider how to remove excess water from our Earth-analogs. 3.2.1. Comparison between EJS and CJS simulations The properties of the planetary systems from the EJS and CJS simulations are summarized in Table 1. Also, Figures 8 and 9 show the snapshots on the a-e plane at the end of all EJS and [Table 1] CJS simulations, respectively. For both EJS and CJS simulations, the total mass and the typical size of planets decrease with increasing gas decay time τdecay because of Type I migration. When the surface density is large, protoplanets rapidly grow and then spiral into the Sun. Therefore, the final total mass MT,final depends on the initial surface density only weakly for a certain τdecay. This trend is particularly evident when τdecay ≥ 2 Myr, because the mass depletion due to Type I migration propagates to the outer asteroidal region for large τdecay. The number of planets Np is usually 3 or 4. In the EJS simulations with τdecay = 5 Myr, however, Np is smaller; in these simulations, the ν5 resonance effectively sweeps most of the remnant mass to its current location (0.61AU), and a planet with dominant mass (0.3 − 0.5 M [Figs. 8 and 9] ) forms there. ⊕ We plot the final planets on the plane of semimajor axis versus mass (Fig. 10). The crosses on the figure represent the current terrestrial planets. The EJS simulations clearly have better 2Smaller timp(≃ 30 Myr) is not ruled out from the terrestrial W/Hf records because the degree of metal-silicate equilibration at the Moon-forming impact is still in debate (Jacobsen et al., 2009). In this case, however, one needs an alternative interpretation for the prolonged differentiation of the Moon (Touboul et al., 2007), other than the late Moon-forming giant impact. [Fig. 10] – 19 – radial mass concentration than the CJS simulations. In the CJS simulations, the effect of the ν5 resonance is negligible so that we inevitably have a massive planet around 2 AU, quite in contrast to the EJS simulations. In both EJS and CJS simulations, sizes of outer planets are smaller when the initial surface density in the asteroidal region is small, as is the case for small MT or large p. In addition, for the EJS simulations, the radial mass concentration is indirectly affected by the initial spatial distribution of planetesimals via their interactions with Jupiter. Asteroidal particles reduce eJ either by angular momentum exchange at the ν5 resonance or by energy exchange in close encounters with Jupiter; in both cases, the decrease in eJ is given as ∆eJ ∼ m/(mJeJ), where mJ is Jupiter's mass. With decreasing eJ, the effect of the ν5 resonance decreases (Nagasawa et al., 2005). If the total mass in the asteroidal region is small due to small MT, large p, or large τdecay, the decrease in eJ is small (see final eJ in Table 1). Such systems can obtain high radial mass concentration since Jupiter's perturbation remains effective. This accounts for the large difference in the radial mass concentration between cases with MT = 5 and 10 M in the EJS simulations (Fig. 10). The ν6 resonance is also important for the radial mass concentration, and it also works effectively when eS is large (we will discuss its effect at Fig. 17 more in detail). Since Saturn's eccentricity eS is primarily determined by the eccentricity forced by Jupiter, eS is usually comparable to or slightly larger than eJ. ⊕ Figure 11 shows the plot of the final planets on the plane of mass versus orbital eccentric- ⊕ ⊕ ⊕ but not in those with MT = 10 M ity. The small orbital eccentricities are well reproduced in the EJS simulations with MT = 5 [Fig. 11] . This difference is closely related to the radial mass M concentration. In the case with MT = 5M , the ν5 resonance effectively sweeps most of the bodies and giant impacts between protoplanets complete early while there remains a sufficient amount of gas and planetesimals. On the other hand, the systems starting with MT = 10 M ⊕ radially expand even after the secular resonance passage and Jupiter continuously perturbs the outer planets. These planets eventually trigger orbital instability and cause giant impacts after both gas and planetesimals significantly deplete. Giant impacts tend to occur very late in the CJS simulations. However, small asteroidal bodies remain for a long time here as well and tend to damp eccentricities of the planets. This is why planets in CJS simulations have moderately small eccentricities both for MT = 5 and 10 M . For the CJS simulations, we found a simple correlation between the mass weighted mean orbital eccentricity, em, and the mass weighted mean semimajor axis, am (Fig. 12). The longer the gas decay time, the larger am, because mass depletion due to Type I migration propagates outward with time. Since Jupiter perturbs outer planets more strongly than inner planets, we have this simple correlation. The CJS simulations always produce too large a value for am, while the EJS simulations well reproduce the current value of am except for runs with τdecay = 5 Myr where am becomes too small. [Fig. 12] ⊕ Figure 13 shows angular momenta and impact velocities of all potentially Moon-forming impacts as functions of time. The most important result is that giant impacts occur most com- monly at ∼ 10 Myr in the EJS simulations whereas at ∼ 100 Myr in the CJS simulations. In the EJS simulations, some early giant impacts occur between protoplanets trapped in the ν5 resonance. It is more common, however, that giant impacts occur slightly later when orbital eccentricities of planets have sufficiently damped after passage of the resonance so that mutual gravitational attraction is enhanced. In the CJS simulations, giant impacts usually occur after [Fig. 13] – 20 – gas completely depletes and planetesimals in the asteroidal region also significantly (but not completely) deplete. The time of the last Moon-forming impact in each simulation is shown in Table 1. In fact, last giant impacts in some of the EJS simulations occur very late. These systems, however, inevitably have either large eccentricities or low radial mass concentrations. When we see a single simulation in Fig. 13, the impact velocity, vimp, is as small as the mutual escape velocity vesc at early time, but vimp/vesc increases with time so that its value for the last impact is notably larger than unity. Some simulations, however, have final impacts with vimp/vesc ≃ 1, which is favorable for the small iron content of the Moon (Canup, 2004, 2008). When the orbital separation of two planets, normalized by the mutual Hill radius, slowly decreases due to their growth or weak perturbation from other planets, an impact between them with vimp/vesc ≃ 1 can occur after the Jacobi integral in Hill's equation slightly exceeds the potential at the Hill's sphere. Orbits of other protoplanets are affected only weakly during the time that two colliding protoplanets interact. On the other hand, if three or larger number of planets interact strongly, their relative velocity increases to the escape velocity. In this case, the √2. Further higher vimp/vesc are possible between small typical impact velocity is vimp/vesc ≃ protoplanets stirred by larger planets or if Jupiter's perturbation is strong. The former type of slow giant impacts usually occur in early stages while the latter type of high-velocity giant impacts occur after depletion of both gas and planetesimals (Fig. 13). Systems with S d and em as small as those for the current system do not usually experience the latter type of impacts. Therefore, not only the small orbital eccentricities of the Earth and Venus but also the small iron content of the Moon are likely to support the presence of some damping forces at the time of the Moon-forming impact. 3.2.2. Cases with larger eccentricities of giant planets: EEJS and CJSECC simulations As shown in Table 1, eJ (and eS) decreases via interactions with particles in the asteroidal region. Remnant planetesimals orbiting outside of Jupiter's orbit may also reduce eJ, although we do not include these bodies in our simulations. Therefore, the initial eJ and eS are likely to have been larger than those adopted in the EJS simulations, if their initial semimajor axes were (nearly) the same as the present values. On the other hand, if their initial semimajor axes were those given in the Nice model, eJ and eS might have been nearly zero throughout the terrestrial planet formation, but larger initial eJ and eS are not excluded. The summary of simulations with larger initial eJ and eS, namely the EEJS and CJSECC simulations, is given in Table 2. The [Table 2] snapshots of planets in the end of simulations in the a-e plane are also shown in Figs. 14 (the EEJS runs) and 15 (the CJSECC runs). [Figs. 14 and 15] In the EEJS simulations, the ν5 resonance is so strong that am usually becomes too small. The number of planets is typically one or two, because protoplanets experience large orbital excursions. None of the simulations have S c as large as the present value, because the semimajor axis of the innermost planet is usually too small. The innermost planet is pushed toward the Sun by outer planets which are swept by the ν5 resonance. The potentially Moon-forming impacts – 21 – usually occur too early and impact velocities are too high (Fig. 16). Clearly, the initial eJ of our EEJS simulations is too high. If eJ is somewhat smaller but still larger than the present value, the ν5 resonance may work much better (Thommes et al., 2008). [Fig. 16] The initial orbital eccentricities of Jupiter and Saturn in the CJSECC simulations are sim- ilar to those in the EJS simulations. The orbital separation between Jupiter and Saturn in the CJSECC simulations is narrower than that in the EJS simulations. This makes the ν5 and ν6 resonances in the CJSECC simulations more distant from the Sun than those in the EJS sim- ulations. Because of this, the averaged am and timp in the CJSECC simulations become larger than those in the EJS simulations (Tables 1 and 2). Figure 17 shows a comparison between a CJSECC and an EJS simulation with the same parameters except for the orbits of Jupiter and Saturn. The ν5 resonance sweeps planetesimals effectively at 3 Myr but no longer works at 5 [Fig. 17] Myr in both cases. In the EJS simulation, however, the ν6 resonance (∼ 2 AU) still sweeps planetesimals toward the terrestrial region. On the other hand, the location of the ν6 resonance in the CJSECC simulation is too far (∼ 3 AU) to contribute to radial mass concentration. In the CJSECC simulation, accretion of outer bodies, which survived the sweeping by the ν5 res- onance, proceeds slowly and this makes the final giant impact late. Surprisingly, 7 out of 10 CJSECC simulations in which potentially Moon-forming impacts occur satisfy the time of the last giant impact suggested from isotope records (Table 2). However, vimp/vesc's of most of these late impacts are too high (Fig. 16) and also S c's of these systems are too small (Table 2). As we have seen, simulation outcomes strongly depend on Jupiter's eccentricity, eJ. Through interactions with small bodies, eJ decreases. Clearly, the effect of Jupiter's perturbation is more important if eJ in the late stage is larger regardless of initial values of eJ. In Fig. 18, we plot the angular momentum deficit S d, the radial mass concentration parameter S c/a2 m, and the time of the last one of potentially Moon-forming impacts timp versus the value of eJ at the end of a simulation, eJ,final (averaged over a few Myr). We find that S c/a2 m more appropriately represents the radial mass confinement than does S c because systems consisting of small planets with large a tend to obtain large S c while bmin is large (particularly in the CJS simulations). Simple linear χ2 fits suggest that S c/a2 m and timp increases/decreases with eJ,final, although the scatter of the simulation data are very large. The problem is that S c/a2 m can be as large as the present value when eJ,final is large whereas timp can be as large as that suggested from isotope records only when eJ,final is small enough. For S d, we fit the data with a quartic function as two competing effects are expected: large eJ means stronger perturbation while quick accretion due to large eJ means stronger damping forces due to remnant gas and planetesimals. We find that S d is nearly flat for small eJ,final and increases with increasing eJ,final when eJ,final is larger than the present value. In an ensemble average sense, S d is usually much larger than the current value, although some simulations have S d consistent or even smaller than the present day value. The dependence of all three properties on eJ,final is stronger for the CJS+CJSECC simulations than for the EJS+EEJS simulations. It is unclear to us, however, whether this is really physically meaningful because fits for the CJS+CJSECC simulations are not very reliable with only a few points at large eJ. [Fig. 18] – 22 – 4. Discussion m in the current system is due to the small masses of Mars and Mercury and the [Fig. 19] None of our simulations succeeded in satisfying all of the three constraints listed in Sec. 3.2. The most serious issue we found is a trade-off between the radial mass concentration parameter, S c/a2 m, and the time of the potentially Moon-forming impact, timp (the upper panel of Fig. 19). The large S c/a2 small orbital separation between the Earth and Venus. It is hard to satisfy both of these condi- tions in a system with large timp. Since the evolution time scale is longer in the outer regions, giant impacts can occur late when a large amount of mass stays in the outer region for a long time. Such a system usually has outer planets more massive than Mars. Also, if a planetary system slowly evolves with weak perturbations from Jupiter (e.g., the CJS simulations), the orbital separation between neighboring planets normalized by the mutual Hill radius inevitably becomes larger than that between the Earth and Venus (26.3) owing to the mutual orbital re- pulsion. If eJ is as large as the present value or larger, S c/a2 m can be as large as today's value because Jupiter tends to radially compress a system. However, accretion completes too rapidly in such a case. There is also a trade-off between the angular momentum deficit S d and timp (the lower panel of Fig. 19). Namely, in order to obtain small S d, a final large impact needs to occur while some damping forces work effectively. Otherwise, S d ends up with much larger than the current value. This trend is particularly clear for the EJS and CJS simulations (filled and open circles). Small S d even without any damping forces seems to be possible but such a case is statistically rare. Our failure indicates that there may be some missing physics in our simulations or flaws with the assumptions that we have adopted. In the following, we suggest three possible scenar- ios/solutions for future studies. In the case with an eccentric Jupiter, a possible solution is that τdecay is much longer and Type I migration is much slower than we assumed. This situation actually corresponds to that adopted in the original dynamical shake up model of Nagasawa et al. (2005) and Thommes et al. (2008). In their simulations (τdecay = 3 and 5 Myr), the giant impacts usually occur at t ≃ 4 − 5 τdecay when the ν5 resonance reaches near 1 AU. Therefore, τdecay > 10 Myr is required for the late Moon-forming impact indicated from isotope records (timp > 50 Myr; Touboul et al. 2007). If τdecay = 10 Myr, the gas density is still ∼ 1% of the minimum mass solar nebular at 50 Myr, and observations suggest that such a gas disk is exceptionally long-living (Pascucci et al. 2006). Slightly smaller timp(≃ 30 Myr) is not excluded from the terrestrial W/Hf records only (Jacobsen et al., 2009), but this still requires a long-surviving disk (τdecay ∼ 6 Myr). Recent theories indicate significant changes in Type I migration rate due to the non-isothermal effect (Baruteau and Masset, 2008; Paardekooper and Mellema, 2008) and the non-linear effect of the co-rotation resonance (Masset et al., 2006; Paardekooper and Paparoizou, 2009). These mechanisms cause slow down of inward migration or even outward migration if the radial entropy gradient is negative for the former mechanism or if the radial surface density gradient is nearly flat or positive for the latter one. Therefore, if a local density maximum exists, planetary migration – 23 – slows down or protoplanets even tend to migrate toward this point (Kretke et al., 2009). A local density maximum can exist at the boundary between the optically thin inner region and the optically thick outer region (i.e., the inner boundary of the so-called dead-zone; Kretke and Lin, 2007; Brauer et al., 2008). Protoplanets might have survived in the protosolar nebular for a long time with such mechanisms. An alternative possibility for the case with an eccentric Jupiter is that the nebular gas inside of Jupiter might have dissipated very rapidly after the formation of Jupiter. This situation corresponds to previous N-body simulations for the late stages without gas (Chambers, 2001; O'Brien et al., 2006; Raymond et al., 2009). Without gas, timp tends to be larger than those in our simulations, because the effect of ν5 is less signifiant. Non-uniform gas dissipation models, such as inside-out and gap-opening models discussed in Nagasawa et al. (2000), probably have similar outcomes, because ν5 passes a certain location after the gas dissipates there. Raymond et al. (2009) showed that some of their EEJS simulations nearly satisfy all the three constraints listed in Sec. 3.2. In fact, these simulations satisfy the small mass of Mars, but not the small orbital separation between the Earth and Venus. This is why the values of S c in all of their simulations are smaller than the current value. This problem might be resolved, however, if they adopt even higher resolution with smaller planetesimals (e.g., Morishima et al., 2008) and inelastic rebounds for high-speed or grazing collisions (Agnor and Asphau, 2004; Asphau et al., 2006). Because these effects are likely to enhance the damping effects and reduce the orbital excursions of planets, the orbital separation between neighboring planets will probably be smaller. Very quick or non-uniform dissipation of gas seems to be favorable, at least, in the asteroidal region in the case with large eJ. As we have shown, the ν5 resonance is so strong in the asteroidal region, except the CJS simulations, that the radial compositional structure in the asteroidal region is destroyed even for τdecay = 1 Myr. The gas dissipation of the order of 0.1 Myr or shorter seems to be necessary for the uniform dissipation model (Nagasawa et al., 2000). Jupiter might have opened a wide gap around its orbit, extending to the entire asteroidal region immediately after its formation (however, τdecay = 1 Myr or larger may be still possible if eJ is small, when ν5 passes the asteroid region, and increases later). In the case of a circularly orbiting Jupiter, the initial planetesimals needed to be radially highly concentrated in order to satisfy a large S c. In other words, the depletion of mass in the asteroidal region needed to occur before the formation of planetesimals. Such a concentration of planetesimals may be possible if planetesimal formation occurs in radially limited regions after the increase of the surface density due to dust migration (Youdin and Shu, 2002; Youdin and Chiang, 2004; Kretke and Lin, 2007; Brauer et al., 2008). Morishima et al. (2008) performed N-body simulations starting from highly concentrated disks in the absence of gas, and found that some of the resulting planetary systems were similar to the current solar system. The problem of their simulations is that accretion usually completes very quickly ( < 10 Myr), as well as our EJS simulations. Nevertheless, one simulation (run 1 of Morishima et al., 2008) has a very late giant impact. This system unfortunately has slightly too large an orbital separation between the two largest planets. Hansen (2009) also made simulations similar to Morishima et al. (2008) and found the mean time of last giant impacts to be 45 Myr. However, the lower limit to the impactor mass (0.02 M ) used for his definition of giant impacts is probably too small ⊕ – 24 – for formation the Moon. Further explorations for planetary accretion starting from radially concentrated disks are necessary. 5. Conclusions We have performed N-body simulations which follow the growth and accretion histories of the terrestrial planets, including the effects of the giant planets and the interaction with gas in laminar disks. We vary initial total mass and spatial distribution of planetesimals, the time scale of dissipation of a gas disk, and the orbits of Jupiter and Saturn. It is a remarkable success of numerical simulations, that planetary systems with global properties similar to the terrestrial planets can be obtained starting with a proto-planetary disk. We are now at the stage where fine details of the models can be compared with solar system constraints. In order to retain a final total mass as massive as the present total mass (∼ 2 Earth mass) in the presence of standard Type I migration, the time scale of dissipation of gas needs to be 1-2 Myr. Using the results from a large number of simulations we compared the evolutionary his- tories of the resulting planetary systems with the following three properties of the terrestrial planets: (1) the high radial mass concentration, (2) the small orbital eccentricities and (3) the very late Moon-forming impact (∼ 100 Myr). The orbital eccentricities of the planets can be as small as their present values if giant impacts finish whilst there is still remnant gas or planetes- imals, as expected. The most serious issue we found was a trade off between the radial mass concentration and the late Moon-forming impact. If the orbital eccentricities of Jupiter are as large as the present value, most of bodies in the asteroidal region are swept up to the terrestrial region owing to inward migration of the secular resonance. Although this mechanism helps planetary systems to have high radial concentrations in the end, giant impacts between proto- planets most commonly occur too early (∼10 Myr). On the other hand, if the orbital eccentricity of Jupiter is close to zero as suggested in the Nice model, the effect of the secular resonance is negligible and a large amount of mass stays for a long period of time in the asteroidal region. Although giant impacts usually occur around 100 Myr, we inevitably have an Earth size planet at around 2 AU in this case. It is very difficult to obtain spatially concentrated terrestrial planets with very late giant impacts, as long as we include all the effects of gas and assume initial disks similar to the minimum mass solar nebular. Acknowledgements We appreciate comments on our manuscript from two anonymous reviewers and the ed- itor, Alessandro Morbidelli. R.M. thanks Sei-ichiro Watanabe for his hospitality during the visit of R.M. and fruitful discussions. Simulations were performed on the Zbox2 and Zbox3 supercomputers at the University of Zurich. – 25 – References Adachi, I., Hayashi, C., Nakazawa, K. 1976. The gas drag effect on the elliptical motion of a solid body in the primordial solar nebula. Prog. Theoret. Phys. 56, 1756–1771. Agnor, C.B., Asphau, E. 2004. Accretion efficiency during planetary collisions. Astrophys. J. 613, L157–L160. Agnor, C.B., Ward, W.R. 2002. Damping of terrestrial-planet eccentricities by density-wave interactions with a remnant gas disk. Astrophys. J. 567, 579–589. Agnor, C.B., Canup, R.M., Levison, H.F. 1999. On the character and consequences of large impacts in the late stage of terrestrial planet formation. Icarus 142, 219–237. All´egre, C.J., Manh`es, G., Gopel, C. 2008. The major differentiation of the Earth at ∼4.45 Ga. Earth Planet. Sci. L. 267, 386–398. Asphau, E., Agnor, C.B., Quentin, W. 2006. Hit-and-run planetary collisions. Nature 439, 7073, 155–160. Baruteau, C., Masset, F. 2008. On the corotation torque in a radiatively inefficient disk. Astro- phys. J. 672, 1054–1067. Bodenheimer, P., Pollack, J.B. 1986. Calculations of the accretion and evolution of giant planets: The effects of solid cores. Icarus 67, 391–408. Boley, A.C. 2009. The two modes of gas giant planet formation. Astrophys. J. 695, L53–L57. Brauer, F., Henning, Th., Dullemond, C.P. 2008. Planetesimal formation near the snow line in MRI-driven turbulent protoplanetary disks. Astrophys. Astron. 487, L1–L4. Cameron, A.G.W., Benz, W. 1991. The origin of the moon and the single impact hypothesis. IV. Icarus 92, 204–216. Canup, R.M. 2004. Simulations of a late lunar-forming impact. Icarus 168, 433–456. Canup, R.M. 2008. Lunar-forming collisions with pre-impact rotation. Icarus 196, 518–538. Canup, R.M., Ward, W.R., Cameron, A.G.W. 2001. A scaling relationship for satellite-forming impacts. Icarus 150, 288–296. Chambers, J.E. 1999. A hybrid symplectic integrator that permits close encounters between massive bodies. Mon. Not. R. Astron. Soc. 304, 793–799. Chambers, J.E. 2001. Making more terrestrial planets. Icarus, 152, 205–224. Chambers, J.E., Cassen, P. 2002. The effects of nebula surface density profile and giant-planet eccentricities on planetary accretion in the inner solar system. Meteo. Planet. Sci. 37, 1523–1540. – 26 – Chambers, J.E., Wetherill, G.W. 1998. Making the terrestrial planets: N-body integrations of planetary embryos in three dimensions. Icarus 136, 304–327 Chambers, J.E., Wetherill, G.W., Boss, A.P. 1996. The stability of multi-planet systems. Icarus 119, 261–268. Cresswell, P., Dirksen, G., Kley, W., Nelson, R.P. 2007. On the evolution of eccentric and inclined protoplanets embedded in protoplanetary disks. Astron. Astrophys. 473, 329– 342. Crida, A., Morbidelli, A. 2007. Cavity opening by a giant planet in a protoplanetary disc and effects on planetary migration. Mon. Not. R. Astron. Soc. 377, 1324–1336. Daisaka, J.K., Tanaka, H., Ida, S. 2006. Orbital evolution and accretion of protoplanets tidally interacting with a gas disk. II. Solid surface density evolution with Type-I migration. Icarus 185, 492–507. D'Angelo, G., Lubow, S.H., Bate, M.R. 2006. Evolution of giant planets in eccentric disks. Astrophys. J. 652, 1698–1714. Duncan, M.J., Levision, H.F., Lee, M.H. 1998. A multiple time step symplectic algorithm for integrating close encounters. Astron. J. 116, 2067–2077. Goldreich, P., Sari, R. 2003. Eccentricity evolution for planets in gaseous disks. Astrophys. J. 585, 1024–1037. Goldreich, P., Tremaine, S. 1980. Disk-satellite interactions. Astrophys. J. 241, 425–441. Goldreich, P., Ward, W.R. 1973. The formation of planetesimals. Astrophys. J. 183, 1051- 1062. Gomes, R., Levison, H.F., Tsiganis, K., Morbidelli, A. 2005. Origin of the cataclysmic Late Heavy Bombardment period of the terrestrial planets. Nature 435, 466–469. Gradie, J., Tedesco, E. 1982. Compositional structure of the asteroid belt. Science 216, 1405– 1407. Hansen, B. 2009. Formation of the terrestrial planets from a narrow annulus. Astrophys. J., 703, 1131-1140. Hayashi, C. 1981. Structure of the solar nebula, growth and decay of magnetic fields and effects of magnetic and turbulent viscosities on the nebula. Suppl. Prog. Theoret. Phys. 70, 35–53. Heppenheimer, T.A. 1980. Secular resonances and the origin of eccentricities of Mars and the asteroids. Icarus 41, 76-88. – 27 – Ida, S., Lin, D.N.C. 2008. Toward a deterministic model of planetary formation. IV. Effects of Type I migration. Astrophys. J. 673, 487–501. Ida, S., Canup, R.M., Stewart, G.R. 1997. Lunar accretion from an impact-generated disk. Nature 389, 353–357. Ikoma, M., Nakazawa, K., Emori, H. 2000. Formation of giant planets: dependences on core accretion rate and grain opacity. Astrophys. J. 537, 1013–1025. Inaba, S., Tanaka, H., Nakazawa, K., Wetherill, G.W., Kokubo, E. 2001. High-accuracy statis- tical simulation of planetary accretion: II. Comparison with N-body simulations. Icarus 149, 235–250. Ito, T., Tanikawa, K. 1999. Stability and instability of the terrestrial protoplanet system and their possible roles in the final stage of planet formation. Icarus, 139, 336–349. Jacobsen, S.B., Remo, J.L., Petaev, M.I., Sasselov, D.D. 2009. Hf-W chronometry and the timing of the giant Moon-forming impact on Earth. 40th Lunar and Planetary Science Conference abstract, A2054. Kokubo, E., Ida, S. 1996. On runaway growth of planetesimals. Icarus 123, 180–191. Kokubo, E., Ida, S. 1998. Oligarchic growth of protoplanets. Icarus 131, 171–178. Kokubo, E., Ida, S. 2000. Formation of protoplanets from planetesimals in the Solar nebula. Icarus 143, 15–27. Kokubo, E., Ida, S. 2002. Formation of protoplanet systems and diversity of planetary systems. Astrophys. J. 581, 666–680. Kokubo, E., Kominami, J., Ida, S. 2006. Formation of terrestrial planets from protoplanets. I. Statistics of basic dynamical properties. Astrophys. J. 642, 1131–1139. Kominami, J., Ida, S. 2002. The effect of tidal interaction with a gas disk on formation of terrestrial planets. Icarus 157, 43–56. Kominami, J., Ida, S. 2004. Formation of terrestrial planets in a dissipating gas disk with Jupiter and Saturn. Icarus 167, 231–243. Kretke, K.A., Lin, D.N.C. 2007. Grain retention and formation of planetesimals near the snow line in MRI-driven turbulent protoplanetary disks. Astrophys. J. 664, L55–L58. Kretke, K.A., Lin, D.N.C., Garaud, P., Turner, N.J. 2009. Assembling the building blocks of giant planets around intermediate-mass stars. Astrophys. J. 690, 407–415. Laskar, J. 1997. Large scale chaos and the spacing of the inner planets. Astron. Astrophys. 317, L75–L78. – 28 – Lissauer, J.J., Hubickyj, O., D'Angelo, G., Bodenheimer, P. 2009. Models of Jupiter's growth incorporating thermal and hydrodynamic constraints. Icarus 199, 338–350. Masset, F.S., D'Angelo, G., Kley, W. 2006. On the migration of protogiant solid cores. Astro- phys. J. 652, 730–745. McNeil, D., Duncan, M., Levison, H.F. 2005. Effects of Type I migration on terrestrial planet formation. Astron. J. 130, 2884–2899. Morbidelli, A., Crida, A. 2007. The dynamics of Jupiter and Saturn in the gaseous protoplan- etary disk. Icarus 191, 158–171. Morishima, R., Schmidt, M.W., Stadel, J., Moore, B. 2008. Formation and accretion history of terrestrial planets from runaway growth through to late time: Implications for orbital eccentricity. Astrophys. J. 685, 1247–1261. Muzerolle, J., Calvet, N., Briceno, C., Hartmann, L., Hillenbrand, L. 2000. Disk accretion in the 10 Myr old T Tauri stars TW Hydrae and Hen 3-600A. Astrophys. J. 535, L47–L50. Nagasawa, M., Lin, D.N.C., Thommes, E.W. 2005. Dynamical shake-up of planetary sys- tems. I. Embryo trapping and induced collisions by the sweeping secular resonance and embryo-disk tidal interaction. Astrophys. J. 635, 578–598. Nagasawa, M., Tanaka, H., Ida, S. 2000. Orbital evolution of asteroids during depletion of the solar nebula. Astron. J. 119, 1480–1497. O'Brien, D.P., Morbidelli, A., Levison, H.F. 2006. Terrestrial planet formation with strong dynamical friction. Icarus 184, 39– 58. Ogihara, M., Ida, S., Morbidelli, A. 2007. Accretion of terrestrial planets from oligarchs in a turbulent disk. Icarus 188, 522–534. Oglvie, G.I., Lubow, S.H. 2003. Saturation of the corotation resonance in a gaseous disk. Astrophys. J. 587, 398–406. Paardekooper, S.-J., Mellema, G. 2008. Growing and moving low-mass planets in non-isothermal disks. Astron. Astrophys. 478, 245–266. Paardekooper, S.-J., Papaloizou, J.C.B. 2009. On corotation torques, horseshoe drag and the possibility of sustained stalled or outward protoplanetary migration. Mon. Not. R. As- tron. Soc. 394, 2283–2296. Papaloizou, J.C.B., Larwood, J.D. 2000. On the orbital evolution and growth of protoplanets embedded in a gaseous disc. Mon. Not. R. Astron. Soc. 315, 823–833. Pascucci, I., and 19 co-authors, 2006. Formation and evolution of planetary systems: Upper limits to the gas mass in disks around Sun-like stars. Astrophys. J. 651, 1177–1193. – 29 – Quinn, T., Tremaine, S., Duncan, M. 1991. A three million year integration of the earth's orbit. Astron. J. 101, 2287–2305. Raymond, S.N., O'Brien, D.P., Morbidelli, A., Kaib, N.A. 2009. Building the terrestrial plan- ets: Constrained accretion in the inner solar system. Icarus 203, 644–662. Raymond, S.N., Quinn, T., Lunine, J.I. 2004. Making other earths: dynamical simulations of terrestrial planet formation and water delivery. Icarus 168, 1–17. Raymond, S.N., Quinn, T., Lunine, J.I. 2006. High-resolution simulations of the final assem- bly of Earth-like planets I. Terrestrial accretion and dynamics. Icarus 183, 265–282. Richardson, D.C., Quinn, T., Stadel, J., Lake, G. 2000. Direct large-scale N-body simulations of planetesimal dynamics. Icarus 143, 45–59. Stadel, J. 2001. Cosmological N-body simulations and their analysis. PhD dissertation, Univ. of Washington, Seattle. Strom, R.G., Malhotra, R., Ito, T., Yoshida, F., Kring, D.A. 2005. The origin of planetary impactors in the inner solar system. Science 309, 1847–1850. Tanaka, H., Ward, W.R. 2004. Three-dimensional interaction between a planet and an isother- mal gaseous disk. II. Eccentricity waves and bending waves. Astrophys. J. 602, 388–395. Tanaka, H., Takeuchi, T., Ward, W.R. 2002. Three-dimensional interaction between a planet and an isothermal gaseous disk. I. Corotation and Lindblad torques and planet migration. Astrophys. J. 565, 1257–1274. Touboul, M., Kleine, T., Bourdon, B., Palme, H., Wieler, R. 2007. Late formation and pro- longed differentiation of the Moon inferred from W isotopes in lunar metals. Nature 450, 1206–1209. Thommes, E., Nagasawa, M., Lin, D.N.C. 2008. Dynamical shake-up of planetary systems. II. N-body simulations of solar system terrestrial planet formation induced by secular resonance sweeping. Astrophys. J. 676, 728–739. Tsiganis, K., Gomes, R., Morbidelli, R., Levision, H.F. 2005. Origin of the orbital architecture of the giant planets of the Solar System. Nature 435, 459–461. Yoshinaga, K., Kokubo, E., Makino, J. 1999. The stability of protoplanet systems. Icarus 139, 328–335. Youdin, A.N., Chiang, E.I. 2004. Particle pileups and planetesimal formation. Astrophys. J. 601, 1109–1119. Youdin, A.N., Shu, F.H. 2002. Planetesimal formation by gravitational instability. Astrophys. J. 580, 494–505. – 30 – Wadsley, J.W., Stadel, J., Quinn, T. 2004. Gasoline: a flexible, parallel implementation of TreeSPH. New Astronomy 9, 137–158. Ward, W.R. 1981. Solar nebula dispersal and the stability of the planetary system. I - Scanning secular resonance theory. Icarus 47, 234-264. Ward, W.R. 1986. Density waves in the solar nebula - Differential Lindblad torque. Icarus 67, 164–180. Ward, W.R. 1997. Protoplanet migration by nebula tides. Icarus 126, 261–281. JS orbits EJS CJS SS τdecay (Myr) 1 1 1 1 2 2 2 2 3 3 3 3 5 5 5 5 1 1 1 1 2 2 2 2 3 3 3 3 5 5 5 5 p MT (ME) 10 5 10 5 10 5 10 5 10 5 10 5 10 5 10 5 10 5 10 5 10 5 10 5 10 5 10 5 10 5 10 5 1 1 2 2 1 1 2 2 1 1 2 2 1 1 2 2 1 1 2 2 1 1 2 2 1 1 2 2 1 1 2 2 Np MT,final (ME) 4.30 3.08 2.44 1.93 1.97 2.28 1.49 1.18 0.94 1.33 0.69 0.78 0.46 0.51 0.31 0.40 4.32 2.62 2.45 2.02 2.50 1.82 1.57 0.98 1.22 1.49 0.93 0.75 0.73 0.81 0.55 0.53 1.98 4 3 3 3 5 4 3 3 4 3 2 3 3 1 2 1 4 4 4 4 4 4 2 3 3 4 3 2 3 4 3 4 4 – 31 – S c bmin 13.8 29.3 31.0 78.3 28.1 17.7 26.5 79.9 69.8 34.5 18.1 74.9 25.8 – 30.1 39.8 40.6 33.7 33.5 34.7 34.0 42.2 23.8 47.2 111.5 42.1 83.1 – 70.2 101.9 – 10.9 17.1 17.5 11.7 13.0 33.7 54.9 72.3 55.7 13.5 43.1 70.4 52.3 19.7 16.7 32.0 89.9 – 35.9 38.3 34.5 43.3 38.1 35.4 51.9 55.9 50.5 35.8 52.5 59.7 56.1 59.1 51.7 61.7 26.3 am (AU) 1.15 0.94 0.97 0.95 1.19 0.82 0.93 0.97 1.19 0.99 0.65 1.13 0.88 0.68 0.74 0.68 1.11 1.00 1.35 1.14 1.16 1.22 1.26 1.49 1.46 1.06 1.39 1.63 1.59 1.21 1.49 1.50 0.90 em S d 0.098 0.063 0.078 0.028 0.114 0.033 0.137 0.036 0.055 0.081 0.016 0.064 0.332 0.285 0.107 0.143 0.058 0.064 0.083 0.071 0.073 0.045 0.077 0.100 0.071 0.050 0.191 0.095 0.080 0.074 0.069 0.117 0.038 0.0074 0.0032 0.0058 0.0013 0.0109 0.0010 0.0119 0.0016 0.0044 0.0145 0.0004 0.0048 0.0599 0.0446 0.0062 0.0113 0.0025 0.0030 0.0054 0.0055 0.0134 0.0016 0.0047 0.0115 0.0081 0.0025 0.0306 0.0094 0.0098 0.0151 0.0167 0.0159 0.0018 timp (Myr) 85.6 12.1 27.5 13.5 32.8 308.4 83.3 13.3 0.6 327.4 – – – – – – 109.7 85.5 117.2 139.5 388.8 46.1 184.4 189.5 122.7 – 506.2 – – – – – 50–150 eJ,final 0.006 0.021 0.032 0.026 0.002 0.028 0.025 0.035 0.022 0.030 0.040 0.035 0.036 0.031 0.038 0.042 0.004 0.004 0.004 0.003 0.007 0.004 0.004 0.003 0.004 0.003 0.005 0.003 0.004 0.004 0.003 0.003 0.048 Table 1: Simulation parameters and outcomes for the EJS and CJS simulations. Table columns are the orbits of Jupiter and Saturn, the gas decay time scale τdecay, the power-law index for the radial gradient of the initial surface density of planetesimals p, the initial total mass of planetesimals MT, the number of planets Np, the final total mass of planets MT,final, the radial mass concentration statistics S c, the mass weighted mean semimajor axis am, the minimum orbital separation between neighboring planets bmin normalized by the mutual Hill radius, the mass weighted mean eccentricity em, the angular momentum deficit S d, the time of the last potentially Moon-forming impacts timp, and the final orbital eccentricity of Jupiter eJ,final. Any orbital elements averaged over a few Myr are used. Each value is boldly marked when 3 ≤ ⊕ ≤ 3.0, S c > 45.0, 0.75 AU < am < 1.05 AU, bmin < 35, em < 0.076, Np ≤ 5, 1.0 ≤ MT,final/M S d < 0.0036, or 50 Myr < timp < 150 Myr. SS represents the solar system. – 32 – JS orbits EEJS CJSECC SS τdecay (Myr) 1 1 1 1 2 2 2 2 3 3 3 3 5 5 5 5 1 1 1 1 2 2 2 2 3 3 3 3 5 5 5 5 p MT (ME) 10 5 10 5 10 5 10 5 10 5 10 5 10 5 10 5 10 5 10 5 10 5 10 5 10 5 10 5 10 5 10 5 1 1 2 2 1 1 2 2 1 1 2 2 1 1 2 2 1 1 2 2 1 1 2 2 1 1 2 2 1 1 2 2 Np MT,final (ME) 3.70 3.41 2.66 1.50 1.71 2.15 0.89 1.23 0.63 0.59 0.67 0.69 0.33 0.67 0.15 0.21 3.72 2.74 2.00 1.96 2.27 2.17 1.43 1.46 1.11 1.44 0.71 0.95 0.62 0.92 0.41 0.50 1.98 3 4 2 1 2 4 2 2 2 1 2 1 1 2 1 1 4 3 3 3 5 4 4 3 3 3 2 2 3 3 2 2 4 S c 16.6 22.6 40.7 – 36.7 31.0 34.0 52.6 38.4 – 46.6 – – bmin 38.1 30.8 59.8 – 64.5 24.8 154.1 72.2 85.1 – 140.0 – – 31.1 49.6 – – 27.0 32.8 42.6 50.8 10.9 20.7 49.9 59.6 64.8 27.3 89.4 85.6 12.4 61.8 67.5 110.4 89.9 – – 35.3 31.7 48.6 56.5 25.0 33.0 32.2 25.0 34.5 65.4 54.3 105.0 121.7 31.6 74.1 57.9 26.3 am (AU) 0.72 0.85 0.86 0.64 0.56 0.75 0.65 0.95 0.66 0.72 0.86 0.73 0.61 0.63 0.72 0.65 1.28 1.11 1.18 0.95 0.87 1.06 1.13 0.89 1.33 1.12 1.02 0.93 0.78 0.87 1.08 1.02 0.90 em S d 0.149 0.053 0.093 0.089 0.111 0.077 0.289 0.196 0.170 0.597 0.478 0.411 0.171 0.061 0.234 0.263 0.041 0.180 0.077 0.045 0.057 0.065 0.053 0.089 0.064 0.076 0.139 0.160 0.202 0.025 0.121 0.061 0.038 0.0219 0.0023 0.0114 0.0113 0.0065 0.0069 0.0427 0.0572 0.0211 0.2090 0.1316 0.0916 0.0159 0.0144 0.0356 0.0377 0.0019 0.0192 0.0076 0.0033 0.0024 0.0038 0.0022 0.0060 0.0219 0.0060 0.0112 0.0132 0.0476 0.0007 0.0190 0.0038 0.0018 timp (Myr) 180.0 62.8 17.6 26.1 263.9 28.3 23.3 12.7 0.6 21.5 22.6 20.9 – – – – 33.4 111.9 50.5 71.0 106.3 110.1 – 16.5 180.4 80.8 – 49.2 – – – – 50–150 eJ,final 0.054 0.054 0.060 0.074 0.067 0.077 0.077 0.046 0.071 0.074 0.082 0.089 0.084 0.074 0.090 0.093 0.004 0.007 0.011 0.014 0.014 0.005 0.015 0.014 0.006 0.013 0.032 0.032 0.013 0.037 0.027 0.029 0.048 Table 2: The same as Table 1, but for the EEJS and CJSECC simulations – 33 – Fig. 1.- Time scales for radial migration τmig and damping of eccentricity τdamp due to gas drag and tidal interaction (τmig = a[(da/dt)drag + (da/dt)tid]−1 and τdamp = e[(de/dt)drag + (de/dt)tid]−1) at 1 AU with Σgas = 2000 g cm−2 and i = 0. The migration and damping rates due to gas drag, (da/dt)drag and (de/dt)drag, are derived after Adachi et al. (1976) with some corrections given in Inaba et al. (2001). The labels to the lines represent e. For e = 0.05, we also plot the time scales , and mGI = 1019 with the enhancement of the gas drag force for m0 = 0.0025 M g (see Sec. 2.2.1). The masses of solar system bodies are also plotted in the upper panel. , m1 = 0.01 M ⊕ ⊕ – 34 – Fig. 2.- Snapshots of an EJS simulation (τdecay = 1 Myr, p = 2, and MT = 5 M ) on the a − e plane. The size of each body is proportional to the radius, except Jupiter's size is modified to the Earth size on this plot. The color represents spin rate ω: ω > ωcrit, ωcrit ≥ ω > 0.5 ωcrit, 0.5 ωcrit ≥ ω > 0, and ω = 0 for red, blue, green, and black bodies, respectively, where ωcrit is the spin rate with which the centrifugal force balances with the gravity at the surface. The long- dashed and short-dashed lines are locations of the ν5 and ν6 resonances, respectively. The black dashed curve is given by a(1 + e) = aJ; particles in the right side of this curve experience orbital crossing with Jupiter. Note the difference in scales of e between the top panels and others. ⊕ – 35 – Fig. 3.- Time evolutions of the total masses of small, big, and all particles (the upper panel) and the mass weighted mean eccentricities of small and big particles (the lower panel) for an EJS simulation. We define mass of a small (big) particle to be below (above) 2 × 1026 g. – 36 – Fig. 4.- The time evolutions of the masses (the upper panel) and the orbital excursions (the lower panel) of surviving planets in an EJS simulation. The numbers of planets are given in the order of their masses in this figure. – 37 – Fig. 5.- The same as Fig. 2, but for a CJS simulation. Except for orbits of Jupiter and Saturn, other parameters are the same as those in Fig. 2. – 38 – Fig. 6.- The same as Fig. 3, but for a CJS simulation. – 39 – Fig. 7.- The same as Fig. 4, but for a CJS simulation. – 40 – Fig. 8.- Snapshots of all EJS simulations on the a − e plane at the end of simulations. In each . The panel the numbers after t, p, and m represent τdecay in units of Myr, p itself, and MT/M colors of bodies represent the spin rates as well as Fig. 2. In the panel at the upper left corner, we plot the current terrestrial planets with dashed circles. ⊕ – 41 – Fig. 9.- The same as Fig. 8, but for the CJS simulations. – 42 – Fig. 10.- Mass vs. semimajor axis for the EJS and CJS simulations. The filled and open symbol are for MT/M ⊕ = 5 and 10. The color represents the gas decay time: τdecay = 1, 2, 3, and 5 Myr for black, red, blue, and green symbols. Circles and squares are for p = 1 and 2. Cross marks represent the current terrestrial planets. – 43 – Fig. 11.- Eccentricity vs. mass for the EJS and CJS simulations. See Fig. 10 for symbols. – 44 – Fig. 12.- Mass weighted mean eccentricity em vs. mass weighted mean semimajor axis am for the EJS and CJS simulations. The dotted vertical and horizontal lines indicate the values for the current terrestrial planets. See Fig. 10 for symbols. – 45 – Fig. 13.- Angular momenta and velocities of potentially Moon-forming impacts as functions of time for the EJS and CJS simulations. See Fig. 10 for symbols. – 46 – Fig. 14.- The same as Fig. 8, but for the EEJS simulations. – 47 – Fig. 15.- The same as Fig. 8, but for the CJSECC simulations. – 48 – Fig. 16.- The same as Fig. 13, but for the EEJS and CJSECC simulations. Some impacts in the EEJS simulations have too large Limp (> 6 LEM) and only their vimp's are shown. – 49 – Fig. 17.- Comparison between the EJS and CJSECC simulations. The same parameters are used for the two simulations except for orbits of Jupiter and Saturn: τdecay = 1 Myr, p = 1, and MT = 5 M . The format for colors and lines is after Fig. 2. ⊕ – 50 – Fig. 18.- Angular momentum deficit S d (upper panel), radial mass concentration parameter S c/a2 m (middle panel), and time of the last one of potentially Moon-forming impacts timp (lower panel) vs. final eccentricity of Jupiter eJ,final. Filled circles, filled squares, open circles, and open squares are results from the EJS, EEJS, CJS, and CJSECC simulations, respectively. The color represents the gas decay time: τdecay = 1, 2, and 3 Myr for black, red, and blue symbols. A fit to a quartic function is applied to log S d, while linear fits are applied to S c/a2 m and log timp for sets of the EJS+EEJS and CJS+CJSECC simulations, respectively. Horizontal and vertical dotted lines indicate the current values (for timp, a possible range indicated from isotope records is between two dotted lines) – 51 – Fig. 19.- Radial mass concentration parameter S c/a2 m (upper panel) and angular momentum deficit S d (lower panel) vs. time of the Moon-forming impact timp. The data are the same as those in Fig. 18.
1704.01541	1	1704	2017-04-05T17:36:51	Masses of Kepler-46b, c from Transit Timing Variations	[ "astro-ph.EP" ]	We use 16 quarters of the \textit{Kepler} mission data to analyze the transit timing variations (TTVs) of the extrasolar planet Kepler-46b (KOI-872). Our dynamical fits confirm that the TTVs of this planet (period $P=33.648^{+0.004}_{-0.005}$ days) are produced by a non-transiting planet Kepler-46c ($P=57.325^{+0.116}_{-0.098}$ days). The Bayesian inference tool \texttt{MultiNest} is used to infer the dynamical parameters of Kepler-46b and Kepler-46c. We find that the two planets have nearly coplanar and circular orbits, with eccentricities $\simeq 0.03$ somewhat higher than previously estimated. The masses of the two planets are found to be $M_{b}=0.885^{+0.374}_{-0.343}$ and $M_{c}=0.362^{+0.016}_{-0.016}$ Jupiter masses, with $M_{b}$ being determined here from TTVs for the first time. Due to the precession of its orbital plane, Kepler-46c should start transiting its host star in a few decades from now.	astro-ph.EP	astro-ph	Submitted to The Astronomical Journal Preprint typeset using LATEX style AASTeX6 v. 1.0 MASSES OF KEPLER-46B,C FROM TRANSIT TIMING VARIATIONS Ximena Saad-Olivera1, David Nesvorný1,2, David Kipping3, and Fernando Roig1 1Observatório Nacional, Rio de Janeiro, José Cristino 77, Rio de Janeiro, RJ 20921-400, Brazil 2Department of Space Studies, Southwest Research Institute, 1050 Walnut Street,Suite 300, Boulder, CO 80302, USA 3Dept. of Astronomy, Columbia University, 550 W 120th St., New York, NY 10027, USA [email protected] Abstract We use 16 quarters of the Kepler mission data to analyze the transit timing variations (TTVs) of the extrasolar planet Kepler-46b (KOI-872). Our dynamical fits confirm that the TTVs of this planet −0.005 days) are produced by a non-transiting planet Kepler-46c (P = 57.325+0.116 (period P = 33.648+0.004 −0.098 days). The Bayesian inference tool MultiNest is used to infer the dynamical parameters of Kepler- 46b and Kepler-46c. We find that the two planets have nearly coplanar and circular orbits, with eccentricities ≃ 0.03 somewhat higher than previously estimated. The masses of the two planets are found to be Mb = 0.885+0.374 −0.016 Jupiter masses, with Mb being determined here from TTVs for the first time. Due to the precession of its orbital plane, Kepler-46c should start transiting its host star in a few decades from now. −0.343 and Mc = 0.362+0.016 Keywords: planetary systems 1. INTRODUCTION More than 3300 exoplanets have been discovered with the Kepler and K2 missions, and almost 16% of these are in multi-transiting planetary systems with at least two confirmed planets (exoplanetarchive.ipac.caltech.edu) The detection of these exoplanets was possible because they have nearly edge-on orbits and pass in front of the star, thus producing a dip in the photometric light curve. The analysis of the transit light curve allows scientists to determine the planetary to stellar radius ratio (RP /R⋆), the inclination of the planetary orbit (iP ), the transit duration (tD), and the time of mid-transit (tC ) (e.g., Seager & Mallén-Ornelas 2003, Carter et al. 2008). In multi-planetary systems, the mutual gravitational perturbation between planets can be detectable under certain conditions. For example, the mid-transit time tC of consecutive transits of the same planet may not occur exactly on constant ephemeris. Instead, due to the gravitational perturbations from planetary companions, tC values can display non linear trends called Transit Timing Variations (TTVs). In some cases, it is also possible to measure variations of tD, known as the Transit Duration Variations (TDVs), which can also be attributed to the gravitational perturbations. The dynamical analysis of TTVs can yield a comprehensive characterization of orbits and also the planetary masses, or at least provide certain limits (Miralda-Escudé 2002, Agol et al. 2005, Holman & Murray 2005). The TTVs have been primarily used to confirm and characterize the systems of multi-transiting planets (Holman et al. 2010, Lissauer et al. 2011, Steffen et al. 2012, Xie 2014), but they can also be used to detect and characterize non-transiting planets (Ballard et al. 2011, Nesvorný et al. 2012, 2013, 2014, Dawson et al. 2014, Mancini et al. 2016). In fact, if the radial velocity measurements are not available, the observation and analysis of TTVs provide the only way currently available to determine masses. When the mass determination is combined with the planetary radius estimate from the transit light curve, this results in a planetary density estimate, and hence allows scientists to develop internal structure models (e.g. Guillot & Gautier 2015). In this work, we focus on the Kepler-46 system (KOI-872). The first analysis of the light curve of Kepler-46 showed the presence of a candidate planet, currently known as Kepler-46b, with a 33.6 day period and radius consistent with a Saturn-class planet (Borucki et al. 2011). Nesvorný et al. (2012) measured and analyzed the TTVs and TDVs of 15 transits from Kepler quarters 1-6. They found that the TTVs can be best explained if Kepler-46b is a member of a two- planet system with the planetary companion, Kepler-46c, having a non-transiting orbit on outside of Kepler-46b. The TTVs analysis provided two possible solutions for the orbital parameters of the two planets, but one of these solutions was discarded by Nesvorný et al. because it produced significant TDVs, which were not observed. The preferred 2 Saad-Olivera et al. Table 1. The mid-transit times tC obtained for Kepler-46b from the analysis of Kepler Quarters 1-16(Model MI ). Cycle tC (BJD) 1 2 3 4 6 7 8 9 11 12 13 14 15 18 19 20 21 22 23 24 25 26 27 29 31 32 33 35 36 37 38 39 40 41 42 55019.6993 ± 0.0012 55053.2959 ± 0.0010 55086.8640 ± 0.0011 55120.4409 ± 0.0018 55187.6880 ± 0.0015 55221.3360 ± 0.0016 55254.9066 ± 0.0011 55288.4664 ± 0.0012 55355.6667 ± 0.0010 55389.3438 ± 0.0011 55422.9309 ± 0.0017 55456.4875 ± 0.0011 55490.0601 ± 0.0009 55590.9283 ± 0.0008 55624.5138 ± 0.0008 55658.0838 ± 0.0011 55691.6593 ± 0.0008 55725.3035 ± 0.0009 55758.9147 ± 0.0009 55792.5562 ± 0.0010 55826.1262 ± 0.0011 55859.6845 ± 0.0017 55893.2891 ± 0.0011 55960.5704 ± 0.0010 56027.7090 ± 0.0010 56061.2821 ± 0.0010 56094.8799 ± 0.0010 56162.1511 ± 0.0011 56195.7346 ± 0.0014 56229.3033 ± 0.0012 56262.8845 ± 0.0013 56296.5319 ± 0.0012 56330.1453 ± 0.0010 56363.7757 ± 0.0010 56397.3426 ± 0.0010 solution indicates a nearly coplanar (Imut ≃ 1◦) and nearly circular (e ≃ 0.01) orbits, and Mc = 0.376+0.021 −0.019 MJ. The mass of Kepler-46b was not constrained from TTVs/TDVs, but an upper limit of Mb < 6 MJ was determined from the stability analysis. Here, we reanalyzed the Kepler-46 system using 35 transits from Kepler quarters. The transit analysis and dynamical fits were executed using MultiNest (Feroz et al. 2009, 2013), which is an efficient tool based on the Bayesian inference method. We were able to better determine the orbital parameters and masses of the two planets. The mass of Kepler- 46b was determined here from TTVs for the first time. The paper is organized as follows: in Section 2, we describe the light curve analysis to obtain the mid-transit times and transit parameters of Kepler-46b. In Section 3, we discuss the dynamical analysis. The results are reported in Section 4. The last section is devoted to the conclusions. 2. LIGHT CURVE ANALYSIS AND TRANSIT TIMES Masses of Kepler-46b,c from TTVs 3 Before a dynamical analysis can be conducted, the transit times and durations of each event need to be inferred. This, in turn, requires us to first detrend the Kepler photometry and then fit the transits with an appropriate model. For this task, we use a similar approach to that described in Teachey et al. (2017), which is essentially a modified version of the approach described in Kipping et al. (2013). The broad overview is that we remove long-term trends from the Kepler simple aperture photometry with the CoFiAM algorithm described in Kipping et al. (2013). Essentially, CoFiAM is a cosine filter designed to not disturb the transit of interest but remove present longer-period trends. A pre-requisite for using CoFiAM is that the time and duration of each transit are already approximately known. Since this system is known to exhibit strong variations in both, this poses a catch-22 problem for us. Following Teachey et al. (2017), we remedy the problem by conducting an initial detrending using approximate estimates for the times and a fixed duration taken from Nesvorný et al. (2014). This data are then fitted (as we will describe later) to provide revised estimates for the times and durations. These revised times and durations are then used as inputs for a second attempt at detrending the original Kepler data using CoFiAM again. This iteration process allows for self-consistent inference of the basic transit parameters. Initially, we take into account the 40 transit events available for Kepler-46b (e.g. Holczer et al. 2016). However, one of such transit (corresponding to cycle 5) is found to have inadequately detrended data, which is identified by visual inspection of the light curve. Four other events (corresponding to cycles 0, 10, 16, and 28) are found to have insufficient temporal coverage1. There are several reasons that this might happen. At first, safe modes and data downlinks cause genuine gaps in the time series. At the next level, any Kepler time stamp with an error flag not set to 0 is removed by CoFiAM. Then, CoFiAM puts a smoothed moving median through the data and looks for 3σ outliers caused by sharp flux changes or some other weird behavior. Stitching problems among the quarters can also cause apparent sharp flux changes, and if we do not have enough usable data around the transit, then the event will also have data chopped this way. At last, the detrending of an event can fail giving a poor detrending, so we perform a second round of outlier cleaning that which may eventually lead to most of the data for that event being removed. In the end, we keep only 35 transits for which visual inspection of the light curve indicates that we would be able to get reliable time estimates (Table 1). The actual fitting process, which is ultimately repeated twice, is conducted using MultiNest coupled to a Mandel & Agol (2002) light curve model. We assume freely fitted quadratic limb-darkening coefficients, using the prescription of Kipping (2013), and allow each transit to have a unique mid-transit time but a common set of basic shape parameters (i.e. a common duration). While 20 of the transits were short cadence, 15 were long cadence requiring re-sampling using the technique of Kipping (2010), for which we used Nresam = 30. Since each transit requires a unique free parameter, this leads to a very large number of free parameters in the final fit (35 just for the times alone). For computational expedience, we split the light curve into 4 segments of 10 transits each, with 5 for the last segment. Each segment is independent of the others, but assumes internally consistent shape parameters. Rather than freely fitting each event, this allows for the transit template to be well constrained such that the mid-transit time precision is improved. We refer to the above model of segments of transits as model MS. As we did in Nesvorný et al. (2014), we also try fitting each transit completely independently, but fixing the limb- darkening coefficients to the same as those used in Nesvorný et al. (2014). This allows for TDVs to be inferred as well as TTVs, but generally increases the credible interval of the TTV posteriors due to the much weaker information about the transit shape available. In what follows, we refer to this model as MI . 3. DYNAMICAL ANALYSIS OF TRANSIT TIMES Following Nesvorný et al. (2012), we proceed by searching for a dynamical model that can explain the measured transit times. At variance with Nesvorný et al., we do not fit the TTVs but we directly fit the mid-transit times, which is expected to be a more accurate procedure. The results presented in the following are based on the mid-transit times from model MI (Table 1). We have performed similar computations using model MS and verified that the results are indistinguishable within their 1-σ uncertainties. To perform the dynamical fit, we assume a model of two planet and we use a modified version of the efficient symplectic integrator SWIFT (Levison & Duncan 1994), adapted to record the mid-transit times of any transiting planet (Nesvorný et al. 2013, Deck et al. 2014). MultiNest is then used to perform the model selection and to estimate the best-fit values of the dynamical parameters with their errors. 1 In particular, the transit of cycle 16 is marked with an outlier code > 32 in Holczer et al. (2016). 4 Saad-Olivera et al. Table 2. Prior distributions of 12 model parameters that we used to obtain the dynamical fit. U[x, y] means a uniform distribution between x and y. The subindex P refers to any of the two planets. Parameter Prior values MP /M⋆ U[0, 0.005] Pb (days) U[33.5, 33.7] Pc (days) U[40, 84] eP bc P (◦) λc (◦) U[0, 0.4] U[1, 20] U[0, 360] U[0, 360] Ωc − Ωb (◦) U[0, 360] δt (d) U[−0.1, 0.1] MultiNest applies the Bayes rule (Appendix A) to determine the values of the 12 parameters of the model that are necessary to reproduce the observed mid-transit times. The method provides the posterior distributions of these 12 parameters, namely, the planet-over-star mass ratios (Mb/M⋆, Mc/M⋆), the orbital periods (Pb, Pc), the eccentricities (eb, ec), the longitudes of periastron (b, c), the mean longitude and impact parameter of the non-transiting planet (λc, bc), the difference of the nodal longitudes of the two planets (Ωc − Ωb), and the difference between the mid-transit time reference epoch (τ = 55053.2826 BJD) and the mid-transit time of the nearest transit (δt). This latter parameter gives us the information about the mean longitude of the transiting body at the reference epoch (λb = λb,0 − 2π δt/Pb, where λb,0 is the mean longitude at mid-transit time). The initial value of the impact parameter of Kepler-46b (bb) is fixed at the value determined from the transit fit. We use the transit reference system from Nesvorný et al. (2012), where the reference plane is the plane defined by b = 0, the origin of longitudes is at the line of sight, and the nodal longitude of Kepler-46b (Ωb) is set to 270◦. The stellar parameters are also adopted from Nesvorný et al. (2012). The priors distributions of the 12 model parameters are given in Table 2. The distributions were chosen as uninfor- mative with uniform priors. The interval limits of these priors are based on results from Nesvorný et al. (2012) that guarantee that the system is constituted of a transiting and a non-transiting planet and is dynamically stable. The orbital period of the transiting body is well known from the transit fit, so we only consider a very small range of priors around the known value. For the period of the non-transiting planet, we consider an interval of priors that includes the two solutions found by Nesvorný et al. (2012). The upper limit in eccentricities is set to only 0.4, since higher values cause the code to become too slow and do not provide additional solutions. The upper limit of the impact parameter of the non-transiting planet corresponds to an inclination iP ∼ 70◦. The calculation of the integrals involved in Eq. (A2) requires the use of a numerical method. MultiNest uses a multi-modal nested sampling technique to efficiently compute the evidence integral, and also provides the posterior distributions of the parameters. In our case, the likelihood function (see Appendix A) is defined as L(dθ, M ) = N Yj=1 exp − N Xi=1 (cid:0)tobs C,i − tcal C,i(cid:1)2 2σ2 i ! , (1) 1 q2πσ2 j where N = 35 is the number of transits; tobs spectively; and σi is the uncertainty of tobs this likelihood function. C,i and tcal C,i are the observed (Table 1) and calculated mid-transit times, re- C,i . According to Eq. (A1), the best-fit parameters are obtained by maximizing 4. RESULTS Our analysis provides two best-fit solutions that correspond to the two solutions reported in Nesvorný et al. (2012). The first solution (S1) was obtained for a uniform distribution of the period prior in the interval Pc = [40, 70] days. This solution has a maximum likelihood of ln LS1 = 183.42, a reduced χ2 S1 = 1.96 (for 35 − 12 = 23 degrees of freedom), and a global evidence of ln ZS1 = 125.23. The reduced χ2 indicates that in principle this is a good fit, but the transit time errors may be underestimated by a factor of ∼ 0.8. Table 3 reports the parameter values corresponding to S1. They were determined from the weighted posteriors Masses of Kepler-46b,c from TTVs 5 calculated by MultiNest. The associated uncertainties are defined by the standard error at the 68.2% confidence level. The orbital elements provided by this solution are astrocentric osculating elements at epoch BJD 55053.2826. The corresponding fit to the TTV signal is shown in Fig. 1. The posterior distributions of parameters, and the correlations between pairs of parameters, are shown in Fig. 2. According to solution S1, Kepler-46b is a Jupiter class planet (Mb ≃ 0.9 MJ), with a density similar or slightly higher than that of Jupiter. The companion is a Saturn class planet (Mc ≃ 0.36 MJ)2 moving on an outer orbit with Pc = 57.325 d. The period ratio Pc/Pb = 1.703 indicates that the two planets are close to the 5:3 mean motion resonance, but their orbits are not resonant. This configuration may be part of the trend identified previously where compact planetary systems appear to avoid exact resonances (e.g., Veras & Ford 2012). At least in some cases, this can be explained by tidal dissipation acting in the innermost planet (Lithwick & Wu 2012, Batygin & Morbidelli 2013). Kepler-46b, however, is probably too far from its host star for the tidal effects to be important. The orbital eccentricities of the two planets are similar, eb, ec ≃ 0.03, and the mutual inclination of the orbits is Imut = 0.43◦, confirming the nearly circular and nearly coplanar nature of the planetary system. The small mutual inclination is indicates that Kepler-46c may become a transiting planet in the near future (see Section 4.1). The second solution (S2) is obtained for the period prior Pc = [80, 84] days. This solution has a global evidence of ln ZS2 = 105.97 and reveals the existence of two modes differing mainly in the mutual inclination between orbits. Specifically, Mode 1 of S2 corresponds to an impact parameter for Kepler-46c of bc ≃ 5 with 0◦ ≤ Ωc ≤ 180◦, while Mode 2 corresponds to bc ≃ 7 with 180◦ ≤ Ωc ≤ 360◦. Mode 1 gives a mutual inclination of Imut ≃ 170◦ implying that the orbit of Kepler-46c is retrograde, while Mode 2 gives Imut ≃ 10◦ meaning that both planets are in prograde orbits. The maximum likelihood and reduced χ2 values of the two modes are: ln LS2 = 157.16, χ2 S2 = 4.25 and ln LS2 = 157.56, χ2 S2 = 4.21, respectively. Table 4 summarizes some characteristics of the solution S2. For both modes, the mass of Kepler-46c would be larger (Mc ≃ 1.9 MJ) than that obtained for the solution S1. The orbital eccentricities of both modes are smaller than for the S1 solution. The orbital period ratio suggested by both modes is Pc/Pb ≃ 2.42, which places the two planets very close to (but not inside of) the 5:2 mean motion resonance. Since a ∆χ2 ∼ 2.25 between solutions S1 and S2 is not statistically significant enough to penalize any of the solutions, we apply Eq. (A3) to compare the global evidences of S1 and S2. The Bayes factor becomes ∆ ln Z = 19.26, suggesting that solution S1 is preferred over S2 with a confidence level of 5.7σ. This argument can be used to rule out solution S2. It is worth stressing that this conclusion is based on the analysis of the TTVs only, while Nesvorný et al. (2012) had to resort to using the TDV constraints to arrive at the same conclusion. In order to test how sensitive is the Bayes factor is to the choice of prior distributions, we redo the dynamical fits restricting the intervals of the eb, ec priors to U[0, 0.2] and the interval of the Pc priors to U[55, 84]. The resulting values of the evidence ln Z of each solution are higher in this case, as expected from Eq. (A2), but the Bayes factor is almost the same: ∆ ln Z = 20.81. This indicates that S1 is still preferred over S2 with a confidence level of 6σ. Then, we can claim that model selection relying on the Bayes factor is not sensitive to the choice of the priors distributions. Finally, it is worth noting that if we perform the dynamical fits using more than 35 transits, the solutions do not change significantly. At least three of the five discarded transits (corresponding to cycles 0, 5, and 10; see Sect. 2) allow us to estimate mid-transit times, even with large errors. The dynamical fits using these 35 + 3 transits provide again the two solutions S1 and S2, with S2 showing two modes, and the Bayes factor favoring S1. All the parameters of these fits are indistinguishable within 1-σ errors with respect to the parameters reported in Tables 3 and 4. We conclude that our two-planet model solution S1 is quite robust. 4.1. Long-term stability Nesvorný et al. (2012) demonstrated that their S1 solution is dynamically stable over 1 Gyr. Since here we obtained slightly larger values of the orbital eccentricities, slightly smaller values of the mutual inclination, and were able to constrain the mass for Kepler-46b, we find it useful to reanalyze the long-term stability of the system. We used the orbits and masses corresponding to the solution S1 (see Table 3) and numerically integrated the orbits over 100 Myr using the SWIFT code and a one-day time step. The orbital evolution of the planets in the first 1500 days of this integration is shown in Figs. 3 and 4. The semi-major axes show short-term oscillations with amplitudes never larger than 0.7%. Due to the conservation of the total angular momentum, the eccentricities show anti-correlated oscillations with a period of ≃ 110 years. We note that, for both planets, these oscillations show amplitudes that 2 This mass value is the same as determined by Nesvorný et al. (2012), and it is in good agreement with the analytic estimate of Deck & Agol (2016). 6 Saad-Olivera et al. Table 3. The best-fit parameters and their errors for the solution S1. The first block reports the transit parameters of Kepler- 46b obtained from the light-curve analysis. The second block reports the orbital parameters and masses obtained from the dynamical fit. Orbital elements are astrocentric at osculating epoch BJD 55053.2826, the reference plane is the plane at which the impact parameter is 0, and the origin of longitudes is at the line of sight. The third block reports the secondary parameters. iP is the inclination of the planet with respect to the sky plane during transit (cos iP ≃ bP R⋆/aP ); Ip is the inclination of the planetary orbit with respect to the reference plane; Imut is the mutual inclination relative to the orbital plane of Kepler-46b. The last block lists the stellar parameters compiled from Nesvorný et al. (2012). Transit fit RP /R⋆ bP Dynamical fit Mp/M⋆ (×10−4) PP (days) eP bP P (◦) λP (◦) ΩP (◦) δt (d) Secondary parameters MP (MJ) aP (au) iP (◦) IP (◦) Imut (◦) RP (RJ) ρ (g cm−3) Stellar parameters M⋆ (M⊙) R⋆ (R⊙) ρ⋆ (g cm−3) log g⋆ a Teff (K) L⋆ (L⊙) MV Age (Gyr) Distance (pc) [M/H] Kepler-46b Kepler-46c 0.0887+0.0010 −0.0012 0.757+0.022 −0.027 9.372+3.941 −3.618 33.648+0.004 −0.005 0.0321+0.0069 −0.0078 3.835+0.057 −0.055 57.325+0.116 −0.098 0.0354+0.0057 −0.0059 0.757+0.022 −0.027 264.2+8.2 −8.9 -- 270 1.483+0.418 −0.322 294.16+8.70 −6.42 338.0+0.3 −0.3 261.4+22.7 −24.3 0.0130+0.0006 −0.0006 -- 0.885+0.374 −0.343 0.362+0.016 −0.016 0.1971+0.0001 −0.0001 0.2811+0.0003 −0.0003 89.04+0.14 −0.14 0.957+0.028 −0.034 -- 88.66+0.26 −0.27 1.35+0.38 −0.29 0.43+0.40 −0.26 0.810+0.035 −0.036 2.069+0.913 −1.136 -- -- Kepler-46 0.902+0.040 −0.038 0.938+0.038 −0.039 1.54+0.22 −0.17 4.447+0.040 −0.035 5155+150 −150 0.556+0.078 −0.070 5.60+0.17 −0.17 9.7+3.7 −3.5 855+68 −65 0.41+0.10 −0.10 ag⋆ is given in c.g.s units. are of the same order of the eccentricity uncertainties reported in Table 3. This implies that the eccentricities will Masses of Kepler-46b,c from TTVs 7 Table 4. Similar to Table 3 but for the two modes of solution S2. Mode 1 Mode 2 Kepler-46b Kepler-46c Kepler-46b Kepler-46c Dynamical fit Mp/M⋆ (×10−4) PP (d) eP bP P (◦) λP (◦) ΩP (◦) Secondary parameters MP (MJ) aP (au) iP (◦) ρ (g cm−3) 8.223+20.522 −6.530 33.609+0.002 −0.003 0.0063+0.0047 −0.0040 20.151+0.930 −0.930 81.504+0.165 −0.501 0.0239+0.0031 −0.0033 7.499+19.497 −5.875 33.610+0.002 −0.003 0.0067+0.0051 −0.0042 0.757 225.9+42.3 −108.2 -- 270 4.962+1.983 −1.153 0.757 124.1+7.8 −12.8 197.2+2.6 −2.5 19.4+8.7 −4.5 229.3+38.6 −102.3 -- 270 20.148+0.890 −0.957 81.520+0.152 −0.479 0.0240+0.0032 −0.0034 7.453+1.854 −1.175 124.0+7.9 −13.2 196.7+2.7 −2.6 207.7+7.5 −4.1 0.777+1.939 −0.617 1.904+0.121 −0.119 0.708+1.842 −0.562 1.886 0.1969+0.0001 −0.0001 0.3556+0.0006 −0.0006 0.1969+0.0001 −0.0001 0.3557+0.0006 −0.0006 89.04+0.14 −0.14 86.51+1.85 −1.90 89.04+0.14 −0.14 84.75+4.31 −4.20 1.815+4.537 −2.042 -- 1.655+4.310 −1.858 -- be indistinguishable at any epoch, within their 1σ uncertainties. The inclinations IP relative to the transit reference plane (see Table 3) also show anti-correlated oscillations with a period of ≃ 83 yr resulting from the precession of the nodal longitudes. In this case, the oscillation amplitude of Kepler-46c is also of the same order of the uncertainty in inclination, but for Kepler-46b the amplitude is about five times larger than the uncertainty. The integration results indicate that the orbits show no signs of chaos, and the planetary system is stable at least over 100 Myr. Our analysis also shows that Kepler-46b should always transit the star, as can be seen from the right panel of Fig. 4. Kepler-46c, on the other hand, is not currently transiting, but it may start to display transits in a few decades. Once it does, Kepler-46 will be a good target for light curve observations that may lead to the confirmation of Kepler-46c and the determination of its radius and density. 5. CONCLUSION We reanalyzed the Kepler-46 planetary system using a larger number of transits (35) than in Nesvorný et al. (2012) and applying the Bayesian inference to perform a dynamical fit to the measured TTVs. We obtained two possible solutions, but the Bayesian evidence allows us to rule out one of them without the use of TDVs as in Nesvorný et al. (2012). The availability of a larger number of transits allows us to determine the mass Kepler-46b, and constrain its density. We confirm that the TTVs signal is well reproduced by a system of two planets on nearly circular and coplanar orbits, with periods of Pb ≃ 33.6 days and Pc ≃ 57.3 days, respectively. This means that the two planets are close to, but not inside of, the 5:3 mean motion resonance. With the radius of Kepler-46b determined from the photometric light curve, Rb ≃ 0.85 RJ, and the new mass determination Mb ≃ 0.9 MJ, we can constrain its density to be ∼ 2 g cm−3. This indicates, consistently with its Jupiter-like mass, that Kepler-46b has a significant gas component. No density estimate can be inferred for the non-transiting companion Kepler-46c, but the estimated mass Mc ≃ 0.36 MJ indicates a Saturn-class planet. Interestingly, our new fit indicates that Kepler-46c should start transiting the host star in a few decades. APPENDIX A. BAYESIAN INFERENCE In order to find the best-fit parameters to the observed mid-transit times, we apply a powerful statistics tool that relies on the Bayes rule: P (θd, M ) = L(dθ, M )π(θ, M ) Z(dM ) (A1) 8 Saad-Olivera et al. 80 60 40 20 0 -20 -40 -60 ] i n m [ s V T T -80 0 5 10 15 20 25 Transit Cycle 30 35 40 45 Figure 1. Transit Timing Variations (black dots) computed as the difference between the observed mid-transit times and the linear ephemeris (obtained by fitting a straight line to the mid-transit times series). The red line corresponds to the best dynamical fit corresponding to the solution S1. Masses of Kepler-46b,c from TTVs 9 Figure 2. The equally weighted posterior distributions of our 12 model parameters (diagonal), and the corresponding correlations between parameters. The mass ratios are given in units of 10−4. 10 Saad-Olivera et al. ] U A [ s x a i j r o a m m e s i ] U A [ s x a i j r o a m m e s i 0.1975 0.1974 0.1973 0.1972 0.1971 0.197 0.1969 0.282 0.2815 0.281 0.2805 0.28 0.2795 0.279 Planet b 0 200 400 600 800 time [day] Planet c 1000 1200 1400 0 200 400 600 800 time [day] 1000 1200 1400 Figure 3. Evolution of the semi-major axes of Kepler-46b (top panel) and Kepler-46c (bottom panel) y t i c i r t n e c c e ] g e d [ n o i t a n i l c n i 0.055 0.05 0.045 0.04 0.035 0.03 0.025 0.02 0.015 1.4 1.3 1.2 1.1 1 0.9 0.8 0.7 0 Masses of Kepler-46b,c from TTVs 11 Planet c Planet b 0 50 100 150 200 250 300 350 400 time [yr] Planet c Planet b 50 100 150 200 250 300 350 400 time [yr] Figure 4. Evolution of the eccentricities (top panel) and inclinations (bottom panel) of Kepler-46b (red full lines) and Kepler- 46c (blue full lines) The inclinations are relative to the transit reference plane. The dotted lines in the bottom panel represent the inclination limits below which the planets are expected to show transits (i.e. bP < 1). 12 Saad-Olivera et al. This rule gives the posterior distribution P (θd, M ) of the parameters θ, for the model M , and gives the data d, in terms of the likelihood distribution function L(dθ, M ) within a given set of prior distribution π(θ, M ). The expression is normalized by the so-called Bayesian evidence Z(dM ). We recall that in this formula, the prior and the evidence represent probability distributions, while the likelihood is a function that generate the data d given the parameter θ. The prior is the probability of the parameters θ that is available before making any observation, in other words, it is our state of knowledge of the parameters of the model before considering any new observational data d. The evidence is the probability of the data d given the model M , integrated over the whole space of parameter θ as defined by the prior distribution. This is also referred to as the marginal likelihood, and it is given by Z(dM ) =Z L(dθ, M )π(θ, M )dθ (A2) For parameter estimation, the Bayesian evidence simply enters as a normalization constant that can be neglected, as in most cases it seeks to maximum posterior values. However, in the case of having competing models, Bayes rule penalizes them via model selection, and in this case, the Bayesian evidence is relevant. For example, consider two models M1 and M2; then, the ratio of the posterior probabilities, also known as posterior odds, is given by P (M1d) P (M2d) = Z(dM1) Z(dM2) π(M1) π(M2) (A3) Therefore, the posterior odds of the two models is proportional to the ratio of their respective evidences, which is called the Bayes factor. The proportionality becomes an equality when the ratio of the prior probabilities (or prior odds) of the two models is 1. This work has been supported by the Coordination for the Improvement of Higher Education Personnel (CAPES), the Brazilian Council of Research (CNPq), NASA's Emerging Worlds program, and Brazil's Science without Borders program. REFERENCES Agol, E., Steffen, J., et al. 2005, MNRAS 359, 567 Ballard, S., Fabrycky, D., et al. 2011, ApJ 743, 200 Batygin, K. & Morbidelli, A. 2013, AJ 145, 1 Borucki, W. J., Koch, D. G., et al. 2011, ApJ 736, 19 Carter, J. A., Yee, J. C., et al. 2008, ApJ 689, 499 Dawson, R. I., Johnson, J. A., et al. 2014, ApJ 791, 89 Deck, K. M., Agol, E., et al. 2014, ApJ 787, 132 Deck, K. M. & Agol, E. 2015, ApJ 802, 116 Feroz, F., Hobson, M. P., Bridges, M. 2009, MNRAS 398, 1601 Feroz, F., Hobson, M. P., et al. 2013, arXiv e-print:1306.2144 Guillot, T. & Gautier, D. 2015, in: Treatise on Geophysics, 2nd Edition, G. Schubert (ed.), Elsevier: Amsterdam, p. 529 Holczer, T., Mazeh, T., et al. 2016, ApJS 225, 9 Holman, M. J. & Murray, N. W. 2005, Science 307, 1288 Holman, M. J., Fabrycky, D. C., et al. 2010, Science 330, 51 Kipping, D. M. 2010, MNRAS 408, 1758 Kipping, D. M., Hartman, J., et al. 2013, ApJ 770, 101 Kipping, D. M. 2013, MNRAS 435, 2152 Levison, H. F. & Duncan, M. J. 1994, Icarus 108, 18 Lissauer, J. J., Fabrycky, D. C., et al. 2011, Nature 470, 53 Lithwick, Y. & Wu, Y. 2012, ApJL 756, L11 Mandel, K. & Algol, E. 2002, ApJL 580, L171 Mancini, L., Lillo-Box, J., et al. 2016, A&A 590, A112 Miralda-Escudé, J. 2002, ApJ 564, 1019 Nesvorný, D., Kipping, D. M., et al. 2012, Science 336, 1133 Nesvorný, D., Kipping, D. M., et al. 2013, ApJ 777, 3 Nesvorný, D., Kipping, D. M., et al. 2014, ApJ 790, 31 Seager, S., & Mallén-Ornelas, G., 2003, ApJ 585, 1038 Steffen, J. H., Fabrycky, D. C., et al. 2012, MNRAS 421, 2342 Teachey, A., Kipping, D. M., et al. 2017, ApJ submitted Veras, D. & Ford, E. B. 2012, MNRAS 420, L23 Xie, J.-W. 2014, ApJS 210, 25
1512.02064	1	1512	2015-12-07T14:33:29	A New Mechanism for Chondrule Formation: Radiative Heating by Hot Planetesimals	[ "astro-ph.EP", "astro-ph.SR" ]	We propose that chondrules are formed by radiative heating of pre-existing dust clumps during close fly-bys of planetesimals with incandescent lava at their surfaces. We show that the required temperatures and cooling rates are easily achieved in this scenario and discuss how it is consistent with bulk aspects of chondritic meteorites, including complementarity and the co-mingling of FeO-poor and FeO-rich chondrules.	astro-ph.EP	astro-ph	A New Mechanism for Chondrule Formation: Radiative Heating by Hot Planetesimals William Herbst Astronomy Department, Wesleyan University, Middletown, CT 06459 [email protected] and James P. Greenwood Earth & Environmental Sciences Department, Wesleyan University, Middletown, CT 06459 [email protected] Received ; accepted 5 1 0 2 c e D 7 . ] P E h p - o r t s a [ 1 v 4 6 0 2 0 . 2 1 5 1 : v i X r a -- 2 -- ABSTRACT We propose that chondrules are formed by radiative heating of pre-existing dust clumps during close fly-bys of planetesimals with incandescent lava at their surfaces. We show that the required temperatures and cooling rates are easily achieved in this scenario and discuss how it is consistent with bulk aspects of chondritic meteorites, including complementarity and the co-mingling of FeO- poor and FeO-rich chondrules. Subject headings: Cosmochemistry; Meteorites; Solar Nebula -- 3 -- Chondrules are small (0.1-2 mm) spheres of igneous rock that make up 30-80% of the volume of primitive meteorites. It is generally agreed that they formed 1 - 3 Myr after the CAIs (Kita & Ushikubo 2012; Kita et al. 2013). Their textures require heating to peak temperatures of 1750-2100 K, but not above, and for only a few minutes (Greenwood & Hess 1996; Hewins & Connolly 1996). Cooling rates are inferred to be rapid, but not too rapid, typically 100-1000 K/hr (Hewins & Radomsky 1990; Hewins et al. 2005; Desch et al. 2012). What could possibly provide the appropriate dose of heat at just the right time and in just the right manner? This question has been labeled as "arguably the the major unresolved question in cosmochemistry" (Palme, Hezel & Ebel 2015). Ciesla (2005) reviewed theories of a decade ago, categorizing them as nebular and planetary. Nebular theories, which probably remain most popular today, assume that pre-chondrule dust clumps are entrained in a H/He gaseous disk and heated by some process that deposits energy in the gas. A variant of this, the X-wind model, uses the gas component of the disk to transport the pre-chondrule material close enough to the Sun to reach the required temperatures and then back to ∼3 AU for incorporation into the chondrite meteoroid. The popularity of these theories comes in part from their ability to predict thermal histories that match fairly well what laboratory experiments tell us is required to form igneous spheres with the observed, mostly porphyritic, textures of actual chondrules. Desch et al. (2012) compare the experimental results with predictions of leading nebular models and conclude that pressure shocks induced by gravitational instabilities within the gaseous disk do a better job than other leading nebular theories, such as planetesimal bow shocks, lightning, or the X-wind model, in matching inferred thermal histories. A serious challenge to all nebular theories comes from the elegant work of Alexander et al. (2008). These authors show convincingly from the distribution of Na2O in -- 4 -- many chondrules that these objects could only have formed at very high densities of solids, ρs, in the range ρs ≈ 10−5 − 10−3 gm cm−3. Since the surface density of solids (Σs) in the minimum mass solar nebula is Σs ≈ 10 gm cm−2, to achieve those number densities requires compressing the disk to a full-width thickness (W) of W = Σs/ρ < 10 km. Obviously this is difficult if the solids are still embedded in a gaseous nebula at the time they are heated to chondrule formation temperatures. In addition, Fedkin et al. (2012) show that significant (unobserved) isotopic signatures would be expected in chondrules formed under any plausible nebular conditions. Suppressing these signatures requires higher solids density, higher gas pressure and/or shorter heating times than nebular shock models can produce. Another challenge comes from the increasingly accurate measurements of chondrule and CAI ages by Connelly et al. (2012) and Bollard, Connelly & Bizzarro (2014). These authors confirm that the epoch of chondrule formation is significantly longer than the epoch of CAI formation, continuing for at least 2-3 Myr. They further substantiate that there was a delay in the formation time of most chondrules of about 1 Myr from t = 0, as measured by the CAIs. This demonstrates that the heating event associated with CAI formation, normally taken to be the hot nebular phase expected as infalling gas from the protostellar envelope dissipates its kinetic energy in transitioning to an accretion disk, is not the same heating event (or events) responsible for chondrule formation. Delaying the main chondrule heating event by 1 Myr and extending it for another 2 Myr before shutting it down forever is a serious challenge to all hypotheses, nebular and planetary alike. Planetary models invoke collisions between planetesimals to form chondrules. The complexity of such a process has stymied our ability to prove by numerical simulation or other means that chondrules with the observed range of properties could actually be made in this manner. While gaining some popularity recently, perhaps because of the difficulties being encountered by nebular models, planetary models have had relatively few adherents. -- 5 -- Two obstacles to any successful planetary model, raised by Ciesla (2005), bear repeating: 1) collisions occur throughout the history of the solar system, yet the chondrule formation time is clearly demarcated, and 2) there is evidence for multiple heating of chondrules. A further challenge to planetary models comes from observations of the phenomenon known as complementarity. Recent studies (Ebel et al. 2008; Palme, Hezel & Ebel 2015) extend the evidence along lines first presented by Wood (1963). It is found that, while bulk compositions of carbonaceous chondrites are consistently solar, chondrules have low Si/Mg and Fe/Mg ratios, while matrix has high ratios. The authors argue that this reflects pre-accretionary processes and conclude, therefore, that the chondrules and the matrix must have formed from a single reservoir of solar composition. An independent origin of chondrules followed by transport and mixing with matrix from elsewhere prior to formation of the meteorite parent body is excluded. At face value, this rules out the X-wind jet nebular hypothesis and severely complicates any planetary model. Nonetheless, recent versions of both nebular and planetary models have been developed, some with the intent of addressing these challenges. On the planetary side there is the idea of collisions of already molten planetesimals proposed by Sanders & Scott (2012) and the impact jetting model of Johnson et al. (2015). On the nebular side, there are modern versions of the pressure shock hypothesis (Hood & Weidenschilling 2012) and current sheets (McNally et al. 2013), but none successfully meets all the challenges of the data. Here we propose a new mechanism for chondrule formation that is neither nebular nor planetary. We assume, as in nebular theories that chondrules are formed from pre-existing dust clumps held together by electrostatic force that are probably part of larger aggregates. But we take seriously the evidence that, by the time the chondrules are heated - producing igneous spheres of the observed textures - there is little or no nebular gas remaining in their environment. Whatever gas is present comes from the evaporation of volatiles within what -- 6 -- we infer to be the pre-chondrite aggregate. The heating agent is radiation from hot lava at the surfaces of planetesimals, probably in the 10-100 km size range. It is known from the properties and ages of iron meteorites that relatively small (r ∼50 km) molten planetesimals were present in the forming solar system within the first ∼1 Myr of its existence (Haack, Rasmussen & Warren 1990; Greenwood et al. 2005; Schersten et al. 2006). The most likely heat source for these objects is 26Al. As Lee, Papanastassiou & Wasserburg (1976, 1977) first showed, its inferred abundance at t = 0 is sufficient to melt planetesimals larger than a few km on a time scale of ∼3 ×105 yr. More recent and detailed models by Sanders & Scott (2012) confirm this scenario, finding that planetesimals with r≥10 km will fully melt within 0.3-2.5 Myr. Based on these studies we may reasonably conclude that there was an abundance of magma present in the pre-planetary disk at the time of chondrule formation, encased in 10-100 km radii planetesimals, where it had been incubating since the epoch of planetesimal formation. From time to time some of this molten material must have reached the surfaces of these significantly overheated planetesimals, resulting in crustal foundering and the appearance of incandescent lava at the surface. Even today in the Solar System we regularly see hot spots at the surface of the most overheated body, Io (Keszthely et al. 2007). Could the radiation from hot surface lava and exposed magma oceans (Greenwood et al. 2005) -- plausibly be the heat source for melting pre-chondrule dust aggregates within highly solids-enriched aggregates? Fortunately, this proposal has sufficiently simple physics that we can robustly predict a thermal history for the chondrules if this is how they formed. As we now show, that thermal history is both independent of the size of the planetesimal providing the heat and an intriguingly good match with what the textural data require. To quantitatively assess the plausibility of this idea we consider a fully molten planetesimal of density, ρ, radius, r, and surface temperature, Ts, where to be useful for -- 7 -- chondrule heating, Ts will need to be equal to, close to, or above the maximum liquidus temperature for chondritic material. For the purposes of this exercise, we will adopt ρ = 3 gm cm−3 and Ts = 2150 K. The choice of density affects the shape of the cooling curves while the choice of Ts, which is based on calculations reported by Hewins & Radomsky (1990), affects how close a fly-by of the surface is needed to heat material sufficiently for chondrule formation. We neglect reflective losses, assuming that all of the radiant energy incident on the dust aggregate will be absorbed within it. Our initial focus is on the individual dust clumps that will form chondrules. Their aggregation, which may form the chondrite, could influence the heating/cooling relation by lengthening the time it takes to cool, but discussion of that point is beyond the scope of this exploratory study. Material passing close enough to the surface of an object with an exposed magma ocean can get heated by radiation to the sub-liquidus temperatures required for chondrule formation. Obviously, the material cannot be heated above the liquidus temperature and since the dust would need to be very close to the surface to reach the liquidus, it is clear that in a scenario like this, a sub-liquidus temperature will be the norm -- as required of any chondrule formation theory. Chondrule-mass objects will quickly come into equilibrium with the radiation field and we assume they do so instantaneously. To estimate cooling rates, we consider the case of a parabolic orbit and neglect gas drag and any effects of re-radiation from surrounding chondrules. Small planetesimals in the 10 -100 km range are unable to hold an atmosphere, although perhaps outgassing of volatiles from the magma or evaporation of the silicates could produce an expanding "exosphere" that should be factored into any more detailed model in the future. Hydrogen and helium associated with the solar nebula may be safely ignored at the high density of solids scenario envisioned here. In this exploratory study we, therefore, need consider only the dynamical effects of gravity. The resulting heating and cooling curves are shown in Figs. 1 and 2. It is evident that, -- 8 -- with the simplest of assumptions and underlying physics, we obtain a predicted thermal history for this model that can be directly compared with observational constraints. As Desch et al. (2012) discuss, constraints on the thermal history of chondrules are the most powerful tool for assessing viability. An important aspect of our model is that porphyritic chondrules, the most abundant texture, require subliquidus temperatures to form (Lofgren 1996) and that is precisely the temperature range the model predicts. The examples shown are for a planetesimal of radius 100 km (Fig. 1) and 10 km (Fig. 2) but the peak temperature and shape of the curves depends only on the ratio x = h/r. Smaller planetesimals provide less acceleration to the passing dust but they also heat smaller volumes and these effects offset each other to produce heating and cooling curves that depend only on the ratio, x, not on r. It is clear from these figures that a simple model of radiative heating by close passage to a planetesimal with a molten lava surface, under the action of gravity alone, accounts for both the peak temperatures and basic cooling rates required to explain the formation of chondrules. There are two ways in which the model thermal history presented in Fig. 2 differs from the most successful nebular models, pressure shocks, discussed by Desch et al. (2012). First, there is no "flash heating" but, instead, a symmetric (with the cooling) and more gentle rise in the temperature as the pre-chondrule material approaches the hot lava surface. Second, the simple model explored here has a generally more rapid cooling rate at later times than the pressure shock models, with temperatures remaining above 1400 K for only an hour or so, compared to one or two days in the pressure shock scenarios. These differences could be significant in testing our hypothesis in the future; compared to other models there does not seem to be much latitude in the thermal history we can accommodate, an aspect of our model that we consider to be a strength. If chondrules are proven to require a heating history significantly different from what is shown in Fig. 2 to form, then our model may be ruled out. At present, the issue is debatable because the main argument for flash heating is -- 9 -- the presence of volatiles within chondrules but if volatiles can evaporate during heating and re-condense during cooling, then the constraint of flash heating is relaxed (Alexander et al. 2008; Desch et al. 2012). Slower cooling rates at later times could be generated in our scenario by including radiation from surrounding matter, but it is not entirely clear to us that experimental data currently requires cooling rates significantly below ∼200 K/hr. We believe, therefore, that the predicted thermal history for this model is consistent with all current constraints on chondrule formation and that future developments on the experimental side may eventually strengthen or falsify our model. The heating model proposed here also provides a natural explanation for processes requiring a range of temperatures seemingly within the same small volume of space, such as the formation of two chemical types of chondrules -- FeO-poor and FeO-rich. Pre-meteorite material further from the surface magma will be cooler than that closer to it for two reasons: 1) the proximity effect, and 2) attenuation of the radiation field by the dust closer in. If the ambient radiation field responsible for heating material is attenuated by a factor e−τ then the equilibrium temperature reached by the solids will be reduced by a factor e−τ /4. At an optical depth of τ = 1 the equilibrium temperature is reduced by 22%, corresponding to several hundred degrees. This is sufficient to account for the difference between chondrule types. The path length within the aggregated dust that is required to cause such attenuation depends on the opacity of the material. We can estimate the path length corresponding to τ = 1 as follows. Assume that the matter is concentrated into spheres of radius, r, and grain density, ρg. The optical depth is given by τ = kl, where k is the volumetric opacity; k−1 is the mean free path, l, of a photon in this material. For absorbers of a single size we may write that k = nσ where n is their number density and σ is the extinction cross-section per absorber. Following Mie theory, we take σ = 2πr2. -- 10 -- Calculating n as ρs/m where m is the mass of each dust grain yields l = k−1 = 2rρg 3ρs It seems safe to say that r ≤ 0.1 cm and ρg ≤ 3 gm cm−3, so l ≤ 0.2/ρs. At a typical density for chondrule formation based on the Na2O results of Alexander et al. (2008), namely ρs ≈ 10−4 gm cm−3, l ≤ 20 meters. If chondrules typically require such densities of solids and if the radiative heating mechanism proposed here operates, one will expect to have a wide temperature range within the formation zone. Of course when full aggregates are considered instead of individual dust clumps, there will be some moderation of these effects. The discussion above suggests that the size of the chondrule formation zone, L, may be much smaller than has previously been considered in nebular models. For example, Desch et al. (2012) argue that L ≫ 103 km is required, based on the percentage of chondrules (5%) that are compound, indicating that they collided while still hot. Adopting the same numbers as in the previous paragraph, and r = 0.03 cm, we can calculate the number density of chondrules in the formation zone as n = ρs 2m ≈ 0.12 cm−1, where we have assumed that half of the solids mass is in matrix material. This, in turn, implies a mean free path for the chondrules of L = (nπr2)−1 = 30 m. The constraint that 5% of chondrules suffer a collision while hot would be satisfied if their relative velocity, vrel, satisfied the constraint nπr2vrelt = 0.05, where t is the duration of the chondrule hot phase. Based on Fig. 2 we take t = 1 hr, which yields vrel = 0.04 cm s−1. This velocity is sufficiently low that there seems little doubt that it could be achieved. Clearly, our picture of chondrule formation suggests a more intimate link with the process of chondrite formation than is the case for nebular models. The expected temperature range over the chondrule formation zone can account for the fact that FeO-poor (Type I) chondrules and FeO-rich (Type II) chondrules -- 11 -- are found intermingled in chondritic meteorites. The FeO-rich chondrules are more oxidized, have more volatiles, have lower liquidus temperatures and have higher ∆17O compositions than the FeO-poor chondrules (Alexander et al. 2008; Ushikubo et al. 2012; Schrader et al. 2013). It has been challenging for models of chondrule formation to account for these fundamental chemical differences in the two main types of chondrules (Villeneuve, Libourel & Souli´e 2012). That these two types of chondrules, which apparently formed in different environments, are found intimately mixed in the chondritic meteorites, has required separate formation locations followed by physical mixing in some later process such as Sanders & Scott (2012), for example, have proposed. On the other hand, the constraint of complementarity requires that the FeO-rich and FeO-poor chondrules, and their associated matrix, formed in close proximity to one another (Palme, Hezel & Ebel 2015). We note further that volatiles lost from the most intensely heated (inner region) of the material could have led to an enhancement of the local oxygen fugacity and other volatiles, such as Na, within the cooler, outer zone. Minor incorporation of oxygen from nebular ices into the FeO-rich chondrules can account for their oxygen isotope differences from FeO-poor chondrules, as nebular ice is expected to be highly enriched in ∆17O (Sakamoto et al. 2007). Can our proposed mechanism account for the large volumetric abundance of chondrules within chondritic meteorites? Yes, if we presume that chondritic meteorites form during the same heating events that form the chondrules. Gravitational focusing will lead to a significant compression of the incoming material. Coupled with the heating experienced by the chondrules and matrix one has the ingredients needed for bulk meteorite formation. Fine grained dust, such as that which composes the matrix, might survive radiative heating at 2000 K since it is too small to efficiently absorb energy emitted near the peak of the black body curve. Coarser grains will evaporate under conditions where highly refractory material and very fine grained material might survive, possibly accounting for the basic -- 12 -- component structure of chondrite meteorites. If the heating and compression required for meteorite formation involves a sufficiently close passage past a hot planetesimal that chondrule formation inevitably occurs, then the fact that chondrules are so abundant in primitive meteorites is explained. We note that, if this line of argument is correct, it means that chondrules are an abundant component of primitive meteorites, but not of the early solar system in general. To summarize, we propose that chondrules are formed when pre-existing dust clumps, probably within aggregates, are heated during close encounters with incandescent lava at the surfaces of planetesimals and that chondritic meteorites are probably often formed in the same heating and compression events. The model accounts or potentially accounts for all of the well-established features of chondrules and chondritic meteorites, notably including their thermal history, age range, complementarity, formation environment and ubiquity within chondrites. In particular, we note the following: 1. Experimental data indicates that chondrules formed from material heated to peak temperatures of at least 1750 - 1820 K, but not exceeding 2100 - 2370 K, for times of order minutes, which then cooled at rates of hundreds or, at most, a few thousands of degrees per hour. As Figs. 1 and 2 show, this is precisely the thermal history our model predicts for material passing within about 0.5 planetesimal radii of the surface, regardless of the size of the planetesimal involved. Unlike nebular models, we do not predict a highly asymmetric "flash" heating. In our model, the retention of volatiles and absence of large isotopic anomalies is due to the high vapor pressures associated with a high density of solids in the pre-chondrite aggregate and rather brief, overall, heating times. 2. The epoch of chondrule formation is readily understood in our model as the time in the history of the Solar System when molten lava was present in abundance throughout the terrestrial planet forming zone. The peak of this epoch is ∼1 Myr after CAI formation, -- 13 -- because it typically takes that long to incubate the energy available in 26Al to temperatures ≥ 2000 K within the interiors of planetesimals with radii of ∼10 km or larger. The epoch is over by ∼3 Myr because the decay of the 26Al fuel relieves the condition of overheating; molten lava becomes increasingly less common at the surfaces of small planetesimals thereafter. 3. Multiple heating events for the same chondrules, as observed, are certainly possible. Not every dust clump or aggregate of them that is heated during close passage to a hot surface needs to form a chondritic meteorite, although most will if complementarity is a widespread feature of chondrites. 4. The model potentially meets the constraints of complementarity since it has a local heat source acting on a reservoir of bulk solar composition. Newly formed chondrules and heated matrix from the same original reservoir may be concentrated within the gravitational wake of hot planetesimals, which we propose as a possible site of chondritic meteorite formation. 5. The radiative heating model easily accommodates the observation that chondrules are formed at a density of solids, ρs ≈ 10−5 −10−3 gm cm−3, since no nebular gas whatsoever is required in the heating process. The existence of FeO-poor and FeO-rich chondrules within the same meteorite may be understood as a consequence of the expected temperature range in the pre-meteorite aggregate caused by variable proximity to the surface radiation and the relatively high opacity in the near-infrared. The high opacity of the material limits the chondrule formation zone to tens of meters, consistent with the frequency of compound chondrules. 6. Chondrule formation experiments should be useful in testing the rather highly constrained and symmetric heating and cooling curves predicted by our model, which are different in significant ways from the predictions of nebular models. -- 14 -- We thank Denton Ebel and his staff at the American Museum of Natural History for their hospitality during visits by each of us and for many helpful lessons on chondrules and chondrites. We thank the editor and referees of this paper for helpful comments during the review process. -- 15 -- REFERENCES Alexander, C. M. O'D., Grossman, J. N. Ebel, D. S. & Ciesla, F. J. The Formation Conditions of Chondrules and Chondrites. Science 320, 1617 (2008) Bollard, J., Connelly, N. & Bizzarro, M. The Absolute Chronology of the Early Solar System Revisited. LPI Contribution No. 1800, id 5234 (2014) Ciesla, F. J. Chondrule-forming Processes -- An Overview. in Chondrites and the Protoplanetary Disk, ASP Conference Series, Vol 341, Edward R. D. Scott and Bo Reipurt, editors (San Francisco: Astronomical Society of the Pacific ), p 811 (2005 ) Connelly, J. N., Bizzarro, M., Krot, A. N., Nordlund, A., Wielandt, D. & Ivanova, M. A. The absolute chronology and thermal processing of solids in the solar protoplanetary Disk. Science 338, p. 651-655 (2012) Desch, S. J., Morris, M. A., Connolly, H. C. & Boss, A. P. The Importance of experiments: Constraints on chondrule formation models. Meteoritics and Planetary Science 47, p. 1139-1156 (2012) Ebel, D. S., Weisberg, M. K., Hertz, J. & Campbell, A. J. Shape, Metal abundance, chemistry and origin of chondrules in the Renazzo (CR) chondrite. Meteoritics and Planetary Science 43, p. 1725-1740 (2008) Fedkin, A. V., Grossman, L., Ciesla, F. J. & Simon, S. N. Mineralogical and isotpoic constraints on chondrule formation from shock wave thermal histories. Geochimica et Cosmochimica Acta, 87, pp. 81-116 (2012) Greenwood, J.P. and Hess, P.C. Congruent Melting Kinetics: Constraints on Chondrule Formation. Chondrules and The Protoplanetary Disk. (eds. Boss, A.P., Hewins, R.H., Jones R.H., and Wasson, J.T.). Cambridge University Press. pp. 205-211 (1996) -- 16 -- Greenwood, R. C., Franchi, I. A., Jambon, A. & Buchanan, P. C. Widespread magma oceans on asteroidal bodies in the early Solar System. Nature 435, 916 (2005) Haack, H., Rasmussen, K. L. & Warren, P. H. Effects of Regolity/Megaregolith Insulation on the Cooling Histories of Differentiated Asteroids. Journal of Geophysical Research 95, pp. 5111-5124 (1990) Hewins, R. H. and Connolly, Jr., H. C Peak Temperatures of Flash-melted Chondrules. Chondrules and The Protoplanetary Disk. (eds. Boss, A.P., Hewins, R.H., Jones R.H., and Wasson, J.T.). Cambridge University Press. pp. 197-204 (1996). Hewins, R. H. & Radomsky, P. M. Temperature conditions for chondrule formation. Meteoritics 25, p. 309-318 (1990) Hewins R. H., Connolly H. C., Lofgren G. E. and Libourel G. Experimental constraints on chondrules formation. In Chondrites and the Protoplanetary Disk, ASP Conference Series (eds. A. N. Krot, E. R. D. Scott and B. Reipurth). Sheridan Books, Ann Arbor, Michigan, pp. 286316 (2005) Hood,L. L. & Weidenschilling, S. J. The planetesimal bow shock model for chondrule formation: a more quantitative assessment of the standard (fixed Jupiter) case. Meteoritics and Planetary Science 47, 1715 (2012) Johnson, B. C., Minton, D. A., Melosh, H. J. & Zuber, M. T. Impact jetting as the origin of chondrules. Nature 517, 339 (2015) Keszthely, L., Jaeger, W., Milazzo, M., Radebaugh, J., Davies, A. G. & Mitchell, K. L. New estimates for Io eruption temperatures: Implications for the interior. Icarus 192, p. 491-502 (2007) Kita, N. T. & Ushikubo, T. Meteoritics & Planetary Science 47, 1108-1119 (2012) -- 17 -- Kita, N. T. et al. Meteoritics & Planetary Science 48, 1383-1400 (2013) Lee, T., Papanastassiou, D. A. & Wasserburg, G. J. Demonstration of 26Mg excess in Allende and evidence for 26Al. Geophysical Research Letters 3, p. 41-44 (1976) Lee, T., Papanastassiou, D. A. & Wasserburg, G. J. Aluminum-26 in the early solar system: Fossil or fuel? The Astrophysical Journal 211, L107 - L110 (1977) Lofgren, G. E. A dynamic crystallization model for chondrule melts. In Chondrules and the Protoplanetary Disk, (eds. R. H. Hewins, R. H. Jones, E. R. D. Scott), Cambridge University Press, Cambridge, UK. p. 187 - 196(1996) McNally, C. P., Hubbard, A., Mac Low, M-M., Ebel, D. S., & D'Alessia, P. Mineral processing by short circuits in protoplanetary disks. The Astrophysical Journal 767, p. L2-L7 (2013) McSween, H. Y. & Huss, G. R. Cosmochemistry (Cambridge, UK; Cambridge University Press), p. 494 (2010) Palme, H., Hezel, D. C. & Ebel, D. S. The origin of chondrules: Constraints from matrix composition and matrix-chondrule complementarity. Earth and Planetary Science Letters 411, 11 (2015) Sakamoto, N., Seto, Y., Itoh, S., Kuramoto, K., Fujino, K., Nagashima, K., Krot, A. N. & Yurimoto, H. Remnants of the Early Solar System Water Enriched in Heavy Oxygen Isotopes. Science 317, 231 (2007) Sanders, I. S. & Scott, E. R. D. The origin of chondrules and chondrites: Debris from low-velocity impacts between molten planetesimals? Meteoritics & Planetary Sciences 47, p. 2170 - 2192 (2012) -- 18 -- Schersten, A. Elliott, T., Hawkesworth, C., Russell, S., & Masarik, J. HfW evidence for rapid differentiation of iron meteorite parent bodies. Earth Planetary Science Letters 241, p. 530 - 542 (2006) Schrader, D. L., Connolly, H. C., Lauretta, D. S., Nagashima, K., Huss, G. R., Davidson, J. & Domanik, K. J. The formation and alteration of the Renazzo-like carbonaceous chondrites II: Linking O-isotope composition and oxidation state of chondrule olivine. Geochimica et Cosmochimica Acta 101, 302 (2013) Shu, F. H., Shang, H., Gounelle, M., Glassgold, A. E. & Lee T. The origins of chondrules and refractory inclusions in chondritic meteorites. The Astrophysical Journal 548, 1029 (2001) Ushikubo, T., Kimura, M. Kita, N. T. & Valley, J. W. Primodial oxygen isotope reservoirs of the solar nebula recorded in chondrules in Acfer 094 carbonaceous chondrite. Geochimica et Cosmochimica Acta 90, 242 (2012) Villeneuve J., Libourel G. & Souli´e C. Relationships between type I and type II chondrules: Implications on chondrule formation processes, Geochimica et Cosmochimica Acta, 160, 277-305, (2015) Wood, J. A. On the origin of chondrules and chondrites. Icarus 2, p. 152-180 (1963) This manuscript was prepared with the AAS LATEX macros v5.2. -- 19 -- Fig. 1. -- Cooling curves for chondrules from our model compared to observed rates of 100- 1000 K/hr. These rates are independent of the planetesimal radius, depending only on its density (see Fig. 2). -- 20 -- Fig. 2. -- Same as Fig. 1 but showing a larger range of time, including the pre-heating phase. These calculations were done for a 10 km radius planetesimal but, again, the form of the curves is independent of the size of the planetesimal.
1907.09121	2	1907	2019-08-19T03:24:15	Near-resonance tidal evolution of the Earth-Moon system influenced by orbital-scale climate change	[ "astro-ph.EP", "physics.ao-ph" ]	We build a conceptual coupled model of the climate and tidal evolution of the Earth-Moon system to find the influence of the former on the latter. An energy balance model is applied to calculate steady-state temperature field from the mean annual insolation as a function of varying astronomical parameters. A harmonic oscillator model is applied to integrate the lunar orbit and Earth's rotation with the tidal torque dependent on the dominant natural frequency of ocean. An ocean geometry acts as a bridge between temperature and oceanic frequency. On assumptions of a fixed hemispherical continent and an equatorial circular lunar orbit, considering only the 41 kyr periodicity of Earth's obliquity $\varepsilon$ and the $M_2$ tide, simulations are performed near tidal resonance for $10^6$ yr. It is verified that the climate can influence the tidal evolution via ocean. Compared with the tidal evolution with constant $\varepsilon$, that with varying $\varepsilon$ is slowed down; the Earth-Moon distance oscillates in phase with $\varepsilon$ before the resonance maximum but exactly out of phase after that; the displacement of the oscillation is in positive correlation with the difference between oceanic frequency and tidal frequency.	astro-ph.EP	astro-ph	Research in Astron. Astrophys. Vol.0 (20xx) No.0, 000 -- 000 http://www.raa-journal.org (LATEX: RAA-2018-0288.tex; printed on August 20, 2019; 1:44) http://www.iop.org/journals/raa Research in Astronomy and Astrophysics Near-resonance Tidal Evolution of the Earth-Moon System Influenced by Orbital-scale Climate Change Nan Wang and Zhi-Guo He Ocean College, Zhejiang University, Hangzhou 310058, China; [email protected] Received 20xx month day; accepted 20xx month day Abstract We build a conceptual coupled model of the climate and tidal evolution of the Earth-Moon system to find the influence of the former on the latter. An energy balance model is applied to calculate steady-state temperature field from the mean annual inso- lation as a function of varying astronomical parameters. A harmonic oscillator model is applied to integrate the lunar orbit and Earth's rotation with the tidal torque dependent on the dominant natural frequency of ocean. An ocean geometry acts as a bridge between temperature and oceanic frequency. On assumptions of a fixed hemispherical continent and an equatorial circular lunar orbit, considering only the 41 kyr periodicity of Earth's obliquity ε and the M2 tide, simulations are performed near tidal resonance for 106 yr. It is verified that the climate can influence the tidal evolution via ocean. Compared with the tidal evolution with constant ε, that with varying ε is slowed down; the Earth-Moon distance oscillates in phase with ε before the resonance maximum but exactly out of phase after that; the displacement of the oscillation is in positive correlation with the difference between oceanic frequency and tidal frequency. Key words: Moon -- planets and satellites: dynamical evolution and stability -- planets and satellites: oceans 1 INTRODUCTION The tidal evolution of the Earth-Moon system is a classic problem, but has not yet been fully solved. The general trend of the tidal evolution has long been confirmed: since the Moon was born 4.5 Gyr ago (Halliday 2008), it has been receding from the Earth with tidal energy dissipated as heat, and Earth's ro- tation has been slowing down with angular momentum transferred to the lunar orbit (Murray & Dermott 1999). When one natural frequency (also called a normal mode) of the ocean and one tidal forcing frequency, both varying over geologic time, come close to each other, the ocean tide gets excited and the dissipation is largely enhanced. This phenomenon of "tidal resonance" speeds up tidal evolution. Currently, the M2 (semidiurnal tide) resonance in the ocean contributes a dissipation of 2.4 TW (Munk 1997) to the total of 3.7 TW (Munk & Wunsch 1998), which is so abnormally high that extrapolating into the past unrealistically puts the Moon where it was born as recently as ∼ 2 Gyr ago (e.g, Touma & Wisdom 1994; Bills & Ray 1999). Because of the uncertainty that tidal resonance brings, to quantita- tively reconstruct the lunar orbit, the history of oceanic natural frequencies has to be acquired. The natural frequencies of the ocean are determined by its geometry (position, shape and depth), which is associated with climate change and continental drift. In order to simulate tidal evolution driven by ocean tide, Hansen (1982) used Laplace's tidal equations to determine the oceanic tidal torque for two ocean/land geographies, but not only the geography but also the ocean depth remained unchanged in simulations. Webb (1980, 1982a,b) developed a model of average ocean, a statistical average over 2 Nan Wang and Zhi-Guo He many hemispherical oceans centered at various positions relative to Earth's axis, to take the change in ocean geometry due to continental drift into account, whereas the ocean shape and depth were constant. Kagan & Maslova (1994) built a stochastic model, which considered the effect of continental drift as explicit fluctuations in oceanic natural frequencies. By solving the timescale problem (e.g., Goldreich 1966), those ocean modelers demonstrated the importance of the ocean to tidal evolution of the Earth- Moon system, but their results are still qualitative without further improvement for decades, and none has involved climate. The timescale of continental drift is 108 yr (Murray & Dermott 1999), and that of orbital-scale climate change is ∼ 105 yr (Berger 2012). Considering only continental drift means the secular effect of climate influence can be neglected, which is based on insufficient evidence. Hence, we simulate the tidal evolution for 106 yr, so that the influence of climate can be investigated and continental drift can be reasonably neglected. Natural quasi-periodicities of the climate over timescales in a vast range have been discovered through geological records. On the orbital-scale of ∼ 105 yr, the glacial-interglacial cycle dominates (Berger 2012). During a glacial, the ice sheets extend towards the equator and the sea level drops; whereas during an interglacial, the ice sheets shrink towards the poles and the sea level rises. According to the Milankovitch theory, that results from the secularly varying orbit and rotation of the Earth per- turbed by the Sun and other planets, and the subsequent variation of insolation distribution on the terres- trial surface (e.g., Berger 1988). This effect on climate of the astronomical parameters (Earth's eccen- tricity, obliquity and climatic precession) is called astronomical forcing. The change in sea level, which the glacial-interglacial cycle is accompanied by, alters the natural frequencies of the ocean and then the state of tidal resonance. For instance, in the last glacial maximum ∼19 -- 26 kyr ago, the sea level drop of about 130 m (Yokoyama et al. 2000; Clark et al. 2009; Lambeck et al. 2014) leads to considerably higher dissipation than at present (Thomas & Sundermann 1999; Egbert et al. 2004; Griffiths & Peltier 2009; Green et al. 2009). Therefore, it is the ocean that acts as the bridge between climate and tidal evolution. In this work, a coupled model of climate and tidal evolution is proposed. An energy balance model (Sellers 1969; Budyko 1969) is applied to simulate the climate in response to astronomical forcing. It is a kind of conceptual model focusing on major climate components and interactions. Though very sim- plified, it is capable of reproducing the glacial variability with ice sheet involved (Huybers & Tziperman 2008; McGehee & Lehman 2012) and frequently used to study the climate stability (see review in North (1984) for early studies; Lin & North 1990; Wagner & Eisenman 2015). A harmonic oscillator model (Munk 1968; Murray & Dermott 1999) is applied to integrate the lunar orbit and Earth's rotation with the oceanic natural frequency given. It is also a conceptual model, where the response of the ocean to the tidal forcing is compared to that of a harmonic oscillator, and is capable of providing a realistic timescale of tidal evolution (Hansen 1982; Kagan & Maslova 1994). It suits the case when there is one dominant oceanic natural frequency and one dominant tidal forcing frequency. In addition, a simplified ocean geometry is assumed to obtain the natural frequency from temperature field. As a preliminary effort to study the influence of climate on tidal evolution, we aim at verifying the existence of the influence and qualitatively observing its nature and mechanism. Therefore, our idealized model and simulation time (106 yr) are appropriate. The period of interest is when the ocean and tidal forcing are near resonance (but not at the resonance maximum), so that the influence can be amplified. Section 2 describes the coupled model and the numerical method, and Section 3 exhibits the results of two sets of simulations in pre- and post-resonance times. A discussion about simulation results and potential improvements is in Section 4. 2 MODEL AND METHOD 2.1 Climate Model 2.1.1 Steady-state temperature field The climate can be altered by ocean/land geography. To study climate change resulting from astronomi- cal forcing, a simple geography is used and taken to be invariant. It is assumed that a single spherical-cap Tidal Evolution Influenced by Climate Change 3 continent is centered at the North Pole, extending to latitude ϕl, and the rest of Earth's surface is cov- ered by ocean. Such a geography is similar to what Mengel et al. (1988) used. Neglecting the vertical structure of the atmosphere for this zonally symmetric planet, a one-dimensional climate model can be applied. Considering horizontal thermal diffusion, outgoing infrared radiation as well as the solar heating being the only external forcing, an energy balance model leads to this governing equation (North & Kim 2017) C(ϕ) ∂T (ϕ, t) ∂t 1 cos ϕ ∂ ∂ϕ [ − D ∂T (ϕ, t) cos ϕ ∂ϕ ] − [A + BT (ϕ, t)] = W (ϕ)α(ϕ). (1) The climate at time t is just characterized by the temperature field T (ϕ) on the surface. In the first term on the left side, C is the effective heat capacity and controls the climate response to perturbations (relaxation time τ = C/B). The capacity over the ocean Cw is larger than that over the land Cl, and so is the relaxation time for ocean τw (a few years) than that for land τl (a month). According to the geography assumed, C(ϕ) =(Cl, Cw. (ϕ > ϕl) (ϕ < ϕl) (2) The second term involving the thermal diffusion coefficient D allows the heat transport from warm areas to cool. The third term allows the infrared radiation to space and A and B are empirical coefficients from satellite observations. The term on the right side determines the solar radiation absorbed. The insolation function W (ϕ) gives the latitudinal distribution of the solar radiation flux delivered to the surface. It is dependent on Earth's orbital status (Sect. 2.1.2). The coalbedo α(ϕ) gives the fraction of radiation absorbed by the surface. Its mean annual form is well represented by α(ϕ) = α0 + α2P2(sin ϕ), (3) where constants α0 and α2 are from satellite observations, and the second-order Legendre polynomial P2(sin ϕ) = (3 sin2 ϕ − 1)/2. The values of the constants mentioned above and the references that provide them are listed in Table 1. Solving Equation 1 also needs the boundary condition ∂T ∂ϕ = 0, (ϕ = ±90◦) (4) implying that there is no net heat flux into the poles. On the assumption of energy balance (the energy absorbed is equal to the energy lost), for every given insolation function W (ϕ), there exists a corresponding steady-state solution T s(ϕ), which any temperature field T (ϕ) of a different profile relaxes to after a time comparable to τ . The relaxation time τ of the climate system is much smaller than the astronomically driven period of mean annual W (ϕ). Therefore, W (ϕ) can be taken as invariant while T (ϕ) is evolving towards the steady state T s(ϕ). For a steady-state temperature field T s(ϕ), the iceline ϕf , the edge of the permanent ice cap, is determined by (5) where the mean annual isotherm Tf = −10◦C (North & Jr. 1979). We note that although the iceline is defined, expressions of the capacity C(ϕ) and coalbedo ¯α(ϕ) are not influenced by it, that is, the ice- albedo feedback is not included in the present work. The iceline position only influences the sea level (Sect. 2.2). T s(ϕf ) = Tf , 2.1.2 Insolation distribution Given the timescale of interest, the mean annual version of the insolation function W (ϕ) is used, with seasonal variation averaged. It is dependent on Earth's semimajor axis a, eccentricity e and obliquity ε (Loutre et al. 2004). Assuming the Sun is a point source, Berger et al. (2010) deduced with elliptical 4 Nan Wang and Zhi-Guo He integrals the total energy available during any time interval of one year on a given unit surface. Based on their result, we deduce the expression of the mean annual insolation function as cos ϕ ) − tan2 ϕ cos2 εΠ(sin2 ε, sin ε sin ε ) − sin2 ϕ cot2 εΠ(cos2 ϕ, cos ϕ cos ϕ )], sin ε )], (ϕ ∈ [0◦, 90◦ − ε)) (ϕ ∈ (90◦ − ε, 90◦)) (ϕ = 90◦ − ε) (ϕ = 90◦) W (ϕ, a, e, ε) = L⊙ cos ϕ L⊙ sin ε 2π3a2√1−e2 [E( sin ε 2π3a2√1−e2 [E( cos ϕ 2π3a2√1−e2 [sin ε + cos2 ε 4π2a2√1−e2 sin ε. cos ϕ ) + tan2 ϕK( sin ε sin ε ) + cot2 εK( cos ϕ 1−sin ε )], ln( 1+sin ε L⊙ L⊙ 2   (6) The solar luminosity L⊙ is a constant, and K, E and Π are the first, second and third complete elliptical integrals, respectively. This expression is valid for 0◦ < ε < 90◦ and 0 < e < 1. Variation of W (ϕ) results from those of the astronomical parameters e and ε (mainly over a timescale of ∼ 105 yr). Based on the numerical solution for Earth's orbit (e.g, Laskar 1988), e and ε can be expressed in trigonometric form as quasi-periodic functions of t: approximation + P{(amplitude)i cos [(frequency)it + (phase)i]} (Berger 2012). In this work, we only consider the most important term of ε, whose period is 41 kyr, in order to avoid the compound influence of its multiple terms and the complicated effect of e and ε simultaneously varying. Thus, e = ¯e and ε(t) = ¯ε + ∆ε cos(γt + ψ), (7) where values of the approximation ¯ε, the amplitude ∆ε and the frequency γ given by Berger & Loutre (1991) are used, which are valid over 1 -- 3 Myr (Berger & Loutre 1992). However, the phase ψ acts as a controllable parameter in our simulations to manifest the influence of itself. 2.2 Ocean Geometry Model An ocean geometry model is needed to connect climate and tidal evolution. An ocean function with a value of 1 over ocean and 0 over land can be defined to characterize the geography modeled in Section 2.1 h(ϕ) =(1, 0. (ϕ ≤ ϕl) (ϕ > ϕl) (8) The iceline ϕf divides the water on Earth's surface into two reservoirs, i.e., sea water in the ocean basin and ice sheet on the part of continent north of the iceline. It is further assumed that the depth of the ocean basin hb, the depth of the sea water hsw and the thickness of the ice sheet his are all uniform. Sea ice is not considered. By simply taking the volume of sea water as its depth times the area of the ocean basin and the volume of the ice sheet as its thickness times the area of ice cover, conservation of mass gives rise to hsw = hb − ρishis(1 − sin ϕf ) ρsw(1 + sin ϕl) , (9) where ρis and ρsw are densities of ice sheet and sea water, respectively, and hb is equal to the maximum of hsw which corresponds to the condition ϕf = 90◦. The oceanic natural frequency σ is determined by the geometry of the ocean. On the assumption of half-wavelength resonance, the ocean basin is simplified as a closed square and σ can be estimated to be π√ghsw l , σ = (10) where l is the ocean width and g is the gravitational acceleration. Typical values of his, ¯hsw (mean of hsw) and l are set (Table 1). The advantage in setting l to a typical value instead of deriving it from the modeled geography is that a realistic σ value can be obtained. Because the geography is invariant, the oceanic frequency is only dependent on climate, i.e., σ only varies with hsw. On the other hand, σ is yet needed in the calculation of tidal torque coefficient Z (Sect. 2.3.2). The climate and tidal evolution are thus connected. Tidal Evolution Influenced by Climate Change 5 2.3 Tidal Evolution Model 2.3.1 Orbital parameters A two-body system consisting of the Earth and Moon with a circular orbit in the Earth's equatorial plane is considered. With the Moon's rotational angular momentum neglected, conservation of angular momentum leads to IΩ + Mrnr2 = H. (11) The first term on the left side is Earth's rotational angular momentum, where I is Earth's rotational inertia and Ω is Earth's rotational speed. The second term is the lunar orbital angular momentum, where n is the lunar angular orbital speed, r is the Earth-Moon distance, and the reduced mass Mr = M⊕MM/(M⊕ + MM) (M⊕ and MM are masses of Earth and Moon, respectively). The total angular momentum H is constant. There are three orbital parameters, r, n and Ω, characterizing the state of tidal evolution. Besides Equation 11, n and r are also linked by Kepler's third law, n2r3 = G(M⊕ + MM), (12) where G is the gravitational constant. Therefore, knowledge of one of r, n and Ω is equivalent to that of them all. The evolution of Ω is determined by tidal torque L (Sect. 2.3.2), which arises because the Earth, carrying its tidal bulge, rotates faster than the Moon orbiting it (Ω > n) The tidal torque acts to decrease Ω, resulting in a transfer of angular momentum from Earth's rotation to lunar orbital motion and a dissipation of energy in Earth. I dΩ dt = −L. (13) 2.3.2 Tidal torque A tide is raised on an elastic body when this body is distorted in the gravity field of another. In our model, the solid Earth and Moon are taken as rigid spheres, while the ocean is a thin deformable layer partially covering the solid Earth. Therefore, only the tidal torque exerted on the Moon by the distorted ocean is present. Furthermore, only the semidiurnal tide M2, the dominant tidal constituent at present, is considered in this work for simplification. Hansen (1982) derived an approximate expression of the secular variation of the tidal constituent torque after averaging on a short timescale (monthly and yearly) and neglecting terms higher than (R⊕/r)8. For the assumed equatorial lunar orbit, that expression becomes L(σ, ω, r) = −L∗ · (rp/r)6 · Im(Z(σ, ω)), where the constant L∗, which carries the dimensionality of the torque, is (14) L∗ = 6 5 πρswR2 ⊕ · GM⊕ · (MM/M⊕)2 · (R⊕/rp)6, (15) rp is the present Earth-Moon distance and R⊕ is the radius of Earth. Additionally, the torque coefficient Z in Equation 14 is a second-degree spherical harmonic expan- sion coefficient of the complex tidal elevation response function. For a static ocean tide whose shape is the same as that of the tide potential, Z degenerates into a static torque coefficient Z static = [hY 2hi − hY hi2/hhi]/hY 2i, (16) where the complex spherical harmonic Y = P 2 Legendre function P 2 defined in Equation 8, and the angled brackets imply an areal integration over the global surface. 2 (cos θ)ei2λ for M2 tide (the unnormalized associated 2 (cos θ) = 3 sin2 θ, θ is colatitude and λ is longitude), the ocean function h is 6 Nan Wang and Zhi-Guo He Table 1: Values of Model Constants. Constant Value Reference A B D Cl Cw α0 α2 ¯e ¯ε ∆ε γ ϕl Tf ρis ρsw his ¯hsw l δ 218 W m−2 North & Kim (2017) 1.90 W m−2 K−1 North & Kim (2017) 0.67 W m−2 K−1 North & Kim (2017) 0.08 yr·B Lin & North (1990) 4.80 yr·B Lin & North (1990) 0.68 North & Kim (2017) −0.20 North & Kim (2017) 0.028707 Berger (1978) 23.333410◦ Berger & Loutre (1991) −1969.00" Berger & Loutre (1991) 31.54068" yr−1 Berger & Loutre (1991) 0◦ present work −10◦C North & Jr. (1979) 0.917 g cm−3 Haynes (2017) 1.037 g cm−3 Hansen (1982) 2 km 4 km 4000 km 0.092 present work Notes: Values given with no reference are typical in reality. For a dynamic ocean tide, following Hansen (1982), the harmonic oscillator model is adopted to derive Z. A driven harmonic oscillator can be described as d2ζ dt2 + δσ dζ dt + σ2ζ = σ2ζ∗meiωt, (17) where ζ is the displacement from equilibrium, δ is the frictional resistance coefficient, σ is the natural frequency of the oscillator, ω is the frequency of the external force and ζ∗m is the limit of maximal dis- placement as ω approaches 0. Its steady-state solution is ζ = ζmeiωt, where the displacement amplitude is ζ∗m ζm = 1 − (ω/σ)2 + iδ(ω/σ) . (18) In an analogy with ocean tides, ζ is taken as the tidal elevation in the ocean, σ is the oceanic natural frequency, ω is the tidal forcing frequency, ζ∗m = Z static and ζm = Z. The tidal torque coefficient is thus derived Z(σ, ω) = . (19) Z static 1 − (ω/σ)2 + iδ(ω/σ) This tidal torque coefficient Z varies with σ and ω, for Z static absolutely determined by geography is constant in our model. The oceanic natural frequency σ is dependent on the state of ocean (Sect. 2.2), while the tidal forcing frequency for M2 tide is If σ and ω are close enough (not strictly equal for a nonzero δ), Im(Z) will be largely enhanced and so will L. Tidal evolution then speeds up, and that is when a tidal resonance is considered to occur. ω(Ω, n) = 2(Ω − n). (20) 2.4 Method of Solution The values of constants involved in our model are listed in Table 1. Because of the simplification of our model, whether their values are precise or not does not affect our qualitative conclusions. Numerical simulations are executed for different initial Earth-Moon distances and phases of Earth's obliquity. The Tidal Evolution Influenced by Climate Change 7 (cid:2013)(cid:4666)(cid:1872)(cid:4667) (cid:1849)(cid:4666)(cid:2030)(cid:482)(cid:1853)(cid:481)(cid:1857)(cid:481)(cid:2013)(cid:4667) (cid:1846)(cid:2929)(cid:4666)(cid:2030)(cid:4667) (cid:2030)(cid:2916) (cid:1860)(cid:149)(cid:153)(cid:4666)(cid:2030)(cid:2916)(cid:4667) (cid:592)(cid:4666)(cid:1860)(cid:149)(cid:153)(cid:4667) (cid:1838)(cid:4666)(cid:592)(cid:481)(cid:2033)(cid:481)(cid:1870)(cid:4667) (cid:1866) (cid:2007) (cid:1870) Fig. 1: Procedure for calculating model variables. The thick arrows represent numerical methods, while the thin arrows signify substitution into analytical expressions. initialization of simulations will be presented in Section 3. The following is the procedure carried out for any instant after the initial time as shown in Figure 1. In every calculation loop characterized by t, obliquity ε is the first to be obtained. The term with the biggest amplitude 0.547◦(Berger & Loutre 1991) in its trigonometric expansion is used to get ε (Eq. 7). The corresponding period is 41090 yr. Then, the insolation function W (ϕ) is derived from ε using Equation 6, with a fixed at its present value and e fixed at its approximation over the last few million years ¯e. An numerical method is applied to derive the steady-state temperature field T s(ϕ) from W (ϕ). Specifically, the differential equation of T (Eq. 1) is discretized by centered finite difference method. The latitude is discretized in intervals of ∆ϕ = π/180, while the time step is given by ∆t/2, where the propagation time ∆t = (∆ϕ2/2)(Cl/D). Stepping forward in time from the initial condition T (ϕ) = 10◦C, the iteration does not cease until the relative error of temperature is less than 10−6, which happens after no more than 3.5τw = 17 yr in our simulations. Thus, a steady-state solution is considered to be reached. Because it is instantly reached compared to the time step in integration for tidal evolution (∼ 103 yr), the steady-state solution T s(ϕ) solved with W (ϕ) given at time t is just taken as the temperature field at that instant. We note that the topography adopted is a hemispherical continent (ϕl = 0◦), and the nonlinearity in C(ϕ) does not introduce any convergence problems in our procedure. Defining Tf = −10◦C, the iceline ϕf (Eq. 5) is then quickly located by linearly interpolating the adjoined latitudes where temperatures are found to just enclose Tf. The sea water depth hsw is calculated from ϕf using Equation 9, where the maximal depth hb is set (in the first loop) to what ensures that the mean depth ¯hsw = 4 km (equal to hsw in the first loop, given ψ = ±90◦) during simulations. The oceanic natural frequency σ is calculated from hsw using Equation 10, where the ocean width is set to 8 Nan Wang and Zhi-Guo He Fig. 2: Variations of tidal frequency ω (solid red curve), Earth's rotational speed Ω (dash-dotted red curve) and lunar orbital speed n (dashed red curve) as functions of Earth-Moon distance r. The mean oceanic natural frequency ¯σ = 1.555 × 10−4 rad s−1 (horizontal solid line) is illustrated for reference. The resonance distance rres = 57.7 R⊕ where ω = σ and the synchronous distance rsyn = 86.9 R⊕ where Ω = n (left and right vertical dotted lines) are indicated. resonance. l = 4000 km so that the mean natural frequency ¯σ = 1.555 × 10−4 rad s−1 is near the present M2 After σ is obtained, a Runge-Kutta-Fehlberg method is applied to integrate the differential equation of Earth's rotational speed Ω (Eq. 13). As pointed out in Section 2.3.1, the lunar orbital parameters Ω, n and r are related by Equation 11 and 12, so n and r can be known with Ω given. In Equation 11, the total angular momentum H = 3.442 × 1034 kg m2 s−1 which is determined by substituting Ω, n and r with present values. The tidal frequency ω is then also known because of its dependence on Ω and n (Eq. 20). Therefore, within each time step of the Runge-Kutta-Fehlberg integrator, while σ is held constant, L can be obtained with r and ω by using Equation 16, 19 and 14 in turn, following the calculation of r and ω from the coinstantaneous Ω. In these equations, the static torque coefficient Z static = 0.5 for a hemispherical continent, the frictional coefficient δ is set to 0.092 in order to ensure a realistic timescale of tidal evolution based on some test simulations and the dimensionality is calculated to be L∗ = 1.998 × 1017 N m. Given Ω at t, what the Runge-Kutta-Fehlberg integrator finds is Ω at the next instant t′. Updating n, r and ω using Ω at t′ immediately follows. Updating the climate and ocean status at t′ starts in the next loop. The above procedure is repeated for every instant until the final time. 3 RESULTS 3.1 Pre-analysis of tidal evolution We first present a pre-analysis of the general trend of tidal evolution based on our model, for that helps in explaining the initialization of the numerical simulations. According to our tidal evolution model, Earth's rotational speed Ω and the lunar orbital speed n can be expressed as functions of the Earth-Moon distance r, and so can the tidal frequency ω (Sect. 2.3). If the oceanic natural frequency σ is constant, the resonance distance rres, where the tidal resonance occurs as ω ≈ σ (for a nonzero dissipation), can be then predicted without simulations. Figure 2 shows that as r increases from 10 R⊕, both Ω and n decrease, which means both one day 2π/Ω and one month 2π/n lengthens. Their difference diminishes until Ω = n, where Earth begins to synchronously rotate (one day is as long as one month) just like the Moon has been doing in reality Tidal Evolution Influenced by Climate Change 9 Table 2: Initialization and Results of Numerical simulations. Case B0 B+ B− A0 A+ A− ψ (◦) +90 −90 +90 −90 ri (R⊕) 57.43 . . . . . . rf (R⊕) 57.61 +4 × 10−6 −6 × 10−6 57.80 57.98 . . . . . . −12 × 10−6 −5 × 10−6 2π/Ωi (h) 21.44 . . . . . . 21.74 . . . . . . 2π/Ωf (h) 21.59 +3 × 10−6 −5 × 10−6 21.89 −9 × 10−6 −4 × 10−6 2π/ni (d) 25.39 . . . . . . 25.63 . . . . . . 2π/nf (d) 25.51 +3 × 10−6 −4 × 10−6 25.75 −8 × 10−6 −3 × 10−6 ωi ωf (10−4 rad s−1) (10−4 rad s−1) 1.571 . . . . . . 1.549 . . . . . . 1.560 −2 × 10−7 +4 × 10−7 1.538 +7 × 10−7 +2 × 10−7 Notes: For Cases B0 and A0, ψ is not needed, because ε is fixed at ¯ε. Initial values substituted by dots are the same as above. Final values for Cases B+ and B− (A+ and A−) are presented as the divergences from those for Case B0 (A0). (though Moon's rotation is neglected in our model) and the tidal evolution ends. According to the con- stants we set, that happens at rsyn = 86.9 R⊕. In addition, as r increases, ω also keeps decreasing until ω = 0, when the resulting tidal torque and dissipation become zero. If the oceanic frequency is fixed at ¯σ = 1.555 × 10−4 rad s−1, the tidal resonance is found to occur at rres ≈ 57.7 R⊕, slightly smaller than the present Earth-Moon distance rp = 60.3 R⊕. This prediction is in accordance with the reality that the oceanic response has been near M2 resonance currently. Near rres, the tidal evolution should greatly speed up with dissipation of the total energy enhanced largely. The rapid decrease in ω at that time gives rise to the rapid pass through resonance. However, if σ is varying, conditions become complicated. Before rres, a decreasing σ delays the resonance while an increasing σ hastens it; after rres, the former extends it while the latter curtails it. Even multiple passes can arise. Those conditions are beyond the scope of the present work. Therefore, although we focus on the near resonance condition, we will simulate periods a while before and after rres, instead of simulating the maximum period. 3.2 Numerical simulations The initialization for our numerical simulations is shown in Table 2. Two sets of simulations, Cases B and Cases A ("before" and "after" resonance maximum respectively), are performed with initial Earth- Moon distance ri = 57.43 and 57.80 R⊕, respectively, which are slightly smaller and larger than the resonance distance rres ≈ 57.7 R⊕ (Sect. 3.1). The initial values of the other two orbital parameters Ωi and ni are derived from ri (Sect. 2.3.1). The corresponding initial tidal frequency ωi = 1.571 × 10−4 and 1.549× 10−4 rad s−1 for Cases B and A, respectively, just enclosing ¯σ = 1.555× 10−4 rad s−1, i.e., the oceanic frequency in the case of invariant ε. Each set includes three cases: Case 0 acts as a control group with constant obliquity ¯ε = 23.33◦; for Cases + and −, ε periodically varies with ψ = +90◦ and −90◦, respectively. The influence of the existence of climate change can thus be examined. All the other parameters are set as described in Section 2.4. We note that our aim in this work is to study the mechanism of climate influence but not the realistic history of lunar orbit. Therefore, although not comparable to the timescale of tidal evolution 109 yr, the simulation time 106 yr fits this aim, and it is not a problem that the cases with varying ε are initialized at the same ri as the case with constant ε. Furthermore, it is therefore reasonable to randomly set the phase of obliquity ψ at any r. We choose +90 and −90◦ in order to maximize the difference between Cases + and −. Cases B start and end before the resonance maximum (Table 2). As shown in Figure 3a, ε increases and decreases from ¯ε at the beginning for Case B+ and Case B−, respectively, and periodically varies till the end. According to Loutre et al. (2004), the mean annual insolation W , being symmetrical between northern and southern hemispheres, varies in phase with ε in the high latitudes but exactly out of phase in the low latitudes. These properties are completely exhibited in Figure 3c and 3d. The steady-state temperature T s, as a response to insolation, has the same properties as W . As shown in Figure 3e and 3f, located in the high latitudes, the iceline ϕf varies in phase with ε for both Cases B+ and B−. Its 10 Nan Wang and Zhi-Guo He Fig. 3: Temporal variations of Earth's obliquity ε (a), sea water depth hsw (b), mean annual insolation function W (ϕ) (c and d) and steady-state temperature field T s(ϕ) (e and f) for both Cases B and A. In panels a and b, the black lines, red curves and blue curves indicate Case 0, + and −, respectively. In panels c and e, contours are red to indicate Case +, while in d and f, those are blue to indicate Case −. Tidal Evolution Influenced by Climate Change 11 Fig. 4: Temporal variations of oceanic natural frequency σ, tidal forcing frequency ω (solid and dashed curves in panel a), and divergence in Earth-Moon distance ∆r (b) for Cases B. In both panels, the colors black, red and blue indicate Cases B0, B+ and B−, respectively. (In panel a, ω for Cases B+ and B− are indistinguishable from B0) Fig. 5: Temporal variations of oceanic natural frequency σ, tidal forcing frequency ω (solid and dashed curves in panel a), and divergence in Earth-Moon distance ∆r (b) for Cases A. In both panels, the colors black, red and blue indicate Cases A0, A+ and A−, respectively. (In panel a, ω for Cases A+ and A− are indistinguishable from A0) mean ¯ϕf = 65.44◦ and the amplitude defined as the maximal deviation from the mean is 0.23◦. Because the sea water depth hsw ∝ sin ϕf (Eq. 9) and the oceanic frequency σ ∝ √hsw (Eq. 10), they are in phase with ε as well. For either Case B+ or B−, hsw oscillates about ¯hsw with an amplitude of 2.98 m (Fig. 3b), and σ oscillates about ¯σ with an amplitude of 0.001 × 10−4 rad s−1 (Fig. 4a). The general trend of tidal evolution with constant σ has been interpreted in Section 3.1. We now list the initial and final values of the lunar orbital parameters in Table 2 and illustrate only the divergence of Earth-Moon distance ∆r for Cases B+ and B− from Case B0 in Figure 4b in order to focus on the climate influence. Three features are commonly observed for Cases B+ and B−. First, ∆r varies in phase with σ and thus in phase with ε. The reason is that during the pre-resonance time when ω > σ, greater σ means greater tidal dissipation, resulting in a leading evolution characterized by a larger r. 12 Nan Wang and Zhi-Guo He Second, the equilibrium point of the ∆r oscillation is not constant but seems to decrease at least on the near side of resonance maximum. Third, the displacement of ∆r from the gradually decreasing equilibrium point expresses a positive correlation with the difference between ω and σ. One distinction between Cases B+ and B− is that the mean of ∆r for the former is larger than the latter. We attribute this distinction to ψ and the beginning behavior of ∆r it determines: the beginning increase/decrease in ∆r for Case B+/B− contributes to a lead/drop lasting for the whole simulation time. Cases A start later than the resonance maximum (Table 2). The evolutions of climate and ocean state (ε, W , T s, ϕf , hsw and σ) for Cases A are totally the same as those for Cases B. However, because ω < σ during the post-resonance time (Fig. 5a), contrary to Cases B, greater σ means smaller dissipation and thus a smaller r. As shown in Figure 5b, ∆r for either Case A+ or A− is therefore exactly out of phase with σ. The second and third features for Cases B, i.e., the general trend of decreasing and positive correlation between ∆r displacement and the difference between ω and σ, also match Cases A. In addition, ψ again results in a larger mean of ∆r for the case where ∆r increases at the beginning (A−) than where it decreases (A+). It is worth mentioning that both as the case with the larger mean ∆r, B+ holds a positive ∆r for the whole time, whereas A− holds for only about 105 yr. Considering the general decreasing trend, the fact that B+ holds a lead over B0 is probably a temporary effect. In summary, the features of our climate-influenced tidal evolution (characterized by ∆r whose positive value means a lead over the evolution with climate unchanged and negative value means a lag) are 1. Given that iceline ϕf is in high latitudes (so that σ is in phase with ε), ∆r varies in phase with ε during the pre-resonance time but exactly out of phase during the post-resonance time. 2. Despite oscillation, the general trend of ∆r is decreasing. 3. The displacement of ∆r oscillation is in positive correlation with the difference between ω and σ. 4. As a whole, ∆r oscillation is shifted upwards or downwards as ∆r increases or decreases at the beginning. 4 DISCUSSION Based on our conceptual coupled model of climate and tidal evolution (Sect. 2), we carried out numer- ical simulations of the near-resonance tidal evolution for an equatorial circular lunar orbit with Earth's obliquity ε periodically varying (Sect. 3.2). Thus, the climate influence on the tidal evolution via ocean is verified. Our conclusions in terms of the influence mechanism are qualitative. The main conclusion is that compared to the case that the climate is invariant, varying climate slows down the evolution accompanied by oscillations. Furthermore, the oscillation is in phase and exactly out of phase with ε before and after the resonance maximum, respectively; and can be enlarged as the difference between the oceanic frequency σ and the tidal frequency ω increases. The above conclusions should be applied with caution. This is not only because of the idealization and the existence of multiple parameters of the model, but also because the simulations are only done for a short instant near the resonance maximum of the whole lunar tidal evolution. Though we focus on the mechanism of climate influence in this work, it should be pointed out that the absolute differences in final orbital parameters found between the cases with varying and invariant climates (Table 2) are insignificant indeed. However, it is still hasty to conclude that the influence of climate change can be neglected. On one hand, the timescale studied here, 106 yr, is very short compared with the timescale of tidal evolution, 109 yr. If the evolution keeps slowing down with climate varying, the secular accumulation may make a difference. On the other hand, the variations of the climate and ocean state produced here are not as big as in reality. The maximal drop of the sea water depth in simulations is 6 m, whereas the sea level drop during the last glacial maximum relative to the present is about 130 m (Clark et al. 2009). If a more realistic model is used, the influence will be enhanced. One important effect that can enhance the climate influence is the "ice-albedo feedback." In the current model, though the ice sheet on the continent is considered, the coalbedo α has no dependence on iceline ϕf . A more realistic way, for example, is to multiply α by 1/2 in latitudes higher than ϕf Tidal Evolution Influenced by Climate Change 13 (Mengel et al. 1988). In this case, as the ice cover spreads, the planetary coalbedo and thus the absorbed solar radiation diminishes, leading to a further drop in temperature accompanied by the spread of ice cover. In other words, a slight change in solar radiation can cause an abrupt climate transition (North 1984). This nonlinear feedback mechanism will be introduced in our future model. Potential subjects of future works include improving the model, determining the quantitative cor- relation between climate variation and the rate of tidal evolution, and generalizing the model to other planet-satellite systems. In addition, considering that it is the normal modes of the liquid part of the Earth that can largely be excited when tidal resonance occurs, the tidal evolutions of terrestrial planets per- turbed by companions in exosolar systems (e.g., Dong & Ji 2013) may also need further investigations when oceans or liquid cores (Liu & Li 2018) are present. Acknowledgements We thank the referee for constructive comments. We are grateful that Professor Kwang-Yul Kim helped in climate simulations. The code of tidal evolution was developed in Nanjing University when the author Nan Wang was under the supervision of Professor Ji-Lin Zhou, and fur- ther improved in Zhejiang University. This work was funded by the Natural Science Foundation of Zhejiang Province (LR16E090001), the National Key Research and Development Program of China (Grant No. 2017YFC0305905), and NSFC-Zhejiang Joint Fund for the Integration of Industrialization and Informatization (Grant No. U1709204). References Berger, A. 1988, Reviews of Geophysics, 26, 624 Berger, A. 2012, A Brief History of the Astronomical Theories of Paleoclimates, ed. A. Berger, F. Mesinger, & D. Sijacki, Climate Change: Inferences from Paleoclimate and Regional Aspects, ed. A. Berger, F. Mesinger, & D. Sijacki (Vienna: Springer Vienna), 107 Berger, A. L. 1978, Journal of the Atmospheric Sciences, 35, 2362 Berger, A., & Loutre, M. F. 1991, Quaternary Science Reviews, 10, 297 Berger, A., & Loutre, M. F. 1992, Earth and Planetary Science Letters, 111, 369 Berger, A., Loutre, M.-F., & Yin, Q. 2010, Quaternary Science Reviews, 29, 1968 Bills, B. G., & Ray, R. D. 1999, Geophysical Research Letters, 26, 3045 Budyko, M. I. 1969, Tellus, 21, 611 Clark, P. U., Dyke, A. S., Shakun, J. D., et al. 2009, Science, 325, 710 Dong, Y., & Ji, J. 2013, Monthly Notices of the Royal Astronomical Society, 430, 951 Egbert, G. D., Ray, R. D., & Bills, B. G. 2004, Journal of Geophysical Research: Oceans, 109, C03003 Goldreich, P. 1966, Reviews of Geophysics, 4, 411 Green, J. A. M., Green, C. L., Bigg, G. R., et al. 2009, Geophysical Research Letters, 36, 5 Griffiths, S. D., & Peltier, W. R. 2009, Journal of Climate, 22, 2905 Halliday, A. N. 2008, Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 366, 4163 Hansen, K. S. 1982, Reviews of Geophysics, 20, 457 Haynes, W. M., ed. 2017, CRC Handbook of Chemistry and Physics, 97th edn. (Boca Raton, FL: CRC Press) Huybers, P., & Tziperman, E. 2008, Paleoceanography, 23, PA1208 Kagan, B. A., & Maslova, N. B. 1994, Earth, Moon, and Planets, 66, 173 Lambeck, K., Rouby, H., Purcell, A., Sun, Y., & Sambridge, M. 2014, Proceedings of the National Academy of Sciences, 111, 15296 Laskar, J. 1988, Astronomy and Astrophysics, 198, 341 Lin, R. Q., & North, G. R. 1990, Climate Dynamics, 4, 253 Liu, M., & Li, L.-G. 2018, Research in Astronomy and Astrophysics, 18, 23 Loutre, M.-F., Paillard, D., Vimeux, F., & Cortijo, E. 2004, Earth and Planetary Science Letters, 221, 1 McGehee, R., & Lehman, C. 2012, SIAM Journal on Applied Dynamical Systems, 11, 684 Mengel, J. G., Short, D. A., & North, G. R. 1988, Climate Dynamics, 2, 127 14 Nan Wang and Zhi-Guo He Munk, W. 1968, Quarterly Journal of the Royal Astronomical Society, 9, 352 Munk, W. 1997, Progress in Oceanography, 40, 7 Munk, W., & Wunsch, C. 1998, Deep Sea Research Part I: Oceanographic Research Papers, 45, 1977 Murray, C. D., & Dermott, S. F. 1999, Solar system dynamics (Cambridge university press) North, G. R. 1984, Journal of the Atmospheric Sciences, 41, 3390 North, G. R., & Jr., J. A. C. 1979, Journal of the Atmospheric Sciences, 36, 1189 North, G. R., & Kim, K.-Y. 2017, Energy Balance Climate Models, Wiley Series in Atmospheric Physics and Remote Sensing (Wiley-VCH Verlag GmbH & Co. KGaA) Sellers, W. D. 1969, Journal of Applied Meteorology, 8, 392 Thomas, M., & Sundermann, J. 1999, Journal of Geophysical Research: Oceans, 104, 3159 Touma, J., & Wisdom, J. 1994, The Astronomical Journal, 108, 1943 Wagner, T. J. W., & Eisenman, I. 2015, Journal of Climate, 28, 3998 Webb, D. J. 1980, Geophysical Journal International, 61, 573 Webb, D. J. 1982a, Geophysical Journal International, 68, 689 Webb, D. J. 1982b, Geophysical Journal International, 70, 261 Yokoyama, Y., Lambeck, K., De Deckker, P., Johnston, P., & Fifield, L. K. 2000, Nature, 406, 713
1908.03468	2	1908	2019-10-17T20:48:25	On the accuracy of symplectic integrators for secularly evolving planetary systems	[ "astro-ph.EP", "astro-ph.IM", "math.DS" ]	Symplectic integrators have made it possible to study the long-term evolution of planetary systems with direct N-body simulations. In this paper we reassess the accuracy of such simulations by running a convergence test on 20Myr integrations of the Solar System using various symplectic integrators. We find that the specific choice of metric for determining a simulation's accuracy is important. Only looking at metrics related to integrals of motions such as the energy error can overestimate the accuracy of a method. As one specific example, we show that symplectic correctors do not improve the accuracy of secular frequencies compared to the standard Wisdom-Holman method without symplectic correctors, despite the fact that the energy error is three orders of magnitudes smaller. We present a framework to trace the origin of this apparent paradox to one term in the shadow Hamiltonian. Specifically, we find a term that leads to negligible contributions to the energy error but introduces non-oscillatory errors that result in artificial periastron precession. This term is the dominant error when determining secular frequencies of the system. We show that higher order symplectic methods such as the Wisdom-Holman method with a modified kernel or the SABAC family of integrators perform significantly better in secularly evolving systems because they remove this specific term.	astro-ph.EP	astro-ph	Mon. Not. R. Astron. Soc. 000, 000 -- 000 (0000) Printed 21 October 2019 (MN LATEX style file v2.2) On the accuracy of symplectic integrators for secularly evolving planetary systems Hanno Rein1,2,3, Garett Brown1,3, Daniel Tamayo4(cid:63) 1 Department of Physical and Environmental Sciences, University of Toronto at Scarborough, Toronto, Ontario M1C 1A4, Canada 2 Department of Astronomy and Astrophysics, University of Toronto, Toronto, Ontario, M5S 3H4, Canada 3 Department of Physics, University of Toronto, Toronto, Ontario, M5S 3H4, Canada, 4 Department of Astrophysical Sciences, Princeton University, Princeton, New Jersey 08544, United States Accepted 2019 October 16. Received 2019 October 16; in original form 2019 August 9. ABSTRACT Symplectic integrators have made it possible to study the long-term evolution of plan- etary systems with direct N -body simulations. In this paper we reassess the accuracy of such simulations by running a convergence test on 20 Myr integrations of the Solar System using various symplectic integrators. We find that the specific choice of metric for determining a simulation's accuracy is important. Only looking at metrics related to integrals of motions such as the energy error can overestimate the accuracy of a method. As one specific example, we show that symplectic correctors do not improve the accuracy of secular frequencies compared to the standard Wisdom-Holman method without symplectic correctors, despite the fact that the energy error is three orders of magnitudes smaller. We present a framework to trace the origin of this apparent para- dox to one term in the shadow Hamiltonian. Specifically, we find a term that leads to negligible contributions to the energy error but introduces non-oscillatory errors that result in artificial periastron precession. This term is the dominant error when determining secular frequencies of the system. We show that higher order symplectic methods such as the Wisdom-Holman method with a modified kernel or the SABAC family of integrators perform significantly better in secularly evolving systems because they remove this specific term. Key words: methods: numerical -- gravitation -- planets and satellites: dynamical evolution and stability 1 INTRODUCTION The range of timescales present in planetary systems is truly astronomical, from a few days to billions of years. Special- ized integrators are therefore needed to accurately calculate trajectories over the entire lifetime of planetary systems. In recent years, the development of such accurate numerical methods and the increase in computing power have made very long-term integrations possible (Laskar & Gastineau 2009). The mixed variable symplectic Wisdom & Holman (1991) integrator (WH) and higher order variants thereof are one of the preferred tools for this purpose (Wisdom et al. 1996; Laskar & Robutel 2001). Recently, Rein et al. (2019b) implemented many of these higher order symplectic integra- (cid:63) NHFP Sagan Fellow c(cid:13) 0000 RAS tors in the freely available REBOUND package. This present paper uses several of these methods to perform a conver- gence study for integrations of the Solar System. Typically authors report the performance of a method by monitoring how well it can conserve energy, as energy should be con- served exactly in a conservative system (Laskar & Robutel 2001; Wisdom 2018; Rein & Tamayo 2015). Although the energy error is a metric involving all planets, it is domi- nated by only a few. We show that some of these methods, despite reducing the energy error by three orders of magni- tude, do not improve the accuracy of the apsidal precession frequencies that drive the solar system's long-term evolu- tion. In this paper, we present a framework which lets us resolve this apparent paradox. We introduce the notation and describe the numeri- cal setup and algorithms in Sect. 2. This includes a short overview of the symplectic integrators that we use, as well 2 Rein et al. as an outline of the modified Fourier transform that we use to determine the secular frequencies of the system. The re- sults of our convergence study of long-term integrations are presented in Sect. 3. We then develop a framework which lets us understand the origin of errors in symplectic meth- ods in Sec. 4. We conclude with a summary and discussion of the implications in Sec. 5. 2 METHODS We use the freely available N -body package REBOUND (Rein & Liu 2012) to perform simulations of the Solar System with all 8 planets. We model the planets as point masses and treat the Earth-Moon system as one particle in our simulation. The initial conditions are taken from the NASA Horizons database, provided by the Solar System Dynamics Group of the Jet Propulsion Laboratory1. The effects of general rel- ativity are approximated by including a 1/r3 term in the potential (Nobili & Roxburgh 1986) using the implementa- tion provided by REBOUNDx (Tamayo et al., in prep2). We integrate the simulations for 20 million years. Al- though the direction in which we integrate does not mat- ter, we here integrate backwards so that we can more eas- ily compare the secular frequencies to those obtained by Laskar et al. (2011). We use the Simulation Archive (Rein & Tamayo 2017) to record 22000 snapshots throughout each integration. Since our goal is to compare simulations with different timesteps to very high precision, we have to ensure that we can record the snapshots at exactly the same times for all simulations without the need for any interpolation or change of the timestep during the integration. We achieve this by choosing the timesteps such that all snapshots oc- cur at integer multiples of the timestep (the same procedure was used by Laskar et al. 2011). Furthermore, we count the number of timesteps taken rather than keeping track of the integration time itself. This ensures that the round-off er- rors originating in adding small numbers (the timestep) to a large number (the current simulation time) do not affect the analysis. We use the IAS15 integrator, a high order Gauss-Radau integrator (Rein & Spiegel 2015), to obtain a true solution to which we can then compare all other simulations. It is im- portant to point out that we only consider it to be the true solution of our model, but not the true solution of the real Solar System. All we need for our convergence study is a well defined model that contains the most important dynamical effects determining the evolution of the Solar System. Other simulations, in particular those by Laskar et al. (2011), are better representations of the real evolution of the Solar Sys- tem because they use more accurate initial conditions and include important physical effects such as tidal evolution, stellar mass loss, and perturbations from minor planets. We come back to why we have confidence that the IAS15 inte- grator indeed provides the true solution to our model later, and also comment on it in Appendix A. 1 https://ssd.jpl.nasa.gov 2 Code and documentation are already available at https:// reboundx.readthedocs.io We run simulations with the exact same initial con- ditions using the IAS15 integrator, the standard WH inte- grator, the WH integrator with first symplectic correctors of order 17, WHC (Wisdom et al. 1996), the WH integra- tor with the lazy implementer's kernel method, WHCKL (Wisdom et al. 1996), and the SABA integrator with four function evaluations and correctors using the lazy imple- menter's method, SABACL4 (Laskar & Robutel 2001; Rein et al. 2019b). Next, we introduce the notation that we use for operator splitting methods and then summarize the main features of the various integrators mentioned. The differences between the various symplectic integrators and specifically their im- plementation in REBOUND are also described in more detail in Rein et al. (2019b). 2.1 Operator Splitting Schemes Let us begin by defining the Poisson bracket {g, h} of two functions g and h of the canonical coordinates (qi, pi) as {g, h} ≡(cid:88) i (cid:18) ∂g ∂qi (cid:19) , ∂h ∂pi − ∂g ∂pi ∂h ∂qi (1) (2) which allows us to write down Hamilton's equations as = {pi, H}. = {qi, H} qi = and pi = − ∂H ∂qi ∂H ∂pi The time derivative of any function g that depends only on p and q can then be written succinctly as g = {g, H}. (3) Let us further introduce LH , an operator which can act on any phase space function g. We call this operator the Lie derivative with respect to the Hamiltonian H and define it as LH g ≡ {g, H}. Assume that the dynamical system we are interested in has 3N coordinates and 3N momenta. The Lie derivative allows us to write Hamilton's equation in a compact way as y = LH y (5) where y(t) ≡ (q1(t), . . . , q3N (t), p1(t), . . . , p3N (t)) are the canonical coordinates and momenta. (4) t Lastly, let us define the formal solution operator ϕ[H] (y0). This operator returns the solution to the differ- ential equation y = LH y at time t with initial conditions y0 given at t = 0. Because we do in general not know how to write down this abstract solution operator, we use an operator split- ting method to approximate it. The idea behind an opera- tor splitting method is to rewrite the differential equation in Eq. 5, to which we do not know the solution, as y = LA y + LB y. By virtue of the linearity of the Lie derivative, it follows that this equation is identical to Eq. 5 if the have H = A+B. This idea of expressing H as a sum of A and B is often referred to as splitting the Hamiltonian. Let us refer to the evolution operators corresponding to the differential equations y = (6) c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 LAy and y = LBy as ϕ[A] respectively. Then, we can write down a second order (leap-frog) operator splitting scheme, Ψt(y0) = ϕ[A] and ϕ[B] ◦ ϕ[A] (7) t/2 ◦ ϕ[B] t t/2(y0). t t As we will see below, this is a second order method which approximates the evolution operator ϕ[H] . Note that in the above notation, the ϕs are operators and we apply the right most operator first. t Next, we use a formal trick to express any solution op- erator ϕ[X] in terms of its corresponding Lie derivative op- erator LX and an exponential (defined as an infinite series in the usual way), for example, ϕ[H] t (y0) = exp (tLH ) Id(y0) = exp (t (LA + LB)) Id(y0). (8) To see that this is indeed a solution to the differential equa- tion in Eq. 5, note that the series expansion of the expo- nential provides a series of differential operators acting on Id(y0). The series is simply the Taylor expansion of the so- lution ϕ[H] around t = 0. We can use this formalism to also express our splitting method from Eq. 7 in terms of expo- nentials of Lie derivative operators, i.e. Ψt(y0) = exp LA exp tLB exp LA Id(y0). (9) (cid:19) (cid:16) (cid:17) (cid:19) (cid:18) t 2 (cid:18) t 2 (cid:16) This allows us to use the symmetric Baker-Campbell- Hausdorff (BCH) identity to combine the three exponentials of non-commuting operators into one exponential: Ψt(y0) = exp − t3 12 [LA, [LA,LB]] tLA + tLB − t3 24 [LB, [LA,LB]] + O(t5) (cid:17) (10) where the commutator [LA,LB] is short hand for LALB − LBLA. Id(y0), If we compare Eq. 8 with Eq. 10, we can see that the differential equations which our splitting scheme is solving are not those originating from the Hamiltonian H. Although the first two terms are equal, Eq. 10 also contains an infinite series of commutators of Lie derivatives. Note that by virtue of the Jacobi identity, we have (11) son brackets. For example, the two terms at order O(cid:0)t3(cid:1) [LA,LB] g(y) = L{A,B} g(y). This allows us to identify the commutators of Lie derivatives in Eq. 10 as Poisson brackets of the Hamiltonians A and B. We can then construct a shadow Hamiltonian H which is the original Hamiltonian H plus an infinite series of Pois- in Eq. 10 correspond to the Poisson brackets t2{A,{A, B}} and t2{B,{A, B}} in the shadow Hamiltonian. Note that all of these terms contain a factor of tn with n (cid:62) 2. In the limit of t → 0 all error terms are therefore con- verging3 to 0. In practice we will choose a timestep smaller than the shortest dynamical timescale in the problem. We indicate that we take a small but finite timestep by replacing t with dt from now on. 3 There is an issue of whether the series actually converges (see Wisdom 2018). This is not important for the discussion in this paper. c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 Integrators for secularly evolving systems 3 2.2 Wisdom-Holman Mapping The Wisdom-Holman (WH) integrator (Wisdom & Holman 1991) is using the scheme given in Eq. 7 together with a spe- cific splitting of the Hamiltonian H. In the WH integrator, part A describes the evolution of planets on Keplerian orbits around the Sun and part B describes the much smaller in- terplanetary forces. The exact solutions ϕ[A] and ϕ[B] of the equations of motions associated with A and B, y = LAy and y = LBy, can be calculated efficiently. ϕ[A] is referred to as the Kepler step and involves solving Kepler's equation. This can be done using a series expansion which can be truncated once machine precision is reached. ϕ[B] is referred to as the kick or interaction step. It updates the velocities of planets by evaluating the planet-planet interaction potential. Let us estimate the size of the additional terms in the shadow Hamiltonian of the WH method. Because the WH integrator splits the Hamiltonian in a dominant Keplerian part A and a perturbation part B, we can introduce a di- mensionless parameter to keep track of this separation of scales by formally replacing B with B, where (cid:28) 1. In the Solar System, the gravitational forces from other plan- ets are roughly a factor of 10−3 smaller than the force from the Sun, thus in this particular case ≈ 10−3. Using this substitution, the two terms in the shadow Hamiltonian in- volving {A,{A, B}} and {B,{A, B}} are of order O(cid:0)dt2(cid:1) and O(cid:0)2dt2(cid:1), respectively. Because a factor of is present in all error terms, the WH integrator performs well in weakly perturbed systems where is small. 2.3 Wisdom-Holman Mapping with Symplectic Correctors The standard WH method is a second order method in dt. Symplectic methods can also be constructed to higher order in the timestep (Yoshida 1990). To improve the accuracy for a fixed dt, one can thus simply choose a higher order method to remove the dominant error terms in Eq. 10. However, this requires more force evaluations per timestep and is therefore slower. Symplectic correctors were introduced by Wisdom et al. (1996) as an alternative to improve the accuracy with effectively no overhead. They can be understood as a clever transformation between the original problem and integra- tor variables which removes terms of order O() from the shadow Hamiltonian. Because the transformations between real and mapping variables only need to be performed when outputs are needed (not every timestep), they incur negligi- ble computational cost. For example a third-order corrector removes the {A,{A, B}} term which is of order O(dt2). A fifth-order corrector removes all the terms up to and includ- ing those of order O(dt4), and so forth. In this paper, we use the symplectic correctors of order 17 and refer to the method as WHC. The now dominant term in the shadow Hamiltonian for small timesteps is {B,{A, B}}, which is of order O(cid:0)2dt2(cid:1). 4 Rein et al. 2.4 Wisdom-Holman Mapping with a modified Kernel orbital parameters. We use the S matrix published by Laskar (1990) to do the transformation to proper modes, If we wish to further improve the WHC method, we have to remove the now leading order term, {B,{A, B}} ∼ O(2dt2). It goes beyond the scope of this paper to describe in de- tail how Wisdom et al. (1996) achieve this using a modi- fied kernel. What is important for our analysis is that this new method, WHCKL, has now a leading order term of O(2dt4) by getting rid of the {B,{A, B}}} term. As sum- marized in Rein et al. (2019b), there are different ways the kernel can be implemented. In this paper, we use the lazy implementer's method. In comparison to the modi- fied kick method WHCKM, it allows for the inclusion of non-Newtonian forces. And in comparison to the composi- tion method WHCKC, it is significantly faster (Rein et al. 2019b). 2.5 SABA integrator family We also test the symplectic integrators of the SABA family (Laskar & Robutel 2001). Specifically, we use the SABACL4 method which uses four force evaluations and one correc- tor per timestep (Rein et al. 2019b). The correctors in SABACL4 are responsible for removing the {B,{A, B}} term in the shadow Hamiltonian. Note that although both Wisdom et al. (1996) and Laskar & Robutel (2001) use the word corrector to describe parts of their integrators, the derivation, interpretation, and implementation of the cor- rectors differ significantly. The SABACL4 method has a leading order error term of O(dt6 + 2dt4). For small timesteps the O(2dt4) term dominates. Most importantly for our later discussion, this method got rid of the {B,{A, B}} error term in the shadow Hamiltonian. 2.6 Frequency Analysis We follow the procedure of Laskar (1988, 1990, 1993, 2003) to determine the secular frequencies of the Solar System us- ing a modified Fourier transform. Specifically we will com- pute the g-modes of the Solar System, which are related to the periastron and eccentricity of the planets. A full descrip- tion of the algorithm is given in the references above (see also Sidlichovsk´y & Nesvorn´y 1996). Here, we only present a short outline of the algorithm. We start by calculating the orbital elements in Jacobi coordinates for all planets at each of our snapshots. For simulations that use a symplectic corrector, we need to ap- ply the correctors beforehand. This is done by calling the synchronize routine in REBOUND. We then compute the com- plex eccentricities, zi(t) = ei(t) exp (ii(t)) i = 1, 2, . . . , 8, (12) where ei(t) and i(t) are the eccentricity and the longitude of periastron of the i-th planet at time t. Next, we perform a linear transformation to a basis that approximately corresponds to the proper modes. Although not absolutely necessary, this transformation makes it easier to identify the dominant secular frequencies in each planet's u(t) = S −1z(t). (13) Since the Solar System's secular frequencies are changing very slowly over time, this transformation should be thought of as an approximation. For our purpose, the precision of this approximation is not relevant as long as we use the same transformation for the entire convergence study. On each proper mode ui(t), we then perform a Fourier transform and find the frequency ωi,fft with the maximum power. This frequency can only be determined to within one Nyquist frequency using a standard Fourier transform. We would need an unreasonably large number of samples (1010) to achieve the accuracy we need for our convergence study (10−10). To get around this, we search in the neighbour- hood of ωi,fft for the frequency ωi,max which maximizes the following scalar product (cid:104)eiωi,max t, ui(t)(cid:105) = eiωi,max t ui(t) h(t)dt, (cid:90) (14) where h(t) is the Hanning window function. The integral is evaluated at the discrete sampling points using the trape- zoidal rule. We refer to the frequency ωi,max as the secular frequency gi. Whereas one can continue with an orthogonal- ization procedure to extract other frequencies in each proper mode, we here only determine one frequency per mode. 3 LONG TERM INTEGRATIONS OF THE SOLAR SYSTEM In this section we present the results of our 20 Myr integra- tions of the Solar System. With these results as a motivation, we then develop a formalism to explain them in Sec. 4. 3.1 Comparing our secular frequencies to previous work As a consistency check, we start by comparing the secular frequencies obtained with our IAS15 simulation to previous work of Laskar et al. (2011) and Spalding et al. (2018). The results are listed in Table 1. We can reproduce the frequen- cies obtained by Laskar et al. (2011) with high precision, with the largest discrepancy of about 0.5% being in g2, the frequency associated with Venus4. The frequencies of the outer planets are all reproduced to four decimal digits or better. We do not expect to reproduce the frequencies of Laskar et al. (2011) exactly as these authors use a different set of ini- tial conditions, treat the Moon as a separate object, include a model of tidal dissipation as well as the effects of some minor planets. Whereas the treatment of these additional effects are important to accurately model the long-term evo- lution of the Solar System, the high level of agreement we find even without including these effects gives us confidence 4 Given the low eccentricity of Venus, it makes sense that this mode is particularly sensitive to changes in the initial conditions. c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 Integrators for secularly evolving systems 5 Table 1. Secular frequencies obtained by our integrations and other authors. All frequencies are given in units of arc-seconds per year. The frequencies obtained in this work have been rounded to the same number of digits as in Laskar et al. (2011) for easier comparison. This work Laskar et al. (2011), La2010a Spalding et al. (2018) g1 5.59 5.59 5.71 g2 g3 g4 g5 g6 g7 g8 7.422 7.453 7.44 17.360 17.368 17.19 17.914 17.916 17.76 4.257519 4.257482 4.26 28.2457 28.2449 28.25 3.088043 3.087946 3.09 0.673301 0.673019 0.67 ergy error in simulations using the WH integrator converges following a dt2 power-law until it reaches machine precision at dt ≈ 1 day. We overplot the expected floor originating from IEEE754 double floating point round-off errors, Efloor ≈ 2 −53(cid:112)Nstep, (15) where Nstep is the number of timesteps required for a 20 mil- lion year integration (Brouwer 1937; Quinn & Tremaine 1990). We cannot expect to achieve a better accuracy than that because floating point numbers on a computer are rep- resented using a finite number of digits (about 16 in decimal, 53 in binary). The above estimate assumes that one round- ing error of ∼ 10−16 occurs after every timestep, ignoring other operations during the timestep. Clearly, every inte- grator has many operations per timestep but this provides a lower limit. The energy error in simulations using the WHC and WHCKL integrators converge much faster and reach ma- chine precision at dt ≈ 8 days (the two curves for WHC and WHCKL are on top of each other). Similarly, the energy er- ror in the SABACL4 simulations reaches machine precision at dt ≈ 20 days. Note that the SABACL4 integrator uses four sub-timesteps and therefore four force evaluations per timestep (plus one corrector step also requiring a force eval- uation). If we were to take this into account and plot the error as a function of the effective timestep, corresponding to one force evaluation per timestep, then the SABACL4 curve would shift to the left. From this plot, we might conclude that the WHC, WHCKL, and SABACL4 integrators are thus very similar in performance and several orders of magnitude better than standard WH for intermediate timesteps. As a reference, we also show in Fig. 1 the relative en- ergy error in the simulation using the IAS15 integrator. Since IAS15 is an adaptive scheme, we plot the average timestep on the horizontal axis. Note that IAS15 uses compensated sum- mation for some crucial parts of the calculation and there- fore achieves an accuracy better than the IEEE754 floor. The fact that the energy error in the IAS15 simulation is significantly smaller than for the other integrators gives us confidence that we can consider it the true solution of our model for all practical purposes (see also Appendix A). Finally, note that the energy error is really not a par- ticularly good metric for the convergence study in the first place. There is only about one decade in the timestep, from dt = 10 days to dt = 1 day, that we can effectively use. For timesteps larger than 10 days, the methods are not well converged because we barely resolve the shortest timescale in the problem, Mercury's orbital period of 88 days. For timesteps smaller than 1 day, all methods show an energy error which is dominated by round-off error, suggesting that Figure 1. Relative energy error in 20 Myr integrations of the Solar System using different integrators. that our simulations are an accurate model which captures the most important dynamics in the Solar System. A further difference between the studies is that the fre- quencies obtained in this work come from a 20 Myr inte- gration; the work of Laskar et al. (2011) use a 20 Myr in- tegration for g1 through g4 and a 50 Myr integration for g5 through g8; and the work of Spalding et al. (2018) use a 450 Myr integration for all their frequencies. We note that although Spalding et al. (2018) also uses the REBOUND integrator package and initial conditions from NASA Horizons, we were not able to reproduce their fre- quencies as accurately as those of Laskar et al. (2011). Pos- sible explanations for this discrepancy might be their treat- ment of general relativity or a difference in their data reduc- tion. Given that a precise integration was not necessary for the conclusions of Spalding et al. (2018) and that we agree with Laskar et al. (2011) to such high precision, we have confidence in this being the right solution. 3.2 Convergence study Next, we present the results of our convergence study of the WH, WHC, WHCKL, and SABACL4 integrators in our 20 Myr integrations of the Solar System. For that, we vary the timestep from 0.05 days, to 100 days and measure var- ious different error metrics for each simulation. For small enough timesteps, all integrators should converge to the true solution. Fig. 1 shows the maximum relative energy error in our simulations. Since we are integrating a conservative system the energy should be conserved exactly by an ideal integrator and we can therefore measure the energy error directly with- out the need to know the exact solution. This is why many studies only look at the energy error. We can see that the en- c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 101100101102effective timestep [days]101410121010108106104relative energy errorIAS15IEEE754 floor, dt1/2WHWHCWHCKLSABA4CL6 Rein et al. Figure 2. Relative semi-major axis error, periastron error, and secular frequency error for all planets in 20 Myr integrations of the Solar System using different integrators. if we further decrease the timestep, we do not increase the accuracy anymore (this conclusion is wrong as we show be- low). Part of the reason why the energy error is not a good metric for Solar System simulations is that it is a global error metric, but the Solar System has a wide range of or- bital periods and a wide range of planet masses. Jupiter's mass is more than 5000 times larger than that of Mercury, but Mercury's orbital period is 50 times shorter. Mercury has the fewest timesteps along its orbit, but because of its low mass, the errors in its trajectory contribute compara- tively little to the total energy of the system. Nevertheless, it is Mercury's orbit which is most likely to undergo macro- scopic instabilities over billions of years (Laskar & Gastineau 2009). Since the energy error makes it hard to isolate poten- tial issues, we plot different quantities in Fig. 2. Specifi- cally, we plot the semi-major axis, periastron, and secular frequency errors associated with each planet. Although all secular modes couple to each planet's orbital elements to some degree via Eq. 13, we associated the secular frequency g1 with Mercury, g2 with Venus, and so forth, following tra- ditional conventions. Unlike for the energy error, we need to compare the integrations to the true solution to calculate the error in the secular frequencies. We here consider the IAS15 integration as the true solution. The semi-major axis error is scaled relative to each planet's initial semi-major axis. Looking at the top row, we can see that the semi-major axis error for all planets follows qualitatively the results for the energy error. The WHC, WHCKL, and SABACL4 inte- grators perform significantly better than the standard WH integrator. This is expected as to leading order the energy er- ror is simply the sum of the semi-major axes error weighted by planet mass. However, looking at the middle and bottom row of Fig. 2, we can see that the periastron errors and the sec- ular frequency errors behave very differently. Using those metrics, the accuracy of the WHC and WH integrators are now suddenly comparable for the inner Solar System, with WHC even having a slightly larger secular frequency error in g1 (bottom left panel) than WH. The other two integrators in our sample, WHCKL and SABACL4, remain significantly better than the WH method. This, the fact that the WHC method is performing so well when looking at the energy error, but so poorly5 when looking at quantities important for the secular evolution, is the main paradox which we will explain in Sec. 4. 5 Relatively speaking, of course. c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 107106105104relative semi-major axis errorMercuryVenusEarthMarsJupiterSaturnUranusNeptune106105104103102periastron error [rad]101100101timestep [days]1010109108107106105104secular frequency error ["/yr]101100101timestep [days]101100101timestep [days]101100101timestep [days]WHWHCWHCKLSABA4CL101100101timestep [days]101100101timestep [days]101100101timestep [days]101100101timestep [days]Integrators for secularly evolving systems 7 Note that the secular frequency error seems to be a particularly well suited metric for this convergence study. We can now clearly see a second order power law for the WHC method when changing the timestep dt over several orders of magnitude. Compare this to the energy error in Fig. 1, where we were not able to see a power law for WHC. Furthermore, note that the simulations which appeared to be dominated by round-off error for small timesteps in Fig. 1, do in fact continue to improve in accuracy if the timestep is reduced further. 4 ANALYSIS In this section we will explain why, compared to the stan- dard WH method, the WHC method improves the energy error but not the secular frequency error. To do this, we need to introduce a bit of notation first. We use an opera- tor based approach, with a notation similar to that in Rein et al. (2019b) and Rein et al. (2019a). For different perspec- tives, we refer the reader to books and papers by Hairer et al. (2006); Laskar & Robutel (2001), and Wisdom (2018) as well as references therein. 4.1 Error operator 3.3 High accuracy of secular frequencies Before we go on to derive a formalism to explain these re- sults, let us reiterate that we use several integrators which are based on conceptually different approaches to solving the evolution of planetary systems. IAS15 can be thought of as a general purpose high-order integrator, using compen- sated summation, variable timesteps, and not making use of the fact that we are integrating a Hamiltonian system. Then there is the SABACL4 integrator which can also be thought of as a very general purpose scheme for nearly integrable systems, not necessarily being derived for only Hamiltonian systems (Laskar & Robutel 2001). And there is the family of WH integrators whose derivation was heavily influenced by the dynamical systems approach (Wisdom 2018). Above, we have shown evidence that all these differ- ent integrators agree on the secular frequencies of the Solar System to a precision of 10−9 arc-seconds per year for small enough timesteps. This might not be surprising. The inte- grators should converge to the same solution, after all. But given that the integrations differ significantly and consist of up to 1011 timesteps, this is nevertheless a reassuring result. Note that determining the secular frequencies to ex- tremely high precision does not immediately lead to new physical insight about the system. In fact, the frequencies have little obvious physical meaning beyond approximately 1 part in 102 for the inner planets and 1 part in 104 for the outer planets in a secular framework because they can be no longer considered constants at higher precision. That is why they are only quoted to a finite precision in Tab. 1. For example, if we were to measure the frequencies in an even slightly shorter or longer integration, they would change by a factor much larger than 10−9, the precision to which we have determine them. However, the precise physical inter- pretation of the secular frequencies does not matter for our convergence study. The way to think about the secular fre- quencies is not as fundamental constants of the Solar Sys- tem. Rather, they are complicated combinations of action and angle variables which are important for the long-term evolution of the system. And because we can measure them so precisely in a numerical simulation, we can use them to verify the behaviour of our numerical methods in an ideal- ized but well defined model of the Solar System. As we have shown, the observed behaviour of some methods in numer- ical experiments is not always what their formal properties might suggest. c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 (16) We are interested in the error of our splitting methods. Let us therefore define an error operator as Et(y0) ≡ ϕt(y0) − Ψt(y0). At every point in phase space, this operator returns the dif- ference between the true solution ϕ and the approximate solution of our splitting method Ψ after some time t. We can express this operator using exponentials from Eqs. 8 and 10, expanding them, and then matching terms of the same power in t. This quickly becomes tedious but can be done by hand for the leading order term that we're inter- ested in. To leading order, we have Et = (cid:16) −(cid:16) + t2 2 (cid:17) (LA + LB)2 Id + tLA + tLB + t3 (LA + LB)6 + O(t4) 6 t2 Id + tLA + tLB + (LA + LB)2 2 t3 (LA + LB)6 − t3 [LA, [LA,LB]] + 6 24 − t3 [LB, [LA,LB]] + O(t4) 12 t3 [LA, [LA,LB]] + 24 t3 12 (cid:17) = [LB, [LA,LB]] + O(t4). (17) Note that all terms up to second order in t cancel because we have a second order splitting Ψ . To save space we have dropped the state on which the operator is acting on from the notation above. 4.2 Approximating the error operator using a composition Next, consider the composition of operators defined by CAAB ≡ ϕ[A] ◦ ϕ[A]−t ◦ CBA ◦ CAB t t t t (18) where CAB CBA t t ≡ ϕ[A] ≡ ϕ[B] t t t ◦ ϕ[B] ◦ ϕ[A] t ◦ ϕ[A]−t ◦ ϕ[B]−t , ◦ ϕ[B]−t ◦ ϕ[A]−t = and (cid:16)CAB t (cid:17)−1 and thus CAAB t 2t ◦ ϕ[B] = ϕ[A] ◦ ϕ[A] ◦ ϕ[B] t ◦ ϕ[A]−t ◦ ϕ[B]−t ◦ ϕ[A]−t ◦ ϕ[B]−t ◦ ϕ[A]−t . t t 8 Rein et al. Note these operators look like commutators of group ele- ments, in contrast to the commutators of Lie derivatives (which are part of an algebra). After repeatedly applying the BCH formula, and ignoring terms of order O(t4) and higher, we can rewrite this composition as t3 [LA, [LA,LB]] + O(t4) CAAB (y0) = exp Id(y0) (cid:17) (cid:16) t Id + t3 [LA, [LA,LB]] Id + O(t4) = (cid:16) (cid:17) (y0). (19) t . (20) (y0), (cid:17) αt + CBAB βt − 2 · Id + O(t4) 24 and β = 1/ 3√ Comparing this to Eq. 17 and using the analogous opera- tor CBAB, we can explicitly calculate the error operator to (cid:16)CAAB leading order in t via Et(y0) = where α = 1/ 3√ 12. Note that this allows us to calculate E without implementing any Lie brackets. We only need the evolution operators ϕ[A] and ϕ[B]. Further- more, if we are only interested in one of the leading order terms in Eq. 17 we can calculate each term individually. The calculation above can be repeated for different splitting methods. For example, the error operator of a first order method can be expressed to leading order by the composi- tion CAB As previously discussed in Sect. 2, due to the Jacobi identity, we have [LA,LB] g(y) = L{A,B} g(y), which allows us to identify the commutator of Lie derivatives with a cor- responding Poisson bracket of Hamiltonians A and B. Thus, we could also compute the operator CAAB by first explicitly calculating the Poisson bracket {A,{A, B}}, then calculat- ing, and finally solving Hamilton's equations. In practice though, this can get tedious very quickly if calculated by hand for non-trivial Hamiltonians6. On the other hand, the composition methods we introduced above are straightfor- ward to implement using only two already existing opera- tors. Note that these compositions are similar to the compo- sitions used by Wisdom et al. (1996) to express their sym- plectic correctors7. To summarize, the composition CAAB that we have in- troduced above is a way to approximate the evolution opera- tor ϕ[{A,{A,B}}] corresponding to evolution under the Hamil- tonian {A,{A, B}} to leading order in t at every point in phase space y0 without the need to calculate any derivatives, no matter how complicated the Hamiltonian is. Explicitly, from Eq. 19 we have t3 L{A,{A,B}}Id(y0) = (cid:16)CAAB (y0) + O(t4) − Id (cid:17) (21) t t and therefore ϕ[{A,{A,B}}] (22) Note that the factor of t−2 acts like a denominator in a (y0) + O(t2). t2 CAAB (y0) = 1 t t 6 An automatic differentiation algorithm could do this, but most likely not very efficiently as Lie derivatives not only contribute derivatives, but also sums over all coordinates. 7 Similar to how these authors construct higher order symplectic correctors, we could improve our approximation of E by going to higher order. However, since we are interested in the leading order to identify relevant terms anyway, going to higher order is not of much use for us here. finite difference. If the result of this operator is dominated by round-off errors, we can easily rescale t on the right hand side of Eq. 22 to avoid these errors. 4.3 Using the error operator We can use the composition operators to approximate the evolution of any function of the phase space variables under any Poisson bracket. Here, we want to estimate the evolution under the Poisson brackets which appear at leading order in the shadow Hamiltonian of a given splitting scheme. For example, we can approximate the change in energy that occurs after one step with the WH method Ψt due to the term [LA, [LA,LB]] in Eq. 17 as follows + O(t2), ∆tE(y0) = H(y0) − H (cid:16)CAAB (cid:17) (y0) (23) αt where H is the Hamiltonian, now used as a function oper- ating on the phase space variables and returning the energy as a scalar. And α is a constant taking into account the constant coefficients which appear in Eq. 17. In fact we can calculate how any quantity changes due to the [LA, [LA,LB]] term. For example the error in the periastron of a planet after a step of size t is given by ∆t(y0) = (y0) − (cid:16)CAAB + O(t2). (cid:17) (y0) (24) αt Compare this to simply calculating the Poisson bracket {A,{A, B}} explicitly. This would give us the change in en- ergy. But to calculate the change in any other quantity, we would need the extra steps of calculating the corresponding equations of motions and then solving them. 4.4 One planet and general relativistic precession Let us start exploring how these tools might help us under- stand the longterm behaviour of symplectic integrators by looking at a very simple test problem consisting of just one planet orbiting a central mass together with general rela- tivistic corrections in the form of an additional 1/r3 term in the potential (Nobili & Roxburgh 1986). We treat the gen- eral relativistic corrections as a perturbation in part B of the Hamiltonian. In this case they are the sole perturbation as there are no other planets. The dominant Keplerian mo- tion of the planet is described by A. In the particular case we are looking at, Mercury and the Sun, the general rela- tivistic corrections are small, leading to a precession rate of 0.5 arc-seconds per year (Park et al. 2017). None of the symplectic methods described above will be able to solve this problem exactly because ϕ[A] and ϕ[B] don't commute and, among others, the Poisson brackets {A,{A, B}} and {B,{A, B}} are non-zero. Let us focus on the standard WH method for now. In Fig. 3, we plot the one-step as well as the cumulative errors of an integration with the WH integrator and a timestep of about 10% of the planet's orbit. Note that in contrast to the one-planet Ke- pler problem without GR, here neither the phase nor the periastron are exact action or angle variables, but combi- nations thereof. As one can see in the first panel, both the one-step and the cumulative energy errors oscillate but show no sign of any secular increase, as one might expect for a c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 Integrators for secularly evolving systems 9 Figure 4. Periastron error in a simulation with one planet and general relativistic corrections. Solid lines are one step errors, dashed lines are cumulative errors. Figure 3. Energy, phase, and periastron error in a simulation with one planet and general relativistic corrections. Solid lines are one step errors, dashed lines are cumulative errors. symplectic method. The same conclusion could be reached by looking at the periastron errors shown in the third panel, which do seem to average out, and no long-term drift is ob- served. However, one can see in the middle panel, the one step phase errors do not average out, leading to the abso- lute value of the cumulative phase error to increase linearly with time over long timescales. In a system such as the Solar System which evolves primarily due to secular interactions on long timescales, it is important to get the periastron re- solved accurately. The phases on the other hand matter in systems which are dominated by mean motion resonances or encounters. Typically, one would estimate the error of the method by looking at Eq. 10 and conclude that the leading order term is O(dt2) coming from {A,{A, B}}. And indeed, this does provide a good estimate of the energy error. However, this is not the only metric we might want to use. For exam- ple, we might reasonably ask how the errors in the planet's other orbital elements behave. As mentioned above, one way to estimate these errors would be to analytically calculate the evolution under {A,{A, B}} and {B,{A, B}}. For this simple one planet test case this is straightforward to do, at least to leading order. But let us use the composition method introduced above which allows us to quickly estimate the er- rors due to the terms in Eq. 10 using any metric we want by using the already implemented operators ϕ[A] and ϕ[B]. We plot the estimate for the periastron error due to the c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 Figure 5. Periastron error in a long-term simulation with one planet and general relativistic corrections using the WH, WHC, and WHCKL integrators. two leading order term in Eq. 10 in Fig. 4. We calculate these estimates using the CAAB and CBAB operators. The top panel shows that the these estimates match the actual errors (Fig. 3) almost perfectly (the curves are hardly visible because they are on top of each other). The contributions from CBAB are significantly smaller than those from CAAB. Note the different scale on the vertical axes, approximately one factor of . Nevertheless, we can observe a qualitatively different behaviour for the errors coming from CBAB: these errors do not cancel out over one orbital period and lead to a secular growth in the cumulative error. We can use these results to predict the accuracy of the WH integrator over very long timescales without any further integrations. Not only that, we can also correctly predict the long-term behaviour of the WH integrator with symplec- tic correctors and the higher order methods WHCKL and SABACL4 using only the results shown in Fig. 4. Over short timescales, the secular error growth from {B,{A, B}} is not important because it is so much smaller. However, if we integrate for long enough, this will eventually become the dominant contribution to the periastron error. Just by looking at our calculation over five orbital periods, 1012energy error1e16One step errorCumulative error1.00.50.00.5phase error [rad]1e8012345time [orbits]0.50.00.5periastron error [rad]1e80.50.00.5periastron error [rad]1e8One step errorOne step error due to {A,{A,B}}Cumulative error due to {A,{A,B}}012345time [orbits]0.02.55.07.5periastron error [rad]1e14One step error due to {B,{A,B}}Cumulative error due to {B,{A,B}}100101102103104105106107time [orbits]101410121010108periastron error [rad]WHWHCWHCKL10 Rein et al. we can estimate the growth rate and thus estimate when this happens. For this specific problem, the error due to {B,{A, B}} will become dominant after ≈ 106 orbital peri- ods. The above calculation was done for the standard WH integrator. However we can also use it to predict the long- term periastron error of other integrators. If we use WHC with symplectic correctors, we remove the {A,{A, B}} term. This leads to a significant improvement of the integrations's accuracy over short timescales because the only contribu- tion to the periastron error is now the {B,{A, B}} term which is several orders of magnitude smaller (compare the top and bottom panels of Fig. 4). But on long timescales, the {B,{A, B}} term, which is still present in the shadow Hamiltonian even when using the WHC integrator, will grow in exactly the same way as it did for the WH integrator with- out correctors. Thus, over long timescales, we do not expect any benefit from using symplectic correctors in reducing the periastron error. We test this by running the same simula- tion for 107 orbits and comparing the periastron error for the WH and WHC integrators to an IAS15 simulation. The results are shown in Fig. 5. We can see that the correc- tors indeed improve the periastron error significantly, but only on short timescales. On long timescales, (cid:39) 106 orbital periods, the errors are identical for the WH and WHC in- tegrators. Given that Mercury's orbital period is only 88 days, 106 orbits correspond to about 250 kyrs, i.e. a short timescale compared to both the lifetime of the system and the timescale on which the dynamics of the Solar System change significantly (a few million years). We also show the periastron error for the WHCKL integrator in Fig. 5. This integrator does not have the {B,{A, B}} term in the shadow Hamiltonian and we there- fore expect it to perform significantly better. A linear growth of the periastron error is still visible at late times, but this is now suppressed by two orders of magnitude. Given that the now leading error term is two orders of magnitude smaller and the timestep is 10% of the orbital period, we can deduce that the now leading order term must be of order O(cid:0)2dt4(cid:1). 4.5 Two planets We now repeat the line of argument from above, but this time with two planets and without general relativistic cor- rections. The two planets correspond to Mercury and Sat- urn. Saturn contributes about 7.3(cid:48)(cid:48) per century to the pre- cession rate of Mercury (Park et al. 2017). Now part A of the Hamiltonian corresponds to the Keplerian motion of the planets and part B corresponds to planet-planet interac- tions. The same one step and cumulative errors of the pe- riastron as before are shown in Fig. 6. And as before, we can see that the contributions to the periastron error due to {A,{A, B}} average out to leading order. The contribu- tions due to {B,{A, B}} on the other hand do not average out and we can see a secular growth, clearly visible after only 60 Mercury periods. Thus, even though the errors due to {B,{A, B}} are significantly smaller than those due to {A,{A, B}}, we expect that they become dominant after ≈ 105 Mercury periods, i.e. 25 kyrs. Figure 6. Periastron error in a simulation with two planets. Solid lines are one step errors, dashed lines are cumulatative errors. Figure 7. Periastron error in a long-term simulation with two planets using the WH, WHC, and WHCKL integrators. Note that once again, we were able to predict when which error dominates without actually running a long-term integration. To verify this, we run the two-planet case for 106 Mercury period (250 kyr) and plot the periastron er- ror in Fig. 7 for simulations using the WH, WHC, and WHCKL integrators. The results confirm our estimates. On short timescales, the symplectic correctors significantly im- prove the periastron error. However, after about a hundred thousand years, the simulations using correctors have the same error in the periastron as those without. We conclude that symplectic correctors do not improve the accuracy of secularly evolving planetary systems over long timescales. We also show the periastron error for the WHCKL inte- grator in Fig. 7. As before, since this integrator does not have the {B,{A, B}} term in the shadow Hamiltonian, it per- forms significantly better. Periastron errors are still present due to higher order terms, but they are significantly sup- pressed. c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 420246periastron error [rad]1e8One step errorOne step error due to {A,{A,B}}Cumulative error due to {A,{A,B}}01234 time [orbits]0.00.51.01.5periastron error [rad]1e12204060One step error due to {B,{A,B}}Cumulative error due to {B,{A,B}}100101102103104105106107time [orbits]10131010107periastron error [rad]WHWHCWHCKL4.6 Implications for long-term integrations of the Solar System We can now come back to explaining the results of the long- term integration of the entire Solar System presented in Sect. 3. As compared to the simple test problems shown above, the full Solar System is a much more complex dy- namical system. Nevertheless, we can now qualitatively un- derstand why the WHC integrator performs so poorly in re- producing the secular frequencies compared to higher order methods despite its good energy conservation properties. This behaviour is simply a result of the {B,{A, B}} term in the shadow Hamiltonian. Although this term leads to only oscillatory behaviour in the semi-major axis error of a planet as well as the global energy error, it leads to a secular drift in the periastron error. If we use a timestep of 8 days, then after only ≈ 25 kyrs, the error in the periastron of Mercury is dominated by the cumulative contributions from the {B,{A, B}} term for both the WH and WHC in- tegrator. And since the periastron is used to determine the secular frequencies, we find no improvement in the secular frequencies despite the fact that other quantities such as the semi-major axes are more precise. The integrators WHCKL and SABACL4 on the other hand get rid of the {B,{A, B}} term and can therefore achieve a significantly better accuracy of the secular fre- quencies. As one can see from Fig.2, the secular frequencies, and therefore the dynamics, in the outer Solar System can be im- proved by using symplectic correctors (bottom row). Specif- ically, the secular frequencies associated with Jupiter and Saturn, g5 and g6, improve by about one order of magni- tude. Note that this is still significantly less than the im- provement in the semi-major axes (top row) which is of or- der ∼ 10−3. The frequencies g7 and g8 improve even more. We attribute this difference in the convergence of secular frequencies in the inner and outer Solar System to two ef- fects. First, the orbital timescales are longer in the outer Solar System. Thus, it will take longer for the error contri- butions from the {B,{A, B}} term to become dominant for the outer planets. After all, over short timescales, the cor- rectors lead to an improvement (see Fig. 7) and we here only integrate for 20 Myrs. Second, the dynamics in the outer So- lar System are also physically different from the dynamics in the inner Solar System. For example, the semi-major axes of the outer planets change significantly (∆a/a ∼ 2 ∼ 10−6) over the timespan of our integrations (Milani et al. 1987). Since symplectic correctors improve the semi-major axes of all planets, we can expect them to also somewhat improve the secular frequencies in the outer Solar System if we de- termine them to such extreme precision as we do. Note that this shows that secular frequencies are indeed a very useful metric to use in a convergence study given that they can capture these effects. There is a simple physical explanation as to why re- moving the {B,{A, B}} term is important for the accu- rate determination of secular frequencies, but removing the {A,{A, B}} term has little effect according to our results. First note that the secular precession frequencies are to first order functions of the planets' semi-major axes and masses c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 Integrators for secularly evolving systems 11 only (see e.g. Exercise 7.1 in Murray & Dermott 2000). How- ever, the semi-major axes of the planets are constant to within O () in a secularly evolving system as already shown by Laplace (1776). Because the {A,{A, B}} term is O (), its effects on the semi-major axes average out (at least to leading order in ). Thus, this term does also not affect the secular frequencies over long timescales (again at least to leading order in ). 5 CONCLUSIONS In this paper, we have shown that the energy error alone is not a sufficient metric when measuring the accuracy of long-term planetary systems which evolve due to secular in- teractions. Specifically, we have presented evidence showing that the Wisdom-Holman integrator with symplectic cor- rectors does not improve the accuracy of secular frequencies in the inner Solar System compared to the WH integrator without symplectic correctors. This is despite the fact that the symplectic correctors improve the energy error by three orders of magnitude. We have presented a framework that helps us under- stand this behaviour. We are now able to trace the origin of the poor performance to a specific term in the shadow Hamiltonian. If we use a higher-order symplectic integrator such as the Kernel method of Wisdom et al. (1996) or the SABACL4 method of Laskar & Robutel (2001) which get rid of this term, then the integrators' performance is improved significantly. The use of composition operators to study a specific term in the shadow Hamiltonian has multiple advantages over an order of magnitude estimate of Poisson brackets. First, it lets us study terms that grow with time but are initially small and obscured by much larger terms that might not grow with time. Second, we can study the effect of any error term on any arbitrary quantity in a simulation, i.e. not just the energy error. This allows us to study cases where the error in the energy is oscillatory, but the error in other quantities such as the periastron error is not, leading to an increasing error with time. The composition method is very simple to implement as it only uses already implemented functions which are part of the standard Wisdom-Holman integrator and is compatible with any position dependent potential. We have further shown that long-term simulations of the Solar System can achieve an extremely high level of agreement in the secular frequencies across many different integrators. This makes secular frequencies a great tool to study the accuracy of symplectic integrators. If we use higher order methods which get rid of the dominant error terms for the periastron precession of planets, then we can achieve a remarkable level of precision. With a timestep of 10 days, corresponding to only 8 timesteps per shortest dynamical timescale (Mercury's orbital period), we can calculate the secular frequency associated with Mercury to a precision of 10−9 arc-seconds per year! However, it is important to reit- erate that the purpose of this work was not an attempt to produce a list of the most accurate secular frequencies for the Solar System, but to study the performance of various 12 Rein et al. symplectic integrators. For the former, we refer the reader to Laskar et al. (2011) who use a much more realistic model of the Solar System than we do. Further, note that simply determining the secular frequencies to extremely high preci- sion does not lead to new physical insight about the system. In fact, the frequencies are not well defined in secular the- ory beyond approximately 1 part in 102 for the inner planets and 1 part in 104 for the outer planets. Providing evidence that numerical methods converge and that physical results do not depend on nuisance parame- ters such as the timestep or the specific numerical integrator used are crucial for any numerical experiment. Most studies that check the convergence of N -body simulations monitor integrals of motion such as the energy. We argue that it is wise to go beyond this and check that the quantities which are physically relevant for the evolution of the system are converged as well. In the case of secularly evolving systems such as the Solar System, the secular frequencies are a natu- ral choice. As we have shown, having an energy error which is dominated by round-off, does not necessarily imply that a simulation is converged. Reducing the timestep even fur- ther can continue to improve the accuracy, not of the energy error of course, but of other quantities. Finally, let us comment on how chaos fits into this pic- ture. It is well known that the Solar System is chaotic with the shortest Lyapunov timescale being about 5 Myr (Laskar 1989). Let us consider two of our 20 Myr simulations with two different high order integrators. If the timestep is very small, say dt = 0.1 days, then the simulations are converged to machine precision, 10−16. Even if the these simulations use the exact same model and the exact same initial con- ditions, they will differ after only one timestep because the two integrators perform slightly different operations and this leads to a round-off error due to finite floating point preci- sion. Let us assume that the two simulations differ by 10−16 after one timestep. We can then think of one of these sim- ulations as a shadow simulation of the other, using slightly perturbed initial conditions. If the system is chaotic, this dif- ference will grow exponentially. Our integration span four e- folding times, thus the initial difference of 10−16 will increase to about 5 · 10−15 by the end of the simulation. But since we introduce round-off errors at every timestep (not just in the first one) the round-off errors grow with time as well. By the end of the simulation the round-off error is ≈ 10−10, see Eq. 15 and Fig. 1. We obtain the secular frequencies to a pre- cision of 10−9. Thus, the frequencies are clearly not affected by chaos. The precision is either limited by round-off errors for small timesteps, or discretization errors coming from the integrator for large timesteps as shown in Fig. 2. But this is only true for this specific simulation. If we integrate for much longer, then the precision to which we can obtain the secular frequencies will eventually be limited by chaos. To summarize, there are four different contributions to the di- vergence of trajectories in our simulations with symplectic integrators: (i) Discretization errors in the form of oscillating error terms which average out over time and lead to a con- stant error, O (1). out, growing linearly with time, O (t). (ii) Discretization errors where the errors do not average t(cid:1), and growing linearly (iii) Floating point round-off errors, growing like a random walk in predominantly action-like variables in an un- in a biased implementation, O (t). For predominantly angle-like variables this can also lead to errors growing biased implementation, O(cid:0)√ as O(cid:0)t1.5(cid:1) or O(cid:0)t2(cid:1), respectively. leading to an exponential divergence, O(cid:0)et(cid:1). (iv) Chaos, with the seed coming from round-off errors, Which contribution dominates depends on the system at hand, the integrator and timestep used, and last but not least the length of the integration because the various contri- butions depend very differently on time, i.e. O (1), O(cid:0)√ t(cid:1), O (t), O(cid:0)t1.5(cid:1), O(t2), O(cid:0)et(cid:1). The tools described in this pa- per help to determine which of these contributions is limiting the accuracy of trajectories in a specific N -body simulation, and in some cases, how to improve the accuracy. ACKNOWLEDGMENTS We thank an anonymous referee for helpful comments. This research has been supported by the NSERC Discovery Grant RGPIN-2014-04553 and the Centre for Planetary Sciences at the University of Toronto Scarborough. Support for this work was provided by NASA through the NASA Hubble Fel- lowship grant HST-HF2-51423.001-A awarded by the Space Telescope Science Institute, which is operated by the Asso- ciation of Universities for Research in Astronomy, Inc., for NASA, under contract NAS5-26555. This research was made possible by the open-source projects Jupyter (Kluyver et al. 2016), iPython (P´erez & Granger 2007), and matplotlib (Hunter 2007; Droettboom et al. 2016). APPENDIX A: TAKING IAS15 AS THE TRUE SOLUTION Let us comment on why Fig. 2 provides further evidence that the IAS15 simulation is indeed much more accurate than any of the other simulation in our sample. This is not to say that the IAS15 algorithm is superior. The difference is partly due to the fact that IAS15 uses compensated summation to reduce round-off errors whereas the other integrators do not have this capability in our implementation (see also Laskar & Robutel 2001 and Wisdom 2018 who use these algorithms with some form of extended precision). Note that all error metrics for all integrators in Fig. 2 are eventually dominated by round-off errors at small timesteps. Specifically, note that the curves are noisy as ex- pected from a random walk. Because we are comparing all simulations shown in the plot to the same IAS15 simulation, this noisy pattern would not show up if the IAS15 simula- tion were less accurate than the WHFast simulations, or if the round-off error came from the IAS15 simulation. If the IAS15 simulation had a larger error than the other simula- tions, then all the curves for the different integrators would be on top of each other on the left side of Fig. 1 and 2. This gives us further confidence that our assumption of using the IAS15 simulation as the true solution is appro- priate. Alternatively, we could have used one of the higher c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 Integrators for secularly evolving systems 13 Rein, H. & Tamayo, D. 2015, MNRAS, 452, 376 -- . 2017, MNRAS, 467, 2377 Rein, H., Tamayo, D., & Brown, G. 2019b, arXiv e-prints, arXiv:1907.11335 Sidlichovsk´y, M. & Nesvorn´y, D. 1996, Celestial Mechanics and Dynamical Astronomy, 65, 137 Spalding, C., Fischer, W. W., & Laughlin, G. 2018, ApJ, 869, L19 Wisdom, J. 2018, MNRAS, 474, 3273 Wisdom, J. & Holman, M. 1991, AJ, 102, 1528 Wisdom, J., Holman, M., & Touma, J. 1996, Fields Insti- tute Communications, Vol. 10, p. 217, 10, 217 Yoshida, H. 1990, Physics Letters A, 150, 262 order symplectic integrations with the smallest timestep as the reference simulation. This would still allow us to deter- mine when the error is dominated by round-off errors. But we would not be able to determine the power law in the round-off error dominated regime. REFERENCES Brouwer, D. 1937, AJ, 46, 149 Droettboom, M., Hunter, J., Caswell, T. A., Firing, E., Nielsen, J. H., Elson, P., Root, B., Dale, D., Lee, J.- J., Seppnen, J. K., McDougall, D., Straw, A., May, R., Varoquaux, N., Yu, T. S., Ma, E., Moad, C., Silvester, S., Gohlke, C., Wrtz, P., Hisch, T., Ariza, F., Cimarron, Thomas, I., Evans, J., Ivanov, P., Whitaker, J., Hobson, P., mdehoon, & Giuca, M. 2016, matplotlib: matplotlib v1.5.1 Hairer, E., Lubich, C., & Wanner, G. 2006, Geometric nu- merical integration: structure-preserving algorithms for ordinary differential equations, Vol. 31 (Springer Science & Business Media) Hunter, J. D. 2007, Computing In Science & Engineering, 9, 90 Kluyver, T., Ragan-Kelley, B., P´erez, F., Granger, B., Bus- sonnier, M., Frederic, J., Kelley, K., Hamrick, J., Grout, J., Corlay, S., et al. 2016, Positioning and Power in Aca- demic Publishing: Players, Agents and Agendas, 87 Laplace, P.-S. 1776, M´emoires de lAcad´emie Royale des Sciences de Paris, 7, 69 Laskar, J. 1988, Astronomy and Astrophysics, 198, 341 Laskar, J. 1989, Nat, 338, 237 Laskar, J. 1990, Icarus, 88, 266 -- . 1993, Physica D: Nonlinear Phenomena, 67, 257 -- . 2003, arXiv:math/0305364 Laskar, J., Fienga, A., Gastineau, M., & Manche, H. 2011, A&A, 532, A89 Laskar, J. & Gastineau, M. 2009, Nat, 459, 817 Laskar, J. & Robutel, P. 2001, Celestial Mechanics and Dynamical Astronomy, 80, 39 Milani, A., Nobili, A. M., & Carpino, M. 1987, A&A, 172, 265 Murray, C. D. & Dermott, S. F. 2000, Solar System Dy- namics (Cambridge University Press) Nobili, A. & Roxburgh, I. W. 1986, in IAU Symposium, Vol. 114, Relativity in Celestial Mechanics and Astrome- try. High Precision Dynamical Theories and Observational Verifications, ed. J. Kovalevsky & V. A. Brumberg, 105 -- 110 Park, R. S., Folkner, W. M., Konopliv, A. S., Williams, J. G., Smith, D. E., & Zuber, M. T. 2017, The Astronom- ical Journal, 153, 121 P´erez, F. & Granger, B. E. 2007, Computing in Science and Engineering, 9, 21 Quinn, T. & Tremaine, S. 1990, AJ, 99, 1016 Rein, H., Hernandez, D. M., Tamayo, D., Brown, G., Eck- els, E., Holmes, E., Lau, M., Leblanc, R., & Silburt, A. 2019a, MNRAS, 485, 5490 Rein, H. & Liu, S.-F. 2012, A&A, 537, A128 Rein, H. & Spiegel, D. S. 2015, MNRAS, 446, 1424 c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000
1111.0667	1	1111	2011-11-02T21:10:57	Self-consistent size and velocity distributions of collisional cascades	[ "astro-ph.EP" ]	The standard theoretical treatment of collisional cascades derives a steady-state size distribution assuming a single constant velocity dispersion for all bodies regardless of size. Here we relax this assumption and solve self-consistently for the bodies' steady-state size and size-dependent velocity distributions. Specifically, we account for viscous stirring, dynamical friction, and collisional damping of the bodies' random velocities in addition to the mass conservation requirement typically applied to find the size distribution in a steady-state cascade. The resulting size distributions are significantly steeper than those derived without velocity evolution. For example, accounting self-consistently for the velocities can change the standard q=3.5 power-law index of the Dohnanyi (1969) differential size spectrum to an index as large as q=4. Similarly, for bodies held together by their own gravity, the corresponding power-law index range 2.88<q<3.14 of Pan & Sari (2005) can steepen to values as large as q=3.26. Our velocity results allow quantitative predictions of the bodies' scale heights as a function of size. Together with our predictions, observations of the scale heights for different sized bodies for the Kuiper belt, the asteroid belt, and extrasolar debris disks may constrain the mass and number of large bodies stirring the cascade as well as the colliding bodies' internal strengths.	astro-ph.EP	astro-ph	Self-consistent size and velocity distributions of collisional cascades Margaret Pan1 and Hilke E. Schlichting2,3,4 [email protected], [email protected] ABSTRACT The standard theoretical treatment of collisional cascades derives a steady- state size distribution assuming a single constant velocity dispersion for all bodies regardless of size. Here we relax this assumption and solve self-consistently for the bodies' steady-state size and size-dependent velocity distributions. Specifically, we account for viscous stirring, dynamical friction, and collisional damping of the bodies' random velocities in addition to the mass conservation requirement typically applied to find the size distribution in a steady-state cascade. The resulting size distributions are significantly steeper than those derived without velocity evolution. For example, accounting self-consistently for the velocities can change the standard q = 3.5 power-law index of the Dohnanyi (1969) differential size spectrum to an index as large as q = 4. Similarly, for bodies held together by their own gravity, the corresponding power-law index range 2.88 < q < 3.14 of Pan & Sari (2005) can steepen to values as large as q = 3.26. Our velocity results allow quantitative predictions of the bodies' scale heights as a function of size. Together with our predictions, observations of the scale heights for different sized bodies for the Kuiper belt, the asteroid belt, and extrasolar debris disks may constrain the mass and number of large bodies stirring the cascade as well as the colliding bodies' internal strengths. 1. Introduction Believed to be a primary mechanism operating in circumstellar dusty debris disks as well as our own Kuiper and asteroid belts, collisional cascades -- the transfer of mass from 1Department of Astronomy, University of California, Berkeley, CA 94720 2Department of Earth and Space Sciences, University of California, Los Angeles, CA 90095 3California Institute of Technology, MC 130-33, Pasadena, CA 91125 4Hubble Fellow -- 2 -- larger to smaller sized bodies via collisions between those bodies -- are ubiquitous in our galaxy. Their widespread occurrence and potential importance in understanding planet for- mation and planet-disk interactions have naturally provoked considerable study. Theoretical treatments predicting the collisional size distribution have included analytic work as well as numerical simulations. The pioneering treatment of Dohnanyi (1969), who analytically cal- culated the size distribution for a steady-state cascade of constant-strength bodies, has been elaborated upon, extended to size-dependent strength laws, and applied to different physi- cal contexts by several authors, including Williams & Wetherill (1994); Tanaka et al. (1996); O'Brien & Greenberg (2003); Kenyon & Bromley (2004); O'Brien & Greenberg (2005); Pan & Sari (2005); Lohne et al. (2008). Some of these also considered non-power-law features in the size distribution such as waves due to the gravity-strength transition (O'Brien & Greenberg 2003, 2005; Pan & Sari 2005) or changes in the fragment size spectrum (Belyaev & Rafikov 2011). Many numerical studies of collisional cascades have also been performed (see, for example, Campo Bagatin et al. 1994; Durda & Dermott 1997; Kenyon & Bromley 2004; Krivov et al. 2005; Kenyon & Bromley 2008; Lohne et al. 2008; Wyatt 2008; Fraser & Kavelaars 2009). The analytic and most of the numerical works on collisional cascades have generally as- sumed that the bodies' velocity dispersion is independent of size and constant in time once the cascade begins. Nevertheless, we expect processes like viscous stirring and collisional damping to affect all bodies' velocities throughout the cascade's lifetime. Here we incorpo- rate velocity evolution processes into our treatment of collisional cascades and find the size spectrum and size-dependent velocity dispersion self-consistently. In §2 we give expressions for the physical processes operating in the cascade. The well-known mass conservation con- dition is the basis of previous work beginning with Dohnanyi (1969), so we summarize it quickly in §2.1. In §2.2 we describe the stirring and damping processes affecting the veloc- ities, and in §2.3 we give expressions for the velocity equilibrium required in a steady state cascade. These are the velocity analogues of the mass conservation conditions of §2.1. In §3 we impose mass conservation and velocity equilibrium together to find the size and velocity power-laws of steady-state cascades. As we explain, in a disk of bodies with a power-law size distribution we expect to see up to three different velocity regimes. We derive velocity and size power laws in all three regimes for both gravity-dominated and strength-dominated bodies. In §4 we compare our analytic results to those of our fragmentation simulations. Finally, in §5 we summarize and discuss our findings. -- 3 -- 2. Size and velocity evolution processes In order to find the size distribution and velocity function self-consistently, we assume a debris disk that occupies an annulus with typical orbital angular frequency Ω. The bodies in the cascade have uniform composition and body mass density ρ, and their sizes r cover the range [rmin, rmax]. We write the differential body size spectrum as dN/dr ∝ r−q and the velocity function as v(r) ∝ rp. We consider in turn how mass conservation in the cascade and velocity evolution via gravitational stirring and collisional damping constrain q and p. Since our primary goal is to clearly delineate the relevant physical processes, we work to order of magnitude throughout. 2.1. Mass conservation We begin with mass conservation, the basis for most analytic cascade treatments to date. Our discussion of mass conservation parallels that of Pan & Sari (2005), and we refer readers to that work for a more detailed description. In a steady-state collisional cascade where mass is conserved in catastrophic collisions, the total mass destroyed per unit time per logarithmic interval in radius must be independent of size. This implies constant = ρr3 · N(r) · = ρr3 · N(r) · N(rB(r)) volume · r2 · vrel N(rB(r))r2Ω area (1) (2) where rB(r) is the size of the smallest body, or bullet, that can destroy a target of size r in a collision and vrel is the typical relative velocity of bullets and targets. The second line follows because we assume isotropic velocities, so that the scale height of the disk is of order vrel/Ω. The volume of Eq. 1 in which the bodies move is the area occupied by the disk midplane times this scale height, so the mass conservation relation for a disk is independent of velocity. We further assume that the way the bodies break is independent of size -- that is, that the shape of the average fragment size distribution is size-independent. We parameterize the bullet-target size relation as a power law rB(r) ∝ rα. Then Eq. 2 yields q = 6 + α 1 + α . (3) The value of α depends on how much energy is lost in the post-impact destruction process. In the gravity regime, we can think of the destruction as a shock induced in the target by the bullet which propagates to the antipode of the impact site. The limiting cases for the shock -- 4 -- propagation are energy conservation and momentum conservation in the shocked material; these give respectively esc(r) ∼ ρr3 ρr3v2 ρr3vesc(r) ∼ ρr3 Bv2(r) −→ α = (5 − 2p)/3 Bv(r) −→ α = (4 − p)/3 . (4) (5) Here we assumed p ≥ 0: p < 0 would in principle require arbitrarily large velocities for arbitrarily small sizes, so we will not consider that case here. Numerical simulations of catastrophic collisions find 1.37 < α < 1.66 with constant collision velocities (see, for ex- ample, Stewart & Leinhardt 2009; Benz & Asphaug 1999, and references therein); this is consistent with the range 4/3 < α < 5/3 for p = 0 which we find above. Together, Eqs. 3 -- 5 imply 22 − p 7 − p 23 − 2p 8 − 2p > q > . (6) The inequalities hold if p < 1, which as we will see in §3 is satisfied. In the strength regime, α depends on the material properties of the body, which are often parameterized as Q∗(r), the energy per unit mass needed to destroy a body of size r. In the strength regime simulations find Q∗(r) ∝ rγ where 0 ≥ γ > −1/2 (Benz & Asphaug 1999; Stewart & Leinhardt 2009). Then ρr3Q∗(r) = ρr3 Bv2(r) −→ α = 1 + γ − 2p 3 where, again, we assume p ≥ 0. With Eq. 3 this gives 21 + γ − 2p 6 + γ − 2p q = . (7) (8) As an example, Dohnanyi (1969) used in effect γ = p = 0 in the strength regime; these immediately yield q = 7/2 in Eq. 8. 2.2. Velocity evolution processes We now consider velocity evolution. Physically, v(r) depends on stirring from larger bodies and damping from collisions with and dynamical friction from smaller bodies. In the following, we explore the stirring-damping balance in detail. Motivated by observations of the asteroid and Kuiper belts, we work in the regime where the typical relative velocity in an encounter between two bodies is larger than either body's Hill velocity, which is of order a body's escape velocity times the one-sixth power of the ratio of the body's mass and the central mass. -- 5 -- We begin by writing expressions for the rates at which viscous stirring, collisional damp- ing, and dynamical friction damping affect a body of size r. The rate at which bodies are If viscously stirred by bodies of size R ≥ r depends on which of them is moving faster. vesc(R) > v(R) > v(r), the rate is 1 dv(r) v(r) dt (cid:12)(cid:12)(cid:12)(cid:12)stir ∼ N(R)R2Ω area v(R) (cid:19)2(cid:18)vesc(R) v(r) (cid:19)2 (cid:18)vesc(R) . This focusing factor applies because we need only double v(r), not necessarily v(R). v(R) < v(r) < vesc(R), we have (9) If 1 dv(r) v(r) dt (cid:12)(cid:12)(cid:12)(cid:12)stir ∼ N(R)R2Ω area v(r) (cid:19)4 (cid:18)vesc(R) . (10) Because vesc(R) ∝ R, the largest bodies in the system do most of the stirring unless q > 5 when v(R) > v(r) or unless q > 7 when v(R) < v(r). We expect these conditions to hold in real systems, so we will assume them throughout. We will show in §3 that this assumption is self-consistent in the cascade. The rate at which bodies of size r are damped by direct collisions with bodies of size s ≤ r is 1 dv(r) v(r) dt N(s)Ω area s3 r . (11) There is no focusing factor here because, as we discuss later, dynamical friction damping is faster than damping by direct collisions only if v(r) ≤ vesc(r) 1. Note that bodies in the cascade must have v(r) ≥ vesc(r) because catastrophic collisions would otherwise be impossible. Eq. 11 implies that if q > 4, collisional damping is dominated by the smallest bodies in the disk, giving s = rmin. (cid:12)(cid:12)(cid:12)(cid:12)damp ∼ When q ≤ 4, collisions between equal-sized bodies dominate2, and Eq. 11 becomes 1 dv(r) v(r) dt (cid:12)(cid:12)(cid:12)(cid:12)damp ∼ N(r)Ω area r2 . (12) However, in the cascade rB(r) ≤ r. Then the collisional destruction rate will be at least as fast as the damping rate of Eq. 12 as long as q > 1; the rates are equal only when rB(r) = r. 1Indeed, if v(r) ≥ vesc(r), dynamical friction is equivalent to elastic direct collisions. 2The q = 4 size spectrum is a marginal case in which collisions with bodies of all sizes should contribute equally to the damping. Since this represents only an order unity correction to the damping rate, the scalings given remain valid. -- 6 -- Simulations of catastrophic collision ejecta indicate that nearly all the kinetic energy relative to the bullet-target center of mass is lost to heat in a catastrophic collision (see, for example, Jutzi et al. 2010). Then bodies whose bullet-target mass ratio is not too small should lose most of their velocity dispersion in a catastrophic collision. If we assume that bodies are indeed damped whenever they are destroyed, then 1 dv(r) v(r) dt (cid:12)(cid:12)(cid:12)(cid:12)damp ∼ (cid:12)(cid:12)(cid:12)(cid:12)coll ∼ 1 dv(r) v(r) dt N(rB(r))r2Ω area . (13) If instead destructive collisions cannot damp effectively -- that is, if the largest fragment retains most of its pre-collision velocity -- then whether or not collisional cooling is effective depends on the age of the disk. Bodies of size r are collisionally cooled only if the disk age is longer than the collision timescale implied by Eq. 12. The dynamical friction damping rate of size r bodies by size s < r bodies also depends on whether v(r) > v(s) or vice versa. By analogy to the expressions of Eqs. 9 and 10 for viscous stirring, the two expressions for dynamical friction are 1 dv(r) v(r) dt (cid:12)(cid:12)(cid:12)(cid:12)df ∼   N(s)Ω area s3 r (cid:18)vesc(r) v(s) (cid:19)2(cid:18)vesc(r) v(r) (cid:19)2 v(s) (cid:19)4 r (cid:18)vesc(r) area s3 N(s)Ω v(s) < v(r) < vesc(r) v(r) < v(s) < vesc(r) . (14) esc(r)/(v(s)v(r))2 As mentioned above, dynamical friction acts faster than direct collisions by v4 if v(r) > v(s) or by (vesc(r)/v(s))4 if v(r) < v(s), so it applies to bodies with v(r) < vesc(r) which have not entered the cascade. The dynamical friction damping rate scales as s4−q−2p if v(r) > v(s) and as s4−q−4p if v(r) < v(s). Then the smallest bodies with s = rmin dominate the damping if q + 2p > 4 when v(r) > v(s) or if q + 4p > 4 when v(r) < v(s). 2.3. Velocity equilibrium With expressions in hand for rates of velocity evolution, we can impose the steady-state condition that stirring and damping balance. In addition to the size and velocity power laws q, p of bodies in the cascade, we consider the analogous power laws q ′, p′ for any bodies of size r > rmax which may be present in the disk but are too large to have entered the cascade. If the bodies in the disk formed through core accretion, we would expect 1 < q ′ < 5 (see, for example, Kenyon & Bromley 2004, 2008; Schlichting & Sari 2011) as well as the q < 5 we already assumed. Motivated by the discussion after Eq. 10, we let the largest bodies in the disk have size R. -- 7 -- We first consider bodies in the cascade. As explained in §2.2, these bodies are vis- cously stirred and collisionally damped. For cascades in which catastrophic collisions damp velocities effectively, velocity equilibrium means 1 dv(r) v(r) dt 1 dv(r) v(r) dt N(R)R2Ω area 0 = ∼ ∼   1 v(r) dt dv(r) (cid:12)(cid:12)(cid:12)(cid:12)stir − (cid:18)vesc(R) (cid:12)(cid:12)(cid:12)(cid:12)coll v(r) (cid:19)2(cid:18)vesc(R) v(R) (cid:19)2 v(r) (cid:19)4 (cid:18)vesc(R) N(R)R2Ω area N(rB(r))r2Ω area N(rB(r))r2Ω area − − (15) v(r) < v(R) < vesc(R) vesc(R) > v(r) > v(R) (16) where we have applied Eqs. 9, 10, and 13. Equivalently, if v(R) > v(r) the ratio of stirring and collision rates is N(R) N(rB(r)) (cid:18)R 1 ∼ ∼ (cid:16)rmax r (cid:17)α(1−q)+2+2p · r (cid:19)2(cid:18) vesc(R) v(r) (cid:19)2 v(R) (cid:19)2(cid:18)vesc(R) max (cid:18) vesc(R) N(rB(rmax)) R2 r2 N(R) v(R) (cid:19)2(cid:18) vesc(R) v(rmax)(cid:19)2 and if v(R) < v(r) this ratio is 1 ∼ (cid:16)rmax r (cid:17)α(1−q)+2+4p · N(R) N(rB(rmax)) R2 r2 v(rmax)(cid:19)4 max (cid:18) vesc(R) . (17) (18) (19) Note that we have transferred the coefficient of rα(1−q) in N(rB(r)) to N(rB(rmax)) in the last step. In Eqs. 18 and 19, the second through last terms on the right-hand side are simply the ratio of the stirring and collision rates for size rmax bodies, those at the top of the cascade: compare them, for example, to the right-hand side of Eq. 17, which is the ratio of stirring and collision rates for size r bodies. This indicates that if the stirring and collision rates for size rmax bodies balance -- which we expect since these bodies have just entered the cascade -- the rest of the cascade will also be in velocity equilibrium if q = 1 + q = 1 + 2 + 2p α 2 + 4p α v(R) > v(r) v(R) < v(r) . (20) In the gravity regime, using Eqs. 4 and 5 together with Eq. 20 gives -- 8 -- 10 + 5p 4 − p 10 + 11p 4 − p > q > > q > 11 + 4p 5 − 2p 11 + 10p 5 − 2p v(R) > v(r) v(R) < v(r) . (21) The inequalities for q hold when −1/2 < p < 1, which as we will see in §3 is satisfied. Similarly, in the strength regime we use Eq. 7 with Eq. 20 to get q = q = 9 + γ + 4p 3 + γ − 2p 9 + γ + 10p 3 + γ − 2p v(R) > v(r) v(R) < v(r) . (22) We next consider a disk in which catastrophic collisions do not damp the velocities. This may occur, for example, if rB(r) ≪ r, in which case the center of mass velocity of a colliding bullet-target pair is dominated by the target velocity. Then conservation of momentum dictates that the velocity of the largest collisional fragment will be quite similar to the target's velocity even if all of the relative kinetic energy between the bullet and target is lost. If q ≤ 4 and if the system's lifetime is at least as long the timescale for two bodies of size rmax to collide, damping occurs through collisions between equal-sized bodies according to Eq. 12. This damping mechanism dominates for all bodies with v(r) > vesc(r). While this condition holds over the entire cascade, it may hold for bodies outside the cascade as well. To see that all bodies in the cascade are included, note that if v(r) < vesc(r), then the impact energy in a collision between equal-sized bodies, ∼ρr3v2(r), is less than the gravitational binding energy ∼ρr3v2 esc(r) of either body. If p > 0, the impact energy in a collision with a smaller bullet is likewise less than the gravitational binding energy of the target. Since both gravity-dominated and strength-dominated bodies require impact energy at least as large as their gravitational binding energies, v(r) > vesc(r) is required in the cascade. However, v(r) > vesc(r) may also apply for some bodies larger than rmax. A calculation entirely analogous to that of Eqs. 16 -- 20 above which uses damping by Eq. 12 rather than Eq. 13, and the size where v(r) ∼ vesc(r) instead than rmax, gives q = 3 + 2p q = 3 + 4p v(R) > v(r) v(R) < v(r) . (23) If q > 4, we substitute Eq. 11 for Eq. 12 in the above calculation. In this case N(r) disappears from the ratio of stirring and damping rates and we get a condition on p alone: p = 1/2 p = 1/4 v(R) > v(r) v(R) < v(r) . (24) -- 9 -- If the cascade lifetime is short compared to the timescale for collisions between bodies of size rmax, then some bodies near the top of the cascade will not have had time to damp. For these undamped bodies, we expect shallower velocity power laws. Finally, we consider any bodies in the disk whose velocities are smaller than their escape velocities. They cannot be part of the cascade, so we expect them to have sizes r > rmax. The equilibrium velocities of these bodies follow from a balance between viscous stirring and dynamical friction. If v(r) > v(R), we equate the stirring rate of Eq. 9 with the damping rates of Eq. 14 to get v(r) v(r) v(s) ∼ (cid:18)N(R) v(s) ∼ (cid:18)N(R) N(s) (cid:19)1/2 R3 N(s) (cid:19)1/4 R3/2 s3/2r3/2 v(s) < v(r) v(s) > v(r) . (25) s3/4r3/4 Here s is the size of bodies which dominate the dynamical friction. Because we have broken power-law size and velocity distributions, and because the power-law breaks do not occur at r, we expect s to be independent of r. If v(r) ∝ rp′, Eq. 25 implies p′ = −3/2 p′ = −3/4 v(s) < v(r) v(s) > v(r) . (26) This is indeed consistent with the v(r) > v(R) we assumed. Note that the kinetic energy per body, ∼ρr3v2(r) ∝ r3+2p′, cannot increase with decreasing body size, so these bodies lie outside the cascade. For the same reason we can neglect any dynamical friction heating effects, which contribute at most an order unity correction. If instead we assume v(r) < v(R) and replace Eq. 9 with Eq. 10 above, no self-consistent solution for p′ is possible. 3. Steady-state size and velocity distributions We now solve simultaneously the mass conservation and velocity stirring/damping bal- ance conditions of §2 to find the steady-state size and velocity distributions in the disk. We first confirm that the steady-state condition -- equivalent to requiring that N(rmax) changes on a timescale long compared to collisions between and stirring of smaller bodies -- is physical. Since the stirring cross-section of size r bodies scales as r−2p, smaller bodies are indeed stirred faster than the largest bodies break. Similarly, since smaller bodies have more total surface area than larger bodies as long as q > 3, smaller bodies break faster than -- 10 -- larger ones. Our assumption of a steady state is therefore reasonable for all bodies smaller than rmax. Said another way, rmax corresponds to the location of the break seen in colli- sional size distributions separating collisional and primordial bodies (O'Brien & Greenberg 2003; Kenyon & Bromley 2004; Pan & Sari 2005). As bodies of size rmax break and N(rmax) decreases, the normalization of the cascade below rmax should follow adiabatically. In addition, we can see that these solutions are stable by considering a perturbed cas- cade. If for any reason stirring becomes faster than catastrophic collisions, the velocities will increase and rB(r) will decrease until the catastrophic collision rate equals the new stirring rate. Similarly, if stirring becomes slower than catastrophic collisions, the velocities will slow and rB(r) will increase until collisions just balance stirring as long as v(r) ≥ vesc(r). The timescale for bodies smaller than some size r to relax to this solution should be of order a few catastrophic collision times for size r bodies. We frame our discussion of the solutions via the velocity stirring/damping equilibria listed in §2.3. They suggest that given a disk in which a single rB(r) power-law relation applies to all bodies, and in which cooling has had time to operate, up to three velocity regimes occur3. First, the largest bodies, which are stirred viscously and damped by dy- namical friction, have velocities that are below their escape velocities but that increase with decreasing size according to Eq. 26. At the size for which the bodies' velocity equals their es- cape velocity, dynamical friction can no longer cool efficiently and the second regime begins. Bodies slightly smaller than this first transition size have velocities faster than both their own escape velocities and the velocity of the largest bodies in the disk. In this regime stir- ring proceeds according to Eq. 10 and damping proceeds by collisions. Third, if the cascade includes sufficiently small bodies, we expect for p > 0 that the smallest bodies' velocities fall below the velocity of the largest body. In this regime stirring proceeds according to Eq. 9 and collisional damping continues. While we would expect the power-law breaks associated with transitions between regimes will produce waves in the size and velocity distributions, we also expect that, on average, the sizes and velocities in each regime will be consistent with the q and p values we find. We discuss these three regimes -- first, v(r) < vesc(r); second, v(r) > vesc(r) and v(R) < v(r); third, v(r) > vesc(r) and v(R) > v(r) -- in turn below. The regime containing the largest bodies of sizes r > rmax is simplest. Because its bodies are not part of the cascade, we cannot constrain their size distribution except to require that it satisfy the conditions for Eq. 10 to hold. Instead we expect that their size distribution 3If significant external stirring has occurred, not all of these three regimes may occur. For example, if all the bodies in the disk have velocities larger than their own escape velocities, dynamical friction will never be important. -- 11 -- N(r) ∝ r1−q′ has not changed since their formation. Regardless of whether the bodies are gravity- or strength-dominated, Eq. 26 gives p = −3/2 , p = −3/4 , 1 < q < 7 1 < q < 7 v(s) < v(r) v(s) > v(r) . (27) The remaining two velocity regimes, v(r) < v(R) and v(r) > v(R), may support cas- cades. We consider cascades with four different categories of rB(r) relations characterized by 1) whether the bodies are gravity- or strength-dominated and 2) whether the bullet-target size ratio is close enough to unity for catastrophic collisions to provide effective cooling. First we assume cooling by catastrophic collisions. This case requires a cascade. In the gravity regime, we solve Eqs. 6 and 20 simultaneously to get √241 17 − 4 11 − √85 4 p = p = 31 − √865 8 p = p = This implies 1 4 , q = , q = √241 − 9 √85 − 3 2 2 , q = √865 − 23 2 , q = 3 for α = for α = for for α = α = 4 − p 3 5 − 2p 3 4 − p 3 5 − 2p 3 v(R) > v(r) (28) v(R) < v(r) . (29) 0.37 < p < 0.45 3.26 > q > 3.11 0.20 < p < 1/4 3.21 > q > 3 v(R) > v(r) v(R) < v(r) . (30) (31) In the strength regime, we likewise solve Eqs. 8 and 20 together for p = 9 + γ −p69 + 6γ + γ2 4 , q = −1 − γ +p69 + 6γ + γ2 2 p = 15 + 2γ −p201 + 36γ + 4γ2 8 , q = −7 − 2γ +p201 + 36γ + 4γ2 2 v(R) > v(r) (32) v(R) < v(r) . (33) -- 12 -- For the range −1/2 < γ ≤ 0, these give 0.090 < p ≤ 0.17 3.82 > q ≥ 3.65 0.054 < p ≤ 0.10 3.78 > q ≥ 3.59 v(R) > v(r) v(R) < v(r) . (34) (35) Now we assume catastrophic collisions cannot damp the velocities significantly, so that the cooling timescale is the time it takes for a given body to collide with a total mass equal to its own. We also assume the lifetime of the disk is longer than this cooling timescale for all bodies with v(r) > vesc. In the gravity regime, these bodies all participate in the cascade: v(rmax) ∼ vesc(rmax). Their steady state sizes and velocities should follow from Eqs. 6 and 23. When α = (4 − p)/3 this gives 6 − √34 2 √34 6 − 4 p = p = √34 √34 , q = 9 − , q = 9 − for α = for α = 4 − p 3 4 − p 3 v(R) > v(r) v(R) < v(r) . (36) We find, however, that no solution with p ≥ 0 and q ≤ 4 is possible when α = (5− 2p)/3. It turns out q > 4 is also impossible in the gravity regime. If q > 4, Eq. 6 implies p > 3/2, and since vesc(r) ∝ r, having p > 1 means that v(r) will fall below vesc(r) at some r, stopping the cascade. Then the maximum α allowed must lie between (4 − p)/3 and (5 − 2p)/3. To find this limiting value, we recast Eqs. 4 and 5 as ρr3vβ esc(r) ∼ ρr3 Bvβ(r) −→ α = 1 + β(1 − p)/3 (37) where 1 < β < 2. With Eqs. 3 and 23, this gives 3 p = p = β ± p36 − 6β + 4β2 4β ± p144 + 12β + 9β2 12 + β 4β 2β v(R) > v(r) v(R) < v(r) . (38) A look at the zeros of dp/dβ shows that p is monotonic for the relevant β, so the limiting α and β should occur at a limiting value of p. For gravity-dominated bodies, 0 ≤ p ≤ 1 as discussed above; for both v(R) > v(r) and v(R) < v(r), the only β between 1 and 2 that satisfies p = 0 or p = 1 is β = 3/2 at p = 0. When p = 0, Eq. 23 gives q = 3. Then the -- 13 -- allowed p, q in the gravity regime are 0 ≤ p < 0.085 3 ≤ q < 3.17 0 ≤ p < 0.042 3 ≤ q < 3.17 v(R) > v(r) v(R) < v(r) . (39) (40) In the strength regime, not all of the bodies with v(r) > vesc(r) can participate in the cascade. For those in the cascade, we first assume q ≤ 4 and solve Eqs. 8 and 23 simultaneously. This gives p = p = 4 + γ −p4 + 16γ + γ2 5 + γ −p19 + 14γ + γ2 4 4 , , q = 10 + γ −p4 + 16γ + γ2 q = 8 + γ −p19 + 14γ + γ2 2 v(R) > v(r) (41) v(R) < v(r) (42) When γ < 0, the only real solutions to Eq. 41 have q > 4, which is inconsistent. The allowed ranges in p, q are p = 1/2 q = 4 1/4 > p ≥ 0.16 4 > q > 3.64 v(R) > v(r) γ = 0 v(R) < v(r) −1/2 < γ ≤ 0 . (43) (44) (45) (46) The q > 4 which arose above when v(R) > v(r) and γ < 0 suggests that we look for a solution where the smallest bodies in the system dominate the collisional damping -- that is, a solution using Eq. 24 instead of Eq. 23. Indeed, Eqs. 8 and 24 together give p = 1/2 , q = 20 + γ 5 + γ v(R) > v(r) (47) and, for −1/2 < γ ≤ 0, p = 1/2 13/3 > q > 4 v(R) > v(r) . (48) For bodies with v(r) > vesc(r) but r > rmax -- those not in the cascade -- the primordial size distribution q ′ applies. As long as q ′ satisfies the conditions on Eq. 9, the velocities follow from this and Eq. 23 if q ′ < 4 or Eq. 24 if q ′ > 4. -- 14 -- Finally, if the collisional cooling timescale is shorter than the age of the cascade, the velocity distribution will be shallower than predicted in the relevant regime above. How much shallower depends on particulars of the stirring timescale and the energy loss per collision. For example, if the kinetic energy lost in a catastrophic collision is so small that the kinetic energy retained by the fragments is larger than the energy they gain via stirring in one collision time, p will instead depend on exactly how much energy is lost in an average collision. In turn the energy loss per collision depends heavily on the bodies' internal structure, which is very poorly constrained (Leinhardt et al. 2008, and references therein). We will not discuss this uncooled regime in detail here. Our results for p and q in all the velocity and strength law regimes discussed in this work are summarized in Table 1. Note that all the size distributions are steeper than those that obtain when fixed velocities are used (p = 0); these are 3.14 > q > 2.88 for the gravity regime (Pan & Sari 2005) and 3.72 > q ≥ 3.5 for the strength regime. The steepening is certainly consistent with smaller velocities for smaller bodies: lower velocities mean larger bullets are needed to break a target of a given size; an increase in bullet size corresponds to a decrease in the number of bullets for q > 0; and a steeper size distribution offsets this decrease. Table 1 also confirms that our assumption p < 1 of §2.1 is self-consistent. 4. Comparison with numerical simulations To test the analytic results above we used a numerical cascade simulation based on the coagulation code of Schlichting & Sari (2011). Because our goal here is to study the dominant physical processes in the cascade -- viscous stirring, collisional and dynamical friction damping, and mass transfer from larger to smaller body sizes -- we neglect factors of order unity in the stirring and damping rates. We study a single belt of bodies orbiting in an annulus about a much more massive star. We take the initial total mass in bodies to be about 1MEarth, and we assume the bodies have bulk density 1 g/cc and follow the mass and velocity evolution of bodies with radii ranging from 1 m to 3000 km, a few times the size of Pluto. As a first test of our velocity evolution theory, we artificially fix the size spectrum in the simulations and allow only the velocities to evolve. In Figure 1 we show as an example the test results with strength-dominated γ = 0 bodies and a fixed q = 3.6 size spectrum. Since we do not allow for catastrophic collisions in this run and since we fix the size spectrum at q < 4, the collisional damping is dominated by collisions between similarly sized bodies as given in Eq. 12. The resulting steady-state velocities obey Eq. 23, which for q = 3.6 means p = 0.3 if v(R) > v(r) and p = 0.15 if v(R) < v(r). We expect the velocities for large bodies -- 15 -- v(R) > v(r) v(R) < v(r) references damping mechanism catastrophic collisions collisions with equal-sized bodies catastrophic collisions collisions with equal-sized bodies 0.37 < p < 0.45 3.26 > q > 3.11 0 ≤ p < 0.085 3 ≤ q < 3.17 0.090 < p ≤ 0.17 3.82 > q ≥ 3.65 p = 1/2 q = 4 0.20 < p < 1/4 3.21 > q > 3 0 ≤ p < 0.042 3 ≤ q < 3.17 0.054 < p ≤ 0.10 3.78 > q ≥ 3.59 1/4 > p > 0.16 4 > q ≥ 3.64 Eqs. 30, 31 Eqs. 39, 40 Eqs. 34, 35 Eqs. 44, 46 Eq. 48 Eq. 27 v(r) > vesc(r): includes all bodies in cascade gravity regime strength regime v(r) < vesc(r): bodies too large for cascade gravity or strength regime collisions with smallest bodies p = 1/2 13/3 > q ≥ 4 -- -- dynamical frictiona -- -- p = −3/2 or p = −3/4 1 < q < 7 Table 1: Summary of velocity power laws p and size power laws q in steady-state for all of the stirring and damping regimes discussed in this work. ap = −3/2 applies when vesc(r) > v(r) > v(s); p = −3/4 applies when v(r) < v(s) < vesc(r). Here v(s) is the velocity of the bodies providing the dynamical friction. -- 16 -- with velocities below their own escape velocities to follow Eq. 26. Our simulations agree well with these numbers. Similarly, we test our mass cascade implementation by artificially fixing the velocity as a function of size and allowing only the size distribution to evolve. Figure 2 shows the results of a test with strength-dominated γ = 0 bodies and velocities fixed to a broken power law with p = 1/4, p = 1/8. For these Eq. 8 gives q = 3.73, q = 3.61; these agree well on average with our simulations. Our simulations also show waves as mentioned in §3; these are induced by the break in the velocity distribution as well as the artificial "breaks" in the mass power law created by the finite range of body sizes in our simulations. Finally, we allow both the size and velocity distributions to evolve in the simulations. Figure 3 shows an example again using strength-dominated bodies with γ = 0. We assumed in this run that the collisional damping of the velocity dispersion is dominated by collisions between like-sized bodies (see Eq. 12). This criterion applies when catastrophic collisions do not damp the velocity dispersion significantly, which may occur for small bullet-to-target ratios. Here the steady-state solution of Eq. 41 applies, and γ = 0 implies p = 1/2, q = 4 when v(R) > v(r) and p = 0.16, q = 3.64 when v(R) < v(r). Again, these agree well with our simulations on average in each of the three different velocity regimes. This model and the results shown in Figure 3 may, for example, apply at the end of protoplanetary growth in a planetesimal disk. Initially, the velocity dispersion is so small that collisions lead to growth. As the largest bodies -- "protoplanets" -- grow, they continue to excite the small planetesimals' velocity dispersion; their velocities grow on the same timescale as the large protoplanets' sizes (for a comprehensive description of this growth phase see Schlichting & Sari (2011)). Once the system reaches an age comparable to the small planetesimals' collision time, but before collisions become destructive, the balance between gravitational stirring and collisional damping determines the planetesimals' velocity dispersion. This phase is similar to the situation shown in Figure 1, but with a mass spectrum that continues to evolve due to planetesimal accretion. Finally, the planetesimals' velocity dispersion is excited sufficiently above their escape velocities that destructive collisions set in. This stage is shown in Figure 3. The mass spectrum now no longer reflects the growth history; instead it is determined by the collisional evolution. 5. Summary We have found self-consistent steady-state solutions for the velocity function and size distribution of collisional cascades in the super-Hill regime. These solutions occur when -- 17 -- q = 3 . 6 5 ´ 1026 1 ´ 1026 5 ´ 1025 1 ´ 1025 5 ´ 1024 1 ´ 1024 g s s a M 0.001 0.01 0.1 1 10 Radius km 100 1000 1 ´ 105 5 ´ 104 s m c v 2 ´ 104 1 ´ 104 5000 5 5 5 p = 0 . 1 p = 0 . 1 p = 0 . 1 p p p = = = - - - 1 1 1 . . . 5 5 5 p = 0.3 p = 0.3 p = 0.3 0.001 0.01 0.1 1 10 Radius km 100 1000 Fig. 1. -- Comparison between analytic results (dashed blue line) and simulations (solid orange curve) for steady-state velocities in a system of strength-dominated bodies with γ = 0. The top panel shows the mass in a given log2 mass bin as a function of radius, which in this run is fixed with q = 3.6; the bottom panel shows the simulations and analytic results for the velocities. Since we do not allow for catastrophic collisions in this run and since we fix the size spectrum at q < 4, the collisional damping is dominated by collisions between similarly sized bodies as given in Eq. 12. There is good agreement in each of three velocity regimes. The smallest bodies, which have velocities greater than their own escape velocities but less than v(R), follow p = 0.3 (see Eq. 23). Larger bodies still small enough to have velocities larger than their own escape velocities, but which have velocities greater than v(R), follow p = 0.15 (see Eq. 23). Finally, the largest bodies have p′ = −3/2 because they are subject to dynamical friction by small bodies with velocity dispersion v(r) < v(R) (see Eq. 26). 5 ´ 1026 2 ´ 1026 1 ´ 1026 5 ´ 1025 2 ´ 1025 1 ´ 1025 g s s a M -- 18 -- 3 3 q = 3 .7 q = 3 .7 q = 3.61 q = 3.61 0.001 0.01 0.1 1 10 Radius km 100 1000 s m c v 1 ´ 105 5 ´ 104 2 ´ 104 1 ´ 104 5000 p = 1 8 p = 1 8 4 4 p = 1 p = 1 0.001 0.01 0.1 1 10 Radius km 100 1000 Fig. 2. -- Comparison between analytic results (dashed blue line) and simulations (solid orange curve) in a steady-state system of strength-dominated bodies with γ = 0 and fixed velocity distribution. The top panel shows the resulting mass spectrum plotted as mass in a given log2 mass bin as a function of radius (solid orange curve) and the corresponding analytic results (dashed blue line); the bottom panel show the fixed velocity distribution. We begin the simulations with an initial size distribution q = 3.6, which evolved to the q = 3.61 for v(r) > v(R) and q = 3.73 for v(r) < v(R) as expected from Eq. 8 with p = 1/4, p = 1/8. Waves are clearly visible as oscillations in the steady-state mass spectrum. We note here that the wavelength of the waves changes as one transitions from the p = 1/4 to the p = 1/8 velocity distribution. This change in wavelength reflects the velocity dependence in the bullet-to-target ratio. -- 19 -- q = 3.6 4 q = 3.6 4 5 ´ 1026 2 ´ 1026 1 ´ 1026 5 ´ 1025 2 ´ 1025 1 ´ 1025 g s s a M q=4 q=4 0.001 0.01 0.1 1 10 Radius km 100 1000 5 ´ 104 s m c v 2 ´ 104 1 ´ 104 5000 0.5 0.5 0.5 p= p= p= 6 6 6 p = 0 . 1 p = 0 . 1 p = 0 . 1 p p p = = = - - - 1 1 1 . . . 5 5 5 2000 0.001 0.01 0.1 1 10 Radius km 100 1000 Fig. 3. -- Comparison between analytic results (dashed blue line) and simulations (solid orange curve) in which the size distributions and velocities are both evolved together. The top panel shows the mass spectrum; the bottom panel shows the velocity distribution. In this run we assumed that the collisional damping of the velocity dispersion is dominated by collisions between like-sized bodies (see Eq. 12). This damping criterion applies when catastrophic collisions do not damp the velocity dispersion significantly, which may occur for small bullet-to-target ratios. The agreement between the simulations and our analytic results in Eq. 41 and Eq. 26 is good on average for the v(R) < v(r) and v(R) > v(r) regimes in both the mass and velocity plots as well as for the v(r) < vesc(r) regime. (see caption of Figure 1 for a description of the regimes). Waves due to both the transition between v(R) > v(r) and v(R) < v(r) and the ends of our simulation range are again visible in the data. -- 20 -- mass conservation is satisfied and when viscous stirring balances velocity damping. Three kinds of velocity equilibrium may occur. For the biggest bodies, which have velocities slower than their escape velocities, viscous stirring and dynamical friction balance. These bodies' velocities increase with decreasing size until the size at which the velocity and escape velocity are equal. Since dynamical friction is inefficient for bodies with velocities faster than their escape velocities, stirring balances damping by direct collisions for all smaller bodies. Bodies just smaller than this first transition size have velocities faster than both their escape veloc- ities and the velocity of the largest bodies in the system. A second transition occurs at the body size whose velocity equals that of the largest bodies in the system. Bodies smaller than this second transition have velocities slower than the largest bodies in the system, so their stirring requires a different cross-section. The resulting size distributions for the gravity- and strength-dominated regimes are steeper than the ones expected with size-independent velocities. We find good agreement between the predictions of our theory and the results of our numerical simulations. To our knowledge, previous analytic treatments of collisional cascades have not consid- ered velocity stirring or damping. Wyatt (2008) and Kennedy & Wyatt (2010) study disks in which the cascade start time depends on orbital radius because the large bodies needed to excite the velocity dispersion and initiate a cascade take longer to accrete at larger orbital radii. However, they do not consider the effects of stirring or damping on colliding bodies' velocities as the cascade proceeds. Kenyon & Bromley (2008) do account for simultaneous velocity and size spectrum evo- lution in their coagulation/fragmentation code. Our results here are not directly comparable to theirs because they do not account for energy lost during catastrophic disruptions and because the largest bodies in their simulations continue to accrete while their collisional cas- cades operate. We plan to extend and modify our calculations to enable comparison with their findings. Other areas for future investigation include incorporating velocity stirring and damping into collisional cascades covering both the gravity and strength regimes as well as the waves induced in the size and velocity power laws due to transitions between regimes. A good knowledge of the size and velocity distributions will also allow us to predict observables such as the dust production rate as a function of time and the scale height of the disk as a function of size or, for the smallest bodies, observing wavelength. Ongoing surveys of the Kuiper and asteroid belts provide observational size distribution and velocity data to which we can compare our results. Kuiper belt surveys indicate that its size spectrum follows a broken power law whose break falls at a body size of several tens of kilometers (Bernstein et al. 2004; Fraser 2009); this break is interpreted as the top of a collisional cascade. Typical Kuiper belt velocities are about 1 km/s, of order 30 times larger -- 21 -- than the escape velocities from the largest bodies in the cascade, so the typical bullet/target size ratio is far from unity. Then cooling by catastrophic collisions should be ineffective. Also, the timescales for the observed Kuiper belt objects (KBOs) to collide with bodies of equal size are longer than the age of the solar system. These KBOs have therefore not had time to cool; we would expect their average velocities should be very similar to those of the primordial KBOs. Indeed, small KBOs' eccentricities and inclinations show no significant trends with size. As for the size distribution, assuming the break exists, surveys find a range of size distributions 1.9 < q < 3.9 for KBOs smaller than the break size (Bernstein et al. 2004; Fraser et al. 2008; Fraser 2009; Schlichting et al. 2009). This is consistent with the 2.88 < q < 3.14 we expect if p ≃ 0 but not strongly constraining. In the asteroid belt, typical relative velocities of ∼5 km/s suggest catastrophic collisions are likewise ineffective at cooling. Surveys of the asteroid belt indicate a size distribution of q ≃ 3.5 for large bodies of H magnitude smaller than about 15, or size larger than ∼1 km (Gladman et al. 2009, and references therein). For smaller bodies, however, the slope becomes shallower; different surveys report slopes ranging from q = 2 to q = 2.8 (Ivezi´c et al. 2001; Yoshida et al. 2003; Yoshida & Nakamura 2007; Wiegert et al. 2007). While the overall size distribution slope is roughly consistent with the expected 2.88 < q < 3.14, we would predict that the average slope steepen for bodies smaller than about 100 m in size. Still, our theory alone suggests several possible causes for waves that might explain the observed break and its location. This again makes strong constraints difficult without further data on smaller bodies. We look forward to future observations of smaller KBOs and asteroids whose cooling time may be shorter than the belts' lifetimes and which will provide a longer size baseline with which to compare our theory. Future survey results of this kind will provide more stringent tests of our results and may shed light on the catastrophic collision process in our solar system. In particular, measurements of the slopes of the size and velocity distributions would provide a direct probe of the bodies' strengths. Similarly, observations of debris disk scale heights as a function of wavelength at millimeter wavelengths, for example with ALMA, would provide direct tests of our velocity power laws as well as constraints on the internal strengths of pebble-sized particles in those disks. It is a pleasure to thank Re'em Sari for helpful discussions. For HS support for this work was provided by NASA through Hubble Fellowship Grant # HST-HF-51281.01-A awarded by the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., for NASA, under contact NAS 5-26555. -- 22 -- REFERENCES Belyaev, M. A., & Rafikov, R. R. 2011, Icarus, 214, 179 Benz, W., & Asphaug, E. 1999, Icarus, 142, 5 Bernstein, G. M., Trilling, D. E., Allen, R. L., Brown, M. E., Holman, M., & Malhotra, R. 2004, AJ, 128, 1364 Campo Bagatin, A., Cellino, A., Davis, D. R., Farinella, P., & Paolicchi, P. 1994, Planet. Space Sci., 42, 1079 Dohnanyi, J. S. 1969, J. Geophys. Res., 74, 2531 Durda, D. D., & Dermott, S. F. 1997, Icarus, 130, 140 Fraser, W. C. 2009, ApJ, 706, 119 Fraser, W. C., & Kavelaars, J. J. 2009, AJ, 137, 72 Fraser, W. C., Kavelaars, J. J., Holman, M. J., Pritchet, C. J., Gladman, B. J., Grav, T., Jones, R. L., Macwilliams, J., & Petit, J.-M. 2008, Icarus, 195, 827 Gladman, B. J., Davis, D. R., Neese, C., Jedicke, R., Williams, G., Kavelaars, J. J., Petit, J.-M., Scholl, H., Holman, M., Warrington, B., Esquerdo, G., & Tricarico, P. 2009, Icarus, 202, 104 Ivezi´c, Z., Tabachnik, S., Rafikov, R., Lupton, R. H., Quinn, T., Hammergren, M., Eyer, L., Chu, J., Armstrong, J. C., Fan, X., Finlator, K., Geballe, T. R., Gunn, J. E., Hennessy, G. S., Knapp, G. R., Leggett, S. K., Munn, J. A., Pier, J. R., Rockosi, C. M., Schneider, D. P., Strauss, M. A., Yanny, B., Brinkmann, J., Csabai, I., Hind- sley, R. B., Kent, S., Lamb, D. Q., Margon, B., McKay, T. A., Smith, J. A., Waddel, P., York, D. G., & the SDSS Collaboration. 2001, AJ, 122, 2749 Jutzi, M., Michel, P., Benz, W., & Richardson, D. C. 2010, Icarus, 207, 54 Kennedy, G. M., & Wyatt, M. C. 2010, MNRAS, 405, 1253 Kenyon, S. J., & Bromley, B. C. 2004, AJ, 128, 1916 -- . 2008, ApJS, 179, 451 Krivov, A. V., Sremcevi´c, M., & Spahn, F. 2005, Icarus, 174, 105 -- 23 -- Leinhardt, Z. M., Stewart, S. T., & Schultz, P. H. 2008, Physical Effects of Collisions in the Kuiper Belt, ed. Barucci, M. A., Boehnhardt, H., Cruikshank, D. P., Morbidelli, A., & Dotson, R. (University of Arizona Press), 195 -- 211 Lohne, T., Krivov, A. V., & Rodmann, J. 2008, ApJ, 673, 1123 O'Brien, D. P., & Greenberg, R. 2003, Icarus, 164, 334 -- . 2005, Icarus, 178, 179 Pan, M., & Sari, R. 2005, Icarus, 173, 342 Schlichting, H. E., Ofek, E. O., Wenz, M., Sari, R., Gal-Yam, A., Livio, M., Nelan, E., & Zucker, S. 2009, Nature, 462, 895 Schlichting, H. E., & Sari, R. 2011, ApJ, 728, 68 Stewart, S. T., & Leinhardt, Z. M. 2009, ApJ, 691, L133 Tanaka, H., Inaba, S., & Nakazawa, K. 1996, Icarus, 123, 450 Wiegert, P., Balam, D., Moss, A., Veillet, C., Connors, M., & Shelton, I. 2007, AJ, 133, 1609 Williams, D. R., & Wetherill, G. W. 1994, Icarus, 107, 117 Wyatt, M. C. 2008, ARA&A, 46, 339 Yoshida, F., & Nakamura, T. 2007, Planet. Space Sci., 55, 1113 Yoshida, F., Nakamura, T., Watanabe, J.-I., Kinoshita, D., Yamamoto, N., & Fuse, T. 2003, PASJ, 55, 701 This preprint was prepared with the AAS LATEX macros v5.2.
1812.07471	1	1812	2018-12-18T16:40:33	The Liquidus Temperature for Methanol-Water Mixtures at High Pressure and Low Temperature, with Application to Titan	[ "astro-ph.EP", "cond-mat.mtrl-sci" ]	Methanol is a potentially important impurity in subsurface oceans on Titan and Enceladus. We report measurements of the freezing of methanol-water samples at pressures up to 350~MPa using a volumetric cell with sapphire windows. For low concentrations of methanol, the liquidus temperature is typically a few degrees below the corresponding ice freezing point, while at high concentrations it follows the pure methanol trend. In the Ice-III regime, we observe several long-lived metastable states. The results suggest that methanol is a more effective antifreeze than previously estimated, and might have played an important role in the development of Titan's subsurface ocean.	astro-ph.EP	astro-ph	Confidential manuscript submitted to JGR-Planets The Liquidus Temperature for Methanol-Water Mixtures at High Pressure and Low Temperature, with Application to Titan A. J. Dougherty1, Z. T. Bartholet1, R. J. Chumsky1, K. C. Delano1∗, X. Huang1, D. K. Morris1† 1Department of Physics, Lafayette College, Easton, Pennsylvania 18042 USA Key Points: • Titan's icy shell and subsurface ocean may be affected by any impurities such as • We measure the freezing point of methanol/water solutions as a function of pressure • Previous estimates may have underestimated the effectiveness of methanol as an an- methanol tifreeze 8 1 0 2 c e D 8 1 . ] P E h p - o r t s a [ 1 v 1 7 4 7 0 . 2 1 8 1 : v i X r a ∗Current address, Dept. of Physics, U. Texas at San Antonio, San Antonio, TX, USA †Current address, Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena, CA, USA Corresponding author: Andrew Dougherty, [email protected] -- 1 -- Confidential manuscript submitted to JGR-Planets Abstract Methanol is a potentially important impurity in subsurface oceans on Titan and Enceladus. We report measurements of the freezing of methanol-water samples at pressures up to 350 MPa using a volumetric cell with sapphire windows. For low concentrations of methanol, the liquidus temperature is typically a few degrees below the corresponding ice freezing point, while at high concentrations it follows the pure methanol trend. In the Ice-III regime, we observe several long-lived metastable states. The results suggest that methanol is a more ef- fective antifreeze than previously estimated, and might have played an important role in the development of Titan's subsurface ocean. 1 Introduction A number of icy satellites likely contain subsurface oceans [Nimmo and Pappalardo, 2016]. On Titan, evidence for a subsurface ocean comes from several sources, including gravity measurements [Iess et al., 2012] and electrical measurements of a Schumann reso- nance [Beghin et al., 2012], as well as computations based on obliquity [Baland et al., 2011] and surface topography [Hemingway et al., 2013; Nimmo and Bills, 2010]. Detailed models of the interior of Titan rely on the properties of aqueous solutions at the low temperatures and high pressures found within such an ocean, and such properties would be affected by the presence of impurities. Important impurities likely include ammo- nia, but methanol may also be present [Kargel, 1992; Hogenboom et al., 1997] and is in- cluded in models such as Neveu et al. [2017], Dunaeva et al. [2016], Lefevre et al. [2014], Fortes [2012], Tobie et al. [2012], Davies et al. [2010], and Deschamps et al. [2010]. More broadly, methanol is also an important volatile found in the outer solar system nebula [Sekine et al., 2014], on the surfaces of some trans-Neptunian objects [Merlin et al., 2012], and in some star-forming regions [Dartois et al., 1999; Shimonishi et al., 2016]. As a potent antifreeze compound, small concentrations of methanol could play an im- portant role in the formation and development of the subsurface ocean [Deschamps et al., 2010]. In addition, the rheology of methanol-containing slurries could have significant im- plications for the structure of possible cryovolcanic flows [Zhong et al., 2009; Davies et al., 2010; Lopes et al., 2007]. Although there is not conclusive evidence for widespread cryovol- canic activity on Titan [Moore and Pappalardo, 2011; Lopes et al., 2013], there are regions for which it may be a reasonable explanation of topographic features [Lopes et al., 2013; Solomonidou et al., 2014], and it is one possible source of replenishment for Titan's atmo- spheric methane [Davies et al., 2016]. On Enceladus, the observations of water-rich plumes [Dougherty et al., 2006; Porco et al., 2006; Waite et al., 2009; Hansen et al., 2011; Hsu et al., 2015] reveal either the pres- ence of a liquid reservoir or highly active melting. The composition of the plumes suggests a liquid reservoir [Hsu et al., 2015; Postberg et al., 2011], and the measured libration is consis- tent with a global subsurface ocean, rather than a localized reservoir [Thomas et al., 2016]. As with Titan, any subsurface ocean would likely contain impurities, such as ammonia and methanol [Kargel, 1992], that act as powerful antifreeze compounds. Small amounts of methanol may have been detected on the surface of Enceladus [Hodyss et al., 2009], as well as in the plume [Waite et al., 2009]. In addition to being a powerful antifreeze, methanol could also play a role in the formation of methane hydrates [McLaurin et al., 2014]. More generally, thermodynamic models of the interiors of icy moons can benefit from an improved understanding of how the addition of small amounts of methanol will change the temperature and pressure-dependent properties of ice [Deschamps et al., 2010; Hsieh and Deschamps, 2015]. In the model of Deschamps et al. [2010], for example, the authors estimated the pressure-dependent behavior of a methanol-water solution by interpolating between the pure water and pure methanol behavior. -- 2 -- Confidential manuscript submitted to JGR-Planets In this work, we present results for the liquidus temperature for methanol concentra- tions between 5 wt.% and 75 wt.%, for pressures from 5 MPa to 350 MPa. We find that inter- polation underestimates the effect of small concentrations of methanol on the freezing point. 2 Materials and Methods 2.1 Materials The phase diagram for methanol/water solutions at atmospheric pressure is shown in Fig. 1. The eutectic concentration is approximately 88 wt.%. Figure 1. Phase diagram for methanol/water, based on compilations by Haghighi et al. [2009], Vuillard and Sanchez [1961], and Miller and Carpenter [1964], and adapted from Kargel [1992]; Ott et al. [1979]; Takaizumi and Wakabayashi [1997]. Phases are solid unless indicated otherwise. The hydrated solid phase is methanol monohydrate [Fortes, 2006; Fortes et al., 2011]. Sample solutions were made from HPLC-grade methanol (EMD Chemicals, 99.8%) and HPLC-grade filtered water (Fisher Scientific). Solutions of approximately 4.9, 9.9, 33.8, and 75 wt.% methanol were studied. (All concentrations are methanol mass divided by to- tal solution mass.) Since methanol is highly hygroscopic, the concentration would change somewhat during the extensive filling and leak-testing process. The beaker of solution was exposed to the atmosphere for periods of hours, and residual water from flushing and clean- ing the connecting tubing could not be completely eliminated. In addition, when leak-testing with partially frozen Ice Ih samples, any leaked fluid would be enriched in methanol. The net effect could be a significant change, on the order of 4%. Accordingly, in each case, once the system was filled and sealed, and leak testing was completed, the concentration of the sample was estimated by extrapolating the low-pressure liquidus temperature to 1 atmosphere, and comparing it with the data in Fig. 1. Those estimated concentrations are the values reported here. The range of concentrations was chosen to explore both the low-concentration regime where the initial solid formed is ice Ih, and the high-concentration regime where the initial solid formed is methanol monohydrate. The former is relevant for the proposed primordial bulk composition of Titan's ocean [Deschamps et al., 2010], while the latter would be rele- vant for the freezing of a trapped subsurface body of liquid, where the concentration would tend towards the eutectic. -- 3 -- 160 180 200 220 240 260 280 0 20 40 60 80 100LiquidH2O + LiquidCH3OH . H2O + H2OCH3OH . H2O + LiquidCH3OH + LiquidT (K)Wt.% MethanolVuillard & Sanchez (1961)Miller & Carpenter (1964)Haghighi et. al (2009)Confidential manuscript submitted to JGR-Planets The range of pressures was limited by the maximum pressure able to be maintained with the apparatus, generally about 350 MPa. For models of Titan that include an outer Ice Ih shell roughly 100 km thick and a subsurface ocean atop an inner high-pressure Ice V or Ice VI layer, such as Fortes [2012] and Dunaeva et al. [2014], this would correspond to the pressure in the ocean at a depth of roughly 120 km below the Ice Ih shell. 2.2 Methods The apparatus and technique are similar to those described in previous papers for mag- nesium sulfate and ammonia-water mixtures [Hogenboom et al., 1995, 1997], enhanced to allow optical access to the sample. The system is shown in Fig. 2. Approximately 1 mL of sample is loaded into a pressure cell that is placed in a cop- per container and immersed in an insulated, temperature-controlled ethanol/water bath. The pressure cell is made from a 316 stainless steel block with four ports, known as a cross (High Pressure Equipment Company #60-HF6). Two opposing ports contain replaceable plugs that have sapphire windows sealed in them with epoxy. The third port contains a plug in which a silicon diode thermometer is installed. The fourth port connects the cell to the pressure sys- tem. Pressure is applied through a pump with ethylene glycol as the pump fluid. Pressure is measured with two Heise gauges: an analog pressure gauge valid for all pressures studied, and a more sensitive digital pressure gauge for pressures below 200 MPa. Figure 2. Diagrams of the pressure system (left) and imaging system (right). The imaged portion of the sample is confined to a 1 mm-wide gap between the two windows at the center of the cross. This system al- lows simultaneous measurements of pressure, temperature, and volume changes, along with optical images of the sample. The sample in the pressure cell is separated from the ethylene glycol pump fluid by a vertical U-tube filled with mercury. A steel capillary tube of constant cross section forms one arm of the U-tube. A small Alnico magnet is placed in the capillary on the interface between the pump fluid and the mercury, and the height of that magnet is measured by a transducer. Changes in the transducer voltage are approximately proportional to changes in sample vol- ume. As long as the sample is mostly liquid, this system allows simultaneous measurements of temperature, pressure, and volume of the sample. -- 4 -- PumpPressuregaugeCrossPlugwithwindowTemperatureprobeCoppercontainerTemperature-controlledbathVolumeTransducermagnetmercuryMethanol/watersolutionPumpfluidCrossTemperatureprobeMirrorCameraMicroscopeObjectiveRelaylensesMirrorLampOpticalFiberConfidential manuscript submitted to JGR-Planets The imaging system consists of a lamp that shines light through an infrared filter and optical fiber that directs the beam horizontally through the sample cell. The infrared filter is used to minimize heating of the sample by the light source. After passing through the cell, the beam is reflected by a 45◦ mirror upward through a matched pair of lenses to a long working distance optical microscope objective coupled to a Pulnix digital camera. The cam- era obtains images of a vertical cross-section of the sample, with 1392 × 1040 pixels and an overall resolution of about 1.7 µm/pixel. The gap between the sapphire windows is ap- proximately 1 mm. Although the camera's field of view does not cover the entire system, we typically observed dissolving or growing crystals corresponding to changes in temperature, pressure, and volume, indicating that the crystal images reflect the phase transitions within the sample. For runs above a temperature of 210 K, the temperature is controlled by a immersion cooler in the ethanol/water bath. The cooling and warming rates are controlled by com- puter. In these viscous solutions, it is important to allow sufficient time for the system to approach thermodynamic equilibrium. Typically, the temperature was increased at a rate of 0.005 K/minute, or even more slowly, to ensure that the system was near thermodynamic equilibrium. For colder temperatures (such as were needed for the 75 wt.% runs), liquid ni- trogen was used to cool the system, and the system warmed up as the nitrogen evaporated. The warming rates in those cases were typically closer to 0.1 K/minute. 3 Data The results for a run with a 4.9 wt.% solution at a nominal pressure of 263 MPa are shown in Fig. 3. The horizontal axis shows temperature, while the vertical axis gives the transducer voltage, which varies approximately linearly with volume. The run started with a homogeneous liquid (point a) at a temperature of 256 K. Upon cooling, the liquid con- tracted gradually until about 237.8 K (point b), where the supercooled sample rapidly froze and many ice crystals were visible in the window. This crystallization was accompanied by a sharp decrease in volume and pressure, indicating that the solid formed was denser than the surrounding liquid. Upon further cooling, the pressure reached about 252 MPa at a temper- ature of 237 K (point c). The sample was then gradually warmed at a rate of 0.005 K/min. As the sample warmed, the volume gradually increased and ice crystals could be observed falling downward in the image. At 245.9 K (point d), the melting was complete, and the now- liquid sample gradually expanded with temperature as it continued to warm (point e). A typi- cal run would last about three days. The liquidus temperature was then estimated by finding the intersection of straight lines fitted to the end of the melting curve (just before point d) and the pure liquid warming curve (after point d). For sufficiently slow warming, the transition was fairly sharp and the uncertainties in the liquidus temperature were typically less than ±0.2 K. For transition tem- peratures below about 210 K, where liquid nitrogen was used for cooling and the warming rates were closer to 0.1 K/min, the transitions were more rounded and the uncertainties were closer to ±1.0 K. Uncertainties in pressure were typically less than ±0.3 MPa for runs below 200 MPa, and about ±1 MPa for runs above 200 MPa, due to the different pressure gauges used. An image of crystals formed at 212.8 MPa, presumably Ice-III, is shown in Fig. 4. The volume decreased when these crystals formed, indicating that the solid is denser than the surrounding liquid. Upon warming, the volume increased and the crystals fell downward, again indicating a denser solid phase. -- 5 -- Confidential manuscript submitted to JGR-Planets Figure 3. Data for a run with 4.9 wt.% methanol at a nominal pressure of 263 MPa. The vertical axis shows transducer voltage (which varies approximately linearly with volume), while the horizontal axis shows temperature. The labelled points are described in the text. The liquidus temperature for this run was 245.9 K. Ice-III crystals in a 4.9 wt.% methanol solution slowly melting at a pressure of 212.8 MPa and Figure 4. temperature of 245.5 K (about 2 K below the liquidus temperature). The image is approximately 2.2 mm across. Gravity points downward. As the crystal melts, the arms detach and fall downward, indicating that the solid is more dense than the surrounding liquid. 4 Results 4.1 Low methanol concentrations The resulting liquidus temperatures for two low-concentration samples (4.9 wt.% and 9.9 wt.%) are shown in Fig. 5, along with the boundaries for various phases of ice. (For the 9.9% sample, a leak in the pressure system prevented the acquisition of data at pressures above about 200 MPa.) In the region between 200 MPa and 350 MPa, we observed three different states, all for the same 4.9 wt.% sample. In one run at 244 MPa, the volume increased upon freezing, and the ice crystals were less dense than the surrounding fluid. The liquidus point for that sam- ple was approximately 243 K, clearly consistent with a metastable continuation of the Ice-Ih -- 6 -- 0.42 0.43 0.44 0.45 0.46 0.47 0.48 0.49 0.5 0.51 236 238 240 242 244 246 248 250 252 254 256 258(a) (b)(c)(e)(d)Transducer Voltage (V) (~Volume)Temperature (K)Confidential manuscript submitted to JGR-Planets Figure 5. Liquidus temperatures for two low-concentration samples as a function of pressure. Shown for comparison are the pure ice melting curves from Wagner et al. [2011] and the phase boundaries for the differ- ent ice phases from Dunaeva et al. [2010]. The uncertainties in the data points are typically smaller than the symbols on the graph. The solid lines through the data are fits described in the text. regime curve. We also observed two different states where the volume decreased upon freez- ing, and the ice crystals were more dense than the surrounding fluid. These states fall on two distinct curves in Fig. 5. The states labeled Ice-III had a lower density than the ones labeled Ice-V, but we were not able to measure absolute densities of solids with this apparatus. The lower-temperature data may follow an extension of the Ice V regime curve, consistent with the formation of metastable Ice V, as suggested by Evans [1967], and are included in the Ice V regime fits. The two states also differed in the experimental protocol necessary for their forma- tion. For pressures below about 240 MPa, the system would spontaneously freeze upon su- percooling into either Ice-Ih or Ice-III (as in Fig. 4). For pressures above 240 MPa, the sys- tem would spontaneously freeze as Ice-V (as in Fig. 3). In order to prepare Ice-III samples at higher pressures, we would first cool at a pressure of approximately 230 MPa until ice crys- tals formed, and then raise the pressure up to the desired target. 4.2 Fits To fit the data, we considered a number of different candidate functional forms. In par- ticular, we tried functions of the form used by Wagner et al. [2011] for modeling the pure ice melting curves, (1) (cid:19) bi(cid:33) (cid:18) T T0 = 1 + p p0 1 − ai (cid:32) 3 1 where the numerical values of the parameters for pure ice are given in Wagner et al. [2011]. However, our data do not cover a wide enough range of temperatures to constrain the parameters, particularly in the Ice III regime, especially since the the values for p0 and T0 are not independently known. -- 7 -- 230 235 240 245 250 255 260 265 270 275 0 50 100 150 200 250 300 350 400Ice IhIce IIIce IIIIce VLiquidTemperature (K)Pressure (MPa)4.9%: Ice-Ih4.9%: Ice-III4.9%: Ice-V9.9%: Ice-IhConfidential manuscript submitted to JGR-Planets We also considered the Simon-Glatzel equation [Simon and Glatzel, 1929], as well as the modified form proposed by Kechin [1995]: (cid:18) (cid:19) a2 T T0 = 1 + ∆p a1 exp(−a3∆p) where ∆p = p − p0. This is similar to including a single term from Eq. 1, but with an ad- ditional exponential decay term a3. However, that term did not significantly improve the quality of the fit, and the values for a3 were consistent with zero. We also considered other forms sometimes used for the melting of pure substances, such as that recommended by Yi- Jing et al. [1982], but they either did not fit as well over the full range of the data, or required more parameters that were typically poorly constrained. The best descriptions of the data were ultimately obtained with simple polynomial fits of the form T = a0 + a1(p − p0) + a2(p − p0)2 (2) (3) The p0 term was arbitrarily fixed at a convenient pressure so that the uncertainties more ac- curately reflected the fit over the pressure range of interest. The coefficients for the curves in Fig. 5 are given in Table 1. Table 1. Coefficients of fits to Eq. 3 for various curves in Fig. 5, along with the corresponding temperature ranges of the data. Conc. Regime T Range (K) 4.9% Ice Ih 4.9% Ice III 4.9% Ice V 9.9% Ice Ih 248 -- 270 248 -- 252 245 -- 254 249 -- 266.5 p0 (MPa) 0 209 350 0 a0 (K) a2 (K/MPa2) a1 (K/MPa) 270.1 ± 0.1 −0.074 ± 0.003 −(1.5 ± 0.1) × 10−4 247.2 ± 0.3 0.078 ± 0.011 −(2.9 ± 0.9) × 10−4 252.7 ± 0.3 0.089 ± 0.006 266.6 ± 0.3 −0.066 ± 0.009 −(1.65 ± 0.5) × 10−4 -- 4.3 Higher methanol concentrations The results for two additional samples, at 33.8 and 75 wt.%, are included in Fig. 6. For the 75 wt.% sample, the solid formed is presumably CH3OH·H2O, as shown in the phase diagram (Fig. 1). The coefficients for the fits are given in Table 2. Table 2. Coefficients of fits to Eq. 3 for the high-concentration curves in Fig. 6 Concentration Regime 33.8% 33.8% 75% Ice Ih Ice II p0 (MPa) 0 209 0 a0 (K) a1 (K/MPa) a2 (K/MPa2) 243.4 ± 0.6 −0.095 ± 0.016 −(9.7 ± 7.5) × 10−5 −(1.7 ± 1.5) × 10−4 221.6 ± 0.6 169.7 ± 0.4 −(2.7 ± 2.3) × 10−5 0.087 ± 0.023 0.098 ± 0.006 5 Discussion Methanol is a potentially important antifreeze. Specifically, we find that the freez- ing point depression due to the addition of methanol is about 3.1 K for a concentration of 4.9 wt.%, 5.8 K for 9.9 wt.%, and grows to about 31 K for a concentration of 33.8 wt.%. For the low concentrations of planetary relevance (< 10 wt.%), we find the freezing point depres- sion f z(x) (in Kelvin) due to the addition of methanol is roughly independent of pressure, at -- 8 -- Confidential manuscript submitted to JGR-Planets Figure 6. Liquidus temperatures for four different concentrations as a function of pressure. The fits are described in the text. In addition to the data shown in Fig. 5, the melting curve of pure methanol [Sun et al., 1988; Wurflinger and Landau, 1977; Gromnitskaya et al., 2004] is included for comparison. least for the two concentrations measured (4.9 wt.% and 9.9 wt.%) in the Ice-Ih regime, and is reasonably represented over that range as a function of percent methanol concentration x by f z(x) = (0.59 ± 0.01)x. (4) In contrast, a linear interpolation between the Ice-Ih and pure methanol curves, as was used in Deschamps et al. [2010], gives a freezing point depression that decreases with in- creasing pressure, and hence tends to underestimate the effect of adding methanol. Specif- ically, for a 4.9 wt.% concentration of methanol and a pressure of 200 MPa, the measured liquidus temperature based on Table 1 yields a freezing point depression of approximately 3.1 K, while a linear interpolation yields a smaller effect of only 1.6 K. Or, equivalently, that same 1.6 K freezing point depression at 200 MPa could be achieved with only a 2.7 wt.% concentration. 6 Conclusions In their modeling of Titan's primordial ocean, Deschamps et al. [2010] found that the inclusion of methanol reduced the solidification temperature, and hence reduced convection and heat transfer in the outer Ice Ih shell. Ultimately, this could lead to an end to crystal- lization and the maintenance of a subsurface ocean. The main contribution from the present experiments is to observe that methanol is an even more effective antifreeze than expected, so that a subsurface ocean might be maintained with an even lower methanol concentration than considered there. More generally, models of the interiors of Titan, Enceladus, and other bodies with subsurface oceans can benefit from more information about the thermodynamic behaviors of aqueous solutions at the relevant temperatures and pressures. For Titan and Enceladus, methanol is one reasonable impurity to include in such models. For other outer solar sys- tem bodies and even exoplanets, incorporation of methanol may similarly lead to a subsur- face ocean being maintained where a pure water ocean might completely freeze. Since many of the of thermodynamic and rheological properties of an ocean and icy shell are strongly -- 9 -- 160 180 200 220 240 260 280 0 50 100 150 200 250 300 350 400Ice IhIce IIIce VLiquidTemperature (K)Pressure (MPa)4.9%: Ice-Ih4.9%: Ice-III4.9%: Ice-V9.9%: Ice-Ih33.8%: Ice-Ih33.8%: Ice-IIPure Methanol75% MethanolConfidential manuscript submitted to JGR-Planets temperature-dependent, it may be worthwhile to include this freezing point depression in fu- ture modeling. -- 10 -- Confidential manuscript submitted to JGR-Planets Acknowledgments Financial support was provided by Lafayette College. Data used in the figures is available at https://data.mendeley.com/datasets/fwpf6t3bxn/1. The authors thank D.L. Hogenboom for assistance in building the pressure system. References Baland, R. M., T. V. Hoolst, M. Yseboodt, and O. Karatekin (2011), Titan's obliquity as ev- idence of a subsurface ocean?, Astron. & Astrophys., 530, 1 -- 6, doi:10.1051/0004-6361/ 201116578. Beghin, C., O. Randriamboarison, M. Hamelin, E. Karkoschka, C. Sotin, R. C. Whitten, J.-J. Berthelier, R. Grard, and F. Simoes (2012), Analytic theory of Titan's Schumann reso- nance: Constraints on ionospheric conductivity and buried water ocean, Icarus, 218(2), 1028 -- 1042, doi:10.1016/j.icarus.2012.02.005. Dartois, E., W. Schutte, T. R. Geballe, K. Demyk, P. Ehrenfreund, and L. d'Hendecourt (1999), Methanol: The second most abundant ice species towards the high-mass protostars RAFGL7009S and W 33A, Astron. Astrophys., 342(2), L32 -- L35. Davies, A. G., C. Sotin, D. L. Matson, J. Castillo-Rogez, T. V. Johnson, M. Choukroun, and K. H. Baines (2010), Atmospheric control of the cooling rate of impact melts and cry- olavas on Titan's surface, Icarus, 208(2), 887 -- 895, doi:10.1016/j.icarus.2010.02.025. Davies, A. G., C. Sotin, M. Choukroun, D. L. Matson, and T. V. Johnson (2016), Cryolava flow destabilization of crustal methane clathrate hydrate on Titan, Icarus, 274, 23 -- 32, doi: 10.1016/j.icarus.2016.02.046. Deschamps, F., O. Mousis, C. Sanchez-Valle, and J. I. Lunine (2010), The role of methanol in the crystallization of Titan's primordial ocean, Astrophys. J., 724(2), 887 -- 894, doi:10. 1088/0004-637X/724/2/887. Dougherty, M., K. Khurana, F. Neubauer, C. Russell, J. Saur, J. Leisner, and M. Burton (2006), Identification of a dynamic atmosphere at Enceladus with the Cassini magnetome- ter, Science, 311(5766), 1406 -- 1409, doi:10.1126/science.1120985. Dunaeva, A. N., D. V. Antsyshkin, and O. L. Kuskov (2010), Phase diagram of H2O: Ther- modynamic functions of the phase transitions of high-pressure ices, Sol Syst. Res., 44(3), 202 -- 222, doi:10.1134/S0038094610030044. Dunaeva, A. N., V. A. Kronrod, and O. L. Kuskov (2014), Models of titan with water-ice shell, rock-ice mantle, and constraints on the rock-iron component composition, Dokl. Earth Sci., 454(1), 89 -- 93, doi:10.1134/S1028334X14010188. Dunaeva, A. N., V. A. Kronrod, and O. L. Kuskov (2016), Physico-chemical models of the internal structure of partially differentiated Titan, Geochem. Int., 54(1), 27 -- 47, doi:10. 1134/S0016702916010043. 4932, doi:10.1063/1.1709255. Evans, L. F. (1967), Selective nucleation of high-pressure ices, J. Appl. Phys., 38(12), 4930 -- Fortes, A. D. (2006), The crystal structure of methanol monohydrate (CD3OD·D2O) at 160K from powder neutron diffraction, Chemical Physics Letters, 431(4), 283 -- 288, doi:10.1016/ j.cplett.2006.09.077. Fortes, A. D. (2012), Titan's internal structure and the evolutionary consequences, Planet. Space Sci., 60(1), 10 -- 17, doi:10.1016/j.pss.2011.04.010. Fortes, A. D., E. Suard, and K. S. Knight (2011), Negative Linear Compressibility and Mas- sive Anisotropic Thermal Expansion in Methanol Monohydrate, Science, 331(6018), 742 -- 746, doi:10.1126/science.1198640. Gromnitskaya, E. L., O. V. Stal'gorova, O. F. Yagafarov, V. V. Brazhkin, A. G. Lyapin, and S. V. Popova (2004), Ultrasonic study of the phase diagram of methanol, JETP Lett., 80(9), 597 -- 601, doi:10.1134/1.1851642. Haghighi, H., A. Chapoy, R. Burgess, S. Mazloum, and B. Tohidi (2009), Phase equilibria for petroleum reservoir fluids containing water and aqueous methanol solutions: Experimen- tal measurements and modelling using the CPA equation of state, Fluid Phase Equilib., -- 11 -- Confidential manuscript submitted to JGR-Planets 278(1-2), 109 -- 116, doi:10.1016/j.fluid.2009.01.009. Hansen, C. J., D. E. Shemansky, L. W. Esposito, A. I. F. Stewart, B. R. Lewis, J. E. Col- well, A. R. Hendrix, R. A. West, J. H. W. Jr., B. Teolis, and B. A. Magee (2011), The composition and structure of the Enceladus plume, Geophys. Res. Lett., 38, L11,202, doi: 10.1029/2011GL047415. Hemingway, D., F. Nimmo, H. Zebker, and L. Iess (2013), A rigid and weathered ice shell on Titan, Nature, 500(7464), 550 -- 552, doi:10.1038/nature12400. Hodyss, R., C. D. Parkinson, P. V. Johnson, J. V. Stern, J. D. Goguen, Y. L. Yung, and I. Kanik (2009), Methanol on Enceladus, Geophys. Res. Lett., 36, L17,103, doi:10.1029/ 2009GL039336. Hogenboom, D., J. Kargel, J. Ganasan, and L. Lee (1995), Magnesium sulfate-water to 400 MPa using a novel piezometer - densities, phase-equilibria, and planetological implica- tions, Icarus, 115(2), 258 -- 277, doi:10.1006/icar.1995.1096. Hogenboom, D., J. Kargel, G. Consolmagno, T. Holden, L. Lee, and M. Buyyounouski (1997), The ammonia-water system and the chemical differentiation of icy satellites, Icarus, 128(1), 171 -- 180, doi:10.1006/icar.1997.5705. Hsieh, W.-P., and F. Deschamps (2015), Thermal conductivity of H2O-CH3OH mixtures at high pressures: Implications for the dynamics of icy super-earths outer shells, J. Geophys. Res. Planets, 120(10), 1697 -- 1707, doi:10.1002/2015JE004883. Hsu, H.-W., F. Postberg, Y. Sekine, T. Shibuya, S. Kempf, M. Horanyi, A. Juhasz, N. Al- tobelli, K. Suzuki, Y. Masaki, T. Kuwatani, S. Tachibana, S. iti Sirono, G. Moragas- Klostermeyer, and R. Srama (2015), Ongoing hydrothermal activities within enceladus, Nature, 519(7542), 207 -- 210, doi:10.1038/nature14262. Iess, L., R. A. Jacobson, M. Ducci, D. J. Stevenson, J. I. Lunine, J. W. Armstrong, S. W. Asmar, P. Racioppa, N. J. Rappaport, and P. Tortora (2012), The tides of Titan, Science, 337(6093), 457 -- 459, doi:10.1126/science.1219631. Kargel, J. S. (1992), Ammonia-water volcanism on icy satellites: Phase relations at 1 atmo- sphere, Icarus, 100, 556 -- 574, doi:10.1016/0019-1035(92)90118-Q. Kechin, V. V. (1995), Thermodynamically Based Melting-Curve Equation, J. Phys.: Con- dens. Matter, 7(3), 531 -- 535, doi:10.1088/0953-8984/7/3/008. Lefevre, A., G. Tobie, G. Choblet, and O. Cadek (2014), Structure and dynamics of Titan's outer icy shell constrained from Cassini data, Icarus, 237, 16 -- 28, doi:10.1016/j.icarus. 2014.04.006. Lopes, R. M. C., K. L. Mitchell, E. R. Stofan, J. I. Lunine, R. Lorenz, F. Paganelli, R. L. Kirk, C. A. Wood, S. D. Wall, L. E. Robshaw, A. D. Fortes, C. D. Neish, J. Radebaugh, E. Reffet, S. J. Ostro, C. Elachi, M. D. Allison, Y. Anderson, R. Boehmer, G. Boubin, P. Callahan, P. Encrenaz, E. Flamini, G. Francescetti, Y. Gim, G. Hamilton, S. Hens- ley, M. A. Janssen, W. T. K. Johnson, K. Kelleher, D. O. Muhleman, G. Ori, R. Orosei, G. Picardi, F. Posa, L. E. Roth, R. Seu, S. Shaffer, L. A. Soderblom, B. Stiles, S. Ve- trella, R. D. West, L. Wye, and H. A. Zebker (2007), Cryovolcanic features on Titan's surface as revealed by the Cassini Titan radar mapper, Icarus, 186(2), 395 -- 412, doi: 10.1016/j.icarus.2006.09.006. Lopes, R. M. C., R. L. Kirk, K. L. Mitchell, A. LeGall, J. W. Barnes, A. Hayes, J. Kargel, L. Wye, J. Radebaugh, E. R. Stofan, M. A. Janssen, C. D. Neish, S. D. Wall, C. A. Wood, J. I. Lunine, and M. J. Malaska (2013), Cryovolcanism on Titan: New results from Cassini RADAR and VIMS, J. Geophys. Res. Planets, 118(3), 416 -- 435, doi:10.1002/jgre.20062. McLaurin, G., K. Shin, S. Alavi, and J. A. Ripmeester (2014), Antifreezes act as catalysts for methane hydrate formation from ice, Angew. Chem., 53(39), 10,429 -- 10,433, doi:10.1002/ anie.201403638. Merlin, F., E. Quirico, M. A. Barucci, and C. de Bergh (2012), Methanol ice on the surface of minor bodies in the solar system, Astron. Astrophys., 544, A20, doi:10.1051/0004-6361/ 201219181. Miller, G. A., and D. K. Carpenter (1964), Solid-liquid phase diagram of the system methanol-water., J. Chem. & Eng. Data, 9(3), 371 -- 373, doi:10.1021/je60022a017. -- 12 -- Confidential manuscript submitted to JGR-Planets Moore, J. M., and R. T. Pappalardo (2011), Titan: An exogenic world?, Icarus, 212(2), 790 -- 806, doi:10.1016/j.icarus.2011.01.019. Neveu, M., S. J. Desch, and J. C. Castillo-Rogez (2017), Aqueous geochemistry in icy world interiors: Equilibrium fluid, rock, and gas compositions, and fate of antifreezes and ra- dionuclides, Geochimica Et Cosmochimica Acta, 212, 324 -- 371, doi:10.1016/j.gca.2017. 06.023. Nimmo, F., and B. G. Bills (2010), Shell thickness variations and the long-wavelength topog- raphy of Titan, Icarus, 208(2), 896 -- 904, doi:10.1016/j.icarus.2010.02.020. Nimmo, F., and R. T. Pappalardo (2016), Ocean worlds in the outer solar system, J. Geophys. Res. Planets, 121(8), 1378 -- 1399, doi:10.1002/2016JE005081. Ott, J., J. Goates, and B. Waite (1979), (solid + liquid) phase-equilibria and solid-hydrate formation in water + methyl, + ethyl, + isopropyl, and + tertiary butyl alcohols, J. Chem. Thermodyn., 11(8), 739 -- 746, doi:10.1016/0021-9614(79)90005-3. Porco, C., P. Helfenstein, P. Thomas, A. Ingersoll, J. Wisdom, R. West, G. Neukum, T. Denk, R. Wagner, T. Roatsch, S. Kieffer, E. Turtle, A. McEwen, T. Johnson, J. Rathbun, J. Vev- erka, D. Wilson, J. Perry, J. Spitale, A. Brahic, J. Burns, A. DelGenio, L. Dones, C. Mur- ray, and S. Squyres (2006), Cassini observes the active south pole of enceladus, Science, 311(5766), 1393 -- 1401, doi:10.1126/science.1123013. Postberg, F., J. Schmidt, J. Hillier, S. Kempf, and R. Srama (2011), A salt-water reservoir as the source of a compositionally stratified plume on enceladus, Nature, 474(7353), 620 -- 622, doi:10.1038/nature10175. Sekine, Y., H. Genda, Y. Muto, S. Stigita, T. Kadono, and T. Matsui (2014), Impact chem- istry of methanol: Implications for volatile evolution on icy satellites and dwarf planets, and cometary delivery to the Moon, Icarus, 243, 39 -- 47, doi:10.1016/j.icarus.2014.08.034. Shimonishi, T., E. Dartois, T. Onaka, and F. Boulanger (2016), VLT/ISAAC infrared spec- troscopy of embedded high-mass YSOs in the Large Magellanic Cloud: Methanol and the 3.47 mu m band, Astron. Astrophys., 585, A107, doi:10.1051/0004-6361/201526559. Simon, F., and G. Glatzel (1929), Bemerkungen zur schmelzdruckkurve, Z. Anorg. Allg. Chem., 178(1), 309 -- 316. Solomonidou, A., M. Hirtzig, A. Coustenis, E. Bratsolis, S. L. Mouelic, S. Rodriguez, K. Stephan, P. Drossart, C. Sotin, R. Jaumann, R. H. Brown, K. Kyriakopoulos, R. M. C. Lopes, G. Bampasidis, K. Stamatelopoulou-Seymour, and X. Moussas (2014), Surface albedo spectral properties of geologically interesting areas on Titan, J. Geophys. Res. Planets, 119(8), 1729 -- 1747, doi:10.1002/2014JE004634. Sun, T., J. Schouten, N. Trappeniers, and S. Biswas (1988), Accurate measurement of the melting line of methanol and ethanol at pressures up to 270 MPa, Berichte Der Bunsen- Gesellschaft-Physical Chemistry Chemical Physics, 92(5), 652 -- 655, doi:10.1002/bbpc. 198800153. Takaizumi, K., and T. Wakabayashi (1997), The freezing process in methanol-, ethanol-, and propanol-water systems as revealed by differential scanning calorimetry, J. Solution Chem., 26(10), 927 -- 939, doi:10.1007/BF02768051. Thomas, P. C., R. Tajeddine, M. S. Tiscareno, J. A. Burns, J. Joseph, T. J. Loredo, P. Helfen- stein, and C. Porco (2016), Enceladus's measured physical libration requires a global sub- surface ocean, Icarus, 264, 37 -- 47, doi:10.1016/j.icarus.2015.08.037. Tobie, G., D. Gautier, and F. Hersant (2012), Titan's bulk composition constrained by Cassini-Huygens: Implication for internal outgassing, Astrophys. J., 752(2), 125, doi: 10.1088/0004-637X/752/2/125. Vuillard, G., and M. Sanchez (1961), Vitrification et cristallisation dans le systeme binaire eau-methanol, Bull. Soc. Chim. France, pp. 1877 -- 1880. Wagner, W., T. Riethmann, R. Feistel, and A. H. Harvey (2011), New Equations for the Sub- limation Pressure and Melting Pressure of H(2)O Ice Ih, J. Phys. Chem. Ref. Data, 40(4), doi:10.1063/1.3657937. Waite, J. H., W. S. Lewis, B. A. Magee, J. I. Lunine, W. B. McKinnon, C. R. Glein, O. Mousis, D. T. Young, T. Brockwell, J. Westlake, M.-J. Nguyen, B. D. Teolis, B. Nie- -- 13 -- Confidential manuscript submitted to JGR-Planets mann, R. L. McNutt, M. Perry, and W.-H. Ip (2009), Liquid water on Enceladus from observations of ammonia and Ar-40 in the plume, Nature, 460(7254), 487 -- 490, doi: 10.1038/nature08153. Wurflinger, A., and R. Landau (1977), Differential thermal analysis at high pressures -- VII: Phase behavior of solid methanol up to 3 kbar, J. Phys. Chem. Solids, 38, 811 -- 814, doi: 10.1016/0022-3697(77)90115-9. Yi-Jing, D., Y. Tsu-Tung, and C. Li-Rong (1982), On The Equation for the Pressure- Dependence of Melting Temperature, J. Phys. D: Appl. Phys., 15(2), 263 -- 265, doi: https://doi.org/10.1088/0022-3727/15/2/011. Zhong, F., K. L. Mitchell, C. C. Hays, M. Choukroun, M. Barmatz, and J. S. Kargel (2009), The rheology of cryovolcanic slurries: Motivation and phenomenology of methanol-water slurries with implications for Titan, Icarus, 202(2), 607 -- 619, doi:10.1016/j.icarus.2009. 03.015. -- 14 --
0910.0468	1	0910	2009-10-02T20:01:39	Forming Jupiter, Saturn, Uranus and Neptune in Few Million Years by Core Accretion	[ "astro-ph.EP" ]	Giant planet formation process is still not completely understood. The current most accepted paradigm, the core instability model, explains several observed properties of the solar system's giant planets but, to date, has faced difficulties to account for a formation time shorter than the observational estimates of protoplanetary disks' lifetimes, especially for the cases of Uranus and Neptune. In the context of this model, and considering a recently proposed primordial solar system orbital structure, we performed numerical calculations of giant planet formation. Our results show that if accreted planetesimals follow a size distribution in which most of the mass lies in 30-100 meter sized bodies, Jupiter, Saturn, Uranus and Neptune may have formed according to the nucleated instability scenario. The formation of each planet occurs within the time constraints and they end up with core masses in good agreement with present estimations.	astro-ph.EP	astro-ph	Forming Jupiter, Saturn, Uranus and Neptune in Few Million Years by Core Accretion Omar G. Benvenuto1,2,3, Andrea Fortier1,3,4, Adri´an Brunini1,3,5 1.- Facultad de Ciencias Astron´omicas y Geof´ısicas, Universidad Nacional de La Plata (UNLP), 1900 La Plata, Argentina. 2.- Member of the Carrera del Investigador Cient´ıfico, Comisi´on de Investigaciones Cient´ıficas de la Provincia de Buenos Aires (CIC), Argentina. 3.- Instituto de Astrof´ısica de La Plata (IALP), CCT-Consejo Nacional de Investigaciones Cient´ıficas y T´ecnicas (CONICET), UNLP, Argentina. 4.- Fellow of CONICET. 5.- Member of the Carrera del Investigador Cient´ıfico, CONICET. Abstract Giant planet formation process is still not completely understood. The current most accepted paradigm, the core instability model, explains several observed properties of the solar system's giant planets but, to date, has faced difficulties to account for a formation time shorter than the observational estimates of protoplanetary disks' lifetimes, especially for the cases of Uranus and Neptune. In the context of this model, and considering a recently proposed primordial solar system orbital structure, we performed numerical calculations of giant planet formation. Our results show that if accreted planetesimals follow a size distribution in which most of the mass lies in 30-100 meter sized bodies, Jupiter, Saturn, Uranus and Neptune may have formed according to the nucleated instability scenario. The formation of each planet occurs within the time constraints and they end up with core masses in good agreement with present estimations. key words: Planetary formation; Accretion; Planetesimals. Correspondence should be directed to: Omar G. Benvenuto, Facultad de Ciencias Astron´omicas y Geof´ısicas de La Plata, Paseo del Bosque S/N (1900) La Plata, Argentina. E-mail: oben- [email protected] 1 Terrestrial planets are widely believed to form in the inner solar nebula through the accretion of a population of centimeter to kilometer sized planetesimals orbiting the early Sun (Wetherill 1990). The core-accretion model states that this process also occurred in the outer solar nebula (Mizuno 1980; Pollack et al. 1996). When a solid embryo has grown to few thousandths of the Earth mass, a gaseous envelope in hydrostatic equilibrium begins to bind to it. This equilibrium is achieved by the balance between the heating due to the energy released by the incoming planetesimals and the gravity of the core. As the core grows and planetesimals exhaust, the gaseous envelope eventually can no longer be supported and contracts. This leads to the last stage of giant planet formation, a short period during which the protoplanet quickly gains the bulk of its total mass (Pollack et al. 1996). On the other hand, the first stage of the process is dominated by the formation of the core. The time scale of core growth is strongly dependent on several factors, the initial disk surface density of solids in the giant planet region, Σ, and the radius of accreted planetesimals being among the most important ones. The first calculations that self-consistently considered both the accretion of gas as well as planetesimals showed that, even prescribing an unrealistically fast core accretion rate, with a surface density of solids of Σ = 10 g cm−2 at 5.2 AU, Jupiter takes 6 million years (My) to complete its formation (Pollack et al. 1996, updated by Hubickyj et al. 2005)1, and Saturn more than 10 My (Pollack et al. 1996). However, according to observations, the lifetime of circumstellar disks at the Jupiter-Saturn region is between 3-8 My, while beyond 10 AU they are expected to last somewhat longer (Hillenbrand 2005). This fact limits the formation time of gaseous planets. Increasing the local density speeds up the formation process, but at the cost of building too massive cores: interior models of Jupiter that agree with all observational constraints show that its core mass is at most 12 M⊕ (Guillot 2005), whereas more recent calculations give core masses of 14 - 18 M⊕ (Militzer et al. 2008). In the case of Saturn, its core mass is between 9-22 M⊕ (Guillot 2005). Recent improved core accretion calculations adopt more adequate solid accretion rates for the cores (Fortier et al. 2007, 2009), or take into account local density patterns in the disk (Klahr & Bodenheimer 2006), but nevertheless they have not solved the time scale problem. Other recent models showed that the time scale for envelope growth depends on the assumed dust grain opacity (Ikoma et al. 2000). Grains entering the protoplanetary envelope may coagulate and settle down quickly. As a consequence, the correct opacity may be far lower than usually believed (Podolak 2003), easing an early envelope contraction that allows the formation of Jupiter in less than 3 My (Hubickyj et al. 2005). However, the time scale to form the core, one of the main problems of the core accretion model, is not substantially modified and, to date, no self-consistent calculations of the opacity in the protoplanet's atmosphere have been performed (Movshovitz & 1This result corresponds to the simulation 10H∞ of Hubickyj et al. (2005) computed with full interstellar opacities. 2 Podolak 2008). Improved solid surface density calculations (Dodson - Robinson et al. 2008) allow to compute formation models of Saturn consistent with circumstellar disk lifetimes. However, their simulations end with a total mass of heavy elements higher than current estimations for Saturn. In all core accretion calculations performed so far, a single sized planetesimal population has been assumed. However, remarkably, the accretion rate of solids is strongly dependent on the planetesimal size. Therefore, self-consistent numerical simulations, in which the core is formed through accretion of planetesimals following a size distribution are in order. At early stages, larger planetesimals grow faster than smaller ones, resulting in a runaway growth of the largest (Greenberg et al. 1978; Wetherill & Stewart 1989; Kokubo & Ida 1996). Runaway bodies then become protoplanetary embryos that grow by the accretion of small planetesimals. When the mass of a protoplanet exceeds a critical value far lower than the Earth mass, runaway growth slows down (Ida & Makino 1993) and switches to oligarchic growth (Kokubo & Ida 1998), in which similar-sized protoplanets dominate the planetesimal system. During the oligarchic regime only protoplanets grow as there is no substantial accretion among planetesimals. Remarkably, solid bodies in the swarm are not of uniform size but span over a wide interval ranging from millimeter to kilometer. Numerical simulations show that the mass distribution of these residual planetesimals may be represented by a single or piecewise power law dnc(m)/dm ∝ n(m) ∝ m−α, where nc(m) is the number of bodies larger than a given mass m and n(m) is the number of bodies in a linear mass bin. For a constant value of α the mass in the interval is R n m dm ∝ m2−α. Depending on the value of α the mass in the interval would be mostly contained in small (α > 2) or large (α < 2) planetesimals. Wetherill and Stewart (1993) studied the evolution of the planetesimal system considering an initial population of planetesimals whose radius is ∼ 10 km that evolved only through collisions and fragmentation. They found that planetesimal size distribution relaxes to a piecewise power law: a population of small planetesimals due to fragmentation (α ∼ 1.7) and a population of large planetesimals that follow an accretive regime (α ∼ 2.5) which, in turn, contained most of the mass of the system. Kokubo and Ida (2000) studied through N-body simulations the evolution of planetesimal size and they found that, in the oligarchic regime, large planetesimals follow a continuous power law distribution with 2 < α < 3. However, these studies do not take into account the effects due to magneto-rotational instability turbulence, which establish the predominance of an erosive rather than an accretive regime for planetesimal growth, nor the existence of dead zones that could favor accretion among planetesimals (Ida et al. 2008). On the other hand, recent laboratory experiments show that reaccumulation of fragmentation debris can lead to the formation of planetesimals (Teiser and Wurm, 2009). These effects could have a substantial impact on the accretion process of planetesimals, specially in the small planetesimal tail of the distribution. Evidently, the primordial planetesimal size distribution in the protoplanetary 3 disk is still an open problem. In our study, for the sake of simplicity we shall consider a continuous power law distribution, adopting for the exponent the value α = 2.5. In view of the present status of the theory of planetesimal growth, we consider plausible such a distribution. Another important point to take into account is that the primordial configuration of the outer solar system was probably not the one we know today. Numerical simulations (Tsiganis et al. 2005) assuming an originally compact planetary system (the Nice model) were able to reproduce the main observational constraints regarding the present orbital structure of the outer solar system, that was achieved after the migration of the fully formed giant planets to their present location. In ∼ 50% of these simulations Uranus and Neptune switch places, explaining naturally why Neptune's mass is larger than that of Uranus. The Nice model implies a solid surface density profile Σ ∝ a−2.168, where a is the distance from the Sun (Desch 2007). In our model, in order to be consistent with the Nice model, we placed Jupiter, Saturn, Uranus and Neptune's embryos at a = 5.5, 8.3, 14, and 11 AU respectively. We calculate the in situ formation of these planets. This assumption, although a simplification of the model, is compatible with recent studies of the formation of planetary systems. Thommes et al. (2008) found that solar system analogs come out if the gas giants do not undergo significant migration and remain in nearly circular orbits. In all our simulations, the initial core and gaseous envelope masses were set to ∼ 10−2 M⊕ and ∼ 10−10 M⊕ respectively. The equations governing the evolution of the gaseous envelope were solved coupled self-consistently to the planetesimal accretion rate, employing a standard finite difference (Henyey) method and detailed constitutive physics as described in our previous work (Fortier et al. 2007, 2009). We adopted the equation of state of Saumon et al. (1995), the grain opacities of Pollack et al. (1985) and, for temperatures above 103 K, the molecular opacities of Alexander & Ferguson (1994). The accretion of planetesimals modify the solid surface density around the planet which, to- gether with their relative velocities, control the solids accretion rate. The relative velocities are calculated assuming the equilibrium between the gravitational perturbations due to the embryo and the dissipation due to gas drag. Since the oligarchic regime dominates the time scale for the formation of the core, we prescribed this accretion rate during the whole simulation (Fortier et al. 2007, 2009). To approximately mimic the effects due to neighboring planets (Hubickyj et al. 2005), after the planet reaches the first maximum in the solid accretion rate, we inhibited a further growth of the local planetesimal surface density, i.e., we no longer allow planetesimals to refill the planetary feeding zone. We assumed that all captured planetesimals sink onto the solid core and that their mass density is equal to 1.5 g cm−3. We approached the continuous planetesimal size distribution with a discrete one of nine sizes from 30 m to 100 km of radius, geometrically evenly spaced. Considering a non homogeneous population is very important as planetesimal dynamics is affected by nebular drag. Small planetesimals suffer a stronger damping of random velocities, 4 increasing this way the accretion rate (Safronov 1969); but they are also subject to faster orbital decay, reducing the planetesimal surface density in the neighborhood of the protoplanet. In the nebular model we are using, the accretion time scale of planetesimals whose radius is r ≥ 30 m is shorter than the corresponding time scale of orbital decay (Chambers 2006; Brunini & Benvenuto 2008). This situation reverts at r ∼ 20 m. Therefore, the giant planets of the solar system should have formed essentially through accretion of planetesimals larger than r ∼ 30 m, as smaller plan- etesimals could not have been efficiently accreted. It is worth mentioning that a nebular model based on the present planetary masses, like the minimum mass solar nebula, does not account for the primordial solid mass comprised in the planetesimals that were not accreted by the planets. Then, the estimations of the nebular surface density adopted in this article only take into account the mass contained in planetesimals that have chances to be accreted. Fig. 1 shows the results for the formation of Jupiter. Allowing for a distribution of planetesimal sizes where the most abundant ones are those of r = 30 − 80 m, Jupiter grows much faster than considering a typical single size population (e.g. r = 10 km). Note that the model corresponding to Σ = 11 g cm−2 forms timely and achieves a core mass compatible with currently accepted estimations. The key ingredient that speeds up planetary growth is the dependence of the cross- section of the planet as a function of the planetesimal size. This, in turn, is determined by the drag force suffered by planetesimals after entering the planetary envelope, which is proportional to the reciprocal of the planetesimal radius (Podolak et al. 1988). If there was no gas bound to the planet, the core would grow only by inelastic collisions with planetesimals. For this process, the cross-section is given by the projected area of the planet enhanced by gravitational focusing. Gas drag modifies planetesimal trajectories, eventually trapping them in the planet, greatly enlarging the cross-section (Pollack et al. 1996; Fortier et al. 2007, 2009). Let us consider the conditions for the formation of the giant planets beyond Jupiter's orbit. Regarding Uranus and Neptune, models consistent with available observations predict a gaseous mass fraction far lower than those of Jupiter and Saturn. The upper limits for the gas content of Uranus and Neptune are of 5.0 M⊕ and 4.7 M⊕ respectively (Podolak et al. 2000), while the absolute lower limit for both of them is of 0.5 M⊕ (Guillot 2005). Fig. 2 shows the results corresponding to the four giant planets of the solar system computed separately, together with plausible values for their core masses (Guillot 2005; Podolak et al. 2000). Here the four giant planets are formed within the estimated lifetime of protoplanetary disks. Other simulations (Alibert et al. 2005) were able to form Jupiter and Saturn in less than 2.5 My but with Saturn undergoing a migration in conflict with the Nice model (Tsiganis et al. 2005) and neglecting the oligarchic regime for the growth of the core. On the other hand, Desch (2007) estimated the growth time for the cores of the four giant planets in the framework of the Nice model. Although he did not consider the presence of the planets' atmosphere, his results support 5 the possibility of the formation of all the giant planets of the solar system in less than 10 My, provided that small planetesimals were the building blocks of the cores. Regarding the mass of the core, as can be seen from Fig. 2, Uranus and Saturn's fall inside the error bar while Jupiter is compatible with the top of its error bar. Neptune's model presents a core smaller than current estimations. Nevertheless, we should remark that Neptune's observations are by far the more inaccurate ones (Podolak et al. 2000). If we consider that the size distribution of planetesimals is represented by 7 species where the minimum radius is r = 100 m, instead of 9 species starting with r = 30 m, the formation time of the four giant planets is lengthened, being the longest one that of Uranus (∼ 13 My). Moreover, if we consider that most of the planetesimal mass lies in kilometer-sized bodies, the formation time of the ice giants results much longer than the estimated lifetime of the protoplanetary disk. In order to evaluate the effects of the above mentioned inhibition on the solids accretion rate, we also computed the growth of the four giant planets without such effect. The results are presented in Fig. 3. In this case, planetary cores are more massive: Jupiter, Saturn, Neptune and Uranus form cores of 20, 19, 13 and 11 M⊕ respectively. The only core mass outside, but near its error bar is that of Jupiter. However, in view of the disagreement between different research groups on the actual value of Jupiter's core mass, we judge our result as acceptable. Also, note that imposing no inhibition makes the whole planetary formation process even somewhat faster. In this study we performed calculations of giant planet formation in the framework of the nucleated instability model, taking into account the oligarchic regime for the growth of the core and a primordial configuration of the solar system according to the Nice model. Also, we considered a size distribution for the accreted planetesimals in which most of the mass lies in the smallest planetesimals of the population. This study shows that the formation of the four giant planets of the solar system is compatible with the observational constraints imposed by dust disk lifetimes. We found that to account for the formation of Jupiter, Saturn, Uranus and Neptune in less than 10 My most of the mass of accreted planetesimals has to be in planetesimals whose radius is ∼< 100 m. We should remark that if we assume a power law with α < 2, planetary embryos would grow on a too long timescale. In this sense, we have chosen particularly favorable conditions for the occurrence of planetary formation within the currently accepted constraints (timescales and core masses). Hopefully, future investigation on the theory of planetesimal formation will provide further insights on the primordial size distribution of planetesimals in the protoplanetary disk. Acknowledgments The authors thank R.E. Martinez and R.H. Viturro for the technical support. References 6 Alexander, D. R., Ferguson, J. W. 1994. Low-temperature Rosseland opacities. Astrophysical Journal 437, 879-891. Alibert, Y., Mousis, O., Mordasini, C., Benz, W. 2005. New jupiter and saturn formation models meet observations. Astrophysical Journal 626, L57-L60. Brunini, A., Benvenuto, O. G. 2008. On oligarchic growth of planets in protoplanetary disks. Icarus 194, 800-810. Chambers, J. 2006. A semi-analytic model for oligarchic growth. Icarus 180, 496-513. Desch, S. J. 2007. Mass distribution and planet formation in the solar nebula. Astrophysical Journal 671, 878-893. Dodson-Robinson, S. E., Bodenheimer, P., Laughlin, G., Willacy, K., Turner, N. J., Beichman, C. A. 2008. Saturn Forms by Core Accretion in 3.4 Myr. Astrophysical Journal Letters, 688, L99 Fortier, A., Benvenuto, O. G., Brunini, A. 2007. Oligarchic planetesimal accretion and giant planet formation. Astronomy and Astrophysics 473, 311-322. Fortier, A., Benvenuto, O. G., Brunini, A. 2009. Oligarchic planetesimal accretion and giant planet formation II. Astronomy and Astrophysics, in press (DOI: 10.1051/0004-6361/200811367). Greenberg, R., Hartmann, W. K., Chapman, C. R., Wacker, J. F. 1978. Planetesimals to planets - Numerical simulation of collisional evolution. Icarus 35, 1-26. Guillot, T. 2005. The interiors of giant planets: Models and outstanding questions. Annual Review of Earth and Planetary Sciences 33, 493-530. Hillenbrand, L. A. 2005. Observational constraints on dust disk lifetimes: implications for planet formation. ArXiv Astrophysics e-prints arXiv:astro-ph/0511083. Hubickyj, O., Bodenheimer, P., Lissauer, J. J. 2005. Accretion of the gaseous envelope of Jupiter around a 5 - 10 Earth-mass core. Icarus 179, 415-431. Ida, S., Makino, J. 1993. Scattering of planetesimals by a protoplanet - Slowing down of runaway growth. Icarus 106, 210-227. Ida, S., Guillot, T., Morbidelli, A. 2008. Accretion and Destruction of Planetesimals in Turbulent Disks. Astrophysical Journal 686, 1292-1301. Ikoma, M., Nakazawa, K., Emori, H. 2000. Formation of giant planets: dependences on core accretion rate and grain opacity. Astrophysical Journal 537, 1013-1025. Klahr, H., Bodenheimer, P. 2006. Formation of giant planets by concurrent accretion of solids and gas inside an anticyclonic vortex. Astrophysical Journal 639, 432-440. Kokubo, E., Ida, S. 1996. On runaway growth of planetesimals. Icarus 123, 180-191. Kokubo, E., Ida, S. 1998. Oligarchic growth of protoplanets. Icarus 131, 171-178. Kokubo, E., Ida, S. 2000. Formation of protoplanets from planetesimals in the solar nebula. Icarus 143, 15-27. 7 Militzer, B., Hubbard, W. B., Vorberger, J., Tamblyn, I., Bonev, S. A. 2008, Astrophysical Journal Letters, 688, L45 Mizuno, H. 1980. Formation of the giant planets. Progress of Theoretical Physics 64, 544-557. Movshovitz, N., Podolak, M. 2008. The opacity of grains in protoplanetary atmospheres. Icarus 194, 368-378. Podolak, M. 2003. The contribution of small grains to the opacity of protoplanetary atmospheres. Icarus 165, 428-437. Podolak, M., Pollack, J. B., Reynolds, R. T. 1988. Interactions of planetesimals with protoplane- tary atmospheres. Icarus 73, 163-179. Podolak, M., Podolak, J. I., Marley, M. S. 2000. Further investigations of random models of Uranus and Neptune. Planetary and Space Science 48, 143-151. Pollack, J. B., McKay, C. P., Christofferson, B. M. 1985. A calculation of the Rosseland mean opacity of dust grains in primordial solar system nebulae. Icarus 64, 471-492. Pollack, J. B., Hubickyj, O., Bodenheimer, P., Lissauer, J. J., Podolak, M., Greenzweig, Y. 1996. Formation of the giant planets by concurrent accretion of solids and gas. Icarus 124, 62-85. Safronov, V.S. 1969. Evolution of of the protoplanetary cloud and formation of the earth and planets. Nauka, Moscow. Saumon, D., Chabrier, G., van Horn, H. M. 1995. An Equation of State for Low-Mass Stars and Giant Planets. Astrophysical Journal Supplement Series 99, 713. Teiser, J., Wurm, G. 2009. High-velocity dust collisions: forming planetesimals in a fragmentation cascade with final accretion. Monthly Notices of the Royal Astronomical Society 393, 1584-1594. Thommes, E. W., Matsumura, S., Rasio, F. A. 2008. Gas disks to gas giants: Simulating the birth of planetary systems. Science 321, 814-817. Tsiganis, K., Gomes, R., Morbidelli, A., Levison, H. F. 2005. Origin of the orbital architecture of the giant planets of the Solar System. Nature 435, 459-461. Wetherill, G. W. 1990. Formation of the earth. Annual Review of Earth and Planetary Sciences 18, 205-256. Wetherill, G. W., Stewart, G. R. 1989. Accumulation of a swarm of small planetesimals. Icarus 77, 330-357. Wetherill, G. W., Stewart, G. R. 1993. Formation of planetary embryos - Effects of fragmentation, low relative velocity, and independent variation of eccentricity and inclination Icarus 106, 190-209. 8 Figure 1: The growth of Jupiter for different surface densities. Thick solid (dashed) lines depict the growth of the total (core) mass for models computed with a size distribution for the accreted planetesimals. Thin lines stand for the case of a single sized population of planetesimals with r = 10 km. The error bars depict the currently accepted values for Jupiter's core mass. G05 denotes the values given by Guillot (2005) while M08 stands for the results presented by Militzer et al. (2008). Σ represents the assumed values for the initial disk surface density of solids in the Jupiter region. Note that the formation is a two step process. First, the core grows by planetesimal accretion, whose gravitational energy release inhibits the gaseous envelope from undergoing a premature contraction. When the neighborhood of the protoplanet is depleted from planetesimals, the core mass gets its asymptotic value and a runaway gas accretion is established. Quickly, the planet reaches its final mass (318 M⊕). Clearly, the considered size distribution of planetesimals yields a formation time scale much shorter than considering a single planetesimal size of r = 10 km. In particular, with a local surface density of 11 g cm−2, we obtain an acceptable core mass value and a comfortably short formation time. Therefore we assume that this result scales the density throughout the entire protoplanetary disk. 9 Figure 2: The growth of the four giant planets. Solid (dashed) lines depict the growth of the total (core) mass for models computed with a size distribution for the accreted planetesimal. The values of Σ correspond to the initial surface density of solids at the position of each embryo. The error bars depict the currently accepted values for the core masses of each planet. For the case of Jupiter, as in Fig. 1, the lower error bar is that of Guillot (2005), whereas the upper one corresponds to the results given by Militzer et al. (2008). Simulations were stopped when the final mass of each planet was reached (approximately 318, 95, 14, and 17 M⊕ for Jupiter, Saturn, Uranus and Neptune respectively). The formation times are: Jupiter 0.44 My, Saturn 1.4 My, Neptune 2.5 My and Uranus 4.75 My. 10 Figure 3: Same as Fig. 2 but without imposing any inhibition on the solid mass accretion rate. 11
1501.00716	1	1501	2015-01-04T20:26:35	COSIMA-Rosetta calibration for in-situ characterization of 67P/Churyumov-Gerasimenko cometary inorganic compounds	[ "astro-ph.EP" ]	COSIMA (COmetary Secondary Ion Mass Analyser) is a time-of-flight secondary ion mass spectrometer (TOF-SIMS) on board the Rosetta space mission. COSIMA has been designed to measure the composition of cometary dust grains. It has a mass resolution m/{\Delta}m of 1400 at mass 100 u, thus enabling the discrimination of inorganic mass peaks from organic ones in the mass spectra. We have evaluated the identification capabilities of the reference model of COSIMA for inorganic compounds using a suite of terrestrial minerals that are relevant for cometary science. Ground calibration demonstrated that the performances of the flight model were similar to that of the reference model. The list of minerals used in this study was chosen based on the mineralogy of meteorites, interplanetary dust particles and Stardust samples. It contains anhydrous and hydrous ferromagnesian silicates, refractory silicates and oxides (present in meteoritic Ca-Al-rich inclusions), carbonates, and Fe-Ni sulfides. From the analyses of these minerals, we have calculated relative sensitivity factors for a suite of major and minor elements in order to provide a basis for element quantification for the possible identification of major mineral classes present in the cometary grains.	astro-ph.EP	astro-ph	COSIMA-‐Rosetta calibration for in-‐situ characterization of 67P/Churyumov-‐ Gerasimenko cometary inorganic compounds. Harald Krüger1, Thomas Stephan2, Cécile Engrand3, Christelle Briois4, Sandra Siljeström5, Sihane Merouane1, Donia Baklouti6, Henning Fischer1, Nicolas Fray7, Klaus Hornung8, Harry Lehto9, François-‐Régis Orthous-‐Daunay10, Jouni Rynö11, Rita Schulz12, Johan Silen11, Laurent Thirkell4, Mario Trieloff13, Martin Hilchenbach1. 1Max-‐Planck-‐Institut für Sonnensystemforschung, Justus-‐von-‐Liebig-‐Weg 3, 37077 Göttingen, Germany 2Department of the Geophysical Sciences, The University of Chicago, 5734 S Ellis Ave, Chicago, IL 60637, USA 3Centre de Sciences Nucléaires et de Sciences de la Matière, CNRS/IN2P3-‐Univ. Paris Sud -‐ UMR8609, Batiment 104, 91405 Orsay campus, France. 4Laboratoire de Physique et Chimie de l'Environnement et de l'Espace (LPC2E), CNRS/ Université d’Orléans, 45071 Orléans, France 5Department of Chemistry, Materials and Surfaces, SP Technical Research Institute of Sweden, Box 857, 50115 Borås, Sweden, 6Institut d'Astrophysique Spatiale, CNRS / Université Paris Sud, Orsay Cedex, France 7LISA, UMR CNRS 7583, Université Paris Est Créteil et Université Paris Diderot, Institut Pierre Simon Laplace, France 8Universität der Bundeswehr LRT-‐7, Werner Heisenberg Weg 39, 85577 Neubiberg, Germany 9University of Turku, Department of Physics and Astronomy, Tuorla Observatory Väisäläntie 20, 21500 Piikkiö, Finland 10Institut de Planétologie et d’Astrophysique de Grenoble, 414, Rue de la Piscine, Domaine Universitaire, 38400 St-‐Martin d’Hères, France 11Finnish Meteorological Institute, Observation services, Erik Palménin aukio 1, FI-‐ 00560 Helsinki, Finland 12ESA – ESTEC, Postbus 299, 2200AG Noordwijk, The Netherlands 13Institut für Geowissenschaften der Universität Heidelberg, Im Neuenheimer Feld 234-‐ 236, 69120 Heidelberg, Germany Corresponding author : [email protected] Manuscript pages: 19 Figures: 3 Tables: 4 1 Abstract Keywords: Corresponding author: Cecile Engrand CSNSM CNRS/IN2P3-‐Univ. Paris Sud Batiment 104 91405 Orsay Campus [email protected] Tel : +33 1 69 15 52 95 COSIMA (COmetary Secondary Ion Mass Analyser) is a time-‐of-‐flight secondary ion mass spectrometer (TOF-‐SIMS) on board the Rosetta space mission. COSIMA has been designed to measure the composition of cometary dust grains. It has a mass resolution m/Δm of 1400 at mass 100 u, thus enabling the discrimination of inorganic mass peaks from organic ones in the mass spectra. We have evaluated the identification capabilities of the reference model of COSIMA for inorganic compounds using a suite of terrestrial minerals that are relevant for cometary science. Ground calibration demonstrated that the performances of the flight model were similar to that of the reference model. The list of minerals used in this study was chosen based on the mineralogy of meteorites, interplanetary dust particles and Stardust samples. It contains anhydrous and hydrous ferromagnesian silicates, refractory silicates and oxides (present in meteoritic Ca-‐Al-‐rich inclusions), carbonates, and Fe-‐Ni sulfides. From the analyses of these minerals, we have calculated relative sensitivity factors for a suite of major and minor elements in order to provide a basis for element quantification for the possible identification of major mineral classes present in the cometary grains. Rosetta, cometary dust, interplanetary dust, micrometeorite, comet, time-‐of-‐flight secondary ion mass spectrometry (TOF-‐SIMS). Comets spend most of their lifetime far away from the sun and are therefore only little affected by solar radiation. In addition, as they are small bodies, they are very likely not altered by internal differentiation. Therefore comets are considered to be among the most primitive objects in the Solar System and might even still contain residuals of the solar nebula. In other words, comets may have preserved refractory and/or volatile interstellar material left over from Solar System formation and can provide key information on the origin of our Solar System. While remote observations allow measurements of collective properties of cometary dust, mass spectrometers flown on spacecraft allow the compositional analysis of individual particles. The latter technique was first introduced on the Giotto and Vega 1/2 missions to comet 1P/Halley (Kissel et al., 1986a; Kissel et al., 1986b). The measurements showed that in comet Halley’s dust, a mineral component is mixed with organic matter in individual grains (Lawler and Brownlee, 1992). Remote observations of comet C/1995 O1 (Hale-‐Bopp) and other bright comets, as well as laboratory analyses of cosmic dust of inferred cometary origin, showed that cometary dust is an unequilibrated, heterogeneous mixture of crystalline and glassy 2 1 Introduction silicate minerals, organic refractory material, and other constituents such as iron sulfide and possibly minor amounts of iron oxides (Bradley, 2005; Crovisier et al., 1997; Dobrică et al., 2012; Hanner and Bradley, 2004, and references therein). The silicates are mostly Mg-‐rich, while Fe is distributed in silicates, sulfides, and FeNi metal. Remote infrared spectra of silicate emission features in comet Hale-‐Bopp have led to identification of the minerals forsterite and enstatite in both, amorphous and crystalline form. This mineralogy is consistent with the composition of chondritic porous anhydrous interplanetary dust particles (CP-‐IDPs) (e.g., Brunetto et al., 2011) and of UltraCarbonaceous Antarctic MicroMeteorites (UCAMMs) (Dobrică et al., 2012). The high D/H ratios of the organic refractory material in these IDPs (Messenger, 2002) and in UCAMMs (Duprat et al., 2010), as well as the physical and chemical structure of glassy silicate grains, suggest a primitive origin of cometary dust. Whether the components are of presolar origin is still a matter of debate. Carbon is enriched relative to CI chondrites; some of the C is in an organic phase (Jessberger et al., 1988). “Ground truth” was provided by the Stardust mission which successfully returned in 2006 samples of dust collected in the coma of comet 81P/Wild 2 (Brownlee, 2014; Brownlee et al., 2006). The bulk of the Stardust samples appear to be weakly constructed mixtures of nanometer-‐sized grains, interspersed with much larger (>1 µm) ferromagnesian silicates, Fe-‐Ni sulfides, Fe-‐Ni metal, and others (Zolensky et al., 2006). The very wide variety of olivine and low-‐Ca pyroxene compositions in comet Wild 2 requires a wide range of formation conditions, probably reflecting very different formation locations in the protoplanetary disk (e.g., Frank et al., 2014). The restricted compositional ranges of Fe-‐Ni sulfides, the wide range for silicates, and the absence of hydrous phases indicate that comet Wild 2 likely experienced little or no aqueous alteration. Less abundant Wild 2 materials include refractory grains such as calcium-‐ aluminum-‐rich inclusions (CAIs), high-‐temperature phases (Brownlee, 2014, and references therein), whose presence appears to require radial transport in the early protoplanetary disk. Spitzer Space Telescope observations of comet 9P/Tempel 1 during the Deep Impact encounter revealed emission signatures that were assigned to amorphous and crystalline silicates, amorphous carbon, carbonates, phyllosilicates, polycyclic aromatic hydrocarbons, water gas and ice, and sulfides (Lisse et al., 2006). Good agreement is seen between the Tempel 1 ejecta spectra, the material emitted from comet Hale-‐Bopp, and the circumstellar material around the young stellar object HD100546 (Malfait et al., 1998). The atomic abundance of the observed material is consistent with solar and CI chondritic abundances. The presence of the observed mix of materials requires efficient methods of annealing amorphous silicates and mixing of high-‐ and low-‐temperature phases over large distances in the early protosolar nebula. In August 2014, the European Space Agency’s spacecraft Rosetta arrived at Jupiter-‐family comet 67P/Churyumov-‐Gerasimenko (hereafter 67P/C-‐G). The Rosetta spacecraft carries eleven scientific instruments to study the nucleus of the comet as well as the gas, plasma, and particle environment in the inner coma as a function of heliocentric distance. On November 12, 2014, the lander spacecraft Philae has performed the first ever landing on a comet nucleus and provided in situ analysis of its physical and compositional properties (Gibney, 2014; Glassmeier et al., 2007, and references therein; Hand, 2014). One of the core instruments of the Rosetta payload is the COmetary Secondary Ion Mass Analyser (COSIMA) that presently collects and analyzes the composition of dust grains in the coma of 67P/C-‐G (Kissel et al., 2009). COSIMA is a high-‐resolution 3 für Sonnensystemforschung (hereafter MPS) time-‐of-‐flight secondary ion mass spectrometry (TOF-‐SIMS) instrument (Vickerman et al., 2013), which uses an indium primary ion beam to analyze the chemical composition of collected cometary grains. The mass resolution is m/Δm ~ 1400 at 50% height (FWHM) of the peak at m/z=100 u. The bombardment of indium ions onto the sample produces secondary ions that are subsequently accelerated into a time-‐of-‐flight mass spectrometer, generating a secondary ion mass spectrum. By switching polarity of the mass spectrometer potentials, COSIMA is able to collect either positive or negative secondary ions. The goal of the COSIMA investigation is the in situ characterization of the elemental, molecular, mineralogical, and possibly isotopic composition of dust in the coma of comet 67P/C-‐G. A twin of the COSIMA instrument flying on board Rosetta is located at the Max-‐ Planck-‐Institut in Göttingen. This instrument serves as a reference instrument (Reference Model, RM) for the COSIMA flight instrument (named COSIMA XM). Pre-‐launch tests have shown that the performances of the RM and the XM are similar. Since the launch of Rosetta in 2004, the RM has been extensively used for laboratory calibration measurements. We have obtained a “library” of COSIMA mass spectra of well prepared and specially selected reference samples. Our reference samples are, among others, pure minerals expected to be present at the comet. These reference spectra will facilitate interpretation of the mass spectra expected from the comet with the COSIMA XM. In this paper, we describe calibration measurements with the COSIMA RM that we performed with a set of mineral samples during recent years. A similar calibration campaign with samples of organic compounds is described in an accompanying paper (Le Roy et al., 2014). For our COSIMA reference measurements, we selected minerals that have either been detected in comets or that were identified in other primitive Solar System materials, namely meteorites (in particular carbonaceous chondrites) or interplanetary dust particles (IDPs) and Antarctic micrometeorites. The selected mineral groups include anhydrous silicates (in particular olivines, pyroxenes, and feldspars of different compositions), hydrated silicates, oxides and hydroxides, carbonates, sulfides, pure elements and alloys (Table 1). For the abundant minerals in comets, in particular anhydrous silicates, more than one sample was measured from the same mineral class (e.g., olivine). The samples were either purchased from a commercial provider (MPS samples -‐ Krantz Mineral Shop in Bonn, Germany) or obtained from collections of the natural history museums in Los Angeles, London, Paris, and Vienna. A few samples were also provided through personal collaborations. The compositions of the mineral samples were either obtained from the literature, or were measured by electron microprobe at Univ. Paris VI, CAMPARIS. Major and minor elements were measured at 15 keV, 10 nA. Oxygen, carbon, and hydrogen were not measured but calculated by stoichiometry (for oxygen) or by difference (for carbon and hydrogen). The corresponding formula were calculated and compared to the theoretical values (Tables 1 and 2). 2 Samples and methods 2.1 Sample selection and determination of compositions. 4 Mineral family Mineral Name General Formula Table 1. Minerals analyzed with COSIMA RM including target types and numbers, and the sample preparation technique. Orthopyroxene Ca-‐poor Px Au blank Enstatite Ca-‐poor Px Ag blank Hypersthene Ag blank Ca-‐poor Px Clinopyroxene Ca-‐rich Px Au blank Diopside Ca-‐rich Px Ag blank Diopside Ca-‐rich Px Ag blank Augite Ag blank Ca-‐rich Px Hedenbergite Ca-‐rich Px Ag blank Forsterite Olivine Ag blank Forsterite Olivine Au blank Olivine Olivine Au blank Zabargad Fayalite Ag blank Olivine Albite Feldspar Ag blank Anorthite Feldspar Ag blank Plagioclase Feldspar Ag blank Plagioclase Feldspar Ag blank Orthoclase Ag blank Feldspar Nepheline Feldspathoid Ag blank Fuchsite Hydr. silicate Ag blank Richterite Hydr. silicate Ag blank Smectite Au blank Hydr. silicate Talc Hydr. silicate Ag blank Dolomite Carbonate Ag blank Calcite Carbonate Ag blank Melilite Ag blank Melilite Åkermanite Melilite Ag blank Ilmenite Oxide Ag blank Magnetite Oxide Ag blank Corundum Oxide Ag blank Ag blank Sphalerite Sulfide Pyrite Sulfide Au blank Pentlandite Sulfide Ag blank Pyrrhotite Sulfide Ag blank Substrate gold -‐-‐-‐ Au blank Substrate silver -‐-‐-‐ Ag blank H calculated by difference #C calculated by difference ♮Provider of the minerals, and sampling location, when available. MPS : Max Planck Institut für Sonnensystemforschung (Göttingen, Germany). CSNSM : Centre de Sciences Nucléaires et de Sciences de la Matière (Orsay France). LPC2E : Laboratoire de Physique et Chimie de l'Environnement et de l'Espace (Orléans France). ISTO : Institut des Sciences de la Terre d'Orléans (France). NHM : Natural History Museum (Vienna Austria, or London 5 UK), MNHN : Museum National d'Histoire Naturelle (Paris France). CRPG : Centre de Recherches Pétrographiques et Géochimiques (Nancy France). MPS – From M. Trieloff (Z31 Zabargad Island Kurat et al., 1993; Trieloff et al., 1997) CSNSM (MM, R2958, Bamle, Norway) CSNSM -‐ Los Angeles Museum MSP – From M. Trieloff (DW918 Witt-‐Eickschen et al., 2003)) CSNSM-‐NHM Vienna (Madagaskar) CSNSM-‐NHM London, BM 1906,382 (Italy) CSNSM-‐NHM London (Daun tuff quarry, Germany) LPC2E-‐ISTO 90407 CSNSM – From A. Revcolevski (Synthetic mineral) CSNSM – From A. Revcolevski (Synthetic mineral) MPS – From M.Trieloff (Z104 Zabargad Island Kurat et al., 1993; Trieloff et al., 1997) CSNSM – From J. Borg CSNSM (MM 118082, Ramona San Diego) CSNSM-‐NHM Vienna, (T. de la Foya, Austria) CSNSM-‐NHM Vienna (Tanzmeister, Austria) CSNSM-‐NHM Vienna (Tanzmeister, Austria) CSNSM-‐NHM London (Moon Stone, Sri Lanka) CSNSM-‐NHM London (York River, Ontario CA) CSNSM MPS (Bancroft Ontario, Canada) CSNSM (Bowling, Le Lamentin, Martinique CSNSM-‐Museum Lauzenac Ariege MPS (Vegarsheien, Norway) MPS (Creel Chihuahua, Mexico) CSNSM-‐MNHN Paris (Vesuvius) CSNSM-‐Dr.Morioka Japan (Synthetic mineral) MPS (Flekkefjord, Norway) MPS (Minas Gerais, Brasil) CSNSM-‐NHM Vienna (Ceylon) CSNSM (Picos de Europa, Spain) CSNSM-‐CRPG CSNSM-‐CRPG CSNSM-‐CRPG -‐-‐-‐-‐-‐-‐-‐-‐ -‐-‐-‐-‐-‐-‐-‐-‐ (Mg,Fe)SiO3 (Mg,Fe)SiO3 (Mg,Fe)SiO3 CaMgSi2O6 CaMgSi2O6 CaMgSi2O6 (Ca,Na)(Mg,Fe,Al,Ti)(Si,Al)2O6 CaFeSi2O6 Mg2SiO4 Mg2SiO4 (Mg,Fe)2SiO4 Fe2SiO4 NaAlSi3O8 CaAl2Si2O8 (Na,Ca)(Si,Al)4O8 (Na,Ca)(Si,Al)4O8 KAlSi3O8 (Na,K)AlSiO4 KAl2(Si3Al)O10(OH,F)2 Na(CaNa)(Mg,Fe)5 [Si8O22](OH)2 Ca0.25(Mg,Fe)3((Si,Al)4O10) (OH)2 nH2O Mg3Si4O10(OH)2 CaMg(CO3)2 CaCO3 (Ca,Na)2(Al,Mg,Fe)(Si,Al)2O7 Ca2Mg[Si2O7] FeTiO3 Fe3O4 Al2O3 [(Zn, Fe)S] FeS2 (Fe,Ni)9S8 Fe(1-‐x)S (x = 0 -‐ 0.17) -‐-‐-‐-‐-‐-‐-‐-‐-‐-‐ -‐-‐-‐-‐-‐-‐-‐-‐-‐-‐ Measured Formula (Mg0.9Fe0.1)Si0.9O3 Mg0.9SiO3 (Mg0.7Fe0.3)SiO3 Ca0.7Al0.1Mg0.9Fe0.1Si1.8O6 CaAl0.1Mg0.9Fe0.1Si2O6 (Ca0.6Na0.4)(Mg0.6Al0.4)Si2O6 (Ca0.9)(Mg0.8Fe0.2)(Si1.8Al0.2)O6 (Ca1.1)(Mg0.3Fe0.5Mn0.1)Si2O6 Mg2SiO4 Mg2SiO4 (Mg1.8Fe0.2)SiO4 Fe1.9SiO4 NaAlSi3O8 CaAl2Si2O8 (Na0.5Ca0.5)(Si2.5Al1.5)O8 (Na0.5Ca0.5)(Si2.5Al1.5)O8 (Na0.3K0.6)AlSi3O8 (Na0.6Ca0.3)AlSiO4 Na0.1K0.5Si3.2Al2.8Fe0.1O10(OH)1.8 Na0.9Al0.3K0.2Ca1.6(Mg4.6Fe0.4) [Si8O21.2(OH)4.8* Ca0.2(Mg0.1Fe2.5)((Si4Al0.1)O10(OH)2 2H2O* Mg3.4Si3.8O10(OH)2 4H2O* Ca(Mg0.8Fe0.2)(CO3)2# Ca1.1CO3# (Ca1.8Na0.1)(Al0.6Mg0.3Fe0.1)(Si1.6Al0.4)O7 Ca2Mg[Si2O7] (Fe0.8Mg0.2)TiO3 Fe2.5O4 (O measured as FeO) Al2O3 ZnS FeS2.0 (Fe4.4Ni4.8Co0.1)S8 FeS -‐-‐-‐-‐-‐-‐-‐-‐ -‐-‐-‐-‐-‐-‐-‐-‐ 41E 49C (111) 496 (150) 41D 49C (111) 49D (114) 496 (150) 497 (136) 496 (150) 420 48C 496 (150) 496 (150) 497 (136) 497 (136) 48B (AG57) 496 (150) 497 (136) 49D (114) 4B0 (147) 422 49C (111) 4AF (142) 4AF (142) 498 (143) 496 (150) 4B0 (147) 4B0 (147) 4B0 (147) 4AF (142) 421 49D (114) 49D (114) 41D/41E 49C (111) Pressing Suspension Suspension Pressing Suspension Suspension Suspension Suspension Suspension Pressing Pressing Suspension Suspension Suspension Suspension Pressing Suspension Suspension Suspension Suspension Pressing Suspension Suspension Suspension Suspension Suspension Suspension Suspension Suspension Suspension Pressing Suspension Suspension No sample No sample Provider (Origin)♮ Target Type Cosima Target Label Preparation Technique Na b.d. b.d. b.d. 0.30 0.20 4.04 0.39 -‐ b.d. -‐ 0.07 7.57 b.d. 3.56 2.67 9.13 0.32 1.90 0.07 b.d. b.d. b.d. 1.03 b.d. b.d. b.d. -‐ -‐ -‐ -‐ -‐ Mg 17.41 18.94 13.67 9.24 9.05 5.90 7.58 3.21 28.36 25.82 0.18 b.d. b.d. b.d. b.d. b.d. 0.20 9.90 0.43 10.27 8.37 b.d. 2.80 8.28 4.05 0.08 -‐ -‐ -‐ -‐ -‐ Al 0.80 0.07 0.90 1.08 1.01 3.66 2.36 -‐ b.d. -‐ 0.12 7.85 15.29 11.64 7.82 14.33 13.82 0.58 0.55 b.d. b.d. b.d. 8.12 b.d. b.d. 0.23 39.99 -‐ -‐ -‐ -‐ Table 2. Composition of the minerals measured by electron microprobe (atomic percent). V Ti Mineral Family Mineral Name Ca-‐poor Px b.d. -‐ Orthopyroxene Ca-‐poor Px b.d. -‐ Enstatite Ca-‐poor Px 0.05 -‐ Hypersthene Ca-‐rich Px 0.16 -‐ Clinopyroxene -‐ b.d. Diopside. (Madagaskar) Ca-‐rich Px Ca-‐rich Px 0.03 -‐ Diopside (Italy) Ca-‐rich Px Augite 0.41 -‐ Ca-‐rich Px Hedenbergite -‐ -‐ Olivine Synthetic Forsterite b.d. -‐ -‐ -‐ San Carlos Olivine Olivine Olivine Fayalite b.d. -‐ Feldspar Albite b.d. -‐ Feldspar Anorthite b.d. -‐ Feldspar Plagioclase b.d. -‐ Feldspar Orthoclase b.d. -‐ -‐ b.d. Feldspathoid Nepheline Hydrated silicate Fuchsite 0.32 -‐ Hydrated silicate Richterite b.d. -‐ Hydrated silicate Smectite b.d. -‐ Hydrated silicate Talc b.d. -‐ b.d. b.d. Dolomite Carbonate Carbonate Calcite b.d. -‐ Melilite Melilite b.d. -‐ Melilite Åkermanite b.d. -‐ Oxide Ilmenite 19.28 -‐ -‐ b.d. Magnetite Oxide Oxide Corundum b.d. -‐ Sulfide Sphalerite -‐ -‐ Sulfide Pyrite -‐ -‐ Sulfide Pentlandite -‐ -‐ Sulfide 0.14 -‐ Pyrrhotite H calculated by difference #C calculated by difference b.d. : below detection limit O 60.77 60.13 59.94 61.91 60.00 59.95 59.86 60.08 57.21 56.54 57.28 61.54 61.55 61.49 61.62 58.45 58.51 55.64 52.13 48.11 59.99 59.56 58.39 58.39 59.66 61.29 60.01 -‐ -‐ -‐ -‐ S -‐ b.d. b.d. b.d. b.d. b.d. b.d. -‐ b.d. -‐ b.d. b.d. b.d. b.d. b.d. b.d. b.d. b.d. b.d. b.d. b.d. b.d. b.d. b.d. b.d. b.d. b.d. 50.71 65.99 46.30 49.63 Si 19.02 20.15 19.34 18.60 19.52 19.95 18.28 20.15 14.43 14.81 14.46 22.95 15.40 19.02 23.01 14.31 15.40 17.17 14.92 11.32 b.d. b.d. 13.28 16.79 b.d. 0.05 b.d. -‐ -‐ -‐ -‐ P -‐ b.d. b.d. -‐ b.d. 0.07 b.d. -‐ b.d. -‐ b.d. b.d. b.d. b.d. b.d. b.d. b.d. b.d. b.d. b.d. b.d. b.d. b.d. b.d. b.d. b.d. b.d. -‐ -‐ -‐ -‐ K -‐ b.d. b.d. -‐ b.d. b.d. b.d. -‐ b.d. -‐ b.d. 0.04 b.d. 0.20 4.80 b.d. 2.25 0.38 0.15 b.d. b.d. b.d. 0.11 b.d. b.d. b.d. b.d. -‐ -‐ -‐ -‐ Ca 0.14 b.d. 0.56 7.20 9.55 5.95 9.12 10.74 b.d. b.d. b.d. 0.05 7.76 4.03 0.09 3.78 b.d. 3.37 0.64 b.d. 10.01 21.34 15.41 16.54 b.d. b.d. b.d. -‐ -‐ -‐ -‐ Cr 0.04 b.d. 0.05 0.09 b.d. b.d. 0.08 -‐ b.d. b.d. b.d. b.d. b.d. b.d. b.d. b.d. 0.05 b.d. b.d. b.d. b.d. b.d. b.d. b.d. 0.05 b.d. b.d. -‐ b.d. b.d. b.d. Mn 0.04 b.d. 0.12 0.03 0.04 0.36 0.04 0.52 b.d. -‐ b.d. b.d. b.d. b.d. b.d. b.d. b.d. 0.03 b.d. b.d. 0.05 b.d. b.d. b.d. 0.11 0.02 b.d. -‐ 0.05 b.d. b.d. Fe 1.75 0.71 5.36 1.35 0.63 0.11 1.88 5.30 b.d. 2.73 27.89 b.d. b.d. 0.07 b.d. b.d. 0.39 0.82 9.34 0.09 1.61 b.d. 0.85 b.d. 16.81 38.33 b.d. 0.05 33.97 25.70 50.21 Co -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ b.d. -‐ -‐ -‐ -‐ -‐ -‐ b.d. 0.35 b.d. Ni 0.04 b.d. b.d. 0.03 b.d. b.d. b.d. -‐ b.d. 0.10 b.d. b.d. b.d. b.d. b.d. b.d. b.d. b.d. 0.03 b.d. -‐ b.d. b.d. b.d. 0.04 b.d. b.d. -‐ b.d. 27.65 0.03 Cu -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ b.d. b.d. b.d. Zn -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ 49.24 -‐ b.d. -‐ H 8.7* 10.2* 21.7* 30.2* C# 20.0# 19.1# Total 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 6 2.2 COSIMA substrates and sample preparation The COSIMA XM is equipped with a set of 72 substrates of different types to collect dust grains in the cometary coma (Genzer and Rynö, 2010): _ 34x Gold black _ 12x Silver black _ 16x Platinum black _ 3x Palladium black _ 7x Silver blank The metallic black were obtained by deposition of nanometer-‐sized grains, formed by condensation from the vapor phase in the case of Au and Ag, or by an electrochemical procedure in the case of Pt and Pd (Hornung et al., 2014). To ease sample preparation for the laboratory calibration campaign, we selected blank silver substrates in most cases and in a few cases blank gold substrates (Table 1). The blank silver targets were cleaned in an ultrasonic bath for 15 min, first in isopropanol and then in distilled water. The gold blank targets were cleaned with acetone, and heated to 900C for 5 min. The heating of the target removes any organics from the target and softens the gold. The samples procured by MPS (see Table 1) were crushed to pieces of several millimeters to one centimeter in size with a jaw crusher. They were then cleaned for 10 min in ethanol followed by 10 min in deionized water using an ultrasonic bath. After inspection by eye and optical microscope to identify pure and clean specimen of the respective mineral, a few selected specimens were then ground in an electrical ball mill down to a smallest grain size of approximately 25 µm. The powder obtained was then sieved with a set of stainless steel sieves with mesh sizes between 25 µm and 200 µm. The fraction with grain size between 25 µm and 50 µm was usually selected for COSIMA measurements. Samples from MPS (Zabargad olivine and clino-‐ and orthopyroxene; Kurat et al., 1993; Trieloff et al., 1997; Witt-‐Eickschen et al., 2003) were prepared by standard mineral separation techniques, i.e., several cycles of hand-‐picking of coarse grained material, crushing, sieving, washing, and occasionally magnetic separation. Samples from CSNSM were washed in acetone, then crushed in an agate mortar and visually examined in order to pick a fragment relevant to the mineral type. Two different types of sample preparation were used (Table 1): suspension (at MPS) or pressing in gold foil (at CSNSM). In the first case, the sieved minerals was suspended in water and applied to the metal target with a pipette in the field of view of a laboratory microscope. Approximately 1 µl was deposited onto the substrate with a pipette, creating a 1 mm droplet on the surface of the target. After evaporation of the solvent, a homogeneous surface coverage was usually visible on the substrate. For the pressing method, small fragments (~ 50 μm) of the minerals were selected using a binocular and pressed into the blank gold substrates using a microcrusher at CSNSM, which is usually used to press micrometeorites or IDPs into gold foils. In the microcrusher, the sample was crushed into the foil with a disk of fused silica previously cleaned for 10 min with ethanol followed by 10 min in deionized water using an ultrasonic bath. 7 2.3 Measurement strategy with the COSIMA RM and selection of best mass spectra Positive and negative secondary ion mass spectra have been obtained for the 31 selected minerals with the COSIMA RM at MPS. Each sample was measured at least at two locations as a set of 4x4 or 5x5 raster with a separation of the raster points of 50 µm. The measurement time per spectrum was typically 5 min. Background spectra were systematically acquired outside of the sample (see below). In order to establish the intensities and intensity ratios for several relevant elements for each mineral analyzed, three to six positive mass spectra were selected for evaluation. The same coordinates on the target were used for selection of positive as well as negative spectra. The mass spectra were selected based on the following criteria: the spectra had to be on the mineral grains and the level of contamination in the spectra had to be as low as possible. Common contaminants found in TOF-‐SIMS analyses include (i) organic compounds such as polydimethylsiloxane (PDMS -‐ a silicone oil) and phthalates, and (ii) ions such as Na+, K+, Cl-‐, SO2-‐, SO3-‐, and HSO4-‐. Mass spectra measured outside of the minerals on the substrate were used as contamination control. Features related to common contaminants for these analyses will be described in Hilchenbach et al. (in prep.), therefore this topic will not be discussed further. Last but not least, particular attention was paid to the characteristic peaks of each mineral: presence of individual elements and correlated elements, as well as their corresponding intensities. Raman spectra of all mineral samples were measured using a laboratory Raman spectrometer (model alpha300 R, WITec, Ulm, Germany) in order to verify the mineral identification. The confocal Raman spectrometer is equipped with a polarized fiber optic coupled 532 nm laser. Raman scattered light and fluorescence emission is transmitted through a beamsplitter, a laser notch filter, a long wavelength filter, and a 50 µm optical fiber to a spectrometer with a Peltier-‐cooled CCD detector. The wavenumber range of the spectrometer is 150 cm-‐1 to 3800 cm-‐1 with 5 cm-‐1 spectral resolution. The microscopic system accommodates three objectives with increasing magnifying power and numeric aperture (10x, 50x, and 100x). Metal targets with mineral grains were placed on the piezo-‐driven x-‐y scan table beneath the objective coupled to the z-‐axis focusing unit. The coarse sample grain surfaces were monitored with a CCD video camera prior to and following the Raman scans with diffraction limited optics. The excitation intensity of the laser system was adjusted prior to the depth scan with a variable slit between the laser and the transmitting optical fiber to maximize the recorded Raman and fluorescence emission while keeping sample alterations due to heating at a minimum. Since the sample composition and therefore absorption and refractive index were not spatially uniform, sample spot deterioration could not be ruled out prior to matrix scans for the whole area. The excitation intensity varies from 0.4 mW to 5 mW for samples with high and low absorption. The effective measurement time interval and laser illumination was 0.2 s for each scan matrix point. For each sample, the spectral data was obtained from two areas 80 x 80 µm2 that had been previously analyzed by COSIMA. Two kinds of Raman analyses were made: first a slice cutting through a selected area was scanned in depth mode along a line parallel to the x-‐axis in the x-‐z-‐plane, and then, an image mode scan was made parallel to the focal plane in an x-‐ y plane encompassing the sample surface. In the latter case, fractions of the rectangular scan matrix were in or out of focus due to the surface roughness of the mineral grain samples. The recorded fluorescence and Raman emission spectra were corrected for 8 2.4 Raman measurements 2.5 Sputtering cosmic ray particle events. Spectra were summed up and averaged for selected adjoining measurement points within each scan to improve the signal-‐to-‐noise ratio. False color images for both scan modes were plotted for selected spectral bandwidth, resulting in depth and image scans each representing other spectral features and thereby allowing the spatial identification and feasible separation of the emission sources. The minerals were identified by comparison of the observed Raman scattered lines with a database of Raman spectra of minerals accessible via the RRUFF Project webpage (RRUFF Project). To clean the mineral surface before analysis, sputtering with a direct current or long pulses is often used in SIMS (Stephan, 2001). During sputtering, the mineral surface is exposed to a much higher ion dose than during analysis when short pulses are used. This increased ion dose efficiently removes any organic molecules and other undesired components covering the mineral surface, thus increasing the ion signals obtained from the mineral. In addition to removing any contaminants, the sputtering also makes the mineral matrix more homogenous and causes amorphization (e.g., Stephan, 2001), which means a more stable ion signal can be obtained during subsequent analyses. The sputtering time needed to obtain these effects depends on the ion beam and current used, and size of area sputtered, however a couple of minutes of sputtering is usually sufficient (Siljeström et al., 2010; Stephan, 2001). Most studies of sputtering of minerals have been done on flat mineral surfaces with either Ar+, C60+, Bi+, Cs+, and O+/-‐, which are the sputter ion beams most frequently found on commercial TOF-‐SIMS instruments (Siljeström et al., 2011; Stephan, 2001). So far, no studies on the sputtering of mineral grains with an indium primary beam have been performed. Therefore, studies on the effects of sputtering of mineral grains had first to be performed in the RM before it can be used on samples collected in space. The sputtering experiments were executed according to the following protocol: 5 min sputtering followed by analysis during 5 min, and this sequence was repeated up to 10 times. The primary emission current for sputtering was 10 µA (continuous beam), and 5 µA for the analysis beam (pulsed beam). We will not further discuss the sputtering experiments on mineral grains performed in the RM, as no significant effect on the stability of the secondary ion beam was demonstrated. Sputtering was, however, useful for cleaning the sample surface from PDMS contamination. Measurements of all mineral samples listed in Tables 1 and 2 were performed with the COSIMA RM instrument at MPS. Representative spectra of a selection of minerals are displayed in Figure 1. Most minerals have a signature in positive secondary ions (Fig. 1a to 1e), and Fe-‐sulfides show the sulfur signature in negative secondary ion spectra (Fig. 1f). Figure 2 shows details of some elemental peaks presenting the discrimination between the inorganic peak and the organic peak present at each mass. 3 Results and discussion 9 Figure 1 : Representative spectra of some minerals analyzed in this study (a-‐e: positive secondary ions, f: negative secondary ions). a) Mg-‐rich olivine (Z104); b) Mg-‐rich Ca-‐poor pyroxene (Enstatite Bamle); c) Ca-‐rich pyroxene (DW918); d) Melilite (Vesuvius); e)Hydrated mineral fuchsite; f) Fe-‐sulfide pyrrhotite. See Tables 1 and 2 for description and compositions of the minerals. 10 ) 0 EE , ( 0 ) = RSF ESI ( ESI ( 0 ) EE 3.1 Relative sensitivity factors In order to quantify SIMS results, to calculate element ratios from secondary ion ratios, relative sensitivity factors (RSFs) are needed. Positive spectra were used for the quantification of the RSFs. Negative spectra are used to demonstrate the presence of S-‐ rich compounds, especially in the case of Fe-‐sulfides. As no normalizing element is present in negative spectra for Fe sulfides, no RSF could be calculated for sulfur. For a known element atomic ratio E/E0, the RSF can be calculated from the secondary ion intensities SI: The elemental E/E0 ratio is known from the composition of the mineral, and the normalizing element E0 is usually one of the most abundant specie (i.e., Si, Mg, or Fe). For silicate minerals, such RSFs are usually reported relative to Si since it is the only element besides O being present in all silicates. Oxygen, on the other hand, has a very low ionization probability for positive secondary ions, which makes it not suitable as a reference element. RSFs are obtained by analyzing standard materials with known elemental composition under the same conditions as the samples to be analyzed (i.e., cometary grains in this case). Following data evaluation steps as described by Stephan (2001), secondary ion ratios for numerous elements relative to Si were obtained. For some mineral samples, either no statistically significant Si element data were available or the Si peaks were compromised by silicone oil contamination. In such cases, we used Mg as a reference element and renormalized the result to Si using an RSF(Mg/Si) of 3.71 (geometric mean of RSF(Mg/Si) values calculated for minerals with reliable Si data). If Si and Mg normalization failed, Fe was used as a reference element, and an RSF(Fe/Si) of 1.81 (geometric mean of RSF(Fe/Si) values calculated for minerals with reliable Si data) was applied. Table 3 shows all results that were used to calculate (geometric) mean values for the RSFs. Figure 3 shows a comparison between RSFs calculated in this study with data from Stephan (2001) for a commercial TOF-‐SIMS instrument that uses a 25 keV 69Ga+ primary ion beam. The general trend for both primary ion species is the same, and ionization probability mainly depends on the ionization energy of a given element. For some elements, only limited data are available, e.g., V, Co, and Ni were only measured reliably in one mineral each. This might explain why the RSFs for Co and Ni do not seem to follow the general trend. Therefore, we recommend for these elements, RSFs that have been calculated from Fe-‐normalized values from Stephan (2001). Table 4 presents mean RSFs normalized to Si, Mg, and Fe. 11 Table 3. Relative sensitivity factors for positive secondary ions normalized to Si obtained with the COSIMA RM instrument at MPS. Fe Ni Sample Co Mn 1.05(18) — Enstatite Bamle — — 1.67(5) — — Hypersthene 3.6(5) 1.74(13) — Clinopyroxene — — 3.3(3) — Diopside Madagascar — — — — Diopside San Marcel — — 1.8(5) — — Augite — 1.87(4) — Hedenbergite — 3.0(2) 2.17(3) — Olivine Zabargad — — 1.07(3) — Fayalite — — — — — Albite — — — Anorthite — — — — Plagioclase (497) — — — — Plagioclase (48B) — — — — — Orthoclase — — — Nepheline — — 1.9(3) — Fuchsite — — 3.34(13) — Richterite — — 0.932(9) — — Smectite (422) — 1.57(2) 2.4(11) Smectite (49C) — — — — Talc — — 1.0(3) — — Dolomite 2.3(7) 5.04(18) — Melilite — — — — Akermanite — — 1.88(5) — Ilmenite — — — 0.73(8) Pentlandite — ≡1.81 Errors referring to the last significant digits are given in parentheses (i.e. RSF(Fe/Si) for Enstatite Bamle = 1.05 ± 0.18). n gives the — Pyrrhotite — — ≡1.81 number of samples measured to calculate the geometric mean values. Two samples on different substrates were analyzed for plagioclase and smectite, respectively. For Co and Ni, recommended values given in italics are derived from literature values (Stephan, 2001). *Numbers for O are calculated from positive spectra, hence the low values, comparable to those of Stephan (2001). Mg Na O Al 2.56(3) — 0.00060(16) — 2.01(5) — 0.0010(3) — 24.2(10) 0.0027(8) — ≡3.71 3.06(5) — 0.0009(3) — 6.3(2) 171(6) 0.0025(5) 9.0(7) 3.77(7) 0.00061(19) 50.1(9) — 3.46(8) 0.00057(17) — — 3.49(3) 0.00039(11) — — 6.3(4) 0.0011(3) — — — 0.00039(13) 18.6(3) 4.25(15) — 0.00042(18) — 3.27(8) — 0.0008(2) 8.28(12) 1.45(6) — 0.00083(14) 18.3(2) 4.08(11) 0.0009(4) — 10.9(4) 3.4(3) 0.0008(2) 10.42(15) — 2.07(5) 0.0010(2) 71.7(14) 4.8(2) 3.03(9) 0.0012(3) 57.7(9) 3.61(6) — 4.9(5) 0.00030(5) 127.0(11) 3.37(6) — 0.00033(8) 5.32(11) 9.4(4) 0.00067(11) — 1.806(19) — — — 0.0010(3) ≡3.71 0.0013(3) 115(2) 9.5(2) 1.14(13) 0.00085(14) 2.29(4) — — 0.0012(5) — — ≡3.71 — — — — — — — — Cr — — 1.8(6) — — — — — — — — — — — — 3.9(4) — — — — — — — — — — Si ≡1 ≡1 — ≡1 ≡1 ≡1 ≡1 ≡1 ≡1 ≡1 ≡1 ≡1 ≡1 ≡1 ≡1 ≡1 ≡1 ≡1 ≡1 ≡1 — ≡1 ≡1 — — — Ca K — — 12.3(4) — 5.24(13) — 7.84(11) — 15.3(6) — 5.60(9) — — — — — — — — — 5.49(11) — 4.69(8) 62(2) — — — 17.6(7) 5.13(9) — — 96.8(19) 75.0(15) 13.9(2) 9.72(10) — 75.7(18) 9.14(14) — — 7.59(11) — 118(5) 10.9(2) — 3.51(5) — — — — — — Ti — 4.6(9) 1.7(8) — — 4.8(2) — — — — — — — — — 2.38(14) — — — — — — — — — — V — — — — — — — — — — — — — — — — — — — — — — — — — 3.5(5) 12 Figure 2: Details of Mg+, Al+, Ca+, and Fe+ peaks showing the discrimination between the inorganic (to the left of the integer mass/charge) and organic peaks (to the right of the integer mass/charge), thus allowing the quantification of relative sensitivity factors, normalized to Si, Mg or Fe. 13 Li Be B C O NaMg Al Si P K Ca Sc Ti V CrMnFe Co Ni Cu 102 101 100 10-1 10-2 3 4 5 6 8 11 12 13 14 15 Atomic Number 19 20 21 22 23 24 25 26 27 28 29 Figure 3: Mean RSFs normalized to Si for major and minor elements versus atomic number obtained from positive ion spectra of various mineral samples. Filled circles are geometric mean values from Table 3. The vertical bars show the range of individual values obtained for different minerals, except for V, Co, and Ni for which they represent statistical errors. Open circles for Co and Ni are derived from literature values (Stephan, 2001). For comparison, RSFs from Stephan (2001) for a commercial TOF-‐SIMS instrument with a 25 keV 69Ga+ primary ion beam and that were obtained from a suite of glass standards are shown as open diamonds. For these, the variation range is shown in gray. ) 1 = S i ( r o t c a F y t i v i t i s n e S e v i t a l e R 10-3 14 n n 24 12 15 11 21 6 14 4 1 2 3 15 1 1 n 24 12 18 11 15 6 14 4 1 2 3 15 1 1 RSF (Fe≡1) 0.00044 +0.00035 –0.00019 18 +35 –12 2.0 +1.4 –0.8 1.8 +2.8 –1.1 0.53 +0.35 –0.21 27 +24 –13 3.9 +3.0 –1.7 1.7 +1.2 –0.7 1.9 ±0.3 1.5 +0.9 –0.6 2.0 +0.4 –0.3 ≡1 0.64 0.40 RSF (Mg≡1) 0.00022 +0.00017 –0.00010 8.3 +12.7 –5.0 ≡1 0.78 +0.89 –0.42 0.27 +0.16 –0.10 13 +10 –6 2.0 +1.2 –0.8 0.90 +1.06 –0.49 0.87 ±0.13 0.62 +0.26 –0.18 0.99 +0.72 –0.42 0.47 +0.30 –0.18 0.30 0.19 Table 4. Recommended COSIMA relative sensitivity factors for positive secondary ions normalized to Si, Mg, and Fe. Element RSF (Si≡1) O 24 0.00080 +0.00060 –0.00034 12 35 +67 Na –23 Mg 18 3.7 +2.2 –1.4 11 3.5 +3.2 Al –1.7 Si 12 ≡1 K 64 +62 6 –31 14 7.6 +4.3 Ca –2.7 Ti 3.1 +2.0 4 –1.2 1 3.5 ±0.5 V Cr 2.6 +1.9 2 –1.1 Mn 2.9 +0.8 3 –0.6 17 1.8 +1.1 Fe –0.7 Co 1.2 1 1 0.72 Ni Sensitivity factors relative to Si were calculated from geometric mean values for data shown in Table 3. For Mg and Fe normalized RSF, individual mineral data were first normalized to these elements, and then geometric mean values were calculated. n gives the number of samples measured to calculate the geometric mean values. For Co and Ni, recommended values given in italics are derived from Fe-‐normalized literature values (Stephan, 2001). One of the main objectives of the Rosetta mission is to characterize the elemental and the mineral compositions of the cometary material. This is also an important objective for COSIMA. As many of the minerals identified in cometary material such as pyroxene and olivine carry the same elemental signal, it will be challenging to differentiate between these classes of minerals. In addition, since the individual mineral grain sizes of cometary material are much smaller than the size of the primary ion beam, mixtures of minerals are measured. Statistical methods like PCA, Corico, KNN, RP, Unscrambler, etc. can be applied to further differentiate between minerals (Engrand et al., 2006; Krüger et al., 2011; Varmuza et al., 2011; Varmuza et al., 2014). COSIMA is a high mass resolution dust analyzer that is able to discriminate the mineral and the organic compounds in the mass spectra of dust particles from comet 67P/Churyumov-‐Gerasimenko (67P/C-‐G). To prepare the scientific return of the COSIMA analyses, we have characterized a series of minerals relevant to cometary matter with the reference model of COSIMA on ground. Relative sensitivity factors of 15 3.2 Identification of minerals 5 Summary Forschungszentrum Seibersdorf, Acknowledgements Seibersdorf, Austria, elements have been derived from these analyses, expressed as ratios normalized to Si, Mg, and Fe. Using COSIMA, we will thus be able to quantify the major element composition of 67P/C-‐G cometary grains, normalized to Si, Mg, or Fe. COSIMA was built by a consortium led by the Max-‐Planck-‐Institut für Extraterrestrische Physik, Garching, Germany in collaboration with Laboratoire de Physique et Chimie de l'Environnement et de l'Espace, Orléans, France, Institut d'Astrophysique Spatiale, CNRS/Université Paris Sud, Orsay, France, Finnish Meteorological Institute, Helsinki, Finland, Universität Wuppertal, Wuppertal, Germany, von Hoerner und Sulger GmbH, Schwetzingen, Germany, Universität der Bundeswehr, Neubiberg, Germany, Institut für Physik, Institut für Weltraumforschung, Österreichische Akademie der Wissenschaften, Graz, Austria and is lead by the Max-‐Planck-‐Institut für Sonnensystemforschung, Göttingen, Germany. The support of the national funding agencies of Germany (DLR, grant 50 QP 1302), France (CNES), Austria, Finland, Sweden (Swedish National Space Board, contract No. 121/11) and the ESA Technical Directorate is gratefully acknowledged. The staff of the national CAMPARIS analytical facility in France (University Pierre et Marie Curie, Paris VI) is acknowledged for helping in the measurement of the mineral compositions by electron microprobe. MT acknowledges support by Klaus Tschira foundation. We thank the Rosetta Science Ground Segment at ESAC, the Rosetta Mission Operations Centre at ESOC and the Rosetta Project at ESTEC for their outstanding work enabling the science return of the Rosetta Mission. Bradley, J.P., 2005. Interplanetary dust particles, In: Davis, A.M., Holland, H.D., Turekian, K.K. (Eds.), Meteorites, Comets and Planets: Treatise on Geochemistry. Elsevier-‐ Pergamon, Oxford, pp. 689-‐711. Brownlee, D., 2014. The Stardust Mission: Analyzing Samples from the Edge of the Solar System. Annual Review of Earth and Planetary Sciences 42, 179-‐205. Brownlee, D., Tsou, P., Aléon, J., Alexander, C.M.O.D., Araki, T., Bajt, S., Baratta, G.A., Bastien, R., Bland, P., Bleuet, P., Borg, J., Bradley, J.P., Brearley, A., Brenker, F., Brennan, S., Bridges, J.C., Browning, N.D., Brucato, J.R., Bullock, E., Burchell, M.J., Busemann, H., Butterworth, A., Chaussidon, M., Cheuvront, A., Chi, M., Cintala, M.J., Clark, B.C., Clemett, S.J., Cody, G., Colangeli, L., Cooper, G., Cordier, P., Daghlian, C., Dai, Z., D'Hendecourt, L., Djouadi, Z., Dominguez, G., Duxbury, T., Dworkin, J.P., Ebel, D.S., Economou, T.E., Fakra, S., Fairey, S.A.J., Fallon, S., Ferrini, G., Ferroir, T., Fleckenstein, H., Floss, C., Flynn, G., Franchi, I.A., Fries, M., Gainsforth, Z., Gallien, J.P., Genge, M., Gilles, M.K., Gillet, P., Gilmour, J., Glavin, D.P., Gounelle, M., Grady, M.M., Graham, G.A., Grant, P.G., Green, S.F., Grossemy, F., Grossman, L., Grossman, J.N., Guan, Y., Hagiya, K., Harvey, R., Heck, P., Herzog, G.F., Hoppe, P., Hörz, F., Huth, J., Hutcheon, I.D., Ignatyev, K., Ishii, H., Ito, M., Jacob, D., Jacobsen, C., Jacobsen, S., Jones, S., Joswiak, D., Jurewicz, A., Kearsley, A.T., Keller, L.P., Khodja, H., Kilcoyne, A.L.D., Kissel, J., Krot, A., Langenhorst, F., Lanzirotti, A., Le, L., Leshin, L.A., Leitner, J., Lemelle, L., Leroux, H., Liu, M.-‐C., Luening, K., Lyon, I., MacPherson, G., Marcus, M.A., Marhas, K., Marty, B., Matrajt, G., McKeegan, K., Meibom, A., Mennella, V., Messenger, 16 References K., Messenger, S., Mikouchi, T., Mostefaoui, S., Nakamura, T., Nakano, T., Newville, M., Nittler, L.R., Ohnishi, I., Ohsumi, K., Okudaira, K., Papanastassiou, D.A., Palma, R., Palumbo, M.E., Pepin, R.O., Perkins, D., Perronnet, M., Pianetta, P., Rao, W., Rietmeijer, F.J.M., Robert, F., Rost, D., Rotundi, A., Ryan, R., Sandford, S.A., Schwandt, C.S., See, T.H., Schlutter, D., Sheffield-‐Parker, J., Simionovici, A., Simon, S., Sitnitsky, I., Snead, C.J., Spencer, M.K., Stadermann, F.J., Steele, A., Stephan, T., Stroud, R., Susini, J., Sutton, S.R., Suzuki, Y., Taheri, M., Taylor, S., Teslich, N., Tomeoka, K., Tomioka, N., Toppani, A., Trigo-‐Rodríguez, J.M., Troadec, D., Tsuchiyama, A., Tuzzolino, A.J., Tyliszczak, T., Uesugi, K., Velbel, M., Vellenga, J., Vicenzi, E., Vincze, L., Warren, J., Weber, I., Weisberg, M., Westphal, A.J., Wirick, S., Wooden, D., Wopenka, B., Wozniakiewicz, P., Wright, I., Yabuta, H., Yano, H., Young, E.D., Zare, R.N., Zega, T., Ziegler, K., Zimmerman, L., Zinner, E., Zolensky, M., 2006. Comet 81P/Wild 2 under a microscope. Science 314, 1711-‐1716. Brunetto, R., Borg, J., Dartois, E., Rietmeijer, F.J.M., Grossemy, F., Sandt, C., Le Sergeant d'Hendecourt, L., Rotundi, A., Dumas, P., Djouadi, Z., Jamme, F., 2011. Mid-‐IR, Far-‐IR, Raman micro-‐spectroscopy, and FESEM-‐EDX study of IDP L2021C5: Clues to its origin. Icarus 212, 896-‐910. Crovisier, J., Leech, K., Bockelée-‐Morvan, D., Brooke, T.Y., Hanner, M.S., Altieri, B., Keller, H.U., Lellouch, E., 1997. The spectrum of comet Hale-‐Bopp (C/1995 O1) observed with the infrared space observatory at 2.9 astronomical units from the Sun. Science 275, 1904-‐1907. Dobrică, E., Engrand, C., Leroux, H., Rouzaud, J.N., Duprat, J., 2012. Transmission electron microscopy of CONCORDIA ultracarbonaceous Antarctic micrometeorites (UCAMMs): mineralogical properties. Geochim. Cosmochim. Acta 76, 68-‐82. Duprat, J., Dobrica, E., Engrand, C., Aléon, J., Marrocchi, Y., Mostefaoui, S., Meibom, A., Leroux, H., Rouzaud, J.N., Gounelle, M., Robert, F., 2010. Extreme deuterium excesses in ultracarbonaceous micrometeorites from central Antarctic snow. Science 328, 742-‐745. Engrand, C., Kissel, J., Krueger, F.R., Martin, P., Silén, J., Thirkell, L., Thomas, R., Varmuza, K., 2006. Chemometric evaluation of time-‐of-‐flight secondary ion mass spectrometry data of minerals in the frame of future in situ analyses of cometary material by COSIMA onboard ROSETTA. Rapid. Commun. Mass Spectrom. 20, 1361-‐1368. Frank, D.R., Zolensky, M.E., Le, L., 2014. Olivine in terminal particles of Stardust aerogel tracks and analogous grains in chondrite matrix. Geochim. Cosmochim. Acta 142, 240-‐259. Genzer, M., Rynö, J., 2010. COSIMA Operation Manual. Version 1.7. Gibney, E., 2014. Philae's 64 hours of comet science yield rich data, Nature News. Nature http://www.nature.com/news/philae-‐s-‐64-‐hours-‐of-‐comet-‐ Publishing Group, science-‐yield-‐rich-‐data-‐1.16374. Glassmeier, K.-‐H., Boehnhardt, H., Koschny, D., Kührt, E., Richter, I., 2007. The Rosetta Mission: Flying Towards the Origin of the Solar System. Space Sci. Rev. 128, 1-‐21. Hand, E., 2014. Philae probe makes bumpy touchdown on a comet. Science 346, 900-‐ 901. Hanner, M.S., Bradley, J.P., 2004. Composition and mineralogy of cometary dust, Comets II, pp. 555-‐564. Hornung, K., Kissel, J., Fischer, H., Mellado, E.M., Kulikov, O., Hilchenbach, M., Kr¸ger, H., Engrand, C., Langevin, Y., Rossi, M., Krueger, F.R., 2014. Collecting cometary dust particles on metal blacks with the COSIMA instrument onboard ROSETTA. Planet. Space Sci. 103, 309-‐317. 17 Jessberger, E.K., Christoforidis, A., Kissel, J., 1988. Aspects of the major element composition of Halley's dust. Nature 332, 691-‐695. Kissel, J., Altwegg, K., Briois, C., Clark, B.C., Colangeli, L., Cottin, H., Czempiel, S., Eibl, J., Engrand, C., Fehringer, H.M., Feuerbacher, B., Fischer, H., Fomenkova, M., Glasmachers, A., Greenberg, J.M., Grün, E., Haerendel, G., Henkel, H., Hilchenbach, M., von Hoerner, H., Höfner, H., Hornung, K., Jessberger, E.K., Koch, A., Krüger, H., Langevin, Y., Martin, P., Parigger, P., Raulin, F., Rüdenauer, F., Rynö, J., Schmid, E.R., Schulz, R., Silén, J., Steiger, W., Stephan, T., Thirkell, L., Thomas, R., Torkar, K., Utterback, N.G., Varmuza, K., Wanczek, K.P., Werther, W., Zscheeg, H., 2009. COSIMA : High resolution time-‐of-‐flight secondary ion mass spectrometer for the analysis of cometary dust particles onboard ROSETTA, In: Schultz, R., et al. (Eds.), ROSETTA : ESA's Mission to the Origin of the Solar System. Springer Science, pp. 201-‐242. Kissel, J., Brownlee, D.E., Buchler, K., Clark, B.C., Fechtig, H., Grun, E., Hornung, K., Igenbergs, E.B., Jessberger, E.K., Krueger, F.R., Kuczera, H., McDonnell, J.A.M., Morfill, G.M., Rahe, J., Schwehm, G.H., Sekanina, Z., Utterback, N.G., Volk, H.J., Zook, H.A., 1986a. Composition of comet Halley dust particles from Giotto observations. Nature 321, 336-‐337. Kissel, J., Sagdeev, R.Z., Bertaux, J.L., Angarov, V.N., Audouze, J., Blamont, J.E., Buchler, K., Evlanov, E.N., Fechtig, H., Fomenkova, M.N., von Hoerner, H., Inogamov, N.A., Khromov, V.N., Knabe, W., Krueger, F.R., Langevin, Y., Leonasv, B., Levasseur-‐ Regourd, A.C., Managadze, G.G., Podkolzin, S.N., Shapiro, V.D., Tabaldyev, S.R., Zubkov, B.V., 1986b. Composition of comet Halley dust particles from VEGA observations. Nature 321, 280-‐282. Krüger, H., Briois, C., Engrand, C., Fischer, H., Hilchenbach, M., Hornung, K., Kissel, J., Silén, J., Thirkell, L., Trieloff, M., Varmuza, K., the COSIMA Team, 2011. Rosetta/COSIMA : Principal Component Analysis applied to time-‐of-‐flight secondary ion mass spectra, EPSC-‐DPS2011, Nantes, p. 725. Kurat, G., Palme, H., Embey-‐Isztin, A., Touret, J., Ntaflos, T., Spettel, B., Brandstätten, F., Palme, C., Dreibus, G., Prinz, M., 1993. Petrology and geochemistry of peridotites and associated vein rocks of Zabargad Island, Red Sea, Egypt. Mineralogy and Petrology 48, 309-‐341. Lawler, M.E., Brownlee, D.E., 1992. CHON as a component of dust from Comet Halley. Nature 359, 810-‐812. Le Roy, L., Bardyn, A., Briois, C., Fray, N., Thirkell, L., Cottin, H., Hilchenbach, M., 2014. COSIMA calibration for the detection and characterisation of the cometary solid Planetary organic matter. (DOI: Science,, 10.1016/j.pss.2014.1008.1015). Lisse, C.M., VanCleve, J., Adams, A.C., A'Hearn, M.F., Fernandez, Y.R., Farnham, T.L., Armus, L., Grillmair, C.J., Ingalls, J., Belton, M.J.S., Groussin, O., McFadden, L.A., Meech, K.J., Schultz, P.H., Clark, B.C., Feaga, L.M., Sunshine, J.M., 2006. Spitzer spectral observations of the Deep Impact ejecta. Science 313, 635-‐640. Malfait, K., Waelkens, C., Waters, L.B.F.M., Vandenbussche, B., Huygen, E., de Graauw, M.S., 1998. The spectrum of the young star HD 100546 observed with the Infrared Space Observatory. Astron. Astrophys. 332, L25-‐L28. Messenger, S., 2002. Deuterium enrichments in interplanetary dust. Planet. Space Sci. 50, 1221-‐1225. RRUFF Project, (http://rruff.info/). Space and In Press 18 Siljeström, S., Lausmaa, J., Hode, T., Sundin, M., Sjövall, P., 2011. Structural effects of C60+ bombardment on various natural mineral samples—Application to analysis of organic phases in geological samples. Applied Surface Science 257, 9199-‐9206. Siljeström, S., Lausmaa, J., Sjövall, P., Broman, C., Thiel, V., Hode, T., 2010. Analysis of hopanes and steranes in single oil-‐bearing fluid inclusions using time-‐of-‐flight secondary ion mass spectrometry (ToF-‐SIMS). Geobiology 8, 37-‐44. Stephan, T., 2001. TOF-‐SIMS in cosmochemistry. Planet. Space Sci. 49, 859-‐906. Trieloff, M., Weber, H.W., Kurat, G., Jessberger, E.K., Janicke, J., 1997. Noble gases, their carrier phases, and argon chronology of upper mantle rocks from Zabargad Island, Red Sea. Geochim. Cosmochim. Acta 61, 5065-‐5088. Varmuza, K., Engrand, C., Filzmoser, P., Hilchenbach, M., Kissel, J., Krüger, H., Silén, J., Trieloff, M., 2011. Random projection for dimensionality reduction -‐ applied to time-‐ of-‐flight secondary ion mass spectrometry data. Analytica Chimica Acta 705, 48-‐55. Varmuza, K., Filzmoser, P., Hilchenbach, M., Krüger, H., Silén, J., 2014. KNN classification — evaluated by repeated double cross validation: Recognition of minerals relevant for comet dust. Chemometrics and Intelligent Laboratory Systems 138, 64-‐71. Vickerman, J.C., Briggs, D., (Eds), 2013. ToF-‐SIMS: Materials analysis by mass spectrometry. SurfaceSpectra, Manchester and IM Publications, Chichester, 732 pp. Witt-‐Eickschen, G., Seck, H.A., Mezger, K., Eggins, S.M., Altherr, R., 2003. Lithospheric Mantle Evolution beneath the Eifel (Germany): Constraints from Sr–Nd–Pb Isotopes and Trace Element Abundances in Spinel Peridotite and Pyroxenite Xenoliths. Journal of Petrology 44, 1077-‐1095. Zolensky, M.E., Zega, T.J., Yano, H., Wirick, S., Westphal, A.J., Weisberg, M.K., Weber, I., Warren, J.L., Velbel, M.A., Tsuchiyama, A., Tsou, P., Toppani, A., Tomioka, N., Tomeoka, K., Teslich, N., Taheri, M., Susini, J., Stroud, R., Stephan, T., Stadermann, F.J., Snead, C.J., Simon, S.B., Simionovici, A., See, T.H., Robert, F., Rietmeijer, F.J.M., Rao, W., Perronnet, M.C., Papanastassiou, D.A., Okudaira, K., Ohsumi, K., Ohnishi, I., Nakamura-‐Messenger, K., Nakamura, T., Mostefaoui, S., Mikouchi, T., Meibom, A., Matrajt, G., Marcus, M.A., Leroux, H., Lemelle, L., Le, L., Lanzirotti, A., Langenhorst, F., Krot, A.N., Keller, L.P., Kearsley, A.T., Joswiak, D., Jacob, D., Ishii, H., Harvey, R., Hagiya, K., Grossman, L., Grossman, J.N., Graham, G.A., Gounelle, M., Gillet, P., Genge, M.J., Flynn, G., Ferroir, T., Fallon, S., Ebel, D.S., Dai, Z.R., Cordier, P., Clark, B., Chi, M., Butterworth, A.L., Brownlee, D.E., Bridges, J.C., Brennan, S., Brearley, A., Bradley, J.P., Bleuet, P., Bland, P.A., Bastien, R., 2006. Mineralogy and petrology of comet 81P/Wild 2 nucleus samples. Science 314, 1735-‐1739. 19
1608.03074	1	1608	2016-08-10T08:16:44	The Venus Hypothesis	[ "astro-ph.EP", "physics.pop-ph" ]	Current models indicate that Venus may have been habitable. Complex life may have evolved on the highly irradiated Venus, and transferred to Earth on asteroids. This model fits the pattern of pulses of highly developed life appearing, diversifying and going extinct with astonishing rapidity through the Cambrian and Ordovician periods, and also explains the extraordinary genetic variety which appeared over this period.	astro-ph.EP	astro-ph	The Venus Hypothesis Author: A. Cartwright 1 Affiliations: 1School of Physics and Astronomy, Cardiff University *Correspondence to: [email protected] Abstract: Current models indicate that Venus may have been habitable. Complex life may have evolved on the highly irradiated Venus, and transferred to Earth on asteroids. This model fits the pattern of pulses of highly developed life appearing, diversifying and going extinct with astonishing rapidity through the Cambrian and Ordovician periods, and also explains the extraordinary genetic variety which appeared over this period. One Sentence Summary: Unanswered questions arising from the sudden appearance and radiation of complex life during the Cambrian and Ordovician periods, are answered by the proposal that these complex life forms evolved on Venus, and were transferred here by lithopanspermia. Habitability of Venus and transport of material from Venus to Earth The theory of lithopanspermia was introduced in the 1980s(1) and is now widely accepted. The head of NASA's search for life on Mars and Europa commented this year "There has almost certainly been exchange of biota between Earth and Mars. Mars is not very far away. Martian rocks from impact can easily reach Earth; probably Earth rocks also reached Mars from impacts in the past" (2). The routes by which life can have been transported between Earth, Mars and the moons of Jupiter have been modelled(3), and plants, bacteria and primitive animals such as tardigrades have been shown to survive the vacuum of space (4-6). The possibility that Venus might have been the source of life, which then travelled to Earth, seems not to be considered by those working in the field, probably because Venus is currently so inhospitable. It may not always have been so, and it is closer than Mars. Current 3d models of the atmosphere of Earth-sized planets orbiting closer to their stars have shown that the habitable zone extends closer to the stars than previously thought. The most recent research brings forward the habitability of Venus to at least 750My BP(7) and the authors speculate that life could have evolved. Current models of the interaction of crust and atmosphere on Venus, acknowledge that a habitable Venus cannot be ruled out (8). Isotope studies(9) also indicate the previous existence of liquid water on Venus. If Venus did evolve to have liquid water, and then develop plate tectonics and a carbon cycle as has been suggested (8), then an Earth-like world could have evolved, with the carbon in the atmosphere being absorbed by weathering and forming carbonate rocks in the crust. It is known that if plate tectonics was active early on a habitable Venus, then it must have ceased by the time the planet developed into the stagnant lid phase (10) with violent resurfacing occurring by about 600-400 My BP (11-14). These age estimates have an uncertainty factor of about 200%, as they are obtained from crater counting models(14). All of the Venusian crust is thought to have melted and then resolidified at that time. Any carbonate rocks existing before this happened would have been reprocessed in the mantle of Venus, releasing their CO2 just as happens on Earth, but on an enormous scale, leading to a runaway greenhouse effect and the evaporation of water, with the concomitant negative implications for life. The stagnant lid period would have seen increasing pressure beneath the crust(10) as heat transfer by conduction through the Venusian crust was no longer efficient enough to balance the radioactive heating within the planet's interior. Occasional volcanic eruptions and impacting asteroids would therefore have resulted in more explosive ejecta launch. This would be a perfect scenario to increase the amount of material being launched from Venus' crust to escape velocity and even to Earth-crossing orbit. The escape velocity for Venus is around 10km/s and an additional impulse of 22km/s would be required for an object to just reach Earth orbit (Hohman Transfer orbit calculation) with a transfer time of only 150 days. Recent models (8) suggest that the resurfacing event would have taken approximately 100 My to complete, indicating that it was taking place between 700My and 400 My BP. Cambrian explosion and Great Ordovician Biodiversity Event. Complex life appeared suddenly in Earth's fossil record with the Cambrian Explosion(15, 16) of 540My BP, when all but one of the known phyla appeared on Earth. This was followed by the Great Ordovician Biodiversity Event (GOBE) in the period up to 460My BP, when diversity greatly increased(17). Land plants also appeared around 470My BP (18). The timing is suggestively close to the resurfacing event on Venus. The nature and rapidity of the Cambrian explosion represent one of the major challenges in the study of evolution on Earth (19). Any model of the Cambrian explosion, and the GOBE, must explain the timing of this period – why did it start, and why did it stop? Why is there a lack of early fossil precursors of species which can be shown by genetic analysis to have evolved 1 billion years ago (20)? In particular, what could have caused all 34 known phyla to appear during the Cambrian or just shortly afterwards, with no new phyla ever evolving subsequently? Why was this period unique? The GOBE presents the additional challenge that timings of significant changes are 'diachronous across groups, environments and regions'(17). A systematic review of all of these problems and possible solutions(21) showed that none of the current environmental, developmental or ecological theories convincingly answer all of the questions. Much work is currently being focused on ocean chemistry and the increase in oxygen levels (22), acting as a trigger to sudden development and the rise of predation, but this is not conclusive at present. The sporadic arrival of complex life from Venus, during the lead up to and early stages of its violent resurfacing, is suggested here to provide an elegant solution to these outstanding questions. The timing of the arrival of material from Venus was dictated by the slowing and cessation of plate tectonics on that planet, leading to the stagnant lid state and eventual violent resurfacing(8, 10), possibly exacerbated by the break up of the L-chondrite parent asteroid(23). The lack of precursor fossils arises because the precursors were left behind on Venus. The geographical inconsistency of the GOBE, with some species suddenly flourishing in one area while elsewhere dying out, would be expected, as some meteorites might deliver new lifeforms, but others might have very destructive effects on established communities. The fantastic genetic diversity, much of which died out never to evolve again, would be predicted to arise on Venus due to its proximity to the Sun. Venus is subjected to twice the intensity of solar radiation and solar wind experienced on Earth, while the tidal derotation has led to very long day lengths(24), and hence a negligible magnetic field. Without the protection of a magnetosphere, life on Venus would have been bombarded by the Solar wind. Exposure to radiation leads to accelerated mutation rates(25) . If life evolved on Venus, therefore, we would predict that it would be wildly more exotic and further advanced than that evolving on Earth. It would be plausible for Venusian and Earth life to share common ancestry, through earlier periods of lithopanspermia (1, 2). Another key difference between Venus and Earth, leading to more genetic variety, would be the smaller Venusian water volume, estimated at only 10% of Earth's(7). If Venus therefore had more land and less ocean, then unlike Earth, with contiguous oceans and islands of land, Venus would be more likely to have contiguous land with large isolated bodies of water. We know from Earth's evolution that islands (New Zealand, Australia, Madagascar…) do not have to be isolated for very long before they develop very distinctive fauna. On Venus, this could have happened in the oceans. The increasing day length on Venus, extending perhaps to years then decades, would give a very definite evolutionary advantage to life evolving extended, deep hibernation states, or diapauses, with high resistance to extreme temperatures and levels of radiation. Some of our most ancient life forms, such as tardigrades(5), nematodes(6) and triops canciformis(26), exhibit just those features, and would therefore be well equipped to survive the transfer to Earth. Life on the dark, night side of Venus, most likely to be hit by impactors and then launch material to Earth, would be deeply dormant, and probably encased in ice. Ideally prepared for interplanetary travel. Evidence of impacts An explosive impact in upper New York state at 540My BP has been detected, indicated by high Iridium levels in the relevant rock formations. This has been explicitly linked to the Cambrian explosion (of life) (27). The Great Ordovician Biodiversity Event has been explicitly linked (28) to the break-up of the large L-chondrite asteroid at 470My BP(23, 29). Remnants of L chondrites embedded in limestone, as well as analysis of extra- terrestrial osmium and chromite levels, have demonstrated the precise coincidence between asteroid flux and biodiversification, in Scandinavia and China(28). The rate of impacts during this period is believed to have increased by 1 to 2 orders of magnitude(28). 5 large impact craters in North America and 6 in Scandinavia have been dated to the Ordovician period(30), and these impactors would have been accompanied by many times their volume of smaller pieces of debris, many of which may have landed in shallow water and survived impact(31). Asteroids do not burn up during their passage through the atmosphere. The outer layers burn away or spallate, but the insides remain cold. Medium sized blocks could have landed in water, perhaps splitting open as they finally bumped into the seabed, as happened with the 950kg Chelyabinsk meteorite(31), and releasing dormant life forms into a habitable new environment. These blocks would consist of rocks, including carbonates, from the crust of Venus. As well as craters and the chemical evidence of extraterrestrial impacts, rocks of the Ordovician period display 'odd lithologies and strange fauna… in China, unusual mini-mounds interrupt the normal succession'(28). Many limestone boulders are present in Scandinavia(32), and there are large Breccia fields throughout this period, which have again been explicitly linked with impacts(33, 34). I suggest that these carbonate boulders, mounds and breccia fields are remnants of Venusian asteroids, forming 'strewn fields' related to the many impact craters formed during the Ordovician. Note that the L-chondrite asteroid break up and subsequent migration of asteroid fragments to Earth-crossing orbits(23) would have led not only to bombardment of Earth, but also Venus. An asteroid fragment travelling on an elliptical orbit with perihelion at Venus orbit and aphelion in the asteroid belt would encounter Venus at a higher velocity than it would Earth. The extra kinetic energy of the Venusian impacts, exacerbated by the high pressure developed under the Venusian crust in the 'stagnant lid ' phase(10), would have had a more dramatic effect on Venus than Earth. Indeed, the L-chondrite break-up and suggested bombardment of Venus may even have contributed to or accelerated the final resurfacing of Venus. Venusian material ejected would be imprinted with traces of L-chondrite impactor material when it arrived at Earth. Chondrites, chromite and osmium signatures would still be found associated with some of the impact sites of Venusian material. Implications for the search for life. I have suggested that life evolved simultaneously on Earth and Venus. On Earth, it evolved very slowly while on the neighbouring, more highly irradiated planet, complexity and variety arose much earlier. That planet became uninhabitable, but, during that process, through violent volcanic eruptions and asteroid impacts, living complex lifeforms transferred to Earth. Here, with our protective magnetosphere, ozone layer and steady climate, life evolved at a more stately pace, but emerged from the sea and developed intelligence. This is an elegant and simple solution to the many complex questions arising from the Cambrian Explosion and GOBE, satisfying the demands of Occam's razor. It does, however, further complicate the search for life in other solar systems, and reduce the odds, if a two stage evolution model is adopted. References and Notes: 1. 2. 3. H. J. Melosh, The Rocky Road to Panspermia. Nature 332, 687-688 (1988). A. Mcewen, paper presented at the STARMUS, Tenerife, Spain, 2 July 2016 2016. R. J. Worth, S. Sigurdsson, C. H. House, Seeding Life on the Moons of the Outer Planets via Lithopanspermia. Astrobiology 13, 1155-1165 (2013). D. Tepfer, A. Zalar, S. Leach, Survival of Plant Seeds, Their UV Screens, and nptII DNA for 18 Months Outside the International Space Station. Astrobiology 12, 517-+ (2012). K. I. Jonsson, E. Rabbow, R. O. Schill, M. Harms-Ringdahl, P. Rettberg, Tardigrades survive exposure to space in low Earth orbit. Curr Biol 18, R729-R731 (2008). D. I. Shapiro-Ilan, I. Brown, E. E. Lewis, Freezing and Desiccation Tolerance in Entomopathogenic Nematodes: Diversity and Correlation of Traits. J Nematol 46, 27-34 (2014). M. J. D. G. Way et al https://arxiv.org/ftp/arxiv/papers/1608/1608.00706.pdf (2016) C. Gillmann, P. Tackley, Atmosphere/mantle coupling and feedbacks on Venus. J Geophys Res-Planet 119, 1189-1217 (2014). T. M. Donahue, J. H. Hoffman, R. R. Hodges, A. J. Watson, Venus Was Wet - a Measurement of the Ratio of Deuterium to Hydrogen. Science 216, 630-633 (1982). 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. N. H. Sleep, Evolution of the mode of convection within terrestrial planets. J Geophys Res-Planet 105, 17563-17578 (2000). R. J. Phillips et al., Impact Craters and Venus Resurfacing History. J Geophys Res-Planet 97, 15923-& (1992). R. G. Strom, G. G. Schaber, D. D. Dawson, The Global Resurfacing of Venus. J Geophys Res-Planet 99, 10899-10926 (1994). G. G. Schaber et al., Geology and Distribution of Impact Craters on Venus - What Are They Telling Us. J Geophys Res-Planet 97, 13257-13301 (1992). K. J. Zhanle, McKinnon, W.B., paper presented at the 1997 Chapman Venus Conference, Aspen, Colorado, 1997. H. B. Whittington, The Burgess Shale. (Published in association with the Geological Survey of Canada by Yale University Press, New Haven, 1985), pp. xv, 151 p. S. Conway Morris, The crucible of creation : the Burgess Shale and the rise of animals. (Oxford University Press, Oxford ; New York, 1998), pp. xxiii, 242 p. B. D. Webby, The great Ordovician biodiversification event. Critical moments and perspectives in earth history and paleobiology (Columbia University Press, New York, 2004), pp. x, 484 p. C. V. Rubinstein, P. Gerrienne, G. S. de la Puente, R. A. Astini, P. Steemans, Early Middle Ordovician evidence for land plants in Argentina (eastern Gondwana). New Phytol 188, 365-369 (2010). D. L. Bruton, Harry Blackmore Whittington, 1916-2010 Obituary. P Geologist Assoc 122, 533-535 (2011). D. Y. C. Wang, S. Kumar, S. B. Hedges, Divergence time estimates for the early history of animal phyla and the origin of plants, animals and fungi. P Roy Soc B-Biol Sci 266, 163-171 (1999). C. R. Marshall, Explaining the Cambrian "explosion" of animals. Annu Rev Earth Pl Sc 34, 355-384 (2006). D. Fox, What Sparked the Cambrian Explosion? Nature 530, 268-270 (2016). D. Nesvorny, D. Vokrouhlicky, A. Morbidelli, W. F. Bottle, Asteroidal source of L chondrite meteorites. Icarus 200, 698-701 (2009). J. Leconte, H. B. Wu, K. Menou, N. Murray, Asynchronous rotation of Earth-mass planets in the habitable zone of lower-mass stars. Science 347, 632-635 (2015). H. J. Muller, Genetic Damage Produced by Radiation. Science 121, 837-840 (1955). A. R. Longhurst, Evolution in the Notostraca. Evolution 9, 84-86 (1955). G. M. Friedman, Explosive bolide impact designates the Cambrian explosion, terminating the Cambrian event in New York. Carbonate Evaporite 22, 178-185 (2007). B. Schmitz et al., Asteroid breakup linked to the Great Ordovician Biodiversification Event. Nat Geosci 1, 49-53 (2008). E. V. Korochantseva et al., L-chondrite asteroid breakup tied to Ordovician meteorite shower by multiple isochron Ar-40-Ar-39 dating. Meteorit Planet Sci 42, 113-130 (2007). Eartj Impact Database,Planetary and Space Science Centre. (University of New Brunswick, 2016), accessed 8 August 2016. M. Gritsevich et al., Orbit, Trajectory, and Recovery of Chelyabinsk Meteorite. Meteorit Planet Sci 48, A146-A146 (2013). Kroger B, Hints L.,Lehnert L., Ordovician reef and mound evolution: the Baltoscandian picture. Geol Mag 152, 1-24 (2016). J. Parnell, Global mass wasting at continental margins during Ordovician high meteorite influx. Nat Geosci 2, 57-61 (2009). C. Alwmark, Shocked quartz grains in the polymict breccia of the Granby structure, Sweden-Verification of an impact. Meteorit Planet Sci 44, 1107-1113 (2009). Acknowledgments: I have been fortunate to be able to access the huge body of work by Birger Schmitz, not all of which can be cited here.
1806.11200	1	1806	2018-06-28T21:27:34	UV Surface Environments and Atmospheres of Earth-like Planets Orbiting White Dwarfs	[ "astro-ph.EP" ]	An Earth-like exoplanet orbiting a white dwarf would be exposed to different UV environments than Earth, influencing both its atmospheric photochemistry and UV surface environment. Using a coupled 1D climate-photochemistry code we model atmospheres of Earth-like planets in the habitable zone of white dwarfs for surface temperatures between 6000 K and 4000 K, corresponding to about 7 billion years of white dwarf evolution, as well as discuss the evolution of planetary models in the habitable zone during that evolution.	astro-ph.EP	astro-ph	UV Surface Environments and Atmospheres of Earth-like Planets Orbiting White Dwarfs Thea Kozakis, Lisa Kaltenegger Carl Sagan Institute, Cornell University, Ithaca, New York, USA and D. W. Hoard Eureka Scientific, Oakland, California, USA ABSTRACT An Earth-like exoplanet orbiting a white dwarf would be exposed to different UV environments than Earth, influencing both its atmospheric photochemistry and UV surface environment. Using a coupled 1D climate-photochemistry code we model atmospheres of Earth-like planets in the habitable zone of white dwarfs for surface temperatures between 6000 K and 4000 K, corresponding to about 7 billion years of white dwarf evolution, as well as discuss the evolution of planetary models in the habitable zone during that evolution. Subject headings: astrobiology, planets and satellites: atmospheres, planets and satellites: ter- restrial planets, stars: evolution, white dwarfs 1. Introduction Exoplanets have not yet been discovered orbiting white dwarfs, but have been found around pulsars, indicating that it is possible to have planetary bodies orbiting stellar rem- nants (Wolszczan & Frail 1992). Exoplanet searches are underway around white dwarfs (WDs) (e.g. Fulton et al. (2014); Fossati et al. (2015); Veras & Gansicke (2015); Xu et al. (2015)), as the similarity with Earth's size should make Earth-sized exoplanets tran- siting WDs easier to detect and character- ize than such planets around much larger main sequence stars (e.g. via atmospheric profile measurements from transit observa- tions, "weather" modeling from orbital light curves, direct measurement of atmospheric constituents with spectroscopy, etc.) (Agol 2011). Studies of close-by WDs with NASA's K2 mission (Howell et al. 2014) constrain the occurrence of Earth-sized habitable zone (HZ) planets in those systems to be < 28% (van Sluijs & Van Eylen 2017). Assuming that an Earth-like planet could form or survive around a WD, the WD cool- ing process will provide a changing luminos- ity as well as UV environment, which affects an orbiting planet's temperature, atmospheric photochemistry, and UV surface flux. The lu- minosity of a cool WD evolves slower than in 1 its initial hot phase, thus cool WDs could pro- vide stable environments for potentially hab- itable planets (Agol 2011). Several studies have suggested that the unique UV environ- ment would be high enough to sustain com- plex chemical processes necessary for Earth- like life, while not being strong enough to damage DNA (McCree 1971; Fossati et al. 2012) using estimates assuming present day Earth atmospheres. Multiple teams have shown that a fraction of WDs show evidence of recent heavy metal pollution, which could signal the existence of either disks or planets (e.g. Koester & Wilken (2006); Koester et al. (2014); Hamers & Portegies Zwart (2016); Klein et al. (2011); Malamud & Perets (2016)). For reviews of WD debris disks and pollution see Jura & Young (2014) and Farihi (2016), and for po- tential post-main sequence planetary evolu- tion see Veras (2016). Transit simulations using an unchanged present-day Earth-analog atmosphere compo- sition suggest that the depth of Earth's biosig- natures around a WD would be very strong and detectable by future missions such as JWST (Loeb & Maoz 2013). However a WD's spectral energy distribution (SED) can be significantly different from our present- day Sun. To model the effects of the WD's emitted flux on the atmosphere as well as the surface UV environment of an exoplanet, we use incident WD SED models at different points of a 0.6 M(cid:12)WD's evolution (Figure 1) in our models between 6000 K and 4000 K (Figure 2). We model planets orbiting WDs at three points in their evolution (Figure 2), with sur- face temperatures at 6000 K, 5000 K and 4000 K, during which time the WD's lumi- nosity does not change significantly com- pared to the early luminosity change in a WD's evolution (see Figure 1). That sur- face temperature change corresponds to about 7 billion years (Gyr) for a 0.6 M(cid:12)WD. We consider planets with eroded atmospheres as well as with higher surface pressure (e.g. super-Earths). We show the atmospheric structure and chemical composition as well as the UV surface fluxes at biologically rele- vant wavelengths. 2. Methods 2.1. Planet model: EXO-Prime We use EXO-Prime (see e.g. Kaltenegger & Sasselov (2010)); a coupled 1D radiative- convective atmosphere code developed for rocky exoplanets. The code is based on itera- tions of a 1D climate model (Kasting & Ack- erman 1986; Pavlov et al. 2000; Haqq-Misra et al. 2008), a 1D photochemistry model (Pavlov & Kasting 2002; Segura et al. 2005, 2007), which are run to convergence (see de- tails in Segura et al. (2005)). EXO-Prime models exoplanet atmospheres and environ- ments depending on the stellar and planetary conditions, including the UV radiation that reaches the surface and the planet's reflection, emission and transmission spectrum. We di- vide the atmosphere in 100 plane parallel lay- ers for our model up to an altitude of 60 km (or a pressure of 1 mbar) using a stellar zenith angle of 60 degrees. Visible and near-IR shortwave fluxes are calculated with a two stream approximation including atmospheric gas scattering (Toon et al. 1989), and longwave fluxes in the IR region are calculated with a rapid ra- diative transfer model (RRTM). A reverse- Euler method within the photochemistry code (originally developed by Kasting et al. (1985)) contains 220 reactions to solve for 55 chemical species. First, we scale the incident WD flux at the top of the model planetary atmospheres to the total integrated flux Earth receives from the Sun (Sef f) to model how the WD's irradiance 2 Fig. 1. -- White dwarf (WD) luminosity evolution calculated using cooling models from Bergeron et al. (2001) for a 0.6 M(cid:12) WD, showing (left) WD luminosity (LW D) versus age of a 0.6 M(cid:12) WD model and (right) the 1 AU equivalent orbital distance, where a planet would receive the same irradiance as Earth, for the same WD model. The three points (6000 K, 5000 K, and 4000 K) indi- cate the WD evolution for the planet models discussed. The inset panels show the same quantities plotted over a larger parameter ranges using the WD's surface temperature on the y-axis instead of the luminosity (see text). Fig. 2. -- Comparisons of the irradiance at the top of the atmosphere (ITOA) for planets orbiting a WD compared to the Earth-Sun system both at high resolution (dashed lines) and binned to the wavelength grid of the climate and photochemistry model resolution (solid lines). All spectra are scaled to the level of present day Earth's irradiance. changes the planetary environment compared to Earth. We then bin the high resolution spectra to the resolution of the wavelength grid of the climate/photochemistry models (both shown in Figure 2). We model plan- ets with surface pressures of 0.3 bar (e.g. 3 200300400500600700800Wavelength (nm)0.00.51.01.52.0Incident Energy (W/m2/nm)a)ITOA 6000 KITOA 5000 KITOA 4000 KITOA Sun150200250300350400Wavelength (nm)10-710-610-510-410-310-210-1100101Incident Energy (W/m2/nm)UVCUVBUVAb)ITOA 6000 KITOA 5000 KITOA 4000 KITOA Suneroded atmosphere), 1 bar (Earth analogue), and 1.5 bar and 2 bar. In our planetary models we keep the plan- etary outgassing rates constant for H2, CH4, CO, N2O, and CH3Cl and maintain the mix- ing ratios of O2 at 0.21 and CO2 at 3.55×10−6 to be able to compare the effect of the irra- diation on the planetary atmosphere, with a varying N2 concentration that is used as a fill gas to reach the set surface pressure of the model (following Segura et al. (2003, 2005); Rugheimer et al. (2013, 2015,b); Rugheimer & Kaltenegger (2018)). Note that by keeping the outgassing rates constant, lower surface pressure atmosphere models initially have slightly higher mixing ratios of chemicals with constant outgassing ratios than higher surface pressure models. 2.2. White Dwarf Model Spectra WDs are unique stellar environments with very high surface gravities and extremely dense interiors and physical conditions. They have no internal heat source, and thus cool off over time (Figure 1). We use WD spectral energy distribution models (SED) calculated as described in Saumon et al. (2014) (Fig- ure 2) as irradiation for the planetary mod- els (see e.g. Bergeron et al. (1997); Kowal- ski & Saumon (2006); Kilic et al. (2009a,b); Giammichele et al. (2012); Saumon et al. (2014) for detailed discussion on WD spec- tra). The WD models assume a pure H atmo- spheric composition with a surface gravity of log g = 8.0 in creating the models for the average mass of WD in the field (0.6 M(cid:12); e.g. Kepler et al. (2016)). The WD radius is then recovered by finding the intersection of the standard surface gravity formula with the WD mass-radius-(temperature) relation for C-O core WDs with thick H layers dis- cussed by Parsons et al. (2017)(see their Fig- ure 9; also see Benvenuto & Althaus (1999); Fontaine et al. (2001); Parsons et al. (2017) extrapolated to T<10,000 K. The radius of the WD in our model is 0.0128±0.0001 R(cid:12) for surface temperatures of 6000 K to 4000 K. WD spectra are similar to black bodies with only Balmer absorption lines for surface tem- peratures greater than 5000 K, where hydro- gen becomes neutral as shown in Figure 2. We use models from Bergeron et al. (2001) for the temperature evolution of the WD (see Figure 1). 2.3. Photochemistry of some biologically interesting species Some atmospheric species exhibit notice- able features in our planet's spectrum as a re- sult directly or indirectly from biological ac- tivity. The main ones are oxygen (O2), ozone (O3), methane (CH4), nitrous oxide (N2O) and methyl chloride (CH3Cl) (see e.g. De- marais et al. (2012); Kaltenegger (2017)). We summarize the most important reactions that influence these species in Earth's atmosphere here. Many reactions in Earth's atmosphere are driven by the Sun's UV flux. Ozone and O2 are created with UV photons through the Chapman reactions (Chapman 1930), O2 + hν → O + O (λ < 240 nm), O + O2 + M → O3 + M, O3 + hν → O2 + O (λ < 320 nm), O3 + O → 2O2, (1) where M is a background molecule such as N2. These reactions are primarily responsible for ozone production on present day Earth. A higher UV flux additionally increases the primary source reactions of tropospheric hy- droxyl (OH) for 300 < λ < 320 nm in the troposphere, O3 + hν → O(1D) + O2, O(1D) + H2O → OH + OH, (2) reducing O3 and H2O. Increased OH is the primary sink for H2, CH4, CH3Cl, and CO abundances. 4 Methane (CH4) is a reducing gas that has a lifetime of 10-12 years in present day Earth's atmosphere (Hougton et al. 2004) due to reac- tions with oxidizing species. It has both nat- ural (termites, wetlands) and anthropogenic sources (rice agriculture, natural gas). It is oxidized via (3) (4) CH4 + 2O2 → CO2 + 2H2O, creating H2O and H2 via photolysis, CH4 + hν → CH2 + H2. Its main sink is due to the reaction, OH + CH4 → CH3 + H2O, (5) in the troposphere, and photolysis in the stratosphere. This reaction also creates H2O in the stratosphere, above where it is shielded by the ozone layer. On Earth both nitrous oxide (N2O) and methyl chloride (CH3Cl) are primarily pro- duced by life, and are depleted by higher amounts of UV. N2O is a greenhouse gas that is extremely effective when trapping heat. It is produced naturally in soil through nitrifica- tion and denitrification, and its anthropogenic source is from agriculture. Photolysis of N2O occurs for λ < 220 nm, and is its main sink on present day Earth. Increased O3, e.g. due to UV irradition, is also a sink of N2O, (6) N2O + O(1D) → 2NO, creating NO. It depletes ozone via NO + O3 → NO2 + O2. (7) On Earth CH3Cl is naturally produced in oceans via light interacting with sea foam chlorine and biomass and in small amounts by phytoplankton. CH3Cl reacts with OH creating Cl, a component of chlorofluorocar- bons which damage the ozone layer. It is converted into chlorine via the reactions, CH3Cl + OH → Cl + H2O, CHCl + hν → CH +Cl, CH3Cl + Cl → HCl + Cl. (8) 5 3. Results 3.1. Time evolution of a white dwarf's habitable zone During the cooling process of a WD, its surface temperature as well as its overall lu- minosity decreases. This, in turn, influences the orbital distance of its HZ. The HZ is the circular region around one or multiple stars in which standing bodies of liquid water could exist on a rocky planet's surface (e.g. Kasting et al. (1993); Kaltenegger & Haghighipour (2013); Haghighipour & Kaltenegger (2013); Kopparapu et al. (2013, 2014); Ramirez & Kaltenegger (2017)) and facilitate the detec- tion of possible atmospheric biosignatures (see e.g. Kaltenegger (2017)). The classical, conservative N2-CO2-H2O HZ is defined by the greenhouse effect of two gases: CO2 and H2O vapor. The inner edge corresponds to the distance where mean surface temperatures exceed the critical point of water (∼647 K, 220 bar), triggering a runaway greenhouse that leads to rapid water loss to space on very short timescales (see Kasting et al. (1993) for details). Toward the outer edge of the classical HZ weathering rates decrease, al- lowing atmospheric CO2 concentrations to increase as stellar insolation decreases. At the outer edge, condensation and scattering by CO2 outstrips its greenhouse capacity, the so-called maximum greenhouse limit of CO2. The cooling of a WD translates into an in- ward shift of the orbital distances of the WD HZ, (Figure 3a). In addition to the decrease in overall irradiance (Sef f) as the WD cools, some of the change in the HZ orbital distance is also caused by the shift of the peak wave- length of emission of the WD SED to redder wavelengths, which heat the surface of the planet more efficiently than bluer light (e.g. Kasting et al. (1993)). A 0.6 M(cid:12)WD spends about 7 Gyr cooling from 6000 K to 4000 K, providing a phase during which its luminos- ity does not rapidly change and thus it could provide temperate conditions for an orbiting planet. The size of the WD HZ is shown in Figure 3a as well as the corresponding irradi- ance of a planet, normalized to the value for Earth (Sef f) shown in Figure 3b. Because of the small size of a WD compared to the Sun, the WD HZ is a factor of ∼100 to ∼1000 times closer to the WD than Earth is to the Sun. An alternative HZ limit that is not based on atmospheric models (like the classical HZ) but on empirical observations of our Solar System is also shown in Figure 3 for com- parison. The inner edge of the empirical HZ is defined by the stellar flux received by Venus when we can exclude the possi- bility that it had standing water on the sur- face (about 1 Gyr ago), equivalent to a stel- lar flux of Sef f = 1.77, corresponding to a distance of 0.75 AU for Earth's current So- lar flux (Kasting et al. 1993). The "early Mars" limit is based on observations suggest- ing that the Martian surface may have sup- ported standing bodies of water ∼3.8 Gyr ago, when the Sun was only 75% as bright as today. For our solar system, Sef f = 0.32 for this limit, corresponding to a distance of ∼1.77 AU (e.g. Kasting et al. (1993)). We first explore the range of conditions for Earth-like planets orbiting WDs of different temperatures seen at one point in time (see Sections 3.2 and 3.3). Then we model two case studies, A and B, as shown in Figure 3, which explore the environment of a planet in the HZ of a WD as it cools (see Section 3.4). 3.2. Planetary photochemistry environ- ments for different stages in a WD's evolution The lack of chromospheric activity for a WD at different stages in its evolution causes photochemical differences in the at- mospheres of planets orbiting it compared 6 to Earth. We modeled planets with different surface pressures, which receive an equiva- lent total irradiance as Earth from the Sun (Sef f) from a WD at three different stages in its evolution: for WD surface tempera- tures of 6000 K, 5000 K and 4000 K. We model planets with surface pressures rang- ing from 2 bar to 0.3 bar. Table 1 summa- rizes the model planet surface temperature and integrated overall ozone column depth (ozone column depth) for the different plan- etary models as well as our Earth model for comparison. Figure 4 shows the changes in temperature as well as the mixing ratio for O3, CH4, H2O, CH3Cl and N2O for the dif- ferent models, compared to Earth. Figure 4 shows that in the model atmo- spheres H2O photolysis increases with higher UV levels in the upper atmosphere, where H2O is not shielded from incoming photons below the ozone layer. N2O is depleted by photolysis, with a decreasing mixing ratio to- ward the top of the atmosphere, but remains well mixed beneath the ozone layer, where it is shielded. All higher surface pressure mod- els show increased amounts of CH4, N2O, and CH3Cl in the upper atmosphere. At- mospheric H2O increases with higher surface temperatures. The WD cooling process provides a chang- ing luminosity as well as UV environment at a set orbital distance, as seen in Figure 1. The decrease in UV incident flux leads to an over- all decrease in the ozone level. Compared to present day Earth's integrated overall ozone column depth, a higher surface pressure at- mosphere receiving similar UV irradiance has a higher ozone column depth, as seen in the values for our planet models with surface pressures above 1 bar orbiting a WD with a surface temperature of 6000 K. Table 1 shows the absolute values for our models as well as present day Earth's. Fig. 3. -- Size of a WD's HZ (left) as well as the corresponding range of WD irradiance, normalized to the value for Earth (Sef f) (right). The shaded green region represents the classical HZ for our WD model (following Kopparapu et al. (2014)). The red (recent Venus) and blue (early Mars) lines show the limits of the empirical HZ (following Ramirez & Kaltenegger (2017)). Two case studies for a planet in the WD's HZ, Case A and B, are indicated in both panels. The black dots indicate the specific WD surface temperatures and planetary distances/Sef f values we model in this study. Table 1: Model summary for Earth-equivalent irradiance Host Teff (K) Present day Earth 6000 6000 6000 6000 5000 5000 5000 5000 4000 4000 4000 4000 Pressure Surface Ozone Column Teff (K) Depth (cm−2) 288.2 280.7 285.6 288.0 289.0 282.4 290.8 295.0 298.1 282.8 294.3 300.4 305.1 5.4×1018 3.6×1018 5.7×1018 6.5×1018 6.9×1018 2.6×1018 3.9×1018 4.2×1018 4.5×1018 1.1×1018 1.8×1018 2.1×1018 2.3×1018 (bar) 1.0 0.3 1.0 1.5 2.0 0.3 1.0 1.5 2.0 0.3 1.0 1.5 2.0 3.2.1. WD photochemistry: Earth-analogue: 1 bar surface pressure models Planet models with a 1 bar surface pres- sure, analogous to Earth, orbiting WDs show 7 0.0000.0050.0100.0150.0200.025Distance from WD (AU)35004000450050005500600065007000Host Teff (K)Case BCase AEarly MarsRecent Venus0.51.01.52.0Seff35004000450050005500600065007000Host Teff (K)Case ACase BEarly MarsRecent VenusFig. 4. -- Temperature and chemical mixing ratios for some biological interesting species for model planets orbiting a WD (blue solid lines) with 3 different surface temperatures (6000 K, 5000 K and 4000 K). Earth's values are shown for comparison (black dashed line). surface temperatures 285.6 K, 290.8 K, 294.3 K for WD surface temperatures of 6000 K, 5000 K and 4000 K respectively. However, the different WD UV environ- ments compared to the Sun's lead to a 5% in- crease in ozone column depth for the 6000 K WD surface temperature, and a 28% and 67% decrease of ozone column depth for the 5000 K and 4000 K WDs compared to Earth. Higher ozone levels cause a higher rate of the R2 reaction, producing OH. Increased amounts of OH for higher UV levels depletes CH4 via the R5 reaction, and similarly CH3Cl via the R8 reaction (see Figure 4). 3.2.2. WD photochemistry eroded atmo- spheres: 0.3 bar surface pressure models Planet models with a 0.3 bar surface pres- sure, analogous to eroded atmospheres, orbit- ing WDs show lower surfaces temperatures than 1.0 bar models, increasing with decreas- ing WD surface temperature from 280.7 K to 282.8 K, respectively. The different UV WD environments, the decreased overall amount of oxygen, as well as the decreased density of 8 150200250300T (K)1020304050T10-710-610-5O3O310-1010-910-810-710-610-5CH4CH410-610-510-410-3H2OH2O10-1010-810-6N2ON2O10-2110-2010-1910-1810-1710-1610-1510-1410-1310-1210-1110-1010-910-8CH3Cl0.3 barCH3Cl150200250300T (K)102030405010-710-610-5O310-910-710-5CH410-610-510-410-3H2O10-1010-810-6N2O10-2110-2010-1910-1810-1710-1610-1510-1410-1310-1210-1110-1010-910-8CH3Cl1.0 bar150200250300T (K)1020304050 Altitude (km)10-710-610-5O310-910-710-510-610-510-410-3H2O10-1010-810-6N2O10-1910-1610-1310-10CH3Cl1.5 bar150200250300T (K)102030405010-710-610-510-910-710-510-610-510-410-3Mixing Ratios10-1010-810-610-1910-1610-1310-102.0 bar6000 K WD5000 K WD4000 K WDSunthe atmosphere lead to a 33%, 52%, and 80% decrease in ozone column depth compared to Earth. Longer mean-free-paths due to lower atmospheric density result in photolysis oc- curring at lower altitudes, causing ozone to form closer to the ground (see Figure 4). 3.2.3. WD photochemistry: Higher surface pressure planets: 1.5 & 2.0 bar sur- face pressure models The 1.5 bar surface pressure models, for higher surface pressure planets, orbiting WDs show surfaces temperatures of 288.0 K, 295.0 K, 300.4 K for WD surface temper- atures of 6000 K, 5000 K and 4000 K re- spectively. The different UV environments compared to the Sun and the increase in over- all oxygen content as well as density of the atmosphere lead to a 20% increase in ozone column depth for the 6000 K WD surface temperature, and a 22% and 61% decrease in ozone column depth compared to Earth for the 5000 K and 4000 K WDs, respectively. Planet models with a 2 bar surface pres- sure show surfaces temperatures of 289.0 K, 298.1 K, 305.1 K for WD surface tempera- tures of 6000 K, 5000 K and 4000 K respec- tively. The different UV environments com- pared to the Sun and the increase in overall oxygen content as well as density of the at- mosphere lead to a 28% increase in ozone column depth for the 6000 K WD surface temperature, and a17%, and 57% decrease in ozone column depth compared to Earth, less than for the 1.5 bar surface pressure case. 3.3. UV surface environment around an evolving WD The WD cooling process provides a chang- ing luminosity as well as UV environment at a set orbital distance, as seen in Figure 2. The UV surface environment for our planetary models from eroded atmospheres to dense atmospheres with increased surface pressure 9 is shown in Figure 5, with integrated fluxes for UVA (315-400 nm), UVB (280-315 nm), and UVC (121.6-280 nm) are shown in Ta- ble 2, with comparisons to the Earth-Sun system's integrated fluxes in Table 3. Note that this model does not take scattering or clouds into account and thus overestimates the amount of UV that reaches the surface. However the comparison between the val- ues and models for the UV environment on present-day Earth gives a clearer picture of the level of UV radiation that reaches the ground, compared to our own planet. High energetic UV is capable of causing harm to biological molecules, like DNA (e.g. Voet et al. (1963); Diffey (1991); Matasunaga et al. (1991); Tevini (1993); Cockell (1998); Ker- win & Remmele (2007)). Present day Earth surface life is protected by the ozone layer, which shields the surface from the most bio- logically dangerous radiation (UVC). 3.3.1. Surface UV environments: Earth 1 bar surface pressure models For comparison we first model the amount of radiation that reaches the Earth's surface (see also Rugheimer et al. (2015)). For our present-day Earth model the integrated UVA ground flux compared to the UVA ITOA flux is 70%. For integrated UVB flux, which is partially shielded by ozone, 11% of the ITOA UVB flux reaches the surface. The UVC flux is almost completely shielded by an ozone layer, and only 5.4×10−18% of the ITOA in- tegrated UVC flux reaches the surface (see Table 5). 3.3.2. WD surface UV environments: Earth- analogue: 1 bar surface pressure models For our 1 bar surface pressure planet mod- els, the amount of UVA flux at the surface compared to the ITOA integrated flux in- creases from 70% to 76% for the 6000 K to 4000 K WD surface temperature cases re- spectively, compared to the 70% for present day Earth models. For UVB, it increases from 11%, to 26% of the ITOA integrated UVB for the 6000 K to 4000 K WD surface temperature cases respectively, compared to 11% for Earth models. The UVC flux is al- most completely shielded by an ozone layer, and is only 1.0×10−18% to 1.2×10−5% of the ITOA integrated UVC flux, for the 6000 K to 4000 K WD surface temperature cases respectively, compared to 5.4×10−18% on present day Earth. 3.3.3. WD surface UV environments: Eroded atmospheres: 0.3 bar surface pres- sure models For our 0.3 bar surface pressure planet models, the amount of integrated UVA sur- face flux compared to the ITOA integrated flux increases from 93% to 98% for the 6000 K to 4000 K WD surface temperature cases, compared to 70% for present day Earth models. For integrated UVB flux there is an increase from 23%to 48% of the ITOA UVB for the 6000 K to the 4000 K WD sur- face temperature cases, higher than the 11% for present day Earth. The total integrated UVC surface flux increases from 1.0×10−9% to 4.3×10−3% of the ITOA flux for the 6000 K to the 4000 K WD surface temper- ature cases, orders of magnitude higher than the 5.4×10−18% for Earth. 3.3.4. WD surface UV environments: higher surface pressure planets: 1.5 bar and 2 bar surface pressure models For our 1.5 bar surface pressure planet models, the amount of integrated surface UVA flux compared to the ITOA integrated flux increases from 60% to 65% for the 6000 K through 4000 K WD surface tem- perature cases, less than the 70% for present day Earth models. For UVB surface flux 10 there was an increase of 7.6% to 19% of the ITOA integrated UVB flux for the 6000 K to 4000 K WD surface temperature cases, com- parable to 11% for Earth models. The UVC ground flux increased from 1.8×10−21% to 9.9×10−7% of the ITOA integrated flux for the 6000 K to the 4000 K WD surface tem- perature cases, respectively, compared to 5.4×10−18% for present day Earth. For our 2 bar surface pressure planet mod- els, the amount of integrated UVA ground flux compared to the ITOA integrated flux increases from 53% to 58% for the 6000 K to 4000 K WD surface temperature cases, respectively, less than 70% for present day Earth models. For UVB integrated ground flux there is an increase of 6.1% to 15% of ITOA UVB integrated flux reaching the sur- face for the 6000 K through 4000 K WD sur- face temperature cases respectively, compa- rable to 11% for present day Earth. The in- tegrated UVC flux to the ground increases from 9.6×10−23% to 1.4×10−7% of the orig- inal integrated ITOA flux for the 6000 K to 4000 K WD surface temperature cases, respectively, compared to 5.4×10−18% for Earth models. The planetary models with the 4000 K WD surface temperature host show the high- est overall UVC ground flux despite having a lower UVC ITOA integrated flux, because of the lower ozone level in these atmospheres, compared to hotter WD models. Overall, Figure 5 shows that the UV sur- face environment for our model planets orbit- ing WDs. The 1 bar surface pressure model for a 6000 K WD surface temperature model receives similar UV surface levels as present day Earth models. Only the 1.5 bar and 2 bar surface pressure models for the same 6000 K WD surface temperature receive a lower UV integrated surface flux than present day Earth. For all other models, the UV surface flux is higher than for present day Earth, especially Fig. 5. -- Surface UV environment for planetary models with higher surface pressure (2 bar) to small planets or planets with eroded atmospheres (0.3 bar) orbiting WDs (solid lines) compared to present day Earth (dashed lines), and the incident (top-of-atmosphere) irradiation for the WD (black solid line). the UVC environment (see Table 3). 3.4. Planetary environments for planets in the HZ through the evolution of a WD We model two case studies, A and B, as shown in Figure 3, which explore the envi- ronment of a planet with a 1 bar surface pres- sure (i.e., an Earth analogue) in the HZ of a WD during its evolution. Case A shows the maximum time a planet can stay in the HZ of the WD as the WD cools from 6000 K to 4000 K. This amounts to ∼6 Gyr for the clas- sical (conservative) HZ and ∼8.5 Gyr for the empirical HZ. Case B focuses on a planet that initially receives the same irradiance as Earth around a WD with a surface temperature of 5000 K. As the WD cools from 6000 K to 4000 K the planet spends ∼4 Gyr in the con- servative HZ and ∼7 Gyr in the empirical HZ. Model parameters and results are shown in Tables 4 and 5, with comparisons to Earth in Table 6 for both cases. UV environments are shown in Figure 6 for Case A and Figure 7 for Case B, and photochemistry is shown in Figure 8. Such a planet could have orbited in the HZ of a cool WD for longer than the Earth has existed. Single-celled life likely emerged on Earth less than 1 Gyr after its formation, with multicellular life following 2.7 Gyr later. 3.4.1. Case A The orbital distance of 0.0085 AU or 1.3 million km leads to a changing illumina- tion by the cooling WD from Sef f of 1.34 for a 6000 K WD surface temperature, to Sef f of 0.64 for a 5000 K WD surface temper- ature, to Sef f of 0.26 for the 4000 K WD surface temperature model. This leads to de- creasing planetary surfaces temperatures of 328.4 K, 249.6 K, 191.9 K for WD surface temperatures of 6000 K, 5000 K and 4000 K, respectively. In our model the planet's surface temper- ature is on average above freezing for only a part of the time the planet spends in the WD's HZ, a few billion years (Figure 1 and Figure 3). However we have not adjusted the CO2 content in our models, thus if a cycle similar to the carbonate silicate cy- cle on Earth existed on such a planet, CO2 concentration should increase, heating the planet's surface temperature and keeping it from freezing. As shortly discussed, a geo- logically active planet is the underlying as- 11 200250300350Wavelength (nm)10-3510-3210-2910-2610-2310-2010-1710-1410-1110-810-510-2101104Incident Energy (W/m2/nm)UVCUVBUVA6000 K WDTOA0.3 bar1.0 bar1.5 bar2.0 barEarthI250300350Wavelength (nm)UVCUVBUVA5000 K WDTOA0.3 bar1.0 bar1.5 bar2.0 barEarthI250300350400Wavelength (nm)UVCUVBUVA4000 K WDTOA0.3 bar1.0 bar1.5 bar2.0 barEarthIsumption of continued habitability during a star's evolution, and is what led to the concept of the HZ (see e.g. Kasting et al. (1993)). For a WD, a similar cycle could be possible, ex- tending the time the surface of such a planet can be above freezing to the full range of its time in the WD's HZ. The different WD UV environments com- pared to the Sun lead to a 29%, 18%, and 82% decrease in overall ozone column depth compared to Earth. As shown in Figure 8, ozone levels are highest for a WD model of 5000 K, providing a slightly higher shield- ing from UV than present day Earth. How- ever, during both the 6000 K and 4000 K WD surface temperature stages, the UVC flux at the surface increases substantially for differ- ent reasons. During the 6000 K WD sur- face temperature period, UV photons with λ < 320 [R1] are available. These can dis- sociate ozone, therefore increasing the UV surface flux. The 4000 K WD model has a comparably high amount of surface UVC ra- diation because low UV irradiation from the 4000 K WD cannot initially produce enough ozone to efficiently shield the surface of the planet. The amount of UVA integrated surface flux compared to the ITOA integrated flux is 71%, 74%, and 78%, for the 6000 K, 5000 K, and 4000 K WD surface temperature cases, comparable to the 70% for model Earth mod- els. For UVB flux, which is partially shielded by ozone, 24%, 15%, and 32% of the in- tegrated ITOA UVB reach the surface, for the 6000 K, 5000 K, and 4000 K WD sur- face temperature cases, compared to 11% for present day Earth models. The UVC flux is almost completely shielded by an ozone layer, and the percentage of the integrated ITOA flux reaching the planetary surface 1.0×10−5%, 2.4×10−14%, and 1.0×10−3%, for the 6000 K, 5000 K, and 4000 K WD surface temperature cases, respectively, com- 12 pared to 5.4×10−18% for present day Earth models. 3.4.2. Case B The orbital distance of the planet is 0.0069 AU in Case B. This corresponds to an illumina- tion of 1.0 Sef f for the 5000 K WD, which evolves to 0.41 Sef f as the WD cools to 4000 K. For Case B planet models the sur- faces temperatures are 290.8 K and 216.9 K for WD surface temperatures of 5000 K and 4000 K, respectively. The different UV envi- ronments compared to the Sun lead to a 39% and a 77% decrease in overall ozone column depth for WD surface temperatures of 5000 K and 4000 K, respectively, compared to Earth. The amount of UVA ground flux compared to the ITOA integrated flux is 73% and 77%, for the 5000 K and 4000 K WD surface tem- perature models, respectively (see Figure 7), comparable for the 70% for present day Earth models. For UVB flux, which is partially shielded ozone, only 16%, and 36% of the integrated ITOA UVB reaches the surface, for the 5000 K and 4000 K WD surface tem- perature cases, compared to 11% for present day Earth models. The UVC flux is al- most completely shielded by an ozone layer, and is 1.1×10−12%, and then 1.1×10−2% of the ITOA integrated UVC flux, compared to 5.4×10−18% for present day Earth models. As the UV levels from the WD decrease, less ozone is produced in the planet's atmo- sphere and thus UVB and UVC surface lev- els increase. Average surface temperature de- creases below freezing for outgassing rates similar to present day Earth. However, as discussed for Case A, if a carbonate-silicate cycle exists, increased amounts greenhouse gases should increase surface temperatures above freezing for an Earth-like planet even with decreasing illumination by the cooling WD. Fig. 6. -- UV surface flux for Case A with a planet orbiting at 0.0085 AU from its host, corre- sponding to Sef f = 1.34 for for the 6000 K model, Sef f = 0.64 for a 5000 K model, and Sef f = 0.26 for 4000 K models. 13 150200250300350400Wavelength (nm)10-710-610-510-410-310-210-1100101erg/cm2/s/UVCUVBUVA6000 K WDSun200250300350400Wavelength (nm)10-3510-3210-2910-2610-2310-2010-1710-1410-1110-810-510-2101Incident Energy (W/m2/nm)UVCUVBUVATOA 6000 K WDTOA Sun6000 K WDSunII150200250300350400Wavelength (nm)10-710-610-510-410-310-210-1100101erg/cm2/s/UVCUVBUVA5000 K WDSun200250300350400Wavelength (nm)10-3510-3210-2910-2610-2310-2010-1710-1410-1110-810-510-2101Incident Energy (W/m2/nm)UVCUVBUVATOA 5000 K WDTOA Sun5000 K WDSunII150200250300350400Wavelength (nm)10-710-610-510-410-310-210-1100101erg/cm2/s/UVCUVBUVA4000 K WDSun200250300350400Wavelength (nm)10-3510-3210-2910-2610-2310-2010-1710-1410-1110-810-510-2101Incident Energy (W/m2/nm)UVCUVBUVATOA 4000 K WDTOA Sun4000 K WDSunII4. Discussion 4.1. How could white dwarf planets form? The mechanisms required for a WD plan- etary system to form as secondary genera- tion objects from a disk or to survive post- main sequence evolution are not well under- stood. During the post main sequence evo- lution of the host into a WD, inner rocky planets (within 1-2 AU) would likely be de- stroyed (e.g. Villaver (2011); Kunitomo et al. (2011); Villaver (2012); Mustill & Villaver (2012); Villaver et al. (2014)), while stellar mass loss would cause semimajor axis ex- pansion for outer planets (e.g.Veras (2016); Ramirez & Kaltenegger (2016)). Exomoons of outer planets could potentially migrate in- ward (Veras 2016; Ramirez & Kaltenegger 2016), although the estimated occurrence rate wouldn't explain the rate of polluted WDs (van Sluijs & Van Eylen 2017; Van Eylen et al. 2017). Second-generation planets could even form from fall-back of debris initially expelled during the post-main sequence evo- lution of the host star (Veras 2016). An inter- esting system is WD1145+017, a WD orbited by a rocky minor planet undergoing tidal dis- integration at the system's Roche limit (Van- derburg et al. 2015). 4.2. Only dry planets? White dwarfs start with an extremely hot phase With no stellar mass loss during the WD phase (and assuming rapid tidal circulariza- tion of its orbit), a planet should remain at the same orbital distance during a WD's evo- lution. Due to the extreme change in lumi- nosity early in a WD's evolution (Figure 1) a temperate planet around a cool WD would have thus experienced an extremely high lu- minosity and corresponding surface tempera- ture early in its history. This would initiate a runaway greenhouse process and extreme water loss on such planets early in their his- 14 tory. Similarly for planets orbiting pre-main sequence M-type stars, which also are more luminous initially and should be able to ini- tiate a runaway greenhouse phase and water loss on a temperate planet that can be found in the main sequence HZ later on (see Ramirez & Kaltenegger (2014); Barnes et al. (2015)), there may be methods of late water delivery occurring in a WD system after to the WD has cooled to a surface temperature that main- tains a slower changing luminosity (e.g. Jura & Xu (2010); Farihi et al. (2013); Malamud & Perets (2016, 2017a,b)). Any Earth-like planet that is orbiting too close to the WD would remain in a runaway greenhouse stage for a certain time, until it entered the WD HZ. Whether or not such planets would have lost all their water at that point would depend strongly on the initial water reservoir, and whether it can be replen- ished. Thus, whether a WD planet can host water and an Earth-like atmosphere, is an open question and will depend on its evolu- tion as well as when the WD planet is formed, from what materials it is made of and whether the possibility of a continuous water delivery exists in WD planetary systems. 4.3. A white dwarf planet's evolution dif- fers from that of Earth A planet orbiting a WD will receive de- creasing overall energy from its host during the WD's cooling process. Its evolution is very different from that of a planet orbiting a main sequence star, whose luminosity in- creases with time, pushing the HZ to wider separations. For a planet around a main se- quence star, the increase in greenhouse gases at the inner edge of the HZ coupled with the increasing luminosity of the host star limit the time it can remain habitable. If a cycle similar to the carbonate silicate cycle could also operate on WD planets, planets on the inner edge of a WD could build up substan- tial greenhouse gas amounts in their atmo- spheres as well; however due to the decreas- ing luminosity of their WD host, such build up of greenhouse gases could help to main- tain warm surface temperature as the WD flux decreases, possibly extending the length of time life could survive on WD planets. 5. Conclusions Our models explore the atmospheric en- vironments of planets orbiting WDs, taking into consideration the changing surface tem- peratures and UV environments of WDs dur- ing their cooling process. We model the atmospheric composition as well as the UV surface environments of Earth-like planets orbiting WDs during dif- ferent points throughout a WD's evolution. Our planet models have surface pressures ranging from 2 bar to 0.3 bar, including Earth-analog planets with 1 bar atmospheres, as well as planets with 0.3 bar surface pres- sure (e.g. eroded atmospheres) and planets with higher surface pressures of 1.5 bar and 2 bar (e.g, super-Earths). The integrated overall ozone column depth is less than on present day Earth for all our model runs, except for the models with sur- faces pressures of 1 bar or above for the WD 6000 K surface temperature model, which provides similar UV to the Sun. The UV surface environment on a planet is controlled primarily by the incoming irradiation and by the planetary atmospheric composition. The UV surface environment for all our planetary models orbiting WDs show increased surface UVC flux up to several orders of magnitude compared to present day Earth, except for the models with surface pressures of 1 bar or above for the 6000 K WD model (see Table 5). The UVC can become substan- tially higher for the cool 4000 K WD surface temperature model runs, making those model 15 planet surface environments harsh for life as we know it. In addition to individual models that rep- resent a planet orbiting a WD at a certain time during the cooling process (i.e., specific WD surface temperatures), we also model two planets through their evolution while the WD cools, representing two possible tracks through a WD's HZ. Both cases show a de- crease of ozone for the three points mod- eled (for WD surface temperature of 6000 K, 5000 K and 4000 K and the corresponding irradiation on the planets) as well as an in- crease in UVC surface flux over time. Due to the extreme change in luminosity early in a WD's evolution a temperate planet around a cool WD would have experienced extremely high luminosity and corresponding surface temperature early in its history, what should initiate a runaway greenhouse process and extreme water loss on such planets early in their history. Whether or not such planets would have lost all their water at that point would depend strongly on the initial water reservoir, and whether it can be replenished, leaving open the question about whether an Earth-like planet could survive around WDs if formed. However due to the extremely fa- vorable size ratio of an Earth-like planet com- pared to a WD, as well as the low luminosity of a WD compared to a main sequence host star, WD exoplanets will make interesting tar- gets for characterization. LK and TK acknowledge support by the Simons Foundation (SCOL # 290357, Kalteneg- ger) and the Carl Sagan Institute. We thank Piotr Kowalski for kindly providing the model spectra of cool WDs used here. REFERENCES Agol, E. 2011, ApJ, 731, L31 Barnes, R., Meadows, V. S., & Evans, N. 2015, ApJ, 814, 91 Benvenuto, O. G., & Althaus, L. G. 1999, MNRAS, 303, 30 Bergeron, P., Leggett, S. K., & Ruiz, M. T. 2001, VizieR Online Data Catalog, 213 Bergeron, P., Ruiz, M. T., & Leggett, S. K. 1997, ApJS, 108, 339 Chapman, S. A. (1930) Theory of Upper- Atmospheric Ozone, Mem. R. Met. Soc., 3, 26, pp 103-125. Cockell, C. S. (1998), Biological Journal of the Linnean Society, 63: 449-457 Demarais, N. J., Yang, Z., Martinez, O., et al. 2012, ApJ, 746, 32 Diffey B.L. Rev Phys Med Biol 1991: 36(3): 299?328. Farihi, J. 2016, New A Rev., 71, 9 Farihi, J., Gansicke, B. T., & Koester, D. 2013, Science, 342, 218 Fontaine, G., Brassard, P., & Bergeron, P. 2001, PASP, 113, 409 Fossati, L., Bagnulo, S., Haswell, C. A., et al. 2012, ApJ, 757, L15 Fossati, L., Bagnulo, S., Haswell, C. A., et al. 2015, Polarimetry, 305, 325 Fulton, B. J., Tonry, J. L., Flewelling, H., et al. 2014, ApJ, 796, 114 Giammichele, N., Bergeron, P., & Dufour, P. 2012, ApJS, 199, 29 Haghighipour, N., & Kaltenegger, L. 2013, ApJ, 777, 166 Hamers, A. S., & Portegies Zwart, S. F. 2016, MNRAS, 462, L84 16 Haqq-Misra, J. D., Domagal-Goldman, S. D., Kasting, P. J., & Kasting, J. F. 2008, As- trobiology, 8, 1127 Houghton, J.T., Meira Filho, L.G., Bruce, J., Lee, H., Callander, B.A., Haites, E., Harris, N., and Maskell, K. (eds.) (1994) Climate Change, 1994: Radiative Forcing of Climate Change and an Evaluation of the IPCC IS92 Emission Scenarios, Cam- bridge University Press, Cambridge, UK. Howell, S. B., Sobeck, C., Haas, M., et al. 2014, PASP, 126, 398 Jura, M., & Xu, S. 2010, AJ, 140, 1129 Jura, M., & Young, E. D. 2014, Annual Re- view of Earth and Planetary Sciences, 42, 45 Kaltenegger, L. 2017, ARA&A, 55, 433 Kaltenegger, L., & Haghighipour, N. 2013, ApJ, 777, 165 Kaltenegger, L. 2010, ApJ, 712, L125 Kaltenegger, L., & Sasselov, D. 2010, ApJ, 708, 1162 Kasting, J. F., Whitmire, D. P., & Reynolds, R. T. 1993, Icarus, 101, 108 Kasting, J. F., & Ackerman, T. P. 1986, Sci- ence, 234, 1383 Kasting, J. F., Holland, H. D., Pinto, J. P. 1985, J. Geophys. Res. 90, 10497?10510 Kasting, J. F., & Ackerman, T. P. 1986, Sci- ence, 234, 1383 Kepler, S. O., Pelisoli, I., Koester, D., et al. 2016, MNRAS, 455, 3413 Kerwin BA, Remmele RL, Jr. 2007. J Pharm Sci. 96(6):1468?1479. 153 Kilic, M., Kowalski, P. M., Reach, W. T., & Parsons, S. G., Gansicke, B. T., Marsh, T. R., von Hippel, T. 2009, ApJ, 696, 2094 et al. 2017, MNRAS, 470, 4473 Kilic, M., Kowalski, P. M., & von Hippel, T. Pavlov, A. A., & Kasting, J. F. 2002, Astrobi- 2009, AJ, 138, 102 ology, 2, 27 Klein, B., Jura, M., Koester, D., & Zucker- man, B. 2011, ApJ, 741, 64 Koester, D., & Wilken, D. 2006, A&A, 453, 1051 Koester, D., Gansicke, B. T., & Farihi, J. 2014, A&A, 566, A34 Kopparapu, R. K., Ramirez, R. M., Schot- telKotte, J., et al. 2014, ApJ, 787, L29 Kopparapu, R. K., Ramirez, R., Kasting, J. F., et al. 2013, ApJ, 770, 82 Kowalski, P. M., & Saumon, D. 2006, ApJ, 651, L137 Kunitomo, M., Ikoma, M., Sato, B., Katsuta, Y., & Ida, S. 2011, ApJ, 737, 66 Loeb, A., & Maoz, D. 2013, MNRAS, 432, 11 Malamud, U., & Perets, H. B. 2017, ApJ, 849, 8 Malamud, U., & Perets, H. B. 2017, ApJ, 842, 67 Malamud, U., & Perets, H. B. 2016, ApJ, 832, 160 Matsunaga, T., Heida, K., and Nikaido, O. (1991), Photochem. Photobiol., 54, 403- 410. 19. McCree K. J., 1971, Agricultural Meteorol- ogy, 9, 191 Mustill, A. J., & Villaver, E. 2012, ApJ, 761, 121 Pavlov, A. A., Kasting, J. F., Brown, L. L., Rages, K. A., & Freedman, R. 2000, J. Geophys. Res., 105, 11981 Ramirez, R. M., & Kaltenegger, L. 2017, ApJ, 837, L4 Ramirez, R. M., & Kaltenegger, L. 2016, ApJ, 823, 6 Ramirez, R. M., & Kaltenegger, L. 2014, ApJ, 797, L25 Rugheimer, S., & Kaltenegger, L. 2018, ApJ, 854, 19 Rugheimer, S., Kaltenegger, L., Segura, A., Linsky, J., & Mohanty, S. 2015, ApJ, 809, 57 Rugheimer, S., Segura, A., Kaltenegger, L., & Sasselov, D. 2015, ApJ, 806, 137 Rugheimer, S., Kaltenegger, L., Zsom, A., Segura, A., & Sasselov, D. 2013, Astro- biology, 13, 251 Saumon, D., Holberg, J. B., & Kowalski, P. M. 2014, ApJ, 790, 50 Segura, A., Meadows, V. S., Kasting, J. F., Crisp, D., & Cohen, M. 2007, A&A, 472, 665 Segura, A., Kasting, J. F., Meadows, V., et al. 2005, Astrobiology, 5, 706 Segura, A., Krelove, K., Kasting, J. F., et al. 2003, Astrobiology, 3, 689 Tevini M (1993), Lewis, Boca Raton, pp 125?154. 17 Toon, O. B., McKay, C. P., Ackerman, T. P., & Santhanam, K. 1989, J. Geophys. Res., 94, 16287 Van Eylen, V., Agentoft, C., Lundkvist, M. S., et al. 2017, arXiv:1710.05398 van Sluijs, L., & Van Eylen, V. 2017, arXiv:1711.09691 Vanderburg, A., Johnson, J. A., Rappaport, S., et al. 2015, Nature, 526, 546 Veras, D., & Gansicke, B. T. 2015, MNRAS, 447, 1049 Veras, D. 2016, Royal Society Open Science, 3, 150571 Villaver, E., Livio, M., Mustill, A. J., & Siess, L. 2014, ApJ, 794, 3 Villaver, E. 2012, IAU Symposium, 283, 219 Villaver, E. 2011, American Institute of Physics Conference Series, 1331, 21 Voet D, Gratzer W.B., Cox R. A., and Doty P.., Biopolymers, Vol. 1, 1963 , pp. 193- 208. Wolszczan, A., & Frail, D. A. 1992, Nature, 355, 145 Xu, S., Ertel, S., Wahhaj, Z., et al. 2015, A&A, 579, L8 This 2-column preprint was prepared with the AAS LATEX macros v5.2. 18 d n u o r g o t % 8 1 - E 4 5 . 9 - E 0 1 . 8 1 - E 0 1 . 1 2 - E 8 1 . 3 2 - E 6 9 . 8 - E 9 2 . 2 1 - E 1 1 . 4 1 - E 2 6 . 5 1 - E 3 4 . 3 - E 4 5 - E 2 1 . 7 - E 9 9 . 7 - E 4 1 . d n u o r G 9 1 - E 9 3 . 0 1 - E 9 1 . 9 1 - E 9 1 . 2 2 - E 3 3 . 3 2 - E 8 1 . 0 1 - E 6 6 . 4 1 - E 5 2 . 5 1 - E 4 1 . 7 1 - E 9 9 . 6 - E 1 7 . 8 - E 1 2 . 9 - E 6 1 . 0 1 - E 4 2 . A O T I d n u o r g o t % d n u o r G A O T I d n u o r g o t % d n u o r G 1 7 . . 5 8 1 . 5 8 1 . 5 8 1 . 5 8 1 3 2 . 3 2 . 3 2 . 3 2 . 7 1 0 . 7 1 0 . 7 1 0 . 7 1 0 . 1 1 3 2 1 1 6 7 . 1 6 . 1 3 6 1 2 1 5 9 . 8 4 6 2 9 1 5 1 2 . 2 4 . 7 5 . 3 5 . 2 0 . 2 0 . 2 0 . 1 0 8 . 0 3 6 . 0 7 4 . 0 5 2 . 0 9 1 . 0 5 1 . 0 9 . 8 1 4 . 2 3 4 . 2 3 4 . 2 3 4 . 2 3 6 . 6 6 . 6 6 . 6 6 . 6 8 9 . 0 8 9 . 0 8 9 . 0 8 9 . 0 0 7 3 9 0 7 0 6 3 5 5 9 3 7 2 6 5 5 8 9 6 7 5 6 8 5 6 . 0 5 1 . 5 0 1 9 . 8 7 8 . 7 6 9 . 9 5 6 . 7 3 6 . 8 2 6 . 4 2 7 . 1 2 2 . 0 1 9 . 7 8 . 6 0 . 6 A O T I 4 . 2 7 2 . 3 1 1 2 . 3 1 1 2 . 3 1 1 2 . 3 1 1 4 . 9 3 4 . 9 3 4 . 9 3 4 . 9 3 4 . 0 1 4 . 0 1 4 . 0 1 4 . 0 1 ) r a b ( 0 . 1 3 . 0 0 . 1 5 . 1 0 . 2 3 . 0 0 . 1 5 . 1 0 . 2 3 . 0 0 . 1 5 . 1 0 . 2 ) 2 / m W ( m n 0 8 2 . - 6 1 2 1 C V U ) 2 / m W ( m n 5 1 3 - 0 8 2 B V U ) 2 / m W ( m n 0 0 4 - 5 1 3 A V U e r u s s e r P h t r a E y a d t n e s e r P 0 0 0 6 0 0 0 6 0 0 0 6 0 0 0 6 0 0 0 5 0 0 0 5 0 0 0 5 0 0 0 5 0 0 0 4 0 0 0 4 0 0 0 4 0 0 0 4 19 s e x u l F d e t a r g e t n I t s o H V U : 2 e l b a T f f e T ) K ( Table 3: UV Integrated fluxes compared to Earth Host Teff Pressure WD UV/ Present day Earth UV UVA (315-400 nm) UVB (280-315 nm) UVC (121.6-280 nm) (K) 6000 6000 6000 6000 5000 5000 5000 5000 4000 4000 4000 4000 (bar) 0.3 1.0 1.5 2.0 0.3 1.0 1.5 2.0 0.3 1.0 1.5 2.0 2.1 1.6 1.3 1.2 0.74 0.57 0.49 0.43 0.20 0.16 0.13 0.12 3.3 1.6 1.1 0.90 0.90 0.45 0.36 0.28 0.21 0.11 0.085 0.067 4.9×108 4.9×10−1 8.6×10−4 4.7×10−5 1.7×109 6.5×104 3.6×103 2.6×102 1.8×1013 5.3×1010 4.2×109 6.2×108 Table 4: Case A & Case B Results Host Sef f (K) 6000 5000 4000 1.3 0.64 0.26 Case A: r = 0.0085 AU Surface Ozone Column Sef f Teff (K) Depth (cm−2) 328.4 249.6 191.9 1.7×1018 4.4×1018 9.6×1017 2.1 1.0 0.41 Case B: r = 0.0069 AU Surface Ozone Column Teff (K) Depth (cm−2) RG∗ 290.8 216.9 3.3×1018 1.3×1018 RG∗ ∗RG indicates a runaway greenhouse state 20 d n u o r g o t % 8 1 − 5 − 4 1 − 3 − 2 1 − 2 − 0 1 × 4 5 0 1 × 0 1 0 1 × 4 2 0 1 × 0 1 0 1 × 1 1 0 1 × 1 1 . . . . . . 6 − 9 1 − d n u o r G 0 1 × 9 3 0 1 × 1 5 2 . . 6 1 − 7 − 4 1 − 6 − 0 1 × 6 3 0 1 × 0 7 0 1 × 5 2 0 1 × 0 5 . . . . A O T I 1 7 . . 7 4 2 2 − 5 1 . 0 1 × 8 6 . 2 − 3 2 . 0 1 × 4 4 . d n u o r g o t % d n u o r G A O T I d n u o r g o t % d n u o r G 1 1 4 2 5 1 2 3 6 1 6 3 2 . 2 2 . 0 1 4 6 . 0 3 1 . 0 2 − 0 . 1 0 1 × 2 . 9 9 . 8 1 4 . 3 4 3 . 4 0 4 . 0 6 . 6 6 2 . 0 0 7 1 7 4 7 8 7 3 7 7 7 6 . 0 5 0 . 7 0 1 9 . 8 1 3 . 3 6 . 8 2 1 . 2 A O T I 4 . 2 7 3 . 1 5 1 4 . 5 2 3 . 4 4 . 9 3 8 . 2 ) 2 / m W ( m n 0 8 2 - . 6 1 2 1 C V U ) 2 / m W ( m n 5 1 3 - 0 8 2 B V U ) 2 / m W ( m n 0 0 4 - 5 1 3 A V U ) K ( f f e T t s o H e s a C 0 5 7 5 0 0 0 6 0 0 0 5 0 0 0 4 0 0 0 5 0 0 0 4 h t r a E y a d t n e s e r P A A A B B 21 n o i t u l o v e s e x u fl d e t a r g e t n I V U : 5 e l b a T Table 6: UV Integrated fluxes evolution Earth comparison UVB 4.6 UVA 2.1 3.7×10−1 6.6×10−2 5.7×10−1 4.2×10−2 Case Host Teff WD UV/Present day Earth UV UVC 6.5×1012 9.3×102 1.8×1012 6.48×104 1.3×1013 2.9×10−1 5.8×10−2 4.5×10−1 4.1×10−2 (K) 6000 5000 4000 5000 4000 A A A B B Fig. 7. -- UV surface flux for Case B: a planet orbiting at 0.0069 AU, corresponding to Sef f = 1.0 for the 5000 K WD surface temperature, and Sef f = 0.41 for the 4000 K WD surface temperature model. At 6000 K the Sef f equals 2.07, placing the planet outside of both the conservative as well as empirical HZ. 22 150200250300350400Wavelength (nm)10-710-610-510-410-310-210-1100101erg/cm2/s/UVCUVBUVA5000 K WDSun200250300350400Wavelength (nm)10-3510-3210-2910-2610-2310-2010-1710-1410-1110-810-510-2101Incident Energy (W/m2/nm)UVCUVBUVATOA 5000 K WDTOA Sun5000 K WDSunII150200250300350400Wavelength (nm)10-710-610-510-410-310-210-1100101erg/cm2/s/UVCUVBUVA4000 K WDSun200250300350400Wavelength (nm)10-3510-3210-2910-2610-2310-2010-1710-1410-1110-810-510-2101Incident Energy (W/m2/nm)UVCUVBUVATOA 4000 K WDTOA Sun4000 K WDSunIIFig. 8. -- Temperature and photochemistry profiles for Case A and B (solid lines), with the Earth- Sun profile for comparison (dashed). 23 150200250300510152025303540T(K)10-910-810-710-610-5O310-1510-1210-9CH410-1510-1010-5H2O10-1210-1010-810-6N2O10-4010-3010-2010-10Case ACH3Cl6000 K5000 K4000 KSun150200250300T (K)510152025303540T(K)10-910-810-710-610-5O310-710-6 Mixing RatiosCH410-1510-1010-5H2O10-1010-810-6N2O10-1310-1010-7Case BCH3Cl
1804.04233	1	1804	2018-04-11T21:37:52	Mars' Growth Stunted by an Early Giant Planet Instability	[ "astro-ph.EP" ]	Many dynamical aspects of the solar system can be explained by the outer planets experiencing a period of orbital instability sometimes called the Nice Model. Though often correlated with a perceived delayed spike in the lunar cratering record known as the Late Heavy Bombardment (LHB), recent work suggests that this event may have occurred much earlier; perhaps during the epoch of terrestrial planet formation. While current simulations of terrestrial accretion can reproduce many observed qualities of the solar system, replicating the small mass of Mars requires modification to standard planet formation models. Here we use 800 dynamical simulations to show that an early instability in the outer solar system strongly influences terrestrial planet formation and regularly yields properly sized Mars analogs. Our most successful outcomes occur when the terrestrial planets evolve an additional 1-10 million years (Myr) following the dispersal of the gas disk, before the onset of the giant planet instability. In these simulations, accretion has begun in the Mars region before the instability, but the dynamical perturbation induced by the giant planets' scattering removes large embryos from Mars' vicinity. Large embryos are either ejected or scattered inward toward Earth and Venus (in some cases to deliver water), and Mars is left behind as a stranded embryo. An early giant planet instability can thus replicate both the inner and outer solar system in a single model.	astro-ph.EP	astro-ph	Icarus; accepted Preprint typeset using LATEX style emulateapj v. 12/16/11 8 1 0 2 r p A 1 1 . ] P E h p - o r t s a [ 1 v 3 3 2 4 0 . 4 0 8 1 : v i X r a MARS' GROWTH STUNTED BY AN EARLY GIANT PLANET INSTABILITY Matthew S. Clement1,, Nathan A. Kaib1, Sean N. Raymond2, & Kevin J. Walsh3 Icarus; accepted ABSTRACT Many dynamical aspects of the solar system can be explained by the outer planets experiencing a period of orbital instability sometimes called the Nice Model. Though often correlated with a perceived delayed spike in the lunar cratering record known as the Late Heavy Bombardment (LHB), recent work suggests that this event may have occurred much earlier; perhaps during the epoch of terrestrial planet formation. While current simulations of terrestrial accretion can reproduce many observed qualities of the solar system, replicating the small mass of Mars requires modification to standard planet formation models. Here we use 800 dynamical simulations to show that an early instability in the outer solar system strongly influences terrestrial planet formation and regularly yields properly sized Mars analogs. Our most successful outcomes occur when the terrestrial planets evolve an additional 1-10 million years (Myr) following the dispersal of the gas disk, before the onset of the giant planet instability. In these simulations, accretion has begun in the Mars region before the instability, but the dynamical perturbation induced by the giant planets' scattering removes large embryos from Mars' vicinity. Large embryos are either ejected or scattered inward toward Earth and Venus (in some cases to deliver water), and Mars is left behind as a stranded embryo. An early giant planet instability can thus replicate both the inner and outer solar system in a single model. Keywords: Mars, Planetary Formation, Terrestrial Planets, Early Instability 1. INTRODUCTION It is widely understood that the evolution of the so- lar system's giant planets play the most important role in shaping the dynamical system of bodies we observe today. When the outer planets interact with an exte- rior disk of bodies, Saturn, Uranus and Neptune tend to scatter objects inward (Fernandez & Ip 1984). To conserve angular momentum through this process, the orbits of these planets move outward over time (Hahn & Malhotra 1999; Gomes 2003). Thus, as the young solar system evolved, the three most distant planets' orbits moved out while Jupiter (which is more likely to eject small bodies from the system) moved in. To explain the excitation of Pluto's resonant orbit with Neptune, Mal- hotra (1993) proposed that Uranus and Neptune must have undergone significant orbital migration prior to ar- riving at their present semi-major axes. Malhotra (1995) later expanded upon this idea to explain the full resonant structure of the Kuiper belt. In the same manner, an or- bital instability in the outer solar system can successfully excite Kuiper belt eccentricities and inclinations, while simultaneously moving the giant planets to their present semi-major axes via planet-planet scattering followed by dynamical friction (Thommes et al. 1999). These ideas culminated in the eventual hypothesis that, as the giant planets orbits diverged after their for- mation, Jupiter and Saturn's orbits would have crossed a mutual 2:1 Mean Motion Resonance (MMR). Known as 1 HL Dodge Department of Physics Astronomy, University of Oklahoma, Norman, OK 73019, USA 2 Laboratoire dAstrophysique de Bordeaux, Univ. Bordeaux, CNRS, B18N, alle Geoffroy Saint-Hilaire, 33615 Pessac, France 3 Southwest Research Institute, 1050 Walnut St. Suite 300, Boulder, CO 80302, USA corresponding author email: [email protected] the Nice (as in Nice, France) Model (Tsiganis et al. 2005; Gomes et al. 2005; Morbidelli et al. 2005), this resonant configuration of the two most massive planets causes a solar system-wide instability, which has been shown to reproduce many peculiar dynamical traits of the solar system. This hypothesis has subsequently explained the overall structure of the Kuiper belt (Levison et al. 2008; Nesvorn´y 2015a,b), the capture of trojan satellites by Jupiter (Morbidelli et al. 2005; Nesvorn´y et al. 2013), the orbital architecture of the asteroid belt (Roig & Nesvorn´y 2015), and the giant planets' irregular satellites, includ- ing Triton (Nesvorn´y et al. 2007). The Nice Model itself has changed significantly since its introduction. In order to keep the orbits of the ter- restrial planets dynamically cold (low eccentricities and inclinations), Brasser et al. (2009) proposed that Jupiter "jump" over its 2:1 MMR with Saturn, rather than mi- grate smoothly through it (Morbidelli et al. 2009a, 2010). Otherwise, the terrestrial planets were routinely excited to the point where they were ejected or collided with one another in simulations. The probability of produc- ing successful jumps in these simulations is greatly in- creased when an extra primordial ice giant was added to the model (Nesvorn´y 2011; Batygin et al. 2012). In successful simulations, the ejection of an additional ice giant rapidly forces Jupiter and Saturn across the 2:1 MMR. Furthermore, hydrodynamical simulations (Snell- grove et al. 2001; Papaloizou & Nelson 2003) show that a resonant chain of giant planets is likely to emerge from the dissipating gaseous circumstellar disk, with Jupiter and Saturn locked in an initial 3:2 MMR resonance (Mas- set & Snellgrove 2001; Morbidelli & Crida 2007; Pierens & Nelson 2008). This configuration can produce the same results as the 2:1 MMR crossing model (Morbidelli et al. 2007). In this scenario, the instability ensues when two giant planets fall out of their mutual resonant config- 2 uration. Such an evolutionary scheme seems consistent with the number of resonant giant exoplanets discovered (eg: HD 60532b, GJ 876b, HD 45364b, HD 27894, Ke- pler 223 and HR 8799 (Holman et al. 2010; Fabrycky & Murray-Clay 2010; Rivera et al. 2010; Delisle et al. 2015; Mills et al. 2016; Trifonov et al. 2017)). Despite the fact that 5 and 6 primordial giant planet configurations are quite successful at reproducing the ar- chitecture of the outer solar system (Nesvorn´y & Mor- bidelli 2012), delaying the instability ∼400 Myr to co- incide with the lunar cataclysm (Gomes et al. 2005) still proves problematic for the terrestrial planets. In- deed, Kaib & Chambers (2016) find only a ∼ 1% chance that the terrestrial planets' orbits and the giant plan- ets' orbits are reproduced simultaneously. Even in sys- tems with an ideal "jump," the eccentricity excitation of Jupiter and Saturn can bleed to the terrestrial plan- ets via stochastic diffusion, leading to the over-excitation or ejection of one or more inner planets (Agnor & Lin 2012; Brasser et al. 2013; Roig & Nesvorn´y 2015). It should be noted, however, that Mercury's uniquely ex- cited orbit (largest mean eccentricity and inclination of the planets) may be explained by a giant planet instabil- ity (Roig et al. 2016). Nevertheless, the chances of the entire solar system emerging from a late instability in a configuration roughly resembling its modern architecture are very low (Kaib & Chambers 2016). This suggests that the instability is more likely to have not occurred in conjunction with the LHB, but rather before the ter- restrial planets had fully formed. Fortunately, many of the dynamical constraints on the problem are fairly im- partial to whether the instability happened early or late (Morbidelli et al. 2018). The Kuiper belt's orbital struc- ture (Levison et al. 2008; Nesvorn´y 2015a,b), Jupiter's Trojans (Morbidelli et al. 2005; Nesvorn´y et al. 2013), Ganymede and Callisto's different differentiation states (Barr & Canup 2010) and the capture of irregular satel- lites in the outer solar system (Nesvorn´y et al. 2007) are still explained well regardless of the specific timing of the Nice Model instability. In addition to perhaps ensur- ing the survivability of the terrestrial system, there are several other compelling reasons to investigate an early instability: 1. Uncertainties in Disk Properties: Since the in- troduction of the Nice Model, simplifying assumptions of the unknown properties of the primordial Kuiper belt have provided initial conditions for N-body simulations. The actual timing of the instability is highly sensitive to the particular disk structure selected (Gomes et al. 2005). Furthermore, numerical studies must approxi- mate the complex disk structure with a small number of bodies in order to optimize the computational cost of simulations. In fact, most N-body simulations do not account for the affects of disk self gravity (Nesvorn´y & Morbidelli 2012). When the giant planets are embedded in a disk of gravitationally self-interacting particles using a graphics processing unit (GPU) to perform calculations in parallel and accelerate simulations (Grimm & Stadel 2014), instabilities typically occur far earlier than what is required for a late instability (Quarles & Kaib, in prep). 2. Highly Siderophile Elements (HSE): A late instability (the LHB) was originally favored because of the small mass accreted by the Moon relative to the Earth after the Moon-forming impact (for a review of these ideas see Morbidelli et al. (2012)). The HSE record from lunar samples indicates that the Earth accreted al- most 1200 times more material, despite the fact that its geometric cross-section is only about 20 times that of the Moon (Walker et al. 2004; Day et al. 2007; Walker 2009). Thus, the flux of objects impacting the young Earth would have had a very top-heavy size distribu- tion (Bottke et al. (2010), however Minton et al. (2015) showed that the pre-bombardment impactor size distri- bution may not be as steep as originally assumed). This distribution of impactors is greatly dissimilar from what is observed today, and favors the occurrence of a LHB. New results, however, indicate that the HSE disparity is actually a result of iron and sulfur segregation in the Moon's primordial magma ocean causing HSEs to drag towards the core long after the moon-forming impact (Rubie et al. 2016). Because the crystallization of the lunar magma took far longer than on Earth, a large dis- parity between the HSE records is expected (Morbidelli et al. 2018). 3. Updated Impact Data: The LHB hypothesis gained significant momentum when none of the lunar im- pactites returned by the Apollo missions were older than 3.9 Gyr (Tera et al. 1974; Zellner 2017). However, recent 40Ar/39Ar age measurements of melt clasts in Lunar me- teorites are inconsistent with the U/Pb dates determined in the 1970s (Fernandes et al. 2000; Chapman et al. 2007; Boehnke et al. 2016). These new dates cover a broader range of lunar ages; and thus imply a smoother decline of the Moon's cratering rate. Furthermore, new high- resolution images from the Lunar reconnaissance Orbiter (LRO) and the GRAIL spacecraft have significantly in- creased the number of old (>3.9 Gyr) crater basins used in crater counting (Spudis et al. 2011; Fassett et al. 2012). For example, samples returned by Apollo 17 that were originally assumed to be from the impactor that formed the Serenitatis basin are likely contaminated by ejecta from the Irbrium basin (Spudis et al. 2011). Because the Serenitatis basin is highly marred by young craters and ring structures, it is likely older than 4 Gyr, and the Apollo samples are merely remnants of the 3.9 Gyr Irbrium event. Here we build upon the hypothesis of an early insta- bility by systematically investigating the effects of the Nice Model occurring during the process of terrestrial planetary formation. Since advances in algorithms sub- stantially decreased the computational cost of N-body integrators in the 1990s (Wisdom & Holman 1991; Dun- can et al. 1998; Chambers 1999), many papers have been dedicated to modeling the late stages (giant im- pact phase) of terrestrial planetary formation. Observa- tions of proto-stellar disks (Haisch et al. 2001; Pascucci et al. 2009) suggest that free gas disappears far quicker than the timescale radioactive dating indicates it took the terrestrial planets to form (Halliday 2008; Kleine et al. 2009). Because the outer planets must clearly form first, the presence of Jupiter is supremely impor- tant when modeling the formation of the inner planets (Wetherill 1996; Chambers & Cassen 2002; Levison & Agnor 2003a). Early N-body integrations of planet for- mation in the inner solar system in 3 dimensions from a disk of planetary embryos and a uniformly distributed sea of planetesimals reproduced the general orbital spac- ing of our 4 terrestrial planets (Chambers & Wetherill 1998; Chambers 2001). However, these efforts system- atically failed to produce an excited asteroid belt and 4 dynamically cold planets with the correct mass ratios (Mercury and Mars are ∼ 5% and ∼ 10% the mass of Earth respectively). Numerous subsequent authors approached these prob- lems using various methods and initial conditions. By accounting for the dynamical friction of small planetes- imals, O'Brien et al. (2006) and Raymond et al. (2006) more consistently replicated the low eccentricities of the terrestrial planets. However, the so-called "Small Mars Problem" proved to be a systematic short-coming of N- body accretion models (Wetherill 1991; Raymond et al. 2009). The vast majority of simulations produce Mars analogs roughly the same mass as Earth and Venus; a full order of magnitude too large (Morishima et al. 2010). However, Mars' mass consistently stayed low when using a configuration of Jupiter and Saturn with present day mutual inclination and eccentricities twice their modern values (Raymond et al. 2009). Because planet-disk inter- actions systematically damp the eccentricities of grow- ing gas planets (Papaloizou & Larwood 2000; Tanaka & Ward 2004), this result presented the problem of re- quiring a mechanism to adequately excite the orbits of the giant planets prior to terrestrial planetary forma- tion. Hansen (2009) then demonstrated that a small Mars could be formed if the initial disk of planetesimals was confined to a narrow annulus between 0.7-1.0 au (in- terestingly, a narrow annulus might also explain the or- bital distribution of silicate rich S-type asteroids (Ray- mond & Izidoro 2017b)). The "Grand Tack" hypothesis provides an interesting mechanism to create these con- ditions whereby a still-forming Jupiter migrates inward and subsequently "tacks" backward once it falls into res- onance with Saturn (Walsh et al. 2011; Brasser et al. 2016). When Jupiter "tacks" at the correct location, the disk of planetesimals in the still-forming inner solar sys- tem is truncated at 1.0 au, roughly replicating an annulus (for a critical review of the Grand Tack consult Raymond & Morbidelli (2014)). Another potential solution to the small Mars problem is local depletion of the outer disk (Izidoro et al. 2014, 2015). However, systems in these studies that placed Jupiter and Saturn on more realistic initially circular or- bits failed to produce a small Mars. Mars' formation could have also been affected by a secular resonance with Jupiter sweeping across the inner solar system as the gaseous disk depletes (Thommes et al. 2008; Bromley & Kenyon 2017). The degree to which this process in- duces a dynamical shake-up of material in the vicinity of the forming Mars and asteroid belt is strongly tied to speed of the resonance sweeping. Furthermore, it is possible that Mars' peculiar mass is simply the result of a low probability event. Indeed, there is a low, but non-negligible probability of forming a small Mars when using standard initial conditions and assuming no prior depletion of the disk (Fischer & Ciesla 2014). However, on closer inspection, many of the "successful" systems in Fischer & Ciesla (2014) are poor solar system analogs for other reasons (such as an extra large planet in the asteroid belt (Jacobson & Walsh 2015)). Additionally, the large masses of Mars analogs produced in N-body integrations could be a consequence of the simplifica- 3 tions and assumptions made by such simulations. The process of how planetesimals form out of small, peb- ble to meter sized bodies is still an active field of re- search. Reevaluating the initial conditions used by N- body accretion models of the giant impact phase may potentially shed light on the origin of Mars' small mass. Kenyon & Bromley (2006) considered this problem us- ing a multi-annulus coagulation code to grow kilometer scale planetesimals. Subsequent authors (Levison et al. 2015; Dr¸azkowska et al. 2016; Raymond & Izidoro 2017b) modeling the accretion of meter sized objects found that dust growth and drift cause solids in the inner disk to be redistributed in to a much steeper radial profile. Fi- nally, multiple studies have demonstrated that more re- alistic, erosive collisions can significantly reduce the mass of embryos during the late stages of terrestrial planet ac- cretion (Kokubo & Genda 2010; Kobayashi & Dauphas 2013; Chambers 2013). Here we investigate an alternative scenario wherein the still forming inner planets are subjected to a Nice Model instability. Though the effect of the Nice Model on the fully formed terrestrial planets is well studied (Brasser et al. 2009; Agnor & Lin 2012; Brasser et al. 2013; Kaib & Chambers 2016; Roig et al. 2016), no investigation to date has performed direct numerical simulations of the effect of the Nice Model instability on the still forming terrestrial planets. Furthermore, our work is motivated by simulations from Lykawka & Ito (2013). The authors found that a low massed Mars could be formed when Jupiter and Saturn (with enhanced eccentricities) were artificially migrated across their mutual 2:1 MMR using fictitious forces. A downfall of this scenario, however, is that it overexcites the orbits of the forming terrestrial planets. Additionally, previous authors (Walsh & Mor- bidelli 2011) have investigated whether smooth migra- tion of Jupiter and Saturn (as opposed to the "jumping Jupiter" model) could produce a small Mars, however the speed of the process was found to be too slow with re- spect to Mars' formation timescale (1-10 Myr; Dauphas & Pourmand (2011)). The work of Walsh & Morbidelli (2011) is perhaps the most similar to that of this paper. However, our work differs greatly in that we consider full instabilities directly, rather than modeling terrestrial evolution while migrating the giant planets with artificial forces. It should also be noted that we do not include the "Grand Tack" hypothesis in our study (in section 5 we argue that it is potentially compatible with the scenario we investigate). 2. MATERIALS AND METHODS Because the parameter space of possible giant planet configurations (number of planets, resonant configura- tion, planet spacing, disk mass and disk spacing) emerg- ing from the primordial gas disk is substantial, as a start- ing point for this work we take two of the most successful five and six giant planet configurations from Nesvorn´y & Morbidelli (2012)1. Both begin with Jupiter and Sat- 1 It should be noted that, with the additional constraint of re- quiring that Neptune migrate to ∼28 au prior to the onset of the instability, a 3:2,3:2,2:1,3:2 resonant configuration for 5 planet sce- narios is more effective (Deienno et al. 2017). Because we are mostly interested in studying the excitation of the inner solar sys- tem, which is largely unaffected by the particular migration of Nep- tune, our results should be relatively independent of the particular 4 Giant Planet Initial Conditions: The columns are: (1) the name of the simulation set, (2) the number of giant planets, (3) the mass of the planetesimal disk exterior to the giant planets, (4) the distance between the outermost ice giant and the planetesimal disks inner edge, (5) the semi-major axis of the outermost ice giant (commonly referred to as Neptune, however not necessarily the planet which completes the simulation at Neptune's present orbit), (6) the resonant configuration of the giant planets starting with the Jupiter/Saturn resonance, and (7) the masses of the ice giants from inside to outside. Table 1 Name NP ln Mdisk (M⊕) n1 n2 5 6 35 20 δr (au) 1.5 1.0 rout (au) 30 30 urn in their 3:2 MMR. Though placing Jupiter and Sat- urn in an initial 2:1 configuration on circular orbits can be highly successful at replicating the correct planetary spacing of the outer solar system, only ∼0.2% such simu- lations sufficiently excite Jupiter's eccentricity (Nesvorn´y & Morbidelli 2012). For this reason we focus on the sce- nario where Jupiter and Saturn emerge from the gas disk locked in a 3:2 MMR. Furthermore, Nesvorn´y & Mor- bidelli (2012) found advantages and disadvantages which are mutually exclusive to both the five and six planet cases. Thus, for completeness we select one five and one six planet setup for our study. We summarize our chosen sets of giant planet initial conditions in table 1. To create a mutual resonant chain of gas giants, we ap- ply an external force to the giant planets which mimics the effects of a gas disk by modifying the equations of motion with forced migration ( a) and eccentricity damp- ing ( e) terms (Lee & Peale 2002). Though the precise underlying physics of the interaction between forming giant planets and a gas disk is not fully understood, this method is employed by many authors studying both the solar system, and observed resonant exoplanets because it consistently produces stable resonant chains (Mat- sumura et al. 2010; Beaug´e & Nesvorn´y 2012). The exact functional forms of a and e depend on the timescale and overall distance of migration desired. We utilize a form of a = ka and e = ke/100 (Batygin & Brown 2010a), where the constant k is adjusted to achieve a migration timescale τmig ∼.1-1 Myr. Because the actual migra- tion rate is a complex function of the properties of the gas disk and relative masses of the planets, achieving a specific migration timescale is of less importance than placing the resonant planets on the proper orbits (Kley & Nelson 2012; Baruteau et al. 2014) We first evolve the gas giants (without terrestrial plan- ets or a exterior disk of planetesimals) using the Mer- cury6 Bulirsch-Stoer integrator (as opposed to the faster hybrid integrator) with a 6.0 day timestep (Chambers 1999; Stoer et al. 2002). The Bulirish-Stoer method is necessary when building resonant configurations because the force on a particle is a function of both the posi- tions and momenta (Chambers 1999; Batygin & Brown 2010b). The planets are initially placed on orbits just outside their respective resonances, and then integrated until the outermost planet's semi-major axis is at the ap- propriate location (table 1). Figure 1 shows an example of this evolution for a simulation in the n1 batch. To verify that the planets are in a MMR, libration about a series of resonant angles is checked for using the method described in Clement & Kaib (2017). After the resonant chain is assembled, we add a disk ice giant resonant configuration selected. anep Resonance Chain Mice (M⊕) (au) 16,16,16 17.4 20.6 8,8,16,16 3:2,4:3,3:2,3:2,3:2 3:2,3:2,3:2,3:2 of 1000 equal massed planetesimals using the inner and outer radii that Nesvorn´y & Morbidelli (2012) showed best meet dynamical constraints for the outer solar sys- tem (table 1). The orbital distribution of the planetesi- mals is chosen in a manner consistent with previous au- thors (Batygin & Brown 2010b; Nesvorn´y & Morbidelli 2012; Kaib & Chambers 2016), and follows an r−1 sur- face density profile. Angular orbital elements (arguments of pericenter, longitudes of ascending node and mean anomalies) for the planetesimals are selected randomly. Eccentricities and inclinations are drawn from near circu- lar and co-planar gaussian distributions (standard devia- tions: σe = .002 and σi = .2◦). These same distributions are utilized throughout our study to maintain all initial eccentricities less than 0.01, and inclinations within 1◦. These systems are integrated using the Mercury6 hybrid integrator (Chambers 1999) with a 20.0 day timestep up until the point when two giant planets first pass within 3 mutual Hill Radii. Because we wish to investigate spe- cific instability times with respect to terrestrial planetary formation, and our resonant configurations can often last for tens of millions of years prior to experiencing an insta- bility, we integrate these systems up until the onset of the instability with an empty inner solar system. Only after this point are the terrestrial planetary formation disks at various stages of evolution embedded in these systems. The method allows us to save computational time, and control at exactly what point the instability occurs dur- ing the giant impact phase. Though the giant planets are already on significantly eccentric orbits at this point (and therefore affecting the terrestrial disk), we find that sys- tems can last for millions of years before experiencing an instability if the simulation is stopped sooner. Moreover, the terrestrial planets are far more sensitive to the evo- lution of Jupiter and Saturn than that of the ice giants. We find that, in the vast majority of simulations, Jupiter and Saturn don't begin evolving substantially until af- ter this first close encounter time. In the vast majority of our simulations, the instability ensues within several thousand years of the first close-encounter time. This is consistent with simulations of planet-planet scattering designed to reproduce giant exoplanet systems (Chatter- jee et al. 2008; Juri´c & Tremaine 2008; Raymond et al. 2010). Because we want to embed terrestrial planetary disks at different stages of development into a giant planet in- stability, we begin by modeling terrestrial disk evolution in the presence of a static Jupiter and Saturn in a 3:2 MMR. We form 100 systems of terrestrial planets using the Mercury6 hybrid integrator (Chambers 1999) and a 6.0 day timestep. Because of the integrator's inability to accurately handle low pericenter passages, objects are considered to be merged with the sun at 0.1 au (Cham- 5 Run Number Disk Mass (M⊕) Disk Inner edge (au) 0-24 25-49 50-74 75-99 5.0 5.0 3.0 3.0 Table 2 0.5 0.7 0.5 0.7 Summary of initial conditions for terrestrial planetary formation simulations. stochasticity of the actual process dominates over the effects of small changes to initial conditions such as em- bryo masses and spacing (Hoffmann et al. 2017). For a summary of these initial conditions, see table 2. In all simulations the outer disk edge is set at 4.0 au. Fur- thermore, we place Jupiter and Saturn in a 3:2 MMR at roughly the same orbital locations as in the n1 and n2 configurations (ice giants are not included for this por- tion of the simulation). As in previous studies, we evolve each system for 200 Myr (Raymond et al. 2009; Kaib & Cowan 2015), outputting a snapshot of each system at 104, 105, 106 and 107 years (these times roughly corre- spond to the time elapsed following gas disk dispersal, which we loosely correlate with the onset of the giant impact phase). These snapshots are then input into the giant planet configurations n1 and n2 described above (therefore each output is used twice, table 1), and inte- grated through the giant planet instability for an addi- tional 200 Myr using the same integrator package and timestep. This gives us 800 different instability simulations. We refer to the two different giant planet configurations (n1 and n2; each containing 400 individual systems) as the simulation "set." We then denote each subset of 100 in- tegrations with a unique instability delay time (104, 105, 106 and 107 years) as "batches." And finally each unique simulation within a batch (100 each) is referred to as a "run." The completed terrestrial formation simulations (with a static Jupiter and Saturn in their pre-instability 3:2 MMR) become our control batch (100 runs). Thus the control batch represents a sample of terrestrial for- mation outcomes in a late Nice Model scenario, where the giant planets remain locked in their mutual resonant configuration until the instability occurs ∼ 400 Myr after the planets finish forming. 2.1. Instability Timing We relate time zero in our simulations (the beginning of our control runs) with the epoch of gas dispersal, which loosely correlates with the beginning of the giant impact phase. However, the exact dynamical state of the ter- restrial disk (for our purposes, the relative abundance of larger planet embryos within the planetesimal sea) at the time of gas disk dispersal is not exactly known. For this reason, the specific instability times we test, and draw conclusions about, are of less importance than the par- ticular dynamical state of the disk. In the subsequent sections of this text, we compare broad characteristics of our early (0.01 and 0.1 Myr) and late (1 and 10 Myr) instability delay times. Though we can confidently make a general connection between these specific instability times and the time elapsed following gas dispersal in the solar system, relating the dynamical state of the terres- trial disk with the timing of the instability is the more Figure 1. An example of the resonant evolution of a system of outer planets in the n1 batch. With the forcing function mimicking interactions with the gaseous disk in place, once a set of planets falls in to a MMR, they remain locked in resonance for the remainder of the evolution. bers 1999, 2001). We choose the simplest initial orbital distributions for objects in the terrestrial forming disk in order to mirror previous studies which assumed no prior disk depletion (Chambers 2001, 2007). Half of the disk mass is in 100 equal massed embryos, with the remainder in 1000 equal massed planetesimals. The spacing of the embryos and planetesimals is selected to achieve a surface density profile that falls off radially as r−3/2. Angular or- bital elements are selected randomly, and eccentricities and inclinations are drawn from near circular, co-planar gaussian distributions (σe = .002 and σi = .2◦). The initial disk mass in simulations numbered 0-49 is set to 5 M⊕. Runs numbered 50-99 begin with 3 M⊕ of mate- rial. Additionally, half of the simulations begin with the inner disk edge at 0.5 au (numbers 0-24 and 50-74) as op- posed to 0.7 au. This gives us 25 simulations with each mass/edge permutation. Additionally, the embryo spac- ing varies between ∼5-12 mutual hill radii (depending on the simulation initial conditions and disk locations), and is consistent with simulations of oligarchic growth of embryos (Kokubo & Ida 1998). Furthermore, terrestrial planet formation is a highly chaotic process in which the 0.00.10.20.30.40.51.541.561.581.601.62Period RatioSaturn:Jupiter0.00.10.20.30.40.51.541.561.581.601.621.64Ice1:Saturn0.00.10.20.30.40.5Time (Myr)1.481.501.521.541.561.581.601.621.64Period RatioIce2:Ice10.00.10.20.30.40.5Time (Myr)1.481.501.521.541.561.581.601.621.641.66Ice3:Ice20.00.10.20.30.40.50123456Resonant Angle (radians)Saturn:Jupiter0.00.10.20.30.40.50123456Ice1:Saturn0.00.10.20.30.40.5Time (Myr)0123456Resonant Angle (radians)Ice2:Ice10.00.10.20.30.40.5Time (Myr)0123456Ice3:Ice26 important conclusion of our work. As we expand upon in the subsequent sections, the later two instability delays we test tend to be more successful than the earlier ones. Therefore our work correlates the giant planet instability timing with a terrestrial disk at a state of evolution that is mostly depleted of small planetesimals, with most of the mass concentrated in a handful of growing planet em- bryos. Subsequently overlaying this timeline on that of Mars' growth inferred from isotopic dating (Dauphas & Pourmand 2011) is difficult because the relationship be- tween gas disk dispersal and CAI (Calcium-Aluminum- rich inclusion) formation is not well known. Furthermore, Marty et al. (2017) presented evidence that cometary bombardment accounted for ∼ 22% of the noble gas concentration in Earth's atmosphere. At first glance, this constraint appears to be slightly at odds with the various delay times we examine in this paper (0.01-10 Myr). We choose to discuss this here in detail because it can potentially be construed to undermine the merit of our study. Because the noble gas makeup of the man- tle is so different from that of the atmosphere, and that of comet 67P, this seems to imply that the onslaught of comets (the timing of which would correlate with the giant planet instability) occurred after the moon form- ing impact. Because the moon forming impact occurred after Mars had completed forming (Kleine et al. 2009; Dauphas & Pourmand 2011), the giant planet instabil- ity could not be the mass-depleting event in the Mars forming region. However this argument does not take in to account the timing of impacts with respect to the Earth's magma ocean phase. It is reasonable to assume that some cometary delivery must have occurred prior to core closure. If fractionalization of Xenon occurred dur- ing the magma ocean phase, the preserved signature in the mantle could very well be different from that in the atmosphere. Because neither the distribution of impact times of primordial Kuiper belt objects (KBOs), nor the fractionalization of Xenon in the magma ocean are well known, drawing a broad conclusion of the timing of the instability from this constraint is difficult. Additionally, delivery itself may have been stochastic; in that an early instability might set the Xenon content of the mantle by delivering many small comets, and then a later impact of a large comet could boost the Xenon fraction in the atmo- sphere. Finally, Xenon isotope trends are not well known over a large enough sample of comets. It is reasonable to expect that comet 67P's specific Xenon concentration would fall somewhere on a continuum when compared to other similar comets. A larger sample of such mea- surements must be made before these conclusions can be applied to the giant planet instability timeline. There- fore, as a starting point for our study, we argue that the most important constraint for instability timing is the survivability of the terrestrial planets, which no scenario to date can ensure. 3. SUCCESS CRITERIA When analyzing the results of our simulations, the pa- rameter space for comparison to the actual solar system is extensive. Furthermore, for many metrics our accuracy is strongly limited by the resolution of our simulations. For example, the planetary embryos used in the major- ity of our simulations begin the integration with a fourth of Mars' present mass. Therefore, a "successful" Mars analog could be formed from as few as 2-3 impacts. For these reasons our criteria must be broad, because we are more interested in looking at statistical consistencies and order of magnitude agreements than perfectly replicating every nuance of the actual solar system. Thus, we focus on 10 broad criteria for replication of both the inner and outer solar systems, which we summarize in table 3. 3.1. The Inner Solar System Because our goal is to look for systems like our own, with particular emphasis on forming Mars analogs, we employ an analysis metric similar to Chambers (2001). A system is considered to meet criterion A if it forms a Mars sized body in the vicinity of Mars' semi-major axis, exterior to two Earth sized bodies. We first check for any planets formed in the region of 1.3-2.0 au, where the inner edge of this region is roughly equal to Mars' current pericenter (∼1.38 au) and the outer limit lies at the inner edge of the asteroid belt. If this planet has a mass less than 0.3 M⊕, is immediately exterior to two planets each with masses greater than 0.6 M⊕, and the system contains no planets greater than 0.3 M⊕ in the asteroid belt, criterion A is satisfied. A separate success criteria (criterion D, section 3.3) filters out sys- tems that finish with an embryo in the Asteroid Belt as unsuccessful. While some authors (Hansen 2009) se- lect 0.2 M⊕ as the upper mass limit for Mars analogs, due to the previously discussed resolution limitation en- countered when using 0.025 M⊕ embryos, we follow the prescription in Raymond et al. (2009) and use 0.3 M⊕ as our limit. Additionally, because our simulations do not take collisional fragmentation in to account (for fur- ther discussion of this phenomenon, see section 5.6), it is possible that the masses of our Mars analogs are some- what over-estimated. Because we set the Venus/Earth minimum mass to 0.6 M⊕ (approximately 75% that of Venus' present mass), this provides an adequate mass disparity of a factor of 2 between the Venus/Earth and Mars analogs. We also look at systems which form three planets of the correct mass (criterion A1), but do not have the correct semi major axes (eg: Mars formed at a semi-major axis greater than 2.0 au). For this crite- rion, we include systems which form no Mars, but do accrete appropriately sized Earth and Venus analogs in the correct locations as being successful. 3.2. The Formation Timescales of Earth and Mars Mars is often thought to have been left behind as a "stranded embryo" (Morbidelli et al. 2000) during the process of planetary formation because the timescale for its accretion inferred from Hf/W dating (.1-10 Myr) is so quick (Nimmo & Agnor 2006; Dauphas & Pourmand 2011). Contrarily, Earth is believed to have formed much slower; of order 50-150 Myr (Touboul et al. 2007; Kleine et al. 2009). There is a significant amount of uncertainty in both of these timescales. The specific timing of the moon forming impact, which is thought to correlate with the last major accretion event on Earth, is still not well known. Unfortunately, these metrics are quite difficult to meet when using standard embryo accretion numeri- cal models. In fact, planets with semi-major axes greater than 1.3 au in our control simulation (which assume no gas giant evolution) almost always form far slower than Code Criterion Actual Value Accepted Value A aM ars A,A1 MM ars A,A1 MV enus A,A1 MEarth τM ars B C D E F G H I I I J τ⊕ MAB ν6 W M F⊕ AMD NGP aGP ¯eGP ¯iGP PSat/PJup 1.52 au 0.107 M⊕ 0.815 M⊕ 1.0 M⊕ 1-10 Myr 50-150 Myr ∼ 0.0004 M⊕ ∼0.09 ∼ 10−3 0.0018 4 5.2/9.6/19/30au 0.046/0.054/0.044/0.01 0.37/0.90/1.02/.67◦ 2.49 7 Justification Inside AB (Raymond et al. 2009) Within ∼ 25% Match Venus 1.3-2.0 au > 0.025, < .3M⊕ >0.6 M⊕ >0.6 M⊕ <10 Myr >50 Myr No embryos (Chambers 2001) <1.0 > 10−4 <0.0036 4 20% <.11 < 2◦ <2.8 Table 3 Order of magnitude (Raymond et al. 2009) (Nesvorn´y & Morbidelli 2012) (Nesvorn´y & Morbidelli 2012) (Nesvorn´y & Morbidelli 2012) (Nesvorn´y & Morbidelli 2012) (Nesvorn´y & Morbidelli 2012) Summary of success criteria for the solar system. The columns are: (1) the semi-major axis of Mars, (2-4) The masses of Mars, Venus and Earth, (5-6) the time for Mars and Earth to accrete 90% of their mass, (7) the final mass of the asteroid belt, (8) the ratio of asteroids above to below the ν6 secular resonance between 2.05-2.8 au, (8) the water mass fraction of Earth, (9) the angular momentum deficit (AMD) of the inner solar system, (10) the final number of giant planets, (11-13) the semi-major axes, time-averaged eccentricities and inclination of the giant planets, (14) the orbital period ratio of Jupiter and Saturn. A complete discussion of the success criteria, background information and justifications is provided in the Supplementary Information. interior planets. With these metrics in mind, we require our Mars analogs accrete 90% of their mass within 10 Myr of the beginning of our terrestrial planetary forma- tion simulations (not the onset of the instability, crite- rion B). Additionally, we require our Earth analogs take at least 50 Myr to accrete 90% of their mass (criterion C) 3.3. The Asteroid Belt Imposing strict constraints on the asteroid belt is diffi- cult because, of the 1000 planetesimals that begin a given simulation, typically only 10 to 30 complete the integra- tion in the asteroid belt region. Furthermore, because the smallest objects in our simulations have masses ∼16 times greater than Ceres, our initial conditions are quite unrealistic for an appropriate study of the asteroid belt. Our ability to model the depletion of the asteroid belt is thus limited. Therefore, we cannot draw any conclusions about the total mass of the asteroid belt as all the parti- cles in our simulation are simply too large. However, the dynamical behavior of our small planetesimals should be roughly similar to that of the larger asteroids in the belt (such as Ceres). Because there are only a few such large asteroids in the actual asteroid belt, it is important that we heavily deplete the region of such objects in our sim- ulations. Several studies have already investigated the effects of the Nice Model on the asteroid belt (O'Brien et al. 2007; Walsh & Morbidelli 2011; Roig & Nesvorn´y 2015; Deienno et al. 2016). A similar study, using tens of thousands of smaller particles in the asteroid belt re- gion will be required to study the particular dynamical constraints our scenario places on the asteroid belt. Fur- thermore, the most successful models for the asteroid belt (Walsh et al. 2011; Raymond & Izidoro 2017b) suc- cessfully reproduce the compositional dichotomy between "S-types" (Silicate rich, moderate albedo asteroids) and "C-types" (low albedo, carbonaceous asteroids making up about 75% of the belt) (Gradie & Tedesco 1982). Im- proving our mass resolution within the asteroid belt in the future will allow us to test this constraint as well. For our purposes, we simply require that no embryos remain in the region (a > 2.0 au). This method (criterion D) is similar to that employed in Chambers (2001) and Raymond et al. (2009). The fact that there are no sig- nificant gaps between observed mean motion and secular resonances in the actual asteroid belt implies an upper limit (about a Mercury mass) for the mass of the largest object in the belt that could survive terrestrial planetary formation (O'Brien & Sykes 2011). Because the total mass of the asteroid belt is thought to have depleted by about a factor of ∼ 104 over the life of the solar system (Petit et al. 2001), a detailed calculation of the actual nu- merical value of what is left over is far beyond the scope of this paper. Moreover, because Gyr timescale modeling of test particles in the asteroid belt indicates that deple- tion is logarithmic over the life of the solar system, the majority of this depletion must happen during the first ∼200 Myr of evolution because loss in the next 4 Gyr is only of order ∼50% (Minton & Malhotra 2010). Additionally, we look at the number of remaining plan- etesimals above and below the ν6 secular resonance be- tween 2.05-2.8 au. In the actual solar system, this ratio is about ∼ .09. Again, due to the small number statis- tics, the ratio inferred from an individual simulation will be very imprecise. Therefore we only require this ratio to be less than one (criterion E). Furthermore, resonance sweeping during giant planet migration and evolution can drastically effect the dy- namical structure of the asteroid belt (Walsh & Mor- bidelli 2011; Minton & Malhotra 2011; Roig & Nesvorn´y 2015). In a slow migration scenario, as Saturn moves outward towards its current semi-major axis, the ν16 sec- ular resonance excites inclinations as it sweeps through the asteroid belt. As Saturn continues to migrate, the ν6 resonance erodes the remaining low inclination, low ec- centricity component. Though the process of resonance sweeping can undoubtedly have an effect on the result- ing mass of Mars (Bromley & Kenyon 2017), the mech- anism can also remove low inclination asteroids which are common (figure 6) in today's asteroid belt (Walsh 8 & Morbidelli 2011). To preserve the structure of the asteroid belt from the effects of resonance dragging, pre- vious authors (Roig & Nesvorn´y 2015) utilized a "jump- ing Jupiter" model instability (wherein Jupiter and Sat- urn "jump" toward their present orbital locations, ide- ally preserving the fragile terrestrial planets and asteroid belt structure). In order for our model to be successful, it must heavily deplete the asteroid belt while still main- taining the low inclination component. 3.4. Water Delivery to Earth Many models trace the origin of Earth's water to the early depletion of the primordial asteroid belt (Raymond et al. 2007, 2009). The actual topic of water delivery to Earth is extensive, with many competing models, and far beyond the scope of this paper (for a more complete discussion of various ideas see Morbidelli et al. (2000), Morbidelli et al. (2012) and Marty et al. (2016)). Un- certainties in the initial disk properties and locations of various snow lines make it challenging for embryo ac- cretion models like our own to confidently quantify the water mass fraction (WMF) of Earth analogs. In addi- tion, the actual bulk water content of Earth is extremely uncertain. Estimates of the mantle's water content range between 0 to tens of oceans (L´ecuyer et al. 1998; Marty 2012; Halliday 2013), awhile the core may contain 0 to nearly 100 (Badro 2014; Nomura et al. 2014). See the re- view by Hirschmann (2006) for a discussion of the differ- ence between the capacity of Earth's water reservoirs and geochemical evidence for the actual water contained in Earth's interior. Furthermore, given the amount of plan- etesimal scattering which occurs when the giant plan- ets grew and migrated during the gas disk phase, the material from which Earth formed during the giant im- pact phase may have already been sufficiently water rich (Raymond & Izidoro 2017a). For our simulations, we first look at the bulk WMF of Earth analogs calculated using an initial water radial distribution similar to that used in Raymond et al. (2009) (equation 1, this assumes that the primordial asteroid belt region was populated by water-rich objects from the outer solar system during the gas disk phase (Raymond & Izidoro 2017a)). Any system which boosts Earth's WMF to greater than 10−4 is considered to satisfy criterion F. We also analyze the percentage of objects Earth analogs accrete from differ- ent sections of the disk. 10−5, 10−3, 10%, W M F = r < 2au 2au < r < 2.5au r > 2.5au (1) 3.5. Angular Momentum Deficit One defining aspect about our solar system is the re- markably low eccentricities and inclinations of the terres- trial planets. Over lengthy integrations, the orbits of all the inner planets but Mercury typically stay extremely low (Quinn et al. 1991; Laskar & Gastineau 2009). This orbital constraint was very difficult for early accretion models to meet (Chambers & Wetherill 1998; Chambers 2001). O'Brien et al. (2006) used dynamical friction to explain how the orbits could stay cold through the stan- dard process of planetary formation. However, it has proven even more challenging to keep eccentricities low when a giant planet instability is considered (Brasser et al. 2009; Kaib & Chambers 2016). Any successful model of terrestrial planetary formation must maintain low orbital excitation in the inner solar system. To mea- sure this in our systems, we measure the angular momen- tum deficit (AMD, criterion F) of each system (Laskar 1997). AMD (equation 2) quantifies the deviation of the orbits in a system from perfectly coplanar, circular or- bits. We follow the same procedure as Raymond et al. (2009) and require our systems maintain an AMD less than twice the value of the modern inner solar system (∼.0018). (cid:80) i mi √ ai[1 −(cid:112)(1 − e2 (cid:80) √ i mi ai AM D = i ) cos ii] (2) 3.6. The Outer Solar System When analyzing the success of our terrestrial plane- tary formation simulations, it is important to consider how dependent our results are on the fate of the outer solar system. Indeed, the chance of our chosen resonant chains reproducing all the important traits of the outer solar system after undergoing an instability is often low. For example, Nesvorn´y & Morbidelli (2012) report only a 33% chance of our n1 configurations finishing the inte- gration with the correct number of outer planets. When all four success criteria in that work are considered, n1 resonant chains only successfully match the outer solar system ∼ 4% of the time. Given the computational cost of our integrations, we are less interested in how often we correctly replicate the orbital architecture of the gi- ant planets, and more concerned with how dependent our results are on the fate of the outer solar system. To quantify the outer solar system, we adopt the same success criteria as Nesvorn´y & Morbidelli (2012). First, criterion H requires that the simulation finish with 4 gi- ant planets. If this is satisfied, criterion I stipulates that the final semi-major axis of each planet be within 20% of the modern location, and the time averaged eccen- tricity and inclination of each planet be less than twice the largest current value in the outer solar system. Fi- nally, criterion J states that the period ratio of Jupiter and Saturn stay less than 2.8. It should be noted that, unlike Nesvorn´y & Morbidelli (2012), we check for crite- rion J independently of whether the other two standards are met. The dynamics of the forming terrestrial plan- ets are far less sensitive to the behavior of the ice giants than they are to that of Jupiter and Saturn. Therefore, we are nearly just as interested in systems that correctly produce Jupiter and Saturn but eject too many ice giants as we are in those that replicate the outer solar system perfectly. 4. RESULTS AND DISCUSSION We provide complete summaries of our results in tables 4, 5 and 6. It should be noted that a small number of our instabilities fail to properly eject an ice giant, and com- plete the integration with greater than four giant planets. Because we are only interested in systems similar to the solar system, we do not include these few outliers in any of our analyses. Table 4 shows the total percentage of systems in each simulation batch which meet our success criteria for the inner solar system. Table 5 summarizes 9 Set A A1 a, mT P mT P B τmars Control n1/.01Myr n2/.01Myr n1/.1Myr n2/.1Myr n1/1Myr n2/1Myr n1/10Myr n2/10Myr 0 3 2 0 6 13 8 12 8 0 15 6 2 10 13 14 20 20 9 31 15 16 20 6 11 3 19 E F C τ⊕ MAB ν6 WMF AMD 86 84 75 79 69 90 87 73 80 53 35 46 39 39 31 42 26 31 87 40 34 27 38 47 44 47 62 G 8 14 9 12 12 13 7 16 8 D 2 41 48 38 50 38 54 48 54 Summary of percentages of systems which meet the various terrestrial planet success criteria established in table 3. It should be noted that, because runs beginning with a disk mass of 3M⊕ were not successful at producing appropriately massed Earth and Venus analogs, criterion A and A1 are only calculated for 5M⊕ systems. The subscripts TP and AB indicate the terrestrial planets and asteroid belt respectively. Table 4 Set n1/.01Myr n2/.01Myr n1/.1Myr n2/.1Myr n1/1Myr n2/1Myr n1/10Myr n2/10Myr H NGP 27 18 14 18 23 22 17 16 I a,e,iGP 14 6 7 3 15 9 9 8 J S:J 47 41 42 38 57 38 36 36 Summary of percentages of systems which meet the various giant planet success criteria established in table 3. The subscript GP indicates the giant planets. Table 5 the percentage of systems satisfying our giant planet suc- cess criteria. We find that our systems adequately repli- cate the outer solar system with frequencies consistent with those reported in Nesvorn´y & Morbidelli (2012). In table 6, we look at our success rates for systems with Jupiter/Saturn configurations most similar to the actual solar system (period ratios less than 2.8). Clearly, the fate of the terrestrial planets is highly dependent on the evolution of the solar system's two giant planets. This is largely due to strong secular perturbations which result from the post-instability excitation of the giant planet orbits. Indeed, when we look at systems which eject all ice giants and finish with a highly eccentric Jupiter and Saturn outside a period ratio of 2.8, we find terrestrial planets which are too few in number, on excited orbits and systematically under-massed. Though we still see these symptoms in some of the systems summarized in table 6, they are noticeably less frequent. 4.1. Formation of a Small Mars An early instability is highly successful at producing a small Mars, regardless of instability timing and the particular evolution of the giant planets. 75% of all our instability systems form either no Mars or a small Mars (less than 0.3 M⊕), as opposed to none of our control runs. Additionally, as shown in table 4, most of our con- trol systems leave at least one embryo in the asteroid belt. In fact, many of these systems even form multiple small planets, or an Earth massed planet in the aster- oid belt. Only 9% of our instability simulations form a planet more massive than Mars in the asteroid belt, as opposed to 65% of our control runs. Clearly a Nice Figure 2. Cumulative distribution of Mars analog masses formed in instability systems and our control batch (note that some sys- tems form multiple planets in this region, here we only plot the largest planet). The vertical line corresponds to Mars' actual mass. All control runs with a Mars analog smaller than 0.3 M⊕ (∼20% of the batch) were unsuccessful in that they also formed a large planet in the asteroid belt Model instability is a highly efficient means of depleting the planetesimal disk region of material outside of 1.3 au. Figure 2 shows the cumulative distribution of the largest planets in each system formed between 1.3 and 2.0 au for our instability sets versus the control batch. The solar system fits in well with this distribution, with slightly greater than half of our systems forming Mars analogs larger than the actual planet. Indeed, the in- stability consistently starves this region of material and produces a small planet. In fact, 22% of our systems pro- duce no planet in the Mars region whatsoever. This is slightly lower for the two earliest instability delays (0.01 and 0.1 Myr) which we test, with 18% of such systems forming no Mars in the 1.3 to 2.0 au region. This is due to the fact that, when the instability occurs, the ratio of the number of planetesimals to embryos, and that of total planetesimal mass to total embryo mass is much higher. In the late instability cases, the majority of the system mass is trapped in several large embryos. The dynamical excitement of the additional planetesimals in the early instability delay cases allows the disk mass to disperse, thereby enhancing the mass of Mars analogs. 10-1100MMars (M⊕)0.00.20.40.60.81.0Cumulative Fraction of Mars AnalogsInstabilityControl10 Set A A1 a, mT P mT P B τmars Control n1/.01Myr n2/.01Myr n1/.1Myr n2/.1Myr n1/1Myr n2/1Myr n1/10Myr n2/10Myr 0 0 5 0 12 26 11 9 20 0 15 5 0 18 26 27 18 33 9 33 17 14 7 12 12 7 27 D E F C τ⊕ MAB ν6 WMF AMD 86 94 80 78 84 95 92 92 87 53 33 60 36 59 29 44 29 52 87 61 56 42 64 65 69 53 77 2 24 30 13 32 20 42 16 25 G 8 15 10 5 11 14 2 25 2 Summary of percentages of systems which meet the various terrestrial planet success criteria established in table 3 AND finish with Jupiter and Saturn's period ratio less than 2.8 (criterion J). It should be noted that, because runs beginning with a disk mass of 3M⊕ were not successful at producing appropriately massed Earth and Venus analogs, criterion A and A1 are only calculated for 5M⊕ systems. The subscripts TP and AB indicate the terrestrial planets and asteroid belt respectively. Table 6 butions for late instabilities are slightly better matches to the actual solar system for two reasons. First, the number of outlying Mars analogs which are larger than ∼ .6M⊕ is substantially less for the later instability delay times. As discussed previously, because these simulations begin with fewer planetesimals, the lack of disk dispersal and de-excitation via dynamical friction between small bodies makes it difficult for the system to accrete a large Mars over the next ∼ 200 Myrs of evolution. Next, these systems tend to form more accurate Earth and Venus analogs. Many of the failed early instability delay sim- ulations are clear examples of the instability's tendency to hinder the formation of Earth and Venus. The larger dispersal of the disk mass profile in these delays consis- tently deprives the Earth and Venus forming regions of material. In these simulations, we often form systems with small (less than ∼ .6M⊕) Venus and Earth analogs, just one total terrestrial planet, or no inner planets at all. 4.2. Strengths of an Early Instability Tables 4 and 6 show the percentage of systems in each batch that meet our success criteria for the inner solar system. Because the solar system very well could have been formed in a low likelihood scenario, it is impor- tant not to place too much weight on meeting specific numerical values and exactly replicating every particular dynamical trait of the actual system. For this reason, we try to keep our success criteria as broad as possible. We consider our systems roughly successful at meet- ing our established criteria. Generally, our systems per- form better than our control runs in almost all categories. Because the unique dynamical state of the solar system represents just one point in a broad spectrum of possible outcomes, it is unreasonable to expect that our simula- tions meet every single success criterion exactly, every time. By these standards, our simulations are successful on most accounts. In particular, an instability is very successful at meeting the requirements for the asteroid belt (criterion D) and the formation timescale of Earth (criterion C). Given the large number of constraints involved in cri- terion A, our success rates of ∼ 5 − 20% are still very encouraging. For this reason, we also use the broader criterion A1 for the orbital architecture of the inner solar system. This metric considers systems that form planets of the correct mass ratios, but incorrect orbital locations, Figure 3. Cumulative mass distribution of embryos and planetes- imals at the beginning of the control simulations (black line), 10 Myr after the early instability delay times (.01 and 0.1 Myr; red line), 100 Myr after the early instability delay times (orange line), 10 Myr after the late instability delay times (1 and 10 Myr; cyan line) and 100 Myr after the late instability delay times (blue line). The green line represents the current mass distribution of the inner solar system. We find that planets in the outer disk (a > 1.3 au) from early instability delay systems (0.01 and 0.1 Myr) where the outer planets meet criterion J accrete ∼ 6 times more material from the inner disk than in late instabilities (1 and 10 Myr). Furthermore, dynamical friction from the higher number of planetesimals de-excites material in the outer disk. Indeed, our early instability delay systems which satisfy criterion J lose an average of 0.25 M⊕ less mass in the outer disk (a > 1.3 au) to ejection or collisions with the Sun than the later delays. Figure 3 shows how a late instability delay results in a dramatically steeper mass distribution profile. In Figure 4 we show the distributions of semi-major axes and masses for the planets we form, compared with our control simulations. Regardless of the instability de- lay time, there is a stark contrast between our simula- tions and the control set. Earth mass planets in the Mars region and beyond are very common in the control sim- ulations, and rarely occur when the system undergoes a Nice Model instability. Though the general trends for all four plots are quite similar, we note that our distri- 0.51.01.52.02.53.03.54.0Semi-Major Axis (au)0.00.20.40.60.81.0Cumulative Mass FractionInitial Control distributionPresent solar system10 Myr after early instability100 Myr after early instability10 Myr after late instability100 Myr after late instability11 Figure 4. Distribution of semi-major axes and masses for all planets formed using 5 M⊕ massed planetesimal disks. The red squares denote the actual solar system values for Mercury, Venus, Earth and Mars. The vertical dashed line separates the Earth and Venus analogs (left side of the line) and the Mars analogs (right side). The top panel shows our control runs and each of the 4 lower plots depict a different instability delay time. 0.51.01.52.0Semi-Major Axis (au)0.250.500.751.001.251.50Mass (M⊕)Control0.51.01.52.0Semi-Major Axis (au)0.250.500.751.001.251.50Mass (M⊕)0.01 Myr0.51.01.52.0Semi-Major Axis (au)0.250.500.751.001.251.50Mass (M⊕)0.1 Myr0.51.01.52.0Semi-Major Axis (au)0.250.500.751.001.251.50Mass (M⊕)1.0 Myr0.51.01.52.0Semi-Major Axis (au)0.250.500.751.001.251.50Mass (M⊕)10.0 Myr12 Figure 5. Semi-Major Axis/Eccentricity plot depicting the evolu- tion of a successful system in the n1/10Myr batch. The size of each point corresponding to the mass of the particle (because Jupiter and Saturn are hundreds of times more massive than the terres- trial planets, we use separate mass scales for the inner and outer planets). The final planet masses are 0.37, 1.0, 0.69 and 0.15 M⊕ respectively. and those which form no Mars but a proper Earth and Venus pair to be successful. When we look at our rates of success for meeting this criterion when Jupiter and Saturn finish the integration within a period ratio of 2.8 (Table 6), we find our later instability delay times are remarkably successful with values closer to ∼ 30%. In these successful scenarios, Mars often forms as a stranded embryo. The simulation begins with multiple bodies of order 0.25-2.5 MM ars in the vicinity of Mars' present or- bit. When the instability ensues, most of these bodies are ejected. 40% of the time, "Mars" undergoes no fur- ther major accretion events with other embryos after the instability simulation begins. Figure 5 shows an example of such an evolution scheme. Notice that after the insta- bility ensues the proto-Venus and proto-Earth continue to accrete material while objects in the Mars forming region do not. 4.2.1. The Asteroid Belt Our simulations are successful at depleting the aster- oid belt because of the dynamical excitation provided by the embryos we place in the belt. The embryos pre-excite the asteroid belt during the evolution leading up to the instability. When the instability ensues, excited plan- etesimals in the belt scatter off the embryos, leading to high mass loss. To test this, we performed a follow-on suite of integrations using our 1 Myr instability control disks, and the Mercury6 hybrid integrator. We place Jupiter and Saturn on orbits corresponding to a period ratio of 1.6, and set an extra ice giant immediately exte- rior to Saturn. When the ice giant scatters and is ejected, Jupiter and Saturn jump. Next, systems where the post- jump period ratio of Jupiter and Saturn is between 2.1 and 2.4 are selected, and integrated for an additional 10 Myr with a code that mimics smooth migration and eccentricity damping on ∼ 3 Myr e-folding timescales us- ing fictitious forces (Lee & Peale 2002). To attain final states similar to Jupiter and Saturn, we shut off migra- tion and eccentricity damping when the two gas giants attain a period ratio above 2.45 and eccentricities be- low 0.06. Through this process, we create a sample of asteroid belts (∼ 20) which experience a pre and post- instability evolution broadly similar to the runs from our original simulations that best matched the currently ob- served orbital architecture of Jupiter and Saturn. By performing 2 sets of runs (embryos and planetesimals and planetesimals only), we are able to test the effects of embryo excitation. Planetesimal only simulations are created by converting all embryos with a > 1.5 au in a given system in to an appropriate number of equal-mass planetesimals with similar semi-major axes, eccentricities and inclinations, and random angular orbital elements. Simulations using embryos and planetesimals lost about twice as much mass beyond 1.5 au over just 10 Myr of evolution as the planetesimal only systems. This is con- sistent with the idea that the dynamical excitement of embryos leads to significantly more mass loss in the as- teroid belt. Our simulations' ability to replicate the asteroid belt population about the ν6 resonance is subject to numerical limitations. Our simulations start with 0.0025 and 0.0015 M⊕ planetesimals, both of which are more massive than the entire present contents of the asteroid belt. Fur- thermore, most simulations finish with between 10 and 30 bodies in the main belt. Such small numbers makes it difficult to discern subtle dynamical features within our individual simulated asteroid belts. Although statis- tics can be improved by co-adding many simulations to examine the general effects of giant planet instabilities (Figure 6), every instability is unique at some level, and dynamical sculpting processes occurring during some in- stabilities may not operate in others. When we co-add the asteroids from all of our criterion J satisfying sim- ulations (those which finish with Jupiter and Saturn's period ratio less than 2.8) and remove objects on planet- crossing orbits, we find a ratio of bodies above to below the ν6 resonance (between 2.05 and 2.8 au) to be ∼ 0.71. However, when we only consider asteroids between 2.05 and 2.5 au, the ratio is a poorer match (2.24). Though neither number is close to the actual ratio (∼ 0.09), the first is quite promising with respect to other numerical modeling attempts. For example, Deienno et al. (2016) imposed a "Grand-Tack" style migration on the asteroid belt and found a ratio of ∼ 1.2. In a similar manner, Walsh & Morbidelli (2011) reported a ratio of ∼ 5.2 in a smooth migration scenario. Due to numerical limitations, further simulations, in- volving tens of thousands of smaller bodies in the as- teroid belt region are required to comprehensively study the detailed effect of an early instability on the asteroid belt. However, an early instability seems to generate an asteroid belt similar to the actual belt in broad strokes. The presence of embryos appears to provide sufficient dynamical excitation to substantially deplete the mass 0.00.10.20.3t=0: Dispersal of the gas disk.0.00.10.20.3t=10 Myr: The instability istriggered.0.00.10.20.3Eccentricityt=20 Myr: Mars growthcomplete. Earth and Venusstill forming.0.00.10.20.3t=210 Myr: Final state of thesystem.110Semi-Major Axis (au)0.00.10.20.30.4Present Solar System.13 the ∼ 10% level after ∼ 2-4 Myr (Dauphas & Pourmand 2011), our 1 Myr instability delays are the most suc- cessful at simultaneously matching the mass distribution of the terrestrial system and the proposed accretion his- tory of Mars. However, the geological accretion history of Mars is inferred relative to CAI formation; the tim- ing relative to gas disk dissipation of which is not fully understood. Providing a means of water delivery to Earth is not a strict requirement for the success of an embryo accretion model. It should be noted that many ideas for how Earth was populated with water exist, several of which have nothing to do with delivery via bodies from the outer solar system (Morbidelli et al. 2000, 2012). In fact, wa- ter delivering planetesimals may have been scattered on to Earth-crossing orbits during the giant planets' growth and migration phase (Raymond & Izidoro 2017a). De- spite the small number statistics involved with using only 1000 initial particles in the Kuiper belt, about half of which typically deplete in the initial phase of integration (before the terrestrial disks are imbedded), we do find 12 instances of Earth analogs accreting objects from this region in our simulations. Interestingly, Earth's noble gases are thought to come primarily from comets, de- spite the fact that comets are likely a minor source of water (Marty et al. 2016). We find that a late instability delay time (1 and 10 Myr) systematically stretches the feeding zone of Earth analogs further in to the terrestrial disk. This broader feeding zone is basically a result of eccentricity excitation of planetesimals (Levison & Ag- nor 2003b). In these cases, mass from the outer disk is able to "leap-frog" its way towards the proto-Earth. In the first phase of evolution (before the instability), form- ing embryos in the middle part of the disk (∼2.0-3.0 au) accrete material from the outermost section of the disk (∼3.0-4.0 au). When the instability ensues, these em- bryos are destabilized, and occasionally scattered inward towards the forming "Earth". Our simulations often leave the inner planets with too large of an AMD. Though most runs only exceed the ac- tual AMD of the solar system by a factor of 2-3, some systems occasionally reach AMDs as high as 10 times the value of the current solar system. Many of these outliers are from integrations where particularity violent instabilities leave behind a system of overly excited giant planets. Even when we remove these instances which are not analogous to the actual solar system, our "success- ful" simulations still tend to possess high AMD values. Often, an overly excited Mars is the source of this or- bital excitation (values of eM ars ∼ .1 − .25 are typical for these systems). The obvious source of this excita- tion is secular interactions with the excited giant plan- ets. One potential solution to this problem might be ac- counting for collisional fragmentation. Chambers (2013) showed that angular momentum exchange resulting from hit-and-run collisions noticeably reduces the eccentricity of planets formed in embryo accretion models. Addition- ally, Jacobson & Morbidelli (2014) showed that the AMD of systems increases as the total amount of initial mass placed in embryos instead of planetesimals increases, and as individual embryo mass decreases. We observe a sim- ilar relationship in overall disk mass loss (section 5.1). Moreover, because of the chaotic nature of the actual so- lar system, its AMD can evolve by as much as a factor of Figure 6. The upper plot shows the inclination distribution of the modern asteroid belt (only bright objects with absolute mag- nitude H < 9.7, approximately corresponding to D > 50 km, are plotted). The bottom plot combines all planetesimals remaining in the asteroid belt region from all instability simulations that form a Mars analog less massive than 3 times Mars' actual mass, and finish with Jupiter and Saturn's period ratio less than 2.8. Grey points correspond to high-eccentricity asteroids on Mars crossing orbits which will be naturally removed during subsequent evolution up to the solar system's present epoch. The vertical dashed lines represent the locations of the important mean motion resonances with Jupiter. The bold dashed lines indicate the current location of the ν6 secular resonance. in the region (most simulations deplete more than 95% of belt material in 200 Myr). Though this does fall short of the required depletion of a factor of ∼ 104 (Petit et al. 2001), our mechanism does produce substantial deple- tion in the asteroid belt when compared with our control runs. More realistic initial conditions and handling of collisions will be required to more accurately model de- pletion in the Asteroid Belt in our model. Nevertheless, embryos remaining in the asteroid belt are extremely rare in our instability systems. By using a full instability, rather than a smooth migration scenario, we avoid drag- ging resonances across the belt, thus broadly preserving its orbital structure. 4.3. Weaknesses of an Early Instability On average, our instability simulations are less suc- cessful at meeting the success criteria for the formation timescale of Mars, the WMF of Earth and the AMD of the terrestrial planets. Reproducing the formation timescale of Mars is a difficult constraint for N-body ac- cretion models of terrestrial planetary formation. A suc- cessful Mars analog in our simulations need only be com- posed of 4 embryos. Meanwhile, the real Mars formed from millions of smaller objects that accreted prior to and during the giant impact phase. This difference must be weighed when considering moderate discrepan- cies between the formation timescales of simulated Mars analogs and the real planet. In fact, ∼ 40% of all our Mars analogs undergo no impacts with other embryos following the instability, and Mars' form on average ∼ 39 Myr faster than their Earth counterparts. Additionally, 6 of the 7 Mars analogs in the criterion A1 satisfying [n2/10 Myr] batch (our most successful simulations), form in under 10 Myr. Because Mars' growth only continued at 010203040ν63:15:27:32:1AsteroidsMars Crossing2.02.22.42.62.83.03.23.4Semi-Major Axis (au)0102030400.00.20.40.60.81.00.00.20.40.60.81.0Inclination (degrees)14 2 in either direction over Gyr timescales (Laskar 1997). 4.4. Varied Initial Conditions Our simulations are broken up into 4 different sets of 25 runs with unique inner disk edge and initial disk mass combinations (Table 2). In half of our simulations, we use a disk mass of 3 M⊕ rather than a more typical choice of ∼ 5M⊕ (Chambers 2001; Raymond et al. 2009). 100% of these systems with lower mass disks fail to meet crite- rion A for correctly replicating the semi-major axes and masses of the terrestrial planets. Using a lower overall disk mass leads to less dynamical friction available to save bodies from loss after the instability. We find that by far the most likely final configuration for these simula- tions is a single Venus analog, occasionally accompanied by a Mars analog. However, we note that the percentages of systems that meet the other 6 success criteria (crite- rion B through G) are roughly similar (within ∼ 5%) for systems of either initial disk mass. We see no noticeable differences between the sets of simulations which truncate the inner planetesimal disk at 0.5 au and those with an inner edge at 0.7 au. Both batches are roughly equally likely (9% and 10% of the time, respectively) to form a Mercury analog (we define this as any planet smaller than 0.2 M⊕ interior to an Earth and a Venus analog). For more discussion on the formation of Mercury, see section 6. Finally, our rates for meeting all success criteria for the inner planets are roughly the same (within ∼ 5%), regardless of the se- lected inner disk edge location. 4.5. Excitation of Jupiter's g5 Mode. The sufficient excitation of Jupiter's g5 mode is another important constraint on the evolution of the giant plan- ets. The current amplitude of the mode, e55 = 0.044, is very important in driving the secular evolution of the solar system (Morbidelli et al. 2009b). Additionally, the amplitude of Saturn's forcing on Jupiter's eccentricity, e56 = 0.016, is important for the long term evolution of Mars and the asteroid belt. Overexciting e56 might lead to a small Mars and a depleted asteroid belt. How- ever, this scenario is not akin to the actual evolution of the solar system. To evaluate the relationship between the g5 mode and the mass of Mars, we integrate all sys- tems which finish with Jupiter and Saturn within a pe- riod ratio of 2.8 for an additional 10 Myr, and perform a Fourier analysis of the additional evolution (Sidlichovsk´y & Nesvorn´y 1996). In Figure 7, we plot the values of e55 and e56 against the masses of Mars analogs produced for these systems in our 10 Myr delayed instabilities. We find that systems with an e55 amplitude greater than that of actual solar system never produce a large Mars analog (greater than 0.3 M⊕). The average Mars analog mass in systems with e55 less than half the solar sys- tem value (0.022) is 1.96 times Mars' mass, compared to 1.03 for systems with e55 > 0.022. Additionally, we see multiple examples of systems where e56 is close to the solar system value, that produce a small Mars. Clearly, the complete excitation of the g5 mode is linked to re- ducing the mass of planets in the Mars forming region. Additionally, in Figure 8, we plot the normalized AMD of Jupiter and Saturn versus the mass of Mars analogs. It is very apparent that the range of possible values is Figure 7. Values of the amplitudes of Jupiter's g5 mode versus the mass of Mars analogs formed for 10 Myr delayed instability systems where the orbital period ratio of Saturn to Jupiter completing the integration less than 2.8. The red stars correspond with the present solar system values. extensive, with the solar system falling well within the range of our results. Therefore, the actual solar system is consistent with our dynamical evolution model. Further- more, Jupiter's excitation also effects Earth and Venus. When we plot the cumulative mass of Earth and Venus against the mass of Mars (Figure 9), we find that many of the systems with similar values to the solar system have correspondingly similar values of e55. 4.6. Impact Velocities Our simulations use an integration scheme where all collisions are assumed to be perfectly accretionary (Chambers 1999). This provides a decent approxima- tion of the final outcome of terrestrial planet formation for low relative-velocity collisions between objects with a large mass disparity. However, higher velocity colli- sions can often be erosive (Genda et al. 2012), particu- larly when the projectile to target mass ratio is closer to unity. Additionally, depending on the parameters of the impact, glancing blows can lead to the re-accretion of either all, some or none of the original projectile (As- phaug et al. 2006; Asphaug 2010; Stewart & Leinhardt 2012). Because of the instability's tendency to excite small planetesimals on to high-eccentricity orbits, the 0.000.020.040.06Jupiter Excitation (e55)0.00.20.40.60.81.0MMars (M⊕)0.000.020.040.06Saturn Excitation (e56)15 lisions (36% by mass as opposed to less than 20% for Earth and Venus analogs). This indicates that our Mars analog masses are most likely over-estimated. Addition- ally, because the effects of hit and run collisions have been shown to reduce the AMD of planets produced (Cham- bers 2013), it is possible that the resulting orbital eccen- tricities and inclinations of our Mars analogs are similarly over-excited. This is encouraging because the excitation of Mars significantly contributes to our systematically high AMDs. 5. CONCLUSIONS In this paper, we have presented 800 direct numeri- cal simulations of a giant planet instability occurring in conjunction with the process of terrestrial planet forma- tion. By timing this violent event within the first ∼100 Myr following the dispersion of the gas disk, the instabil- ity scenario no longer requires a mechanism to prevent the destabilization and loss of the fully formed terres- trial planets terrestrial planets (such as the "Jumping Jupiter" model). When we scrutinize our fully formed systems against a wide range of success criteria, we note multiple statistical consistencies between our simulated planets and the actual terrestrial system. First, our Mars analogs are more likely to form small and quickly. In fact, 75% of all our instabilities form either no planet in the Mars region whatsoever, or an appropriately sized Mars. Additionally, cases where the instability is delayed 1-10 Myr after the beginning of the giant impact phase tend to be more successful than earlier timings (<1 Myr). In many of these runs, the instability itself sets the geo- logical formation timescale of Mars. Thus, an early gi- ant planet instability provides a natural explanation for how Mars survived the process of planet formation as a "stranded embryo." We find that our simulated asteroid belts are largely depleted of mass when compared with our control set of simulations, and seldom form a planet in the belt region. Furthermore, the broad orbital distribution of the aster- oid belt seems to be well matched when we co-add the remaining asteroids from all of our simulations. Because the instability itself is inherently chaotic, each resulting system of giant planets has slightly different orbital char- acteristics. When we filter out systems where the giant planets orbits most closely resemble those in the actual solar system, we find higher rates of success among the corresponding terrestrial systems. At first glance, certain geochemical and dynamical con- straints somewhat conflict with an instability occurring 1-10 Myr after gas disk dispersal. New isotopic data from comet 67P suggests that ∼22% of Earth's atmospheric noble gases were delivered via cometary impacts after the Earth had fully formed (Marty et al. 2017). Because the giant planet instability is the most likely source of such a cometary onslaught, this seems to suggest that the in- stability occured after the conclusion of terrestrial planet formation. However, considerable uncertainty remains in the interpretation of this noble gas signature. Another potential conflict with our result is related to the mod- ern Kuiper belt. The classical Kuiper belt population on high inclination orbits can be explained if Neptune ini- tially migrated in a slow, smooth fashion for at least 10 Myrs after gas disk dispersal before being interrupted by the giant planet instability (Nesvorn´y 2015a). Though Figure 8. The AMD of Jupiter and Saturn (normalized to the actual solar system value) versus the mass of Mars analogs formed for all instability systems. The red star denotes the solar system values. Figure 9. Values of the total mass of Earth and Venus analogs versus the mass of Mars analogs formed in instability systems where the value of Jupiter and Saturn's period ratio finished the simu- lation less than 2.8. The color of each point corresponds to the amplitude of Jupiter's g5 mode. The blue star denotes actual solar system values. collisional velocities in our simulations are often quite large (occasionally in excess of 10 times the mutual es- cape velocity). Because of this, it is very important to consider the effects of collisional fragmentation of bodies when analyzing our results. To check our simulations for erosive collisions, we use a code which determines the collision type from the colli- sion speed and impact angle by following the parameter space of gravity dominated impacts mapped by Stewart & Leinhardt (2012). We find that erosive collisions do occur with the forming planets in our simulations. How- ever, they are infrequent, and comprise less than ∼ 5% of all collisions and less than ∼ 1% by mass. Erosive col- lisions occur at similar rates for Earth, Venus and Mars analogs, and are almost always planetesimal-on-embryo impacts. We do note, however, that our Mars analogs undergo a significantly higher number of hit and run col- 0.00.20.40.60.81.01.2MMars (M⊕)10-1100101102AMDJS0.00.20.40.60.81.0MMars (M⊕)0.51.01.52.0MVenus+Earth (M⊕)10-310-210-1e5516 such a timing matches our longest delay, pushing the in- stability time later leads to a conflict with constraints on Mars' accretion history (Dauphas & Pourmand 2011). However, our understanding of the Kuiper belt's origins and Mars' formation timescale are subjects of ongoing study and continually evolving. Given this, we believe this is not enough to rule out the premise of our model, especially because it is able to replicate so many features of the inner solar system. The Nice Model explains a number of aspects of the outer solar system. We have shown that, if it occurred within 10 Myr of the dissipation of the gaseous disk, the instability produces inner solar system analogues that match many important observational constraints regard- ing the formation of Mars and the asteroid belt. In con- trast, simulations lacking an instability consistently yield Mars analogs that are too massive, form too slowly, and are surrounded by over-developed asteroid belts. By in- cluding a giant planet instability, these same simulations show a dramatic decrease in the mass and formation timescale of Mars, and adequate depletion in the asteroid belt. 6. FUTURE WORK Our simulations are insufficient to study the large scale structure of the asteroid belt. Simulations utilizing tens of thousands of smaller particles in the asteroid belt are necessary to test if this model can correctly match the orbital distribution of known asteroids. Additionally, the large scale distribution of "S-types" and "C-types" is an important dynamical constraint which must be ac- counted for (Gradie & Tedesco 1982; Bus & Binzel 2002; DeMeo & Carry 2013). Moreover, as our knowledge of as- teroids continues to undoubtedly expand through further missions such as Asteroid Redirect, Dawn, OSIRIS-Rex and Lucy, so will the number of constraints. Another major shortcoming of our project is that we do not account for the collisional fragmentation of col- liding bodies. Higher velocity collisions can be erosive, rather than accretionary. Given the highly excited orbits which are produced by the Nice Model instability in our simulations, accounting for the fragmentation of bodies is supremely important. Furthermore, Chambers (2013) showed that accounting for hit-and-run collisions in em- bryo accretion models can result in producing planets on colder orbits than when using standard integration schemes. Consistent with most previous N-body simula- tions of the late stages of terrestrial planetary formation, our integrations fail to produce Mercury analogs with any reasonable consistency (Chambers 2001; O'Brien et al. 2006; Raymond et al. 2009; Kaib & Cowan 2015). Per- haps accounting for the fragments of material ejected in high velocity collisions with Venus analogs might provide insight into understanding this problem. Finally, the capture of Mars' trojan satellites (some of which are rare, olivine-rich "A-type" asteroids) is un- doubtedly affected by the strong excitation of orbits we see in the proto-Mars region of our simulations (Tabach- nik & Evans 1999). The trojans of Mars are the only such objects in the inner solar system with orbits stable over Gyr timescales. Their unique compositions repre- sent a potential observational constraint for N-body in- tegrations of terrestrial planetary formation (Polishook et al. 2017). ACKNOWLEDGMENTS M.S.C. and N.A.K. thank the National Science Foun- dation for support under award AST-1615975. S.N.R. thanks the Agence Nationale pour la Recherche for support via grant ANR-13-BS05-0 0 03-0 02 (grant MOJO). K.J.W. thanks NASA's SSERVI program (In- stitute of the Science of Exploration Targets) through institute grant number NNA14AB03A. The majority of computing for this project was performed at the OU Supercomputing Center for Education and Research (OSCER) at the University of Oklahoma (OU). Ad- ditional analyses and simulations were done using re- sources provided by the Open Science Grid (Pordes et al. 2007; Sfiligoi et al. 2009), which is supported by the National Science Foundation award 1148698, and the U.S. Department of Energy's Office of Sci- ence. Control simulations were managed on the Nielsen Hall Network using the HTCondor software package: https://research.cs.wisc.edu/htcondor/. This research is part of the Blue Waters sustained-petascale comput- ing project, which is supported by the National Science Foundation (awards OCI-0725070 and ACI-1238993) and the state of Illinois. Blue Waters is a joint effort of the University of Illinois at Urbana-Champaign and its National Center for Supercomputing Applications (Bode et al. 2013; Kramer et al. 2015). REFERENCES Agnor, C. B., & Lin, D. N. C. 2012, ApJ, 745, 143 Asphaug, E. 2010, Chemie der Erde / Geochemistry, 70, 199 Asphaug, E., Agnor, C. B., & Williams, Q. 2006, Nature, 439, 155 Badro, J. 2014, Annual Review of Earth and Planetary Sciences, 42, 231 Barr, A. C., & Canup, R. M. 2010, Nature Geoscience, 3, 164 Baruteau, C., Crida, A., Paardekooper, S.-J., et al. 2014, Protostars and Planets VI, 667 Batygin, K., & Brown, M. E. 2010a, ApJ, 716, 1323 -. 2010b, ApJ, 716, 1323 Batygin, K., Brown, M. E., & Betts, H. 2012, ApJ, 744, L3 Beaug´e, C., & Nesvorn´y, D. 2012, ApJ, 751, 119 Bode, B., Butler, M., Dunning, T., et al. 2013, in Contemporary High Performance Computing, Chapman & Hall/CRC Computational Science (Chapman and Hall/CRC), 339–366 Boehnke, P., Harrison, T. M., Heizler, M. T., & Warren, P. H. 2016, Earth and Planetary Science Letters, 453, 267 Bottke, W. F., Walker, R. J., Day, J. M. D., Nesvorny, D., & Elkins-Tanton, L. 2010, Science, 330, 1527 Brasser, R., Matsumura, S., Ida, S., Mojzsis, S. J., & Werner, S. C. 2016, ApJ, 821, 75 Brasser, R., Morbidelli, A., Gomes, R., Tsiganis, K., & Levison, H. F. 2009, A&A, 507, 1053 Brasser, R., Walsh, K. J., & Nesvorn´y, D. 2013, MNRAS, 433, 3417 Bromley, B. C., & Kenyon, S. J. 2017, AJ, 153, 216 Bus, S. J., & Binzel, R. P. 2002, Icarus, 158, 146 Chambers, J. E. 1999, MNRAS, 304, 793 -. 2001, Icarus, 152, 205 -. 2007, Icarus, 189, 386 -. 2013, Icarus, 224, 43 Chambers, J. E., & Cassen, P. 2002, Meteoritics and Planetary Science, 37, 1523 Chambers, J. E., & Wetherill, G. W. 1998, Icarus, 136, 304 Chapman, C. R., Cohen, B. A., & Grinspoon, D. H. 2007, Icarus, 189, 233 Chatterjee, S., Ford, E. B., Matsumura, S., & Rasio, F. A. 2008, ApJ, 686, 580 Clement, M. S., & Kaib, N. A. 2017, Icarus, 288, 88 Dauphas, N., & Pourmand, A. 2011, Nature, 473, 489 Day, J. M. D., Pearson, D. G., & Taylor, L. A. 2007, Science, 315, 217 17 Deienno, R., Gomes, R. S., Walsh, K. J., Morbidelli, A., & Nesvorn´y, D. 2016, Icarus, 272, 114 Masset, F., & Snellgrove, M. 2001, MNRAS, 320, L55 Matsumura, S., Thommes, E. W., Chatterjee, S., & Rasio, F. A. Deienno, R., Morbidelli, A., Gomes, R. S., & Nesvorn´y, D. 2017, 2010, ApJ, 714, 194 AJ, 153, 153 Mills, S. M., Fabrycky, D. C., Migaszewski, C., et al. 2016, Delisle, J.-B., Correia, A. C. M., & Laskar, J. 2015, A&A, 579, Nature, 533, 509 A128 DeMeo, F. E., & Carry, B. 2013, Icarus, 226, 723 Dr¸azkowska, J., Alibert, Y., & Moore, B. 2016, A&A, 594, A105 Duncan, M. J., Levison, H. F., & Lee, M. H. 1998, AJ, 116, 2067 Fabrycky, D. C., & Murray-Clay, R. A. 2010, ApJ, 710, 1408 Fassett, C. I., Head, J. W., Kadish, S. J., et al. 2012, Journal of Minton, D. A., & Malhotra, R. 2010, Icarus, 207, 744 -. 2011, ApJ, 732, 53 Minton, D. A., Richardson, J. E., & Fassett, C. I. 2015, Icarus, 247, 172 Morbidelli, A., Brasser, R., Gomes, R., Levison, H. F., & Tsiganis, K. 2010, AJ, 140, 1391 Geophysical Research (Planets), 117, E00H06 Morbidelli, A., Brasser, R., Tsiganis, K., Gomes, R., & Levison, Fernandes, V. A., Burgess, R., & Turner, G. 2000, Meteoritics H. F. 2009a, A&A, 507, 1041 and Planetary Science, 35, 1355 Fernandez, J. A., & Ip, W.-H. 1984, Icarus, 58, 109 Fischer, R. A., & Ciesla, F. J. 2014, Earth and Planetary Science Letters, 392, 28 Genda, H., Kokubo, E., & Ida, S. 2012, ApJ, 744, 137 Gomes, R., Levison, H. F., Tsiganis, K., & Morbidelli, A. 2005, Nature, 435, 466 Gomes, R. S. 2003, Icarus, 161, 404 Gradie, J., & Tedesco, E. 1982, Science, 216, 1405 Grimm, S. L., & Stadel, J. G. 2014, ApJ, 796, 23 Hahn, J. M., & Malhotra, R. 1999, AJ, 117, 3041 Haisch, Jr., K. E., Lada, E. A., & Lada, C. J. 2001, ApJ, 553, L153 Halliday, A. N. 2008, Philosophical Transactions of the Royal Society of London Series A, 366, 4163 -. 2013, Geochim. Cosmochim. Acta, 105, 146 Hansen, B. M. S. 2009, ApJ, 703, 1131 Hirschmann, M. M. 2006, Annual Review of Earth and Planetary Sciences, 34, 629 -. 2009b, A&A, 507, 1041 Morbidelli, A., Chambers, J., Lunine, J. I., et al. 2000, Meteoritics and Planetary Science, 35, 1309 Morbidelli, A., & Crida, A. 2007, Icarus, 191, 158 Morbidelli, A., Levison, H. F., Tsiganis, K., & Gomes, R. 2005, Nature, 435, 462 Morbidelli, A., Lunine, J. I., O'Brien, D. P., Raymond, S. N., & Walsh, K. J. 2012, Annual Review of Earth and Planetary Sciences, 40, 251 Morbidelli, A., Nesvorny, D., Laurenz, V., et al. 2018, Icarus, 305, 262 Morbidelli, A., Tsiganis, K., Crida, A., Levison, H. F., & Gomes, R. 2007, AJ, 134, 1790 Morishima, R., Stadel, J., & Moore, B. 2010, Icarus, 207, 517 Nesvorn´y, D. 2011, ApJ, 742, L22 -. 2015a, AJ, 150, 73 -. 2015b, AJ, 150, 68 Nesvorn´y, D., & Morbidelli, A. 2012, AJ, 144, 117 Nesvorn´y, D., Vokrouhlick´y, D., & Morbidelli, A. 2007, AJ, 133, Hoffmann, V., Grimm, S. L., Moore, B., & Stadel, J. 2017, 1962 MNRAS, 465, 2170 Holman, M. J., Fabrycky, D. C., Ragozzine, D., et al. 2010, -. 2013, ApJ, 768, 45 Nimmo, F., & Agnor, C. B. 2006, Earth and Planetary Science Science, 330, 51 Letters, 243, 26 Izidoro, A., Haghighipour, N., Winter, O. C., & Tsuchida, M. 2014, ApJ, 782, 31 Nomura, R., Hirose, K., Uesugi, K., et al. 2014, Science, 343, 522 O'Brien, D. P., Morbidelli, A., & Bottke, W. F. 2007, Icarus, 191, Izidoro, A., Raymond, S. N., Morbidelli, A., & Winter, O. C. 434 2015, MNRAS, 453, 3619 O'Brien, D. P., Morbidelli, A., & Levison, H. F. 2006, Icarus, 184, Jacobson, S. A., & Morbidelli, A. 2014, Philosophical Transactions of the Royal Society of London Series A, 372, 0174 Jacobson, S. A., & Walsh, K. J. 2015, Washington DC American Geophysical Union Geophysical Monograph Series, 212, 49 Juri´c, M., & Tremaine, S. 2008, ApJ, 686, 603 Kaib, N. A., & Chambers, J. E. 2016, MNRAS, 455, 3561 Kaib, N. A., & Cowan, N. B. 2015, Icarus, 252, 161 Kenyon, S. J., & Bromley, B. C. 2006, AJ, 131, 1837 Kleine, T., Touboul, M., Bourdon, B., et al. 2009, Geochim. Cosmochim. Acta, 73, 5150 Kley, W., & Nelson, R. P. 2012, ARA&A, 50, 211 Kobayashi, H., & Dauphas, N. 2013, Icarus, 225, 122 Kokubo, E., & Genda, H. 2010, ApJ, 714, L21 Kokubo, E., & Ida, S. 1998, Icarus, 131, 171 Kramer, W., Butler, M., Bauer, G., Chadalavada, K., & Mendes, C. 2015, in High Performance Parallel I/O, ed. Prabhat & Q. Koziol (CRC Publications, Taylor and Francis Group), 17–32 Laskar, J. 1997, A&A, 317, L75 Laskar, J., & Gastineau, M. 2009, Nature, 459, 817 L´ecuyer, C., Grandjean, P., Barrat, J.-A., et al. 1998, Geochim. Cosmochim. Acta, 62, 2429 Lee, M. H., & Peale, S. J. 2002, ApJ, 567, 596 Levison, H. F., & Agnor, C. 2003a, AJ, 125, 2692 -. 2003b, AJ, 125, 2692 Levison, H. F., Kretke, K. A., Walsh, K. J., & Bottke, W. F. 2015, Proceedings of the National Academy of Science, 112, 14180 Levison, H. F., Morbidelli, A., Van Laerhoven, C., Gomes, R., & Tsiganis, K. 2008, Icarus, 196, 258 Lykawka, P. S., & Ito, T. 2013, ApJ, 773, 65 Malhotra, R. 1993, Nature, 365, 819 -. 1995, AJ, 110, 420 Marty, B. 2012, Earth and Planetary Science Letters, 313, 56 Marty, B., Avice, G., Sano, Y., et al. 2016, Earth and Planetary Science Letters, 441, 91 39 O'Brien, D. P., & Sykes, M. V. 2011, Space Sci. Rev., 163, 41 Papaloizou, J. C. B., & Larwood, J. D. 2000, MNRAS, 315, 823 Papaloizou, J. C. B., & Nelson, R. P. 2003, MNRAS, 339, 983 Pascucci, I., Apai, D., Luhman, K., et al. 2009, ApJ, 696, 143 Petit, J.-M., Morbidelli, A., & Chambers, J. 2001, Icarus, 153, 338 Pierens, A., & Nelson, R. P. 2008, A&A, 482, 333 Polishook, D., Jacobson, S. A., Morbidelli, A., & Aharonson, O. 2017, Nature Astronomy, 1, 0179 Pordes, R., Petravick, D., Kramer, B., et al. 2007, Journal of Physics: Conference Series, 78, 012057 Quinn, T. R., Tremaine, S., & Duncan, M. 1991, AJ, 101, 2287 Raymond, S. N., Armitage, P. J., & Gorelick, N. 2010, ApJ, 711, 772 Raymond, S. N., & Izidoro, A. 2017a, Icarus, 297, 134 -. 2017b, Science Advances, 3, e1701138 Raymond, S. N., & Morbidelli, A. 2014, in IAU Symposium, Vol. 310, Complex Planetary Systems, Proceedings of the International Astronomical Union, 194–203 Raymond, S. N., O'Brien, D. P., Morbidelli, A., & Kaib, N. A. 2009, Icarus, 203, 644 Raymond, S. N., Quinn, T., & Lunine, J. I. 2006, Icarus, 183, 265 -. 2007, Astrobiology, 7, 66 Rivera, E. J., Laughlin, G., Butler, R. P., et al. 2010, ApJ, 719, 890 Roig, F., & Nesvorn´y, D. 2015, AJ, 150, 186 Roig, F., Nesvorn´y, D., & DeSouza, S. R. 2016, ApJ, 820, L30 Rubie, D. C., Laurenz, V., Jacobson, S. A., et al. 2016, Science, 353, 1141 Sfiligoi, I., Bradley, D. C., Holzman, B., et al. 2009, in (IEEE Publishing), 428–432 Snellgrove, M. D., Papaloizou, J. C. B., & Nelson, R. P. 2001, A&A, 374, 1092 Spudis, P. D., Wilhelms, D. E., & Robinson, M. S. 2011, Journal of Geophysical Research (Planets), 116, E00H03 Marty, B., Altwegg, K., Balsiger, H., et al. 2017, Science, 356, Stewart, S. T., & Leinhardt, Z. M. 2012, ApJ, 751, 32 1069 18 Stoer, J., Bartels, R., Gautschi, W., Bulirsch, R., & Witzgall, C. Tsiganis, K., Gomes, R., Morbidelli, A., & Levison, H. F. 2005, 2002, Introduction to Numerical Analysis, Texts in Applied Mathematics (Springer New York) Tabachnik, S., & Evans, N. W. 1999, ApJ, 517, L63 Tanaka, H., & Ward, W. R. 2004, ApJ, 602, 388 Tera, F., Papanastassiou, D. A., & Wasserburg, G. J. 1974, Earth and Planetary Science Letters, 22, 1 Nature, 435, 459 Sidlichovsk´y, M., & Nesvorn´y, D. 1996, Celestial Mechanics and Dynamical Astronomy, 65, 137 Walker, R. J. 2009, Chemie der Erde / Geochemistry, 69, 101 Walker, R. J., Horan, M. F., Shearer, C. K., & Papike, J. J. 2004, Earth and Planetary Science Letters, 224, 399 Thommes, E., Nagasawa, M., & Lin, D. N. C. 2008, ApJ, 676, 728 Thommes, E. W., Duncan, M. J., & Levison, H. F. 1999, Nature, Walsh, K. J., & Morbidelli, A. 2011, A&A, 526, A126 Walsh, K. J., Morbidelli, A., Raymond, S. N., O'Brien, D. P., & 402, 635 Touboul, M., Kleine, T., Bourdon, B., Palme, H., & Wieler, R. 2007, Nature, 450, 1206 Trifonov, T., Kurster, M., Zechmeister, M., et al. 2017, ArXiv e-prints, arXiv:1706.00509 Mandell, A. M. 2011, Nature, 475, 206 Wetherill, G. W. 1991, Science, 253, 535 -. 1996, Ap&SS, 241, 25 Wisdom, J., & Holman, M. 1991, AJ, 102, 1528 Zellner, N. E. B. 2017, Origins of Life and Evolution of the Biosphere, arXiv:1704.06694
1911.05577	1	1911	2019-11-13T16:17:36	Constraining the magnitude of climate extremes from time-varying instellation on a circumbinary terrestrial planet	[ "astro-ph.EP", "physics.ao-ph" ]	Planets that revolve around a binary pair of stars are known as circumbinary planets. The orbital motion of the stars around their center of mass causes a periodic variation in the total instellation incident upon a circumbinary planet. This study uses both an analytic and numerical energy balance model to calculate the extent to which this effect can drive changes in surface temperature on circumbinary terrestrial planets. We show that the amplitude of the temperature variation is largely constrained by the effective heat capacity, which corresponds to the ocean-to-land ratio on the planet. Planets with large ocean fractions should experience only modest warming and cooling of only a few degrees, which suggests that habitability cannot be precluded for such circumbinary planets. Planets with large land fractions that experience extreme periodic forcing could be prone to changes in temperature of tens of degrees or more, which could drive more extreme climate changes that inhibit continuously habitable conditions.	astro-ph.EP	astro-ph	Haqq-Misra et al., Journal of Geophysical Research - Planets, accepted on 13 November 2019 Constraining the magnitude of climate extremes from time-varying instellation on a circumbinary terrestrial planet Jacob Haqq-Misra1, Eric T. Wolf2, William F. Welsh3, Ravi Kumar Kopparapu4, Veselin Kostov4, and Stephen R. Kane5 1Blue Marble Space Institute of Science, Seattle, Washington, USA. 2University of Colorado Boulder, Boulder, Colorado, USA, USA. 4NASA Goddard Space Flight Center, Greenbelt, Maryland, USA. 5University of California Riverside, Riverside, California, USA. 3San Diego State University, San Diego, California, USA. Abstract Planets that revolve around a binary pair of stars are known as circumbinary planets. The orbital motion of the stars around their center of mass causes a periodic variation in the total instellation incident upon a circumbinary planet. This study uses both an analytic and numerical energy balance model to calculate the extent to which this effect can drive changes in surface temperature on circumbinary terrestrial planets. We show that the amplitude of the temperature variation is largely constrained by the effective heat capacity, which corresponds to the ocean-to-land ratio on the planet. Planets with large ocean fractions should experience only modest warming and cooling of only a few degrees, which suggests that habitability cannot be precluded for such circumbinary planets. Planets with large land fractions that experience extreme periodic forcing could be prone to changes in temperature of tens of degrees or more, which could drive more extreme climate changes that inhibit continuously habitable conditions. 1 Introduction The recent discovery of giant planets in the habitable zone of close binary stars (e.g., Welsh and Orosz [2018] and references therein) has raised the possibility that such systems could host terrestrial planets within the liquid water habitable zone. The circumstellar liquid water "habitable zone" has received broad attention for single-star systems (e.g., Kasting et al. [1993]; Abe et al. [2011]; Kopparapu et al. [2013]; Kopparapu et al. [2014], but terrestrial planets in binary systems may also be able to retain stable atmospheres with habitable surface conditions. Previous studies have calculated the boundaries of the habitable zone in circumbinary systems (also known as P-type systems [Dvorak, 1984]) by combining the spectral energy distributions of the host stars to determine the orbital range that can maintain stable climates for an Earth-like planet [Kane and Hinkel, 2012; Haghighipour and Kaltenegger, 2013; Forgan, 2013; Cuntz, 2013, 2015; Wang and Cuntz, 2019a,b; Georgakarakos and Eggl, 2019]. Such studies demonstrate that circumbinary systems could host dynamically stable planets that maintain habitable conditions, so long as the planet has a sufficiently dense atmosphere and a method for recycling carbon between the atmosphere and interior (i.e., plate tectonics). Circumbinary planets have also been argued to have enhanced habitability prospects due to reduced stellar activity and XUV radiation [Mason et al., 2013]. A circumbinary planet experiences changes in instellation as the binary pair orbit one another. The orbital separation of the binary pair causes a periodic variation in incident radiation, which changes both the intensity and distribution of the incident spectral energy distribution [Forgan et al., 2015]. In some cases more extreme variations in incident energy can occur when one star eclipses another (as seen by the planet); however, while these eclipses are global in extent, they only have a duration of hours. The longer period variation in radiation from the binary pair is a time-dependent factor that could alter a planet's ability to sustain habitable conditions, even if it is otherwise situated Corresponding author: Jacob Haqq-Misra, [email protected] -- 1 -- Haqq-Misra et al., Journal of Geophysical Research - Planets, accepted on 13 November 2019 within the habitable zone. We refer to this periodic change in energy incident upon the planet as the circumbinary "gyration effect." Case studies of existing circumbinary systems suggest that this gyration effect may only exert a modest effect on global climate. (Such results corroborate other studies that have generally found modest deviations from the mean flux approximation when considering variations in a planet's orbital eccentricity [Dressing et al., 2010; Bolmont et al., 2016; Way and Georgakarakos, 2017].) May and Rauscher [2016] examine this periodic variation in instellation for Neptune-like circumbinary planets in the Kepler-47 system by using both an energy balance model and an idealized general circulation model (GCM) to calculate the maximum temperature changes expected. May and Rauscher [2016] find that Neptune-like circumbinary planets would experience no more than one percent variation in temperature from the gyration effect. Popp and Eggl [2017] use a GCM to calculate climate variations for hypothetical Earth-like planets in the Kepler-35 system and find that the global mean surface temperature changes by a few degrees at most. These results suggest that the gyration effect is unlikely to preclude habitability for planets within the circumbinary habitable zone. However, these previous efforts investigated planets with a uniform surface with a large heat capacity, which is appropriate for an ocean-covered planet or Neptune analog but may underestimate temperature variations on an Earth-like planet. In this study, we attempt to determine the extent to which the gyration effect is capable of driving significant changes in temperature for dynamically-stable circumbinary planets that orbit within the habitable zone. We address cases where the planets transit main-sequence stars, i.e., cases similar to the Kepler circumbinary planets. We use an analytic energy balance model to show that the effective heat capacity of a planet exerts the greatest control on the magnitude of the temperature variation that results from the circumbinary gyration effect. We then use a numerical energy balance model to show that the amplitude of time-dependent changes in temperature is greatest for planets with a large land fraction (and thus a lower effective heat capacity). Our results suggest that circumbinary planets with a lower ocean-to-land surface fraction that undergo strong periodic forcing are more likely to experience extreme climate change. 2 Maximum variation from periodic forcing As the distances between a circumbinary planet and its host stars continuously vary, the planet experiences a change in instellation throughout its orbit. The amplitude of the change depends on the orbits of the binary stars and the planets (specifically the luminosity of each star, semi-major axes, and eccentricities; throughout this paper we assume co-planar orbits). The timescale of the change depends on the orbital periods of the binary, the planet, and long-term precession of the planet's orbit. Of the known Kepler circumbinary planets, the shortest period binary is that of Kepler-47 with a period of 7.4 days, with planets b, c, and d on 49.5, 303, and 187 day orbits [Orosz et al., 2019]. The longest period host binary is that of Kepler-16, with an orbital period of 41 days [Doyle et al., 2011] and a planet on a ∼ 229 day orbit. Note that the longer the period of the binary, the lower the probability of transit, and so the more difficult it is to detect a circumbinary planet. Current observations thus do not provide much constraint on the maximum orbital period of the binary. However, the stability criteria of Holman and Wiegert [1999] requires that the binary must have a period shorter than approximately one third of the planet's orbital period in order for the planet to be dynamical stable (assuming a circular planetary orbit). In this study, we apply this stability criterion between the binary orbital period, Pbin, and the orbital period of the circumbinary planet, Pcbp, in order to constrain the extent of temperature variation on Earth-like circumbinary planets. The maximum variation in instellation spans a wide range across the observed Kepler cir- cumbinary systems. The most variable is Kepler-34b, which is not in the habitable zone (too hot). Kepler-34b is the most eccentric of known transiting circumbinary planets (Welsh et al. [2012]), and its eccentric orbit (ecbp = 0.182) is a significant factor driving changes in instellation. Kepler-35b, also not in the habitable zone, has a much lower eccentricity than Kepler-34b (ecbp = 0.042), which leads to a semiamplitude of about 15%. The circumbinary systems that are in the habitable zone (Kepler-16, 47, 453, 1647) have similar instellation semiamplitudes, less than 20% on a timescale -- 2 -- Haqq-Misra et al., Journal of Geophysical Research - Planets, accepted on 13 November 2019 of a few planetary orbits. However, on a longer timescale (∼ decades to centuries) planet orbit precession can play a significant role: Kepler-16b has the most circular orbit (ecbp = 0.0069) of the known circumbinary planets, but experiences a range in semiamplitude variation from about 17% to 28% due to evolution of its orbit. This suggests that a semiamplitude of 30% is certainly reasonable for a nearly circular orbit, and higher values as the eccentricity increases. The most eccentric Kepler circumbinary planet is the non-transiting planet KIC 7821010 b with e ∼ 0.35 [Priv. Comm.]. Similar to that of Kepler-34 b, its change in instellation is ∼90%. But unlike Kepler-34 b, KIC 7821010 b orbits within the habitable zone (most of the time) with an average instellation equal to ∼ 68% of the solar flux. Figure 1. The orbital period of the planet (top left), flux amplitude (top right), binary orbital period (bottom left), and insolation synodic period at the planet's orbit (bottom right) depend on the binary separation and the mass of the secondary star. Calculations assume a 1 M(cid:12) primary and a planet in a circular circumbinary orbit with a time mean instellation of 1360 W m−2. The dark curve indicates the maximum binary separation that permits a stable orbitng planet. The maximum flux amplitude of 55% occurs for a 0.5 M(cid:12) secondary star, which corresponds to a planetary orbital period of about 305 days, a binary orbital period of about 100 days, and an insolation synodic period of about 150 days on the planet. -- 3 -- Orbital Period (days)0.20.40.60.81.0Secondary star mass (M0)0.10.20.30.40.50.60.7Binary Separation (AU)300310325325350375400Primary Star = 1 M0Flux Amplitude (%)0.20.40.60.81.0Secondary star mass (M0)0.10.20.30.40.50.60.7Binary Separation (AU)5101020203030404050607080Binary Period (days)0.20.40.60.81.0Secondary star mass (M0)0.10.20.30.40.50.60.7Binary Separation (AU)255075100125150175200Insolation Synodic Period (days)0.20.40.60.81.0Secondary star mass (M0)0.10.20.30.40.50.60.7Binary Separation (AU)255075100125150200250300350400500UnstableStableUnstableStableUnstableStableUnstableStableHaqq-Misra et al., Journal of Geophysical Research - Planets, accepted on 13 November 2019 We first estimate the maximum amplitude of the gyration effect for a hypothetical circumbinary system with a planet that receives the same time-mean value of stellar insolation as Earth today. We keep the mass of the primary star fixed at 1 M(cid:12) and vary the mass of the secondary from 0.1 to 1 M(cid:12), with a binary separation ranging from 0.1 to 0.75 AU. We simplify our calculations by assuming a non-eccentric orbit and solve Kepler's laws using stellar mass-luminosity functions [Pecaut and Mamajek, 2013; Pecaut et al., 2012]. The top left panel of Fig. 1 shows the orbital period of a planet in such a circumbinary system (following Kepler's laws), where the incident stellar flux on the planet has been fixed at the present-day Earth value of 1360 W m−2. The black curve indicates the maximum binary separation that permits a planet to remain in a stable orbit at this incident flux, following the Pbin ≈ Pcbp/3 stability criteria [Holman and Wiegert, 1999]. The amplitude of flux variation from the circumbinary pair is shown in the top left panel of Fig. 1, which indicates a maximum flux amplitude of about 55% with a secondary mass of about 0.5 M(cid:12) and a binary separation of about 0.5 AU. We next determine the period of variation in instellation that corresponds to this hypothetical circumbinary system with a 1 M(cid:12) primary and 0.5 M(cid:12) secondary. The maximum flux amplitude occurs with a binary orbital period of about 100 days, as shown in the bottom left panel of Fig. 1. The planet in such a system has an orbital period of about 305 days, so the period of variation in flux is determined by the combined motion of the binary pair and circumbinary planet. The resulting insolation synodic period is shown in the bottom right panel of Fig. 1, which indicates a period of about 150 days. In our model calculations that follow, we will consider the response of a terrestrial planet in such a circumbinary system, with the period of the circumbinary gyration effect at 150 days and the maximum amplitude of flux variation at 50%. We specifically choose this extreme instellation variation scenario to examine the maximum response of the circumbinary planet's climate, and to answer the question, "Are climate variations due to the gyration effect negligible or significant?" But note that for developing our intuition, via the analytic energy balance calculations in the next section, we use circular orbits. Much more extreme instellation variations are possible for highly eccentric orbits. 3 Analytic energy balance model This section examines the amplitude of temperature variation due to the time-dependent cir- cumbinary gyration effect. The analytic energy balance calculations shown below demonstrate that the effect typically causes temperature to oscillate by only a few degrees for most cases, which is consistent with previous results [May and Rauscher, 2016; Popp and Eggl, 2017]. At the same time, planets with a much lower effective heat capacity may experience tens of degrees of temperature change. The rate of change of surface temperature, dT/dt, on a terrestrial planet depends upon the incoming stellar energy, S, the planetary Bond albedo, α, and the outgoing infrared radiative flux, FI R. We can express this relationship as C dT dt = S (1 − α) − FI R, (1) where C is effective heat capacity of the surface and atmosphere. A simple representation of FI R is the linear parameterization FI R = A + BT, where A and B are infrared flux constants (with T in Celsius). For this problem, we are interested in a periodic variation of the incoming stellar energy. This serves as an analogy to the changes in flux for a circumbinary planet when its host stars orbit one another. We choose a simple periodic function: S = S0 (1 + κ sin ωt) , (2) where S0 is a constant value of incident stellar flux, ω is the angular frequency of flux variation (with ω = 2π/Pbin), and κ is the amplitude of flux variation. This function is constructed to begin with a flux of S = S0 at t = 0 and recover a single-star solution when κ = 0. We acknowledge that this sinusoidal representation of stellar flux implies a circular planetary orbit (ecbp = 0), which would be dynamically unstable in an actual circumbinary system; however, it is possible for planet orbits -- 4 -- Haqq-Misra et al., Journal of Geophysical Research - Planets, accepted on 13 November 2019 to remain stable with small non-zero eccentricities. (One way to extend this analytic model would be to develop the forcing as a real-valued Fourier series [Popp and Eggl, 2017], with a solution that corresponds to the sum of several modes. The single-mode solution developed in this paper indicates the qualitative behavior expended from more realistic, multi-modal circumbinary orbits.) We will proceed by emphasizing that these calculations are intended to place theoretical limits on the impact of time-varying instellation on climate by using an idealized representation of the circumbinary flux variation. We can then write Eq. (1) as dT dt C = S0 (1 − α)[1 + κ sin(ωt)] − (A + BT) . (3) We can solve this ordinary differential equation in order to obtain an expression for T (t) that depends upon C, ω, and κ. We assume that α is constant, analogous to an ice-free and ocean-covered planet (i.e., no ice-albedo feedback). The analytic solution to Eq. (3) is given in the Appendix, which yields an expression for the steady-state solution: T (t) ≈ S0 (1 − α) κ cos(ωt) + T0. ωC (4) The amplitude of Eq. (4) represents the magnitude of temperature perturbations from circumbinary forcing. It is noteworthy that the infrared flux constants A and B are absent from Eq. (4), with the effective heat capacity as the primary planetary property that affects the amplitude of the gyration effect. Although the mean temperature T0 is determined by properties of the atmosphere's composi- tion (through the greenhouse effect, as described by A and B), the amplitude of circumbinary change in this analytic model is insensitive to changes in the greenhouse effect. In order to apply Eq. (4) to the scenario of a circumbinary planet in the habitable zone, we assume Earth-like conditions of T0 = 15 ◦C and C = 2.1 × 108 J m−2 ◦C−1. We are also interested in small perturbations in temperature that result from variations in instellation as the binary pair rotates, so we ignore any changes in albedo due to ice growth and assume a constant value of α. We obtain a value for albedo by setting Eq. (3) to a steady state (dT/dt = 0) with a fixed single star (ω = 0) and solving for α when T = T0, which gives α = 1−(A + BT0)/S0 ≈ 0.31. (This expression for albedo assumes Earth-like values for the thermal emitted flux parameters, A = 203 W m−2 and B = 2 W m−2 ◦C−1, which are based on northern hemisphere observations [North et al., 1981].) We use this value of α to evaluate Eq. (4). Given the assumptions that went into this calculation, it should not be surprising that this is in fact similar to the measured value of α ≈ 0.29 for Earth (e.g. Stephens et al. [2015]). Examples of solutions given by Eq. (4) are shown in Fig. 2, which includes the time variation in stellar flux (top panel) and steady-state temperature (bottom panel) for several cycles. These cases assume Pbin = 150 days (ω ≈ 0.04 rad day−1), with a single star control case (κ = 0) and two circumbinary cases with small amplitude (κ = 0.2) and extreme amplitude (κ = 0.5). The variation in mean planet temperature shows a lag relative to the variations in stellar flux. The lag timescale, tlag, is the difference in the time between the maxima of the stellar forcing function from Eq. (2) and temperature from Eq. (13), which we can express as 1 ω + arctan B ωC ≈ π 2ω = Pbin 4 , 2 tlag = (5) where we have made the simplifying assumption ωC (cid:29) B. The lag time for climate thus depends only upon the circumbinary orbital period, rather than properties of the planet's radiative transfer variable described by B. At larger circumbinary orbital periods, the lag timescale increases and enables a delayed response of climate to periodic changes in instellation. The Pbin = 150 day circumbinary period considered in Fig. 2 corresponds to a lag timescale of tlag = 38 days, which is long enough to induce a periodic temperature change of one to several degrees. (cid:18) π (cid:19) These calculations indicate that the gyration effect of periodic variation about the mean in incident stellar flux is capable of driving periodic change in planetary temperature. The angular frequency of flux variation, ω, (or equivalently, the period Pbin) determines the lag between flux -- 5 -- Haqq-Misra et al., Journal of Geophysical Research - Planets, accepted on 13 November 2019 Figure 2. Steady state variations with time using the analytic EBM of the relative stellar flux (top) and mean planet temperature (bottom) for a fictitious 1 M(cid:12) primary and 0.5 M(cid:12) secondary case with Pbin = 150 days. Colored curves show a single star control case with κ = 0 (dashed black) and circumbinary cases with κ = 0.2 (blue) and κ = 0.5 (green). variation and temperature response, which can substantially alter the magnitude of variation in T. In addition, the two parameters that influence temperature in this model are the amplitude of circumbinary forcing, κ, and the effective heat capacity, C. Physically, a larger value of κ corresponds to a larger flux variation from the binary pair. A larger value of C corresponds to a planet with greater average heat capacity that is less sensitive to time-varying changes in energy (such as a planet with a large fraction of ocean coverage). The maximum temperature variation that can reasonably be attained through this model is given by the amplitude of the cosine term in Eq. (4). This maximum amplitude is shown in Fig. 3 as a "heat map" plot over the parameter space of C and κ, with Pbin = 150 days, The largest changes in temperature occur for planets with low effective heat capacity and a large amplitude of periodic forcing. Most of the parameter space shows only a few degrees total warming (and, analogously, the same amount of cooling), which corroborate previous studies [May and Rauscher, 2016; Popp and Eggl, 2017]. This suggests that the gyration effect is unlikely to substantially undermine the habitability of an Earth-like planet under the most typical conditions. But the upper-left quadrant of Fig. 3 also indicates that tens of degrees of variation could be possible for atmospheres with low effective heat capacity that experience large variations in stellar forcing. Under extreme conditions, circumbinary planets are capable of experiencing dramatic shifts in temperature. This simple analytic model shows that a planet's susceptibility to the circumbinary gyration effect is primarily controlled by its effective heat capacity. In the case of terrestrial planets, effective heat capacity is lowest for a dry land planet and increases for planets with greater ocean coverage and depth. The coefficients of the outgoing infrared radiative flux, A and B, can be altered to describe atmospheres of various compositions, based upon the net greenhouse warming, but these parameters do not affect the steady-state gyration effect described by Eq. (4). The gyration effect may therefore affect planetary habitability under certain scenarios. Circumbinary planets similar to Earth in terms of effective heat capacity (i.e., a similar ocean to land distribution), with low to moderate variation in instellation, may not experience significant variations in surface temperature. The gyration effect -- 6 -- Haqq-Misra et al., Journal of Geophysical Research - Planets, accepted on 13 November 2019 Figure 3. Maximum temperature change using the analytic EBM after reaching a steady state, shown over a parameter space of effective heat capacity and the amplitude of periodic forcing. The white circle indicates the heat capacity and constant forcing of present-day Earth. cannot preclude habitability on such planets. Conversely, terrestrial planets with lower effective heat capacity (i.e., a lower ocean to land ratio) under larger instellation variation may find their global climates dominated by the changing flux from the binary. These extreme circumbinary systems may find themselves driven into climate regimes that diminish habitable conditions. 4 Numerical energy balance model The analytic EBM used in the previous section demonstrates the significance of effective heat capacity as the primary planetary property that determines the response to circumbinary forcing. The analytic model assumes a fixed value of albedo, which thereby negates any ice-albedo feedback that could occur on such a planet; however planets that experience a circumbinary gyration effect with magnitude of a few degrees or greater could experience periodic changes in ice coverage that alter albedo and thus affect the total energy balance. The analytic model also constrains the planet to a single point, which neglects the contribution of meridional energy transport from equator to poles. We therefore continue our examination of the circumbinary gyration effect using a numerical EBM that accounts for ice-albedo feedback, meridional energy transport, and the effect of topography on effective heat capacity. The numerical EBM [Haqq-Misra, 2014] is modified to include the circumbinary forcing term from Eq. (2). The energy balance equation for this model can then be written in terms of latitude, θ, as: C ∂T ∂t = S0 (1 − α)[1 + κ sin(ωt)] − (A + BT) + D cos θ . (6) (cid:18) 1 cos θ ∂ ∂θ (cid:19) ∂T ∂θ The last term in Eq. (6) accounts for the efficiency of meridional energy transport by using D as a diffusive parameter, with D = 0.38 W m−2 K−1 in this study as a fixed value applicable to present-day Earth conditions. Albedo is defined as a function of temperature so that α = 0.3 for unfrozen land or ocean (T ≥ 263.15 K) and α = 0.663 for any frozen surface (T ≤ 263.15 K). (Note that this threshold is ten degrees below freezing to indicate the formation of permanent surface ice.) The thermal emitted flux parameters are set to A = 203 W m−2 and B = 2 W m−2 ◦C−1. The EBM decomposes the planet into 18 equally-spaced latitudinal zones, assumes a starting surface temperature profile, and then numerically solves Eq. (6) by using a time step of ∆t = 8.64 × 103 s = 1 day. This model assumes a circular orbit for purposes of comparing directly with the analytic solution, while noting that planets -- 7 -- Haqq-Misra et al., Journal of Geophysical Research - Planets, accepted on 13 November 2019 in circumbinary systems will necessarily have non-zero eccentricity. We initially keep planetary obliquity fixed at zero degrees in order to eliminate any seasonal effects, although we later consider the effect of seasons (following Gaidos and Williams [2004]) on circumbinary planets with Earth-like axial tilt. Effective heat capacity, C, is defined in the numerical EBM as a function of both latitude and temperature, which represents the contributions from unfrozen land, unfrozen ocean, and ice-covered surface. Letting fo and fi represent the respective fraction of ocean and ice at each latitudinal zone, we can write the zonally averaged heat capacity as C = (1 − fi) Cl + fo [(1 − fi) Co + fiCi] , (7) where Cl = 5.25 × 106 J m−2 K−1 is the heat capacity of continental land, Co = 40 Cl = 2.1 × 108 J m−2 K−1 is the heat capacity over a wind-mixed 50 m ocean layer, and Ci = 2 Cl = 1.05 × 107 J m−2 K−1 is the heat capacity over ice [Fairén et al., 2012]. Eq. (7) enables the numerical EBM to vary effective heat capacity according to surface conditions on the planet as a function of latitude. Figure 4. Steady state variations of mean planet temperature with time using the numerical EBM for a fictitious 1 M(cid:12) primary and 0.5 M(cid:12) secondary case with Pbin = 150 days. Colored curves show a single star control case with κ = 0 (dashed black) and circumbinary cases with κ = 0.2 (blue) and κ = 0.5 (green), all with zero obliquity. Solid curves indicate global aquaplanet conditions with no topography, while dashed curves indicate an equatorial supercontinent as topography. 4.1 Periodic temperature variations Steady-state solutions of Eq. (6) are shown as mean temperature in Fig. 4 for the same Pbin = 150 day circumbinary system considered with the analytical model. The single star control case is shown as a black line on Fig. 4, along with circumbinary cases with moderate forcing (κ = 0.2, blue curves) and strong forcing (κ = 0.5, green curves). Solid curves correspond to global aquaplanet conditions with no land ( fo = 1.0), while dashed curves include topography as an equatorial supercontinent with ocean at the poles ( fo = 0.7). The aquaplanet cases plotted in Fig. 4 are comparable with the analytic solutions shown in Fig. 2, which show a temperature oscillation of about a degree for κ = 0.2 and about three degrees for κ = 0.5. The amplitudes of these aquaplanet solutions with the numerical EBM are slightly greater than the analytic EBM as the result of enhanced climate sensitivity from the inclusion of ice-albedo feedback and meridional energy transport. The equatorial supercontinent case shows a much larger amplitude of variation of about four degrees for κ = 0.2 and nearly ten degrees for κ = 0.5, due to the reduction in effective heat capacity along the planet's equatorial belt. -- 8 -- Haqq-Misra et al., Journal of Geophysical Research - Planets, accepted on 13 November 2019 Figure 5. Steady state variations of the latitudinal distribution of planet temperature with time using the numerical EBM for a fictitious 1 M(cid:12) primary and 0.5 M(cid:12) secondary case with Pbin = 150 days. All calculations assume strong circumbinary forcing (κ = 0.5) and zero obliquity. Periodicity in temperature is evident for global aquaplanet conditions with no topography (top), Earth-like topography (middle), and an equatorial supercontinent as topography (bottom). The choice of topography affects the amplitude and timing of temperature changes, as shown in Fig. 5 for the circumbinary case with strong forcing (κ = 0.5) and zero obliquity. The top row shows the latitudinal distribution of temperature with time for an aquaplanet, comparable to the solid green line on Fig. 4. The bottom row likewise shows latitudinal temperature with time for a planet with equatorial supercontinent topography, comparable to the dashed green line on Fig. 4. The middle row includes Earth topography, which assumes fo = 0.7 but distributes the land across the northern and southern hemispheres. All cases show warm conditions along the equator and frozen conditions at latitudes greater than about 50 degrees. The aquaplanet and Earth topography cases both include a substantial fraction of ocean along the equator, which gives a higher heat capacity that allows equatorial latitudes to remaining above freezing over a complete circumbinary period. The equatorial supercontinent case has complete land cover and thus a much lower heat capacity along the equator, which also remains above freezing at lower latitudes but shows a much greater amplitude of oscillation in temperature along the equator. -- 9 -- Haqq-Misra et al., Journal of Geophysical Research - Planets, accepted on 13 November 2019 Figure 6. Steady state variations of the latitudinal distribution of planet temperature change with time using the numerical EBM for a fictitious 1 M(cid:12) primary and 0.5 M(cid:12) secondary case with Pbin = 150 days. All calculations assume strong circumbinary forcing (κ = 0.5) and zero obliquity. Periodic changes in temperature are evident for global aquaplanet conditions with no topography (top), Earth-like topography (middle), and an equatorial supercontinent as topography (bottom). The temperature difference is calculated by taking the three cases from Fig. 5 and subtracting the single-star temperature solutions from the EBM with the same respective topography. The transient warming and cooling induced by the circumbinary effect is illustrated in Fig. 6. This figure shows the temperature change obtained by taking the three cases from Fig. 4 and subtracting the corresponding single-star temperature solutions of the numerical EBM. The difference shown in Fig. 6 indicates symmetric temperature variation of a few degrees for the aquaplanet (top row), with the greatest variation occurring at polar latitude where ice caps form, and asymmetric temperature variation for Earth topography (middle row) that reflects the larger concentration of land area in the northern hemisphere. The equatorial supercontinent topography (bottom row) shows changes of more than ten degrees along the equator, with relatively little variation in the northern and southern oceans. This is because the efficiency of the diffusive energy transport is enhanced over land compared to ocean (from Eqs. (6) and (7)), which results in an increase in the amplitude of temperature variation on areas with large land fractions and a a decrease in variation over large -- 10 -- Haqq-Misra et al., Journal of Geophysical Research - Planets, accepted on 13 November 2019 oceans. The lack of a seasonal cycle also causes higher latitudes to experience a lower amplitude of stellar forcing, which contributes to a smaller variation in temperature. Figure 7. Steady state variations of the latitudinal distribution of planet temperature change with time using the numerical EBM for a fictitious 1 M(cid:12) primary and 0.5 M(cid:12) secondary case with Pbin = 150 days. All calculations assume strong circumbinary forcing (κ = 0.5) and 23.5◦ obliquity. Periodic changes in temperature are evident for global aquaplanet conditions with no topography (top), Earth-like topography (middle), and an equatorial supercontinent as topography (bottom). The temperature difference is calculated by subtracting the single-star temperature solutions from the circumbinary solutions from the EBM with the same respective topography. This analysis has so far been restricted to a planet with zero obliquity, which keeps the maximum extent of incident stellar radiation focused along the equator. By contrast, a non-zero obliquity will result in a cycle that shifts the maximum in circumbinary variation from the equator toward the poles with the seasons [May and Rauscher, 2016]. We therefore consider a set of calculations similar to those shown in Fig. 6 but with Earth-like obliquity. The temperature change obtained for a circumbinary planet with 23.5◦ obliquity is shown in Fig. 7 for aquaplanet, Earth-like, and equatorial supercontinent topographies and strong circumbinary forcing (κ = 0.5). The aquaplanet case in Fig. 7 (top row) shows greater temperature change near the poles compared to the corresponding case in Fig. 6, due to the seasonal extremes experienced at the planet's poles. Likewise, the Earth topography case (middle row) shows amplified temperature change near the poles compared to the -- 11 -- Haqq-Misra et al., Journal of Geophysical Research - Planets, accepted on 13 November 2019 zero obliquity case, both of which are more extreme than the aquaplanet case due to the presence of continents in both hemispheres. The equatorial supercontinent case in Fig. 7 (bottom row) shows the most similarity with the corresponding case in Fig. 6, although the 23.5◦ obliquity case still shows seasonal variation of temperature near the poles. These results emphasize the strong control exerted by the ocean-to-land fraction on the temperature variation that results on an Earth-like planet from the circumbinary gyration effect. The analytic EBM indicates that the amplitude of temperature variation from the circumbinary gyration effect is sensitive to the effective heat capacity (Fig. 3), but the numerical EBM illustrates that the latitudinal profile of heat capacity also exerts a strong control on the maximum amplitude of warming (Figs. 6 and 7). This occurs not only because of the reduced heat capacity of land compared to ocean but also because diffusive meridional energy transport is enhanced over land. This feature may also be relevant for planetary habitability, as large land areas (and thus any land-based life) on circumbinary planets are be more likely to experience temperature variations from the gyration effect than large oceans. 4.2 Climate bistability The bistability of Earth-like climates is a well-known feature of EBMs [North et al., 1981; Caldeira and Kasting, 1992] that also appears in many GCMs [DeConto and Pollard, 2003; Ishiwatari et al., 2007; Voigt and Marotzke, 2010; Wolf et al., 2017], with both warm and ice-covered solutions available at a given value of stellar flux. The left panels of Fig. 8 illustrate this classic hysteresis loop for 0◦ obliquity (top) and 23.5◦ obliquity (bottom), with the blue curve representing equilibrium EBM solutions initialized from ice-covered conditions and the red curve indicating solutions initialized from ice-cap conditions. The dashed lines show transitions from an ice-covered to ice-free state (dashed red) and from an ice-cap to ice-covered state (dashed blue). In general Fig. 8 describes a planet's climate state in terms of the relative warming and the previous climate state. This hysteresis loop indicates that large ice caps on Earth can remain stable until approximately 30 degrees, after which the planet falls into global glaciation. This transition is not immediately reversible but instead requires a significant increase in relative warming (either a change in stellar flux or, equivalently in this case, greenhouse gas forcing) before deglaciation can occur. On Earth, deglaciation from a global snowball condition seems to have occurred at least once, during the Neoproterozoic Era [Hoffman et al., 1998]. The 0◦ obliquity case in Fig. 8 also shows additional ice-free climate states in pink; this is known as the "small ice cap" solution, which describes the threshold at which a warming climate causes a small polar ice cap to become unstable and vanish. This small ice cap instability has been observed in other EBMs [North, 1984; Huang and Bowman, 1992] as well as some GCMs [Lee and North, 1995; Winton, 2006]; however, the small ice cap is not as ubiquitous in climate models as the more prominent large ice cap instability. For the 0◦ obliquity case, the small ice cap instability occurs at approximately 80 degrees, after which the planet transitions to an ice-free state. This transition is also not immediately reversible; if relative warming were to decrease on such a planet, then the climate would gradually cool and eventually fall into a large ice cap state. Circumbinary planets show reduced bistability compared to their single star counterparts, as shown in the middle and right columns of Fig. 8. The cases with moderate forcing (κ = 0.2) show a reduction in the width of the hysteresis loop, which indicates a lower threshold for a frozen planet to deglaciate. This narrowing is even more pronounced in the strong forcing cases (κ = 0.5), which features an abrupt transition from a glacial state to a warm, but nearly ice-cap, state. The temperature variation from the circumbinary gyration effect causes this narrowing of the hysteresis loop by reducing both the glaciation and deglaciation thresholds as the climate periodically exceeds mean temperature at all latitudes. The 0◦ obliquity cases show a wider hysteresis loop than the 23.5◦ obliquity cases, as the latter includes seasonal forcing that cause additional melting at the summer pole [Gaidos and Williams, 2004; May and Rauscher, 2016]. The increase in obliquity also causes the small ice cap instability to vanish in this model, even though the moderate forcing (κ = 0.2) circumbinary case retains a small ice cap solution at zero obliquity. -- 12 -- Haqq-Misra et al., Journal of Geophysical Research - Planets, accepted on 13 November 2019 Figure 8. Hysteresis plots using the numerical EBM showing climate bistability for a planet with Earth-like topography orbiting a single star (κ = 0, left), a binary pair with moderate forcing (κ = 0.2, middle), and a binary pair with strong forcing (κ = 0.5, right). The top row shows cases with 0◦ obliquity and the bottom row shows cases with 23.5◦ obliquity. Climate states indicate solutions of the EBM when initialized with ice-covered (blue), ice-cap (red), and ice-free (pink) conditions. Dashed lines show discontinuous transitions between climate states. The width of the hysteresis loop between ice-covered and ice-free states narrows as the circumbinary gyration effect increases or as obliquity increases. The zero obliquity cases may be less realistic for actual circumbinary planets, as the the orbital properties of the host stars and as the presence of other planets in the system can place constraints on a planet's obliquity. In particular, spin-orbit coupling in such a system can cause a circumbinary planet to enter a Cassini State that causes large variations in obliquity [Kostov et al., 2014; Forgan et al., 2015]. Increasing the obliquity beyond the present-day Earth value considered in this study would further increase the magnitude periodic temperature variation (Fig. 7) and reduce the width of the hysteresis loop (Fig. 8). Earth-like circumbinary planets within the liquid water habitable zone may may therefore be less likely to be globally glaciated and more likely to exist in an ice-free or ice-cap state. 5 Discussion and Conclusions We find that the habitability of circumbinary planets depends upon the ocean-to-land fraction of the surface, as well as the amplitude of forcing from the gyration effect. Terrestrial planets with larger land fractions than Earth, or a greater surface area of land at equatorial latitudes, should experience much greater shifts in temperature than planets with large ocean fractions. The lower effective heat capacity of land as well as the enhanced efficiency of meridional energy transport make such areas susceptible to greater changes in temperature from variations in instellation. This behavior occurs because the atmospheres of planets we have considered are largely transparent to incident instellation (which is mostly at short wavelengths), with atmospheric greenhouse gases absorbing primarily at longer infrared wavelengths. Such an assumption is generally valid for a wide range of plausible terrestrial planets, even with atmospheres several times more dense than present-day Earth. Circumbinary planets are expected to show reduced bistability, due to both the gyration effect as well as high obliquity (Fig. 8), which may suggest that such planets may more easily deglaciate from a frozen snowball state. -- 13 -- Haqq-Misra et al., Journal of Geophysical Research - Planets, accepted on 13 November 2019 Our model calculations support the results of Popp and Eggl [2017], showing small temperature variations from the gyration effect, when we examine a planet with global ocean coverage (Figs. 6 and 7, top row). But when we include Earth-like land coverage or an equatorial supercontinent (Figs. 6 and 7, middle and bottom rows, respectively), we find that the lower effective heat capacity increases the extent of surface temperature variation. Such planets that experience a low to moderate variation in the circumbinary flux are likely to experience changes on a timescale analogous to a seasonal cycle, which may not necessarily preclude habitability. Any Earth-like circumbinary planets that are eventually discovered may therefore be attractive candidates for spectroscopic characterization in the quest to discover potential biosignatures with the next generation of space telescopes [Kiang et al., 2018]. Circumbinary planets that lack large oceans are prone to experience a greater amplitude of temperature variation. The equatorial supercontinent cases shown in Figs. 5, 6 and 7 exhibit periodic variation on a scale analogous to seasonal cycles but do not freeze along the equator, so habitability cannot be precluded on such planets. However, the equatorial supercontinent case is not a pure land planet and includes ocean at the poles, with fo = 0.7. The numerical EBM became numerically unstable at lower values of lower values of fo, which indicates the high sensitivity of this model to the planet's effective heat capacity. Planets with much larger land fractions could experience even greater variation in temperature that could lead to freezing conditions along the equator. While such changes could pose challenges for biology, the freeze-thaw cycle on such planets could instead provide a selection pressure that helps to drive biological evolution. The magnitude of temperature change on such planets will help predict whether such behavior is a detriment to habitability and will require more detailed calculations using a GCM tailored to the particular system. M-dwarf stars emit a substantial fraction of radiation at infrared wavelengths, so a planet orbiting a circumbinary pair of M-dwarf stars would be able to absorb a greater fraction of incident instellation with its atmosphere. A planet with a sufficiently dense atmosphere around such a pair of low-mass stars might even absorb enough incoming radiation that periodic variations on the surface are damped and overall less sensitive to the ocean fraction. This would also drive complex behavior, such as changes in the substellar cloud deck, that cause further feedbacks on climate. Testing this scenario will require calculations with a GCM that include stellar spectra from a low-mass circumbinary pair. This study has explored the extent of periodic temperature variation on Earth-like circumbinary planets; however, diurnal variations in climate cannot be captured with either the analytic or numerical EBM. Similarly, the representation of topography in the numerical EBM is based only on the ocean- to-land fraction at each latitude band, with no regard to the distribution of land with longitude. Previous GCM studies of circumbinary planets have assumed uniform surface conditions, and focus on spatially averaged quantities, but localized warming effects over the diurnal cycle and across continents could limit habitability in certain regions on the surface; for example, the surface habitability limits suggested by Sherwood and Huber [2010] are based upon the ability for humans and other mammals to efficiently dissipate metabolic heat. Further work with GCMs will therefore provide quantitative constraints on the expected climates of terrestrial circumbinary planets that experience extreme forcing from their binary star host. Appendix: Analytic Energy Balance Model This appendix provides the analytic solution of the energy balance model equation with sinusoidal forcing in Eq. (3). We begin by writing the homogeneous part of Eq. (3), dT dt + B C T = 0. (8) The solution to Eq. (8) gives Tg (t) = X exp(−Bt/C), where X is a constant. This is the general solution to Eq. (3). We next find the particular solution of Eq. (3) by assuming a solution in the form Tp (t) = Y sin(ωt) + Z cos(ωt) + W, where Y, Z, and W are constants. Substituting this expression -- 14 -- Haqq-Misra et al., Journal of Geophysical Research - Planets, accepted on 13 November 2019 for Tp (t) into Eq. (3) gives [BY − ωCZ] sin(ωt) + [BZ + ωCY] cos(ωt) = S0 (1 − α)[1 + κ sin(ωt)] − A − BW . (9) Setting the constant terms of Eq. (9) equal to each other gives the value of W, as W = S0 (1 − α) − A B . (10) The constants Y and Z can be found by equating the coefficients of the trigonometric functions on both sides of Eq. (9), which gives the values (11) Y = S0 (1 − α) κB ω2C2 + B2 , −S0 (1 − α) κωC (cid:18)−Bt (cid:19) C and The full solution to Eq. Tg (t) + Tp (t). We can now write this solution as Z = (12) (3) is given as the sum of the general and particular solutions, T (t) = ω2C2 + B2 . T (t) = X exp + Y sin(ωt) + Z cos(ωt) + W . (13) The values of Y, Z, and W are known. To find the value of X, we assume T (t) = T0 at t = 0. This boundary condition reduces Eq. (13) to T0 = X + Z + W, which gives A B . ω2C2 + B2 − 1 X = T0 + S0 (1 − α) (14) κωC (cid:20) + B (cid:21) This solution can be verified by substituting Eq. (13) into Eq. (3) using the respective values of W, X, Y, and Z from Eqs. (10), (14), (11), and (12). We can simplify Eq. (13) by examining the steady state solution as t becomes large and the general solution to Eq. (3) vanishes, Tg(t) ≈ 0. The particular solution to Eq. (3), Tp(t), can be reduced by noting that ωC (cid:29) B for the range of paramters considered in this study. By first rewriting √ the particular solution as Tp = steady state solution to Eq. (3): (cid:1) + W, we arrive at an expression for the Y 2 + Z2 cos(cid:0)ωt − arctan Z T (t) ≈ S0 (1 − α) κ Y ωC cos(ωt) + T0. (15) Acknowledgments The data used for this paper can be accessed at https://doi.org/10.6084/m9.figshare.10279790. The authors gratefully acknowledge funding from the NASA Habitable Worlds program under award 80NSSC17K0741. J.H., E.T.W., and R.K.K. also acknowledge funding from the Virtual Planetary Laboratory under NASA award 80NSSC18K0829, and R.K.K. and V.K. acknowledge funding from the Sellers Exoplanet Environments Collaboration. W.F.W. gratefully thanks John Hood Jr. for his support of exoplanet research at SDSU. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of NASA. References Abe, Y., A. Abe-Ouchi, N. H. Sleep, and K. J. Zahnle (2011), Habitable zone limits for dry planets, Astrobiology, 11(5), 443 -- 460. Bolmont, E., A.-S. Libert, J. Leconte, and F. Selsis (2016), Habitability of planets on eccentric orbits: Limits of the mean flux approximation, Astronomy & Astrophysics, 591, A106. Caldeira, K., and J. F. Kasting (1992), Susceptibility of the early earth to irreversible glaciation caused by carbon dioxide clouds, Nature, 359(6392), 226. -- 15 -- Haqq-Misra et al., Journal of Geophysical Research - Planets, accepted on 13 November 2019 Cuntz, M. (2013), S-type and p-type habitability in stellar binary systems: A comprehensive ap- proach. i. method and applications, The Astrophysical Journal, 780(1), 14. Cuntz, M. (2015), S-type and p-type habitability in stellar binary systems: a comprehensive approach. ii. elliptical orbits, The Astrophysical Journal, 798(2), 101. DeConto, R. M., and D. Pollard (2003), Rapid cenozoic glaciation of antarctica induced by declining Doyle, L. R., J. A. Carter, D. C. Fabrycky, R. W. Slawson, et al. (2011), Kepler-16: A Transiting atmospheric co 2, Nature, 421(6920), 245. Circumbinary Planet, Science, 333, 1602. Dressing, C. D., D. S. Spiegel, C. A. Scharf, K. Menou, and S. N. Raymond (2010), Habitable climates: the influence of eccentricity, The Astrophysical Journal, 721(2), 1295. Dvorak, R. (1984), Numerical experiments on planetary orbits in double stars, in The Stability of Planetary Systems, pp. 369 -- 378, Springer. Fairén, A., J. Haqq-Misra, and C. McKay (2012), Reduced albedo on early mars does not solve the climate paradox under a faint young sun, Astronomy & Astrophysics, 540, A13. Forgan, D. (2013), Assessing circumbinary habitable zones using latitudinal energy balance mod- elling, Monthly Notices of the Royal Astronomical Society, 437(2), 1352 -- 1361. Forgan, D. H., A. Mead, C. S. Cockell, and J. A. Raven (2015), Surface flux patterns on planets in circumbinary systems and potential for photosynthesis, International Journal of Astrobiology, 14(3), 465 -- 478. Gaidos, E., and D. Williams (2004), Seasonality on terrestrial extrasolar planets: inferring obliquity and surface conditions from infrared light curves, New Astronomy, 10(1), 67 -- 77. Georgakarakos, N., and S. Eggl (2019), On the enlargement of habitable zones around binary stars in hostile environments, Monthly Notices of the Royal Astronomical Society: Letters, 487(1), L58 -- L60. Haghighipour, N., and L. Kaltenegger (2013), Calculating the habitable zone of binary star systems. ii. p-type binaries, The Astrophysical Journal, 777(2), 166. Haqq-Misra, J. (2014), Damping of glacial-interglacial cycles from anthropogenic forcing, Journal of Advances in Modeling Earth Systems, 6(3), 950 -- 955. Hoffman, P. F., A. J. Kaufman, G. P. Halverson, and D. P. Schrag (1998), A neoproterozoic snowball Holman, M. J., and P. A. Wiegert (1999), Long-term stability of planets in binary systems, The Huang, J., and K. P. Bowman (1992), The small ice cap instability in seasonal energy balance models, earth, science, 281(5381), 1342 -- 1346. Astronomical Journal, 117(1), 621. Climate dynamics, 7(4), 205 -- 215. Ishiwatari, M., K. Nakajima, S.-i. Takehiro, and Y.-Y. Hayashi (2007), Dependence of climate states of gray atmosphere on solar constant: From the runaway greenhouse to the snowball states, Journal of Geophysical Research: Atmospheres, 112(D13). Kane, S. R., and N. R. Hinkel (2012), On the habitable zones of circumbinary planetary systems, Kasting, J. F., D. P. Whitmire, and R. T. Reynolds (1993), Habitable zones around main sequence The Astrophysical Journal, 762(1), 7. stars, Icarus, 101(1), 108 -- 128. Kiang, N. Y., S. Domagal-Goldman, M. N. Parenteau, D. C. Catling, Y. Fujii, V. S. Meadows, E. W. Schwieterman, and S. I. Walker (2018), Exoplanet biosignatures: At the dawn of a new era of planetary observations, Astrobiology. Kopparapu, R. K., R. Ramirez, J. F. Kasting, V. Eymet, T. D. Robinson, S. Mahadevan, R. C. Terrien, S. Domagal-Goldman, V. Meadows, and R. Deshpande (2013), Habitable zones around main-sequence stars: new estimates, The Astrophysical Journal, 765(2), 131. Kopparapu, R. K., R. M. Ramirez, J. SchottelKotte, J. F. Kasting, S. Domagal-Goldman, and V. Eymet (2014), Habitable Zones around Main-sequence Stars: Dependence on Planetary Mass, ApJL, 787, L29, doi:10.1088/2041-8205/787/2/L29. Kostov, V. B., P. R. McCullough, J. A. Carter, M. Deleuil, R. F. Díaz, D. C. Fabrycky, G. Hebrard, T. C. Hinse, T. Mazeh, J. A. Orosz, et al. (2014), Kepler-413b: a slightly misaligned, neptune-size transiting circumbinary planet, The Astrophysical Journal, 784(1), 14. -- 16 -- Haqq-Misra et al., Journal of Geophysical Research - Planets, accepted on 13 November 2019 Lee, W.-H., and G. North (1995), Small ice cap instability in the presence of fluctuations, Climate dynamics, 11(4), 242 -- 246. Mason, P. A., J. I. Zuluaga, J. M. Clark, and P. A. Cuartas-Restrepo (2013), Rotational Synchro- nization May Enhance Habitability for Circumbinary Planets: Kepler Binary Case Studies, ApJL, 774, L26, doi:10.1088/2041-8205/774/2/L26. May, E., and E. Rauscher (2016), Examining tatooine: Atmospheric models of neptune-like cir- cumbinary planets, The Astrophysical Journal, 826(2), 225. North, G. R. (1984), The small ice cap instability in diffusive climate models, Journal of the atmospheric sciences, 41(23), 3390 -- 3395. North, G. R., R. F. Cahalan, and J. A. Coakley Jr (1981), Energy balance climate models, Reviews of Geophysics, 19(1), 91 -- 121. Orosz, J. A., W. F. Welsh, N. Haghighipour, B. Quarles, D. R. Short, S. M. Mills, S. Satyal, G. Torres, E. Agol, D. C. Fabrycky, et al. (2019), Discovery of a third transiting planet in the kepler-47 circumbinary system, The Astronomical Journal, 157(5), 174. Pecaut, M. J., and E. E. Mamajek (2013), Intrinsic colors, temperatures, and bolometric corrections of pre-main-sequence stars, The Astrophysical Journal Supplement Series, 208(1), 9. Pecaut, M. J., E. E. Mamajek, and E. J. Bubar (2012), A revised age for upper scorpius and the star formation history among the f-type members of the scorpius-centaurus ob association, The Astrophysical Journal, 746(2), 154. Popp, M., and S. Eggl (2017), Climate variations on earth-like circumbinary planets, Nature com- munications, 8, 14,957. Sherwood, S. C., and M. Huber (2010), An adaptability limit to climate change due to heat stress, Proceedings of the National Academy of Sciences, 107(21), 9552 -- 9555. Stephens, G. L., D. O'Brien, P. J. Webster, P. Pilewski, S. Kato, and J.-l. Li (2015), The albedo of earth, Reviews of Geophysics, 53(1), 141 -- 163, doi:10.1002/2014RG000449. Voigt, A., and J. Marotzke (2010), The transition from the present-day climate to a modern snowball earth, Climate dynamics, 35(5), 887 -- 905. Wang, Z., and M. Cuntz (2019a), S-type and p-type habitability in stellar binary systems: A comprehensive approach. iii. results for mars, earth, and super-earth planets, The Astrophysical Journal, 873(2), 113. Wang, Z., and M. Cuntz (2019b), S-type and p-type habitable zones of stellar binary systems: Effect of the planetary mass, Research Notes of the AAS, 3(5), 70. Way, M. J., and N. Georgakarakos (2017), Effects of variable eccentricity on the climate of an earth-like world, The Astrophysical Journal Letters, 835(1), L1. Welsh, W. F., and J. A. Orosz (2018), Two suns in the sky: The Kepler circumbinary planets, Handbook of Exoplanets, Edited by Hans J. Deeg and Juan Antonio Belmonte. Springer Living Reference Work, ISBN: 978-3-319-30648-3, 2017, id. 34. Welsh, W. F., J. A. Orosz, J. A. Carter, D. C. Fabrycky, et al. (2012), Transiting circumbinary planets Kepler-34 b and Kepler-35 b, Nature, 481, 475 -- 479, doi:10.1038/nature10768. Winton, M. (2006), Does the arctic sea ice have a tipping point?, Geophysical Research Letters, 33(23). Wolf, E. T., A. L. Shields, R. K. Kopparapu, J. Haqq-Misra, and O. B. Toon (2017), Constraints on climate and habitability for earth-like exoplanets determined from a general circulation model, The Astrophysical Journal, 837(2), 107. -- 17 --
1710.10110	1	1710	2017-10-27T12:55:51	Chaotic dynamics around cometary nuclei	[ "astro-ph.EP", "nlin.CD" ]	We apply a generalized Kepler map theory to describe the qualitative chaotic dynamics around cometary nuclei, based on accessible observational data for five comets whose nuclei are well-documented to resemble dumb-bells. The sizes of chaotic zones around the nuclei and the Lyapunov times of the motion inside these zones are estimated. In the case of Comet 1P/Halley, the circumnuclear chaotic zone seems to engulf an essential part of the Hill sphere, at least for orbits of moderate to high eccentricity.	astro-ph.EP	astro-ph	Chaotic dynamics around cometary nuclei Jos´e Lagesa, Ivan I. Shevchenkoa,b,c, Guillaume Rollina aInstitut UTINAM, Observatoire des Sciences de l'Univers THETA, CNRS, Universit´e de Bourgogne-Franche-Comt´e, Besan¸con 25030, France bPulkovo Observatory, RAS, 196140 Saint Petersburg, Russia cLebedev Physical Institute, RAS, 119991 Moscow, Russia Abstract We apply a generalized Kepler map theory to describe the qualitative chaotic dynamics around cometary nuclei, based on accessible observational data for five comets whose nuclei are well-documented to resemble dumb-bells. The sizes of chaotic zones around the nuclei and the Lyapunov times of the motion inside these zones are estimated. In the case of Comet 1P/Halley, the circumnuclear chaotic zone seems to engulf an essential part of the Hill sphere, at least for orbits of moderate to high eccentricity. Keywords: Comets, nucleus; Comets, dynamics; Comet Halley; Celestial mechanics; Resonances, orbital 1. Introduction Asteroids and cometary nuclei quite often have bilobed (dumb-bell) shapes; a spectacular example of such a shape is offered by the recent radar imaging of the near-Earth asteroid 2014 JO251. The orbital dynamics around bodies with complex gravity fields (Chauvineau et al., 1993; Scheeres et al., 1996; Scheeres et al., 1998; Petit et al., 1997; Scheeres, 2002; Bartczak and Breiter, 2003; Mysen et al., 2006; Olsen, 2006; Mysen and Aksnes, 2007; Feng et al., 2017) and in particular around contact-binary solid bodies (Marchis et al., 2014; Feng et al., 2016), was explored thoroughly in the last two decades (see a brief review in Lages et al., 2017). Here we use a generalized Kepler map technique (Lages et al., 2017) to de- scribe the global dynamics around cometary nuclei known to be bilobed. We recall that the Kepler map is a two-dimensional area-preserving map describing the eccentric circumbinary motion of a massless particle. The motion is de- scribed in terms of increments in particle's energy and orbital period measured at its pericenter and apocenter passages. The Kepler map was verified to be Email addresses: [email protected] (Jos´e Lages), [email protected] (Ivan I. Shevchenko), [email protected] (Guillaume Rollin) 1https://www.jpl.nasa.gov/news/news.php?feature=6817 Preprint submitted to Icarus October 30, 2017 a powerful tool to study resonant and chaotic orbital dynamics of comets, in particular Comet Halley (Petrosky, 1986; Chirikov and Vecheslavov, 1986, 1989; Rollin et al., 2015). One should emphasize that here we apply this technique to describe the orbital motion around cometary nuclei, in particular around the nucleus of Comet Halley. This outlines the general character of the technique. Also the Kepler map has been applied to very different domains, from strong microwave ionization of excited hydrogen atoms (Casati et al., 1988) and au- toionization of molecular Rydberg states (Benvenuto et al., 1994) to capture of dark matter by the solar system (Lages and Shepelyansky, 2013) and by grav- itating binaries in general (Rollin et al., 2015), also to description of transfer trajectories of spacecrafts (Ross and Scheeres, 2007). The employed Kepler map formalism allows one to straightforwardly esti- mate the characteristics of the chaotic zones around the cometary nuclei using simple analytical formulas, avoiding numerical integrations. What is more, in contrast to numerical integrations, it provides a direct physical insight in the dynamical problem of circumbinary motion: one is able to directly see which resonances overlap or interact; which Lyapunov timescales can be expected; how the chaotic diffusion in the energy variable may proceed; etc. In Lages et al. (2017), the Kepler map has been generalized to describe the motion of a massless particle in the gravitational field of a rotating irregularly shaped body modelled by a non-symmetric dumb-bell. This generalization was achieved by introduction of an additional parameter, ω, responsible for the arbitrary rate of rotation of the "central binary". Analytical expressions for the coefficients of the "kick function", representing the energy increment for the test particle per orbital revolution, were derived. In this article, we use the new technique introduced in Lages et al. (2017) to describe the qualitative chaotic dynamics around bilobed cometary nuclei, based on the data for five comets whose nuclei are well documented to be dumb-bells. Note that the majority of cometary nuclei whose shapes are well documented are in fact dumb-bells. The sizes of chaotic zones around the nuclei and the Lyapunov times of the motion inside these zones are estimated. The nucleus of Comet Halley is addressed in particular; we show that, due to its relatively slow rotation, a huge zone of chaos is generated around this object. 2. Bilobed cometary nuclei Many minor bodies in the Solar system are "potato"-shaped, roughly re- sembling ellipsoids, but it is not infrequent that asteroids and cometary nuclei resemble dumb-bells, i.e., they are more like dumb-bells than ellipsoids. There- fore, one can describe the body as a dumb-bell straightforwardly. Remarkably, this is the case for 5 comets out of 8 that have the shapes of their nuclei well- documented (Jorda et al., 2016). A well-known example is the nucleus of Comet 67P/Churyumov–Gerasi- menko, the target of the Rosetta mission (Jorda et al., 2016). Tantalizingly, Comet 1P/Halley, famous for its well-studied chaotic orbital dynamics and which was the first object studied within the Kepler map formalism (Chirikov 2 Table 1: Bilobed cometary nuclei: basic observational data Comet d, km M , kg m1/m2 P , h Refs. 67P/C–G 1P/Halley 2.62 7.7 9.982 × 1012 2.2 × 1014 3.4 12.404 Jorda et al. (2016) 2.6 176.4 Mer´enyi et al. (1990) Stooke and Abergel (1991) Schleicher et al. (2015) 8P/Tuttle 19P/Borrelly 5.0 4.0 4 × 1014 a 2 × 1013 2.14 11.4 Harmon et al. (2010) 3.5 25 Lamy et al. (1998) 103P/Hartley 1.2 2.2 × 1011 3.3 Soderblom et al. (2004) Oberst et al. (2004) Buratti et al. (2004) Britt et al. (2004) 18.2 Harmon et al. (2011) Thomas et al. (2013) Belton et al. (2013) a In the absence of observational data, the mean mass density of 8P/Tuttle has been assumed to be 0.5 g cm−3, equal to the mean density of 67P/C-G. and Vecheslavov, 1986, 1989), also has a dumb-bell nucleus (Mer´enyi et al., 1990; Stooke and Abergel, 1991). Other comets with well-documented dumb-bell nu- clei are 8P/Tuttle (Harmon et al., 2010), 19P/Borrelly (Oberst et al., 2004), and 103P/Hartley (Thomas et al., 2013). Such peculiar appearance is explained by non-uniform erosion of the surface of a cometary nucleus when it passes close to the Sun (Jewitt et al., 2003), or, alternatively, by "soft collisions" of single bodies (Rickman, H. et al., 2015). Table 1 summarizes the basic observational data about the above cited cometary nuclei: d is the separation of the centers of mass of the lobes; m2/m1 is the ratio of masses of the lobes (m2 < m1); M is the total mass of the nucleus; P is the observed rotation period of the nucleus. Let us take the example of the nucleus of Comet 67P/Churyumov–Gerasi- menko which from the observational data presented in Jorda et al. (2016) can be described as an aggregate of two merged bodies with the ratio of masses m1/m2 (cid:39) 3.4; and their centers of mass are separated by d (cid:39) 2.62 km. The total mass and the rotation period of the nucleus are, respectively, 9.982 × 1012 Kg (Patzold et al., 2016) and 12.404 h (Jorda et al., 2016). The Keplerian rate of rotation of a binary with masses m1, m2, and size d is straightforwardly calculated using the third Kepler's law (see, e.g., Mur- 3 ray and Dermott (1999), eq. (2.22)). Then, it is straightforward to calculate that the rotation rate ω of Comet 67P/Churyumov–Gerasimenko satisfies the relation ω (cid:39) 0.73ω0. Analogously, we have ω/ω0 (cid:39) 0.055, 0.33, 0.48, and 1.04 for 1P/Halley, 8P/Tuttle, 19P/Borrelly, and 103P/Hartley. Therefore, for the studied cometary nuclei, the rotation rate ω ranges from 0.055ω0 to 1.04ω0. However note that ω > ω0 are not generally probable, as a rubble-pile object would disintegrate at such high rates of rotation. 3. The dumb-bell map For the clarity of the subsequent presentation, let us review the Kepler map techniques. In Petrosky (1986) and Chirikov and Vecheslavov (1986, 1989), the classi- cal Kepler map was introduced as a tool for description of the chaotic motion of comets in eccentric orbits. The model consists in the assumption that the main perturbing effect of a planet is concentrated when the comet is close to the perihelion of its orbit. This effect is defined by the phase of encounter with the planet. The Kepler map has a single parameter. Its analytical formula was first derived in the restricted planar three-body problem framework in Pet- rosky (1986) and Petrosky and Broucke (1988). In Shevchenko (2011), it was demonstrated that the Kepler map, including analytical formulae for its param- eter, can be derived by quite elementary methods, based on the Jacobi integral formalism. Following a procedure analogous to that presented in Shevchenko (2011), the dumb-bell map can be straightforwardly derived by allowing for an arbitrary rate of rotation of the "central binary" (Lages et al., 2017). Consider the motion of a passively gravitating particle in the planar re- stricted three-body problem "m1–m2–particle", where the two masses m1 (cid:29) m2 are connected by a massless rigid rod, thus forming an asymmetric dumb-bell. Note that further on we consider solely the case of prograde (with respect to the dumb-bell rotation) orbits of the particle; analysis of the retrograde case is analogous. We choose an inertial Cartesian coordinate system with the origin at the dumb-bell's center of mass. Unless otherwise stated we express physical quan- tities is the following units: the dumb-bell size d, i.e., the distance between centers of mass of the m1 and m2 lobes, is set to equal to unity, d = 1; we set the product G(m1 + m2) = 1; consequently the angular frequency of the Keplerian orbital motion of the two lobes (i.e., the motion if the two masses m1 and m2 were unbound) is ω0 =(cid:112)G(m1 + m2)/d3 = 1. The motion of the particle with coordinates (x, y) is described by the differential equations x = ν x1 − x r3 13 + µ x2 − x r3 23 , y = ν y1 − y r3 13 + µ y2 − y r3 23 , (1) with x1 = −µ cos[ω(t − t0)], y1 = −µ sin[ω(t − t0)], x2 = ν cos[ω(t − t0)], y2 = ν sin[ω(t − t0)], 4 ri3 = (cid:112)(xi − x)2 + (yi − y)2 is the distance between the particle and the mi where (x1, y1) and (x2, y2) are the coordinates of m1 and m2, respectively; mass with i = 1, 2; µ = m2/(m1 + m2) is the reduced mass of m2, ν = 1 − µ is the reduced mass of m1; and t0 is an arbitrary time fixing the phase of the dumb-bell at t = 0. The quantity ω is an additional parameter, with respect to the usual equations of motion in the planar restricted three-body problem; this parameter is responsible for the arbitrary rotation frequency of the dumb-bell. From (1) and following either the methodology exposed in Shevchenko (2011) or the potential gravity based methodology exposed in Roy and Haddow (2003), it is possible to obtain (Lages et al., 2017), for any rate of revolution of the central binary ω and any reduced mass µ, the energy gain of the particle after a passage at the pericenter. The expression of the particle's energy kick function reads (Lages et al., 2017) ∆E (µ, q, ω, φ) (cid:39) W1 (µ, q, ω) sin (φ) + W2 (µ, q, ω) sin (2φ) , (2) where φ is the phase of the dumb-bell when particle is at pericenter, q is the pericenter distance, and where the first harmonic coefficient reads (cid:18) (cid:18) (cid:19) , (3) (4) (cid:19) . W1 (µ, q, ω) (cid:39) µν(ν − µ)21/4π1/2ω5/2q−1/4 exp − 23/2 3 ωq3/2 and the second harmonic coefficient reads W2 (µ, q, ω) (cid:39) −µν215/4π1/2ω5/2q3/4 exp − 25/2 3 ωq3/2 The quasi-constancy of q is an important issue considered, in particular, by Shevchenko (2015) in the framework of the Jacobi constant formalism. At a (cid:29) d, where a is the orbiting particle's semimajor axis (measured in the units of the semimajor axis of the central binary), the pericentric distance q is practically constant. In the framework of the restricted three-body problem, if one writes down the expression for the tertiary's energy increment ∆E together with the expression for the increment of perturber's phase angle φ between two consecutive passages of the tertiary at the pericenter, one obtains the usual Kepler map (Chirikov and Vecheslavov, 1986, 1989; Petrosky, 1986). In the case of arbitrary rate of rotation of the central binary, the map takes the form (Lages et al., 2017) Ei+1 = Ei + ∆E (µ, q, ω, φi) , φi+1 = φi + 2πω2Ei+1−3/2, (5) where Ei is the particle's energy at the ith passage at apocenter, and φi is the dumb-bell phase at the ith passage of the particle at pericenter. Apart from the analysis of the global dynamical behavior, the Kepler map can be used to reproduce individual trajectories of a modeled dynamical system, but only to a certain extent. Generally, a comparative analysis of any map's 5 performance versus a corresponding direct numerical integration for an indi- vidual trajectory does not make sense on the time intervals much greater than the trajectory's Lyapunov time TL, due to the essential sensitive dependence of the chaotic trajectories on the initial conditions. Therefore, in the case of the Kepler map, such a comparison for any chaotic trajectory does not make sense at all, because, according to Shevchenko (2007), the Kepler map's TL ∼ 1 in the units of the map iterations; thus, any chaotic trajectories computed by different methods, though with the same initial conditions, would substantially diverge already on the timescale of several map iterations. (Of course, this di- vergence does not influence any comparison of the global dynamical behaviors, which should be the same statistically.) On the other hand, it is worthwhile to check the performance of the Kepler map, versus a direct numerical inte- gration, taking as initial conditions those for an individual regular trajectory. Our computations show that a good accordance is observed indeed in this case: in particular, the deviations in the orbital energy do not exceed 0.1% on the time intervals as high as 103 periods of the binary; what is more, they do not increase but just oscillate. One should outline that the problem of numerical precision of the Kepler map in reproducing of individual trajectories has not yet been studied at all (though many authors used it to study statistics of chaotic trajectories), to our knowledge; this issue deserves a thorough separate study. 4. Borders of chaos domain Let us estimate the size of the chaotic zone generated by the rotating gravi- tating dumb-bell-shaped body, in application to cometary nuclei. The analytical method for this estimation (Shevchenko, 2015) is based on the Kepler map ap- proach developed for gravitating non-bound binaries, such as, e.g., binary stars. In Shevchenko (2015), analytical expressions for the energy kick functions ∆E and then estimations for chaos borders have been obtained in the cases of: (a) highly asymmetric binaries, µ (cid:28) 1, giving ∆E(µ, q, φ) (cid:39) W (µ, q) sin (φ) with W (µ, q) = W1(µ (cid:28) 1, q, ω0 = 1) and (b) equal mass binaries, µ = 1/2, giving ∆E(q, φ) (cid:39) W (q) sin (2φ) with W (q) = W2(µ = 1/2, q, ω0 = 1). For the five cometary nuclei, presented in Section 2, and considered as con- tact solid binaries, the reduced mass µ is moderate and does not vary much from a nucleus to another, since µ = 1/(1 + m1/m2) ranges from ≈ 0.22 (for 19P/Borrely) to ≈ 0.32 (for 8P/Tuttle ) (see Table 1). The cometary nuclei rotation rate ω can be very different from the Keplerian angular frequency ω0, since ω ranges from ω (cid:28) ω0 to ω (cid:39) ω0 (see Table 2). Combining (3) and (4), one has W1(µ, q, ω)/W2(µ, q, ω) (cid:39) − (ν − µ) 2−7/2q−1 exp . For a pericenter distance greater than the dumb-bell size, q > d, we observe that the amplitude W2 dominates over the amplitude W1 for the angular frequencies of rotation such that ω/ω0 (cid:28) (d/q)3/2. In this regime of slow rotation, the bor- der of chaos has been analytically estimated for rotating contact solid binaries 3 ωq3/2(cid:17) (cid:16) 23/2 6 (Lages et al., 2017) as Ecr = −∆Ecr (cid:39) −A(µν)2/5ω7/5q3/10 exp (cid:16)−Bωq3/2(cid:17) , (6) where A = 213/1032/5π3/5K−2/5 and B = 27/2/15. The width of the chaotic component around the separatrix (situated at E = 0) is ∆Ecr; the particles in the chaotic component can not dynamically diffuse below E < Ecr. Conversely, for ω/ω0 (cid:29) (d/q)3/2, W1 dominates over W2. For the sake of completeness, we give here an estimate of the border of chaos in this regime of fast rotation. Con- sequently, dropping the second harmonic term in (2) and using the substitution E = W1y, φ = x, the map (5) is reducible to where yi+1 = yi + sin xi, xi+1 = xi + λyi+1−3/2, λ = 2−1/2πωW −3/2 1 . (7) (8) As the dumb-bell map (7) can be, locally in (x, y) phase space, approximated by the standard map (Chirikov, 1979; Shevchenko, 2014), one finds for the location of the chaos border (cid:18) 3λ (cid:19)2/5 ycr = 2KG , (9) where KG = 0.971635406 . . . (Shevchenko, 2007). Using Eqs. (3) and (8) for W1 and λ, one obtains the half-width of the chaotic layer around the separatrix (y = 0) ∆Ecr = Ecr = W ycr (cid:39) A[µν(ν − µ)]2/5ω7/5q−1/10 exp (cid:16)−Bωq3/2(cid:17) , (10) where A = 2−1/232/5π3/5K and B = 25/2/15. The particle's critical eccen- tricity ecr, following from the relation ∆Ecr = −Ecr = 1/2acr = (1 − ecr)/2q, is −2/5 G (11) where ∆Ecr is given by Eq. (10). The orbits with e (cid:38) ecr(ω, q) are chaotic. Further on, the critical curve (11) will be superimposed on constructed stability diagrams. ecr = 1 − 2q∆Ecr, As mentioned already in the comment to Eqs. (7)-(8), the Kepler map can be linearized in the energy variable so that to approximate the motion locally by the standard map. This procedure provides powerful analytical means to estimate the local properties of the chaotic motion: local diffusion rates, local Lyapunov timescales, proximity to resonances, etc. 7 Figure 1: Stability diagrams of the orbital motion around four cometary nuclei ordered by increasing ω: 8P/Tuttle (top left, ω (cid:39) 0.33ω0), 19P/Borrelly (top right, ω (cid:39) 0.48ω0), 67P/Churyumov-Gerasimenko (bottom left, ω (cid:39) 0.73ω0), 103P/Hartley (bottom right, ω (cid:39) 1.04ω0). The chaotic domain is shown by the reddish area. Chaos is determined by computing the maximum Lyapunov exponent Λ for a trajectory with initial orbital elements (q, e). Number of iterations is 106. The critical curve (11) is shown taking into account only the first harmonic term in ∆E (2) with amplitude W1 (3) (dashed black line) and taking into account only the second harmonic in ∆E (2) with amplitude W2 (4) (solid black line). The vertical white dash-dotted line marks the pericenter q0 at which W2 = W1. For q < q0 (q > q0), W2 > W1 (W2 < W1). The ratio W2/W1 ranges from 10.2 at q = 4d to 0.211 at q = 8d for 8P/Tuttle, from 5.72 at q = 3d to 0.0305 at q = 7d for 19P/Borrelly, from 3.4 at q = 2.5d to 0.0155 at q = 5.5d for 67P/Churyumov-Gerasimenko, from 2.64 at q = 2d to 0.00801 at q = 4.5d for 103P/Hartley. 8 00.10.20.30.40.50.60.70.80.944.555.566.577.58eq/d00.10.20.30.40.50.60.70.80.944.555.566.577.588P/Tuttle00.10.20.30.40.50.60.70.80.933.544.555.566.57eq/d00.10.20.30.40.50.60.70.80.933.544.555.566.5719P/Borrelly00.10.20.30.40.50.60.70.80.92.533.544.555.5eq/d00.10.20.30.40.50.60.70.80.92.533.544.555.5 67P/Churyumov-Gerasimenko00.10.20.30.40.50.60.70.80.922.533.544.5eq/d00.10.20.30.40.50.60.70.80.922.533.544.5103P/Hartley-5-4-3-2-101Log10Λ5. Dynamics around nuclei of Comets 5.1. Stability diagrams Let us construct stability diagrams, using the dumb-bell map (5), by comput- ing the Lyapunov exponents on a fine grid of initial conditions (e, q). The Lya- punov exponents are calculated by means of iterating concurrently the dumb- bell map and its tangent map (see the general method described in Chirikov, 1979). Fig. 1 presents the stability diagrams for particles orbiting cometary nu- clei of 8P/Tuttle, 19P/Borrelly, 67P/Churyumov-Gerasimenko and 103P/Har- tley whose physical parameters are given in Section 2. For any given initial conditions (e, q), the motion is regarded as chaotic if the maximum Lyapunov exponent Λ is non-zero. From Fig. 1 we see that a common peculiarity is that the border delimiting the chaotic domain (reddish area) and the regular domain (bluish area) is ragged. Here the most prominent teeth of instability correspond to integer p:1 and half-integer p + 1 2 :1 resonances. The prominence of the both integer and half-integer teeth at the ragged border, well visible, e.g., in the 8P/Tuttle case, is explained by the fact that the two lobes of the considered cometary nuclei are comparable, i.e., m1 ∼ m2. On the stability diagrams (Fig. 1), we superimpose the critical curves ecr(q) (11) computed for the dynamical regimes with ω/ω0 (cid:29) (d/q)3/2, using (10) as the expression for ∆Ecr in (11), and ω/ω0 (cid:28) (d/q)3/2, using (6) as the expres- sion for ∆Ecr in (11). From Fig. 1, we clearly see that the critical curve for ω/ω0 (cid:29) (d/q)3/2 (dashed line) approximately describes the smoothed border of the chaotic domain at large pericenter distances q and high eccentricities e. Conversely, the critical curve for ω/ω0 (cid:28) (d/q)3/2 (solid line) approximately describes the smoothed border of the chaotic domain at lower pericenter dis- tances and small eccentricities. These agreements testify the adequacy of the analytical approximation for chaos border (Section 4). However we note that this approximation becomes less accurate at ω (cid:38) ω0 as e.g., illustrated by the 103P/Hartley case. 5.2. Circumnuclear chaotic zone Let us define the central chaotic zone as the region in q where even the particles that start in circular orbits (e = 0) move chaotically. From Fig. 1, we retrieve the known fact that, µ being similar for all the cometary nuclei, the extent of the central chaotic zone significantly increases as the rotation rate ω slows down (Lages et al., 2017). As an illustration, here (Fig. 1) the radius of the central chaotic zone ranges from q (cid:39) 3d around 103P/Hartley (ω (cid:39) 1.04ω0) to q (cid:39) 5d around 8P/Tuttle (ω (cid:39) 0.33ω0). As the chaotic border is determined by the critical eccentricity ecr(q), the solution of the equation ecr(q) = 0 at q > 1 can be taken as the radius of the chaotic zone around the cometary nuclei. Fig. 2 shows the extent of the central chaotic zone as a function of the rotation rate ω. In the 0.3 (cid:46) ω/ω0 (cid:46) 1 region, which comprises the rotation rates of 8P/Tuttle, 19P/Borrelly, 67P/Churyumov-Gerasimenko, and 103P/Hartley, we clearly see 9 Figure 2: Extent of the central chaotic zone around a cometary nucleus as a function of the rotation rate ω. Upper panel: the red domain corresponds to the central chaotic zone extent for the dumb-bell symmetric case µ = 1/2; the blue domain is the domain of stable orbits. The borders of the central chaotic zone for µ (cid:39) 0.22 (the mass parameter of 19P/Borrelly) and µ (cid:39) 0.32 (the mass parameter of 8P/Tuttle) are shown by the white dashed line and the white dash-dotted line, respectively. The rotation rates ω of the five cometary nuclei described in Section 2 are represented by vertical white lines. Locations of the integer resonance 1:1 and the half-integer resonances 1:2 and 2:1 are represented by red lines. Bottom panels: schematic presentations of resonances and the chaotic zones (at e = 0) around the nucleus of 1P/Halley (bottom left), a model cometary nucleus with rotation rate ω = 0.1ω0 and mass parameter µ = 0.5 (bottom middle), and the nucleus of 67P/Churyumov-Gerasimenko (bottom right). The centers of mass of the cometary nuclei are located at the centers of these panels. 10 that at µ from ≈ 0.22 to ≈ 0.32 the extents of the chaotic zone are almost the same (see the white dashed and dash-dotted curves in Fig. 2 in the range 0.3 (cid:46) ω/ω0 (cid:46) 1) and are even very close to the extent of the chaotic zone in the µ = 1/2 symmetric case. For these 4 cometary nuclei a generic illustration of the central chaotic zone is given in the bottom right panel of Fig. 2. In the 0.1 (cid:46) ω/ω0 (cid:46) 0.3 region, a zone of regular orbits exists surrounding the close vicinity of the cometary nucleus which is consequently insulated from an annular chaotic zone. This configuration is explained by the fact that the ecr(q) function has two roots in the 0.1 (cid:46) ω/ω0 (cid:46) 0.3 region, these two roots giving the inner and outer radii of the central chaotic zone as illustrated in the bottom middle panel of Fig. 2 for a model cometary nucleus rotating with ω = 0.1ω0 and having the mass parameter µ = 1/2. In the ω (cid:46) 0.1ω0 region, according to the upper panel of Fig. 2, the above defined central chaotic zone does not exist. This is in accordance with the stability diagram (see Fig. 3, left panel) for the particles orbiting a dumb-bell with the characteristics of the 1P/Halley nucleus (ω (cid:39) 0.055ω0, µ (cid:39) 0.28). In Fig. 3, left panel, a particle put initially in a circular orbit has regular dynamics if the initial radius does not correspond to integer or half-integer resonances (corresponding to the resonant teeth reaching the line ecr = 0 in Fig. 3, left panel). The central chaotic zone, defined as the domain in q for which any initial eccentricity implies chaos, is absent as illustrated for the 1P/Halley nucleus in the bottom left panel of Fig. 2. However for higher eccentricities the chaotic domain is quite extended with e.g., a width of ∼ 10d for e (cid:39) 0.5 and ∼ 20d for e (cid:39) 0.8. Usually the Kepler map is used to describe the highly eccentric orbital mo- tion around a binary composed of a primary and a secondary acting as a per- turber, e.g., the chaotic dynamics of 1P/Halley around the Sun and Jupiter (Chirikov and Vecheslavov, 1989; Rollin et al., 2015). Also the Kepler map was derived and used for ω a3/2 > ω0 d3/2, where a = −1/2E, to describe the molec- ular Rydberg states with a rotating dipole core (Casati et al., 1988; Benvenuto et al., 1994). In these two cases of application, the orbital period of the tertiary is greater than the binary rotation period, and the usually considered dynamics within the Kepler map framework lies above the 1:1 resonance line in Fig. 2. According to Fig. 2 the motion of a body orbiting close to the 1P/Halley nucleus may take place below this resonance line. Consequently in order to check the relevance of application of the dumb-bell map (5) below the 1:1 resonance line, i.e., for ω a3/2 < ω0 d3/2, we have numerically integrated the equations of motion of a particle around a dumb-bell with the 1P/Halley nucleus characteristics. We have obtained the corresponding stability diagram (Fig. 3, right panel) which exhibits a ragged chaotic border qualitatively similar to that in the stability diagram (Fig. 3, left panel) obtained iterating the dumb-bell map (5). Note that in these two stability diagrams, the Lyapunov exponents are measured in different units. 11 Figure 3: Stability diagrams of the orbital motion around 1P/Halley (ω (cid:39) 0.055, µ (cid:39) 0.28) obtained (left panel) by iterating the dumb-bell map (5) and (right panel) by integrating the equations of motion of a particle around a contact-binary with 1P/Halley's physical character- istics. The chaotic domain is shown by the reddish area. Chaos is determined by computing the maximum Lyapunov exponent Λ for a trajectory with initial orbital elements (q, e). The critical curve (11) is shown taking into account only the first harmonic term in ∆E (2) with amplitude W1 (3) (dashed black line) and taking into account only the second harmonic term in ∆E (2) with amplitude W2 (4) (solid black line). Left panel: the number of iterations of the dumb-bell map (5) is 106. The Lyapunov exponent Λ has the physical dimension of the inverse of the number of iterations of the map. Right panel: the numerical integration of the equations of motion have been performed over the time duration 104T0 where T0 = 2π/ω0. The Lyapunov exponent Λ has the physical dimension of the inverse of time, and (ΛP )−1 expresses the Lyapunov time in the units of the rotation period P of the nucleus. 12 00.10.20.30.40.50.60.70.80.9234567891011121314151617181920eq/d00.10.20.30.40.50.60.70.80.9234567891011121314151617181920-5-4-3-2-1 0 1Log10 Λ1P/Halley00.10.20.30.40.50.60.70.80.9234567891011121314151617181920eq/d00.10.20.30.40.50.60.70.80.9234567891011121314151617181920-3-2.5-2-1.5-1-0.5Log10 (ΛP)1P/HalleyFigure 4: Left panel: Contour plot showing the difference, Λ − CK, between the analytical expression of the Lyapunov exponent Λ(µ, q, ω) (12) and the Chirikov constant CK (cid:39) 2.2; for mass parameters µ = 0.1 (dashed lines), µ = 0.28 (mass parameter of 1P/Halley, solid lines), and µ = 0.5 (dot-dashed lines). Each color is associated to a value of Λ − CK. Right panel: We show data from the "1P/Halley" stability diagram (left panel of Fig. 3) but selecting only initial conditions (e, q) leading to a Lyapunov exponent 2.1 < Λ < 2.3. 5.3. The Lyapunov times, the chaotic zone radii, and the Hill radii Roughly speaking, the inverse of the Lyapunov exponent obtained at ini- tial conditions (q, e) by iterating the dumb-bell map (5) gives the number of iterations needed to observe the onset of chaos. As shown in Section 4, the dumb-bell map (5) can be rewritten as the original Kepler map (7) for which the Lyapunov exponent inside the around–the–separatrix chaotic component is known (Shevchenko, 2007) to be equal to Λ(λ) = CK − 3/λ, (12) where CK (cid:39) 2.2 is the Chirikov constant and λ is the adiabaticity parameter of the Kepler map (7). Here, in the case of rotating small bodies the adiabaticity parameter reads λ = 21/2πωW2−3/2 (Lages et al., 2017). Even in the case of 1P/Halley, with the slowest rotation rate considered in this study, chaos is ∼ non-adiabatic since the Kepler map parameter is still very high, λ ∼ ωW 100 (cid:29) 1 (from Lages et al. (2017) W2 ∼ 10−2(dω0)2 for ω (cid:39) 0.05ω0 and q > d). Consequently the Lyapunov exponent in the chaotic domain is expected to be approximately Λ (cid:39) CK (cid:39) 2.2 for any set of the physical parameters of this study as is shown by Fig. 4, left panel. We have verified, from Fig. 1 and Fig. 3, left panel, that the Lyapunov exponent is indeed approximately 2.2 for the chaotic orbits around each of the five cometary nuclei (see Fig. 4, right panel as an illustration for the 1P/Halley case). −3/2 2 The inverse of the Lyapunov exponent obtained by the direct numerical integration of Newton's equations gives, by the order of magnitude, the time (in 13 0.111012345678910Λ-CK0.1110ω/ω012345678910q/dΛ-CK0.1110ω/ω012345678910q/dΛ-CK0.1110ω/ω012345678910q/dΛ-CK10-110-110-110-210-210-210-210-310-310-310-310-310-310-610-610-610-610-610-610-60.1110ω/ω012345678910q/d00.10.20.30.40.50.60.70.80.9234567891011121314151617181920eq/d00.10.20.30.40.50.60.70.80.92345678910111213141516171819201P/HalleyTable 2: Bilobed cometary nuclei: sizes of chaotic zones and the Lyapunov times Comet ω/ω0 e (cid:39) 0 Rch, km e (cid:39) 0.5 RHill, km Rch/RHill, % e (cid:39) 0 e (cid:39) 0.5 TL, h TL/P 67P/C–G 1P/Halley 8P/Tuttle 19P/Borrelly 103P/Hartley 0.73 0.055 0.33 0.48 1.04 9 – 25 16 3.4 11 31-108 a 32.5 21 4.2 320 200 620 300 52 3% – 4% 5% 7% 3.3% 16-54% 5.2% 7% 8.1% 35 230 28 66 56 2.9 1.3 2.5 2.6 3.1 a For the 1P/Halley nucleus we give the inner and outer radii of the annular chaotic zone at e (cid:39) 0.5. constant natural time units) needed at given initial conditions (e, q) for chaos to develop. The value of the Lyapunov exponent for an orbit with initial (e, q) in Fig. 3 (right panel) can be straightforwardly estimated using the corresponding value from Fig. 3 (left panel) divided by the corresponding mean orbital period. Table 2 summarizes the radii of the chaotic zone Rch assessed directly from the (e, q) stability diagrams for the cometary nuclei (Figs. 1 and 3). Except the case of 1P/Halley, the chaotic zone radii at e (cid:39) 0 and at e (cid:39) 0.5 are similar in each case. Let us compare these radii to the corresponding Hill radii. In the planar circular restricted three-body problem, the Hill radius is given by RHill ≈ a(cid:12) (M/3M(cid:12))1/3 where a(cid:12) is the semi-major axis of a comet's orbit around the Sun. In the elliptic problem, a "pericenter scaling" for RHill is given by RHill ≈ q(cid:12) (M/3M(cid:12))1/3, where q(cid:12) is a comet's perihelion distance (Hamilton and Burns, 1992). The calculated Hill radii RHill are presented in Table 2. From Table 2, we see that that the typical size of 1P/Halley's chaotic zone Rch is the largest one in the sample, and it seems to engulf an essential part of the Hill sphere, at least for the orbits of moderate and high eccentricity (38% of the chaotic zone overlaps with the Hill sphere at e (cid:39) 0.5). This is an outcome of the slowness of rotation of the nucleus. As mentioned above, ω (cid:38) ω0 are not generally probable, as a rubble-pile ob- ject would disintegrate at such high rates of rotation. For Comet 103P/Hartley, ω ≈ ω0; therefore, the nucleus may formally be on the brink of disintegration. The chaotic zone of this object is formally minimal in size in the sense that at ω < ω0 the zone would be larger. Following the general approach presented in (Shevchenko, 2007, section 5), the local Lyapunov time TL can be estimated, for a satellite assumed to move within the chaotic domain around a nucleus. It is approximately equal to the ratio of the satellite's orbital period and the map's Lyapunov exponent. The map's Lyapunov exponent averaged over the chaotic layer does not depend prac- 14 tically on the adiabaticity parameter, if the latter is high enough; see Shevchenko (2007). The local Lyapunov time is estimated taking the satellite's orbital pe- riod Torb corresponding to the radius of the central chaotic zone (a = Rch) at e (cid:39) 0.5 as given in Table 2, as this gives the maximum possible Torb for the chaotic motion at zero and moderate eccentricities, and, consequently, the maximum possible local Lyapunov time. These estimations of the typical local Lyapunov time for each cometary nucleus are given in Table 2. One can readily see that the Lyapunov times do not differ substantially, ranging from ∼ 1 d to ∼ 10 d. The relative range is even less, if the Lyapunov times are measured in the units of the rotation period of the cometary nuclei; then they range from 1.3 to 3.1. For the satellites with smaller orbits (a < Rch), the local Lyapunov time is smaller, and, consequently, the motion is less predictable. Among the five objects listed in Table 1, four out of the five had been visited and observed closely in their vicinities by space probes: 1P/Halley (by Vega-1, Vega-2, Giotto in 1986), 19P/Borrelly (by Deep Space 1 in 2001), 67P/C–G (by Rosetta 1 in 2014–16), and 103P/Hartley (by Deep Impact EPOXI mission in 2010). No satellites of the nuclei had been observed (apart from appar- ently replenishable clouds of "grains") during these close rendezvous, although meticulous surveys had been done in some cases; see Bertini et al. (2015) and references therein. The observed absence of cometary satellites is apparently in agreement with our theoretical findings on the large extents of the circumnuclear chaotic zones. Among these four nuclei, two objects were observed in such a detail as to allow for considering the close-to-nucleus mass transfer and dynamics of various ejecta, which include, in particular, quite large (decimetre sized) chunks of water ice (A'Hearn et al., 2011; Keller et al., 2017). Thus, there exists a source of ma- terial which can be introduced in orbits around nuclei, as theoretically described in Fulle (1997) and Scheeres and Marzari (2000). However, our theoretical find- ings imply that, even in the absence of any forces other than gravitational, no long-lived orbits of any material may sustain inside the circumnuclear chaotic zones. One may say that the rotating irregularly-shaped nuclei clean up their vicini- ties. This clearing is analogous in some way to the clearing of the "Wisdom gap" in the coorbital vicinities of a planet that is massive enough. (On Wisdom's coorbital chaotic layer, see, e.g., Murray and Dermott 1999.) The dynamical difference is that the Wisdom gap is formed by the overlap of accumulating first- order mean motion resonances in the coorbital vicinity of a planet, whereas the circumnuclear chaotic zone is formed by the overlap of accumulating integer and half-integer orbit-spin resonances with the rotating nucleus, as discussed in Section 5.1. Therefore, our theoretical expectation is that no long-lived satellites, or any long-lived non-replenishable halos formed of large particles, can be normally observed to exist around bilobed cometary nuclei, within the defined boundaries of the circumnuclear chaotic zone. 15 6. Conclusions We have used the generalized "dumb-bell" Kepler map, introduced recently in Lages et al. (2017), to describe the qualitative chaotic dynamics around bilobed cometary nuclei. The analysis has been based on the data for five comets whose nuclei are well-documented to resemble dumb-bells. As lobes' masses are comparable, m1 ∼ m2, the chaotic zone's configuration and extents depend mostly on the cometary nuclei rotation rate ω. The sizes of chaotic zones around the nuclei and the Lyapunov times of the motion inside these zones have been estimated. In the case of Comet 1P/Halley, the chaotic zone seems to en- gulf an essential part of the Hill sphere, at least for orbits of moderate to high eccentricity. Therefore, simple analytical formulas, based on the Kepler map formalism, allow one to straightforwardly estimate the sizes of chaotic zones around the dumb-bell cometary nuclei and the Lyapunov times of the motion inside these zones. In practice, the obtained numerical estimates of the characteristics of the chaotic zones around the five considered objects allow one to judge where (in the space of orbital parameters) any small natural satellites cannot be expected to orbit any of these objects, or where any artificial satellite cannot be put in a stable orbit. On the other hand, the knowledge of the Lyapunov times allows one to judge on which timescales the coordinates of a satellite orbiting in the chaotic zone are predictable. Acknowledgements We are most thankful to the referees for important remarks. We are grateful to Dima L. Shepelyansky for valuable remarks, comments, and discussions. I.I.S. benefited from a grant of Bourgogne-Franche-Comt´e region. I.I.S. was supported in part by the Russian Foundation for Basic Research (project No. 17-02- 00028) and the Programmes of Fundamental Research of the Russian Academy of Sciences "Fundamental Problems of Nonlinear Dynamics" and "Fundamental Problems of the Solar System Study and Exploration". References A'Hearn, M.F., Belton, M.J.S., Delamere, W.A., Feaga, L.M., Hampton, D., Kissel, J., Klaasen, K.P., McFadden, L.A., Meech, K.J., Melosh, H.J., Schultz, P.H., Sunshine, J.M., Thomas, P.C., Veverka, J., Wellnitz, D.D., Yeomans, D.K., Besse, S., Bodewits, D., Bowling, T.J., Carcich, B.T., Collins, S.M., Farnham, T.L., Groussin, O., Hermalyn, B., Kelley, M.S., Kelley, M.S., Li, J.Y., Lindler, D.J., Lisse, C.M., McLaughlin, S.A., Merlin, F., Protopapa, S., Richardson, J.E., Williams, J.L., 2011. EPOXI at Comet Hartley 2. Science 332, 1396–1400. URL: http://science. sciencemag.org/content/332/6036/1396, doi:10.1126/science.1204054, arXiv:http://science.sciencemag.org/content/332/6036/1396.full.pdf. 16 Bartczak, P., Breiter, S., 2003. Double Material Segment as the Model of Irreg- ular Bodies. Celest. Mech. Dyn. Astron. 86, 131–141. Belton, M.J., Thomas, P., Li, J.Y., Williams, J., Carcich, B., A'Hearn, M.F., McLaughlin, S., Farnham, T., McFadden, L., Lisse, C.M., Collins, S., Besse, S., Klaasen, K., Sunshine, J., Meech, K.J., Lindler, D., 2013. The complex spin state of 103P/Hartley 2: Kinematics and orientation in space. Icarus 222, 595 – 609. URL: http://www.sciencedirect.com/science/article/ pii/S0019103512002667, doi:10.1016/j.icarus.2012.06.037. Benvenuto, F., Casati, G., Shepelyansky, D.L., 1994. Chaotic autoionization of molecular Rydberg states. Phys. Rev. Lett 72, 1818–1821. doi:10.1103/ PhysRevLett.72.1818. Bertini, I., Guti´errez, P. J., Lara, L. M., Marzari, F., Moreno, F., Pajola, M., La Forgia, F., Sierks, H., Barbieri, C., Lamy, P., Rodrigo, R., Koschny, D., Rickman, H., Keller, H. U., Agarwal, J., A'Hearn, M. F., Barucci, M. A., Bertaux, J.-L., Cremonese, G., Da Deppo, V., Davidsson, B., De- bei, S., De Cecco, M., Ferri, F., Fornasier, S., Fulle, M., Giacomini, L., Groussin, O., Guttler, C., Hviid, S. F., Ip, W.-H., Jorda, L., Knollenberg, J., Kramm, J. R., Kuhrt, E., Kuppers, M., Lazzarin, M., Lopez Moreno, J. J., Magrin, S., Massironi, M., Michalik, H., Mottola, S., Naletto, G., Ok- lay, N., Thomas, N., Tubiana, C., Vincent, J.-B., 2015. Search for satel- lites near comet 67P/Churyumov-Gerasimenko using rosetta/osiris images. A&A 583, A19. URL: https://doi.org/10.1051/0004-6361/201525979, doi:10.1051/0004-6361/201525979. Britt, D., Boice, D., Buratti, B., Campins, H., Nelson, R., Oberst, J., Sandel, B., Stern, S., Soderblom, L., Thomas, N., 2004. The mor- phology and surface processes of Comet 19P/Borrelly. Icarus 167, 45 – 53. URL: http://www.sciencedirect.com/science/article/pii/ S0019103503002823, doi:10.1016/j.icarus.2003.09.004. Buratti, B., Hicks, M., Soderblom, L., Britt, D., Oberst, J., Hillier, J., 2004. Deep Space 1 photometry of the nucleus of Comet 19P/Borrelly. Icarus 167, 16 – 29. URL: http://www.sciencedirect.com/science/article/ pii/S0019103503002732, doi:10.1016/j.icarus.2003.05.002. Casati, G., Guarneri, I., Shepeliansky, D.L., 1988. Hydrogen atom in monochro- IEEE Journal of matic field: Chaos and dynamical photonic localization. Quantum Electronics 24, 1420–1444. doi:10.1109/3.982. Chauvineau, B., Farinella, P., Mignard, F., 1993. Planar orbits about a triaxial body - Application to asteroidal satellites. Icarus 105, 350–384. doi:10.1006/ icar.1993.1134. Chirikov, B.V., 1979. A universal instability of many-dimensional URL: http: 263–379. oscillator systems. Physics Reports 52, 17 //www.sciencedirect.com/science/article/pii/0370157379900231, doi:10.1016/0370-1573(79)90023-1. Chirikov, B.V., Vecheslavov, V.V., 1986. Chaotic dynamics of Comet Halley. INP Preprint 86–184. Institute of Nuclear Physics, Novosibirsk. (Available at http://www.quantware.ups-tlse.fr/chirikov/publbinp.html). Chirikov, B.V., Vecheslavov, V.V., 1989. Chaotic dynamics of Comet Halley. Astron. Astrophys. 221, 146–154. Feng, J., Noomen, R., Hou, X., Visser, P., Yuan, J., 2017. 1:1 ground-track resonance in a uniformly rotating 4th degree and order gravitational field. Celest. Mech. Dyn. Astron. 127, 67–93. URL: http://dx.doi.org/10.1007/ s10569-016-9717-9, doi:10.1007/s10569-016-9717-9. Feng, J., Noomen, R., Visser, P., Yuan, J., 2016. Numerical analysis of orbital motion around a contact binary asteroid system. Advances in Space Research 58, 387–401. doi:10.1016/j.asr.2016.04.032. Fulle, M., 1997. Injection of large grains into orbits around comet nuclei. Astron. Astrophys. 325, 1237–1248. Hamilton, D.P., Burns, J.A., 1992. Orbital stability zones about aster- Icarus 96, 43–64. URL: http://www.sciencedirect.com/science/ oids. article/pii/001910359290005R, doi:10.1016/0019-1035(92)90005-R. Harmon, J.K., Nolan, M.C., Giorgini, J.D., Howell, E.S., 2010. Radar Icarus 207, observations of 8P/Tuttle: 499–502. URL: http://www.sciencedirect.com/science/article/pii/ S001910350900517X, doi:10.1016/j.icarus.2009.12.026. A contact-binary comet. Harmon, J.K., Nolan, M.C., Howell, E.S., Giorgini, J.D., Taylor, P.A., 2011. Radar Observations of Comet 103P/Hartley 2. Astrophys. J. 734, L2. URL: http://stacks.iop.org/2041-8205/734/i=1/a=L2. Jewitt, D., Sheppard, S., Fern´andez, Y., 2003. 143P/Kowal-Mrkos and the Shapes of Cometary Nuclei. Astron. J. 125, 3366. URL: http://stacks. iop.org/1538-3881/125/i=6/a=3366. Jorda, L., Gaskell, R., Capanna, C., Hviid, S., Lamy, P., Durech, J., Faury, G., Groussin, O., Guti´errez, P., Jackman, C., Keihm, S., Keller, H., Knollenberg, J., Kuhrt, E., Marchi, S., Mottola, S., Palmer, E., Schloerb, F., Sierks, H., Vincent, J.B., A'Hearn, M., Barbieri, C., Rodrigo, R., Koschny, D., Rick- man, H., Barucci, M., Bertaux, J., Bertini, I., Cremonese, G., Deppo, V.D., Davidsson, B., Debei, S., Cecco, M.D., Fornasier, S., Fulle, M., Guttler, C., Ip, W.H., Kramm, J., Kuppers, M., Lara, L., Lazzarin, M., Moreno, J.L., Marzari, F., Naletto, G., Oklay, N., Thomas, N., Tubiana, C., Wenzel, K.P., 2016. The global shape, density and rotation of Comet 67P/Churyumov- Gerasimenko from preperihelion Rosetta/OSIRIS observations. Icarus 277, 18 257 – 278. URL: http://www.sciencedirect.com/science/article/pii/ S0019103516301385, doi:10.1016/j.icarus.2016.05.002. Keller, H.U., Mottola, S., Hviid, S.F., Agarwal, J., Kuhrt, E., Skorov, Y., Otto, K., Vincent, J.B., Oklay, N., Schroder, S.E., Davidsson, B., Pajola, M., Shi, X., Bodewits, D., Toth, I., Preusker, F., Scholten, F., Sierks, H., Barbieri, C., Lamy, P., Rodrigo, R., Koschny, D., Rickman, H., A'Hearn, M.F., Barucci, M.A., Bertaux, J.L., Bertini, I., Cremonese, G., Da Deppo, V., Debei, S., De Cecco, M., Deller, J., Fornasier, S., Fulle, M., Groussin, O., Guti´errez, P.J., Guttler, C., Hofmann, M., Ip, W.H., Jorda, L., Knollenberg, J., Kramm, J.R., Kuppers, M., Lara, L.M., Lazzarin, M., Lopez-Moreno, J.J., Marzari, F., Naletto, G., Tubiana, C., Thomas, N., 2017. Seasonal mass transfer on the nucleus of comet 67P/ChuyumovGerasimenko. Monthly Notices of the Royal Astronomical Society 469, S357–S371. URL: http://dx.doi.org/10. 1093/mnras/stx1726, doi:10.1093/mnras/stx1726. Lages, J., Shepelyansky, D.L., 2013. Dark matter chaos in the Solar system. MN- RAS 430, L25–L29. URL: http://mnrasl.oxfordjournals.org/content/ 430/1/L25.abstract, doi:10.1093/mnrasl/sls045. Lages, J., Shepelyansky, D.L., Shevchenko, I.I., 2017. Chaotic zones around rotating small bodies. Astron. J. 153, 272. URL: http://stacks.iop.org/ 1538-3881/153/i=6/a=272, doi:10.3847/1538-3881/aa7203. Lamy, P.L., Toth, I., Weaver, H.A., 1998. Hubble Space Telescope observations of the nucleus and inner coma of comet 19P/1904 Y2 (Borrelly). Astron. Astrophys. 337, 945–954. Marchis, F., Durech, J., Castillo-Rogez, J., Vachier, F., Cuk, M., Berthier, J., Wong, M.H., Kalas, P., Duchene, G., van Dam, M.A., Hamanowa, H., Viikinkoski, M., 2014. The Puzzling Mutual Orbit of the Binary Trojan Asteroid (624) Hektor. Astrophys. J. 783, L37. doi:10.1088/2041-8205/ 783/2/L37, arXiv:1402.7336. Mer´enyi, E., Foldy, L., Szego, K., T´oth, I., Kondor, A., The landscape of Comet Halley. //www.sciencedirect.com/science/article/pii/001910359090194E, doi:10.1016/0019-1035(90)90194-E. Icarus 86, 9–20. 1990. URL: http: Murray, C.D., Dermott, S.F., 1999. Solar system dynamics. Cambridge Uni- versity Press, Cambridge. Mysen, E., Aksnes, K., 2007. On the dynamical stability of the Rosetta orbiter. II. Astron. Astrophys. 470, 1193–1199. doi:10.1051/0004-6361:20077472. Mysen, E., Olsen, Ø., Aksnes, K., 2006. Chaotic gravitational zones around a regularly shaped complex rotating body. Planet. Space Sci. 54, 750–760. doi:10.1016/j.pss.2006.04.005. 19 Oberst, J., Giese, B., Howington-Kraus, E., Kirk, R., Soderblom, L., Bu- ratti, B., Hicks, M., Nelson, R., Britt, D., 2004. The nucleus of Comet Borrelly: Icarus 167, 70 – 79. URL: http://www.sciencedirect.com/science/article/pii/ S0019103503002707, doi:10.1016/j.icarus.2003.05.001. a study of morphology and surface brightness. Olsen, Ø., 2006. Orbital resonance widths in a uniformly rotating second de- gree and order gravity field. Astron. Astrophys. 449, 821–826. doi:10.1051/ 0004-6361:20054451. Patzold, M., Andert, T., Hahn, M., Asmar, S.W., Barriot, J.P., Bird, M.K., Hausler, B., Peter, K., Tellmann, S., Grun, E., Weissman, P.R., Sierks, H., Jorda, L., Gaskell, R., Preusker, F., Scholten, F., 2016. A homogeneous nucleus for comet 67P/Churyumov-Gerasimenko from its gravity field. Nature 530, 63–65. doi:10.1038/nature16535. Petit, J., Durda, D., Greenberg, R., Hurford, T., Geissler, P., 1997. The long- term dynamics of Dactyl's orbit. Icarus 130, 177–197. doi:10.1006/icar. 1997.5788. Petrosky, T.Y., 1986. Chaos and cometary clouds in the solar system. Physics Letters A 117, 328 – 332. URL: http://www.sciencedirect.com/science/ article/pii/0375960186906730, doi:10.1016/0375-9601(86)90673-0. Petrosky, T.Y., Broucke, R., 1988. Area-preserving mappings and deterministic chaos for nearly parabolic motions. Celestial Mechanics 42, 53–79. Rickman, H., Marchi, S., A'Hearn, M. F., Barbieri, C., El-Maarry, M. R., Guttler, C., Ip, W.-H., Keller, H. U., Lamy, P., Marzari, F., Massironi, M., Naletto, G., Pajola, M., Sierks, H., Koschny, D., Rodrigo, R., Barucci, M. A., Bertaux, J.-L., Bertini, I., Cremonese, G., Da Deppo, V., Debei, S., De Cecco, M., Fornasier, S., Fulle, M., Groussin, O., Gutirrez, P. J., Hviid, S. F., Jorda, L., Knollenberg, J., Kramm, J.-R., Kuhrt, E., Kuppers, M., Lara, L. M., Lazzarin, M., Lopez Moreno, J. J., Michalik, H., Sabau, L., Thomas, N., Vincent, J.-B., Wenzel, K.-P., 2015. Comet 67P/Churyumov-Gerasimenko: Constraints on its origin from OSIRIS observations. Astron. Astrophys. 583, A44. URL: https://doi.org/10.1051/0004-6361/201526093, doi:10. 1051/0004-6361/201526093. Rollin, G., Haag, P., Lages, Symplectic map descrip- 1017 tion of Halley's – 1022. URL: http://www.sciencedirect.com/science/article/pii/ S0375960115001048, doi:10.1016/j.physleta.2015.02.001. Physics Letters A 379, J., 2015. comet dynamics. Rollin, G., Lages, J., Shepelyansky, D.L., 2015. Chaotic enhancement of dark matter density in binary systems. Astron. Astrophys. 576, A40. URL: http://dx.doi.org/10.1051/0004-6361/201425576, doi:10.1051/ 0004-6361/201425576. 20 Ross, S.D., Scheeres, D.J., 2007. Multiple Gravity Assists, Capture, and Escape in the Restricted Three-Body Problem. SIAM Journal on Applied Dynamical Systems 6, 576–596. doi:10.1137/060663374. Roy, A., Haddow, M., in a hard binary due to distant encounters. Celest. Mech. Dyn. Astron. 87, 411– 435. URL: http://dx.doi.org/10.1023/B:CELE.0000006767.34371.2f, doi:10.1023/B:CELE.0000006767.34371.2f. 2003. Energy change Scheeres, D., Ostro, S., Hudson, R., DeJong, E., Suzuki, S., 1998. Dynamics of Orbits Close to Asteroid 4179 Toutatis. Icarus 132, 53 – 79. URL: http://www.sciencedirect.com/science/article/pii/ S001910359795870X, doi:http://dx.doi.org/10.1006/icar.1997.5870. Scheeres, D.J., 2002. Stability of Binary Asteroids. Icarus 159, 271–283. doi:10. 1006/icar.2002.6908. Scheeres, D.J., Marzari, F., 2000. Temporary orbital capture of ejecta from comets and asteroids: Application to the Deep Impact experiment. Astron. Astrophys. 356, 747–756. Scheeres, D.J., Ostro, S.J., Hudson, R.S., Werner, R.A., 1996. Orbits Close to Asteroid 4769 Castalia. Icarus 121, 67–87. doi:10.1006/icar.1996.0072. Schleicher, D.G., Bair, A.N., Sackey, S., Stinnett, L.A.A., Williams, R.M.E., Smith-Konter, B.R., 2015. The evolving photometric lightcurve of Comet 1P/Halley's coma during the 1985/86 apparition. Astron. J. 150, 79. URL: http://stacks.iop.org/1538-3881/150/i=3/a=79. Shevchenko, I.I., 2007. On the Lyapunov exponents of the asteroidal motion subject to resonances and encounters, in: Valsecchi, G.B., Vokrouhlick´y, D., Milani, A. (Eds.), Near Earth Objects, our Celestial Neighbors: Opportunity and Risk (Proc. IAU Symp. 236). Cambridge Univ. Press, Cambridge, pp. 15-29., pp. 15–30. doi:10.1017/S174392130700302X. Shevchenko, I.I., 2011. The Kepler map in the three-body problem. New Astron- omy 16, 94–99. URL: http://www.sciencedirect.com/science/article/ pii/S1384107610000874, doi:10.1016/j.newast.2010.06.008. Shevchenko, I.I., 2014. Lyapunov exponents in resonance multiplets. Physics Letters A 378, 34–42. URL: http://www.sciencedirect.com/science/ article/pii/S0375960113009882, doi:10.1016/j.physleta.2013.10.035. Shevchenko, I.I., 2015. Chaotic zones around gravitating binaries. Astrophys. J. 799, 8. URL: http://stacks.iop.org/0004-637X/799/i=1/a=8. Soderblom, L., Boice, D., Britt, D., Brown, R., Buratti, B., Kirk, R., Lee, M., Nelson, R., Oberst, J., Sandel, B., Stern, S., Thomas, N., Yelle, R., 2004. Imaging Borrelly. Icarus 167, 4 – 15. URL: http://www.sciencedirect.com/ science/article/pii/S0019103503002665, doi:10.1016/j.icarus.2003. 07.008. 21 Stooke, P.J., Abergel, A., 1991. Morphology of the nucleus of Comet P/Halley. Astron. Astrophys. 248, 656–668. Thomas, P., A'Hearn, M.F., Veverka, J., Belton, M.J., Kissel, J., Klaasen, K.P., McFadden, L.A., Melosh, H.J., Schultz, P.H., Besse, S., Carcich, B.T., Farnham, T.L., Groussin, O., Hermalyn, B., Li, J.Y., Lindler, Shape, den- D.J., Lisse, C.M., Meech, K., Richardson, J.E., 2013. sity, and geology of the nucleus of Comet 103P/Hartley 2. Icarus 222, 550 – 558. URL: http://www.sciencedirect.com/science/article/pii/ S0019103512002163, doi:10.1016/j.icarus.2012.05.034. 22
1702.03315	1	1702	2017-02-10T20:12:11	Constraints on Climate and Habitability for Earth-like Exoplanets Determined from a General Circulation Model	[ "astro-ph.EP" ]	Conventional definitions of habitability require abundant liquid surface water to exist continuously over geologic timescales. Water in each of its thermodynamic phases interacts with solar and thermal radiation and is the cause for strong climatic feedbacks. Thus, assessments of the habitable zone require models to include a complete treatment of the hydrological cycle over geologic time. Here, we use the Community Atmosphere Model from the National Center for Atmospheric Research to study the evolution of climate for an Earth-like planet at constant CO2, under a wide range of stellar fluxes from F-, G-, and K-dwarf main sequence stars. Around each star we find four stable climate states defined by mutually exclusive global mean surface temperatures (Ts); snowball (Ts < 235 K), waterbelt (235 K < Ts < 250 K), temperate (275 K < Ts < 315 K), and moist greenhouse (Ts > 330 K). Each is separated by abrupt climatic transitions. Waterbelt, temperate, and cooler moist greenhouse climates can maintain open-ocean against both sea-ice albedo and hydrogen escape processes respectively, and thus constitute habitable worlds. We consider the warmest possible habitable planet as having Ts ~ 355 K, at which point diffusion limited water-loss could remove an Earth ocean in ~1 Gyr. Without long timescale regulation of non-condensable greenhouse species at Earth-like temperatures and pressures, such as CO2, habitability can be maintained for an upper limit of ~2.2, ~2.4 and ~4.7 Gyr around F-, G- and K-dwarf stars respectively due to main sequence brightening.	astro-ph.EP	astro-ph	Title: Constraints on Climate and Habitability for Earth-like Exoplanets Determined from a General Circulation Model. Short Title: Constraints on Climate and Habitability Authors: Eric T. Wolf1, Aomawa L. Shields 2,3,4,7, Ravi K. Kopparapu5,6,7,8, Jacob Haqq- Misra7,8, Owen B. Toon1 Affiliations: 1Laboratory for Atmospheric and Space Physics, Department of Atmospheric and Oceanic Sciences University of Colorado, Boulder, Colorado, USA. 2University of California, Irvine, Department of Physics and Astronomy, 4129 Frederick Reines Hall, Irvine CA 92697, USA 3University of California, Los Angeles, Department of Physics and Astronomy, Box 951547, Los Angeles, CA 90095, USA 4Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138, USA 5NASA Goddard Space Flight Center 6Department of Astronomy, University of Maryland 7NASA Astrobiology Institute's Virtual Planetary Laboratory, P.O. Box 351580, Seattle, WA 98195, USA 8Blue Marble Space Institute of Science, 1001 4th Ave, Suite 3201, Seattle, Washington 98154, USA Corresponding Author Information: Eric Wolf Laboratory for Atmospheric and Space Physics 3665 Discovery Drive Campus Box 600 University of Colorado Boulder, CO 80303-7820 [email protected] 240-461-8336 Abstract Conventional definitions of habitability require abundant liquid surface water to exist continuously over geologic timescales. Water in each of its thermodynamic phases interacts with solar and thermal radiation and is the cause for strong climatic feedbacks. Thus, assessments of the habitable zone require models to include a complete treatment of the hydrological cycle over geologic time. Here, we use the Community Atmosphere Model from the National Center for Atmospheric Research to study the evolution of climate for an Earth-like planet at constant CO2, under a wide range of stellar fluxes from F-, G-, and K-dwarf main sequence stars. Around each star we find four stable climate states defined by mutually exclusive global mean surface temperatures (Ts); snowball (Ts ≤ 235 K), waterbelt (235 K ≤ Ts ≤ 250 K), temperate (275 K ≤ Ts ≤ 315 K), and moist greenhouse (Ts ≥ 330 K). Each is separated by abrupt climatic transitions. Waterbelt, temperate, and cooler moist greenhouse climates can maintain open-ocean against both sea-ice albedo and hydrogen escape processes respectively, and thus constitute habitable worlds. We consider the warmest possible habitable planet as having Ts ~ 355 K, at which point diffusion limited water-loss could remove an Earth ocean in ~1 Gyr. Without long timescale regulation of non-condensable greenhouse species at Earth-like temperatures and pressures, such as CO2, habitability can be maintained for an upper limit of ~2.2, ~2.4 and ~4.7 Gyr around F-, G- and K-dwarf stars respectively due to main sequence brightening. 1. Introduction: Detecting Earth-like extrasolar planets is one of the primary objectives of ongoing and future exoplanetary observational surveys. Upcoming missions like the James Webb Space Telescope (Gardner et al. 2006), the Transiting Exoplanet Survey Satellite (Ricker et al. 2014), and others, will greatly improve our ability to detect and then begin characterizing terrestrial planets in the habitable zones of other stars. Still, observations will remain sparse compared to solar system objects and thus climate models are needed to interpret and understand these remote data. The conventional definition of the habitable zone requires liquid water to be extant and abundant at the surface continuously for at least several billion years in order for advanced life to evolve (Hart 1979). By definition, the fate of habitable worlds is inextricably tied to water and its associated feedbacks on the climate system. Thus, at its heart, the study of the habitable zone for Earth-like planets is the study of the fundamental evolutionary processes of water-rich terrestrial planetary atmospheres, touching upon end-member states that are characterized either by uncontrolled sea ice albedo or by water vapor greenhouse feedbacks. The sea ice albedo feedback can lead to rapid cooling whereupon the oceans become completely frozen over. Conversely, the water-vapor greenhouse feedback can lead to abrupt warming, water-rich atmospheres, and the total of loss of the oceans due to hydrodynamic escape or a thermal runaway. For many years, the leading descriptions of long timescale climatological evolution and the habitable zone have originated from energy balance and one-dimensional radiative-convective models (Budyko 1969; Hart 1979; Kasting et al. 1993; Selsis et al. 2007; Kopparapu et al. 2013). These works have shaped our thinking regarding the evolution of planetary climates. However, these models miss important feedbacks within the climate system caused by atmospheric dynamics, sea ice, clouds, and relative humidity. Only recently have three-dimensional (3-D) climate system models become commonly used to place limits on the habitable zone (Abe et al. 2011; Boschi et al. 2013; Leconte et al. 2013; Shields et al. 2013, 2014, 2016; Yang et al. 2013, 2014; Wolf and Toon, 2014, 2015; Godolt et al. 2015; Kopparapu et al. 2016; Popp et al. 2016). In both 3-D and lower-dimensional models, the first objective has been to study an Earth-like planet, as Earth has the only climate system that is well observed and Earth is the only confirmed habitable world. As a first step, modern 3-D climate systems models have been applied to study an Earth-like exoplanet as it evolves across the habitable zone due to changing stellar luminosity. 3-D models allow for a self-consistent treatment of water in the climate system, including water vapor, clouds, surface ice, and oceans, and their respective spatial and temporal distributions about the planet. The presence or absence of each phase of water significantly affects the radiative energy budget of the planet, and thus the climate. However, there remains uncertainty amongst different 3-D Earth climate system models with regard to clouds (Flato et al. 2013), convection (Del Genio et al. 2016) and radiative transfer (Yang et al. 2016). For instance, amongst leading climate models, the increase in global mean surface temperature of the Earth in response to a doubling of CO2 varies between 2.1 and 4.7 K (Andrews et al. 2012). Differences can become more significant for exoplanetary problems where the implied forcings tend to be larger (e.g. see Fig. 7a in Popp et al., 2016). We still have much to learn both scientifically and technically, as we apply our 3-D models to the new and exotic atmospheres of extrasolar planets. Here we present simulations from a state-of-the-art 3-D climate system model of an Earth-like planet with a fixed amount of CO2 in its atmosphere around F-, G-, and early K-dwarf main sequence stars, over a wide range of stellar fluxes. We do not consider M-dwarf star systems in this study. Habitable zone planets around these low- mass stars are likely to be tidally locked, which has a profound impact on planetary climate (Yang et al. 2013; Yang et al. 2014; Leconte et al. 2015; Kopparapu et al. 2016; Way et al. 2016). Planets in the habitable zone of F-, G-, and early K-dwarf stars, as studied here, fall outside the tidal locking radius and thus can rotate rapidly, like we observe for Earth and Mars. Venus, however, falls outside the tidal locking radius but is a slow rotator due to atmospheric tides. Here, we determine the climate of rapidly rotating terrestrial planets under varying stellar fluxes, and provide constraints on the habitable zone under fixed CO2 conditions. The interaction of atmospheric circulation, water vapor, clouds, and surface ice all play critical roles in modulating climate, and controlling sharp positive feedbacks. While earlier studies have similarly mapped climate as a function of stellar insolation using energy balance or 1-D models, to our knowledge this is the first such study to use an advanced, 3-D climate system model to attempt to map the entire range of habitable climates, complete from snowball to moist greenhouse, around numerous types of main sequence stars. 2. Methods Here we use the Community Atmosphere Model version 4 (CAM4) from the National Center for Atmospheric Research (Neale et al. 2010). We build upon the prior work of Shields et al. (2013; 2014) and Wolf & Toon (2015) with new and complimentary simulations, facilitating a comprehensive description of the evolution of Earth-like climate through the habitable zones of F-, G-, and K-dwarf stars. Simulations of warm climates (i.e. those approaching a moist greenhouse) and cold climates (i.e. those approaching a snowball glaciation) follow the specific modeling methods described in detail in Wolf and Toon (2015) and Shields et al. (2013, 2014) respectively. Different setups for warm and cold simulations sets are used in order to combine new simulations with prior simulations of Wolf and Toon (2015) and Shields et al. (2013, 2014), saving considerable computational expense. In total, we use data from ~45 previous simulations from Shields et al. (2013; 2014) and Wolf & Toon (2015), and ~45 new simulations, yielding a complete picture of habitable zone climates. There are some differences in the specific configuration of CAM4 used for warm and cold simulations sets, including resolution, ocean heat transport, land area assumptions, and the radiative transfer module used in the calculation (see Table 1). Of importance, cold simulations assume zero ocean heat transport and a global ocean, which allows for a transition to snowball Earth to occur at higher stellar fluxes than if some ocean heat transport is included (Poulsen et al. 2001; Pierrehumbert et al. 2011). Simulations approaching a moist greenhouse assume present day Earth continents and present day ocean heat transport. Cold simulations use the native CAM radiative transfer scheme found in CAM versions 4 and earlier (Ramanthan & Downey, 1986; Briegleb, 1992), while warm simulations use a correlated-k radiative transfer scheme (Wolf & Toon, 2013). Both configurations use identical atmosphere, ocean, and sea ice physics (Neale et al. 2010). However, despite these noted differences, results from each configuration are well in agreement where simulations overlap, for conditions near present day Earth surface temperatures (see Sections 3.1 and Fig. 2). All simulations assume an Earth-like planet, with Earth's mass and radius, a 50- meter deep mixed-layer thermodynamic ("slab") ocean, and a 1-bar N2 atmosphere with present day CO2 concentrations but no ozone. Prognostic bulk microphysical parameterizations for condensation, precipitation, and evaporation control atmospheric water vapor, liquid cloud, and ice cloud condensate fields (Rasch and Kristjánsson, 1998). Deep convection (i.e., moist penetrative) is treated using the parameterization of Zhang and McFarlane (1995). Shallow convective overturning is treated by the parameterization of Hack (1994). We use stellar spectra from F2V (σ Bootis HD 128167), G2V (the Sun,), and K2V (ε Eridani HD 22049 and synthetic) stars (Segura et al. 2003; Pickles, 1998). In Fig. 1, we show F-, G-, and K- dwarf spectra normalized to 1360 W m-2. These stars have effective temperatures (Teff) of 6954 K, 5778 K, and 5084 K. In general F-, G-, and K-dwarf stellar classifications span 6,000 ≤ Teff ≤ 7500 K, 5200 ≤ Teff ≤ 6000 K, and 3,700≤ Teff ≤ 5200 K respectively (Habets & Heintze, 1981). We do not consider photochemistry here, however the stellar type, stellar activity, and atmospheric oxygen concentration are all critical for determining the UV radiation hazard for the planet surface, and have been studied elsewhere (Segura et al. 2003; Rugheimer et al. 2015). For each planet, we assume Earth-like orbital characteristics, with an orbital period of 1 Earth year, zero eccentricity, and 23.5° obliquity. Differences in the semi- major axis and orbital period that arise due to the respective mass and luminosity of each star are not incorporated. However, changes to the orbital periods do not appreciably affect global mean climate for rapidly rotating planets within the habitable zones of F-, G- and early K-dwarf stars (Godolt et al. 2015). For this study we assumed a 24-hour rotation rate for all simulations. This is reasonable, as habitable zones studied here are outside the tidal spin-down region of their respective host stars (Leconte et al. 2015). 3. Results 3.1 Control Simulations First, we compare a set of standard atmospheres around F-, G-, and K-dwarf stars, using the model default present day Earth solar insolation of 1361.27 W m-2 (Table 2). In these simulations we assume the warm model configuration (Table 1), with a continental configuration and implied ocean heat transport that matches the present day Earth. The spectra emitted from cooler stars is relatively redder (Fig. 1), and thus interacts more strongly with atmospheric clouds, CO2 and water vapor, as well as surface water ice and snow. Surface water ice (Dunkle & Bevans, 1956), as well as atmospheric clouds, CO2 and water vapor (Kasting et al. 1993, Selsis et al. 2007) strongly absorb in the near- infrared. The end result is that the spectrally averaged all-sky planetary albedo is lower for an Earth-like planet around cooler stars (Table 2). Thus it takes less stellar flux to warm an Earth-like planet around cooler stars. This albedo change is well established from simpler models and our results concur (Kasting et al. 1993; Kopparapu et al, 2013; Yang et al. 2014). The clear-sky (i.e. without clouds) albedo is reduced by a factor of ~2 between F- and K-dwarf control cases, owing to decreased Rayleigh scattering and increased overlap of the incident stellar spectra with near-infrared water vapor absorption bands. Meanwhile, the change in cloud albedo is relatively small across each different stellar type, varying by only about ~6% (analogously a cloud albedo change of ~0.01). Nonetheless, in each control simulation the overall climate is not radically changed due to altering only the stellar spectra but not the stellar energy input (Table 2). Climate does not switch states, remaining (generally) like the present-day Earth, dominated by open ocean but with some sea ice at the poles. The strength of the greenhouse effect and the global mean cloud fractions remain fairly similar for each control case. Note the greenhouse effect is given in units of temperature in order to facilitate comparison with Godolt et al. (2015), and is calculated using the Stefan- Boltzmann law from the difference in upwelling longwave radiation between the surface and the top-of-the-atmosphere implied by clouds and absorbing gases respectively. Differences in water vapor, cloud water, and sea-ice fractions come as no surprise, as these quantities are strongly dependent on the planetary temperature. Differences in the evolution of sea-ice, convection, clouds, and the distribution of relative humidity are significant drivers of planetary climate, and serve to amplify initial radiative perturbations due to albedo changes that are implied by changing the spectral energy distribution. For the F-dwarf star, cooling caused initially by increased atmospheric scattering and surface reflectivity is then amplified by the sea-ice albedo feedback. For the K-dwarf star, warming caused initially by increased absorption by water vapor and the surface is then amplified by the water vapor greenhouse, and is also linked to cloud feedbacks. Interestingly, our results shown in Table 2, exhibit remarkably less sensitivity to changes in the stellar spectra compared to a similar set of simulations conducted with the 5th version of the European Center Hamburg GCM (ECHAM5) from Godolt et al. (2015). Irradiated by the same F- and K-dwarf spectra, global mean surface 2-meter air temperatures reach 280.1 K and 294.1 K in CAM4, while in ECHAM5 they reach 273.6 K and 334.9 K respectively. Note that Popp et al. (2016) similarly find that ECHAM is significantly more sensitive than CAM4 to increasing stellar flux from the Sun. The striking differences in climate between these simulations highlights the need for detailed model intercomparison to ascertain why the models diverge. Without further work, we cannot determine if the model differences arise purely from differences in radiative transfer, or whether sea-ice and cloud feedbacks are more to blame. For simulations near the present day surface temperatures, model configurations for warm and cold simulations (see Table 1) yield very similar results, with mean surface temperatures generally within ~2 K for temperate conditions (see Fig. 2). Under the present day stellar flux from the Sun, a G-dwarf star, the cold (warm) configurations yield a global mean surface temperature of 287.3 K (289.0 K). Under insolation from an F-dwarf star, a 5% increase in the solar constant above the present day is required to reach approximately modern day surface temperatures of 288.4 K (287.1 K) for the cold (warm) configuration. For the K-dwarf star case, a 2% reduction in the solar constant yields 290.3 K (291.6 K) for the cold (warm) configurations. However, at colder temperatures differences emerge between the warm and cold configurations due primarily to ocean heat transport , which is assumed not to occur in the cold configuration. At the present day solar flux, F-dwarf simulations become cold (241.6 K) when no ocean heat transport is included. This behavior in response to turning off ocean heat transport is in agreement with Godolt et al. (2015). 3.2 Multiple climate states Fig. 2 shows the evolution of global mean surface temperature (Ts) and climate sensitivity for an Earth-like planet with fixed CO2 as a function of the relative stellar flux from F-, G-, and K-dwarf stars. Relative stellar flux is defined as the ratio between the incident stellar flux on the planet, and that received by Earth at present day (S0), taken here to be 1360 W m-2. Thus in Fig. 2 a relative stellar flux (S/S0) of 1.0 equals an incident stellar flux on the planet of 1360 W m-2, approximately matching the present day Earth value of 1361 W m-2. Four stable climatic regimes are indicated by shaded regions in Fig. 2; snowball (Ts ≤ 235 K), waterbelt (235 K ≤ Ts ≤ 250 K), temperate (275 K ≤ Ts ≤ 315 K) and moist greenhouse (Ts ≥ 330 K). Stable climates are in equilibrium, having balanced incoming and outgoing radiation and have no systematic temperature drift. The evolution of climate around each type of star is qualitatively similar. Note that here we consider the moist greenhouse climate state to be defined by the radiative-convective state of the atmosphere, as described by Wolf and Toon (2015). Wolf & Toon (2015) found that the climate undergoes an abrupt transition between temperate and moist greenhouse states, characterized by the closing of the 8 to 12 µm water vapor window region, increased solar absorption in the near-infrared water vapor bands, and the subsequent stabilization of the low atmosphere against convection. We discuss water- loss rates versus the depletion of the ocean inventory separately below. Under a relaxed constraint, where any amount of surface water may constitute a habitable planet, waterbelt, temperate and cooler moist greenhouse states are habitable. However, in our surveys of the stars we seek not just to find habitable zone planets, but to find those that are preferably within the temperate climate regime, where the vast majority of the planet is ice-free and temperatures are similar to the Earth presently. The four stable climate regimes are separated by sharp climatic transitions, indicated by maxima in climate sensitivity (numbered in Fig. 2d, e, f). Climate sensitivity is the change in global mean surface temperature for a given change in radiative forcing, here from incrementally increasing or decreasing the incident stellar flux. The sharp climatic transitions are triggered by interactions between sea-ice, water vapor, as well as clouds and radiation, whereupon a small change in solar forcing can beget a large change in Ts. Interestingly, for some temperature ranges, 250 K ≤ Ts ≤ 275 K and 315 K ≤ Ts ≤ 330 K, stable climate states never occur in our model. These unallowable temperature regions are caused by uncontrolled sea-ice albedo, and uncontrolled water-vapor greenhouse feedbacks respectively. Each cause climate sensitivity to spike (Fig. 2). If 250 K ≤ Ts ≤ 275 K, climatic stability can be re- established either by warming into the temperate state where sea-ice is trapped at the poles, or by cooling into the waterbelt state which is stabilized by albedo contrasts between bare sea ice and snow covered areas formed when global ice sheets encroach into the subtropical desert zone (e.g. Abbot et al. 2011). If 315 K ≤ Ts ≤ 330 K, climate stability can be re-established either by cooling into the temperate state where the planet's surface can efficiently cool to space through the water vapor window region, or by warming into the moist greenhouse state which is stabilized by a reduction to relative humidity and the formation of an upper atmosphere cloud deck (Wolf & Toon, 2015). While these unallowable regions appear similarly for an Earth-like planet around each star, it is unclear how sensitive these regions may be to other choices of model parameters. Earth presently exists in a region where two climate states are possible. In Earth's present temperate state (perhaps fortuitously) climate sensitivity is near a minimum against both positive and negative radiative forcings (Fig. 2e). The relative long-term stability of Earth's climate may be circumstantial evidence that terrestrial climates preferentially relax towards climate sensitivity minima. As shown in Fig. 2 and Shields et al. (2014), there is strong hysteresis between the snowball and temperate climate states, which has long been recognized from simple climate models (Budyko, 1967). The solar constant must be raised to high levels to escape from a snowball, but at much lower solar constants temperate climate states are stable, but only if the climate was initially warm. Thus there is a range of solar fluxes at which climate exhibits bistability, with both snowball or temperate and waterbelt states being possibly stable at given stellar fluxes. The actual state depends on the initial conditions, and thus the planet's evolutionary history. For planets around F-, G-, and K-dwarf stars, Earth-like planets exhibit bistability for relative stellar fluxes (S/S0) of 0.99 – 1.14, 0.92 – 1.06, and 0.88 – 0.98 respectively. Note that the extent of the bistable region is encouragingly quite similar to that found from 3-D models of intermediate complexity (Lucarini et al. 2010; Boschi et al. 2013), which found a bistable range of 0.93 – 1.05 for Earth around the Sun. Bistability is not found in our calculations between temperate and the moist greenhouse climate states. However, Popp et al. (2016) find evidence of a small (i.e. contained in a ~2% change in stellar flux) hysteresis between temperate and moist greenhouse climates using an idealized version of the ECHAM6 climate model. The runaway greenhouse provides the most generous bounds for the inner edge of the habitable zone for an Earth-like planet. Kasting et al. (1993) also defined the more conservative "moist greenhouse" inner edge of the habitable zone based on water-loss to space from moist atmospheres, which may occur at lower stellar fluxes than are needed to induce a thermal runaway. Kasting et al. (1993) define this inner edge constraint to occur where diffusion limited escape causes the entirety of Earth's oceans (1.4 ×1024 g H2O) to be lost to space in a period of time that approaches the present age of the Earth. In practical terms, this threshold is reached when the stratospheric H2O volume mixing ratio equals 3x10-3. Here we adopt a marginally more stringent constraint, assuming that an Earth ocean of water must be lost within ~1 Gyr. This ensures that the water-loss timescale is meaningful even for F-dwarf stars, which live little more than half as long as our Sun (Rushby et al. 2013). This escape rate occurs when the stratospheric water vapor volume mixing ratio exceeds ~7×10-3, which occurs in our model when Ts ~ 355 K around each star (Fig. 2). However, note that Earth-like planets with Ts ~ 350 K will desiccate within about ~8 Gyr, which is near the main sequence lifetime of G-dwarf stars and significantly less than that of K-dwarf stars. Fig. 3 shows the model top (~0.2 mb) temperature and water vapor mixing ratio as a function of mean surface temperature, as climate warms around each star. The temperature controls the amount of water vapor in the upper atmosphere. At low Ts the upper atmosphere is noticeably warmer around redder stars due to the effect of increased absorption by water vapor in the near-infrared, coupled with inefficient radiative cooling aloft. However, this trend becomes muted for increasing Ts as the atmospheres become increasingly water-rich, thermally opaque, and convective to high-altitudes (Wolf & Toon, 2015). The time-scale to lose Earth's oceans falls off dramatically as the mean surface temperature increases. By the time Ts ~ 360 K, the oceans would be lost in only several hundred million years. Such atmospheres would be short-lived relative to stellar lifetimes, and would transition into dry planets (Abe et al. 2011). However, relatively cooler moist greenhouse atmospheres, with 330 K ≤ Ts ≤ 350, have upper atmosphere water vapor volume mixing ratios between ~10-6 and 10-3, and thus can retain an Earth ocean of water for tens to hundreds of billion of years. More detailed hydrodynamic escape calculations from these atmospheres would better our understanding of habitable lifetimes of planets near the inner edge of the habitable zone. Furthermore, on other planets hydrogen escape rates may vary due to factors we have not considered such as differences in the exobase temperature, the mean molecular weight of the atmosphere, the gravitational force, stellar activity, and photochemistry. 3.3 Circumstellar Climate Zones Based on the climate results shown in Fig. 2, we can define circumstellar climate zones for Earth-like planets at modern-day CO2 levels around F-, G-, and K-dwarf stars (Fig. 4). Here, circumstellar climate zones provide a more detailed description of habitable planetary climates than does the habitable zone. The habitable zone is based on the existence of liquid water, but does not take into account such issues as the ability of organisms to survive at a given temperature, or the climatological history of the planet. Significant differences in climate zones exist depending upon whether one assumes warm (i.e. ocean covered, Fig. 4a) or cold (i.e. completely ice covered, Fig. 4b) initial conditions. Following the practice of some recent habitable zone studies (Selsis et al. 2007; Zsom et al. 2013; Kopparapu et al. 2013, 2014; Yang et al. 2014), based on our climate modeling results we determine parametric relationships between the relative star. Equations 1 and 2 are valid for stars with Teff between 4900 K and 6600 K, 1) 𝑆!"#$%&'=𝑎+𝑏𝑇∗+𝑐𝑇∗! 2) 𝑇∗=𝑇!""−5780 K incident stellar flux received by an Earth-like planet at modern CO2 that yields a given climate state (Sclimate) noted in Fig. 4, and the stellar effective temperature (Teff) of the host where coefficients a, b, and c, are given in Table 3 for each particular climate. The results of Sclimate using Table 3 describe circumclimate zones that are plotted and labeled in Fig. 4. All circumstellar climate zones found in Fig. 2, are shown in Fig. 4 as shaded regions bounded by solid lines. Additional climate states of interest are indicated by dashed lines. First, moving from left to right in Fig. 4a (from high Ts and high solar flux), we delineate with a dashed line the inner edge of the habitable zone according to the ~1 Gyr water-loss criteria (Ts ~ 355 K) discussed in section 3.1. Next we mark with a solid orange line the radiative-convective transition from a moist greenhouse climate state into a temperate state, as described by Wolf and Toon (2015), whereupon Ts abruptly drops from ~330 K to ~315 K. Next we mark the heat stress limit for mammalian biological functioning. Sherwood and Huber (2010) argue that if Ts ≥ 300 K, then the majority of the Earth's human population would be subject to prolonged periods of lethal heat stress. While technological and biological adaption may facilitate survival at these hotter temperatures, life and society as we know them would be threatened. Next, we mark the climatic conditions of the present day Earth (Ts ~289 K) around each star. In Fig. 4a we define with a solid dark blue line, the boundary where an Earth-analog planet with identical CO2 would transition from the temperate regime into a waterbelt (i.e. ice lines reaching the tropics), a condition only accessible via cooling from a warmer state. Finally, we define with a solid light blue line the transition into a snowball Earth state via reduced solar forcing. If formed initially warm, a planet may access a temperate zone with ΔS/S0 ~ 0.16 − 0.20 (equivalently a ~218 – 272 W m-2 change in total solar insolation received by the planet) as marked by the green shaded region in Fig. 4a. Here, ΔS/S0 is the width of the climate zone in units of relative stellar flux and varies for planets around different effective temperature stars. The habitable zone, including the waterbelt state (for warm start only) up to the water-loss threshold, spans ΔS/S0 ~ 0.24 − 0.34 (~326 – 462 W m-2) for initially warm planets. In the absence of CO2 changes, if a planet is initially cold (i.e. fully ice covered), its habitable zone is constrained by the solar deglaciation and water-loss limits indicated in Fig. 4b, and there is a relatively small range of possible habitable states. Note that a present day Earth-like climate cannot be accessed from a cold start (Fig. 4b, see also Fig. 2). Moving from right to left (from low Ts and low solar flux), solar driven deglaciation is indicated by a light blue line in Fig. 4b. The temperate climate zone is significantly narrower with 290 K ≤ Ts ≤ 315 K, and the waterbelt state is skipped entirely from a cold start. There is no difference between the inner edge of the habitable zone for cold and warm start cases. By the time that climate has warmed to moist greenhouse and water- loss thresholds, snow and ice have long since vanished from the planet, and no memory of the cold initial conditions remain. Thus, the temperate climate regime spans only ΔS/S0 ~ 0.06 − 0.11 (~82 – 150 W m-2) as marked by the green shaded region in Fig. 4b. The full width of the habitable zone for initially cold planets is marked from solar deglaciation to water-loss limits, and spans ΔS/S0 ~ 0.14 − 0.19 (~190 – 258 W m-2). It is interesting to note that the habitable zone is wider for a cold initial planet around a K- dwarf star. This is because solar driven deglaciation is more effective around relatively redder stars, due to the low near-infrared albedo of snow and ice (Joshi & Haberle 2012; Shields et al. 2014). 3.4 Habitable lifetimes Another consideration in Fig. 2 and Fig. 4 is the time that is spent in the habitable zone. Main sequence stars brighten over the course of their lifetime, thus the radiation received by an orbiting planet increases over time (Iben 1967). However, the main sequence lifetime and rate of luminosity evolution depends upon the stellar type. If we assume that the rate of luminosity evolution is linear in time, we can then make simple estimates for the lifetime of climate zones (τclimate), calculated as the time needed for the stellar luminosity to evolve corresponding to the maximum width of each climate zone in terms of the relative stellar flux (ΔS/S0) shown in Fig. 4. We use equation 3: 3) 𝜏!"#$%&'= ∆!/!! !!"#!! !!!"#$!! ×𝜏!" where Ltms/L0 is the luminosity at the end of the star's main sequence lifetime normalized to the present solar luminosity, Lzams/L0 is the zero age main sequence luminosity normalized to the present solar luminosity, and 𝜏!" is the total main sequence lifetime of the star. Ltms/L0 and Lzams/L0 are calculated using the parametric fits given by equations 3 and 5 in Guo et al. (2009). Following Rushby et al. (2013, equations 7 and 8), the total main sequence lifetime can be computed as: 4) 𝜏!"= 10.9∗!!!! where M is the mass of the star, M0 is the mass of the Sun, and 10.9 is the main sequence lifetime of our Sun in billions of years. Our calculation of τclimate assumes that the planet initially has the lowest temperature allowed for a given climate zone, and is then warmed via main sequence brightening through the climate zone. Thus τclimate is the maximum length of time that an Earth-like planet, at fixed CO2, could remain within a given climate zone under the influence of the main sequence brightening. Beginning its life at a higher temperature would decrease τclimate, while a draw down of CO2 could possibly lengthen τclimate. In Fig. 5 we consider τclimate for the temperate climate zone, and for the habitable zone in total, including waterbelt and moist greenhouse states with Ts below 355 K. Viewed in this fashion it is clear that lower effective temperature stars (i.e. the K-dwarf) provide a more stable climatic environment because τclimate for K-dwarfs is about double that of the F- and G-dwarf stars. While habitable planets around K-dwarf stars are more sensitive to changes in the stellar flux than planets around F- and G-dwarfs (Fig. 2a,d), their long main sequence lifetimes (~19.8 Gyr for ε Eridani for instance) and thus slower temporal luminosity evolution across the main sequence bestows a significant advantage for evolution of life. The most optimal scenario is for initially warm planets. Terrestrial planets are believed to have been formed hot from accretion, and with an initially molten surface before their earliest atmospheres cooled and condensed. Thus, even though stellar luminosity increases in time, terrestrial planets likely begin their earliest histories in a hot state, and could then access a wider temperate zone, and undergo waterbelt states upon first cooling (Fig. 4). Still, a waterbelt state may be a low probability occurrence due to the narrow range of allowable stellar fluxes. Alternatively, waterbelt states could also occur if a planet formed with a larger primordial CO2 inventory than assumed here, which is then drawn down over time by weathering processes, allowing the planet to cool (Abbot et al. 2011). The maximum time spent in the habitable zone is ~2.2, ~2.4, and ~4.7 Gyr for F-, G-, and K-dwarf planets respectively, possible only for warm start scenarios. Note that life has existed on Earth for at least 3.8 gyr (Nisbet & Sleep, 2001), significantly longer than the 2.4 Gyr lifetime noted here for a G-dwarf star. Long-lived habitable conditions for Earth are most likely due to a stronger CO2 greenhouse early in Earth's history. See section 4 for more discussion. Cold start cases apply to initially frozen worlds subject to increased stellar fluxes whether by stellar evolution, or possibly planetary migration. For cold initial conditions, temperate climate states around F- and G-dwarf stars may last only ~500 Myr, if a draw down of CO2 is not invoked to mitigate warming. Interestingly, the habitable and temperate climate zone lifetimes are only about ~10 − 20% shorter for F-dwarf planets compared to the G-dwarf, despite the F-dwarf having a total main sequence lifetime that is ~40% shorter (6.4 Gyr versus 10.9 Gyr). Temperate Earth-like planets around F-dwarf stars benefit from their bluer stellar spectra, which is more effectively scattered and less readily absorbed by the near-IR water vapor bands, thus allowing for a temperate climate zone that exists under a wider range of relative stellar fluxes (Fig. 2 and Fig. 4). Thus we should not ignore habitable zone planets around F- dwarf stars, due to their muted climate sensitivity. 4. Discussion In this study we have fixed CO2 at present day values, thus Fig. 2 and Fig. 4 illustrate a relative range of habitable climates for a single choice for non-condensable greenhouse species. Couched in these terms, the habitable zone appears quite narrow for both warm and cold start scenarios. The wide range of stellar fluxes possible for the conventional habitable zone, as described in Kopparapu et al. (2013), is reliant on the strong greenhouse effect from multi-bar CO2 atmospheres to set the outer edge of the habitable zone. It is conventionally thought that CO2 can change on many planets and provide a stabilizing feedback to the climate system through the temperature dependent action of the carbon/silicate cycle. If the temperature increases for some reason, weathering rates will increase, leading to removal of CO2 from the atmosphere/ocean system and sequestration into sea floor carbonates such as limestone, which cools the planet by weakening the total greenhouse effect. Subduction of carbonate-rich sea floor on planets with plate tectonics and subsequent volcanic outgassing recycles the sea floor carbonates and supplies CO2 to the atmosphere to balance the loss from weathering over geologically long periods of time. Of course, silicate weathering requires the presence of continents and plate tectonics, both of which are uncertain. However, if a planet has weak volcanic outgassing, perhaps in tandem with high rates of formation of seafloor carbonates, a cold planet may not be able accumulate sufficient CO2 to deglaciate the planet independently of the stellar flux. Mars, for instance, currently resides within the habitable zone of Kopparapu et al. (2013), but it was unable to retain sufficient CO2, to escape from its present frozen state. Kadoya and Tajika (2015) and Haqq-Misra et al. (2016) argue for Earth that if the paleo-CO2 outgassing rates were less than on Earth presently, carbon/silicate cycles may not have been able to prevent a snowball glaciation for most of Earth's history. Furthermore, planets entering a snowball phase may oscillate between frozen and thawed states, with a frequency dependent on the rate of outgassing (Tajika, 2007; Mills et al., 2011; Driscoll & Bercovici, 2013; Haqq-Misra et al., 2016). Note while the early Earth was indeed habitable for nearly its entire existence, implying higher CO2 outgassing rates in the distant past to sustain a warm climate, there is geological evidence for periodic snowball glaciations up through the Neoproterozoic period, with the last being ~635 Myr ago, when the solar constant was at ~94% of its present day value (Kirschvink et al. 2000; Pierrehumbert et al. 2011). Interestingly, while simple life emerged billions of years earlier, complex life did not emerge until the quasi-periodic snowball events ceased (Hoffman et al., 1998). While here we emphasize CO2 and carbon/silicate cycles, there could be numerous other gases and particles in planetary atmospheres that could impact their climate that we have not considered, including N2, H2, CH4, organic hazes, various sulfur compounds, and a host of others. In Fig. 6 we summarize habitable zone calculations published to date using 3-D climate models of Earth-like planets, along with the widely used values of Kopparapu et al. (2013; 2014) that are based on 1-D radiative-convective model calculations. Fig. 6a shows constraints on the inner edge of the habitable zone derived from climate models with solid lines and diamonds. In theory, the inner edge of the habitable zone for a water-rich planet should not involve high levels of CO2. As the climate warms, enhanced silicate weathering should draw down CO2 to relatively low levels. Each simulation shown assumes a CO2 mixing ratio equal to the modern Earth, along with 1 bar N2 as the background gas. Numerous studies have calculated the inner edge of the habitable zone for Earth around the Sun (Teff = 5778 K) using nationally supported 3-D climate system models (Abe et al. 2011; Leconte et al. 2013; Wolf &Toon, 2014, 2015; Yang et al. 2014; Popp et al. 2016). The models vary in the location of the inner edge of the habitable zone by ~15% (S/S0). The models also differ in the predicted end state of the atmosphere. Leconte et al. (2013), using the LMD generic climate model, predict Earth will enter a runaway greenhouse and no moist greenhouse state is possible. On the contrary both CAM (Wolf & Toon, 2015) and ECHAM (Popp et al. 2016) models predict that a climatologically stable moist greenhouse state with significant water-loss marks the inner edge of the habitable zone for Earth. Note that the results from this study using a modified version of CAM4 appear quite similar to those of Yang et al. (2014), who used CAM3 (marked "fast" on Fig. 6). However, the similarity between the two results is somewhat deceiving. Yang et al. (2014) derived the inner edge of the habitable zone at a point where the model becomes numerically unstable (Ts ~ 310 K). Here numerical improvements to the convection scheme allow us to simulate much hotter temperatures, and define a true inner edge of the habitable zone by water-loss from moist greenhouse atmospheres. Note also that the position of the inner edge determined by Yang et al. (2014) and Wolf & Toon (2015) is also affected by differing properties of the radiative transfer model used (see Yang et al. 2016). In all panels of Fig. 6 light yellow and light blue shaded regions indicate regions in space that are inside and outside of the tidal locking radius respectively, following Edson et al. (2011) and Haqq-Misra & Kopparapu (2015). Tidal locking is expected to be most important for planets towards the inner edge of the habitable zone around low mass stars. Planets located within the tidal locking radius are strongly influenced by the host star gravity, and should exist in synchronous or resonant orbital-rotational configurations. Although, tidal spin down may also be dependent upon atmospheric thickness and thermal tides (Leconte et al. 2015). Planets located in the blue region are unconstrained and can rotate rapidly as does Earth and Mars. There is a large difference between constraints for the inner edge of the habitable zone for rapidly rotating planets (Yang et al. 2014; Leconte et al. 2013; Godolt et al. 2015; Wolf and Toon, 2015) and tidally locked planets, which tend to be more slowly rotating (Yang et al. 2014; Kopparapu et al. 2016). As first described by Yang et al. (2013), strong convection at the substellar point of slow rotators can create thick clouds that substantially raise the planet's albedo, thus allowing liquid surface water to exist under large stellar fluxes. Shown here are calculations from Yang et al. (2014) using CAM3 which assumed a fixed orbital-rotational period of 60 Earth days (marked "slow" on Fig. 6a). However, in reality the orbital-rotation period and the stellar flux received by a tidally locked planet are dependent on the mass and luminosity of the stars, and should vary between ~10 and 65 days for planets near the inner edge of the habitable zone around late-K and early M stars. A subsequent revision of the inner edge of the habitable zone for these slow rotators was published by Kopparapu et al. (2016) using CAM4 and self-consistent orbital periods for both high and low metallicity stars. Changing the planetary rotation rate self-consistently has important consequences for atmospheric dynamics, cloud fields, and ultimately the global mean albedo and climate. Seamlessly connecting the habitable zones for rapid and slow rotators is not trivial. Near Teff ~ 4500 K, the inner edge of the habitable zone for rapid rotators approaches the tidal locking radius. Here one may find a transition region, between the fast and slow rotator limits, dependent upon planet-star tidal interactions and the precise rotation rate of the planet in question. It is clear from Fig. 6a that the evolution of climate may critically depend on the evolution of the planetary rotation rate due to tidal interactions with the host star and also moons. In Fig. 6b we summarize constraints for the outer edge of the habitable zone and for habitable climates at low stellar fluxes and high-CO2, with dashed lines and triangles. We include our estimates for the outer edge of the habitable zone based on snowball glaciation and solar driven deglaciation limits for an Earth-like planet with modern CO2. Note in Fig. 6b, the "snowball" limit marks the transition between habitable and globally ice covered states due to decreasing solar insolation (see also Fig. 4a). The "deglaciation" limit marks the transition between globally ice covered and habitable states triggered by increasing solar insolation (see also Fig. 4b). Kasting et al. (1993) first postulated the so-called maximum CO2 greenhouse limit for the outer edge of the habitable zone using a 1D radiative convective model. This limit has recently been revised for initially warm (Kopparapu et al. 2013), and initially ice covered planets (Haqq-Misra et al. 2016), using similar methodology with a 1D model. To date no 3-D simulations have calculated the maximum CO2 greenhouse limit. However, several 3D studies have explored paleo-Earth, paleo-Mars, and high-CO2 exoplanets scenarios, which may serve as useful steps towards determining the outer edge of the habitable zone around different stars (Wordsworth et al. 2011; Urata & Toon, 2013; Forget et al. 2013; Wordsworth et al. 2013; Wolf & Toon, 2014; Shields et al. 2016). Several of these data points are included on Fig. 6b along with the CO2 burden required to yield a habitable climate. More work is needed in defining the maximum-CO2 greenhouse for terrestrial planets using 3-D models. Finally, Fig. 6c combines habitability constraints from Fig 6a and 6b onto the same plot. It is clear that without the ability to regulate CO2 or other greenhouse gases, the habitable zone for rapidly rotating planets is quite narrow. Still, the effect of slowing rotation as an equally large effect in widening the total habitable zone. Information regarding planetary rotation rate, CO2 cycling, and the ability of a planet to retain its atmosphere against escape are equally as important for determining habitability, as is the incident stellar flux. 4. Conclusions Here we have used a 3-D climate system model to explore the effect of changing stellar fluxes on climate for an Earth-like exoplanet around F-, G-, and early K-dwarf main sequence stars, assuming a fixed amount of CO2. For these stars, the inner edge of the habitable zone lies beyond the tidal-locking radius, and thus planets are free to maintain a rapid rotation rate. Earth-like planets in the habitable zone are subject to four stable climate states (snowball, waterbelt, temperate, and moist greenhouse), each separated by sharp climatic transitions which are triggered by the changing thermodynamic partitioning of water in the climate system. Without allowing for the build up or removal of non-condensable greenhouse gas species such as CO2, the range of relative stellar fluxes that permit temperate climates (i.e. 275 ≤ Ts ≤ 315 K) is quite narrow; ΔS/S0 ~ 0.06 − 0.11 for initially frozen planets, and ΔS/S0 ~ 0.16 − 0.20 for initially warm planets. The range of habitable climates becomes marginally wider if we generously allow both waterbelt and cooler moist greenhouse climates (Ts ≤ 330 K) to be included as habitable worlds; ΔS/S0 ~ 0.14 − 0.19 for initially cold planets, and ΔS/S0 ~ 0.24 − 0.34 for initially warm planets. For cold initial conditions, planets around K- dwarf stars have the widest habitable zones due to deglaciation of sea-ice at lower stellar fluxes. For warm initial conditions, planets around F-dwarf stars have the widest habitable zones due primarily to a muted climate sensitivity across a broad temperate climate zone. These variations in the solar constant represent only a small fraction of the change in solar constant over the host star's main sequence lifetime. Amongst our studied systems, the K-dwarf stars provides the longest lived habitable climates due to their lengthy main sequence lifetimes, and thus relatively slow rate of main sequence brightening. The reader is also reminded that results presented in this work in Fig. 2, Fig. 4, and Fig. 6 are derived from a single three-dimensional climate system model. Differences in radiative transfer, convection, clouds, ocean heat transport, sea ice, and other processes can vary across different 3-D models, and can lead to significant differences in the resultant climates. Furthermore, we have only studied two parameters (stellar flux and spectrum) in detail. The computational expense of modern climate models requires a focused approach to the study of parameter spaces relevant to habitable extrasolar planets. Lower dimensional models retain significant value by allowing multi- dimensional parameter sweeps with relative ease. The standard approximation taken here of the proverbial Earth-like exoplanet is now well worn. We hope these simulations mark an appropriate starting point for intercomparison amongst current climate models for Earth-like planets around various stars, before continuing towards habitable planets having more exotic characteristics. Model intercomparison is needed to constrain the origin of differences found in various simulations some of which are noted in Fig. 6. The differences might arise from different model parameterizations of radiative transfer, clouds, convection, large-scale dynamics, or some other process. While much effort in the literature has been given to defining the effect of the solar constant on habitable climates, it is clear that the geological and/or biological regulation of non-condensable greenhouse species is of equal or possibly greater importance to planetary habitability, by allowing the habitable zone to be extended much further away from the host star. Lastly, to complete our picture of the inner edge of the habitable zone, future work might focus on the 4500 K to 5000 K effective temperature regime, where the inner edge of the habitable zone for rapid rotators approaches the tidal- locking radius. Acknowledgements. E.T.W. and O.B.T. acknowledge support from NASA Planetary Atmospheres Program award NNX14AH17G. A.S. acknowledges support from the National Science Foundation under Award No. 1401554, and from the University of California President's Postdoctoral Fellowship Program. R.K.K., E.T.W., and J.H. acknowledge funding from the NASA Habitable Worlds program under award NNX16AB61G. R.K.K. acknowledge funding from NASA Astrobiology Institute's Virtual Planetary Laboratory team, supported by NASA under cooperative agreement NNH05ZDA001C. This work utilized the Janus supercomputer which is supported by the National Science Foundation (award CNS-0821794) and the University of Colorado at Boulder. We would like to acknowledge high-performance computing support from Yellowstone (ark:/85065/d7wd3xhc) provided by NCAR's Computational and Information Systems Laboratory, sponsored by the National Science Foundation. This work was performed as part of the NASA Astrobiology Institute's Virtual Planetary Laboratory Lead Team, supported by the National Aeronautics and Space Administration through the NASA Astrobiology Institute under solicitation NNH12ZDA002C and Cooperative Agreement Number NNA13AA93A. References Abe, Y., Abe-Ouchi, A., Sleep, N. H., & Zahnle, K. J. 2011, AsBio, 11(5), 443 Abbot, D. S., Voigt, A., & Koll, D. 2011, JGR, 116, D18103 Andrews, T., Gregory, J.M., Webb, M.J., & Taylor, K.E. 2012, GRL, 39, L09712 Boschi, R., Lucarini, V., & Pascale, S. 2013, Icar, 226, 1724 Boyajian, T. S., McAlister, H. A., van Belle, G., et al. 2012, ApJ, 757, 112 Budyko, M. I. 1969, Tell, 21, 611 Butler, R. P., Wright, J. T., Marcy, G. W., et al. 2006, ApJ, 646, 505 Briegleb, B. P. 1992, JGR, 97, 7603 Del Genio, A. D. 2016, ArXiv e-prints [arXiv:1603.07424] Driscoll, P. & Bercovici, D. 2013, Icar, 226, 1447 Dunkle, R. V. & Bevans, J. T. 1956, J. Meteorol., 13, 212 Edson, A., Lee, S., Bannon, P., et al. 2011, Icar, 212, 1 Flato, G., Marotzke, J., Abiodun, B., et al. 2013, in Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change, eds. T. Stocker, D. Qin, G.-K. Plattner, et al. (Cambridge, UK, New York, NY, USA: Cambridge University Press), 741 Forget, F., Wordsworth, R., Millour, E., et al. 2013, Icar, 222, 81 Gardner, J. P., Mather, J. C., Clampin, M., et al. 2006, SSRv, 123, 485 Godolt, M., Grenfell, J. L., Haman-Reinus, A., et al. 2015, P&SS, 111, 62 Godolt, M., Grenfell, J.L., Kitzmann, D., et al. 2016, A&A, 592, A36 Hack, J. J. 1994, JGR, 99, 5551 Habets, G. M. H. J. & Heintze, J. R. W. 1981, A&AS, 46, 193 Hansen, J., Sato, M., Ruedy, R., et al. 2005, JGR, 110, D18104 Hart, M. H. 1979, Icar, 37, 351 Haqq-Misra, J. & Kopparapu, R. K. 2015, MNRAS, 446, 428 Haqq-Misra, J., Kopparapu, R. K., Batalha, N. E., et al. 2016, ApJ, 827, 2 Hoffman, P. F., Kaufman, A. J., Kalverson, G. P., & Schrag, D. P. 1998, Sci, 281, 1342 Iben, I. 1967, ARA&A, 5, 571. Joshi, M. M. & Haberle, R. M. 2012, AsBio, 12, 3 Kadoya S., & Tajika, E. 2015, ApJL 815, L7 Kasting, J. F., Pollack, J. B., & Ackerman, T. P. 1984, Icar, 57, 335 Kasting, J. F. 1988, Icar, 74, 462-494 Kasting, J. F., Whitmire, D. P. & Reynolds, R. T. 1993, Icar, 101, 108 Kirschvink, J. L., Gaidos, E. J, Bertani, L. E. et al. 2000, PNAS, 97, 1400 Kitzmann, D., Alibert, Y., Godolt, M., et al. 2015, MNRAS 452, 3752 Kopparapu, R. K., Ramirez, R., Kasting, J. F., et al. 2013, ApJ, 765, 131 Kopparapu, R. K, Ramirez, R. M., Kotte, J. S., et al. 2014, ApJL, 787, L29. Kopparapu, R. K., Wolf, E. T., Haqq-Misra, J., et al. 2016, ApJ 819, 84 Leconte, J., Forget, F., Charnay, B. et al. 2013 Natur, 504, 268 Leconte, J., Wu, H., Menou, K., & Murray, M. 2015, Sci, 347,632 Lucarini, V., Fraedrich, K., & Lunkeit, F. 2010, QJRMS, 136(646), 2 Mills, B., Watson, A. J., Goldblatt, C., et al., 2011, Nat. Geosci., 4, 861 Neale, R. B., Richter, J. H., Conley, A. J., et al. 2010, NCAR/TN-486+STR Nisbet, E. G. & Sleep, N. H. 2001, Natur, 409, 1083 Pickles, A. J. 1998, PASP, 110, 863 Pierrehumbert, R. T., Abbot, D. S., Voigt, A. & Koll, D. 2011, AREPS, 39, 417 Popp, M., Schmidt, H., & Marotzke, J. 2016, NatCo, 7, 10627 Poulsen, C. J., Pierrehumbert, R. T., & Jacob, R. L. 2001, GRL, 28, 1575 Ramanathan, V. & Downey, P. 1986, JGR, 91, 8649 Rasch, P. J. & Kristjánsson, J. E. 1998, J. Clim, 11, 1587 Ricker, G. R., Winn, J. N., Vanderspek, R., et al. 2014, Proc. SPIE, 9143, 914320 Rugheimer, S., Segura, A., Kaltenegger, L., & Sasselov, D. 2015, ApJ, 806, 137 Rushby, A. J., Claire, M. W., Osborn, H., & Watson, A. J. 2013, AsBio, 13(9), 833 Segura, A., Krelove, K. Kasting, J. F., et al. 2003, AsBio, 3, 689 Selsis, F., Kasting, J. F., Levrard, B., et al. 2007, A&A, 476, 1373 Sherwood, S. C. & Huber, M. 2010, PNAS 107(21), 9552 Shields, A. L., Meadows, V. S., Bitz, C. M., et al. 2013, AsBio, 13(8), 715. Shields, A. L., Bitz, C. M., Meadows, V. S., et al. 2014, ApJL, 785, L9 Shields, A. L., Barnes, R., Agol, E., et al. 2016, AsBio, 16(6), 443 Tajika, E., 2007, EP&S, 59, 293 Urata, R. A. & Toon, O. B. 2013, Icar, 226, 229 Walker, J. C. G., Hays, P. B., & Kasting, J. F. 1981, JGR, 86, 9776 Way, M. J., Del Genio, A. D., Kiang, N. Y., et al. 2016, GRL, 43, ? Wolf, E. T. & Toon, O. B. 2013, AsBio, 13(7), 1 Wolf, E. T. & Toon, O. B. 2014, AsBio, 14(3), 241 Wolf, E. T. & Toon, O. B. 2014, GeoRL, 41, 167 Wolf, E. T. & Toon, O. B. 2015 JGRD, 120, 5775 Wordsworth, R. D., Forget, F., Selsis, F., et al. 2011, ApJL, 733, L48 Wordsworth, R., Forget, F., Millour, E., et al. 2013, Icar, 222, 1 Yang, J., Cowan, N., & Abbot, D. S., 2013, ApJL, 771, L45 Yang, J., Boue, G., Fabrycky, D. C., & Abbot, D. S. 2014, ApJL, 787, L2 Yang, J., Leconte, J., Wolf, E. T. et al. 2016, ApJ, 826, 222 Zhang, G. J. & McFarlane, N. A. 1995, Atmosphere-ocean 33(3), 407 Zsom, A., Seager, S., de Wit, J., & Stamenković, V. 2013, ApJ 778, 109 Figure Captions Figure 1: Empirical stellar spectra for F-, G-, and K-dwarf stars used in this study, normalized to a total flux of 1360 W m-2. Figure 2: The evolution of global mean surface temperatures (panels a,b,c) and climate sensitivity (panels d,e,f) for Earth-like planets around F-, G-, and K-dwarf main sequence stars as a function of relative stellar flux (S/S0). The four stable climate regions are labeled and indicated by shaded regions in the top panels. Different color lines are associated with different simulation sets. Red lines indicate simulations starting from modern Earth conditions, under a positive solar forcing (i.e. warming). Green lines indicate simulations starting from modern Earth conditions, under a negative solar forcing (i.e. cooling). Finally, blue lines indicate simulations starting from a globally glaciated state, under a positive solar forcing. Numbers in the bottom panels mark peaks in climate sensitivity and thus represent specific climatic transitions; (1) from waterbelt to snowball, (2) from temperate to waterbelt, (3) from snowball to temperate, (4) temperate to moist greenhouse, and (5) towards a runaway greenhouse. The black dashed line marks the climate where diffusion limited water-loss could remove an Earth ocean of water within about 1 Gyr. Figure 3: The model top temperature (a) and the model top water vapor volume mixing ratio as a function of mean surface temperature for planets around F-, G- and K-dwarf stars. Note the model top pressure is ~0.2 mb. Ocean loss timescales are calculated as the time for diffusion limited escape to remove an Earth ocean of water from the planet. Figure 4: Circumstellar climate zones as a function of relative stellar flux for Earth-like planets at constant CO2. The top panel assumes an initial state that is warm (i.e. liquid water covers the surface). The bottom panel assumes an initial state of a completely ice- covered planet. Figure 5. The lifetime of habitable (solid lines) and temperate (dashed lines) climate zones driven by main sequence brightening, for warm (red) and cold (blue) initial conditions respectively, as a function of the stellar effective temperature. These values represent the maximum possible time for these phases of climates to exist, without invoking a draw-down of CO2 to stabilize climate against continued warming. Figure 6: Constraints on the inner edge (a), outer edge (b), and total habitable zone (c) determined from recent modeling studies. In all panels, tidally locked planets reside in the light yellow shaded region while planets in the blue shaded region can rotate rapidly. Solid lines and diamonds are used to mark constraints on the inner edge of the habitable zone. Dashed lines and triangles are used to mark constraints on the outer edge of the habitable zone. Table 1 Model configurations Moist physics Horizontal resolution Vertical levels Model top (mb) Radiative transfer3 Longwave range (µm) Shortwave range (µm) Continents Ocean heat transport Ocean albedo, visible/infrared Sea ice albedo, visible/infrared Snow albedo, visible/infrared CO2 (ppm) 0.07/0.06 0.67/0.3 0.8/0.68 1Used for simulations into and out of snowball states, see Shields et al. (2013) 2Used for simulations of moist greenhouse states, see Wolf & Toon (2015) 3See Yang et al. (2016) for a comparison of these two codes. Table 1: Summary of "cold" and "warm" model configurations. Each uses the same core atmosphere model, with the same model physics, except where noted above. 400 Cold1 CAM4 2° x 2.5° 26 3 Native 5.0 – 1000 0.2 – 5.0 None None Warm2 CAM4 4° x 5° 45 0.2 Correlated--‐k 2.5 --‐ 1000 0.2 --‐12.2 Present day Earth Present day Earth 0.07/0.06 0.68/0.3 0.91/0.63 367 F K G 289.1 288.0 0.329 0.152 0.176 0.114 0.038 37.6 28.5 9.1 26.5 0.115 0.015 66.3 34.8 27.0 47.2 7.5 0.042 281.4 280.1 0.387 0.216 0.171 0.143 0.073 35.3 25.0 10.3 15.7 0.093 0.018 69.5 35.7 29.9 49.6 16.1 0.081 294.8 294.1 0.284 0.119 0.166 0.104 0.015 39.5 31.3 8.2 41.2 0.129 0.013 64.3 37.7 23.3 43.4 3.7 0.021 Table 2 Star (type) Surface Temperature (K) 2--‐Meter Air Temperature (K) Albedo, All--‐sky Albedo, Clear--‐sky Albedo, Cloud Albedo, Surface Albedo, Rayleigh Greenhouse Effect, Total (K) Greenhouse Effect, Clear--‐sky (K) Greenhouse Effect, Cloud (K) Water Vapor Column (Kg m--‐2) Cloud Water Column (Kg m--‐2) Cloud Ice Column (Kg m--‐2) Cloud Fraction, Total (%) Cloud Fraction, Low (%) Cloud Fraction, Middle (%) Cloud Fraction, High (%) Sea Ice Fraction Relative to Ocean (%) Snow Depth (m) Table 2: Global and annual mean quantities from control simulations using the warm configuration, which includes ocean heat transport identical to the modern Earth. a b 1.39815×10-4 8.48102×10-5 6.81156×10-5 8.43240×10-5 5.54441×10-5 7.27318×10-5 1.07307×10-4 c 3.12706×10-8 3.89303×10-8 2.12922×10-8 2.61308×10-8 7.18032×10-9 9.82310×10-10 -1.14135×10-8 Table 3 Differentiations of Circumstellar Climate Water-loss 1 Gyr (Ts = 355 K)1 1.19645 1.11892 Temperate to Moist Greenhouse Transition Present Day Earth Conditions (Ts = 289 K)2 1.00014 Biological Heat Stress (Ts = 300 K) 1.06666 Temperate to Waterbelt Transition2 0.96011 Waterbelt to Snowball Transition2 0.92515 Snowball to Temperate Transition3 1.05521 1The water-loss limit is the nominal inner edge of the habitable zone. 2Only accessible from warm start conditions 3Only realizable from cold start conditions Table 3:. Coefficients a, b, and c to be used in equation 1 to calculate circumstellar climates zones in units of relative stellar flux for rapidly rotating Earth-like planets. These parameterizations were fit to our model calculations. Valid for stars with 4900 K ≥ Teff ≥ 6600 K. F2V HD128167, Teff = 6594 K G2V Sun, Teff = 5778 K K2V HD22049, Teff = 5084 K 2000 2000 1500 1500 1000 1000 500 500 0 0 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 Wavelength (µm) Figure 1 3000 3000 2500 2500 ) 1 - m µ 2 - m W ( e c n a d a R i Figure 2 ) K ( e r u t a r e p m e t e c a f r u S 350 300 250 200 K-dwarf a moist greenhouse water-loss ~1 Gyr temperate waterbelt snowball 3 1 2 5 4 G-dwarf b e 3 1 2 4 5 0.6 0.8 1.2 Relative stellar flux (S/S0) 1.0 1.4 F-dwarf c f 12 3 4 5 0.6 0.8 1.2 Relative stellar flux (S/S0) 1.0 1.4 100.0 d 10.0 1.0 0.1 0.6 0.8 1.2 Relative stellar flux (S/S0) 1.0 1.4 ) 1 - ) 2 - m W ( K ( y t i v i t i s n e s e t a m i l C Figure 3 220 200 180 F G K a ) K ( e r u t a r e p m e t p o t l e d o M 160 140 ~80 Myr ~800 Myr ~8 Gyr ~80 Gyr ocean loss timescales 120 100 100 10-2 10-4 10-6 10-8 b o i t a r i i g n x m e m u o v O H p o l 2 300 Mean surface temperature (K) 320 340 360 t l e d o M 10-10 10-12 280 water-loss ~1 Gyr heat stress present day moist greenhouse warm start temperate snowball waterbelt water-loss ~1 Gyr heat stress snowball moist greenhouse cold start 1.3 5000 1.4 temperate 1.2 Relative stellar flux (S/S0) 1.1 1.0 0.9 0.8 0.8 0.8 Figure 4 6500 a 6000 5500 5000 6500 b 6000 5500 ) K ( e r u t a r e p m e t e v i t c e f f e r a l l t e S ) K ( e r u t a r e p m e t e v i t c e f f e r a l l t e S temperate zone, warm start habitable zone, warm start temperate zone, cold start habitable zone, cold start 1 1 2 2 3 5 5 3 (cid:111)climate (Gyr) 4 4 6 6 7 7 8 8 Figure 5 ) K ( e r u t a r e p m e t e v i t c e f f e r a l l t e S 6500 6500 6000 6000 5500 5500 5000 5000 0 0 Yang et al. (2014) slow rotators TidalLockingRadius This study water-loss Kopparapu et al. (2013) water-loss Abe et al. (2011) desert planets Abe et al. (2011) aqua planets Leconte et al. (2013) Popp et al. (2016) Yang et al. (2014) fast rotators 1.5 Haqq-Misra et al. (2016) maximum greenhouse Wolf & Toon (2014) 0.2 bar CO2 1.0 0.5 Kopparapu et al. (2013) maximum greenhouse Shields et al. (2016) 5 bar CO2 Wordsworth et al. (2011) 10 bar CO2 0.4 0.6 0.2 0.2 0.2 0.8 1.5 Relative stellar flux (S/S0) 1.0 0.5 Figure 6 7000 F 6000 G 5000 K 4000 M 3000 7000 F 6000 G 5000 K 4000 M 3000 a b c ) K ( t e r u a r e p m e t e v i t c e f f e r a l l t e S ) K ( e r u t a r e p m e t e v i t c e f f e r a l l e t S ) K ( e r u t a r e p m e t e v i t c e f f e r a l l e t S Kopparapu et al. (2016) low and high metallicity Inner edge boundaries 2.5 2.0 This study snowball This study deglaciation TidalLockingRadius Outer edge boundaries 1.2 1.0 TidalLockingRadius 7000 F 6000 G 5000 K 4000 M Habitable zone boundaries 3000 2.0 inner edge outer edge 2.5
1808.07055	2	1808	2018-11-05T18:19:58	On the possibility of detecting ultra-short period exoplanets with LISA	[ "astro-ph.EP", "gr-qc" ]	Cunha et al. (2018) recently reexamined the possibility of detecting gravitational waves from exoplanets, claiming that three ultra-short period systems would be observable by LISA. We revisit their analysis and conclude that the currently known exoplanetary systems are unlikely to be detectable, even assuming a LISA observation time $T_{\rm obs}=4$ yrs. Conclusive statements on the detectability of one of these systems, GP Com b, will require better knowledge of the system's properties, as well as more careful modeling of both LISA's response and the galactic confusion noise. Still, the possibility of exoplanet detection with LISA is interesting enough to warrant further study, as gravitational waves could yield dynamical properties that are difficult to constrain with electromagnetic observations.	astro-ph.EP	astro-ph	On the possibility of detecting ultra-short period exoplanets with LISA Kaze W. K. Wong1(cid:63), Emanuele Berti1,2, William E. Gabella3, Kelly Holley- Bockelmann3,4 1Department of Physics and Astronomy, Johns Hopkins University, Baltimore, MD 21218 USA 2Department of Physics and Astronomy, The University of Mississippi, University, MS 38677, USA 3Vanderbilt University, Nashville, TN, USA 4Fisk University, Nashville, TN, USA Accepted ; Received ; in original form ABSTRACT Cunha et al. (2018) recently reexamined the possibility of detecting gravitational waves from exoplanets, claiming that three ultra-short period systems would be observable by LISA. We revisit their analysis and conclude that the currently known exoplanetary systems are unlikely to be detectable, even assuming a LISA observation time Tobs = 4 yrs. Conclusive statements on the detectability of one of these systems, GP Com b, will require better knowledge of the system's properties, as well as more careful modeling of both LISA's response and the galactic confusion noise. Still, the possibility of exoplanet detection with LISA is interesting enough to warrant further study, as gravitational waves could yield dynamical properties that are difficult to constrain with electromagnetic observations. Key words: exoplanets - gravitational waves The idea of using space-based gravitational-wave (GW) observations with LISA to detect exoplanets was proposed almost 20 years ago. At the time only about 20 such systems were known. Even taking into account that eccentric systems could produce significant GW power at higher harmonics, and that some of these exoplanets could resonantly excite the oscillation modes of the star they are orbiting, none of them was found to be detectable (Ferrari et al. 2000; Berti & Ferrari 2001a; Berti & Ferrari 2001b). However, the number of known exoplanets is now in the thousands and exoplanet surveys point to a very large pop- ulation of planetary systems in our Galaxy, with more than one planet per star on average (Cassan et al. 2012) and free- floating planets outnumbering the stars (Mr´oz et al. 2017). Many of these planetary systems are dramatically different than our own, with hot Jupiters, highly eccentric and in- clined orbits, as well as entire systems of tightly-packed in- ner planets. Such a rich and varied population of exoplane- tary systems strains our current understanding of planetary system formation and evolution. A few years ago Ain et al. (2015) showed that the stochastic GW background produced by these systems would peak at ∼ 10−5 Hz, with character- istic amplitude about two orders of magnitude below LISA's (cid:63) [email protected] c(cid:13) 0000 The Authors sensitivity, though as the exoplanet discovery space expands, our estimates of this background will evolve. Cunha et al. (2018) recently revisited the possibility of detecting exoplanets with LISA. They computed the char- acteristic strain for some ultra-short period exoplanets from an online catalog†, and claimed that three systems (GP Com b, V396 Hya b, and J1433 b) have characteristic GW strains large enough to be observable using the original LISA de- sign (Larson et al. 2000, henceforth "Classic LISA") in one year of integration, ignoring the galactic confusion noise: cf. Fig. 2 of Cunha et al. (2018). In Table 1 we collected all relevant known properties (to the best of our knowledge) for these three systems. Note that the companions of GP Com b and V396 Hya b have masses in the exoplanet range, but they are donors of AM CVn-type interacting binaries (Kupfer et al. 2016), while J1433 b consists of an irradiated brown-dwarf companion to an accreting white dwarf (Hern´andez Santisteban et al. 2016). Therefore the classification of these three binaries as exoplanetary systems is, at best, debatable. Given the GW strain amplitude h(t), the characteristic strain hc for a monochromatic circular binaries with orbital frequency forb = 2π/P emitting GWs at frequency f = 2forb over an observation time Tobs can be defined as hc = † http://exoplanet.eu/catalog/ 8 1 0 2 v o N 5 . ] P E h p - o r t s a [ 2 v 5 5 0 7 0 . 8 0 8 1 : v i X r a 2 K.W.K. Wong, E. Berti, W.E. Gabella, K. Holley-Bockelmann Table 1. Parameters of the most promising exoplanetary systems for GW detection (note that, as discussed in the text, the classification of these systems as exoplanets is questionable). All parameters are taken from the online exoplanet catalog http://exoplanet.eu/catalog/, with the exception of quantities labeled with † (from Gaia Collaboration 2018), ‡ (from Kupfer et al. 2016), § (from Hern´andez Santisteban et al. 2016), and ∗(from Cunha et al. 2018). Here DL, Mstar[M(cid:12)] and Mplanet[MJ] denote the luminosity distance, mass of the star in solar masses, and mass of the planet in Jupiter masses, while (¯θS , ¯φS ), ι and P denote the sky location (in ecliptic coordinates), inclination and orbital period of the binary. Name GP Com b V396 Hya b J1433 b DL [pc] 72.83 ± 0.32† 93.51 ± 1.29† 224.52 ± 10.22† 0.435‡ 0.345‡ Mstar[M(cid:12)] Mplanet[MJ] 26.2 ± 16.6 18.3 ± 12.2 57.1 ± 0.7 0.8 ± 0.07§ ¯θS [deg] 23.00† -14.50† 23.89† ¯φS [deg] 187.72† 205.73† 212.37† ι[deg] 55.5 ± 22.5 52 ± 27 84.36 P [days] 0.032 0.045∗ 0.054 candidates plotted along with (cid:112)f Sn(f ), where Sn(f ) is the ef- Figure 1. Characteristic strain hc of the loudest exoplanetary fective non-sky averaged noise power spectral density for Clas- sic LISA without galactic confusion noise (Larson et al. 2000, dashed red), as adopted in (Cunha et al. 2018); Classic LISA with galactic confusion noise (solid red); and the current LISA design with galactic confusion noise (Robson et al. 2018, solid black). The galactic confusion background and hc are computed assum- ing Tobs = 2 yrs. Cyan dots with error bars correspond to the non-sky averaged SNR, allowing for uncertainties on the source parameters; brown inverted triangles correspond to the sky- and orientation-averaged SNR. waveform following Cutler (1998), so that h(t) is given by h(t) = (πf )2/3 A(t) √ 3 2 × cos 2M5/3 DL (cid:18)(cid:90) t (cid:19) 2πf (t (cid:48) (cid:48) )dt + ϕp(t) + ϕD(t) , (2) 0 where f (t(cid:48)) is given in equation (1.3) of Poisson & Will (1995). Here A(t), ϕp(t) and ϕD(t) are the amplitude modu- lation, polarization phase and Doppler phase due to LISA's motion (see Appendix A for details). For a binary with com- ponent masses (m1, m2) and total mass M = m1 + m2 the waveform depends on nine parameters: luminosity dis- tance DL, chirp mass M = η3/5M , symmetric mass ratio η = m1m2/M 2, time of coalescence tc, phase of coalescence φc, sky location (¯θS, ¯φS) and orbital angular momentum di- rection (¯θL, ¯φL). The overbar means that the sky location and binary orientation angles are defined in ecliptic coordi- nates. In order to give an estimate of the possible range of SNR, for each source we create Monte Carlo samples (cid:104) 2f(cid:82) Tobs 0 dt h(t)2(cid:105)1/2 (cid:90) Tobs (Moore et al. 2015). In Fig. 1 we follow the conventions established in Robson et al. (2018) -- cf. e.g. their Fig. 6 -- to plot the characteristic strain along with the effective non-sky averaged noise power spectral density of various LISA designs for two readout channels, related to the sky-averaged noise power spectral density by Sn(f ) = n (f ) (Robson et al. 2018).‡ Brown triangles correspond 10 SSA 3 to the sky-averaged characteristic strain (Robson et al. 2018, solid black), while cyan error bars correspond to the range of hc consistent with uncertainties in the source parameters (cf. Table 1). The case for detectability of these three systems with either the current or Classic LISA design based on a characteristic strain calculation is, at best, inconclusive. As discussed in Robson et al. (2018), plots of the char- acteristic strain hc are useful as rough assessments of de- tectability, but any conclusions must ultimately be based on a signal-to-noise ratio (SNR) calculation. For monochro- matic sources, the SNR is defined as ρ = (hh)1/2, where (hh) = 2 Sn(f ) 0 dt h(t)2. (1) To claim detectability, the source of interest must have SNR ρ larger than a certain threshold, which for monochro- matic systems is usually taken to be ρthr = 5 (Kupfer et al. 2018). This is somewhat optimistic: the Mock LISA Data Challenges suggest that ρthr is likely to be larger than 5 (B(cid:32)laut et al. 2010). Crowder & Cornish (2007) even report undetected sources with ρ ∼ 10, though this will likely improve with more research in GW data analysis. Unfortunately, Cunha et al. (2018) did not quantify the SNR of these systems. Furthermore, they used the outdated "Classic LISA" noise curve (Larson et al. 2000) and they did not take into account the fact that galactic binaries produce a significant confusion noise, which is important at the fre- quencies of interest for exoplanetary systems. Here we revisit their analysis for the three planetary systems that are most promising for GW detection. We use updated parameters for these systems (including uncertainties, when available) and we adopt the most recent estimates for the LISA sensitivity curve, including galactic confusion noise. The parameters of the three systems under consideration are listed in Table 1. We model the motion of the LISA detector and compute the SNR using a nonspinning, quasicircular time-domain ‡ We remark that this convention differs from the conventions used in Cutler (1998) and Berti et al. (2005), where the SNRs coming from the strain amplitudes hα (α = 1, 2) in the two chan- nels are added in quadrature and Sn(f ) = 3 20 SSA n (f ). GP Com bV396 Hya bJ1433 bOn the possibility of detecting ultra-short period exoplanets with LISA 3 Table 2. SNR for the loudest sources considered in (Cunha et al. 2018), using the noise power spectral density for Classic LISA (Larson et al. 2000, columns 2, 3 and 4) and the current LISA design (Robson et al. 2018, columns 5 and 6). The second row indicates whether we included galactic confusion noise or not. The third row lists the assumed observation time Tobs (in yrs). Numbers in square brackets are the maximum and minimum SNRs consistent with parameter uncertainties for the given source. In round parentheses we report the sky location and orientation averaged SNR. Classic LISA LISA Confusion Tobs (yrs) GP Com b V396 Hya b J1433 b No 1 5.56(cid:2)13.91 1.21(cid:2)2.04 1.12(cid:2)1.61 (cid:3) (6.20) (cid:3) (1.17) (cid:3) (1.63) 0.97 0.14 0.41 No 2 8.05(cid:2)19.37 1.73(cid:2)3.01 1.52(cid:2)2.28 (cid:3) (8.76) (cid:3) (1.65) (cid:3) (2.30) 1.38 0.19 0.55 S , φeq based on the parameter uncertainties listed in Table 1. Our waveforms depend on the sky location in the solar system barycenter frame, while the sky location (θeq S ) and incli- nation ι are given in equatorial coordinates (electromagnetic observations do not give information on the polarization an- gle ψ). In order to translate the waveform from the solar sys- tem barycenter frame to an Earth-centered frame, we must solve for the geometric angles in ecliptic coordinates as func- tions of geometric angles in equatorial coordinates. Translat- ing the sky location from ecliptic coordinates to equatorial coordinates is trivial, but the mapping from the orbital an- gular momentum direction to the inclination angle is more complicated. Therefore we draw samples in the LISA (solar system barycenter frame) coordinates, compute the SNR, and display the maximum and minimum SNRs which are consistent with the parameter uncertainties of each source. Our results, which we have checked to be in agreement with the sky-location and orientation averaged results of Robson et al. (2018), are shown in Table 2. If we fix the detectability threshold at ρthr = 5, none of the currently known systems has ρ > ρthr, even assuming coherent integration over the nominal LISA mission lifetime, i.e. Tobs = 4 yrs (Amaro-Seoane et al. 2017). GP Com b -- whose companion is a donor in an AM CVn-type interact- ing binary (Kupfer et al. 2016), so it can hardly be classified as an exoplanet -- would be marginally detectable with the "Classic LISA" design, and it is marginally detectable by the current LISA design in four years only if we consider the most optimistic SNR values allowed by parameter un- certainties. A more reliable assessment of the detectability of this system will require better knowledge of the system's properties, as well as more careful modeling of LISA's re- sponse and of the galactic confusion noise (see e.g. Timpano et al. 2006). For V396 Hya b and J1433 b, the SNR is always lower than the detection threshold. Detection thresholds can be lowered if we incorporate information from electromag- netic measurements into the GW search, but a quantitative assessment of this issue is beyond the scope of this paper (see e.g. Shah & Nelemans 2014). The search for ultra-short period exoplanets is certainly an exciting scientific target for LISA. We hope that our considerations will motivate further work to optimize data analysis methods, to reduce the noise power spectral den- sity at low frequencies, and to improve our understanding of the galactic confusion noise. It will be interesting to model the exoplanet parameter space that would be detectable by LISA (including galactic exoplanets and brown dwarf popu- Yes 2 2.29(cid:2)5.51 0.56(cid:2)0.98 0.54(cid:2)0.80 0.06 0.39 (cid:3) (2.49) (cid:3) (0.54) (cid:3) (0.81) 0.20 Yes 2 2.03(cid:2)4.87 0.52(cid:2)0.92 0.50(cid:2)0.74 0.35 0.06 (cid:3) (2.21) (cid:3) (0.50) (cid:3) (0.75) 0.18 Yes 4 3.31(cid:2)8.05 0.82(cid:2)1.37 0.73(cid:2)1.11 0.54 0.09 (cid:3) (3.62) (cid:3) (0.76) (cid:3) (1.11) 0.27 lations) to better understand the potential of GW observa- tions and their complementarity with respect to traditional detection methods. ACKNOWLEDGMENTS K.W.K.W. and E.B. are supported by NSF Grants No. PHY-1841464 and AST-1841358, and by NASA ATP Grant 17-ATP17-0225. We thank the referee (Neil Cornish), Quentin Baghi, Robert Caldwell, Tyson Littenberg, Travis Robson, Ira Thorpe, Nadia Zakamska, Hsiang-Chih Hwang, Kevin Schlaufman and all members of the NASA LISA Study Team for useful discussions. APPENDIX A: ANTENNA PATTERN In this Appendix we write down, for completeness, the an- tenna pattern expressions used in our non angle-averaged SNR calculation. Following Cutler (1998), we denote the LISA-based coordinate system by unbarred quantities, while barred quantities refer to the fixed ecliptic coordinate sys- tem. The amplitude modulation in equation 2 is given by (cid:114)(cid:104) 2(cid:105)2 A(t) = 1 + (L · n) 2 + 4(L · n) 2 F+ F×2, (A1) where L and −n are the unit vector along the binary's or- bital angular momentum and the GW direction of propa- gation, respectively. The pattern functions F+ and F× are defined as F+(θS, φS, ψS) = (1 + cos2θS) cos 2φS cos 2ψS 1 2 − cos θS sin 2φS sin 2ψS, F×(θS, φS, ψS) = 1 2 (1 + cos2θS) cos 2φS sin 2ψS + cos θS sin 2φS cos 2ψS. (A2) The angles (θS, φS) specify the source location, while ψS denotes the the polarization angle: tan ψS(t) = L · z − (L · n)(z · n) n · (L × z) , (A3) where z is the unit normal to the LISA detector plane. 4 K.W.K. Wong, E. Berti, W.E. Gabella, K. Holley-Bockelmann The scalar products can be written as z · n = cos θS, L · z = L · n = cos ¯θL cos ¯θS + sin ¯θL sin ¯θS cos( ¯φL − ¯φS), sin ¯θL cos( ¯φ(t) − ¯φL), cos ¯θL − 3 2 √ 1 2 (A4) (A5) (A6) Kupfer T., Steeghs D., Groot P. J., Marsh T. R., Nelemans G., Roelofs G. H. A., 2016, MNRAS, 457 Kupfer T., et al., 2018, MNRAS, 480, 302 Larson S. L., Hiscock W. A., Hellings R. W., 2000, Phys. Rev. D, 62, 062001 Moore C. J., Cole R. H., Berry C. P. L., 2015, Classical and Quantum Gravity, 32, 015014 Mr´oz P., et al., 2017, Nature, 548, 183 Poisson E., Will C. M., 1995, Phys. Rev. D, 52, 848 Robson 2018, Cornish N., Liu T., C., preprint, (arXiv:1803.01944) Shah S., Nelemans G., 2014, ApJ, 790, 161 Timpano S. E., Rubbo L. J., Cornish N. J., 2006, Phys. Rev. D, 73, 122001 and n · (L × z) = √ 3 √ 2 3 2 − − given by 1 2 sin ¯θL sin ¯θS sin( ¯φL − ¯φS) (cid:1) cos ¯φ(t)(cid:0)cos ¯θL sin ¯θS sin ¯φS − cos ¯θS sin ¯θL sin ¯φL sin ¯φ(t)(cid:0)cos ¯θS sin ¯θL cos ¯φL − cos ¯θL sin ¯θS cos ¯φS (cid:1) . The polarization and Doppler phases in equation 2 are (cid:34) ϕp(t) = tan −1 ϕD(t) = 2πf c 2(L · n)F×(t) 2 (1 + (L · n) )F+(t) R sin ¯θS cos( ¯φ(t) − ¯φS), (cid:35) (A7) (A8) (A9) where R = 1AU and ¯φ(t) = ¯φ0 + 2πt/T . Here T = 1 yr is the orbital period of LISA, and ¯φ0 is a constant specifying the detector's location at time t = 0. Assuming no precession of the orbital angular momen- tum, the time-dependent LISA related angles (θS, φS, ψS) can be expressed in terms of the time-independent angles defined in the ecliptic coordinates (¯θS, ¯φS, ¯θL, ¯φL) through the following relations: cos ¯θS − 2πt T sin ¯θS cos( ¯φ(t) − ¯φS), φS(t) = α0 + cos θS(t) = (A10a) 3 2 √ 1 2 −1 + tan 3 cos ¯θS + sin ¯θS cos( ¯φ(t) − ¯φS) 2 sin ¯θS sin( ¯φ(t) − ¯φS) (cid:20)√ (cid:21) , (A10b) where α0 is a constant specifying the orientation of the de- tector arms at t = 0. We set α0 = 0 and ¯φ0 = 0 in our calculations, but we checked that varying α0 and ¯φ0 has an insignificant effect on the SNR as long as the observation period Tobs (cid:38) 1 yr. REFERENCES Ain A., Kastha S., Mitra S., 2015, Phys. Rev. D, 91, 124023 Amaro-Seoane P., et al., 2017, preprint, (arXiv:1702.00786) Berti E., Ferrari V., 2001a, ICTP Lect. Notes Ser., 3, 371 Berti E., Ferrari V., 2001b, Phys. Rev. D, 63, 064031 Berti E., Buonanno A., Will C. M., 2005, Phys. Rev. D, 71, 084025 B(cid:32)laut A., Babak S., Kr´olak A., 2010, Phys. Rev. D, 81, 063008 Cassan A., et al., 2012, Nature, 481, 167 Crowder J., Cornish N. J., 2007, Classical and Quantum Gravity, 24, S575 Cunha J. V., Silva F. E., Lima J. A. S., 2018, MNRAS, 480, L28 Cutler C., 1998, Phys. Rev. D, 57, 7089 Ferrari V., Berti E., D'Andrea M., Ashtekar A., 2000, Interna- tional Journal of Modern Physics D, 9, 495 Gaia Collaboration 2018, VizieR Online Data Catalog, 1345 Hern´andez Santisteban J. V., et al., 2016, Nature, 533, 366
1006.1643	1	1006	2010-06-08T19:48:50	Stochastic orbital migration of small bodies in Saturn's rings	[ "astro-ph.EP", "astro-ph.IM" ]	Many small moonlets, creating propeller structures, have been found in Saturn's rings by the Cassini spacecraft. We study the dynamical evolution of such 20-50m sized bodies which are embedded in Saturn's rings. We estimate the importance of various interaction processes with the ring particles on the moonlet's eccentricity and semi-major axis analytically. For low ring surface densities, the main effects on the evolution of the eccentricity and the semi-major axis are found to be due to collisions and the gravitational interaction with particles in the vicinity of the moonlet. For large surface densities, the gravitational interaction with self-gravitating wakes becomes important. We also perform realistic three dimensional, collisional N-body simulations with up to a quarter of a million particles. A new set of pseudo shear periodic boundary conditions is used which reduces the computational costs by an order of magnitude compared to previous studies. Our analytic estimates are confirmed to within a factor of two. On short timescales the evolution is always dominated by stochastic effects caused by collisions and gravitational interaction with self-gravitating ring particles. These result in a random walk of the moonlet's semi-major axis. The eccentricity of the moonlet quickly reaches an equilibrium value due to collisional damping. The average change in semi-major axis of the moonlet after 100 orbital periods is 10-100m. This translates to an offset in the azimuthal direction of several hundred kilometres. We expect that such a shift is easily observable.	astro-ph.EP	astro-ph	Astronomy&Astrophysicsmanuscript no. main March 1, 2018 c(cid:13) ESO 2018 0 1 0 2 n u J 8 . ] P E h p - o r t s a [ 1 v 3 4 6 1 . 6 0 0 1 : v i X r a Stochastic orbital migration of small bodies in Saturn's rings Hanno Rein and John C. B. Papaloizou University of Cambridge, Department of Applied Mathematics and Theoretical Physics, Centre for Mathematical Sciences, Wilberforce Road, Cambridge CB3 0WA, UK e-mail: [email protected] Preprint online version: March 1, 2018 ABSTRACT Many small moonlets, creating propeller structures, have been found in Saturn's rings by the Cassini spacecraft. We study the dynam- ical evolution of such 20-50m sized bodies which are embedded in Saturn's rings. We estimate the importance of various interaction processes with the ring particles on the moonlet's eccentricity and semi-major axis analytically. For low ring surface densities, the main effects on the evolution of the eccentricity and the semi-major axis are found to be due to collisions and the gravitational interac- tion with particles in the vicinity of the moonlet. For large surface densities, the gravitational interaction with self-gravitating wakes becomes important. We also perform realistic three dimensional, collisional N-body simulations with up to a quarter of a million particles. A new set of pseudo shear periodic boundary conditions is used which reduces the computational costs by an order of magnitude compared to previous studies. Our analytic estimates are confirmed to within a factor of two. On short timescales the evolution is always dominated by stochastic effects caused by collisions and gravitational interaction with self- gravitating ring particles. These result in a random walk of the moonlet's semi-major axis. The eccentricity of the moonlet quickly reaches an equilibrium value due to collisional damping. The average change in semi-major axis of the moonlet after 100 orbital periods is 10-100m. This translates to an offset in the azimuthal direction of several hundred kilometres. We expect that such a shift is easily observable. Key words. Saturn -- planetary rings -- moons -- satellites -- celestial mechanics -- collisional physics -- N-body simulations 1. Introduction Small bodies, so called moonlets, which are embedded in Saturn's rings, can create propeller shaped structures due to their disturbance of the rings. These have been predicted both analytically and numerically [Spahn and Sremcevi´c, 2000, Sremcevi´c et al., 2002, Seiss et al., 2005]. Only recently, they have been observed by the Cassini spacecraft in both the A and B ring [Tiscareno et al., 2006, 2008]. These 20 m − 100 m sized bodies can migrate within the rings, similar to proto-planets which migrate in a proto-stellar disc. Depending on the disc properties and the moonlet size, this can happen in either a smooth or in a stochastic (random walk) fashion. We refer to those migration regimes as type I and type IV, respectively, in analogy to the terminology in disc-planet in- teractions. Crida et al. [2010] showed that there is a laminar type I regime that might be important on very long timescales. This migration is qualitatively different to the migration in a pressure supported gas disc. However, the migration of moonlets in the A ring is generally dominated by type IV migration, at least on short timescales. In this paper, we study the type IV migration regime in de- tail. The full mutual interaction of the ring particles with the moonlet and its consequent induced motion are considered both analytically and numerically. We first review the basic equations governing the moonlet and the ring particles in a shearing box approximation in Sect. 2. Then, in Sect. 3, we estimate the eccentricity damping timescale due to ring particles colliding with the moonlet and ring particles on horseshoe orbits as well as the effect of particles on circulat- ing orbits. In Sect. 4 we estimate the excitation of the moonlet eccentricity caused by stochastic particle collisions and gravita- tional interactions with ring particles. This enables us to derive an analytic estimate of the equilibrium eccentricity. In Sect. 5 we discuss and evaluate processes, such as colli- sions and gravitational interactions with ring particles and self- gravitating clumps, that lead to a random walk in the semi-major axis of the moonlet. We describe our numerical code and the initial conditions used in Sect. 6. We perform realistic three dimensional N-body simulations of the ring system and the moonlet, taking into ac- count a moonlet with finite size, a size distribution of ring par- ticles, self-gravity and collisions. The results are presented in Sect. 7. All analytic estimates are confirmed both in terms of qualitative trends and quantitatively to within a factor of about two in all simulations that we performed. We also discuss the long term evolution of the longitude of the moonlet and its ob- servability, before summarising our results in Sec. 8. 2. Basic equations governing the moonlet and ring particles We adopt a local right handed Cartesian coordinate system with its origin being in circular Keplerian orbit with semi-major axis a and rotating uniformly with angular velocity Ω. This orbit co- incides with that of the moonlet when it is assumed to be un- perturbed by ring particles. The x axis coincides with the line joining the central object of mass Mp and the origin. The unit vector in the x direction, ex, points away from the central ob- ject. The unit vector in the y direction ey points in the direction 2 H. Rein and J.C.B. Papaloizou: Stochastic orbital migration of moonlets in Saturn's rings b r e t e m a r a p t c a p m i (a) (b) (c) (d) (e) x y Lagrange point L1 Roche lobe Lagrange point L2 Fig. 1. Trajectories of ring particles in a frame centred on the moonlet. Particles accumulate near the moonlet and fill it's Roche lobe. Particles on trajectories labelled a) are on circulating orbits. Particles on trajectories labelled b) collide with other particles in the moonlet's vicinity. Particles on trajectories labelled (c) collide with the moonlet directly. Particles on trajectories labelled (d) are on horseshoe orbits. Particles on trajectories labelled (e) leave the vicinity of the moonlet. of rotation and the unit vector in the z direction, ez points in the vertical direction being normal to the disc mid-plane. In general, we shall consider a ring particle of mass m1 in- teracting with the moonlet which has a much larger mass m2. A sketch of three possible types of particle trajectory in the vicin- ity of the moonlet is shown in Fig. 1. These correspond to three distinct regimes, a) denoting circulating orbits, b) and c) denot- ing orbits that result in a collision with particles in the vicinity of the moonlet and directly with the moonlet respectively, and d) denoting horseshoe orbits. All these types of trajectory occur for particles that are initially in circular orbits, both interior and exterior to the moonlet, when at large distances from it. (1) Approximating the gravitational force due to the central ob- ject by its first order Taylor expansion about the origin leads to Hill's equation, governing the motion of a particle of mass m1 of the form r = −2Ωez × r + 3Ω2(r · ex)ex − ∇Ψ/m1, where r = (x, y, z) is position of a particle with mass m1 and Ψ is the gravitational potential acting on the particle due to other objects of interest such as the moonlet. The square amplitude of the epicyclic motion E2 can be de- fined through E2 = Ω−2 x2 + (2Ω−1y + 3x)2. Note that neither the eccentricity e, nor the semi-major axis a are formally defined in the local coordinate system. However, in the absence of interaction with other masses (Ψ = 0), E is conserved, and up to first order in the eccentricity we may make the identification E = e a. We recall the classical definition of the eccentricity e in a coordinate system centred on the central object (2) w × (s × w) Ω2a3 − e =(cid:12)(cid:12)(cid:12)(cid:12)(cid:12) s s(cid:12)(cid:12)(cid:12)(cid:12)(cid:12) . (3) Here, s is the position vector (a, 0, 0) and w is the velocity vector of the particle relative to the mean shear associated with circu- lar Keplerian motion as viewed in the local coordinate system plus the circular Keplerian velocity corresponding to the orbital frequency Ω of the origin. Thus 3 2 (4) Ωx + aΩ, z! . w = x, y + All eccentricities considered here are very small (∼ 10−8 − 10−7) so that the difference between the quantities defined through use of Eqs. 2 and 3 is negligible, allowing us to adopt E as a measure of the eccentricity throughout this paper. Let us define another quantity, A, that is also conserved for non interacting particle motion, Ψ = 0, which is the x coordinate of the centre of the epicyclic motion and is given by A = 2Ω−1 y + 4x. (5) We identify a change in A as a change in the semi-major axis a of the particle, again, under the assumption that the eccentricity is small. 2.1. Twointeractingparticles We now consider the motion of two gravitationally interact- ing particles with position vectors r1 = (x1, y1, z1) and r2 = (x2, y2, z2). Their corresponding masses are m1 and m2, respec- tively. The governing equations of motion are r1 = −2Ωez × r1 + 3Ω2(r1 · ex)ex − ∇r1 Ψ12/m1 r2 = −2Ωez × r2 + 3Ω2(r2 · ex)ex + ∇r1 Ψ12/m2, where the interaction gravitational potential −Gm1m2/ r1 − r2. The position vector of the centre of mass of the two particles is given by is Ψ12 (6) (7) = ¯r = m1r1 + m2r2 m1 + m2 . (8) The vector ¯r also obeys Eq. 1 with Ψ = 0, which applies to an isolated particle. This is because the interaction potential does not affect the motion of the centre of mass. We also find it useful to define the vector Ei = (Ω−1 xi, 2Ω−1 yi + 3xi) i = 1, 2. (9) H. Rein and J.C.B. Papaloizou: Stochastic orbital migration of moonlets in Saturn's rings 3 Then, consistently with our earlier definition of E, we have Ei = Ei. The amplitude of the epicyclic motion of the centre of mass ¯E is given by (m1 + m2)2 ¯E2 = m2 1E2 (10) Here φ12 is the angle between E1 and E2. It is important to note that ¯E is conserved even if the two particles approach each other and become bound. This is as long as frictional forces are in- ternal to the two particle system and do not affect the centre of mass motion. 2 + 2m1m2E1E2 cos(φ12). 1 + m2 2E2 3. Effects leading to damping of the eccentricity of the moonlet We begin by estimating the moonlet eccentricity damping rate associated with ring particles that either collide directly with the moonlet, particles in its vicinity, or only interact gravitationally with the moonlet. 3.1. Thecontributionduetoparticlesimpactingthemoonlet m2 (11) m2! E2, m1 + m2 E2 ≃ 1 − Particles impacting the moonlet in an eccentric orbit exchange momentum with it. Let us assume that a ring particle, identified with m1 has zero epicyclic amplitude, so that E1 = 0 far away from the moonlet. The moonlet is identified with m2 and has an initial epicyclic amplitude E2. The epicyclic amplitude of the centre of mass is therefore m1 ¯E = where we have assumed that m1 ≪ m2. The moonlet is assumed to be in a steady state in which there is no net accretion of ring particles. Therefore, for every particle that either collides directly with the moonlet or nearby particles bound to it (and so itself becomes temporarily bound to it), one particle must also escape from the moonlet. This happens pri- marily through slow leakage from locations close to the L1 and L2 points such that most particles escape from the moonlet with almost zero velocity (as viewed from the centre of mass frame) and so do not change its orbital eccentricity. However, after an impacting particle becomes bound to the moonlet, conservation of the epicyclic amplitude associated with the centre of mass motion together with Eq. 11 imply that each impacting particle will reduce the eccentricity by a factor 1 − m1/m2. It is now an easy task to estimate the eccentricity damping timescale by determining the number of particle collisions per time unit with the moonlet or particles bound to it. To do that, a smooth window function Wb+c(b) is used, being unity for impact parameters b that always result in an impact with the moonlet or particles nearby that are bound to it, being zero for impact parameters that never result in an impact. The number of particles impacting the moonlet per time unit, dN/dt, is obtained by integrating over the impact parameter with the result that dN dt We note that allowing b to be negative enables impacts from both sides of the moonlet to be taken into account. Σ Ω b Wb+c(b) db. m1Z ∞ (12) 3 2 −∞ 1 = Therefore, after using Eq. 11 we find that the rate of change of the moonlet's eccentricity e2, or equivalently of it's amplitude of epicyclic motion E2, is given by 3 m2Z ∞ dE2 2 dt Σ Ω b Wb+c(b) db, = − E2 τe,collisions = − (13) E2 −∞ where τe,collisions defines the circularisation time arising from col- lisions with the moonlet. We remark that the natural unit for b is the Hill radius of the moonlet, rH = (m2/(3Mp))1/3a, so that the dimensional scaling for τe,collisions is given by τ−1 e,collisions ∝ G Σ r−1 which we find to also apply to all the processes for modifying the moonlet's eccentricity discussed below. If we assume that Wb+c(b) can be approximated by a box function, being unity in the interval [1.5rH, 2.5rH] and zero elsewhere, we get H Ω−1 (14) τ−1 e,collisions = 2.0 G Σ r−1 H Ω−1. (15) 3.2. Eccentricitydampingduetheinteractionofthemoonlet withparticlesonhorseshoeorbits The eccentricity of the moonlet manifests itself in a small oscil- lation of the moonlet about the origin. Primarily ring particles on horseshoe orbits will respond to that oscillation and damp it. This is because only those particles on horseshoe orbits spend enough time in the vicinity of the moonlet, i.e. many epicyclic periods. In appendix A, we calculate the amplitude of epicyclic mo- tion E1 f (or equivalently the eccentricity e1 f ) that is induced in a single ring particle in a horseshoe orbit undergoing a close ap- proach to a moonlet which is assumed to be in an eccentric orbit. In order to calculate the circularisation time, we have to consider all relevant impact parameters. We begin by noting that each par- ticle encounter with the moonlet is conservative and is such that for each particle, the Jacobi constant, applicable when the moon- let is in circular orbit, is increased by an amount m1Ω2E2 1 f /2 by the action of the perturbing force, associated with the eccentric- ity of the moonlet, as the particle passes by. Because the Jacobi constant, or energy in the rotating frame, for the moonlet and the particle together is conserved, the square of the epicyclic ampli- tude associated with the moonlet alone changes by E2 1 f m1/m2. Accordingly the change in the amplitude of epicyclic motion of the moonlet E2, consequent on inducing the amplitude of epicyclic motion E1 f in the horseshoe particle, is given by ∆E2 Note that this is different compared to Eq. 11. Here, we are deal- ing with a second order effect. First, the eccentric moonlet ex- cites eccentricity in a ring particle. Second, because the total epicyclic motion is conserved, the epicyclic motion of the moon- let is reduced. m1 m2 E2 1 f . 2 = − (16) Integrating over the impact parameters associated with ring particles and taking into account particles streaming by the moonlet from both directions by allowing negative impact pa- rameters gives dE2 2 dt (cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)horseshoe = −Z ∞ −∞ 3ΣE2 1 f Ω b Wd(b) db 2m2 ≡ − 2E2 2 τe,horseshoe ,(17) where τe,horseshoe is the circularisation time and, as above, we have inserted a window function, which is unity on impact pa- rameters that lead to horseshoe orbits, otherwise being zero. Using E1 f given by Eq. A.20, we obtain τ−1 e,horseshoe = 9 128 Σr2 H Ω m2 Z ∞ −∞ I2η4/3Wd(cid:16)(2ηrh)1/6(cid:17) dη. (18) 4 H. Rein and J.C.B. Papaloizou: Stochastic orbital migration of moonlets in Saturn's rings For a sharp cutoff of Wd(b) at bm = 1.5rH we find the value of the integral in the above to be 2.84. Thus, in this case we get τ−1 e,horseshoe However, note that this value is sensitive to the value of bm adopted. For bm = 1.25rH, tc is a factor of 4.25 larger. = 0.13 G Σ r−1 H Ω−1. (19) 3.3. Theeffectsofcirculatingparticles The effect of the response of circulating particles to the gravi- tational perturbation of the moonlet on the moonlet's eccentric- ity can be estimated from the work of Goldreich and Tremaine [1980] and Goldreich and Tremaine [1982]. These authors con- sidered a ring separated from a satellite such that co-orbital ef- fects were not considered. Thus, their expressions may be ap- plied to estimate effects due to circulating particles. However, we exclude their corotation torques as they are determined by the ring edges and are absent in a local model with uniform az- imuthally averaged surface density. Equivalently, one may sim- ply assume that the corotation torques are saturated. When this is done only Lindblad torques act on the moonlet. These tend to excite the moonlet's eccentricity rather than damp it. Eq. 70 Goldreich and Tremaine [1982] by the integral over impact parameter and switch to our notation to obtain dE2 ring mass replace of the We the in (20) 2 Wa(b) db, m2 Σ G2 Ω−3 b−5 E2 = 9.55Z ∞ −∞ where Wa(b) is the appropriate window function for circulating particles. Assuming a sharp cutoff of Wa(b) at bm, we can eval- uate the integral in Eq. 20 to get 1 τe,circ = − H Ω−1, = 0.183 G Σ r−1 1 E2 where we have adopted bm = 2.5rH, consistently with the simu- lation results presented below. We see that, although τe,circ < 0 corresponds to growth rather than damping of the eccentricity, it scales in the same manner as the circularisation times in Sects. 3.1 and 3.2 scale (see Eq. 14). (21) This timescale and the timescale associated with particles on horseshoe orbits, τe,horseshoe, are significantly larger than the timescale associated with collisions, given in Eq. 15. We can therefore ignore those effects for most of the discussion in this paper. 2 dt (cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)circ dt (cid:12)(cid:12)(cid:12)(cid:12)(cid:12)circ dE2 4. Processes leading to the excitation of the eccentricity of the moonlet In Sec. 3 we assumed that the moonlet had a small eccentric- ity but neglected the initial eccentricity of the impacting ring particles. When this is included, collisions and gravitational in- teractions of ring particles with the moonlet may also excite its eccentricity. 4.1. Collisionaleccentricityexcitation To see this, assume that the moonlet initially has zero eccentric- ity, or equivalently no epicyclic motion, but the ring particles do. As above we consider the conservation of the epicyclic mo- tion of the centre of mass in order to connect the amplitude of the final epicyclic motion of the combined moonlet and ring parti- cle to the initial amplitude of the ring particle's epicyclic motion. dt (cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)collisions m2! E1. m1 + m2 E1 ≃ m1 This gives the epicyclic amplitude of the centre of mass after one impact as m1 ¯E = Assuming that successive collisions are uncorrelated and occur stochastically with the mean time between consecutive encoun- ters being τce, the evolution of ¯E is governed by the equation d ¯E2 (22) (23) m2!2 = m1 1iτ−1 hE2 ce , where τ−1 ce = dN/dt can be expressed in terms of the surface density and an impact window function Wb+c(b) (see Eq. 12). The quantity hE2 1i is the mean square value of E1 for the ring particles. Using Eq. 2, this may be written in terms of the mean squares of the components of the velocity dispersion relative to the background shear, in the form 1 + 4(y1 + 3Ωx1/2)2i. 1i = Ω−2h( x2 hE2 (24) We find in numerical simulations he1i ∼ 10−6 for almost all ring parameters. This value is mainly determined by the coefficient of restitution [see e.g. Fig 4 in Morishima and Salo, 2006]. 4.2. Stochasticexcitationduetocirculatingparticles dt (cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)circulating particles Ring particles that are on circulating orbits, such as that given by path (a) in Fig. 1, exchange energy and angular momen- tum with the moonlet and therefore change its eccentricity. Goldreich and Tremaine [1982] calculated the change in the ec- centricity of a ring particle due to a satellite. We are interested in the change of the eccentricity of the moonlet induced by a ring of particles and therefore swap the satellite mass with the mass of a ring particle. Thus, rewriting their results (Eq. 64 in Goldreich and Tremaine [1982]) in our local notation without reference to the semi-major axis, we have ∆E2 (25) Supposing that the moonlet has very small eccentricity, it will receive stochastic impulses that cause its eccentricity to undergo a random walk that will result in E2 2 increasing linearly with time, so that we may write dE2 1 G2 Ω−4 b−4. 2 = 5.02 m2 (26) 2 2d(1/tb), Wa(b)∆E2 = Z ∞ −∞ where d(1/tb) is the mean encounter rate with particles which have an impact parameter in the interval (b, b + db). Wa(b) is the window function describing the band of particles in circulating orbits. Setting d(1/tb) = 3ΣΩbdb/(2m1), we obtain dE2 (27) Wa(b)m1b−3ΣG2Ω−3db. = 7.53Z ∞ −∞ For high surface densities an additional effect can excite the ec- centricity when gravitational wakes occur. The Toomre parame- ter Q is defined as 2 dt (cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)circulating particles Q = ¯vΩ πGΣ , (28) where ¯v is the velocity dispersion of the ring particles1. When the surface density is sufficiently high such that Q approaches unity, 1 The Toomre criterion used here was originally derived for a flat gaseous disc. To make use of it we replace the sound speed by the radial velocity dispersion of the ring particles. H. Rein and J.C.B. Papaloizou: Stochastic orbital migration of moonlets in Saturn's rings 5 transient clumps will form in the rings. The typical length scale of these structures is given by the critical Toomre wave length λT = 2π2GΣΩ−2 [Daisaka et al., 2001], which can be used to estimate a typical mass of the clump: mT ∼ πλ2 Whenever strong clumping occurs, mT should be used in the above calculation instead of the mass of a single particle m1: dE2 TΣ = 4π5G2Σ3Ω−4. (29) (30) W′a(b)mTb−3ΣG2Ω−3db, 2 = 7.53Z ∞ −∞ where W′a(b) ≈ Wa(b) is the appropriate window function. For typical parameters used in Sect. 7, this transition occurs at Σ ∼ 200kg/m2. dt (cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)circulating clumps 4.3. Equilibriumeccentricity Putting together the results from this and the previous section, we can estimate an equilibrium eccentricity of the moonlet. Let us assume the eccentricity e2, or the amplitude of the epicyclic motion E2, evolves under the influence of excitation and damp- ing forces as follows dE2 2 dt + τ−1 e,horseshoe 2 2 e,circ(cid:17) E2 dt (cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)circulating particles . 2 + + e,collisions + τ−1 dE2 = −2(cid:16)τ−1 dE2 dt (cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)collisions dt (cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)circulating clumps dE2 + 2 (31) The equilibrium eccentricity is then found by setting the above equation equal to zero and solving for E2. To make quantitative estimates we need to specify the win- dow functions Wa(b), Wb+c(b) and Wd(b) that determine in which impact parameter bands the particles are in (see Fig. 1). To com- pare our analytic estimates to the numerical simulations pre- sented below, we measure the window functions numerically. Alternatively, one could simply use sharp cutoffs at some multi- ple of the Hill radius. We already made use of this approxima- tion as a simple estimate in the previous sections. The results may vary slightly, but not significantly. H However, as the window functions are dimensionless, it is possible to obtain the dimensional scaling of E2 by adopting the length scale applicable to the impact parameter to be the Hill radius rH and simply assume that the window functions are of order unity. As already indicated above, all of the circularisation times scale as τe ∝ Ω rH G−1 Σ−1, or equivalently ∝ m2/(Σr2 Ω). Assuming the ring particles have zero velocity dispersion, the eccentricity excitation is then due to circulating clumps or parti- cles and the scaling of dE2 2/dt due to this cause is given by Eqs. 27 and 30 by dE2 dt ∝ mir−2 where mi is either m1 or mT . We may then find the scaling of the equilibrium value of E2 from consideration of Eqs. 31 and 13 as (33) E2 ∝ This means that the expected kinetic energy in the non circu- Ω2 which can be viewed as lar motion of the moonlet is ∝ mir2 H ΣG2Ω−3, mi m2 (32) r2 H. H 2 stating that the non circular moonlet motion scales in equiparti- tion with the mass mi moving with speed rH Ω. This speed ap- plies when the dispersion velocity associated with these masses is zero indicating that the shear across a Hill radius replaces the dispersion velocity in that limit. In the opposite limit when the dispersion velocity exceeds the shear across a Hill radius and the dominant source of eccen- tricity excitation is due to collisions, Eq. 31 gives m2E2 = m1hE2 1i so that the moonlet is in equipartition with the ring particles. Results for the two limiting cases can be combined to give an expression for the amplitude of the epicyclic motion excited in the moonlet of the form m2Ω2E2 = m1Ω2hE2 where Ci is a constant of order unity. This indicates the tran- sition between the shear dominated and the velocity dispersion dominated limits. 1i + CimiΩ2r2 H , (34) (35) 5. Processes leading to a random walk in the semi-major axis of the moonlet We have established estimates for the equilibrium eccentricity of the moonlet in the previous section. Here, we estimate the random walk of the semi-major axis of the moonlet. In contrast to the case of the eccentricity, there is no tendency for the semi- major axis to relax to any particular value, so that there are no damping terms and the deviation of the semi-major axis from its value at time t = 0 grows on average as √t for large t. Depending on the surface density, there are different effects that dominate the contributions to the random walk of the moon- let. For low surface densities, collisions and horseshoe orbits are most important. For high surface densities, the random walk is dominated by the stochastic gravitational force of circulat- ing particles and clumps. In this section, we try to estimate the strength of each effect. 5.1. Randomwalkduetocollisions m1 m1 + m2 A1 ∼ Let us assume, without loss of generality, that the guiding centre of the epicyclic motion of a ring particle is displaced from the orbit of the moonlet by A1 in the inertial frame, whereas in the co-rotating frame the moonlet is initially located at the origin with A2 = 0. When the impacting particle becomes bound to the moonlet, the guiding centre of the epicyclic motion of the centre of mass of the combined object, as viewed in the inertial frame, is then displaced from the original moonlet orbit by ∆ ¯A = This is the analogue of Eq. 22 for the evolution of the semi-major axis. For an ensemble of particles with impact parameter b, the average centre of epicyclic motion is hA1i = b. Thus we can write the evolution of d ¯A2 dt ¯A due to consecutive encounters as m2!2 = m1 1i τ−1 hA2 3 2 Z ∞ Σ Ω b3 Wb+c(b) db, 2 (cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)collisions m1 m2 A1. m1 m2 (37) (36) (38) −∞ ce = which should be compared to the corresponding expression for the eccentricity in Eq. 23. 6 H. Rein and J.C.B. Papaloizou: Stochastic orbital migration of moonlets in Saturn's rings 5.2. Randomwalkduetostochasticforcesfromcirculating particlesandclumps Particles on a circulating trajectory (see path (a) in Fig. 1) that come close the moonlet will exert a stochastic gravitational force. The largest contributions occur for particles within a few Hill radii. Thus, we can crudely estimate the magnitude of the specific gravitational force (acceleration) due to a single particle as fcp = G m1 (2rH)2 . (39) When self-gravity is important, gravitational wakes (or clumps) have to be taken into account, as was done in Sec. 4.2. Then a rough estimate of the largest specific gravitational forces due to self gravitating clumps is shear t o p l a n e t main box shear y auxiliary boxes fcc = G mT (2rH + λT )2 , (40) x where 2rH has been replaced by 2rH + λT . This allows for the fact that λT could be significantly larger than rH, in which case the approximate distance of the clump to the moonlet is λT . Following the formalism of Rein and Papaloizou [2009], we define a diffusion coefficient as D = 2τch f 2i, where f is an ac- celeration, τc is the correlation time and the angle brackets de- note a mean value. We here take the correlation time associated with these forces to be the orbital period, 2πΩ−1. This is the nat- ural dynamical timescale of the systems and has been found to be a reasonable assumption from an analysis of the simulations described below. By considering the rate of change of the en- ergy of the moonlet motion, we may estimate the random walk in ¯A due to circulating particles and self-gravitating clumps to be given by [Rein and Papaloizou, 2009]: d ¯A2 dt (41) (cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)circulating particles (cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)circulating clumps d ¯A2 dt = 4Dcp Ω−2 = 16π Ω−3h f 2 cpi (2rH)2!2+ = 16π Ω−3* G m1 = 4Dcc Ω−2 = 16π Ω−3h f 2 cci (2rH + λT )2!2+ . = 16π Ω−3* G mT (42) (43) (44) This is only a crude estimate of the random walk undergone by the moonlet. In reality several additional effects might also play a role. For example, circulating particles and clumps are clearly correlated, the gravitational wakes have a large extent in the azimuthal direction and particles that spend a long time in the vicinity of the moonlet have more complex trajectories. Nevertheless, we find that the above estimates are correct up to a factor 2 for all the simulations that we performed (see below). 5.3. Randomwalkduetoparticlesinhorseshoeorbits Finally, let us calculate the random walk induced by particles on horseshoe orbits. Particles undergoing horseshoe turns on opposite sides of the planet produce changes of opposite sign. Encounters with the moonlet are stochastic and therefore the semi-major axis will undergo a random walk. A single particle with impact parameter b will change the semi-major axis of the moonlet by ∆ ¯A = 2 m1 + m2 (45) m1 b. Fig. 2. Shearing box, simulating a small patch of a planetary ring system. The particles that leave an auxiliary box in the az- imuthal (y) direction reenter the same box on the other side and get copied into the main box at the corresponding location. All auxiliary boxes are equivalent and there are no curvature terms present in the shearing sheet. Analogous to the analysis in Sec. 5.1, the time evolution of then governed by d ¯A2 dt b2 τ−1 he (cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)horseshoe m2!2 = 4 m1 2 Z ∞ m1 m2 = 6 −∞ Σ Ω b3 Wd(b) db. ¯A is (46) (47) Note that this equation is identical to Eq. 38 except a factor 4, as particles with impact parameter b will leave the vicinity of the moonlet at −b. 6. Numerical Simulations We perform realistic three dimensional simulations of Saturn's rings with an embedded moonlet and verify the analytic esti- mates presented above. The nomenclature and parameters used for the simulations are listed in Table 1. 6.1. Methods The gravitational forces are calculated with a Barnes-Hut tree [Barnes and Hut, 1986]. The numerical scheme is similar to that used by Rein et al. [2010]. Additionally, we implemented a sym- plectic integrator for Hill's equations [Quinn et al., 2010]. This turned out to be beneficial for accurate energy conservation at almost no additional cost when the eccentricity of the moonlet (∼ 10−8) was small and integrations were undertaken over many hundreds of dynamical timescales. To further speed up the calculations, we run two coupled simulations in parallel. The first adopts the main box which in- corporates a moonlet. The second adopts an auxiliary box which initially has the same number of particles but which does not contain a moonlet. This is taken to be representative of the unper- turbed background ring. We describe this setup as enabling us to adopt pseudo shear periodic boundary conditions. We consider the main box together with eight equivalent auxiliary boxes to H. Rein and J.C.B. Papaloizou: Stochastic orbital migration of moonlets in Saturn's rings 7 Name EQ5025 EQ5050 EQ20025 EQ20050 EQ40025 EQ40050 EQ40050DT EQ5050DTW EQ20050DTW EQ40050DTW Σ 50 kg/m2 50 kg/m2 200 kg/m2 200 kg/m2 400 kg/m2 400 kg/m2 400 kg/m2 50 kg/m2 200 kg/m2 400 kg/m2 dt r2 4s 25 m 4s 50 m 4s 25 m 4s 50 m 4s 25 m 50 m 4s 50 m 40s 50 m 40s 50 m 40s 50 m 40s Lx × Ly N 1000 m × 1000 m 7.2k 1000 m × 1000 m 7.2k 1000 m × 1000 m 28.8k 1000 m × 1000 m 28.8k 1000 m × 1000 m 57.6k 1000 m × 1000 m 57.6k 1000 m × 1000 m 57.6k 2000 m × 2000 m 28.8k 2000 m × 2000 m 115.2k 2000 m × 2000 m 230.0k Table 1. Initial simulation parameters. The second column gives the surface density of the ring. The third gives the moonlet radius. The fourth column gives the time step. The fifth and sixth columns give the lengths of the main box as measured in the xy-plane and the number of particles used, respectively. be stacked as illustrated in Fig. 2. All auxiliary boxes are identi- cal copies whose centres are shifted according to the background shear as in a normal shearing sheet. On account of this motion auxiliary boxes are removed when their centres are shifted in azimuth by more than 1.5Ly from the centre of the main box and then reinserted in the same orbit on the opposite side of the main box so that the domain under consideration is prevented from shearing out. If a particle in the main box crosses one of its boundaries, it is discarded. If a particle in the auxiliary box crosses one of its boundaries, it is reinserted on the other side of this box, according to normal shear periodic boundary con- ditions. But in addition it is also copied into the corresponding location in the main box. We describe this procedure as applying pseudo shear periodic boundary conditions. In a similar calculation, Lewis and Stewart [2009] use a very long box (about 10 times longer than the boxes used in this pa- per) to ensure that particles are completely randomised between encounters with the moonlet. We are not interested in the long wavelength response that is created by the moonlet. Effects that are most important for the moonlet's dynamical evolution are found to happen within a few Hill radii. Using the pseudo shear periodic boundary conditions, we ensure that incoming particles are uncorrelated and do not contain prior information about the perturber. This setup speeds up our calculations by more than an order of magnitude. The gravity acting on a particle in the main box, which also contains the moonlet, is calculated by summing over the parti- cles in the main box and all auxiliary boxes. The gravity acting on a particle in the auxiliary box is calculated the standard way, by using ghost boxes which are identical copies of the auxiliary box. Tests have indicated that our procedure does not introduce unwanted fluctuations in the gravitational forces. The moonlet is allowed to move freely in the main box. However, in order to prevent it leaving the computational do- main, as soon as the moonlet has left the innermost part (de- fined as extending one eighth of the box size), all particles are shifted together with the box boundaries, such that the moonlet is returned to the centre of the box. This is possible because the shearing box approximation is invariant with respect to transla- tions in the xy plane (see Eq. 1). Collisions between particles are resolved using the instan- taneous collision model and a velocity dependent coefficient of restitution given by Bridges et al. [1984]: ǫ(v) = min 0.34 · 1cm/s!−0.234 vk , , 1 (48) Fig. 3. Snapshots of the particle distribution and the moonlet (blue) in simulation EQ40050DTW. The wake is much clearer than in Fig. 4 as the box size is twice as large. where vk is the impact speed projected on the axis between the two particles. The already existing tree structure is reused to search for nearest neighbours [Rein et al., 2010]. 6.2. Initialconditionsandtests Throughout this paper, we use a distribution of particle sizes, r1, ranging from 1m to 5m with a slope of q = −3. Thus dN/dr1 ∝ r−q 1 . The density of both the ring particles and the moonlet is taken to be 0.4 g/cm2. This is at the lower end of what has been assumed reasonable for Saturn's A ring [Lewis and Stewart, 2009]. The moonlet radius is taken to be either 50 m or 25 m. We found that using a larger ring particle and moonlet density only leads to more particles being bound to the moonlet. This effectively increases the mass of the moonlet (or equivalently its Hill radius) and can therefore be easily scaled to the formalism presented here. Simulating a gravitational aggregate of this kind is computationally very expensive, as many more collisions have to be resolved each time-step. 8 H. Rein and J.C.B. Papaloizou: Stochastic orbital migration of moonlets in Saturn's rings (a) Σ = 200 kg/m2, r2 = 25 m (b) Σ = 200 kg/m2, r2 = 50 m (c) Σ = 400 kg/m2, r2 = 25 m (d) Σ = 400 kg/m2, r2 = 50 m Fig. 4. Snapshots of the particle distribution and the moonlet (blue) for different surface densities after 25 orbits. The wake created by the moonlet is much longer than the size of the box and therefore hardly visible in these images. H. Rein and J.C.B. Papaloizou: Stochastic orbital migration of moonlets in Saturn's rings 9 The initial velocity dispersion of the particles is set to 1 mm/s for the x and y components and 0.4 mm/s for the z component. The moonlet is placed at a semi-major axis of a = 130000 km, corresponding to an orbital period of P = 13.3 hours. Initially, the moonlet is placed in the centre of the shearing box on a circular orbit. We run simulations with a variety of moonlet sizes, sur- face densities and box sizes. The dimensions of each box, as viewed in the (x, y) plane, are specified in Table 1. Because of the small dispersion velocities, the vertical motion is automati- cally strongly confined to the mid-plane. After a few orbits, the simulations reach an equilibrium state in which the velocity dis- persion of particles does not change anymore. We ran several tests to ensure that our results are converged. Simulation EQ40050DT uses a ten times larger time step than simulation EQ40050. Simulations EQ5050DTW, EQ20050DTW and EQ40050DTW use a box that is twice the size of that used in simu- lation EQ40050. Therefore, four times more particles have been used. No differences in the equilibrium state of the ring particles, nor in the equilibrium state of the moonlet have been observed in any of those test cases. A snapshot of the particle distribution in simulation EQ40050DTW is shown in Fig. 3. Snapshots of simulations EQ20025, EQ20050, EQ40025 and EQ40050, which use the smaller box size, are shown in Fig. 4. In all these cases, the length of the wake that is created by the moonlet is much longer than the box size. Nevertheless, the pseudo shear peri- odic boundary conditions allow us to simulate the evolution of the moonlet accurately. 7. Results 7.1. Impactbands The impact band window functions Wa(b), Wb+c(b) and Wd(b) are necessary to estimate the damping timescales and the strength of the excitation. To measure those, each particle that enters the box is labelled with it's impact parameter. The possi- ble outcomes are plotted in Fig. 1: (a) being a circulating parti- cle which leaves the box on the opposite side, (b) representing a collision with other ring particles close to the moonlet where the maximum distance from the centre of the moonlet has been taken to be twice the moonlet radius, (c) being a direct collision with the moonlet and finally (d) showing a horseshoe orbit in which the particle leaves the box on the same side that it entered. The impact band (b) is considered in addition to the impact band (c) because some ring particles will collide with other ring particles that are (temporarily) bound to the moonlet. Thus, these collisions take part in the transfer the energy and momentum to the moonlet. Actually, if the moonlet is simply a rubble pile of ring particles as suggested by Porco et al. [2007], then there might be no solid moonlet core and all collisions are in the im- pact band (b). Fig. 5 shows the impact bands, normalised to the Hill ra- dius of the moonlet rH for simulations EQ20050 and EQ200252. For comparison, we also plot sharp cutoffs, at 1.5rH and 2.5rH. Our results show no sharp discontinuity because of the velocity dispersion in the ring particles which is ∼ 5 mm/s. This corre- sponds to an epicyclic motion of ∼ 40 m or equivalently ∼ 0.8 rH and ∼ 1.6 rH (for r2 = 50m and r2 = 25m, respectively), which 2 Pi=a,b,c,d Wi & 1 because all particles are shifted from time to time when the moonlet is to far away from the origin (see Sect. 6.1). In this process, particles with the same initial impact parameter might become associated with the more than one impact band. Name Numerical results EQ5025 EQ5050 EQ20025 EQ20050 EQ40025 EQ40050 18.9 27.6 9.0 12.9 2.8 5.5 Analytic results collisions 18.4 33.8 3.9 8.5 2.1 4.4 horseshoe 1712 3424 428 856 214 428 Table 2. Eccentricity damping timescale τe of the moonlet in units of the orbital period. The second column lists the simula- tion results. The third and fourth column list the analytic esti- mates of collisional and horseshoe damping timescales, respec- tively. explains the cutoff width of approximately one Hill radii. The impact bands are almost independent of the surface density and depend only on the Hill radii and the mean epicyclic amplitude of the ring particles, or, more precisely, the ratio thereof. 7.2. Eccentricitydampingtimescale To measure the eccentricity damping timescale, we first let the ring particles and the moonlet reach an equilibrium and integrate them for 200 orbits. We then change the velocity of the moonlet and the ring particles within 2rH. The new velocity corresponds to an eccentricity of 6 · 10−7, which is well above the equilib- rium value. We then measure the decay timescale τe by fitting a function of the form ed(t) = heeqi +(cid:16)6 · 10−7 − heeqi(cid:17) e−t/τe . The results are given in Table 2. They are in good agreement with the estimated damping timescales from Sect. 3.1 and 3.2, show- ing clearly that the most important damping process, in every simulation considered here, is indeed through collisional damp- ing as predicted by comparing Eq. 15 with Eqs. 19 and 21. (49) 7.3. Meanmoonleteccentricity The mean eccentricity of the moonlet is measured in all simula- tions after several orbits when the ring particles and the moonlet have reached an equilibrium state. To compare this value with the estimates from Sect. 2 we set Eq. 31 equal to zero and use the analytic damping timescale listed in Table 2. The analytic es- timates of the equilibrium eccentricity are calculated for each ex- citation mechanism separately to disentangle their effects. They are listed in the third, fourth and fifth column in Table 3. The sixth column lists the analytic estimate for the mean eccentricity using the sum of all excitation mechanisms. For all simulations, the estimates are correct within a factor of about 2. For low surface densities, the excitation is dominated by individual particle collisions. For larger surface densities, it is dominated by the excitation due to circulating self gravitat- ing clumps (gravitational wakes). The estimates and their trends are surprisingly accurate, as we have ignored several effects (see below). 7.4. Randomwalkinsemi-majoraxis The random walk of the semi-major axis a (or equivalently the centre of epicyclic motion A in the shearing sheet) of the moon- let in the numerical simulations are measured and compared to 10 H. Rein and J.C.B. Papaloizou: Stochastic orbital migration of moonlets in Saturn's rings Wd(b) 1 n o i t c a r f 0.75 0.5 0.25 0 0 r2=50m Wa(b) Wd(b) r2=25m Wa(b) Wb+c(b) Wc(b) Wb(b) Wb+c(b) Wc(b) Wb(b) 1 2 3 4 5 0 1 2 3 4 5 impact parameter [rHill] impact parameter [rHill] Fig. 5. Impact bands for simulation EQ20050 (left) and EQ20025 (right). The vertical lines indicate sharp cutoffs at 1.5rH and 2.5rH. the analytic estimates presented in Sect. 5. We ran one simula- tion per parameter set. To get a statistically meaningful expres- sion for the average random walk after a given time A(∆t), we average all pairs of A(t) and A(t′) for which t − t′ = ∆t. In other words, we assume the system satisfies the Ergodic hypothesis. A(∆t) then grows like √∆t and we can fit a simple square root function. This allows us to accurately measure the average growth in A after ∆t = 100 orbits by running one simulation for a long time and averaging over time, rather than running many simulations and performing an ensemble average. The measured values are given in the second column of Table 4. The values that correspond to the analytic expressions in Eqs. 47, 38, 42 and 44 are listed in columns three, four, five and six, respectively. For low surface densities, the evolution of the random walk is dominated by collisions and the effect of particles on horse- shoe orbits. For large surface densities, the main effect comes from the stochastic gravitational force due to circulating clumps (gravitational wakes). 7.5. Longtermevolutionandobservability The change in semi-major axis is relatively small, a few tens of meters after 100 orbits (= 50 days). The change in longitude that corresponds to this is, however, is much larger. We there- fore extend the discussion of Rein and Papaloizou [2009], who derive the time evolution of orbital parameters undergoing a ran- dom walk, to the time evolution of the mean longitude. This is of special interest in the current situation, as in observations of Saturn's rings, it is much easier to measure the shift in the longi- tude of an embedded moonlet than the change in its semi-major axis. The time derivative of the mean motion is given by n = − 3 2rGM a5 a = − 3 Fθ a , (50) where Fθ is the time dependent stochastic force in the θ (or in shearing sheet coordinates, y) direction and we have assumed a nearly circular orbit, e ≪ 1. The mean longitude is thus given by ] m k [ t e s f f o l a h t u m z a i 200 150 100 50 0 -50 -100 -150 -200 Σ=50kg/m2 r2=25m Σ=50kg/m2 r2=50m Σ=200kg/m2 r2=25m Σ=200kg/m2 r2=50m Σ=400kg/m2 r2=25m Σ=400kg/m2 r2=50m 0 0.1 0.2 0.3 0.4 0.5 time [years] Fig. 6. Offset in the azimuthal distance due to the random walk in the semi-major axis a measured in simulations EQ5025, EQ5050, EQ20025, EQ20050, EQ40025 and EQ40050. Individual moon- lets may show linear, constant or oscillatory growth. On average, the azimuthal offset grows like t3/2. the double integral λ(t) = Z t 0 (cid:0)n0 + ∆n(t′)(cid:1) dt′ 3Fθ(t′′) = n0t −Z t 0 Z t′ 0 a dt′′dt′. (51) (52) The root mean square of the difference compared to the unper- turbed orbit (n0) is (∆λ)2 = D(λ(t) − nt)2E t, t′, t, t′′′ = = 0 & 9DF2 θEa2 9Fθ(t′′)Fθ(t′′′′) a2 dt′′′′dt′′′dt′′dt′ t, t′, t, t′′′ & 0 g(t′′ − t′′′′) dt′′′′dt′′′dt′′dt′ (53) (54) (55) H. Rein and J.C.B. Papaloizou: Stochastic orbital migration of moonlets in Saturn's rings 11 Name Numerical results EQ5025 EQ5050 EQ20025 EQ20050 EQ40025 EQ40050 6.1 · 10−8 4.3 · 10−8 6.4 · 10−8 4.9 · 10−8 11.3 · 10−8 7.2 · 10−8 collisions 4.5 · 10−8 1.6 · 10−8 4.6 · 10−8 1.6 · 10−8 4.6 · 10−8 1.6 · 10−8 Analytic results circulating particles 1.9 · 10−8 1.4 · 10−8 1.7 · 10−8 1.4 · 10−8 2.5 · 10−8 1.5 · 10−8 circulating clumps 0.3 · 10−8 0.2 · 10−8 2.0 · 10−8 1.6 · 10−8 8.4 · 10−8 4.9 · 10−8 total 4.9 · 10−8 2.1 · 10−8 5.3 · 10−8 2.7 · 10−8 9.9 · 10−8 5.4 · 10−8 Table 3. Equilibrium eccentricity heeqi of the moonlet. The second column lists the simulation results. The third, fourth and fifth column list the analytic estimates of the equilibrium eccentricity assuming a single excitation mechanism. The last column lists the analytic estimate of the equilibrium eccentricity summing over all excitation mechanisms. Name Numerical results EQ5025 EQ5050 EQ20025 EQ20050 EQ40025 EQ40050 22.7m 12.6m 38.1m 22.7m 78.1m 45.6m horseshoe 9.0m 4.5m 18.1m 13.6m 25.6m 12.8m collisions 10.5m 4.8m 25.1m 9.6m 36.5m 13.4m Analytic results circulating particles 17.6m 4.4m 17.6m 4.4m 17.6m 4.4m total circulating clumps 22.4m 0.3m 7.9m 0.1m 38.9m 15.7m 4.8m 14.7m 86.6m 100.7m 31.4m 36.7m Table 4. Change in semi-major axis of the moonlet after 100 orbits, A(∆t = 100 orbits). The second column lists the simula- tion results. The third till sixth columns list the analytic estimates of the change in semi-major axis assuming a single excitation mechanism. The last column lists the total estimated change in semi-major axis summing over all excitation mechanisms. = 9DF2 θEa2 −2τ4 +(cid:16)2τ3t + 2τ4 + τ2t2(cid:17) e−t/τ + 1 3 τt3! (56) where g(t) is the auto-correlation function of Fθ, which has been assumed to be exponentially decaying with an auto-correlation time τ, thus g(t) = exp(−t/τ). In the limit t ≫ τ, the above equation simplifies to (∆λ)2 = 3DF2 θE τ a2 t3 = (∆n)2 t2 6 . (57) On shorter timescale, the other terms in Eq. 56 are important, leading to linear and oscillatory behaviour. For uncorrelated noise (e.g. collisions) Eq. 57 is replaced by (∆λ)2 = 3 h∆vi2 a2 t3/τ, (58) where h∆vi is the average velocity change per impulsive event and τ is taken to be the average time between consecutive events. Instead of a stochastic migration, let us also assume a lami- nar migration with constant migration rate τa, defined by τa = a a 3 2 n n . = − (59) As above, we can calculate the longitude as a function of time as the following double integral λ(t) = Z t 0 (cid:0)n0 + ∆n(t′)(cid:1) dt′ = n0t −Z t 0 Z t′ 0 3 2 n τa dt′′dt′ = n0t − and thus (∆λ)2 = 9 16 n2 τ2 a t4 = (∆n)2 t2 4 . 3 4 n τa t2 (60) (61) (62) Note that Eq. 62 is an exact solution, whereas Eq. 57 is a statisti- cal quantity, describing the root mean square value in an ensem- ble average. In the stochastic and laminar case, (∆λ)2 grows like ∼ t3 and ∼ t4, respectively. This behaviour allows to discrim- inate between a stochastic and laminar migration by observing ∆λ over an extended period of time. Results from individual simulations (i.e. not an ensemble av- erage) are plotted in Fig. 6. Here, ∆λ is expressed in terms of the azimuthal offset relative to a Keplerian orbit. One can see that for an individual moonlet, the shift in azimuth can appear to be linear, constant or oscillating on short timescales (see curve for simulation EQ20025). This is partly because of the lower order terms in Eq. 56. Though, on average, the azimuthal offset grows very rapidly, as ∼ t3/2 (see Eq. 57) for t ≫ τ. 8. Conclusions In this paper, we have discussed the dynamical response of an embedded moonlet in Saturn's rings to interactions with ring particles both analytically and by the use of realistic three dimen- sional many particle simulations. Both the moonlet and the ring particle density were taken to be 0.4 g/cm2. Moonlets of radius 25 m and 50 m were considered. Particle sizes ranging between 1 m and 5 m were adopted. We estimated the eccentricity damping timescale of the moonlet due to collisions with ring particles and due to the response of ring particles to gravitational perturbations by the moonlet analytically. We found the effects due to the response of particles on horseshoe and circulating orbits are negligible. On the other hand the stochastic impulses applied to the moonlet by circulating particles were found to cause the square of the eccen- tricity to grow linearly with time as did collisions with particles with non zero velocity dispersion. A balance between excitation and damping processes then leads to an equilibrium moonlet ec- centricity. We also estimated the magnitude of the random walk in the semi-major axis of the moonlet induced by collisions with indi- vidual ring particles and the gravitational interaction with parti- 12 H. Rein and J.C.B. Papaloizou: Stochastic orbital migration of moonlets in Saturn's rings cles and gravitational wakes. There is no tendency for the semi- major axis to relax to any particular value, so that there are no damping terms. The deviation of the semi-major axis from its value at time t = 0 grows on average as √t for large t. From our simulations we find that the evolution of the eccen- tricity is indeed dominated by collisions with ring particles. For large surface densities (more than 200kg/m2) the effect of grav- itational wakes also becomes important, leading to an increase in the mean steady state eccentricity of the moonlet. When the particle velocity dispersion is large compared to ΩrH, we obtain approximate energy equipartition between the moonlet and ring particles as far as epicyclic motion is concerned. Similarly, the random walk in the semi-major axis was found to be dominated by collisions for low surface densities and by gravitational wakes for large surface densities. In addition we have shown that on average, the difference in longitude ∆λ of a stochastically forced moonlet grows with time like t3/2 for large t. The distance travelled within 100 orbits (50 days at a dis- tance of 130000km) is, depending on the precise parameters, of the order of 10-100m. This translates to a shift in longitude of several hundred kilometres. The shift in λ is not necessarily monotonic on short timescales (see Fig. 6). We expect that such a shift should be easily detectable by the Cassini spacecraft. And indeed, Tiscareno et al. [2010] report such an observation, al- though the data is not publicly available yet, at the time when this paper was submitted. There are several effects that have not been included in this study. The analytic discussions in Sec. 3, Sec. 4 and Sec. 5 do not model the motion of particles correctly when their trajectories take them very close to the moonlet. Material that is temporarily or permanently bound to the moonlet is ignored. The density of both the ring particles and the moonlet are at the lower end of what is assumed to be reasonable [Lewis and Stewart, 2009]. This has the advantage that for our computational setup only a few particles (with a total mass of less than 10% of the moonlet) are bound to the moonlet at any given time. In a future study, we will extend the discussion presented in this paper to a larger variety of particle sizes, moonlet sizes and densities. Acknowledgements. We thank Aur´elien Crida for reading an early draft of this paper. Hanno Rein was supported by an Isaac Newton Studentship and St John's College Cambridge. Simulations were performed on the computational facili- ties of the Astrophysical Fluid Dynamics Group and on Darwin, the Cambridge University HPC facility. with the cylindrical coordinates of the particle and moonlet be- ing (r, φ) and (R, Φ) respectively. This may be expanded correct to first order in R/r in the form Gm1m2 Gm1m2R cos(φ − Φ) . r r2 − Ψ1,2 = − The moonlet undergoes small amplitude epicyclic oscillations such that X = E2 cos(Ωt + ǫ), Y = −2E2 sin(Ωt + ǫ) where e is its small eccentricity and ǫ is an arbitrary phase. Then Ψ1,2 may be written as (A.4) Gm1m2 r − Ψ′1,2, Ψ1,2 = − where (A.5) Gm1m2r E2(x cos(Ωt + ǫ) − 2y sin(Ωt + ǫ)) Ψ′1,2 = − gives the part of the lowest order interaction potential associated with the eccentricity of the moonlet. (A.6) r3 We here view the interaction of a ring particle with the moon- let as involving two components. The first, due to the first term on the right hand side of Eq. A.4 operates when E2 = 0 and re- sults in standard horseshoe orbits for ring particles induced by a moonlet in circular orbit. The second term, Eq. A.6, perturbs this motion when E2 is small. We now consider the response of a ring particle undergoing horseshoe motion to this perturbation. In doing so we make the approximation that the variation of the leading order potential Eq. A.4 due to the induced ring particle perturbations may be neglected. This is justified by the fact that the response induces epicyclic oscillations of the particle which are governed by the dominant central potential. A.2. Responsecalculation Setting x → x + ξx, y → y + ξy, where ξ is the small response displacement induced by Eq. A.6 and linearising Eqs. A.2, we obtain the following equations for the components of ξ d2ξx dt2 − 2Ω d2ξy dt2 + 2Ω dξy dt dξx dt ∂Ψ′1,2 ∂x and 1 = 3Ω2ξx − m1 ∂Ψ′1,2 ∂y = − 1 m1 , . (A.7) (A.8) Appendix A: Response calculation of particles on From these we find a single equation for ξx in the form horseshoe orbits A.1. Interactionpotential d2ξx dt2 + Ω2ξx = − 1 m1 ∂Ψ′1,2 ∂x −Z 1 m1 ∂Ψ′1,2 ∂y dt = F. (A.9) Due to some finite eccentricity, the moonlet undergoes a small oscillation about the origin. Its Cartesian coordinates then be- come (X, Y), with these being considered small in magnitude. The components of the equation of motion for a ring particle with Cartesian coordinates (x, y) ≡ r1 are d2x and dt2 − 2Ω d2y + 2Ω dt2 where the interaction gravitational potential due to the moonlet is 1 m1 ∂Ψ1,2 = 3Ω2x − dy dt dx dt ∂Ψ1,2 ∂x = − 1 m1 (A.1) (A.2) ∂y , Ψ1,2 = − Gm1m2 (r2 + R2 − 2rR cos(φ − Φ))1/2 (A.3) When performing the time integral on the right hand side of the above, as we are concerned with a potentially resonant epicyclic response, we retain only the oscillating part. The solution to Eq. A.9 which is such that ξ vanishes in the distant past (t = −∞) when the particle is far from the moonlet may be written as ξx = α cos(Ωt) + β sin(Ωt), where F sin(Ωt) Ω dt and α = −Z t β = Z t −∞ −∞ F cos(Ωt) dt. Ω (A.10) (A.11) (A.12) H. Rein and J.C.B. Papaloizou: Stochastic orbital migration of moonlets in Saturn's rings 13 After the particle has had its closest approach to the moonlet and moves to a large distance from it it will have an epicyclic oscillation with amplitude and phase determined by Ω Ω dt. −∞ −∞ (A.13) dt and F sin(Ωt) F cos(Ωt) α∞ = −Z ∞ β∞ = Z ∞ In evaluating the above we note that, although F vanishes when the particle is distant from the moonlet at large t, it also oscil- lates with the epicyclic angular frequency Ω which results in a definite non zero contribution. This is the action of the co-orbital resonance. It is amplified by the fact that the encounter of the particle with the moonlet in general occurs on a horseshoe libra- tion time scale which is much longer than Ω−1. We now consider this unperturbed motion of the ring particles. (A.14) A.3. Unperturbedhorseshoemotion The equations governing the unperturbed horseshoe motion are Eqs. A.2 with Ψ1,2 = Ψ0 1,2 Gm1m2 r . = − (A.15) We assume that this motion is such that x varies on a time scale much longer than Ω−1 so that we may approximate the first of these equations as x = −2/(3Ω)(dy/dt). Consistent with this we also neglect x in comparison to y in Eq. A.15. From the second equation we than find that d2y dt2 = 3 m1 ∂Ψ0 1,2 ∂y , which has a first integral that may be written (A.16) , + 4 = − (A.17) 9Ω2b2 6Gm2 y dt!2 dy where as before b is the impact parameter, or the constant value of x at large distances from the moonlet. The value of y for which the horseshoe turns is then given by y = y0 = 24Gm2/(9Ω2b2). At this point we comment that we are free to choose the origin of time t = 0 such that it coincides with the closest approach at which y = y0. Then in the approximation we have adopted, the horseshoe motion is such that y is a symmetric function of t. A.4. Evaluationoftheinducedepicyclicamplitude Using Eqs. A.6, A.9 and A.17 we may evaluate the epicyclic amplitudes from Eq. A.14. In particular we find (3x2 + 12y2) ! . 5 − Gm2 E2 cos(Ωt + ǫ) r3 r2 (A.18) F = − When evaluating Eq. A.14, consistent with our assumption that the epicyclic oscillations are fast, we average over an orbital pe- riod assuming that other quantities in the integrands are fixed, and use Eq. A.17 to express the integrals with respect to t as integrals with respect to y. We then find α∞ = −A'∞ sin ǫ and β∞ = −A'∞ cos ǫ where y0 p6Gm2y0 E2 A∞ = Z ∞ 6Ωr3p1 − y0/y Ω2 5 − (cid:16) 8Gm2 · y − 1 (cid:16) 1 r2 y0(cid:17) + 12y2(cid:17)  dy, (A.19) 0 y0 y2 0r−3 Ω2 I, Gm2 E2 y3 0 with r2 = 8Gm2/(3Ω2y0)(1− y0/y) + y2. From this we may write (A.20) A∞ = E1 f = where E1 f is the final epicyclic motion of the ring particle and the dimensionless integral I is given by I = s Ω2y3 6Gm2 Z ∞ 5 − (cid:16) 8Gm2 · We remark that the dimensionless quantity η = 8Gm2/(3Ω2y3 0) = 8(rH/y0)3, with rH being the Hill radius of the moonlet. It is re- lated to the impact parameter b by η = 2−6(b/rH)6. Thus for an impact parameter amounting to a few Hill radii, in a very ap- proximate sense, η is of order unity, y0 is of order rH and the induced eccentricity E1 f is of order E2. References J. Barnes and P. Hut. A Hierarchical O(NlogN) Force-Calculation Algorithm. p1 − y0/y y − 1 (cid:16) 1 r2 y0(cid:17) + 12y2(cid:17) (A.21)  dy. Ω2 Nature, 324:446 -- 449, December 1986. F. G. Bridges, A. Hatzes, and D. N. C. Lin. Structure, stability and evolution of Saturn's rings. Nature, 309:333 -- 335, May 1984. A. Crida, J. Papaloizou, H. Rein, S. Charnoz, and J. Salmon. Migration of a moonlet in a ring of solid particles: Theory and application to saturn's pro- pellers. AJ, 2010. Submitted. H. Daisaka, H. Tanaka, and S. Ida. Viscosity in a Dense Planetary Ring with Self-Gravitating Particles. Icarus, 154:296 -- 312, December 2001. P. Goldreich and S. Tremaine. Disk-satellite interactions. ApJ, 241:425 -- 441, October 1980. P. Goldreich and S. Tremaine. The dynamics of planetary rings. ARA&A, 20: 249 -- 283, 1982. M. C. Lewis and G. R. Stewart. Features around embedded moonlets in Saturn's Icarus, 199: rings: The role of self-gravity and particle size distributions. 387 -- 412, February 2009. R. Morishima and H. Salo. Simulations of dense planetary rings. IV. Spinning self-gravitating particles with size distribution. Icarus, 181:272 -- 291, March 2006. C. C. Porco, P. C. Thomas, J. W. Weiss, and D. C. Richardson. Saturns Small Inner Satellites: Clues to Their Origins. Science, 318:1602 -- , December 2007. T. Quinn, R. P. Perrine, D. C. Richardson, and R. Barnes. A Symplectic Integrator for Hill's Equations. AJ, 139:803 -- 807, February 2010. H. Rein, G. Lesur, and Z. M. Leinhardt. The validity of the super-particle approx- imation during planetesimal formation. A&A, 511:A69+, February 2010. H. Rein and J. C. B. Papaloizou. On the evolution of mean motion resonances through stochastic forcing: fast and slow libration modes and the origin of HD 128311. A&A, 497:595 -- 609, April 2009. M. Seiss, F. Spahn, M. Sremcevi´c, and H. Salo. Structures induced by small moonlets in Saturn's rings: Implications for the Cassini Mission. Geophys. Res. Lett., 32:11205 -- +, June 2005. F. Spahn and M. Sremcevi´c. Density patterns induced by small moonlets in Saturn's rings? A&A, 358:368 -- 372, June 2000. M. Sremcevi´c, F. Spahn, and W. J. Duschl. Density structures in perturbed thin cold discs. MNRAS, 337:1139 -- 1152, December 2002. M. S. Tiscareno, J. A. Burns, M. M. Hedman, and C. C. Porco. The Population of Propellers in Saturn's A Ring. AJ, 135:1083 -- 1091, March 2008. M. S. Tiscareno, J. A. Burns, M. M. Hedman, C. C. Porco, J. W. Weiss, L. Dones, D. C. Richardson, and C. D. Murray. 100-metre-diameter moonlets in Saturn's A ring from observations of propeller structures. Nature, 440:648 -- 650, March 2006. M. S. Tiscareno, J. A. Burns, M. Sremcevic, K. Beurle, M. M. Hedman, N. J. Cooper, A. J. Milano, M. W. Evans, C. C. Porco, J. N. Spitale, and J. W. Weiss. Directly Observing the Orbital Evolution of Disk-embedded Masses. In Bulletin of the American Astronomical Society, volume 41 of Bulletin of the American Astronomical Society, pages 939 -- +, May 2010.
1501.04189	1	1501	2015-01-17T11:10:46	Gravitational quantization of exoplanet orbits in HD 40307, $\mu$ Ara, Kepler-26, Kepler-62, and Kepler-275: Comparing predicted orbits	[ "astro-ph.EP" ]	According to the so-called "global polytropic model", hydrostatic equilibrium for a planetary system leads to the Lane-Emden differential equation. Solving this equation in the complex plane, we obtain polytropic spherical shells defined by succesive roots of the real part $\mathrm{Re}(\theta)$ of the Lane-Emden function $\theta$. Such shells provide hosting orbits for the members of a planetary system. Within the framework of the global polytropic model, we study the exoplanet systems HD 40307, $\mu$ Ara, Kepler-26, Kepler-62, and Kepler-275. We give emphasis on comparing our results with observations, and with orbit predictions derived from the "generalized Titius-Bode relation".	astro-ph.EP	astro-ph	Gravitational quantization of exoplanet orbits in HD 40307, µ Ara, Kepler-26, Kepler-62, and Kepler-275: Comparing predicted orbits Vassilis S. Geroyannis Department of Physics, University of Patras, Greece [email protected] June 10, 2021 Abstract According to the so-called "global polytropic model", hydrostatic equi- librium for a planetary system leads to the Lane -- Emden differential equa- tion. Solving this equation in the complex plane, we obtain polytropic spherical shells defined by succesive roots of the real part Re(θ) of the Lane -- Emden function θ. Such shells provide hosting orbits for the mem- bers of a planetary system. Within the framework of the global poly- tropic model, we study the exoplanet systems HD 40307, µ Ara, Kepler-26, Kepler-62, and Kepler-275. We give emphasis on comparing our results with observations, and with orbit predictions derived from the "general- ized Titius -- Bode relation". Keywords: exoplanets; global polytropic model; planets: orbits; quan- tized orbits; stars: individual (HD 40307, µ Ara, Kepler-26, Kepler-62, Kepler-275) 1 Introduction Gravitational quantization of orbits in systems of planets or satellites has been considered in three recent investigations ([1], [2], [3]). In the framework of classical mechanics, such systems are assumed to be in hydrostatic equilibrium, governed by the Lane -- Emden differential equation. We solve this equation in the complex plane applying the so-called "complex plane strategy" [5]. The resulting solution is the complex Lane -- Emden function θ. Polytropic spherical shells defined by succesive roots of the real part Re(θ) seem to provide hosting orbits for planets or satellites. In fact, there is only one parameter to be adjusted for a particular polytropic configuration defined by θ: the polytropic index n of the central body. Alternative studies regarding quantization of orbits of planets or satellites are cited and discussed in [1] and [2]. 1 In the present paper, we emphasize on comparing our results with: (i) cor- responding observations, and (ii) orbit predictions of undetected exoplanets, calculated by the so-called "generalized Titius -- Bode relation" ([4], Secs. 1 -- 3 and references therein; the method is analyzed in Sec. 4). Table 2 of [4] shows orbit predictions for 29 exoplanet systems; in this table, there are both "interpo- lated predictions" concerning undetected planets with distances less than that of the outermost planet, and "extrapolated predictions" regarding undetected planet(s) being next to the outermost detected one(s). For the purpose of ex- amining, as best we can, the issue (ii) above, we focus on 5 exoplanet systems with the larger numbers of predictions: HD 40307 (4+1 predictions), µ Ara (4+1 predictions), Kepler-26 (5+1 predictions), Kepler-62 (6+1 predictions), and Kepler-275 (6+1 predictions). 2 Polytropic Models Simulating Host Stars, and Polytropic Shells Hosting Exoplanets Preliminary concepts for the global polytropic model can be found in [1] (Sec. 3). The complex-plane strategy and the complex Lane -- Emden function θ are anal- ysed in [5] (Sec. 3.1). We will use hereafter the definitions and symbols adopted in [1]. The real part ¯θ(ξ) of θ(ξ) has a first root ξ1 = ¯ξ1 + i ξ0, which is the radius of the central body. Next to ξ1, there is a second root ξ2 = ¯ξ2 + i ξ0 with ¯ξ2 > ¯ξ1, a third root ξ3 = ¯ξ3 + i ξ0 with ¯ξ3 > ¯ξ2, etc. The polytropic spherical shells S2, S3, . . . , are defined by the pairs ( ¯ξ1, ¯ξ2), ( ¯ξ2, ¯ξ3), . . . , respectively. A shell Sj can be considered as a place appropriate for hosting a planet. For this purpose, the most proper orbit radius αj ∈ [ ¯ξj−1, ¯ξj] is that at which ¯θ takes its maximum value inside Sj, max¯θ[Sj] = ¯θ(αj + i ξ0). (1) In the case of two or three planets hosted inside the same shell Sj, there are two further proper orbits with radii αLj and αRj, such that αLj < αj < αRj, at which ¯θ becomes equal to its average value inside Sj, avg¯θ[Sj] = ¯θ(αLj + i ξ0) = ¯θ(αRj + i ξ0). (2) So, planets inside Sj can be hosted on orbits with radii αLj, αj, and αRj. In [2] (Sec. 2), we have developed an algorithm for computing the optimum polytropic index nopt of a star, which hosts a number of planets. This algorithm, called A[n], can be applied to NP members P1, P2, . . . , PNP of a system with NP prescribed distances A1 < A2 < · · · < ANP from the central body. The host stars of the exoplanet systems HD 40307 ([6], [7], [8]), µ Ara ([9], [10], [11]), Kepler-26 (KOI-250; [12], [13]), Kepler-62 (KOI-701; [14], [15], [16]), and Kepler-275 (KOI-1198; [17]) are Sun-like stars. We expect therefore that the optimum values of the polytropic index n for modeling such stars are about 2 n ∼ 3 ([18], Sec. 6.1 and references therein; [1], Sec 3 and references therein). Accordingly, we apply the algorithm A[n] to an array {ni} with elements ni = 2.500 + 0.001 (i − 1), i = 1, 2, . . . , 1001. (3) The 1001 complex IVPs, counted in this equation, are solved by the Fortran code DCRKF54 [19], which is a Runge -- Kutta -- Fehlberg code of fourth and fifth order modified for solving complex IVPs of high complexity in their ODEs, along contours prescribed as continuous chains of straight-line segments (for details on the usage of DCRKF54, see [1], Sec. 4; see also [3], Sec. 3). We will hereafter quote the real parts of the complex orbit radii, since these quantities have the physical interest in this study. For simplicity, we will drop overbars denoting these real parts. 3 Numerical Results and Discussion All astrophysical data used for comparing our results with exoplanet observa- tions are from the site http://exoplanet.eu. In the following tables, the first root ξ1 of θ, coinciding with the radius of the host star, is expressed in both "classical polytropic units" (cpu) -- in such units, the length unit is equal to the polytropic parameter α ([1], Eq. (3b)) -- and solar radii R⊙. All other orbit radii are expressed in AU. 3.1 The HD 40307 system Results for the HD 40307 system are shown in Table 1. For this system, the minimum sum of absolute percent errors is ∆min(cid:18)nopt(HD 40307) = 2.531; qb = 5, qc = 6, qd = 8, qe = 9, qf = 10, qg = 15(cid:19) ≃ 37.9. (4) Smaller error is that for g's distance, ≃ 2.2%, and larger one is that for f's distance, ≃ 16%. For the most massive planet, d, the error is ≃ 3.3%. The average error for the computed orbit radii of the 6 planets in HD 40307 is ≃ 6.3%. Regarding the large error quoted for f's distance, an interesting conjecture is to associate this distance with the mean-density orbit αR10 = 0.2274 AU, provided that the maximum-density orbit α10 is already occupied by another planet not yet observed. In view of this conjecture, the error for f's distance drops to ≃ 7.9%, the minimum sum of absolute percent errors drops to ≃ 30%, and the average error is reduced to ≃ 5.1%. Table 2 of [4] includes 5 predictions of planet orbits for the HD 40307 system. To compare these predictions with orbit radii computed by A[n], we present in Table 2 all internal shells, i.e. shells located before the outermost occupied shell 3 No 15, which seem to be unoccupied according to the up-to-now observations. Comparison is also made for the external shell No 17, within which an extrap- olated planet prediction seems to exist according to [4]. The sum of absolute percent differences between corresponding estimates is ≃ 32%, giving an average difference ≃ 6.3%. The larger difference, ≃ 12%, is found for the shell No 12. On this issue, it seems interesting to apply again the conjecture discussed above. In particular, the prediction P3 = 0.33 AU can be associated with the orbit αL12 = 0.3496 AU, leading to a smaller difference ≃ 5.9%. However, if, eventually, the average- density orbit αL12 ≃ 0.35 AU hosts a planet, then the maximum-density orbit α12 ≃ 0.37 AU should already host a planet. 3.2 The µ Ara system As seen in Table 3, the optimum case for the µ Ara system gives minimum sum of absolute percent errors ∆min(cid:18)nopt(µ Ara) = 2.554; qc = 5, qd = 14, qb = 18, qe = 31(cid:19) ≃ 5.39. (5) Smaller error is that for e's distance, ≃ 0.06%, which is also the most massive planet; while larger error is that for c's distance, ≃ 3.9%. The average error for the computed distances of the 4 planets in µ Ara is ≃ 1.3%. Regarding predictions of planetary orbits, Table 2 of [4] shows 5 such cases for the µ Ara system. Table 4 presents shells providing respective hosting or- bits computed by A[n]. The sum of absolute percent differences of associated estimates is ≃ 27%, which leads to an average difference ≃ 5.4%. The larger discrepancy, ≃ 15%, appears in the shell No 8. Alternatively, we can conjecture, as above, that the prediction P2 = 0.29 AU is associated with the orbit αR8 = 0.2714 AU, deriving a smaller difference ≃ 6.4%. Accordingly, if, eventually, the average-density orbit αR8 ≃ 0.27 AU hosts a planet, then the maximum-density orbit α8 ≃ 0.25 AU should already host a planet. 3.3 The Kepler-26 (KOI-250) system For the Kepler-26 system, Table 5 reveals the optimum case ∆min(cid:18)nopt(Kepler−26) = 2.531; qd = 5, qb = 7, qc = 8, qe = 11(cid:19) ≃ 18.57. (6) Here, smaller error is that for b's distance, ≃ 2.9%, while larger one is that for e's distance, ≃ 7.2%. The average error in the computed distances of the 4 planets in Kepler-26 is ≃ 4.6%. Table 2 in [4] shows 6 predictions of planet orbits for the Kepler-26 system. In Table 6, we present all unoccupied shells according to up-to-now observations, located before the outermost occupied shell No 11. The external shell No 12, for which an extrapolated planet prediction is made in [4], is also included in 4 this table. The sum of absolute percent differences of respective estimates is ≃ 39%, leading to an average difference ≃ 6.5%. The larger discrepancy, ≃ 16%, arises in the shell No 9. This shell hosts the orbit αL9 ≃ 0.16 AU associated with the prediction P4 = 0.14 AU, and the orbit α9 ≃ 0.17 AU associated with the prediction P5 = 0.17 AU. The latter almost coincides with its counterpart, while the former is in large discrepancy with its counterpart (that quoted above). Since this discrepancy is ∼ 3 times the average difference, we adjudge this issue in favour of the orbit αL9 ≃ 0.16 AU computed by the algorithm A[n], and we thus adopt this orbit in the place of the prediction P4 = 0.14 AU obtained by the Titius -- Bode relation. The second larger discrepancy, ≃ 13%, appears in the shell No 12. We can again conjecture that the prediction P6 = 0.27 AU is associated with the orbit αL12 = 0.2881 AU, giving an appreciably smaller difference ≃ 6.7%. In view of such a conjecture, if, eventually, the orbit αL12 ≃ 0.29 AU is occupied by a planet, then the orbit α12 ≃ 0.30 AU should be already occupied by a planet. 3.4 The Kepler-62 (KOI-701) system For the Kepler-62 system, Table 7 gives the optimum case ∆min(cid:18)nopt(Kepler−62) = 2.773; qb = 5, qc = 6, qd = 7, qe = 11, qf = 14(cid:19) ≃ 20.36. (7) Smaller error is that for b's distance, ≃ 0.08%, while larger error is that for c's distance, ≃ 14%. The error for the most massive planet, e, is ≃ 0.42%. The average error in the computed distances of the 5 planets in Kepler-62 is ≃ 4.1%. Regarding comparisons with the 7 predicted orbits of the Kepler-62 system in Table 2 of [4], Table 8 reveals that the sum of absolute percent differences of respective estimates is ≃ 44%, which gives an average difference ≃ 6.2%. The larger discrepancy, ≃ 14%, arises for the orbit α13 associated with the prediction P6 = 0.55 AU. Instead, we can again conjecture that the prediction P6 is associ- ated with the orbit αL13 = 0.5870 AU, resulting in a smaller difference ≃ 6.7%. Accordingly, if, eventually, the orbit αL13 ≃ 0.59 AU is occupied by a planet, then the orbit α13 ≃ 0.63 AU should be also occupied by a planet. In fact, we have tacitly invoked this conjecture for the prediction P5 = 0.33 AU, which has been associated with the orbit αR10 instead of the orbit α10 = 0.2992 AU (taking the latter in the place of the former, we find a difference ≃ 9.3%, a total difference ≃ 53%, and an average difference ≃ 7.5%). It is worth noting here that the shells No 5 and No 7, hosting the planets b and d on their maximum-density orbits α5 and α7, do also provide the average- density orbits αR5 and αR7, which are associated with the predictions P1 and P2, respectively. We can consider this issue as the "complementary case" of the conjecture discussed above. In particular, if we assume for a while that the planets b and d have not been observed yet, then our eventual conclusion 5 that αR5 corresponds to P1 and αR7 corresponds to P2, does also lead to the conclusion that α5 and α7 should be already occupied by planets. 3.5 The Kepler-275 (KOI-1198) system For the exoplanet system of Kepler-275, Table 9 gives the optimum case ∆min(cid:18)nopt(Kepler−275) = 2.837; qb = 4, qc = 5, qd = 6(cid:19) ≃ 10.5. (8) Smaller error is that for b's distance, ≃ 0.42%, while larger one is that for c's distance, ≃ 5.4%, which is also the most massive planet. The average error in the computed distances of the 3 planets in Kepler-275 is ∼ 3.5%. Table 2 of [4] gives 7 predictions for the Kepler-275 system. As seen in Table 10, the sum of absolute percent differences of respective estimates is ≃ 58%, giving an average difference ≃ 8.3%. We now focus our attention on two cases with large discrepancies. The first case has to do with the prediction P1 = 0.03 AU and its associated orbit αR2 ≃ 0.022 AU. The resulting large discrepancy ≃ 25% (∼ 3 times the average discrepancy) leads us to adjudge this issue in favour of the orbit αR2. Second, likewise, the large discrepancy ≃ 15% (∼ 2 times the average discrepancy) between the prediction P6 = 0.17 AU and its associated orbit αR5 ≃ 0.14 AU turns to be in favour of the latter. Thus we adopt the orbits αR2 and αR5 computed by the algorithm A[n] in the place of the predictions P1 and P6 obtained by the Titius -- Bode relation. Since however αR2 is an average-density orbit, adopting it as a planet's orbit prediction, presumes that the maximum-density orbit α2 should have priority for hosting a planet. It is worth noting here that, for the similar case of αR5, the maximum-density orbit α5 is already occupied by the planet c (Table 9). Finally, it is of particular interest the fact that, in order to associate three orbits with the predictions P2 = 0.04 AU, P3 = 0.05 AU, and P4 = 0.06, we need to employ all three available orbits in the shell No 3, i.e. αL3 ≃ 0.038 AU, α3 ≃ 0.051 AU, and αR3 ≃ 0.061 AU, respectively. Accordingly, any further prediction in the interval [∼ 0.03 AU, ∼ 0.07 AU] should be rejected. 3.6 Further discussion on the HD 40307 and Kepler-26 systems Tables 1 and 5 show that HD 40307 and Kepler-26 both have optimum poly- tropic index nopt = 2.531. This means that all respective orbit radii are equal when expressed in polytropic units. Equivalently, all respective ratios of orbit radii are equal; for example, ξ8(cid:19)HD 40307 (cid:18) ξ5 (cid:18) α5 α8(cid:19)HD 40307 = (cid:18) ξ5 = (cid:18) α5 ξ8(cid:19)Kepler−26 α8(cid:19)Kepler−26 = 0.3925, = 0.3593, (9) (10) 6 etc. Next, the orbits α5 and α8 are occupied in both systems. In addition, the orbit αR5 is found to be in satisfactory agreement with respective predictions of [4] in both systems. Furthermore, the orbits αL6, α6, αL9, and α9 in the Kepler-26 system are found to be in satisfactory agreement with respective predictions of [4]; while the orbits α6 and α9 in the HD 40307 system are occupied. This means in turn that the average-density orbits αL6 and αL9 could be also predictions for the HD 40307 system, but they are "missing" from Table 2 of [4]. Relevant worth noting cases are: (i) the orbit α7 occupied in the Kepler- 26 system, and being in satisfactory agreement with respective prediction of [4] for the HD 40307 system; and (ii) the orbit α11 occupied in the Kepler-26 system, and being empty in [4]; so, the latter orbit could be a prediction for the HD 40307 system, but it is "missing" from Table 2 of [4]. Finally, the following cases seem to be also "missing predictions" from Ta- ble 2 of [4]: (i) the orbit α10 in Kepler-26, while this orbit is occupied in HD 40307; (ii) the orbit α13 in Kepler-26, while this orbit is quoted as pre- diction for HD 40307 in [4]; (iii) the orbit α15 in Kepler-26, while this orbit is occupied in HD 40307; and (iv) the orbit α17 in Kepler-26, while this orbit is quoted as prediction for HD 40307 in [4]. 4 Conclusions We have examined 5 exoplanetary systems with a total of 22 detected exoplanets. Their orbit radii computed by A[n] exhibit an average deviation ∼ 4% from the corresponding observed distances. In the systems examined, there are 50 unoccupied internal shells; excluding the 26 internal shells of the µ Ara system, this total drops to 24. These 24 internal shells, each providing 3 hosting orbits, can accomodate a total of 24 × 3 = 72 exoplanets. We have compared 30 of our predictions with respective ones given in Table 2 of [4], estimated by the generalized Titius -- Bode relation, and we have found an average discrepancy ∼ 6%. In 5 cases, we have applied a conjecture asserting that, eventually, hosting orbits are average-density orbits instead of respective maximum-density orbits, provided that the latter should be already occupied. Finally, in 3 cases with large discrepancies (≥ 15%), we have adopted our predictions in the place of the corresponding predictions of [4]. 7 Table 1: The HD 40307 system: central body S1, i.e. the host star HD 40307, and polytropic spherical shells of the planets b, c, d, e, f, g. For successive shells Sj and Sj+1, inner radius of Sj+1 is the outer radius of Sj. All radii are expressed in AU, except for the host's radius ξ1. Percent errors %Ej in the computed orbit radii αj are given with respect to the corresponding observed radii Aj, %Ej = 100 × (Aj − αj )/Aj. Parenthesized signed integers following numerical values denote powers of 10. Host star HD 40307 -- Shell No nopt ξ1 (cpu) ξ1 (R⊙) 1 2.531 (+00) 5.4317(+00) 7.16 (−01) b -- Shell No Inner radius, ξ4 Outer radius, ξ5 Orbit radius, αb = α5 c -- Shell No Outer radius, ξ6 Orbit radius, αc = α6 d -- Shell No Inner radius, ξ7 Outer radius, ξ8 Orbit radius, αd = α8 e -- Shell No Outer radius, ξ9 Orbit radius, αe = α9 f -- Shell No Outer radius, ξ10 Orbit radius, αf = α10 g -- Shell No Inner radius, ξ14 Outer radius, ξ15 Orbit radius, αg = α15 5 3.7372(−02) 6.4264(−02) 4.8999(−02) 6 9.4481(−02) 8.2084(−02) 8 1.2142(−01) 1.6372(−01) 1.3639(−01) 9 2.0736(−01) 2.0636(−01) 10 2.5487(−01) 2.0838(−01) 15 5.3513(−01) 6.3922(−01) 5.8682(−01) A %E 4.68(−02) 4.70(+00) 7.99(−02) 2.73(+00) 1.321(−01) 3.25(+00) 1.886(−01) 9.42(+00) 2.47(−01) 1.56(+01) 6.(−01) 2.20(+00) 8 Table 2: The HD 40307 system: unoccupied shells, or unoccupied orbits αLj, αj, αRj within occupied shells. Pj denote predicted orbit radii of planets not yet observed, according to Table 2 of [4]; external shells, i.e. shells next to the outermost occupied shell, are included in this table only when such predictions are available. Percent differences %Dj in the computed orbit radii αj are given with respect to the corresponding predicted radii Pj, %Dj = 100×(Pj −αj)/Pj. Other details as in Table 1. P %D Shell No Inner radius, ξ1 Outer radius, ξ2 Orbit radius, α2 Shell No Outer radius, ξ3 Orbit radius, α3 Shell No Outer radius, ξ4 Orbit radius, α4 Shell No Outer radius, ξ5 Orbit radius, αR5 Shell No Inner radius, ξ6 Outer radius, ξ7 Orbit radius, α7 Shell No Inner radius, ξ10 Outer radius, ξ11 Orbit radius, α11 Shell No Outer radius, ξ12 Orbit radius, α12 Shell No Outer radius, ξ13 Orbit radius, α13 Shell No Outer radius, ξ14 Orbit radius, α14 External Shell No Inner radius, ξ16 Outer radius, ξ17 Orbit radius, α17 2 3.3310(−03) 1.1002(−02) 7.0832(−03) 3 2.0234(−02) 1.4473(−02) 4 3.7372(−02) 2.5603(−02) 5 6.4264(−02) 5.7002(−02) 7 9.4481(−02) 1.2142(−01) 1.0609(−01) 11 2.5487(−01) 3.2477(−01) 2.8608(−01) 12 3.9435(−01) 3.6973(−01) 13 4.4884(−01) 4.1694(−01) 14 5.3513(−01) 4.8061(−01) 17 7.2715(−01) 8.2144(−01) 7.3545(−01) 9 6.(−02) 5.00(+00) 1.1(−01) 3.55(+00) 3.3(−01) 1.20(+01) 4.4(−01) 5.24(+00) 7.8(−01) 5.71(+00) Table 3: The µ Ara system: central body S1, i.e. the host star µ Ara, and polytropic spherical shells of the planets c, d, b, e. Other details as in Table 1. Host star µ Ara -- Shell No nopt ξ1 (cpu) ξ1 (R⊙) 1 2.554 (+00) 5.4898(+00) 1.245 (+00) c -- Shell No Inner radius, ξ4 Outer radius, ξ5 Orbit radius, αc = α5 d -- Shell No Inner radius, ξ13 Outer radius, ξ14 Orbit radius, αd = α14 b -- Shell No Inner radius, ξ17 Outer radius, ξ18 Orbit radius, αb = α18 e -- Shell No Inner radius, ξ30 Outer radius, ξ31 Orbit radius, αe = α31 5 6.6323(−02) 1.1495(−01) 8.7360(−02) 14 8.2970(−01) 9.7484(−01) 9.1279(−01) 18 1.4513(+00) 1.6443(+00) 1.5075(+00) 31 5.0053(+00) 5.4708(+00) 5.2381(+00) A %E 9.094(−02) 3.94(+00) 9.210(−01) 8.91(−01) 1.5(+00) 5.03(−01) 5.235(+00) 5.84(−02) 10 Table 4: The µ Ara system: unoccupied shells for which predicted orbit radii Pj are given for planets not yet observed, according to Table 2 of [4]. Other details as in Table 2. P %D Shell No Inner radius, ξ5 Outer radius, ξ6 Orbit radius, α6 Shell No Inner radius, ξ7 Outer radius, ξ8 Orbit radius, α8 Shell No Inner radius, ξ10 Outer radius, ξ11 Orbit radius, α11 Shell No Inner radius, ξ23 Outer radius, ξ24 Orbit radius, α24 External Shell No Inner radius, ξ39 Outer radius, ξ40 Orbit radius, α40 6 1.1495(−01) 1.6901(−01) 1.4882(−01) 8 2.1725(−01) 2.9575(−01) 2.4604(−01) 11 4.6303(−01) 5.9148(−01) 5.2105(−01) 24 2.8301(+00) 3.1098(+00) 3.0123(+00) 40 8.9572(+00) 9.4795(+00) 9.1042(+00) 1.6(−01) 6.99(+00) 2.9(−01) 1.52(+01) 5.2(−01) 2.02(−01) 2.9(+00) 3.87(+00) 9.17(+00) 7.18(−01) 11 Table 5: The Kepler-26 system: central body S1, i.e. the host star Kepler-26, and polytropic spherical shells of the planets d, b, c, e. Other details as in Table 1. Host star Kepler-26 -- Shell No nopt ξ1 (cpu) ξ1 (R⊙) 1 2.531 (+00) 5.4317(+00) 5.9 (−01) d -- Shell No Inner radius, ξ4 Outer radius, ξ5 Orbit radius, αd = α5 b -- Shell No Inner radius, ξ6 Outer radius, ξ7 Orbit radius, αb = α7 c -- Shell No Outer radius, ξ8 Orbit radius, αc = α8 e -- Shell No Inner radius, ξ10 Outer radius, ξ11 Orbit radius, αe = α11 5 3.0795(−02) 5.2955(−02) 4.0376(−02) 7 7.7855(−02) 1.0005(−01) 8.7423(−02) 8 1.3491(−01) 1.1239(−01) 11 2.1002(−01) 2.6762(−01) 2.3574(−01) A %E 3.9(−02) 3.53(+00) 8.5(−02) 2.85(+00) 1.07(−01) 5.03(+00) 2.2(−01) 7.15(+00) 12 Table 6: The Kepler-26 system: unoccupied shells, or unoccupied orbits αLj, αj, αRj within occupied shells. Pj denote predicted orbit radii of planets not yet observed, according to Table 2 of [4]. Other details as in Table 2. P %D Shell No Inner radius, ξ1 Outer radius, ξ2 Orbit radius, α2 Shell No Outer radius, ξ3 Orbit radius, α3 Shell No Outer radius, ξ4 Orbit radius, α4 Shell No Outer radius, ξ5 Orbit radius, αR5 Shell No Outer radius, ξ6 Orbit radius, αL6 Orbit radius, α6 Shell No Inner radius, ξ8 Outer radius, ξ9 Orbit radius, αL9 Orbit radius, α9 Shell No Outer radius, ξ10 Orbit radius, α10 External Shell No Inner radius, ξ11 Outer radius, ξ12 Orbit radius, α12 2 2.7448(−03) 9.0659(−03) 5.8367(−03) 3 1.6674(−02) 1.1926(−02) 4 3.07954(−02) 2.10978(−02) 5 5.2955(−02) 4.6971(−02) 6 7.7855(−02) 5.9882(−02) 6.7639(−02) 9 1.3491(−01) 1.7087(−01) 1.6284(−01) 1.7005(−01) 10 2.1002(−01) 1.7171(−01) 12 2.6762(−01) 3.2496(−01) 3.0467(−01) 13 5.(−02) 6.06(+00) 6.(−02) 7.(−02) 1.97(−01) 3.37(+00) 1.4(−01) 1.7(−01) 1.63(+01) 2.94(−02) 2.7(−01) 1.28(+01) Table 7: The Kepler-62 system: central body S1, i.e. the host star Kepler-62, and polytropic spherical shells of the planets b, c, d, e, f. Other details as in Table 1. Host star Kepler-62 -- Shell No nopt ξ1 (cpu) ξ1 (R⊙) 1 2.773 (+00) 6.1063(+00) 6.3 (−01) b -- Shell No Inner radius, ξ4 Outer radius, ξ5 Orbit radius, αb = α5 c -- Shell No Outer radius, ξ6 Orbit radius, αc = α6 d -- Shell No Outer radius, ξ7 Orbit radius, αd = α7 e -- Shell No Inner radius, ξ10 Outer radius, ξ11 Orbit radius, αe = α11 f -- Shell No Inner radius, ξ13 Outer radius, ξ14 Orbit radius, αf = α14 5 4.3481(−02) 7.6678(−02) 5.5342(−02) 6 1.1442(−01) 1.0545(−01) 7 1.5633(−01) 1.2283(−01) 11 3.5889(−01) 4.4702(−01) 4.4511(−01) 14 6.7326(−01) 7.8344(−01) 7.1679(−01) A %E 5.53(−02) 7.65(−02) 9.29(−02) 1.35(+01) 1.2(−01) 2.36(+00) 4.27(−01) 4.24(−01) 7.18(−01) 1.69(−01) 14 Table 8: The Kepler-62 system: unoccupied shells, or unoccupied orbits αLj, αj, αRj within occupied shells. Pj denote predicted orbit radii of planets not yet observed, according to Table 2 of [4]. Other details as in Table 2. P %D Shell No Inner radius, ξ1 Outer radius, ξ2 Orbit radius, α2 Shell No Outer radius, ξ3 Orbit radius, α3 Shell No Outer radius, ξ4 Orbit radius, α4 Shell No Outer radius, ξ5 Orbit radius, αR5 Shell No Inner radius, ξ6 Outer radius, ξ7 Orbit radius, αR7 Shell No Outer radius, ξ8 Orbit radius, α8 Shell No Outer radius, ξ9 Orbit radius, α9 Shell No Outer radius, ξ10 Orbit radius, α10 Orbit radius, αR10 Shell No Inner radius, ξ11 Outer radius, ξ12 Orbit radius, α12 Shell No Outer radius, ξ13 Orbit radius, α13 External Shell No Inner radius, ξ14 Outer radius, ξ15 Orbit radius, α15 2 2.9309(−03) 1.2620(−02) 6.2399(−03) 3 2.6848(−02) 2.3207(−02) 4 4.3481(−02) 3.0100(−02) 5 7.6678(−02) 6.6406(−02) 7 1.1442(−01) 1.5633(−01) 1.4004(−01) 8 2.1787(−01) 1.8604(−01) 9 2.7712(−01) 2.5403(−01) 10 3.5889(−01) 2.9922(−01) 3.2838(−01) 12 4.4702(−01) 5.4706(−01) 4.4893(−01) 13 6.7326(−01) 6.2825(−01) 15 7.8344(−01) 9.4463(−01) 8.4677(−01) 15 7.(−02) 5.13(+00) 1.5(−01) 6.64(+00) 2.(−01) 6.98(+00) 2.6(−01) 2.30(+00) 3.3(−01) 4.90(−01) 5.5(−01) 1.42(+01) 9.2(−01) 7.96(+00) Table 9: The Kepler-275 system: central body S1, i.e. the host star Kepler-275, and polytropic spherical shells of the planets b, c, d. Other details as in Table 1. Host star Kepler-275 -- Shell No nopt ξ1 (cpu) ξ1 (R⊙) 1 2.837 (+00) 6.3115(+00) 1.38 (+00) b -- Shell No Inner radius, ξ3 Outer radius, ξ4 Orbit radius, αb = α4 c -- Shell No Outer radius, ξ5 Orbit radius, αc = α5 d -- Shell No Outer radius, ξ6 Orbit radius, αd = α6 4 7.1357(−02) 1.1188(−01) 9.8416(−02) 5 1.8278(−01) 1.2481(−01) 6 3.0822(−01) 2.3432(−01) A %E 9.8(−02) 4.24(−01) 1.32(−01) 5.44(+00) 2.24(−01) 4.61(+00) Table 10: The Kepler-275 system: unoccupied shells, or unoccupied orbits αLj, αj, αRj within occupied shells. Pj denote predicted orbit radii of planets not yet observed, according to Table 2 of [4]. Other details as in Table 2. P %D Shell No Inner radius, ξ1 Outer radius, ξ2 Orbit radius, α2 Orbit radius, αR2 Shell No Outer radius, ξ3 Orbit radius, αL3 Orbit radius, α3 Orbit radius, αR3 Shell No Outer radius, ξ4 Orbit radius, αL4 Shell No Outer radius, ξ5 Orbit radius, αR5 Shell No Outer radius, ξ6 Orbit radius, αR6 2 6.4201(−03) 2.9258(−02) 1.3834(−02) 2.2430(−02) 3 7.1357(−02) 3.8197(−02) 5.1166(−02) 6.0524(−02) 4 1.1188(−01) 8.2954(−02) 5 1.8278(−01) 1.4480(−01) 6 3.0822(−01) 2.6992(−01) 16 3.(−02) 2.52(+01) 4.(−02) 5.(−02) 6.(−02) 4.51(+00) 2.33(+00) 8.73(−01) 8.(−02) 3.69(+00) 1.7(−01) 1.48(+01) 2.9(−01) 6.92(+00) References [1] V. Geroyannis, F. Valvi and T. Dallas, International Journal of Astronomy and Astrophysics, 2014, 4, 464 (2014). http://dx.doi.org/10.4236/ijaa.2014.43042 [2] V. S. Geroyannis, arXiv:1410.5844v2 [astro-ph.EP] (2014). [3] V. S. Geroyannis, arXiv:1411.5390v1 [astro-ph.EP] (2014). [4] T. Bovaird and C. H. Lineweaver, arXiv:1304.3341v4 [astro-ph.EP] (2013). [5] V. S. Geroyannis and V. G. Karageorgopoulos, New Astronomy, 28, 9 (2014). [6] J. C. B. Papaloizou and C. Terquem, Monthly Notices of the Royal Astro- nomical Society, 405, 573 (2010). [7] M. Tuomi, G. Anglada-Escud´e, E. Gerlach, et al., arXiv:1211.1617v1 [astro- ph.EP] (2012). [8] C. Yuan-Yuan, Z. Ji-Lin, and M. Yue-Hua, arXiv:1408.4228v1 [astro- ph.EP] (2014). [9] K. Go´zdziewski, M. Konacki, and A. Maciejewski, The Astrophysical Jour- nal, 622, 1136 (2005). [10] F. Pepe, A. C. M. Correia, M. Mayor, et al., Astronomy and Astrophysics, 462, 769 (2007). [11] R. A. Wittenmyer, C. G. Tinney, S. J. O'Toole, et al., arXiv:1011.4720v1 [astro-ph.EP] (2010). [12] J. H. Steffen, D. C. Fabrycky, E. B. Ford, et al., arXiv:1201.5412v1 [astro- ph.EP] (2012). [13] S. E. Dodson-Robinson, arXiv:1204.2275v1 [astro-ph.EP] (2012). [14] W. J. Borucki, E. Agol, F. Fressin, et al., arXiv:1304.7387v1 [astro-ph.EP] (2013). [15] S. R. Kane, arXiv:1401.3349v1 [astro-ph.EP] (2014). [16] E. Bolmont, S. N. Raymond, J. Leconte, et al., arXiv:1410.7428v1 [astro- ph.EP] (2014). [17] J. F. Rowe, S. T. Bryson, G. W. Marcy, et al., arXiv:1402.6534v1 [astro- ph.EP] (2014). [18] G. P. Horedt, Polytropes: Applications in Astrophysics and Related Fields (Kluwer Academic Publishers, New York, 2004). [19] V. S. Geroyannis and F. N. Valvi, International Journal of Modern Physics C, 23, 5 (2012). 17
1604.01413	2	1604	2016-05-05T21:14:31	Doppler Monitoring of five K2 Transiting Planetary Systems	[ "astro-ph.EP" ]	In an effort to measure the masses of planets discovered by the NASA {\it K2} mission, we have conducted precise Doppler observations of five stars with transiting planets. We present the results of a joint analysis of these new data and previously published Doppler data. The first star, an M dwarf known as K2-3 or EPIC~201367065, has three transiting planets ("b", with radius $2.1~R_{\oplus}$; "c", $1.7~R_{\oplus}$; and "d", $1.5~R_{\oplus}$). Our analysis leads to the mass constraints: $M_{b}=8.1^{+2.0}_{-1.9}~M_{\oplus}$ and $M_{c}$ < $ 4.2~M_{\oplus}$~(95\%~conf.). The mass of planet d is poorly constrained because its orbital period is close to the stellar rotation period, making it difficult to disentangle the planetary signal from spurious Doppler shifts due to stellar activity. The second star, a G dwarf known as K2-19 or EPIC~201505350, has two planets ("b", $7.7~R_{\oplus}$; and "c", $4.9~R_{\oplus}$) in a 3:2 mean-motion resonance, as well as a shorter-period planet ("d", $1.1~R_{\oplus}$). We find $M_{b}$= $28.5^{+5.4}_{-5.0} ~M_{\oplus}$, $M_{c}$= $25.6^{+7.1}_{-7.1} ~M_{\oplus}$ and $M_{d}$ < $14.0~M_{\oplus} $~(95\%~conf.). The third star, a G dwarf known as K2-24 or EPIC~203771098, hosts two transiting planets ("b", $5.7~R_{\oplus}$; and "c", $7.8~R_{\oplus}$) with orbital periods in a nearly 2:1 ratio. We find $M_{b}$= $19.8^{+4.5}_{-4.4} ~M_{\oplus}$ and $M_{c}$ = $26.0^{+5.8}_{-6.1}~M_{\oplus}$.....	astro-ph.EP	astro-ph	DRAFT VERSION MAY 9, 2016 Preprint typeset using LATEX style emulateapj v. 8/13/10 DOPPLER MONITORING OF FIVE K2 TRANSITING PLANETARY SYSTEMS FEI DAI1†, JOSHUA N. WINN1, SIMON ALBRECHT2, PAMELA ARRIAGADA3, ALLYSON BIERYLA3, R. PAUL BUTLER4, JEFFREY D. CRANE5, TERUYUKI HIRANO7, JOHN ASHER JOHNSON4, AMANDA KIILERICH2, DAVID W. LATHAM4, NORIO NARITA8,9,10, GRZEGORZ NOWAK11,12, ENRIC PALLE11,12, IGNASI RIBAS13, LESLIE A. ROGERS14,15, ROBERTO SANCHIS-OJEDA16,15, STEPHEN A. SHECTMAN5, JOHANNA K. TESKE4,5, IAN B. THOMPSON5, VINCENT VAN EYLEN2,1, ANDREW VANDERBURG4, ROBERT A. WITTENMYER17,18, LIANG YU1 Draft version May 9, 2016 ABSTRACT In an effort to measure the masses of planets discovered by the NASA K2 mission, we have conducted precise Doppler observations of five stars with transiting planets. We present the results of a joint analysis of these new data and previously published Doppler data. The first star, an M dwarf known as K2-3 or EPIC 201367065, has three transiting planets ("b", with radius 2.1 R⊕; "c", 1.7 R⊕; and "d", 1.5 R⊕). Our analysis leads to the mass constraints: Mb = 8.1+2.0 −1.9 M⊕ and Mc < 4.2 M⊕ (95% conf.). The mass of planet d is poorly constrained because its orbital period is close to the stellar rotation period, making it difficult to disentangle the planetary signal from spurious Doppler shifts due to stellar activity. The second star, a G dwarf known as K2-19 or EPIC 201505350, has two planets ("b", 7.7 R⊕; and "c", 4.9 R⊕) in a 3:2 mean-motion resonance, as well as a shorter-period planet ("d", 1.1 R⊕). We find Mb= 28.5+5.4 −7.1 M⊕ and Md < 14.0 M⊕ (95% conf.). The third star, a G dwarf known as K2-24 or EPIC 203771098, hosts two transiting planets ("b", 5.7 R⊕; and "c", 7.8 R⊕) with orbital periods in a nearly 2:1 ratio. We find Mb= 19.8+4.5 −6.1 M⊕. The fourth star, a G dwarf known as EPIC 204129699, hosts a hot Jupiter for which we measured the mass to be 1.857+0.081 −0.081 MJup. The fifth star, a G dwarf known as EPIC 205071984, contains three transiting planets ("b", 5.4 R⊕; "c", 3.5 R⊕; and "d", 3.8 R⊕), the outer two of which have a nearly 2:1 period ratio. We find Mb= 21.1+5.9 Subject headings: planetary systems −5.9 M⊕, Mc < 8.1 M⊕ (95% conf.) and Md < 35 M⊕ (95% conf.). - planets and satellites: −4.4 M⊕ and Mc = 26.0+5.8 individual (K2-3 (EPIC 201367065), K2-19 (EPIC 201505350), EPIC 204129699, K2-24 (EPIC 203771098), EPIC 205071984) - techniques: radial velocities composition - −5.0 M⊕, Mc= 25.6+7.1 stars: 6 1 0 2 y a M 5 . ] P E h p - o r t s a [ 2 v 3 1 4 1 0 . 4 0 6 1 : v i X r a 1 Department of Physics, and Kavli Institute for Astrophysics and Space Research, Massachusetts Institute of Technology, Cambridge, MA 02139, USA† [email protected] 2 Stellar Astrophysics Centre, Department of Physics and Astronomy, Aarhus University, Ny Munkegade 120, DK-8000 Aarhus C, Denmark 3 Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138, USA 4 Carnegie Institution of Washington, Department of Terrestrial Mag- netism, 5241 Broad Branch Road, NW, Washington DC, 20015-1305, USA 5 The Observatories of the Carnegie Institution of Washington, 813 Santa Barbara Street, Pasadena, CA 91101, USA 6 Institut d'Astrophysique et de Géophysique, Université de Liège, Al- lée du 6 Août 17, Sart Tilman, 4000 Liège 1, Belgium 7 Department of Earth and Planetary Sciences, Tokyo Institute of Tech- nology, 2-12-1 Ookayama, Meguro-ku, Tokyo 152-8551, Japan 8 Astrobiology Center, National Institutes of Natural Sciences, 2-21-1 Osawa, Mitaka, Tokyo 181-8588, Japan 9 National Astronomical Observatory of Japan, 2-21-1 Osawa, Mitaka, Tokyo 181-8588, Japan 10 SOKENDAI (The Graduate University for Advanced Studies), 2-21- 1 Osawa, Mitaka, Tokyo 181-8588, Japan 11 Instituto de Astrofísica de Canarias (IAC), 38205 La Laguna, Tener- ife, Spain 12 Departamento de Astrofísica, Universidad de La Laguna (ULL), 38206 La Laguna, Tenerife, Spain 13 Institut de Ciències de l'Espai (CSIC-IEEC), Carrer de Can Magrans, Campus UAB, 08193 Bellaterra, Spain 14 Department of Earth and Planetary Science, University of California, Berkeley, CA 94720, USA 15 NASA Sagan Fellow 16 Department of Astronomy, University of California, Berkeley, CA 94720, USA 17 School of Physics and Australian Centre for Astrobiology, University of New South Wales, Sydney 2052, Australia 18 Computational Engineering and Science Research Centre, University of Southern Queensland, Toowoomba, Queensland 4350, Australia 2 Dai et al. 1. INTRODUCTION The characterization of a planet starts with the measure- ment of its mass and radius. Without these two quantities, we cannot even begin to answer the most basic questions about the planet's internal structure, atmospheric composition and formation history. Of particular interest are planets with radii in the range of 1–4 R⊕, known as "super-Earths" or "sub- Neptunes." Despite being the most frequently occurring ex- oplanets within 1 AU of solar-type stars (e.g. Petigura et al. 2013; Howard et al. 2012; Fressin et al. 2013), they have no known counterparts in our solar system. Moreover, traditional core accretion models have trouble explaining why these plan- ets did not undergo runaway accretion that would have led to the formation of gas giants (Mizuno 1980; Rafikov 2006; Bodenheimer & Lissauer 2014; Inamdar & Schlichting 2015; Lee & Chiang 2015). Detailed characterization of these plan- ets will shed light on these mysteries. Only a few dozen Doppler mass measurements of planets smaller than Neptune have been reported (see, e.g., Marcy et al. 2014; Howard et al. 2013; Pepe et al. 2013; Dressing et al. 2015; Motalebi et al. 2015; Gettel et al. 2016; Weiss et al. 2016). The number has been limited because of the rel- ative faintness of the host stars of the known transiting plan- ets. The ongoing NASA K2 mission (Howell et al. 2014) is gradually providing a larger sample of stars which are bright enough to be amenable to precise Doppler follow-up obser- vations. In this paper, we present new Doppler observations using three different high-resolution spectrographs. By com- bining these data with previously reported Doppler data (Al- menara et al. 2015; Petigura et al. 2016; Grziwa et al. 2015), we placed mass constraints on the planets in five K2 systems: K2-3, K2-19, K2-24, EPIC 204129699 and EPIC 205071984. This paper is organized as follows. In Section 2, we present our transit candidate search pipeline and follow-up target se- lection which led us to the five K2 systems. Section 3 summa- rizes the instruments and the details of our Doppler observa- tions. Section 4 describes the methods we used to analyze the Doppler observations. Section 5 contains the results of our analysis for each planet candidate host. Section 6 discusses the implications of our results in a broader context. 2. K2 PHOTOMETRY AND TARGET SELECTION The five systems presented in this paper have all been re- ported previously. They were also independently identified by the K2 collaboration in which many of us participate, which is known as ESPRINT ("Equipo de Seguimiento de Plane- tas Rocosos INterpretando sus Tránsitos" or the "Follow-up team of rocky planets via the interpretation of their transits"). The production of calibrated and detrended K2 light curves by the ESPRINT collaboration was described by Sanchis-Ojeda et al. (2015). The light curves were searched for transiting planet candidates with two algorithms. We employed the Box-Least-Squares routine (Kovács et al. 2002; Jenkins et al. 2010) using the optimal frequency sampling described by Ofir (2014) and Vanderburg et al. (2016). We also employed a Fast-Fourier-Transform method (Sanchis-Ojeda et al. 2014) suitable for detecting short-period planet candidates (< 1 day). We then removed false positives by looking for the alternat- ing eclipse depths associated with the primary and secondary eclipses of eclipsing binaries, and secondary eclipses deep enough that they could only be caused by a secondary star rather than a planet. Candidates that passed these initial tests were then selected for Doppler follow-up observations based on their scientific interest and measurement feasibility. Regarding scientific interest, we preferentially selected systems containing sub- Neptunes (R < 4 R⊕), because their internal compositions and formation pathways are poorly understood. In order to esti- mate the planetary radius, we combined the measured tran- sit depth with the stellar parameters derived from broadband photometries. Systems with multiple transiting planet can- didates, especially those close to mean-motion resonances, were given higher priority in our selection process. Multi- candidate systems are less likely to be false positives (Latham et al. 2011; Lissauer et al. 2011) and transit timing variation (TTV) analysis of these systems may unveil the orbital config- urations and offer an independent way of measuring masses. With the consideration of measurement feasibility, we only selected targets brighter than V ≈ 13. Another factor affect- ing the feasibility of Doppler mass measurements is the level of stellar variability of the host star. Stellar activity such as spots and plages introduces radial-velocity perturbations on a characteristic timescale of the stellar rotation period. We tried to anticipate the level of rotation-induced radial velocity perturbations using the approximation: × δF F δRVrot = 2πR(cid:63) Prot (1) where δF/F is the observed fractional photometric variation and Prot is the rotation period, both of which were estimated from the K2 data. We did not pursue systems for which δRVrot was significantly higher than the anticipated Doppler signal. Using these criteria, we selected a few targets per K2 cam- paign for Doppler follow-up observations. 3. DOPPLER OBSERVATIONS The Doppler observations presented in this paper were ob- tained with three spectrographs: the High Accuracy Radial- velocity Planet Searcher (HARPS) at the ESO La Silla 3.6m telescope, the Carnegie Planet Finder Spectrograph (PFS) on the 6.5m Magellan/Clay Telescope at Las Campanas Obser- vatory in Chile, and the Tillinghast Reflector Echelle Spectro- graph (TRES) on the 1.5m Tillinghast telescope at the Smith- sonian Astrophysical Observatory's Fred L. Whipple Obser- vatory on Mt. Hopkins in Arizona. HARPS is an échelle spectrograph that employs the simultaneous-reference method to achieve precise Doppler observations with long-term stability (Mayor et al. 2003). The spectral coverage of HARPS spans 378-691 nm and it has a spectral resolution of ≈115,000. The instrument is sealed in a vacuum container to maintain stability against temper- ature and pressure fluctuations. We used the Fabry-Perot etalon for simultaneous wavelength calibration. The expo- sure times ranged from 20-45 min. The signal-to-noise ra- tio (SNR) of the stellar spectra obtained near a wavelength range of 5500 Å ranged from 20-50. Radial velocities and uncertainties for our targets were derived using the standard HARPS Data Reduction Software. PFS employs an iodine gas cell to superimpose well- characterized absorption features onto the stellar spectrum. The iodine absorption lines help to establish the wavelength scale and instrumental profile (Crane et al. 2010). The de- tector was read out in the standard 2× 2 binned mode. Ex- posure times ranged from 15-20 min, giving a SNR of 50- 140 pixel−1 and a resolution of about 76,000 in the vicinity of the iodine absorption lines. An additional iodine-free spec- Doppler Monitoring of five K2 Transiting Planetary Systems 3 FIG. 1.- Measured radial velocity of K2-3 (open circles are HARPS data from Almenara et al. (2015); black circles are new PFS data) and the best- fitting model (red line) assuming circular orbits. The other colored lines show the contributions to the model curve from individual planets. trum with higher resolution and higher SNR was obtained for each star, to serve as a template spectrum for the Doppler anal- ysis. The relative radial velocities were extracted from the spectrum using the techniques of Butler et al. (1996). The in- ternal measurement uncertainties (ranging from 2.5-4 m s−1) were estimated based on the scatter in the results to fitting individual 2 Å sections of the spectrum. TRES is a fiber-fed échelle spectrograph with resolving power of 44,000 and wavelength coverage 390-910 nm. Wavelength calibration is achieved using exposures of a Thorium-Argon hollow cathode lamp through the science fiber before and after each observation. The relative veloci- ties reported for EPIC 204129699 in this paper were derived by cross-correlating the individual observations against the strongest observation, which thus defines the velocity zero point. The mean SNR per resolution element for the ten obser- vations reported here was 40, the average exposure time was 10 minutes, and the average internal error estimate, derived from the scatter of the velocities for the individual échelle orders in each observation, was 15 m s−1. Some of the or- ders were not included in the velocity determinations, either because the orders were contaminated by telluric lines intro- duced by the Earth's atmosphere, or the exposure level in the order was too weak, typically at the shortest wavelengths. 4. RADIAL VELOCITY ANALYSIS We modeled the Doppler measurements for each system as the sum of Keplerian radial velocity signals, one for each planet candidate identified in the K2 photometry. We fixed the orbital periods and times of inferior conjunction at the values derived from K2 photometry. We allowed the Doppler semi- amplitude K induced by each planet to be a free parameter, with a uniform prior. We note in particular that we did not require K to be positive. If we knew our model to be correct in all relevant details, and had a fundamental understanding of the uncertainties in the data points, the proper Bayesian approach would be to use a prior on K that precludes nega- tive values because they correspond to unphysical solutions (negative planet masses). However, given that our model may fail to include significant effects arising from stellar activity or additional planets, we opted not to place a prior on K that would rule out negative values. In this way, when we find K < 0 we will know the model is missing important sources of radial-velocity variation. For each system we performed two fits: one in which the orbits were assumed to be circular, and one with eccentric or- FIG. 2.- Radial velocity as a function of time since mid-transit, for each of the planets in the K2-3 system (open circles are HARPS data from Almenara et al. (2015); black circles are new PFS data). In each case, the modeled contributions of the other two planets has been removed, before plotting. √ bits. In the first case, the radial velocity perturbation due to each planet is specified only by its Doppler semi-amplitude K. For eccentric models, the eccentricity e and argument of periastron ω are also needed. To guard against the bias to- wards non-zero eccentricity (Lucy & Sweeney 1971), we used esin ω. For each obser- the fitting parameters vatory, we included a constant velocity offset γ and a jitter parameter σjit to subsume additional astrophysical and instru- mental sources of apparent radial-velocity variation in excess of our internally-estimated measurement uncertainties. We also tested whether the inclusion of a constant acceleration term γ improved the model fit. ecos ω and √ We adopted the following likelihood function:  (cid:113) Nins(cid:89) NRV,k(cid:89) k=1 i=1 L = (cid:34) 1 i + σ2 2π(σ2 jit, k) exp −[RV (ti) −M(ti)]2 jit, k) i + σ2 2(σ2 (cid:35) , (2) 020406080100Time (days, BJD 2,457,045)1510505101520Radial Velocity (m s−1)bcdSumPFSHARPS0246810Time since midtransit (days)1510505101520Radial Velocity (m s−1)planet b05101520Time since midtransit (days)10505101520Radial Velocity (m s−1)planet c010203040Time since midtransit (days)10505101520Radial Velocity (m s−1)planet d4 Dai et al. FIG. 3.- The bisector span (BIS) and radial velocity variations from HARPS, for the K2-3 system. The color coding shows the SNR of the stel- lar spectra obtained near a wavelength range of 5500 Å. The uncertainty in the BIS was assumed to be twice the uncertainty in the radial velocity mea- surement. The Pearson correlation coefficient and the corresponding p-value were calculated to be 0.36 and 0.018. This suggests the presence of stellar activity signal, as the p-value indicates that the observed degree of correla- tion has only a 2% chance of being produced by uncorrelated noise. See text for details. FIG. 4.- Measured radial velocity of K2-19 (open circles are new HARPS data; black circles are new PFS data), and the best-fitting model (red line) as- suming circular orbits. The other colored lines show the contributions to the model curve from individual planets. To account for radial velocity pertur- bations from stellar activity, the data points within each 12-day interval were grouped together, and allowed to shift up or down by a constant velocity spe- cific to the group (see text for details). where RV (ti) is the measured radial velocity at time ti; M(ti) is the calculated radial velocity at time ti for a particular choice of model parameters; σi is the internal measurement uncer- tainty; σjit, k is the jitter for the kth instrument; NRV,k is the number of observations obtained with the kth instrument and Nins is the number of instruments involved for each system. Uniform priors were adopted for all model parameters. We maximized the likelihood using the Nelder-Mead ("Amoeba") method as implemented in the Python scipy package. To determine the parameter uncertainies and covari- ances, we employed a Markov Chain Monte Carlo method. We used the affine-invariant ensemble sampler proposed by Goodman & Weare (2010) and implemented in the Python package emcee (Foreman-Mackey et al. 2013). We started 100 walkers in a Gaussian ball surrounding the best-fitting model parameters as obtained from the maximum likelihood estimation above. We stopped the walkers after 5000 links and checked the convergence by ensuring that the Gelman- FIG. 5.- Radial velocity as a function of time since mid-transit, for each of the planets in the K2-19 system (open circles are new HARPS data; black circles are new PFS data). For each planet, the modeled contributions of the other two planets has been removed, before plotting. Rubin potential scale reduction factor (Gelman et al. 1993) dropped below 1.03. We report the median of the marginal- ized posterior distribution, and defined the uncertainty inter- val based on the 16% and 84% percentile levels of the cumu- lative distribution. We assessed the statistical significance of eccentric orbits and constant acceleration terms using the Bayesian Informa- tion Criterion, BIC = −2× log(Lmax) + N log(M), where Lmax is the maximum likelihood, N is the number of parameters and M is the number of observations (Schwarz 1978; Lid- dle 2007). As another measure of the significance of various models, we also report the root-mean-square (RMS) of the radial-velocity residuals for comparison with the RMS of the original radial velocity measurements. To test whether the best-fitting model is likely to be dynam- ically stable, we calculated the orbital separations in units of mutual Hill radius: . (3) ∆ = aout − ain RH 504030201001020BIS (m s−1)1510505101520RV-mean RV (m s−1)30354045SNR05101520403020100102030Radial Velocity (m s−1)7075Time (days, BJD 2,457,050)135140145150155bcdSumPFSHARPS012345678Time since midtransit (days)3020100102030Radial Velocity (m s−1)planet b024681012Time since midtransit (days)3020100102030Radial Velocity (m s−1)planet c0.00.51.01.52.02.5Time since midtransit (days)3020100102030Radial Velocity (m s−1)planet dDoppler Monitoring of five K2 Transiting Planetary Systems 5 FIG. 6.- The bisector span (BIS) and radial velocity variations from HARPS for the K2-19 system. The color coding shows the SNR of the stellar spectra obtained near a wavelength range of 5500 Å. The uncertainty in the BIS was assumed to be twice the uncertainty in the radial velocity measure- ment. No correlation was observed. FIG. 8.- Radial velocity as a function of time since mid-transit, for each planet in the K2-24 system (open circles are new HARPS data; black circles are new PFS data; triangles are HIRES data from Petigura et al. (2016)). In each case, the modeled contributions of the other planet has been removed, before plotting. FIG. 7.- Measured radial velocity of K2-24 (open circles are new HARPS data; black circles are new PFS data; triangles are HIRES data from Petigura et al. (2016)), and the best-fitting model (red line) assuming circular orbits and an additional constant acceleration. The gray line shows the constant ac- celeration term. The other colored lines show the contributions to the model curve from individual planets. (cid:20) Mout + Min (cid:21)1/3 (aout + ain) RH = 3 M(cid:63) 2 . (4) where Min and Mout are the masses of the inner and outer planets and ain and aout are their semi-major axes. For two- √ planet systems, we used the analytical criterion for stability: ∆ > 2 3 (Gladman 1993). For systems with three planets, we used heuristic criterion for long-term dynamical stability proposed by Fabrycky et al. (2014): ∆1 + ∆2 > 18 where ∆1 and ∆2 are calculated respectively for the inner and outer pair of planets. By modeling the observed Doppler shifts as the sum of Keplerian radial velocity signals, we implicitly assumed that gravitational interaction between planets can be ignored. We tested and ultimately justified this assumption by experiment- ing with a fully dynamical (N-body) model obtained from the 4th order Hermite integration scheme that is available on the Systemic console (Meschiari et al. 2009). By maximizing the likelihood function with this dynamical model, we found that the best-fitting system parameters were consistent with those derived from our simpler multi-Keplerian model. We then ex- amined the deviations between the radial velocities calculated FIG. 9.- The bisector span (BIS) and radial velocity variations from HARPS for the K2-24 system. The color coding shows the SNR of the stel- lar spectra obtained near a wavelength range of 5500 Å. The uncertainty in the BIS was assumed to be twice the uncertainty in the radial velocity mea- surement. The Pearson correlation coefficient and the corresponding p-value were calculated to be -0.14 and 0.70. The large p-value suggests no signifi- cant correlation between BIS and radial velocity variations. in the dynamical model and the radial velocities in the Ke- plerian models. Over the timespan of our observations, the maximum deviation occurred for the K2-19 system and had a numerical value of 0.47 m s−1, which was much smaller than the uncertainties in the Doppler data (≈3.8 m s−1). For all the other systems, the deviations were an order of magnitude 0102030405060BIS (m s−1)201510505101520RV-mean RV (m s−1)20202525SNR020406080100Time (days, BJD 2,457,190)15105051015Radial Velocity (m s−1)bcγSumPFSHARPSHIRES05101520Time since midtransit (days)1050510Radial Velocity (m s−1)planet b010203040Time since midtransit (days)105051015Radial Velocity (m s−1)planet c101520253035404550BIS (m s−1)151050510RV-mean RV (m s−1)3035404550SNR6 Dai et al. smaller. Some of the Doppler observations were conducted during an expected transit of one of the planets. Due to the scarcity and relatively large uncertainties of Doppler data, we did not attenmpt to model the Rossiter-McLaughlin (RM) effect in detail. Rather, we assessed the likely amplitude of the RM effect, and excluded from consideration the data for which the RM effect would make a significant contribution to the overall radial velocity. We estimated the RM amplitude with the following equation: √ (cid:19)2 × v sin i, 1 − b2 × (cid:18) Rp R(cid:63) ∆RVRM = (5) where b is the impact parameter of the transit, Rp/R(cid:63) is the radius ratio between the planet and the host star, and v sin i is the projected rotational velocity of the stellar photosphere. For planets for which ∆RVRM was at least comparable to the measurement uncertainty, we chose to omit the data points taken during the calculated transit intervals. The planets that were affected were K2-3b, K2-19b, EPIC 205071984b and EPIC 205071984c. In order to test whether the observed Doppler signal could be associated with stellar activity, we plotted the measured Doppler shifts against the bisector span (BIS, defined by Queloz et al. 2001) derived from the cross-correlation func- tions of the obtained HARPS spectra. (Bisector data is not available for the other spectrographs.) We also calculated the Pearson correlation coefficient between the radial veloc- ity and BIS, and the corresponding p-value which is roughly the probability that an uncorrelated process would produce a data set with a correlation coefficient at least as large as the observed coefficient. The BIS varies as a result of the de- formation of stellar absorption lines induced by stellar activ- ity. The gravitational pull of a planet should not change the BIS. Therefore, a strong correlation between BIS and Doppler shifts suggests that the observed Doppler shifts are the result of stellar activity rather than the gravitational influence of the planet. 5. INDIVIDUAL TARGETS 5.1. K2-3 K2-3 or EPIC 201367065 is a nearby M0V star with three transiting planets (Crossfield et al. 2015). The planets are on 10.1, 24.6 and 44.6 day orbits with radii of 2.18± 0.30 R⊕, 1.85 ± 0.27 R⊕ and 1.51 ± 0.23 R⊕. The outermost planet may lie within the habitable zone. The host star is relatively bright (V = 12.17± 0.01) and small (R(cid:63) = 0.561± 0.068 R(cid:12)), thus making the system a favorable target for the James Webb Space Telescope (JWST). These planets have been validated by Crossfield et al. (2015), Sinukoff et al. (2015) and Beich- man et al. (2016) through adaptive optics imaging and Spitzer transit observations. We observed K2-3 with PFS from January 28th to April 11th, 2015. We gathered a total of 31 spectra. The typical internally-estimated measurement uncertainty was 2.5 m s−1. Initially we modeled only the PFS data, and the results were not constraining. For all three planets, we could only place upper bounds on the masses: Mb < 15.1 M⊕ (95% conf.) Mc < 6.3 M⊕ (95% conf.) Md < 21.1 M⊕ (95% conf.). Nev- ertheless, the PFS data did serve as an independent check of the measurements reported by Almenara et al. (2015): Mb = 8.4± 2.1 M⊕, Mc = 2.1+2.1 −3.5 M⊕. Their results were based on the analysis of the 66 HARPS Doppler −1.3 M⊕ and Md = 11.1+3.5 FIG. 10.- Top.-Measured radial velocity of EPIC 204129699 (stars are FIES data from Grziwa et al. (2015); open circles are HARPS data from Grziwa et al. (2015); black circles are new PFS data; triangles are new TRES data). The gray rectangle illustrates the best-fitting K value. (Because of the short orbital period, the model curve varies too rapidly to be plotted clearly over the entire time range.) Bottom.-The radial-velocity residuals, after sub- tracting the best-fitting model assuming a circular orbit. FIG. 11.- Radial velocity of of EPIC 204129699 as a function of time since mid-transit of planet b (stars are FIES data from Grziwa et al. (2015); open circles are HARPS data from Grziwa et al. (2015); black circles are new PFS data; triangles are new TRES data). measurements with an average uncertainty of 2.9 m s−1. For all three planets, the PFS results were consistent with the HARPS results. We then performed a joint analysis using both the PFS and HARPS datasets. We started with the simplest model, assum- ing circular orbits. The radial-velocity residuals of the best- fitting model showed substantial temporal correlation. Specif- ically, the most prominent peak in the Lomb-Scargle peri- odogram of the radial velocity data occurred at 20.6 ± 1.6 days, with a false alarm probability (FAP) of 5.6 × 10−3. This same peak remained the strongest in the periodogram of the radial-velocity residuals after subtracting the best-fitting 4002000200Radial Velocity (m s−1)PFSHARPSTRESFIES020406080100120140Time (days, BJD 2,457,122)6040200204060Residual (m s−1)0.00.20.40.60.81.01.2Time since midtransit (days)3002001000100200300Radial Velocity (m s−1)planet bDoppler Monitoring of five K2 Transiting Planetary Systems 7 FIG. 12.- Measured radial velocity of EPIC 205071984 (open circles are new HARPS data; black circles are new PFS data) and the best-fitting circu- lar model (red line). The contributions from each planet are also plotted as colored lines. The gray solid line represent a constant acceleration term. model. Meanwhile, the stellar rotation period is 40± 10 days, based on the Lomb-Scargle periodogram of K2 light curve. This suggests that the temporal correlation in the radial- velocity residuals might be caused by radial velocity pertur- bations induced by stellar activity. The BIS and the observed radial velocity variations from HARPS also showed substan- tial correlation (see Fig 3). The Pearson correlation coefficient and the corresponding p-value were calculated to be 0.36 and 0.018, indicating that the observed degree of correlation has only a 2% chance of being produced by uncorrelated noise. To account for the effects of stellar variability, we tried the following strategy. We added to the model a series of sinu- soidal functions of time with periods corresponding to the lowest harmonics of the stellar rotation period (Prot, Prot/2, Prot/3, etc.). We imposed a prior on Prot (40± 10 days). The amplitudes and phases of the sinusoids were allowed to float freely. In order to decide on the number of harmonics to in- clude, we used the BIC as a determinant. The model with the lowest BIC was selected. After experimenting with different numbers of sinusoids to model the stellar activity signal, we found that none of the models with sinusoidal stellar activity signals yielded a lower BIC number than the original model. In other words, the models with additional sinusoidal vari- ability representing stellar activity did not improve the fit by enough to justify the increase in the number of parameters. Additionally, the orbital period of K2-3d and the stellar rota- tion period (45 and 40 days) are close to one another. Thus, the amplitude of the radial velocity signal from K2-3d and the amplitude of rotation-induced radial velocity perturbation are highly degenerate. A much longer time series would be required to disentangle the signals from K2-3d and stellar ac- tivity. Therefore, we reverted to the original model (see Fig. 1 and 2) and concluded that the mass of K2-3d cannot be con- strained with the data at hand. The results for the other plan- ets are Mb = 8.1+2.0 −1.9 M⊕ and Mc = < 4.2 M⊕ (95% conf.). The standard deviation of the radial velocities was 5.5 m s−1. After subtracting the best-fitting three-planet model, the stan- dard deviation of the residuals was reduced to 4.7 m s−1. We then allowed for non-zero eccentricity. The mass con- straints changed to: Mb = 7.7+2.0 −2.0 M⊕, Mc = < 12.6 M⊕ (95% conf.). See Table 1. The eccentricity of planet b was con- strained as eb = 0.21+0.15 −0.12, while the eccentricities of the planet c and d were unconstrained. The circular model had a more favorable BIC than the eccentric model. 5.2. K2-19 FIG. 13.- Radial velocity as a function of time since mid-transit, for each of the planets in the EPIC 205071984 system (open circles are new HARPS data; black circles are new PFS data). In each case, the modeled contributions of the other two planets has been removed, before plotting. In the best-fitting model, planet c has a negative mass, an unphysical result that probably arises from astrophysical or systematic noise. K2-19 or EPIC 201505350 is a G9V star with three tran- siting planet candidates. The outer two planets ("b", with ra- dius 7.74± 0.39 R⊕; and "c", 4.86+0.62 −0.44 R⊕) have periods of 7.9 days and 11.9 days. They are within or near a 3:2 mean- motion resonance (MMR). These two planets were first re- ported by Foreman-Mackey et al. (2015). Vanderburg et al. (2016) and Sinukoff et al. (2015) later revealed a third planet candidate in the system: "d", with a radius of 1.14± 0.13 R⊕ and an orbital period of 2.5 days. Note that the planet can- didates were named based on the order in which they were discovered, rather than orbital distance. K2-19b and K2- 19c have been validated using transit-timing variations and adaptive-optics imaging (Armstrong et al. 2015; Narita et al. 2015; Sinukoff et al. 2015). The transit candidate K2-19d has not been validated in this manner, but the false positive proba- bility is expected to be low because it is associated with a star already known to have multiple transiting planets. 020406080Time (days, BJD 2,457,185)201001020Radial Velocity (m s−1)bcdγSumPFSHARPS02468Time since midtransit (days)20151050510Radial Velocity (m s−1)planet b05101520Time since midtransit (days)105051015Radial Velocity (m s−1)planet c051015202530Time since midtransit (days)15105051015Radial Velocity (m s−1)planet d8 Dai et al. the grouping approach (see Fig 4 and 5). This is because the grouping approach produced a better fit to the data as indi- cated by a ∆BIC of 43 for 68 data points. Additionally, the sinusoidal approach may not be appropriate for the case of K2-19, because the rotation-induced radial velocity perturba- tions might not remain coherent over the observation span of half a year (≈8 rotation cycles). The radial velocity signals of K2-19b and c were clearly −1.6 m s−1 and Kc = detected with semi-amplitudes Kb = 9.6+1.8 7.5+2.1 −2.1 m s−1. Our analysis set an upper bound of Kd < 6.9 m s−1 (95% conf.) on the semi-amplitude of the inner- most and smallest planet K2-19d. The standard deviation of the measured radial velocities was 14.3 m s−1, while the standard deviation of the residuals after subtracting the best- fitting model was reduced to 10.2 m s−1. The planetary masses were then calculated from the semi-amplitudes, orbital peri- ods, and the stellar mass. The masses of K2-19b-d were re- spectively constrained as Mb = 28.5+5.4 −7.1 M⊕ and Md < 14.0 M⊕ (95% conf.). We tested the long-term dy- namical stability using Equations 3 and 4. The best-fitting model has ∆1 + ∆2 ≈ 28 > 18, consistent with stability. For comparison, we also report the various results without apply- ing the 12-day averaging method in Table 2. −5.0 M⊕, Mc = 25.6+7.1 We also tried an eccentric model. The masses of planets b and c increased slightly while the constraint for planet d remained as an upper bound (see Table 2). The eccentricities of planet b and c were both consistent with zero, with upper bounds of eb < 0.66 (95% conf.) and ec < 0.82 (95% conf.). Obviously the eccentricity of planet d was not constrained, either. 5.3. K2-24 Petigura et al. (2016) reported the discovery of K2-24, a G9V star with two sub-Saturn planets close to (or within) a 2:1 MMR. The orbital periods are 21 and 42 days. They mea- sured the planetary radii to be Rb = 5.68± 0.56 R⊕ and Rc = 7.82±0.72 R⊕ by fitting transit models to the K2 light curves. They were also able to validate the planets with adaptive- optics imaging and Doppler spectroscopy. They measured the masses, Mb = 21.0± 5.4 M⊕ and Mc = 27.0± 6.9 M⊕, based on Doppler data obtained with Keck/HIRES. They obtained 32 spectra with a typical internally-estimated radial velocity uncertainty of 1.7 m s−1. We observed K2-24 with PFS from June 25th to July 3rd, 2015, obtaining a total of 16 spectra. The average internally- estimated radial velocity uncertainty was 1.7 m s−1. We also monitored K2-24 with HARPS from June 17th to Septem- ber 11th, 2015, obtaining 10 spectra. The average internally- estimated radial velocity uncertainty was 2.1 m s−1. We ini- tially modeled the PFS and HARPS data only, as an indepen- dent check of the results reported by Petigura et al. (2016). This led to mass constraints of Mb = 30.0+9.1 −9.3 M⊕ and Mc < 21.3 M⊕ (95% conf.). Both of these are consistent with the HIRES results. We then proceeded to perform a joint anal- ysis with the HIRES data, yielding mass constraints of Mb = 19.8+4.5 −6.1 M⊕ (see Fig. 7 and 8). Petigura et al. (2016) found it necessary to include a constant acceler- ation term, γ = −22.5± 9.2 m s−1yr−1, to obtain a satisfactory fit. This may indicate the presence of an additional compan- ion in the system. Our joint analysis led to reduced signifi- cance of this acceleration: γ = −12+10 −10 m s−1yr−1. In particular, the HARPS data which spanned ≈3 months seem to disfa- −4.4 M⊕ and Mc = 26.0+5.8 FIG. 14.- The BIS and radial velocity variations from HARPS for the EPIC 205071984 system. The color coding shows the SNR of the stellar spectra obtained near a wavelength range of 5500 Å. The uncertainty in the BIS is taken as twice the uncertainty of the radial velocity measurement. The Pearson correlation coefficient between RV and BIS were calculated to be −0.26 with p = 0.094. We observed K2-19 from January 28th to July 3rd, 2015, with PFS. We obtained a total of 61 spectra. The average internally-estimated radial velocity uncertainty was 5.0 m s−1. We also observed K2-19 with HARPS from June 12th to June 30th, 2015, obtaining 8 spectra. The average internally- estimated radial velocity uncertainty was 3.8 m s−1. Our radial velocity analysis started with the simplest model, as- suming circular orbits for each planet candidate. The radial- velocity residuals showed strong temporal correlation, sim- ilar to the case of K2-3. Specifically, the most prominent peak in the Lomb-Scargle periodogram of the radial veloc- ity data occurred at 21.6 ± 1.1 days with FAP = 1.7 × 10−3. This is similar to the stellar rotation period of 20.4±2.7 days, which was estimated based on the K2 photometry. Thus, the 21.6±1.1 day signal in the RV data seems likely to be caused by stellar activity. We did not observe a significant correlation between the measured RV and the BIS (see Fig 6) but the test is not conclusive, given that only a few HARPS data points were obtained. To account for the rotation-induced radial velocity pertur- bations, we experimented with the following strategies. We separated the Doppler observations into groups of 12-day du- ration and allowed each group to have an independent radial velocity offset. A similar grouping strategy was employed by Howard et al. (2013) to disentangle the planetary signal of Kepler-78b (Pb ≈ 0.35 days) from the stellar activity sig- nal (Prot ≈ 12.5 days). This grouping strategy is most effec- tive when there is a clear separation in timescales between the planetary signal and the stellar activity signal. This separation is not as extreme for K2-19 as it was for Kepler-78b. The stel- lar rotation period of K2-19 is about 21 days, while the orbital periods of the planets are 2.5, 8 and 12 days. Nonetheless, the stellar rotation period of 21 days is still about twice as long as the longest orbital period of the planets. The choice of grouping in 12-day intervals was made such that each group covered at least one orbital period of the outermost planet. Therefore it should largely preserve the planetary signals. We also tried to model the stellar activity signal explicitly as a series of sinusoidal functions at the lowest harmonics of the stellar rotation period, but for our final analysis we adopted 2010010203040BIS (m s−1)201510505101520RV-mean RV (m s−1)2530354045SNRDoppler Monitoring of five K2 Transiting Planetary Systems 9 vor a constant acceleration (see Fig. 7). When we allowed for non-zero eccentricity for both planets, the best-fitting planet masses increased slightly (see Table 3). The eccen- tricity constraint for the inner planet's orbit was 0.24+0.10 −0.11 (Pe- tigura et al. 2016, 0.24+0.11 −0.11). The eccentricity of the the outer planet's orbit was consistent with zero with an upper bound of < 0.58 (95% conf.) (Petigura et al. 2016, < 0.39 (95% conf.)). We adopted the circular models as they were favored by a ∆BIC of 14. We checked if the observed radial velocity could be caused by stellar activity. The Pearson correlation coefficient and the corresponding p-value were calculated to be −0.14 and 0.70. The large p-value suggests that there is no significant correlation between BIS and radial-velocity variations (see Fig 9). The standard deviation of the radial velocity data was 6.2 m s−1, while the standard deviation of the residu- als after subtracting the best-fitting circular model reduced to 3.2 m s−1. The separation between the two planets, in units of their mutual Hill radius, is ∆ ≈ 16.9 (as defined in Equations 3 and 4). This is well above the minimum value of 2 3 that is needed for dynamical stability, according to the criterion of Gladman (1993). √ 5.4. EPIC 204129699 EPIC 204129699 is G7V star with a 1.26-day hot Jupiter. Grziwa et al. (2015) modeled the K2 photometry and found that the transit trajectory of the planet likely grazes the stellar limb (b = 0.9 − 1.05). They also measured the planetary mass to be Mb = 1.774 ± 0.079 MJup using Doppler observations obtained with FIES and HARPS. We obtained 7 PFS spec- tra from June 26th to July 3rd, 2015, and 10 TRES spectra from April 10th to May 11th, 2015. The average internally- estimated radial velocity uncertainties for PFS and TRES were 1.5 m s−1 and 16 m s−1, respectively. To improve on our knowledge of the stellar parameters for EPIC 204129699, we also obtained a high-SNR spectrum with the High Dispersion Spectrograph (HDS) on the Sub- aru 8.2m telescope on May 30, 2015. From the measurement of equivalent widths of iron lines (Takeda et al. 2002), we es- timated the stellar atmospheric parameters as Teff = 5384± 46 K, logg = 4.410 ± 0.060, and [Fe/H] = 0.21 ± 0.03. We also analyzed the 10 TRES spectra using the procedure de- scribed by Buchhave et al. (2012) and Buchhave & Latham (2015). The results (Teff = 5445± 49 K, logg = 4.52± 0.10, and [Fe/H] = 0.16± 0.08) were consistent with the HDS val- ues. We thus opted to combine these measurements to obtain a weighted average of Teff = 5412± 33 K, logg = 4.44± 0.05, and [Fe/H] = 0.20 ± 0.03. These atmospheric parameters were then converted into estimates for the stellar mass and radius using the empirical relations presented by Torres et al. (2010). Using these relations and Monte-Carlo sampling (Hi- rano et al. 2012a) we obtained M(cid:63) = 1.000 ± 0.064M(cid:12) and R(cid:63) = 0.986± 0.070R(cid:12). These stellar parameters are consis- tent with the Yonsei-Yale stellar-evolutionary models (Yi et al. 2001). By analyzing our PFS and TRES data, we measured the mass of EPIC 204129699b to be Mb = 1.856 ± 0.084 MJup. This is consistent with the results of Grziwa et al. (2015). We then analyzed our PFS and TRES data jointly with the published FIES and HARPS data from Grziwa et al. (2015). The result for the planet mass is Mb = 1.857 ± 0.081 MJup. The RMS of the measured radial velocities was 262 m s−1, whereas the RMS of residual radial velocities was reduced to 18.2 m s−1. Given the short tidal circularization timescale that is expected for such a close-in orbit, the assumption of a cir- cular orbit seems justified. Indeed, our joint analysis favored the circular model over the eccentric model by ∆BIC of 15 for 25 data points, and placed an upper limit on the eccentricity as < 0.027 (95% conf.). See Fig. 10 and 11 and Table 4. 5.5. EPIC 205071984 EPIC 205071984 is a G9V star with three transiting planet candidates reported by Vanderburg et al. (2016) and Sinukoff et al. (2015). The inner planet is on a 9-day orbit with a radius of 5.38 ± 0.35 R⊕. The outer two planets have radii of 3.48 ± 0.97 R⊕ and 3.75 ± 0.40 R⊕ and they are within or close to a 3:2 MMR, with orbital periods of 20.7 and 31.7 days. Adaptive-optics images obtained by Sinukoff et al. (2015) revealed several nearby faint sources; however, through a careful analysis of the K2 pixel data, those authors demonstrated that the dimming events cannot be associated with those faint sources, and are most likely associated with EPIC 205071984. We monitored EPIC 205071984 with PFS from June 25th to July 3rd, 2015, obtaining 6 spectra. The average internally- estimated radial velocity uncertainty was 2.4 m s−1. We also observed EPIC 205071984 with HARPS from April 5th to July 7th, 2015, obtaining 43 spectra. The average internally- estimated radial velocity uncertainty was 2.6 m s−1. Our analysis started with the simplest model assuming circular orbits. The residuals suggested a positive linear trend with time. We added a term to the model representing a constant acceleration, which improved the model fit by ∆BIC ≈ 12 (see Figs.12 and 13). The resulting mass for planet b was Mb = 21.1+5.9 −5.9 M⊕ corresponding to a 3.5σ detection. We did not detect planets c and d. In fact, the best-fitting model sug- gested an unphysical negative mass for planet c (see Fig. 13). We therefore only report upper limits for planet c and d: Mc < 8.1 M⊕ and Md < 35.0 M⊕ both at a 95% confidence level. The RMS of the measured radial velocities was 7.9 m s−1, whereas the RMS of residual radial velocities was reduced to 4.2 m s−1. The Pearson correlation coefficient between RV and BIS is −0.26 with p = 0.094 (see Fig. 14). −9.7 m s−1yr−1. The constant acceleration term was constrained as γ = 34.0+9.9 If this trend is associated with another companion in the system, we can estimate the order of mag- nitude of M sin i as function of semi-major axis a. Assuming a circular orbit and M sin i (cid:28) M(cid:63), then γ ≈ G M sin i/a2. The data suggest that the companion's mass and semi-major axis are such that (cid:16) a (cid:17)2 1 AU M sin i ≈ 60 M⊕ . (6) We then tried allowing for eccentric orbits for the three planets. The mass of planet b increased to 25.0+7.4 −7.0 M⊕, while the upper bounds on planet c and d also increased slightly (see Table 5). We found that the eccentricity of planet b was con- sistent with zero with an upper bound of <0.43 (95% conf.). Due to non-detection of the signals from planet c and d, their eccentricities were unconstrained. 6. DISCUSSION Fig. 15 shows all of the reported measurements of mass and radius for planets smaller than 4R⊕, from our own sample and from the literature. To convey the statistical significance of the measurements, we have set the opacity of the data points 10 Dai et al. Doppler observations. A much longer time series would be required. K2-19 presents a rare opportunity to compare the results of the RV method and the TTV method for planetary mass measurement. So far the majority of exoplanets have their mass measured with one or the other of these two methods. The TTV method seems to be yielding systematically smaller masses than the RV method (Weiss & Marcy 2014). To illus- trate this, we have used different colors in Fig. 15 to represent mass measurements based on the TTV and Doppler meth- ods. Many of the planetary systems unveiled by TTV analysis have surprisingly low densities (Lissauer et al. 2013; Schmitt et al. 2014; Masuda 2014; Jontof-Hutter et al. 2014). Stef- fen (2015) attributed the discrepancy to selection effects: for a given planet size, the RV method tends to pick up the more massive planets, while the TTV method is less strongly bi- ased toward massive planets. This is because the proximity to MMR amplifies the TTV signal, thus making smaller planets detectable. Lee & Chiang (2015) offered a related explana- tion. Since the TTV method is more sensitive to planets in MMR, and convergent disk migration tends to produce plan- ets in resonant pairs, it is possible that the TTV planets are more likely to have undergone disk migration. Consequently, the TTV planets tend to originate further out in the protoplan- etary disk. The outer, thus colder and optically thinner part of the disk is more conducive for the formation of low den- sity planets ("super-puffs"), because the gaseous atmosphere of these planets can readily cool and contract thereby allow- ing more gas to be accreted. Therefore, TTV planets more likely have lower densities. As always it is useful to have at least some systems where more than one measurement technique can be applied. The comparison of TTV and RV measurements for a single system has only been achieved in a few cases: Kepler-11 (Lissauer et al. 2013; Weiss et al. 2015), WASP-47 (Becker et al. 2015; Dai et al. 2015) and Kepler-89 (Hirano et al. 2012b; Masuda et al. 2013; Weiss et al. 2013). For the former two systems, the TTV and Doppler methods agree within reported uncer- tainties. However, for Kepler-89 the TTV mass measurement by Masuda et al. (2013) is lower than the Doppler mass mea- surements reported by both Hirano et al. (2012b) and Weiss et al. (2013). In the former case the discrepancy is 1σ and in the latter case it is nearly 4σ. K2-19 offers another opportunity to compare TTV and RV mass determinations. The system has been analyzed with the TTV method by Narita et al. (2015) and Barros et al. (2015). Using the analytical TTV model presented by Deck & Agol (2015), Narita et al. (2015) found Mc = 21.4± 1.9 M⊕. This is in agreement with our RV mass: Mc = 25.6+7.1 −7.1 M⊕. Barros et al. (2015) derived masses for K2-19b and c by analyzing the short-term synodic signal ("chopping") in the TTV of the system with a photodynamical model. 20 The results were Mb = 44± 12 M⊕ and Mc = 15.9± 7.0 M⊕. These results were based on an assumed stellar mass of 0.918+0.086 −0.070 M(cid:12), whereas we assumed 0.93± 0.05 M(cid:12) based on work by Sinukoff et al. (2015). To facilitate comparison, we rescaled our results un- der the same assumption for the stellar mass as Barros et al. 20 Vanderburg et al. (2016) and Sinukoff et al. (2015) recently revealed an- other planet in the system, K2-19d a super-Earth on a 2.5 day orbit. With its presence in the system, the photodynamical model may need to be revised. However, given that planet d is much smaller and far from mean-motion res- onance with the other two planets, its effect on other two planets should be quite small. FIG. 15.- The mass radius diagram of sub-Neptune planets. The yellow circles indicate the mass measurements from TTV analysis; the black circles indicate the mass measurements from Doppler method. The opacity indicates the quadrature sum of the SNR of the mass and radius measurement of each planet. The mass measurements and upper limits (95% conf.) presented in this paper are annotated with error bars and red arrows. The theoretical mass- radius relationships from Zeng & Sasselov (2013) are also plotted. "Rock" corresponds to a composition of 100% MgSiO3. in proportion to the quadrature sum of the SNR of the radius and mass measurements for each planet. Darker symbols cor- respond to higher SNR. We have also plotted theoretical mass- radius relationships for various compositions, from Zeng & Sasselov (2013). K2-3 is a favorable target for JWST owing to the fact that the host star is relatively bright (Ks = 8.561±0.023) and small (R(cid:63) = 0.561 ± 0.068 R(cid:12)). Our analysis has confirmed the mass measurement of the innermost planet K2-3b by Alme- nara et al. (2015): they found Mb = 8.4± 2.1 M⊕ while we find Mb = 8.1+2.0 −1.9 M⊕. Because we adopted a slightly differ- ent stellar mass from that reported by Almenara et al. (2015), a more direct comparison is between the measured velocity semi-amplitudes K. Almenara et al. (2015) reported Kb = 3.60± 0.87 m s−1, and our result of Kb = 3.36+0.75 −0.74 m s−1 is consistent with theirs. The mass and radius of K2-3b are rem- iniscent of 55 Cnc e. Both planets have dimensions of roughly ≈8 M⊕ and ≈2 R⊕. Using the model of Zeng et al. (2015), the measured mass and radius of K2-3b suggests that the planet may contain up to 60% H2O with 40% MgSiO3. Should the presence of H2O be confirmed by spectroscopic follow-up, interesting questions will be raised. How did the planet acquire its water? Did it form beyond the snow line and migrated inward? Or did it form in situ and acquire its water from cometary impacts? For K2-3c, our analysis placed a upper bound of <4.2 M⊕ with 95% confidence. This upper bound is not strong enough to allow for firm conclusions about the composition, but it is at least suggestive of a composition similar to or less dense than pure rock. As we noted in the previous section, the prox- imity of the orbital period of K2-3d and the stellar rotation period (45 days and 40 days) makes it difficult to disentangle the planetary signal of K2-3d and the stellar activity signal in 1234571020Planet Mass (M⊕)0.51.01.52.02.53.03.54.04.5Planet Radius (R⊕)Zeng & Sasselov 2013Solar SystemMass from RVMass from TTVThis workIronRockWaterK2-3cEPIC 205071984cK2-19dK2-3bDoppler Monitoring of five K2 Transiting Planetary Systems 11 (2015), finding the RV masses to be Mb = 28.1+5.3 = 25.3+7.0 by 1.2σ and 0.9σ, respectively, for planets b and c. −4.9 M⊕ and Mc −7.0 M⊕. These differ from the photodynamical masses In addition to the planetary masses, Barros et al. (2015) = 0.42 ± 0.12. The ratio also reported the mass ratio Mc Mb can be directly constrained based on the observed synodic chopping signal for mildly eccentric, nearly coplanar systems (Nesvorný & Vokrouhlický 2014; Deck & Agol 2015). In contrast, in our RV analysis, the mass ratio is relatively poorly constrained ( Mc Mb = 0.90+0.39 −0.35). The low mean density (0.334+0.081 −0.077 g cm−3) of K2-19b sug- gests a large envelope-to-core mass ratio. The low surface gravity of K2-19b coupled with its relatively large size makes it a potentially favorable target for transmission spectroscopy. K2-19c has a size of 4.86+0.62 −0.44 R⊕ and a mean density of 1.18+0.57 −0.57 g cm−3. The low density suggests that the planet most likely contains a subsantial H/He envelope. Due to the small size of planet d, our current dataset only placed an up- per limit on its mass. More observations are needed before we can discuss the nature of planet d. Petigura et al. (2016) found the masses of K2-24 b and c to be Mb = 21.0± 5.4 M⊕ and Mc = 27.0± 6.9 M⊕. Based on the masses, radii and irradiation levels of the two plan- ets, they were able to estimate the relative envelope-to-core mass ratio by fitting the data with models by Lopez & Fort- ney (2014) for the interior compisition and thermal evolution. They found that the two planets have similar core masses: 17.6± 4.3 M⊕ and 16.1± 4.2 M⊕, while having different en- velope mass fractions of 24± 8% and 48± 9%. The results of their analysis posed intriguing questions about the forma- tion scenarios of the two planets. How did the planets avoid runaway core accretion with core masses of ≈16 M⊕? Why did they end up with such disparate envelope-to-core mass ratios? Our analysis has confirmed and refined the mass mea- surements of the two planets: Mb = 19.8+4.5 −4.4 M⊕ and Mc = 26.0+5.8 −6.1 M⊕, and weakened the detection of a constant ac- celeration term reported by Petigura et al. (2016). However, it does not alter the basic picture painted by Petigura et al. (2016). EPIC 205071984b has very similar properties as K2- 24b. Again, its low density suggests a substantial gaseous envelope. The upper limits for EPIC 205071984 c and d, Mc < 8.1 M⊕ and Md < 35.0 M⊕ both at a 95% confidence level, are reasonable for Neptune-sized planets. The mass of EPIC 204129699b was measured with rela- tively high precision: 1.857± 0.081 MJup. However, due to the 30 min time averaging of the K2 photometric data and the grazing trajectory of the transiting planet, the uncertainties in the transit parameters are larger than usual (Grziwa et al. 2015). As a result, the planetary radius (and mean density) are relatively poorly constrained. Given that the host star is relatively bright (V ≈ 11), a priority for future work should be ground-based transit observations with better time sampling, which will help to break the degeneracies. We thank Kento Masuda for helpful discussions. We thank Saul Rappaport and Tushar Shrotriya for their contributions to the HARPS observing proposal. We are grateful to Xavier Bonfils, Nicola Astudillo and collaborators, with whom we exchanged observing time at ESO's 3.6m telescope, facilitat- ing our observations. We thank Jason Eastman, Emmanuel Jehin and Michaël Gillon for scheduling ground based tran- sit observations with LCOGT and TRAPPIST. Work by FD and JNW was supported by the Transiting Exoplanet Survey Satellite mission. NN acknowledges supports by the NAOJ Fellowship, Inoue Science Research Award, and Grant-in-Aid for Scientific Research (A) (JSPS KAKENHI Grant Number 25247026). This work was performed [in part] under con- tract with the California Institute of Technology (Caltech)/Jet Propulsion Laboratory (JPL) funded by NASA through the Sagan Fellowship Program executed by the NASA Exoplanet Science Institute. AV is supported by the National Science Foundation Graduate Research Fellowship, Grant No. DGE 1144152. Australian access to the Magellan Telescopes was supported by the Australian Federal Government through the Department of Industry and Science. Facilities: Magellan:Clay (Planet Finder Spectrograph) REFERENCES Almenara, J. M., Astudillo-Defru, N., Bonfils, X., et al. 2015, A&A, 581, L7 Armstrong, D. J., Santerne, A., Veras, D., et al. 2015, A&A, 582, A33 Barros, S. C. C., Almenara, J. M., Demangeon, O., et al. 2015, MNRAS, 454, 4267 Becker, J. C., Vanderburg, A., Adams, F. C., Rappaport, S. A., & Schwengeler, H. M. 2015, ApJ, 812, L18 Beichman, C., Livingston, J., Werner, M., et al. 2016, ArXiv e-prints, 381 arXiv:1603.01934 Bodenheimer, P., & Lissauer, J. J. 2014, ApJ, 791, 103 Buchhave, L. A., & Latham, D. W. 2015, ApJ, 808, 187 Buchhave, L. A., Latham, D. W., Johansen, A., et al. 2012, Nature, 486, 375 Butler, R. P., Marcy, G. W., Williams, E., et al. 1996, PASP, 108, 500 Crane, J. D., Shectman, S. A., Butler, R. P., et al. 2010, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 7735, Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, 773553 Crossfield, I. J. M., Petigura, E., Schlieder, J. E., et al. 2015, ApJ, 804, 10 Dai, F., Winn, J. N., Arriagada, P., et al. 2015, ApJ, 813, L9 Deck, K. M., & Agol, E. 2015, ArXiv e-prints, arXiv:1509.08460 Dressing, C. D., Charbonneau, D., Dumusque, X., et al. 2015, ApJ, 800, 135 Fabrycky, D. C., Lissauer, J. J., Ragozzine, D., et al. 2014, ApJ, 790, 146 Foreman-Mackey, D., Hogg, D. W., Lang, D., & Goodman, J. 2013, PASP, 125, 306 Foreman-Mackey, D., Montet, B. T., Hogg, D. W., et al. 2015, ApJ, 806, 215 Fressin, F., Torres, G., Charbonneau, D., et al. 2013, ApJ, 766, 81 Gettel, S., Charbonneau, D., Dressing, C. D., et al. 2016, ApJ, 816, 95 778, 185 Grziwa, S., Gandolfi, D., Csizmadia, S., et al. 2015, ArXiv e-prints, arXiv:1510.09149 Hirano, T., Sanchis-Ojeda, R., Takeda, Y., et al. 2012a, ApJ, 756, 66 Hirano, T., Narita, N., Sato, B., et al. 2012b, ApJ, 759, L36 Howard, A. W., Marcy, G. W., Bryson, S. T., et al. 2012, ApJS, 201, 15 Howard, A. W., Sanchis-Ojeda, R., Marcy, G. W., et al. 2013, Nature, 503, Howell, S. B., Sobeck, C., Haas, M., et al. 2014, PASP, 126, 398 Inamdar, N. K., & Schlichting, H. E. 2015, MNRAS, 448, 1751 Jenkins, J. M., Caldwell, D. A., Chandrasekaran, H., et al. 2010, ApJ, 713, Jontof-Hutter, D., Lissauer, J. J., Rowe, J. F., & Fabrycky, D. C. 2014, ApJ, L87 785, 15 Kovács, G., Zucker, S., & Mazeh, T. 2002, A&A, 391, 369 Latham, D. W., Rowe, J. F., Quinn, S. N., et al. 2011, ApJ, 732, L24 Lee, E. J., & Chiang, E. 2015, ApJ, 811, 41 Liddle, A. R. 2007, MNRAS, 377, L74 Lissauer, J. J., Ragozzine, D., Fabrycky, D. C., et al. 2011, ApJS, 197, 8 Lissauer, J. J., Jontof-Hutter, D., Rowe, J. F., et al. 2013, ApJ, 770, 131 Lopez, E. D., & Fortney, J. J. 2014, ApJ, 792, 1 Lucy, L. B., & Sweeney, M. A. 1971, AJ, 76, 544 Marcy, G. W., Isaacson, H., Howard, A. W., et al. 2014, ApJS, 210, 20 Masuda, K. 2014, ApJ, 783, 53 Masuda, K., Hirano, T., Taruya, A., Nagasawa, M., & Suto, Y. 2013, ApJ, Mayor, M., Pepe, F., Queloz, D., et al. 2003, The Messenger, 114, 20 Dai et al. TABLE 1 SYSTEM PARAMETERS OF K2-3 3896± 189 4.72± 0.13 −0.32± 0.13 0.601± 0.089 0.561± 0.068 12.17± 0.01 b c d(cid:63) 10.05403+0.0026 −0.0025 1913.4189+0.0011 −0.0011 2.14+0.27 −0.26 24.6454+0.0013 −0.0013 1912.2786+0.0026 −0.0027 1.72+0.23 −0.22 44.5631+0.0063 −0.0055 1926.2232+0.0037 −0.0043 1.52+0.21 −0.20 < 1.3 (95% con f .) < 4.2 (95% con f .) < 4.5 (95% con f .) 3.36+0.75 −0.74 8.1+2.0 −1.9 4.5+2.0 −2.0 5.7+1.0 −0.8 2.97+0.58 −0.53 3.26+0.81 < 3.9 (95% con f .) −0.79 7.7+2.0 −2.0 < 12.6 (95% con f .) 4.3+2.0 −2.0 < 13.5 (95% con f .) 0.21+0.15 Unconstrained −0.12 5.4+1.1 −0.9 2.74+0.65 −0.63 1.60+0.80 −0.91 5.2+2.6 −3.0 5.6+3.6 −3.9 Ref. B C B B B A B B B A A A A A A A A A A A A A A A A A A A 1.91+0.74 −0.74 7.5+3.0 −3.0 11.7+6.7 −6.7 2.9+1.5 −1.4 11.3+5.9 −5.8 18+12 −12 0.65+0.17 −0.43 5.0+1.7 −1.6 19.8+6.8 −6.6 31+17 −16 12 Parameter Stellar Parameters Teff (K) log g (dex) [Fe/H] (dex) M(cid:63) ( M(cid:12)) R(cid:63) (R(cid:12)) Apparent V mag Planetary Parameters Transit Model P (days) t0 (BJD-2454900) Rp (R⊕) Circular RV model K (m s−1) Mp ( M⊕) ρ (g cm−3) σjit, PFS (m s−1) σjit, HARPS (m s−1) Eccentric RV model K (m s−1) Mp ( M⊕) ρ (g cm−3) e σjit, PFS (m s−1) σjit, HARPS (m s−1) Circular RV model including stellar activity K (m s−1) Mp ( M⊕) ρ (g cm−3) σjit, PFS (m s−1) σjit, HARPS (m s−1) Period of stellar activity (days) Amplitude of stellar activity (m s−1) Phase of stellar activity (rad) 2.87+0.72 −0.67 6.9+1.9 −1.7 3.8+1.8 −1.7 5.0+1.7 −1.6 2.16+0.62 −0.64 39.7+0.9 −1.0 3.9+1.4 −1.5 2.1+2.8 −1.6 NOTE. - A: This Work; B: Crossfield et al. (2015); C: Sinukoff et al. (2015). (cid:63): The radial velocity result for planet d is highly degenerate with stellar activity signal. See text for details. Meschiari, S., Wolf, A. S., Rivera, E., et al. 2009, PASP, 121, 1016 Mizuno, H. 1980, Progress of Theoretical Physics, 64, 544 Motalebi, F., Udry, S., Gillon, M., et al. 2015, A&A, 584, A72 Narita, N., Hirano, T., Fukui, A., et al. 2015, ApJ, 815, 47 Nesvorný, D., & Vokrouhlický, D. 2014, ApJ, 790, 58 Ofir, A. 2014, A&A, 561, A138 Pepe, F., Cameron, A. C., Latham, D. W., et al. 2013, Nature, 503, 377 Petigura, E. A., Howard, A. W., & Marcy, G. W. 2013, Proceedings of the National Academy of Science, 110, 19273 Petigura, E. A., Howard, A. W., Lopez, E. D., et al. 2016, ApJ, 818, 36 Queloz, D., Henry, G. W., Sivan, J. P., et al. 2001, A&A, 379, 279 Rafikov, R. R. 2006, ApJ, 648, 666 Sanchis-Ojeda, R., Rappaport, S., Winn, J. N., et al. 2014, ApJ, 787, 47 Sanchis-Ojeda, R., Rappaport, S., Pallè, E., et al. 2015, ApJ, 812, 112 Schmitt, J. R., Agol, E., Deck, K. M., et al. 2014, ApJ, 795, 167 Schwarz, U. J. 1978, A&A, 65, 345 Sinukoff, E., Howard, A. W., Petigura, E. A., et al. 2015, ArXiv e-prints, arXiv:1511.09213 Steffen, J. H. 2015, ArXiv e-prints, arXiv:1510.04750 Takeda, Y., Ohkubo, M., & Sadakane, K. 2002, PASJ, 54, 451 Torres, G., Andersen, J., & Giménez, A. 2010, A&A Rev., 18, 67 Vanderburg, A., Latham, D. W., Buchhave, L. A., et al. 2016, ApJS, 222, 14 Weiss, L. M., & Marcy, G. W. 2014, ApJ, 783, L6 Weiss, L. M., Marcy, G. W., & Isaacson, H. T. 2015, in American Astronomical Society Meeting Abstracts, Vol. 225, American Astronomical Society Meeting Abstracts, 409.05 Weiss, L. M., Marcy, G. W., Rowe, J. F., et al. 2013, ApJ, 768, 14 Weiss, L. M., Rogers, L. A., Isaacson, H. T., et al. 2016, ArXiv e-prints, arXiv:1601.06168 Yi, S., Demarque, P., Kim, Y.-C., et al. 2001, ApJS, 136, 417 Zeng, L., & Sasselov, D. 2013, PASP, 125, 227 Zeng, L., Sasselov, D., & Jacobsen, S. 2015, ArXiv e-prints, arXiv:1512.08827 Doppler Monitoring of five K2 Transiting Planetary Systems 13 Parameter Stellar Parameters Teff (K) log g (dex) [Fe/H] (dex) v sin i (km s−1) M(cid:63) ( M(cid:12)) R(cid:63) (R(cid:12)) Apparent V mag Planetary Parameters Transit Model P (days) t0 (BJD-2454900) Rp (R⊕) TABLE 2 SYSTEM PARAMETERS OF K2-19 5430± 60 4.63± 0.07 0.10± 0.04 < 2 0.93± 0.05 0.86± 0.04 13.00± 0.01 b 7.91940± 0.00005 1913.3837± 0.0003 7.74± 0.39 c 11.90715± 0.00150 1917.2755± 0.0051 4.86+0.62 −0.44 d 2.50856± 0.00041 1908.9207± 0.0086 1.14± 0.13 Circular RV model (with 12-day averaging) K (m s−1) Mp ( M⊕) ρ (g cm−3) σjit, PFS (m s−1) σjit, HARPS (m s−1) Circular RV model (without 12-day averaging) K (m s−1) Mp ( M⊕) ρ (g cm−3) σjit, PFS (m s−1) σjit, HARPS (m s−1) 9.6+1.8 −1.6 28.5+5.4 −5.0 0.334+0.081 −0.077 10.0+1.4 −1.2 2.2+2.8 −1.7 9.0+1.7 −1.6 26.2+5.0 −4.8 0.307+0.075 −0.072 11.8+1.5 −1.2 1.9+2.3 −1.3 7.5+2.1 −2.1 25.6+7.1 −7.1 1.18+0.57 −0.57 < 6.9 (95% conf.) < 14.0 (95% conf.) < 51.4 (95% conf.) 6.3+1.9 −2.0 20.9+6.5 −6.6 0.96+0.52 −0.53 < 7.7 (95% conf.) < 15.6 (95% conf.) < 57.3 (95% conf.) Eccentric RV model K (m s−1) Mp ( M⊕) ρ (g cm−3) e σjit, PFS (m s−1) σjit, HARPS (m s−1) NOTE. - A: This work. B: Sinukoff et al. (2015) 7.8+2.9 < 9.7 (95% conf.) −3.2 26.5+9.8 < 9.6 (95% conf.) −10.8 1.26+0.67 −0.70 < 35.2 (95% conf.) Unconstrained < 0.66 (95% conf.) < 0.82 (95% conf.) 11.42+2.3 −2.4 31.8+6.7 −7.0 0.374+0.095 −0.089 9.7+2.4 −1.5 2.2+3.1 −1.6 Ref. B B B B B B B B B B A A A A A A A A A A A A A A A A 14 Dai et al. TABLE 3 SYSTEM PARAMETERS OF K2-24 Parameter Stellar Parameters Teff (K) log g (dex) [Fe/H] (dex) v sin i (km s−1) M(cid:63) ( M(cid:12)) R(cid:63) (R(cid:12)) Apparent V mag 5743± 60 4.29± 0.07 0.42± 0.04 < 2 1.12± 0.05 1.21± 0.11 11.07± 0.13 Planetary Parameters Transit Model P (days) t0 (BJD-2454900) Rp (R⊕) b 20.8851± 0.0003 2005.7948± 0.0007 5.68± 0.56 c 42.3633± 0.0006 2015.6251± 0.0004 7.82± 0.72 Circular RV model K (m s−1) Mp ( M⊕) ρ (g cm−3) γ (m s−1 yr−1) σjit, PFS (m s−1) σjit, HARPS (m s−1) σjit, HIRES (m s−1) Eccentric RV model K (m s−1) Mp ( M⊕) ρ (g cm−3) e γ (m s−1 yr−1) σjit, PFS (m s−1) σjit, HARPS (m s−1) σjit, HIRES (m s−1) NOTE. - A: This Work; B: Petigura et al. (2016). 4.4+1.0 −1.0 26.0+5.8 −6.1 0.30+0.11 −0.11 5.4+1.3 −1.1 30.7+7.1 −6.3 0.35+0.13 −0.12 < 0.58 (95% conf.) 4.25+0.95 −0.94 19.8+4.5 −4.4 0.59± 0.22 −12+10 −10 3.7+1.0 −0.8 3.9+2.1 −1.8 3.57+0.75 −0.59 5.0+1.0 −1.0 23.3+4.8 −4.7 0.69± 0.25 0.24+0.10 −0.11 −14.9+8.3 −8.1 3.43+0.96 −0.71 3.6+2.0 −1.6 3.13+0.67 −0.57 Ref. B B B B B B B B B B A A A A A A A A A A A A A A A Doppler Monitoring of five K2 Transiting Planetary Systems 15 SYSTEM PARAMETERS OF EPIC 204129699 TABLE 4 Parameter Stellar Parameters Teff (K) log g (dex) [Fe/H] (dex) v sin i (km s−1) M(cid:63) ( M(cid:12)) R(cid:63) (R(cid:12)) Apparent V mag 5412± 34 4.44± 0.05 0.20± 0.03 2.6+1.0 −1.0 1.000± 0.064 0.986± 0.070 10.775± 0.023 Planetary Parameters Transit Model P (days) t0 (BJD-2454900) Rp (R⊕) b 1.257850± 0.000002 2291.70889± 0.00024 Unconstrained Circular RV model K (m s−1) Mp ( M⊕) ρ (g cm−3) σjit, PFS (m s−1) σjit, HARPS (m s−1) σjit, TRES (m s−1) σjit, FIES (m s−1) Eccentric RV model K (m s−1) Mp ( M⊕) ρ (g cm−3) e σjit, PFS (m s−1) σjit, HARPS (m s−1) σjit, TRES (m s−1) σjit, FIES (m s−1) 349.5+3.1 −3.7 590.1+25.7 −25.9 Unconstrained 14.9+8.6 −4.2 9+24 −6 28+12 −8 3.3+5.1 −2.3 348.5+3.4 −3.6 564.2+8.3 −9.5 Unconstrained < 0.027 (95% con f .) 16+11 −5 7+20 −5 29+12 −8 3.3+5.3 −2.4 Ref. A A A A A A A B B B A A A A A A A A A A A A A A A NOTE. - A: This Work; B: Grziwa et al. (2015). 16 Dai et al. SYSTEM PARAMETERS OF EPIC 205071984 TABLE 5 Parameter Stellar Parameters Teff (K) log g (dex) [Fe/H] (dex) v sin i (km s−1) M(cid:63) ( M(cid:12)) R(cid:63) (R(cid:12)) Apparent V mag 5315± 60 4.43± 0.07 0.00± 0.04 < 2 0.87± 0.04 0.87± 0.05 12.332± 0.017 Planetary Parameters Transit Model P (days) t0 (BJD-2454900) Rp (R⊕) b 8.99218± 0.00020 2000.9258± 0.0009 5.38± 0.35 c 20.65614± 0.00598 1999.4306± 0.010 3.48+0.97 −0.42 d 31.71922± 0.00236 903.7846± 0.0031 3.75± 0.40 Circular RV model K (m s−1) Mp ( M⊕) ρ (g cm−3) γ (m s−1 yr−1) σjit, PFS (m s−1) σjit, HARPS (m s−1) Eccentric RV model K (m s−1) Mp ( M⊕) ρ (g cm−3) e γ (m s−1 yr−1) σjit, PFS (m s−1) σjit, HARPS (m s−1) NOTE. - A: This work; B: Sinukoff et al. (2015) < 2.1 (95% conf.) < 8.1 (95% conf.) < 1.1 (95% conf.) < 7.8 (95% conf.) < 35.0 (95% conf.) < 3.6 (95% conf.) < 4.3 (95% conf.) < 16.5 (95% conf.) < 2.1 (95% conf.) Unconstrained < 8.9 (95% conf.) < 40.3 (95% conf.) < 4.2 (95% conf.) Unconstrained 7.1+2.0 −2.0 21.1+5.9 −5.9 0.74+0.25 −0.25 34.0+9.9 −9.7 5.4+3.5 −2.1 3.37+0.69 −0.58 8.6+2.6 −2.4 25.0+7.4 −7.0 0.88+0.31 −0.30 < 0.43 (95% conf.) 34+13 −13 7.2+4.3 −2.8 3.26+0.75 −0.62 Ref. B B B B B B B B B B A A A A A A A A A A A A A Doppler Monitoring of five K2 Transiting Planetary Systems 17 TABLE 6 RELATIVE RADIAL VELOCITY OF K2-3 BJD 2457050.75983 2457051.80129 2457052.84973 2457054.80341 2457055.83610 2457061.76506 2457062.82087 2457064.74697 2457065.75519 2457066.72884 2457067.74709 2457068.74840 2457069.75326 2457118.61730 2457118.62796 2457118.63900 2457119.62004 2457119.63088 2457119.64192 2457120.60464 2457120.61554 2457120.62655 2457121.58823 2457121.59876 2457121.61042 2457122.59521 2457122.60651 2457122.61786 2457123.60398 2457123.61534 2457123.62646 RV (m s−1) Unc. (m s−1) 2.51 2.35 2.61 4.38 2.81 3.02 2.33 2.63 2.57 2.99 2.74 2.14 2.51 2.34 2.56 2.64 1.95 2.24 2.23 1.91 2.02 1.88 2.65 2.55 2.46 2.26 2.21 2.59 2.56 2.22 2.19 −8.34 0.00 −0.10 2.80 −3.97 13.48 −2.22 −4.85 −5.97 −9.21 −10.36 −8.44 1.86 0.27 0.72 3.64 4.87 −2.53 1.86 17.54 6.90 1.24 0.75 −1.31 −3.45 −2.62 4.05 −3.06 4.08 −4.13 0.31 Source 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 NOTE. - 0: PFS 18 Dai et al. RELATIVE RADIAL VELOCITY OF K2-19 TABLE 7 RV (m s−1) Unc. (m s−1) BIS. (m s−1) Source 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 BJD 2457185.51808 2457186.50677 2457187.53841 2457188.54280 2457189.57695 2457190.57602 2457191.59452 2457203.50361 2457198.47072 2457198.52715 2457199.51333 2457199.52804 2457200.52892 2457200.54451 2457203.52075 2457203.53537 2457204.49376 2457204.50816 2457205.50352 2457205.51810 2457206.49605 2457206.50796 RV (m s−1) Unc. (m s−1) BIS. (m s−1) 41.05 43.01 29.47 16.00 37.20 28.18 29.63 3.74 0.01 −4.75 −8.46 −1.20 −1.17 11.82 −11.02 −10.79 −6.69 −15.31 −1.28 −9.33 10.03 −29.66 −47.19 −5.84 −20.38 −20.85 −5.42 11.04 10.19 3.19 4.83 3.55 3.50 3.56 3.11 3.99 4.73 5.13 4.53 4.49 5.21 5.14 7.79 5.43 6.68 4.48 4.80 4.69 4.48 4.13 4.49 Source 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 BJD −8.88 2457050.85856 −15.46 2457051.81505 −29.96 2457052.86521 0.96 2457054.81634 3.82 2457055.85049 20.63 2457061.70681 17.55 2457061.79848 30.88 2457061.81294 6.73 2457062.77601 13.98 2457062.79007 10.52 2457062.80443 8.89 2457064.76307 17.62 2457064.77760 30.78 2457064.79190 33.43 2457065.76985 23.95 2457065.78424 30.89 2457065.79889 26.22 2457066.75799 30.65 2457066.77275 4.93 2457066.78723 −10.04 2457067.76192 −22.25 2457067.77624 −0.95 2457067.79082 14.88 2457068.76326 2.66 2457068.77757 −6.31 2457068.79244 −23.29 2457069.76793 −11.48 2457069.78244 −12.98 2457069.79654 −13.28 2457118.65185 −1.10 2457118.66628 0.00 2457118.68049 −10.47 2457119.65552 −7.98 2457119.67002 10.92 2457119.68417 6.03 2457120.63937 7.85 2457120.65355 0.24 2457121.62451 8.99 2457121.63942 19.60 2457121.65395 6.10 2457122.62934 −16.70 2457122.64080 −6.75 2457122.65206 −16.07 2457123.63854 4.63 2457123.66345 −5.18 2457123.67839 −21.64 2457124.67522 NOTE. - 0: PFS; 1: HARPS 5.19 3.42 4.76 6.52 4.55 6.77 4.50 5.05 4.57 4.42 5.10 4.15 4.27 4.92 4.99 4.94 4.83 5.42 6.15 7.64 5.16 5.54 5.19 5.04 5.19 5.14 5.11 5.13 5.01 4.09 4.78 4.24 3.85 3.89 4.36 3.72 5.47 4.29 3.82 3.88 5.34 4.84 5.11 5.56 6.12 5.88 8.47 Doppler Monitoring of five K2 Transiting Planetary Systems 19 RELATIVE RADIAL VELOCITY OF K2-24 TABLE 8 BJD 2457198.61435 2457198.69058 2457199.60735 2457199.76091 2457200.70251 2457200.71696 2457202.64632 2457202.66105 2457203.55168 2457203.56610 2457204.67286 2457204.68711 2457205.60288 2457205.61494 2457206.63194 2457206.64440 2457190.62372 2457190.81012 2457191.63661 2457191.64804 2457249.50031 2457253.46724 2457266.49748 2457267.54626 2457270.48326 2457277.47435 RV (m s−1) Unc. (m s−1) BIS. (m s−1) 9.03 9.35 3.53 5.34 −3.16 4.18 0.00 −5.46 8.16 2.72 −3.11 −0.53 0.87 −0.73 −1.82 −0.84 1.86 3.51 −2.34 4.43 −4.52 4.10 −7.77 −3.18 −1.91 5.83 1.79 1.91 1.77 1.79 1.43 1.56 1.75 1.91 1.43 1.52 1.54 1.53 1.54 1.57 1.86 1.69 1.33 1.68 2.55 2.41 3.79 1.52 2.48 1.56 1.47 1.86 23.05 35.98 19.09 21.81 37.23 37.28 32.82 22.12 41.09 26.13 Source 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 RELATIVE RADIAL VELOCITY OF EPIC 204129699 TABLE 9 BJD 2457199.59189 2457201.60623 2457202.63203 2457203.63081 2457204.65786 2457205.59027 2457206.61979 2457122.88109 2457144.86865 2457145.82832 2457146.83816 2457148.89470 2457149.88110 2457150.88949 2457151.88311 2457152.83368 2457153.85085 RV (m s−1) Unc. (m s−1) 1.64 2.22 1.33 1.34 1.21 1.28 1.46 13.52 16.65 14.09 14.24 14.03 24.69 16.81 14.66 14.12 13.52 −284.80 289.83 368.56 0.00 −277.36 −24.70 347.99 −487.18 199.73 −132.03 −449.60 −63.91 224.15 −3.44 −410.65 −340.50 0.00 Source 0 0 0 0 0 0 0 2 2 2 2 2 2 2 2 2 2 NOTE. - 0: PFS; 1: HARPS NOTE. - 0: PFS; 2: TRES 20 Dai et al. RELATIVE RADIAL VELOCITY OF EPIC 205071984 TABLE 10 RV (m s−1) Unc. (m s−1) BIS. (m s−1) 2.31 2.56 2.16 2.40 2.09 2.67 2.65 2.34 2.73 4.14 3.49 2.97 2.84 3.18 1.66 2.41 2.22 3.13 2.19 1.73 3.83 2.62 3.66 2.19 1.71 1.53 1.53 1.78 1.67 1.73 2.04 2.18 2.22 3.27 2.14 2.42 3.20 3.98 3.69 4.51 2.74 1.89 1.59 2.31 2.50 2.99 2.15 1.92 3.48 14.16 8.36 1.37 1.26 7.83 1.12 13.94 0.83 12.95 19.73 11.51 9.47 16.38 5.09 17.25 10.10 -8.32 -9.69 22.02 9.36 32.01 7.80 4.34 19.98 21.29 6.21 4.07 0.45 24.39 -2.81 22.16 14.21 13.14 8.10 11.57 15.52 4.78 19.41 8.17 -6.49 15.49 7.98 -4.74 Source 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 BJD 2457198.67460 2457199.74554 2457200.68722 2457204.72413 2457205.64081 2457206.66905 2457185.60690 2457185.67718 2457185.81974 2457185.84975 2457186.59974 2457186.66981 2457186.79015 2457186.82363 2457187.67459 2457187.74382 2457187.78138 2457187.81506 2457188.63849 2457188.69403 2457188.75958 2457188.79906 2457188.83141 2457189.64108 2457189.69856 2457189.75643 2457189.78890 2457189.82294 2457190.64915 2457190.68400 2457190.71716 2457190.74988 2457190.78420 2457191.66878 2457191.70642 2457191.76095 2457191.79493 2457191.82491 2457192.65964 2457198.72143 2457198.80982 2457199.60136 2457199.68346 2457211.70008 2457253.48543 2457264.51816 2457267.56388 2457270.50147 2457278.56859 NOTE. - 0: PFS; 1: HARPS −13.95 −24.28 −12.24 0.00 2.07 2.37 10.69 10.92 9.38 3.16 1.56 8.09 0.48 6.96 5.04 2.94 0.31 8.21 6.46 5.69 −8.45 −1.28 10.79 −1.58 −1.46 −3.27 −8.02 −8.48 −6.05 −2.64 −7.75 −3.93 −8.97 −10.48 −13.28 −4.02 −4.92 −0.74 −2.13 −2.16 −11.76 −9.32 −13.58 11.06 −0.60 6.03 15.81 −0.73 12.00
1605.03941	1	1605	2016-05-12T19:25:47	A Deep Search for Additional Satellites around the Dwarf Planet Haumea	[ "astro-ph.EP", "astro-ph.IM" ]	Haumea is a dwarf planet with two known satellites, an unusually high spin rate, and a large collisional family, making it one of the most interesting objects in the outer solar system. A fully self-consistent formation scenario responsible for the satellite and family formation is still elusive, but some processes predict the initial formation of many small moons, similar to the small moons recently discovered around Pluto. Deep searches for regular satellites around KBOs are difficult due to observational limitations, but Haumea is one of the few for which sufficient data exist. We analyze Hubble Space Telescope (HST) observations, focusing on a ten-consecutive-orbit sequence obtained in July 2010, to search for new very small satellites. To maximize the search depth, we implement and validate a non-linear shift-and-stack method. No additional satellites of Haumea are found, but by implanting and recovering artificial sources, we characterize our sensitivity. At distances between $\sim$10,000 km and $\sim$350,000 km from Haumea, satellites with radii as small as $\sim$10 km are ruled out, assuming an albedo ($p \simeq 0.7$) similar to Haumea. We also rule out satellites larger than $\gtrsim$40 km in most of the Hill sphere using other HST data. This search method rules out objects similar in size to the small moons of Pluto. By developing clear criteria for determining the number of non-linear rates to use, we find that far fewer shift rates are required ($\sim$35) than might be expected. The non-linear shift-and-stack method to discover satellites (and other moving transients) is tractable, particularly in the regime where non-linear motion begins to manifest itself.	astro-ph.EP	astro-ph	Search for Additional Satellites around Haumea Preprint typeset using LATEX style emulateapj v. 01/23/15 A DEEP SEARCH FOR ADDITIONAL SATELLITES AROUND THE DWARF PLANET HAUMEA Luke D. Burkhart1,2, Darin Ragozzine1,3,4, Michael E. Brown5 Search for Additional Satellites around Haumea ABSTRACT Haumea is a dwarf planet with two known satellites, an unusually high spin rate, and a large collisional family, making it one of the most interesting objects in the outer solar system. A fully self-consistent formation scenario responsible for the satellite and family formation is still elusive, but some processes predict the initial formation of many small moons, similar to the small moons recently discovered around Pluto. Deep searches for regular satellites around KBOs are difficult due to observational limitations, but Haumea is one of the few for which sufficient data exist. We analyze Hubble Space Telescope (HST) observations, focusing on a ten-consecutive-orbit sequence obtained in July 2010, to search for new very small satellites. To maximize the search depth, we implement and validate a non-linear shift-and-stack method. No additional satellites of Haumea are found, but by implanting and recovering artificial sources, we characterize our sensitivity. At distances between ∼10,000 km and ∼350,000 km from Haumea, satellites with radii as small as ∼10 km are ruled out, assuming an albedo (p ≃ 0.7) similar to Haumea. We also rule out satellites larger than &40 km in most of the Hill sphere using other HST data. This search method rules out objects similar in size to the small moons of Pluto. By developing clear criteria for determining the number of non-linear rates to use, we find that far fewer shift rates are required (∼35) than might be expected. The non-linear shift-and-stack method to discover satellites (and other moving transients) is tractable, particularly in the regime where non-linear motion begins to manifest itself. Keywords: Kuiper belt objects: individual (Haumea) -- methods: data analysis -- techniques: image processing 1. INTRODUCTION The dwarf planet Haumea (Rabinowitz et al. 2006), its two moons (Brown et al. 2005, 2006), and its col- lisional family (Brown et al. 2007) provide important constraints on the formation of Kuiper belt and outer solar system. This well-studied object is the fastest- rotating large body in the solar system (Rabinowitz et al. 2006) with rotational variability in color (Lacerda et al. 2008; Lacerda 2009), an unexpectedly high density (Lockwood et al. 2014), and large albedo (Elliot et al. 2010). It has two moons on dynamically excited orbits (Ragozzine & Brown 2009, hereafter RB09) which have scaled mass ratios and distances similar to the Earth- Moon system. Dynamical, photometric, and spectro- scopic observations of objects in the vicinity of Haumea clearly indicate a collisional family of icy fragments with similarly high albedos (Ragozzine & Brown 2007; Schaller & Brown 2008; Snodgrass et al. 2010). How- ever, though the expected dispersion velocity of these fragments is of order several hundred meters per sec- ond, the observed dispersion is well constrained within ∼150 m s−1. The apparent lack of high velocity ejecta is confirmed by observational surveys and dynam- [email protected] 1 Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138, USA 2 Yale University, Department of Physics, 217 Prospect St, New Haven, CT 06511, USA 3 University of Florida, 211 Bryant Space Science Center, Gainesville, FL 32611, USA 4 Florida Institute of Technology, Department of Physics and Space Sciences, 150 West University Boulevard, Melbourne, FL 32901, USA 5 California Institute of Technology, Division of Geological and Planetary Sciences, MC 150-21, Pasadena, CA 91125, USA ical studies (e.g., Fraser & Brown 2012; Lykawka et al. 2012; Volk & Malhotra 2012), though it is possible that some high velocity ejecta would be unrecognizable dy- namically (Marcus et al. 2011) and/or compositionally (Carry et al. 2012; Brown et al. 2012). There is no simple high-probability formation scenario that naturally explains all of these observational con- straints: Haumea's rapid near-breakup rotation rate, the two moons on distant and dynamically warm orbits, and a collisional family that is an order of magnitude smaller in velocity dispersion than expected. Though multiple explanations and variations have been pro- posed (e.g., Levison et al. 2008; Schlichting & Sari 2009; Leinhardt et al. 2010; Marcus et al. 2011; Ortiz et al. 2012; ´Cuk et al. 2013), none have adequately and self- consistently explained all of the unique features of this interesting system and its family. Attempting to place the formation of the Haumea sys- tem in context with other similar systems in the Kuiper belt quickly leads to comparisons with Kuiper belt ob- jects (KBOs) of similar sizes, particularly Eris, Pluto, and Makemake. Of these, Pluto is the best understood due to a wealth of observational data and the recent flyby by the New Horizons Mission (Stern et al. 2015). Furthermore, there are similiarities between some of the theories for the formation of Haumea's satellites (e.g. Leinhardt et al. 2010) and for the formation of Pluto's satellites (e.g. Canup 2011): both suggest a relatively large impactor with very low incoming velocities that undergo a grazing collision to form a satellite system. With the discovery of a retinue of small satellites exte- rior to Charon's orbit -- now dubbed Styx, Nix, Ker- beros, and Hydra -- there is a renewal in interest in 2 observational constraints on the formation of the Pluto system (Weaver et al. 2006; Showalter et al. 2011, 2012; Showalter & Hamilton 2015). Standard explanations for the formation of Nix and Hydra were already problematic (Ward & Canup 2006; Lithwick & Wu 2008), and the characteristics of Styx and Kerberos are even more puz- zling (Pires dos Santos et al. 2012; Kenyon & Bromley 2014; Cheng et al. 2014b). For example, in the cur- rent orbital configuration, the dynamical stability of Styx requires that Charon's eccentricity at its present semi-major axis was never above ∼0.035, using the circumbinary stability criterion of Holman & Wiegert (1999, see Equation 3). Thus, the discovery of Styx combined with dynamical stability immediately pre- cludes some of the more extreme proposed orbital histories of Cheng et al. (2014a) if Styx formed con- currently with Charon (Cheng et al. 2014b). Long- term dynamical stability can also place some of the best constraints on the masses of these small moons (Youdin et al. 2012; ´Cuk et al. 2013; Porter & Stern 2015; Showalter & Hamilton 2015). The discovery of small moons around Pluto and their ability to add constraints to the understanding of this system suggests that all asteroid and KBO binaries and triples be searched for additional moons. We recommend the continuation of this standard practice, even when an initial companion is identified. For KBOs, satellite searches are observationally difficult for multiple reasons. First, acquiring data of sufficient depth and resolution to identify faint moons of faint KBOs usually requires a considerable amount of time at the best telescopes in the world, such as the Hubble Space Telescope (HST) or 8-10 meter class telescopes with Laser Guide Star Adaptive Optics. The only KBOs with large amounts of continu- ous high-quality data are Pluto and Haumea. Second, the discovery of small moons can be frustrated by their a priori unknown satellite orbital motion dur- ing long exposures. Faint, fast-moving moons can then evade detection even with the best data, using standard analysis methods. Therefore, an enhanced methodology to search for faint moving moons is required. In an attempt to better understand the formation of the Haumea system, we use a large set of consec- utive HST observations to perform a search for very small moons around Haumea similar to those discov- ered around Pluto (§2). To search for faint, fast-moving moons well below the single-exposure limit, we imple- mented the non-linear shift-and-stack method proposed by Parker & Kavelaars (2010, hereafter PK10) for the discovery of KBOs (§3). Adapted to the problem of find- ing additional satellites, this method was both efficient and effective. Though no additional satellites of Haumea were detected (§4), with careful characterization of this null detection, we set strong limits on the size and lo- cation of possible undiscovered moons (§5) and discuss the implications for understanding of Haumea's satellite system (§6). 2. OBSERVATIONS In determining the ideal set of observations for a deep satellite search, a balance must be struck between includ- ing the largest number of observations and considering the motion of putative satellites during the total observa- tional baseline. The standard stacking method of adding images that have been co-registered to the position of the primary to enhance sensitivity to faint satellites is limited to observational arcs where the satellite's relative posi- tion remains within a region not much larger than ∼1 Point Spread Function (PSF) Full Width at Half Max- imum (PSF FWHM). Our use of non-linear shift-and- stack can mitigate this problem significantly and allows us to perform a sensitive search on longer timescales. In particular, our HST Program 12243 observed with a wide filter for 10 consecutive orbits and is an excellent dataset for a deep satellite search; with the technique discussed below, we can search these observations even for close- in satellites which traverse a significant fraction of an orbit during the 15-hour baseline, corresponding to sev- eral PSF widths. These observations are our main focus as they are clearly the best for a deep satellite search (§2.1); however, we also inspected other observations for the additional satellites of Haumea (§2.2). 2.1. HST 12243: 10 Orbits in July 2010 HST Program 12243 obtained 10 orbits of observations over the course of ∼15 hours in July 2010. This pro- gram used the Wide Field Camera 3 (WFC3) UVIS im- ager with the F350LP (long-pass) filter. The primary goal of these observations was the the detection of a mu- tual event between the inner moon Namaka and Haumea (RB09). In order to produce high cadence time series photometry of the proposed mutual event (and to avoid saturation), exposures were limited to ∼45 seconds. To prevent a costly memory buffer download, only a 512x512 subarray of the full WFC3 camera was used with a field of view of ∼20.5 arcseconds. HST tracked at Haumea's rate of motion (except for controlled dithering) to maintain its position near the center of the field of view throughout the observations. The geocentric distance to Haumea at time of observations was 50.85 AU. At this distance, 1" corresponds to 36900 km, 1 WFC3 pixel (.04 arcseconds) corresponds to 1475 km, and the entire subarray field of view corresponds to ∼750000 km. Parts of the last two orbits were affected by the South Atlantic Anomaly. This caused a portion of the these orbits to lose data entirely, and another portion was severely affected with cosmic rays and loss of fine point- ing precision. The most offensive frames were discarded for the purpose of the satellite search, leaving 260 indi- vidual exposures. The center of Haumea was identified by eye in combi- nation with a 2-d Gaussian fitting routine. With these well-determined preliminary Haumea locations, all im- ages were co-registered to Haumea's position. In this Haumea-centric frame, cosmic rays and hot pixels were identified by significant changes in brightness at a par- ticular position using robust median absolute deviation filters. A detailed and extensive image-by-image investi- gation of cosmic rays by eye confirmed that this method was very accurate at identifying cosmic rays and other anomalies. Furthermore, the automatic routine did not flag the known objects (Haumea and its two moons: Hi'iaka and Namaka) as cosmic rays, nor were any other specific localized regions identified for consistent masking (i.e., putative additional satellites were not removed). TinyTim software6 was used to generate local point- 6 http://www.stsci.edu/hst/observatory/focus/TinyTim spread function (PSF) models. As in RB09, these PSF models were then fit to the three known objects using standard χ2 minimization techniques (Markwardt 2009). This identified the best-fit locations and heights of the scaled PSFs, with bad pixels masked and thus not in- cluded in the χ2 calculation. The astrometric positions of the known satellites relative to Haumea seen by this method were in clear accord with their projected or- bital motion from RB09. Despite Namaka's nearness to Haumea (which was purposely chosen, as the goal of the observations was to observe a mutual event), it is easily distinguishable in the first few orbits. The best fit PSFs are visually inspected and found to be good fits for all images. These PSFs are then subtracted, removing a large portion of the three sig- nals, but leaving non-negligible residuals; these residu- als are caused by imperfect PSFs (note that Haumea may be marginally resolved in some images) and stan- dard Poisson noise. Using the updated best-fit cen- ters of Haumea, these PSF-subtracted images are re- coregistered to Haumea's best-fit location (though the additional shifts from the preliminary 2-d Gaussian cen- ters are very small). As described below, these same PSF-subtracted images are used to perform the non- linear shift-and-stack. Moons with negligble motion relative to Haumea can be identified in this image by stacking all the observa- tions in the Haumea-centered reference frame (includ- ing fractional pixel shifts implemented by IDL's fshift). Throughout this investigation, we create stacks by per- forming a pixel-by-pixel median of the images, which is less sensitive to cosmic rays, bad pixels, and errors in the PSF subtraction of the known bodies. This results in a small decrease in sensitivity, as, assuming white noise, the noise level grows a factor of p π 2 faster when using the median over the mean. This will correspond to only a ∼ 20% difference in brightness sensitivity, or a ∼ 10% difference in radius, and we find this acceptable so that the other effects mentioned above can be mitigated. A portion of this stacked image around Haumea is shown in Figure 1. Detailed investigation of this deep stack by eye by each co-author yielded no clear satellite candidates. The median stacks were also automatically searched us- ing the IDL routine find, which uses a convolution filter with FWHM of 1.6 pixels to identify positive brightness anomalies. Detections with SNR of 5 or greater were ex- amined; none were found that were consistent with an additional satellite (e.g., having PSF-like shape). Scal- ing from the SNR of Haumea and using a more conser- vative detection limit SNR of 10, this places a limit on non-moving satellites as fainter than about V ≃ 27.1, corresponding to Haumea satellites with radii less than about 8 km (see §5). 2.2. Other Observations HST has observed Haumea during many programs for multiple reasons. Program 11518 was proposed to obtain astrometry of both moons and is 5 independent orbits of observations spread over 2 weeks (RB09). Although it would be interesting to investigate the possibility of combining these data in a long-baseline non-linear shift- and-stack, given the existence of other more sensitive datasets, we investigated only the single-orbit median stacked images. Motion during a single 45-minute HST 3 Figure 1. A portion of the median stack of 260 images from 10 orbits of HST WFC3 data (Program 12243). Individual images are co-registered to be stationary in the Haumea-centric frame, with best-fit TinyTim PSFs for Haumea, Hi'iaka and Namaka subtracted. The brightness has been stretched significantly to highlight the residuals. These residuals will limit sensitivity near Haumea, but the diffraction spikes and the majority of the PSFs have been removed. Above and to the left of Haumea lie the residu- als from Hi'iaka which is 1.23" away (45600 km projected distance) in this stack. Vertical darker columns are due to minor uncorrected pixel sensitivity. The image is aligned so that Astronomical North is up. orbit is small compared to the PSF width, even for the shortest satellite orbital periods. HST Program 11971 was 5 consecutive orbits and HST Program 12004 was 7 consecutive orbits, both attempts to observe the last satellite-satellite mutual events. The latter program was within a few weeks of the HST 4th Servicing Mission but was still executed. Unfortunately, for 6.5 of the 7 orbits, the STIS shutter was closed and no on-sky data were taken. For the 5-orbit Wide Field Planetary Camera 2 ob- servation of Program 11971, we median-stacked images centered on Haumea and searched for additional sources by eye and using IDL's find as described above. We investigated stacks of individual orbits and of the entire 5-orbit sequence and found no sources consistent with ad- ditional satellites. Though the non-linear shift-and-stack method below could fruitfully be applied to these obser- vations, the WFC3 observations are considerably deeper and we opted to focus on our best dataset. Finally, we obtained some long-duration (∼5 hours) observations of Haumea using the Laser Guide Star Adaptive Optics system at the Keck Observatory. Co- registered stacks of this data also showed no clear addi- tional satellites, though the known satellites were very easily detected. 3. METHODS For the detection of faint bodies, with signal-to-noise ratio (SNR) per image of .5, a useful approach is the co- addition ("stacking") of multiple images. With the 260 images in our dataset, this method can increase the SNR √260 ≈ 13, thereby searching for satellites with by ∼q 2 radii ∼ √13 ≈ 3.6 times as small as could be detected in π a single image. If the object does not remain apparently stationary (within .1 FWHM) over the course of the observation, the simple co-addition will result in insufficient overlap between images to yield the expected increase in SNR. 4 If the motion of the object is known, images can be first shifted to compensate for this motion, and the im- ages added with the object localized regaining nearly the full sensitivity: this is the meaning of "shift-and- stack". Linear searches with shift-and-stack have been used to discover satellites in the past (Kavelaars et al. 2004; Holman et al. 2004; Showalter et al. 2013) al- though these searches did not need to use the non-linear shift-and-stack method we employ below. In a situation where the motion is unknown, such as a broad search for KBOs, a large set of possible paths on the sky can be considered, and each path independently used as the basis for a shift-and-stack, as described by PK10. A composite image results from each proposed orbital path (which we call a "sky track"), and each stack can be searched for faint satellites which emerge from the noise due to shifting the image accurately enough to (mostly) compensate for its motion. To minimize statistical false-positives and to increase computational tractability, it is important to identify a near-minimal number of sky tracks that will faithfully reproduce all the possible motions without performing redundant searches. PK10 suggested an algorithm for identifying the most important non-redundant set of sky tracks, which we fruitfully employ: generate a large num- ber of random sky tracks based on the full range of ex- pected motion (within desired search parameters) and then remove tracks that are similar to one another. We have adapted this technique for our search. There is a distinction between a general KBO search and a satellite search, which, it is important to note, is largely ignored in the method presented here. This distinction is that, in a broad KBO search, a sky track could be valid for any part of an image; that is, there is little correlation between position and motion. This is not the case for a satellite orbiting a given primary, in which a specific motion only applies to a small spa- tial region. The more highly curved tracks are, the more specific to a particular region they are -- a curved or- bital arc translated to the other side of Haumea would not make physical sense. The method described below involves shifting and stacking the entirety of each image, and searching the whole of the composite image, when in fact the track upon which the shift-and-stack is based applies to only a small subset of each image. In addi- tion to the computational cost of shifting and searching larger images than is necessary, this overuse of the im- ages could potentially result in an increase in statistical false-positives. However, neither of these effects mani- fest in a noticeable way -- neither computation time nor an abundance of false-positives limit our search method. This suggests that we are near the optimal minimum number of sky tracks searched, or have at least reached an acceptably small number. Discussed in greater detail below, an overview of our search algorithm is as follows: 1. Generate a large bank of physically reasonable pu- tative sky tracks by randomly selecting from plau- sible Keplerian satellite orbital parameters. 2. Fit each sky track with non-linear polynomials in time (shift rates). If the shift rates for two distinct sky tracks are similar enough (quantified below), discard one. 3. Continue searching for sky tracks until a nearly- complete non-redundant set is identified. 4. For each track, create a composite image. This is done by overlaying the dataset (in our case, 260 images) upon itself, with the images shifted by the appropriate shift rates such that an object on that track will appear in the same place in each image. Co-add the images into one composite. 5. Search each composite image for satellite candidate sources. The use of non-linear polynomial fits allows the shift rates to more accurately capture curved orbits than sim- ple linear fits. For the motion of even the fastest de- tectable Haumean satellites over the timescale of our ob- servations, we find that quadratic fits to the x and y posi- tions are always sufficient. Note that the polynomial fits are included for convenience in describing the sky tracks; the actual positions of a putative satellite could be used, but the difference between the actual positions and the best-fit quadratic approximate was negligible. Including non-linear rates is often expected to greatly expand the number of dissimilar shift rates to the point of compu- tational impracticality, but we find that an appropriate criterion for similarity of shift rates easily permits the inclusion of quadratic rates. 3.1. Generation of Sky Tracks In a typical shift-and-stack search for KBOs, the puta- tive sky tracks are selected from a grid of the six degrees of freedom needed to describe an object in a Keplerian orbit (PK10, Bernstein et al. 2004). For the purposes of a KBO satellite, particularly that of a primary with other known satellites, it is convenient to instead sample the space of Keplerian orbital elements relative to the primary: semi-major axis (a), eccentricity (e), inclina- tion (i), longitude of the ascending node (Ω), argument of periapse (ω), and mean anamoly at epoch (M ). Sam- pling in this space allows for direct control over the types of orbits that are searched, making it straightforward to exclude unphysical motions. In our case, we also bene- fit from a well-known mass of the primary; if this is not known, a variety of plausible values could be sampled for the generation of sky tracks. For this search, a and e were randomly sampled from orbits with semi-major axes between 5310 and 368000 km and eccentricities less than 0.5, while the orbital angles i, Ω, ω and M were al- lowed to assume any value. All parameters were chosen from the sample space uniformly, with the exception of a, which was sampled on a log scale to increase the like- lihood of sampling an orbit in the regime of fast-moving satellites. The lower bound on a is a constraint imposed by our sensitivity of detection. This limit corresponds to 3.75 pixels (15 milliarcseconds) on the WFC3, at which dis- tance from the center of Haumea the subtraction noise is considerable enough to make reliable detections diffi- cult (see Figure 1). The upper bound on a is set much larger than the semi-major axes needed to shift images (as opposed to investigation of the unshifted stack). At a distance where the satellite's maximum velocity would cause it to travel less than one PSF FWHM over the course of the 15 hour observational period (here ∼27 m s−1), shift-and-stack is unnecessary, giving the upper limit of a ≃ 150, 000 in our search. This semi-major axis is ∼3 times the semi-major axis of Hi'iaka, whose motion in these frames is detectable, but .0.5 pixels. For the upper limit on a, we doubled this number to be conser- vative. Much of this orbital parameter space can be excluded on physical grounds, reducing the number of shift rates necessary to well-sample the space. Any putative orbit which crossed paths with the known satellites was re- jected, as was any orbit with periapsis less than 3000 km. These weak restrictions on orbital elements did not appreciably affect the selection of shift-and-stack param- eters and additional tests (described below) show that we are sensitive to objects on practically any orbit with semi-major axis &10000 km. 3.2. Non-linear Fitting and Shift Rate Similarity Having created a bank of physically plausible orbits, we then generate a set of shift rates with which to create composite images to search for satellites. Orbital pa- rameters were converted into sky coordinates relative to Haumea, right ascension (∆RA) and declination (∆Dec), for each image, as described in RB09. We assumed an in- stantaneous Keplerian orbit for the position of the satel- lite as this is an excellent approximation over the course of our observations. In our case, orbital acceleration was quite important, as we desired to search orbital periods down to ∼40 hours, of which the 15-hour observational arc is a sizable fraction. Therefore, the sky positions ∆RA and ∆Dec were fit with quadratic polynomials in time, which we found were sufficient to accurately de- scribe the non-linear motion in every case. In order to minimize the number of sky tracks, we elim- inated tracks which were similar to one another, as sug- gested by PK10. To determine if two tracks were similar, we focused on the final requirement that the shift rates localize the flux of satellite so that it can be identified in the stacked image. If the flux of a satellite traveling along the second orbit would be well-localized by the shift rates of the first, then there would be considerable overlap of the flux between images when shifted according to the rates of the first. This criterion can be quantified by cal- culating the overlap fraction between two shift-and-stack rates using the reasonable assumption that the WFC3 PSF is nearly Gaussian, with FWHM of .067 arcseconds (≈1.7 pixels). For a pair of shift rates, the overlap was defined for each image in the dataset as the integral of the product of two such Gaussians separated by the dif- ference in the two rates (∆RA and ∆Dec) at the time of that image. We call this the overlap between two orbits as it is calculated from the product of two overlapping PSFs, but it is distinct from the concept of the overlap in co-added images. If the median overlap (normalized to 1 for perfect coincidence) was greater than a pre-specified threshold, it was considered that a sufficient fraction of the flux of the proposed satellite would have been col- lected by the stack of an existing sky track, and the new track was rejected as unnecessary. The goal is to build up a bank of sky tracks known to be mutually distinct. After accepting the first track into the bank, each subsequent track was compared to the previ- ously selected tracks in the bank using the above overlap 5 criterion. We experimented with different overlap thresh- old criteria and found the overall results mostly insensi- tive to the specific value chosen. In general, we required a median overlap of less than 0.7 with each previously accepted shift-and-stack tracks to accept the proposed track as distinct enough to add to the bank. By drawing from a large set of orbits covering the de- sired search space, this method efficiently builds a bank of mutually distinct shift rates that are also the most relevant (PK10). However, unlike a grid search, random orbital draws can continue indefinitely. Thus, we also re- quire a "stopping criterion" to decide when the bank is large enough for practical use. To determine the upper limit on the number of necessary shift rates, we noted that the sample space saturates quickly; that is, the rate of acceptance drops off drastically after 10-15 shift rates are chosen. Consequently, the number of orbits rejected between successive accepted rates grows very quickly. Our criterion for a dense sampling was that the number of rejected rates between successive accepted rates was at least equal to the total number of rates rejected so far. Put another way, the selection was stopped when the acceptance of each new shift rate required doubling the number of sampled orbits, which typically occurred after testing hundreds of thousands of random orbits. This is an exponentially slow process, which suggests that once past this threshold, we have reached the limits of our rate selection method. Practically speaking, we found that this stopping criterion still generated enough shift rates to recover injected sources with a complete variety of orbits. Together with the above ranges for satellite orbital pa- rameters, this method yielded only ∼35 sufficiently dis- tinct orbits over which to search. Considering the size of the parameter space (linear and quadratic terms for shifts in both x and y directions), it might seem sur- prising that so small a set of potential orbits spans the space. However, a large portion of the space consists of short, almost entirely linear tracks, where the quadratic corrections are of limited importance. The only strongly quadratic orbits are very near to Haumea, which also have large linear rates. In other words, there are strong correlations between the allowed linear and quadratic coefficients of physically-plausible tracks. The result is a relatively small number of non-linear shift rates that efficiently cover the desired search space (PK10); these tracks are illustrated in Figure 2. Contrasted with the orbital element motivated sampling presented here, the number of shift rates in the case of a quadratic sky mo- tion grid search would have been much larger. 3.3. Creation of Composite Image With our bank of non-degenerate sky tracks, we can now perform the non-linear shift-and-stack procedure. Each track corresponds to a specific set of ∆RA and ∆Dec values of a putative satellite relative to Haumea. We used adxy, a routine from the IDL Astro Library which uses astrometric data from the image headers, to convert these sky coordinates into on-image pixel posi- tions, thus yielding the desired pixel shifts. The prepared images were shifted (including fractional pixel shifts implemented by IDL's fshift) and stacked using the pixel-by-pixel median of the images as de- scribed above. In preparation for the automated search, 6 Figure 2. The non-linear shift-and-stack rates. Arcs show the displacement of images (relative to position at the middle image) over the course of the 15 hour observation. Each arc represents a different shift rate or "sky track." Horizontal and vertical axes show differential right ascension (∆RA) and declination (∆Dec) in arcseconds and pixels (1 pixel = 0.04 arcseconds = 1475 km). The circle at bottom left has diameter of .067 arcseconds, the FWHM of WFC3's PSF. Following the method suggested by PK10, we gener- ate random non-linear shift rates out of Keplerian orbital elements. We reject as duplicates rates for which the overlapping PSFs would catch at least 70% of the flux if moving at the same rate as an ex- isting orbit (see §3.2). This method requires only ∼35 non-linear rates to cover the vast majority of parameter space. The "sky tracks" associated with these rates are mostly symmetric about the origin as seen above, with slight asymmetries arising from the projection of eccentric orbits into the skyplane, and the variation in orbital speed throughout an orbit. Almost all rates are sub- stantially quadratic, which shows the importance of the non-linear approach. As can be seen, the use of quadratic shift rates allowed us to probe the region near Haumea where satellites would exe- cute sizable fractions of an orbit during the 15-hour observation. Implantation of artificial sources on orbits randomly drawn from the same Keplerian elements showed an excellent recovery rate (see Figures 3 and 4). many images were investigated in detail by eye. 3.4. Sensitivity To test the sensitivity of this method, artificial sources were implanted into the images with a range of ran- dom brightnesses. Their positions and rates of motion were determined by orbits randomly drawn from the same space mentioned above (but without restriction of non-crossing orbits with the known moons). The im- plants were generated by scaling from the actual PSF of Haumea (when brightest). This source was implanted into the images at the pixel positions corresponding to the randomly-chosen orbit. A subimage of 200 x 200 pixels was used for the search: the outer reaches of this subimage have objects that are practically not moving (see §3.1) and any object beyond this region would have the same detectability threshold as stationary objects. This was done for a large number of orbits, with a new set of images created for each. Stacks were generated in the exact same manner as the real images, with the same ∼35 shift rates making new median stacks for each new set of images. These stacks were inspected using the same automated search routine (IDL's find). To distin- gush detections of implanted sources from the detection of the three known bodies, we examined the output of the search for the sets of stacks with no sources implanted. All detections here were due to known bodies, and the positions were used to establish a mask with three re- gions, one for each known body, to reject detections that were not due to implanted sources. In this way, detec- tions could be automatically classified as a recovery of an implant or as a false positive due to the known bodies. These automated classifications were extensively verified with an investigation that included searching by eye and found to be very robust. Due to the application of a threshhold SNR by the find routine, objects in the vicinity of Haumea, while still far enough and bright enought to be seen by eye, may be rejected by the routine itself (not our masks). The presence of the primary nearby leads to an artificially high computed background noise level, which reduces the computed SNR significantly, causing the object to appear below threshhold. Any stacks with sources at risk of being left undetected due to this effect were searched by eye by multiple coauthors, and any that were detected in this processes were considered to be recovered for the purposes of our results, shown below. The success rates of finding implanted objects places constraints on additional satellites of Haumea: any re- covered implanted source represents a satellite that we can say with reasonable certainty is not present in the Haumea system. 4. RESULTS The implantation and successful recovery of faint mov- ing sources clearly indicated the effectiveness of our non- linear shift-and-stack method. Nevertheless, we did not detect any additional satellites around Haumea and no candidate satellites were found that were worthy of ad- ditional investigation. A careful characterization of this null result is able to place strong limits on the brightness and separation of undiscovered Haumean satellites. These limits are sum- marized by Figures 3 and 4, which show the results of our search for each implanted source. The source is either re- covered, rejected (for being too close to one of the three known objects, usually Haumea), or "missed" because was too faint to be detected, or because it fell off the 200 x 200 subimage that was searched. Note that the "re- jected" category is primarily composed of objects that we not clearly detected by the automated routine, but were detected in a blind search by eye by multiple coau- thors; these consistent entirely of objects that are .0.2" (5 pixels) from Haumea. Figure 3 plots semi-major axis against brightness of the sources as a fraction of that of Haumea, while Figure 4 shows the projected distance (in arcseconds) of the moving sources versus the brightness. Assuming the same albedo (p ≃ 0.7) as Haumea, the rel- ative brightness corresponds to the radius of a spherical satellite, which is also shown. 5. DISCUSSION The constraints on undiscovered Haumean satellites can be divided into three categories based on orbital semi-major axis: close-in satellites (a . 10000 km), in- termediate satellites (10000 km . a . 350000 km), and distant satellites (a & 350000 km). 5.1. Limits on Close-in Satellites 7 Figure 3. (Color online) Results from the sensitivity survey. The figure shows implanted sources that were either recovered (blue stars), not recovered (red crosses), or recovered but rejected due to confusion with existing sources (green squares). The horizon- tal axis is the semi-major axis of implanted objects in thousands of kilometers. The left-hand vertical axis is brightness relative to Haumea (when brightest). The right-hand vertical axis is radius of a spherical satellite assuming an albedo (∼0.7) similar to Haumea. Diamonds represent know satellites Namaka and Hi'iaka; the ver- tical lines are guides to the eye at their respective semi-major axes. Purple triangles represent the moons of Pluto -- Nix, Hydra and Kerberos -- according to brightness relative to the primary. (The smallest moon Styx, with brightness approximately 6 × 10−6 that of Pluto, is below the range of brightness represented on this fig- ure.) Because of differences in geocentric distance and albedo, the approximate radius does not directly apply to these three points. Figure 4 is similar but shows distance in projected separation in- stead of semi-major axis. Bounds on brightness and semi-major axis of were chosen as described in §3.1. The unrecovered implan- tations at semi-major axis & 200 × 103 km are not found because their distance from Haumea often places them outside the subim- ages searched. This figure shows that satellites with radii as low as ∼8 km would be detectable in much of the space searched, and that our lower detection limit on semimajor axis is limited by the properties of the dataset, not by the sensitivity of the non-linear shift-and-stack technique. Nix and Hydra-like objects would be de- tected around Haumea, while Styx and Kerberos-like objects would still be too faint, mostly due to Haumea's further distance (50 AU compared to Pluto's 30 AU). At a semi-major axis of ∼10000 km, the maximum sep- aration of a satellite from Haumea would be 6.9 WFC3 pixels. Within 7 pixels (. 4 PSF FWHM) of Haumea, it is very difficult to recover objects due to imperfect subtraction of Haumea's PSF. It is possible that an em- pirical PSF subtraction would perform better for recov- ering very close-in satellites, but we do not consider such an approach here. As can be seen in Figure 4, there is the expected anti-correlation between the brightness of an object that can be recovered and the separation from Haumea: close in, only brighter objects can be found. However, there are dynamical reasons to expect that this region is nearly devoid of satellites. Due to Haumea's highly triaxial shape, the orbital region near Haumea is strongly perturbed and long-term stable orbits are dif- ficult to maintain. According to Scheeres (1994), pe- riods less than about 10 times the spin period are un- likely to be stable due to primary-spin-satellite-orbit res- onances. In Haumea's case, this is exacerbated by the additional effects of tidal evolution and other dynam- ically excited satellites (Canup et al. 1999; ´Cuk et al. 2013; Cheng et al. 2014b). An orbital period that is 10 times the spin period corresponds to a semi-major axis of about 5000 km (about 5 times the long-axis radius of Haumea). While about twice as distant as the Roche ra- Figure 4. (Color online) Results from the sensitivity survey. The lower horizontal axis is the sky-planet projected separation from Haumea in arcseconds, while the upper axis gives approximate pro- jected distance from Haumea in thousands of kilometers. The vertical axes are brightness and radius of implanted sources, as described in the caption to Figure 3. Symbols connected by hor- izontal lines show the maximum and minimum apparent distance from Haumea of the implanted object during the 15-hour "obser- vation." As in Figure 3, implanted sources were either recovered (blue stars), not recovered (red crosses), or recovered but rejected due to confusion with existing sources (green squares). Note that implantations at separation greater than 4 arcseconds are unre- covered because they fall outside the region of the ∼4" subimages that were searched (see §5.3). The vertical dashed line is a guide to the eye for this rough cutoff. No sources were implanted at separations larger than 10 arcseconds, corresponding to the upper limit on semi-major axis shown in Figure 3. Diamonds represent Hi'iaka and Namaka as they appear in the observaion; Namaka's separation is only given for the first four orbits where its pres- ence is measured reliably enough for precise astrometry; as these observations were designed to catch Namaka in a mutual event, its projected separation would approach very low values if all ten orbits were included. dius, for long-term dynamical stability, we consider this the inner limit. Even if satellites were originally found in such short orbits, it is possible that long-term tidal evolution would have moved them to a more detectable distance. A de- tailed analysis by ´Cuk et al. (2013) calls into question the idea first proposed that the satellites tidally evolved out- wards from orbits near the Roche lobe. While extensive tidal evolution might not have taken place, it is worth noting that scaling the tidal evolution from the proper- ties of the other satellites (Brown et al. 2005), indicates that even for the smallest satellites we could have de- tected (which evolve the shortest distance due to tides), tidal evolution would have placed them near or beyond the ∼5000 km detection threshold. There remains a range of semi-major axes from 5000- 10000 km that could potentially harbor very small un- detected satellites which would be somewhat protected from dynamical and tidal instability. By lying well within Haumea's PSF, these satellites also generally evade de- tection. Furthermore, some satellites would not have been detected if they had an orbital phase placing them at undetectably small distances (although this is miti- gated somewhat by observations at a variety of times). Overall, it is difficult to hide stable inner (a .10000 km) Haumean satellites with radii &30 km. 5.2. Intermediate Satellites At semi-major axes between about 10000 and 350000 km lies the region of satellites near where the other two moons are detected (at semi-major axes of 25600 km for 8 Namaka and 49900 km for Hi'iaka, RB09). At this dis- tance, contamination from Haumea is negligible and the main limitation to detecting satellites is insufficient SNR or falling beyond the edge of the image. By using the non-linear shift-and-stack we maximize the search depth, particularly closer to Haumea. The search depth can be reported as relative brightness (in magnitudes and flux) and as the radius of a spher- ical satellite assuming the same albedo as Haumea. As is usual for such deep searches, the recovery rate is a function of magnitude (Figure 3,4). We reliably detect satellites at -9.2 magnitudes (0.0002 relative brightness, radius of 10 km), our recovery rate is roughly 50% at -10 magnitudes (0.0001 relative brightness, radius of 8 km), and our best case recovery is at -10.4 magnitudes (0.00007 relative brightness, radius of 6 km). Follow- ing typical practice, we summarize the recovery depth using the 50% recovery rate. Note that it is possible that the albedo of the satellites is even higher; using Haumea family member 2002 TX300's measured albedo of 0.9 (Elliot et al. 2010) instead of Haumea's presumed 0.7 albedo (Lockwood et al. 2014) would imply a radius detection threshold of only ∼7 km (or ∼5 km in the best case). While close approaches to Hi'iaka and Namaka as pro- jected on the sky would result in a missed detection for faint objects, this is generally unlikely (even for orbits coplanar with the known satellites which are near edge- on, RB09). Close approaches to Hi'iaka and Namaka are negligibly unlikely to happen at more than one epoch7, thus any missed detection would be mitigated for moder- ately bright objects by the non-detection of satellites in other datasets. Thus, we expect that this region of the Haumean system does not contain undiscovered satellites larger than ∼8 km in radius. Our results compare favorably with the current state of knowledge regarding the small satellites of Pluto. From the New Horizons flyby, we now have detailed knowledge of the albedoes (about 0.5) and sizes of the small satel- lites: ∼10 km for Styx and Kerberos and ∼40 km for Nix and Hydra (Stern et al. 2015). As Figure 3 shows, we predict that a satellite of apparent magnitude relative to Haumea similar to that of Hydra or Nix around Pluto (-8.7 and -9.2 magnitudes respectively) would fall above our detection limit. With the higher expected albedo (0.7) of Haumean satellites, we would have detected ob- jects as large as Styx and Kerberos. We conclude that Haumea very likely does not contain small satellites sim- ilar to Pluto's. 5.3. Distant Satellites Satellites with semi-major axes beyond 350000 km may not have been detected in the Program 12243 WFC3 data due to the small field-of-view employed for the subarray observations. Other HST and Keck observations that were not as deep covered a larger area and were also searched for satellites. We estimate that satellites larger than about 40 km in radius (again assuming an albedo similar to Haumea's) would have been detected even sev- eral tens of arcseconds away by, e.g., the WFPC2 obser- 7 Unlike irregular satellites of the giant planets, long-term tidal stability precludes Hi'iaka or Namaka from being binaries them- selves. vations (with a field of view of 162"). Because the motion of sattelites in this region is negligible over the relevant timescales, the shift-and-stack method is not necessary. Using half the size of Haumea's Hill sphere at per- ihelion as an estimate of the full region of stable satel- lites (Shen & Tremaine 2008), the semi-major axis of the most distant stable satellites would be about 4.6 × 106 km or 124". About half of this volume has been covered down to 40 km in radius. For comparison, the Program 12243 deep observations covered separations up to about 10" around Haumea or 350000 km. Thus, this limit on very small intermediate- range satellites corresponds to about 0.5% of the stable region radius. 6. CONCLUSIONS By efficient application of the PK10 method for non- linear shift-and-stack and recovery of known implanted sources, we have strongly limited the possibility of un- detected satellites in orbit around Haumea. As Figure 4 shows, we detect no satellites larger than ∼8 km in radius with separations between 10000 and 350000 km. This same region around Pluto contains Charon and 4 small satellites which, by size, would all have been de- tected in this search. Nearer to Haumea, diffraction limits make distinguish- ing small satellites difficult, but there are dynamical rea- sons to expect that this region is mostly unpopulated. Further from Haumea, other observations would have detected satellites larger than ∼40 km in radius within much of the entire region of possible stable satellites. Sig- nificant improvement in the detection limits on smaller satellites would require extensive observations that are unlikely in the foreseeable future until, perhaps, deep observations with the James Webb Space Telescope. Though Pluto contains multiple small moons and some formation theories (e.g., Lithwick & Wu 2008) predict them in the Haumea system, we find no additional Haumean moons. Considering upper limits from other studies (summarized in Table 1), Nix/Hydra analogues would have been discovered if present around Makemake and they would be near the detection threshold around Eris. As the properties of the dwarf planet satellite sys- tems differ significantly, it was not anticipated that Pluto's small satellites would necessarily find counter- parts around Haumea, though it seems that Makemake may have a satellite of similar size (Parker et al. 2016). Our null result affirms that, for the time being, Pluto is the only known KBO with a retinue of small satel- lites, though such could have been detected or nearly de- tected around all four dwarf planets. This implies that the satellite systems may result from somewhat differ- ent formation pathways, although all the dwarf planet satellites are probably connected with a collisional for- mation. Pluto's small satellite system may be connected with Charon since, from a dynamical perspective, the other dwarf planet satellites are more like small moons compared to the near-equal-sized Pluto-Charon binary. We demonstrate that the non-linear shift-and-stack is a valuable tool for satellite searches. Utilizing the ap- plication techniques developed herein, this method can sufficiently capture the nonlinearity of the orbits of fast- moving satellites close to the primary. We have ap- Summary of Estimated Properties of Dwarf Planet Satellites Table 1 Object Satellite Relative Brightness Hsat a Vsat (magnitudes) a Radiusb (km) ac Ref (103 km) 9 Haumea Haumea Haumea Haumea Haumea Pluto Pluto Pluto Pluto Pluto Eris Eris Eris Makemake Makemake Hi'iaka Namaka "close" upper limit "intermediate" upper limit "distant" upper limit Charon Hydra Nix Kerberos Styx Dysnomia "close" upper-limit "distant" upper-limit S/2015 (136472) 1 upper-limit −3.3 −4.6 −6.7 −10.0 −6.2 −2.6 −8.7 −9.2 −12 −13 −6.7 −5.8 −8.2 −7.8 −10 3.4 4.7 6.8 10.1 6.3 1.9 8.0 8.5 11 12 5.5 4.6 7.0 7.4 9.6 20.5 21.8 23.9 27.6 23.4 16.6 22.7 23.2 26 27 25.4 24.5 26.9 24.7 26.9 200 150 30 8 40 350 41 35 12 11 60 80 30 25 8 50 26 .10 10-350 &350 20 64 49 59 42 37 &18 &37 ∼100 &30 References. -- (1) RB09 (Ragozzine & Brown 2009) (2) §4, this paper Showalter et al. (2011) Parker et al. (2016) (5) Showalter et al. (2012) (6) Brown & Schaller (2007) (3) Weaver et al. (2006) (7) Brown (2008) 1 1 2 2 2 3 3 3 4 5 6 6 6 8 7,8 (4) (8) Note. -- Magnitudes and semi-major axes of bodies in KBO systems. The relative magnitude of the faintest detectable bodies in our search is -10, comparable to that of Hydra and Nix. For Eris and Makemake, values are more approximate and/or interpolated from published estimates. We do not list the large number of KBO binaries (e.g. Noll et al. 2008) or KBO triple 1999TC36 (Benecchi et al. 2010) since the formation of these systems appears to be distinct from processes associated with dwarf planets. In particular, these binaries tend to be nearly equal brightness without known small additional companions. a Approximate absolute magnitude (H) or approximate apparent magnitude in a typical optical filter (V ) of the satellite. These are calculated combining the relative magnitude with the absolute and typical apparent magnitudes of the KBOs from JPL Horizons. These are meant mostly for illustration purposes and generally have significant uncertainties of .1 magnitude. b Radius estimate in kilometers, listed for illustration purposes only. Quoted radii for the highly ellipsoidal small satellites of Pluto are volumetric means (S. Porter, pers. comm.). Note that these have albedoes of 0.5, somewhat less than assumed for Haumea's moons. For simplicity and ease of inter-comparison, observed moons of Eris and Makemake are given an estimated albedo of 0.7 like the Haumea moons. The actual albedo and size of these moons is not well constrained. c Approximate semi-major axis in units of thousands of kilometers. For upper-limits, this is the approximate range of semi-major axes where this limit applies. The discovery of S/2015 (136472) 1 by Parker et al. (2016) within the magnitude and distance "upper-limit" quoted by Brown (2008) is easily attributed to the difficulty of detecting moons with small semi-major axes and/or edge-on orbits in single-epoch observations when the actual on-the-sky separation is often small enough to render the moon indistinguishable from the primary (Parker et al. 2016). The upper-limits reported here should be understood with that caveat. plied this technique to the regime of searching for sub- threshold satellites around Haumea, but it could also be used for other long-observation datasets (PK10). Be- sides discovery of new moons, it has promise for im- proving astrometric parameters for known faint moving satellites (e.g., precovery observations of Styx and Ker- beros). The tractability of the non-linear shift-and-stack also promotes the possibility of applying this to the gen- eral search for KBOs, as originally proposed by PK10. Other applications for improving sensitivity are also pos- sible, e.g. searching for moving exoplanets in direct imag- ing campaigns (Males et al. 2013). To facilitate further analyses, all data and source codes used in this project are available upon request. The sensitivity and tractability of the method pre- sented in this work suggests that, when appropriate, it should be applied to other satellite searches in the so- lar system. The non-detection of small satellites around Haumea increases our understanding of this intriguing object and contributes to the our understanding of the formation and evolution of multiple KBO systems. We thank Alex Parker, Danielle Hastings, and the anonymous referee for discussions and suggestions that improved the manuscript. DR acknowledges the sup- port of a Harvard Institute for Theory and Computation Fellowship. This work is based on NASA/ESA Hubble Space Telescope Program 12243. Support was provided by NASA through grants HST-GO-12243 from the Space Telescope Science Institute (STScI), which is operated by the Association of Universities for Research in Astron- omy, Inc., under NASA contract NAS 5-26555. REFERENCES Benecchi, S. D., Noll, K. S., Grundy, W. M., & Levison, H. F. 2010, Icarus, 207, 978 Bernstein, G. M., Trilling, D. E., Allen, R. L., Brown, M. E., Holman, M., & Malhotra, R. 2004, AJ, 128, 1364 Brown, M. E. 2008, The Largest Kuiper Belt Objects (The Solar System Beyond Neptune), 335 -- 344 Brown, M. E., Barkume, K. M., Ragozzine, D., & Schaller, E. L. 2007, Nature, 446, 294 Brown, M. E., Bouchez, A. H., Rabinowitz, D., Sari, R., Trujillo, C. A., van Dam, M., Campbell, R., Chin, J., Hartman, S., Johansson, E., Lafon, R., Le Mignant, D., Stomski, P., Summers, D., & Wizinowich, P. 2005, ApJL, 632, L45 Brown, M. E., & Schaller, E. L. 2007, Science, 316, 1585 Brown, M. E., Schaller, E. L., & Fraser, W. C. 2012, AJ, 143, 146 10 Brown, M. E., van Dam, M. A., Bouchez, A. H., Le Mignant, D., Campbell, R. D., Chin, J. C. Y., Conrad, A., Hartman, S. K., Johansson, E. M., Lafon, R. E., Rabinowitz, D. L., Stomski, Jr., P. J., Summers, D. M., Trujillo, C. A., & Wizinowich, P. L. 2006, ApJL, 639, L43 Canup, R. M. 2011, AJ, 141, 35 Canup, R. M., Levison, H. F., & Stewart, G. R. 1999, AJ, 117, 603 Carry, B., Snodgrass, C., Lacerda, P., Hainaut, O., & Dumas, C. 2012, A&A, 544, A137 Cheng, W. H., Lee, M. H., & Peale, S. J. 2014a, Icarus, 233, 242 Cheng, W. H., Peale, S. J., & Lee, M. H. 2014b, ArXiv e-prints ´Cuk, M., Ragozzine, D., & Nesvorn´y, D. 2013, AJ, 146, 89 Elliot, J. L., Person, M. J., Zuluaga, C. A., Bosh, A. S., Adams, E. R., Brothers, T. C., Gulbis, A. A. S., Levine, S. E., Lockhart, M., Zangari, A. M., Babcock, B. A., Dupr´e, K., Pasachoff, J. M., Souza, S. P., Rosing, W., Secrest, N., Bright, L., Dunham, E. W., Sheppard, S. S., Kakkala, M., Tilleman, T., Berger, B., Briggs, J. W., Jacobson, G., Valleli, P., Volz, B., Rapoport, S., Hart, R., Brucker, M., Michel, R., Mattingly, A., Zambrano-Marin, L., Meyer, A. W., Wolf, J., Ryan, E. V., Ryan, W. H., Morzinski, K., Grigsby, B., Brimacombe, J., Ragozzine, D., Montano, H. G., & Gilmore, A. 2010, Nature, 465, 897 Fraser, W. C., & Brown, M. E. 2012, ApJ, 749, 33 Holman, M. J., Kavelaars, J. J., Grav, T., Gladman, B. J., Fraser, W. C., Milisavljevic, D., Nicholson, P. D., Burns, J. A., Carruba, V., Petit, J.-M., Rousselot, P., Mousis, O., Marsden, B. G., & Jacobson, R. A. 2004, Nature, 430, 865 Holman, M. J., & Wiegert, P. A. 1999, AJ, 117, 621 Kavelaars, J. J., Holman, M. J., Grav, T., Milisavljevic, D., Fraser, W., Gladman, B. J., Petit, J.-M., Rousselot, P., Mousis, O., & Nicholson, P. D. 2004, Icarus, 169, 474 Kenyon, S. J., & Bromley, B. C. 2014, AJ, 147, 8 Lacerda, P. 2009, AJ, 137, 3404 Lacerda, P., Jewitt, D., & Peixinho, N. 2008, AJ, 135, 1749 Leinhardt, Z. M., Marcus, R. A., & Stewart, S. T. 2010, ApJ, 714, 1789 Levison, H. F., Morbidelli, A., Vokrouhlick´y, D., & Bottke, W. F. 2008, AJ, 136, 1079 Lithwick, Y., & Wu, Y. 2008, ArXiv e-prints Lockwood, A. C., Brown, M. E., & Stansberry, J. 2014, Earth Moon and Planets, 111, 127 Lykawka, P. S., Horner, J., Mukai, T., & Nakamura, A. M. 2012, MNRAS, 421, 1331 Males, J. R., Skemer, A. J., & Close, L. M. 2013, ApJ, 771, 10 Marcus, R. A., Ragozzine, D., Murray-Clay, R. A., & Holman, M. J. 2011, ApJ, 733, 40 Markwardt, C. B. 2009, in Astronomical Society of the Pacific Conference Series, Vol. 411, Astronomical Data Analysis Software and Systems XVIII, ed. D. A. Bohlender, D. Durand, & P. Dowler, 251 Noll, K. S., Grundy, W. M., Chiang, E. I., Margot, J.-L., & Kern, S. D. 2008, Binaries in the Kuiper Belt (The Solar System Beyond Neptune), 345 -- 363 Ortiz, J. L., Thirouin, A., Campo Bagatin, A., Duffard, R., Licandro, J., Richardson, D. C., Santos-Sanz, P., Morales, N., & Benavidez, P. G. 2012, MNRAS, 419, 2315 Parker, A. H., Buie, M. W., Grundy, W. M., & Noll, K. S. 2016, ArXiv e-prints Parker, A. H., & Kavelaars, J. J. 2010, PASP, 122, 549 Pires dos Santos, P. M., Morbidelli, A., & Nesvorn´y, D. 2012, Celestial Mechanics and Dynamical Astronomy, 114, 341 Porter, S. B., & Stern, S. A. 2015, ArXiv e-prints Rabinowitz, D. L., Barkume, K., Brown, M. E., Roe, H., Schwartz, M., Tourtellotte, S., & Trujillo, C. 2006, ApJ, 639, 1238 Ragozzine, D., & Brown, M. E. 2007, AJ, 134, 2160 -- . 2009, AJ, 137, 4766 Schaller, E. L., & Brown, M. E. 2008, ApJL, 684, L107 Scheeres, D. J. 1994, Icarus, 110, 225 Schlichting, H. E., & Sari, R. 2009, ApJ, 700, 1242 Shen, Y., & Tremaine, S. 2008, AJ, 136, 2453 Showalter, M. R., de Pater, I., French, R. S., & Lissauer, J. J. 2013, in AAS/Division for Planetary Sciences Meeting Abstracts, Vol. 45, AAS/Division for Planetary Sciences Meeting Abstracts, 206.01 Showalter, M. R., & Hamilton, D. P. 2015, Nature, 522, 45 Showalter, M. R., Hamilton, D. P., Stern, S. A., Weaver, H. A., Steffl, A. J., & Young, L. A. 2011, IAU Circ., 9221, 1 Showalter, M. R., Weaver, H. A., Stern, S. A., Steffl, A. J., Buie, M. W., Merline, W. J., Mutchler, M. J., Soummer, R., & Throop, H. B. 2012, IAU Circ., 9253, 1 Snodgrass, C., Carry, B., Dumas, C., & Hainaut, O. 2010, A&A, 511, A72 Stern, S. A., Bagenal, F., Ennico, K., Gladstone, G. R., Grundy, W. M., McKinnon, W. B., Moore, J. M., Olkin, C. B., Spencer, J. R., Weaver, H. A., Young, L. A., Andert, T., Andrews, J., Banks, M., Bauer, B., Bauman, J., Barnouin, O. S., Bedini, P., Beisser, K., Beyer, R. A., Bhaskaran, S., Binzel, R. P., Birath, E., Bird, M., Bogan, D. J., Bowman, A., Bray, V. J., Brozovic, M., Bryan, C., Buckley, M. R., Buie, M. W., Buratti, B. J., Bushman, S. S., Calloway, A., Carcich, B., Cheng, A. F., Conard, S., Conrad, C. A., Cook, J. C., Cruikshank, D. P., Custodio, O. S., Dalle Ore, C. M., Deboy, C., Dischner, Z. J. B., Dumont, P., Earle, A. M., Elliott, H. A., Ercol, J., Ernst, C. M., Finley, T., Flanigan, S. H., Fountain, G., Freeze, M. J., Greathouse, T., Green, J. L., Guo, Y., Hahn, M., Hamilton, D. P., Hamilton, S. A., Hanley, J., Harch, A., Hart, H. M., Hersman, C. B., Hill, A., Hill, M. E., Hinson, D. P., Holdridge, M. E., Horanyi, M., Howard, A. D., Howett, C. J. A., Jackman, C., Jacobson, R. A., Jennings, D. E., Kammer, J. A., Kang, H. K., Kaufmann, D. E., Kollmann, P., Krimigis, S. M., Kusnierkiewicz, D., Lauer, T. R., Lee, J. E., Lindstrom, K. L., Linscott, I. R., Lisse, C. M., Lunsford, A. W., Mallder, V. A., Martin, N., McComas, D. J., McNutt, R. L., Mehoke, D., Mehoke, T., Melin, E. D., Mutchler, M., Nelson, D., Nimmo, F., Nunez, J. I., Ocampo, A., Owen, W. M., Paetzold, M., Page, B., Parker, A. H., Parker, J. W., Pelletier, F., Peterson, J., Pinkine, N., Piquette, M., Porter, S. B., Protopapa, S., Redfern, J., Reitsema, H. J., Reuter, D. C., Roberts, J. H., Robbins, S. J., Rogers, G., Rose, D., Runyon, K., Retherford, K. D., Ryschkewitsch, M. G., Schenk, P., Schindhelm, E., Sepan, B., Showalter, M. R., Singer, K. N., Soluri, M., Stanbridge, D., Steffl, A. J., Strobel, D. F., Stryk, T., Summers, M. E., Szalay, J. R., Tapley, M., Taylor, A., Taylor, H., Throop, H. B., Tsang, C. C. C., Tyler, G. L., Umurhan, O. M., Verbiscer, A. J., Versteeg, M. H., Vincent, M., Webbert, R., Weidner, S., Weigle, G. E., White, O. L., Whittenburg, K., Williams, B. G., Williams, K., Williams, S., Woods, W. W., Zangari, A. M., & Zirnstein, E. 2015, Science, 350, aad1815 Volk, K., & Malhotra, R. 2012, Icarus, 221, 106 Ward, W. R., & Canup, R. M. 2006, Science, 313, 1107 Weaver, H. A., Stern, S. A., Mutchler, M. J., Steffl, A. J., Buie, M. W., Merline, W. J., Spencer, J. R., Young, E. F., & Young, L. A. 2006, Nature, 439, 943 Youdin, A. N., Kratter, K. M., & Kenyon, S. J. 2012, ApJ, 755, 17
1709.02412	1	1709	2017-09-07T19:11:44	A secular increase in continental crust nitrogen during the Precambrian	[ "astro-ph.EP" ]	Recent work indicates the presence of substantial geologic nitrogen reservoirs in the mantle and continental crust. Importantly, this geologic nitrogen has exchanged between the atmosphere and the solid Earth over time. Changes in atmospheric nitrogen (i.e. atmospheric mass) have direct effects on climate and biological productivity. It is difficult to constrain, however, the evolution of the major nitrogen reservoirs through time. Here we show a secular increase in continental crust nitrogen through Earth history recorded in glacial tills (2.9 Ga to modern), which act as a proxy for average upper continental crust composition. Archean and earliest Palaeoproterozoic tills contain 66 $\pm$ 100 ppm nitrogen, whereas Neoproterozoic and Phanerozoic tills contain 290 $\pm$ 165 ppm nitrogen, whilst the isotopic composition has remained constant at ~4$\permil$. Nitrogen has accumulated in the continental crust through time, likely sequestered from the atmosphere via biological fixation. Our findings support dynamic, non-steady state behaviour of nitrogen through time, and are consistent with net transfer of atmospheric N to geologic reservoirs over time.	astro-ph.EP	astro-ph	A secular increase in continental crust nitrogen during the Precambrian Benjamin W Johnson and Colin Goldblatt Recent work indicates the presence of substantial geologic nitrogen reservoirs in the mantle and con- tinental crust. Importantly, this geologic nitrogen has exchanged between the atmosphere and the solid Earth over time. Changes in atmospheric nitrogen (i.e., atmospheric mass) have direct effects on climate and biological productivity. It is difficult to constrain, however, the evolution of the ma- jor nitrogen reservoirs through time. Here we show a secular increase in continental crust nitrogen through Earth history recorded in glacial tills (2.9 Ga to modern), which act as a proxy for average upper continental crust composition. Archean and earliest Palaeoproterozoic tills contain 66±100 ppm nitrogen, whereas Neoproterozoic and Phanerozoic tills contain 290±165 ppm nitrogen, whilst the isotopic composition has remained constant at ∼4(cid:104). Nitrogen has accumulated in the continen- tal crust through time, likely sequestered from the atmosphere via biological fixation. Our findings support dynamic, non-steady state behaviour of nitrogen through time, and are consistent with net transfer of atmospheric N to geologic reservoirs over time. Introduction The evolution of the Earth System N cycle and the distribution of N in the Earth over the planet's history are not well constrained (Zerkle and Mikhail, 2017). Nitrogen moves between different reservoirs in the Earth system including the atmosphere, biosphere, and geosphere (Marty, 1995; Boyd, 2001; Busigny et al., 2003, 2011). Changes in the distribution of N among the major reservoirs of the Earth (mantle, crust, and atmosphere), have direct effects on planetary habitability. Biologic productivity based on N- fixing can be limited under very low N2 partial pressures (Klingler et al., 1989), and the amount and speciation of N in the atmosphere affect temperature through direct or indirect greenhouse warming (Goldblatt et al., 2009; Byrne and Goldblatt, 2015; Wordsworth and Pierrehumbert, 2013). Higher-N2 atmospheres can enhance the effectiveness of greenhouse gasses (Goldblatt et al., 2009; Wordsworth and Pierrehumbert, 2013), potentially providing a solution to the Faint Young Sun Para- dox (Sagan and Mullen, 1972; Fuelner, 2012). Specifically, pressure-broadening (Goldblatt et al., 2009) 1 of CO2 by an atmosphere with 2-3 fold more N2 can provide warming consistent with constraints on at- mospheric CO2 content in the Archean (Sheldon, 2006). It is difficult to assess this, and other hypotheses of changing atmospheric mass (Som et al., 2012, 2016; Barry and Hilton, 2016), through direct mea- surements of palaeoatmospheric conditions. Another approach is to constrain the history of geologic N reservoirs. One such reservoir is the continental crust. Current estimates for the amount of N in the modern con- tinental crust range from 0.25 present atmospheric N mass (PAN, or 4×1018 kg N) (Rudnick and Gao, 2014) to 0.5 PAN (Goldblatt et al., 2009; Johnson and Goldblatt, 2015). These estimates rely on measure- ments of individual rock types, which are then weighted by their proportion in the crust. For comparison, estimates of N in the Earth's interior range from 1 to 7 PAN in the Bulk Silicate Earth and 50 PAN in the core (Johnson and Goldblatt, 2015, and references therein). Modern subducted N is estimated to be 5×10−10 PAN per year (Johnson and Goldblatt, 2015) with non-arc outgassing of 1.75×10−11 PAN per year (Cartigny and Marty, 2013). The estimates of crustal N content may be biased, though, due to the effects of differential chemical weathering and alteration. In addition, these approaches offer no temporal resolution. As an alternative approach, we present measurements of glacial tills through time as a proxy for the upper continental crust. Large glaciers and ice sheets erode a wide variety of rock types, and resulting glacial till will represent an average composition of the crust over which they erode. Thus, integration of many samples of glacial till can act as a proxy for average upper continental crust composition. This approach was first utilized by Goldschmidt (1933), but has since been used to estimate the upper continental crust composition of both Phanerozoic, juvenile crust (Canil and Lacourse, 2011) as well as the composition of the crust through time (Gaschnig et al., 2016). Physical weathering and erosion by a glacier should not impart any isotopic fractionation on the samples. In addition, while weathering can produce locally distinct δ15N values (Boyd, 2001), it is expected that large glaciers will represent an average composition, which will integrate local variation. While biologic N cycling (Gruber and Galloway, 2008) has been a topic of research for well over a hundred years (Breneman, 1889), the geologic N cycle and exchange of N between the atmosphere and solid Earth have received far less attention. Some modelling efforts suggested near steady-state N concentrations in the crust, mantle, and atmosphere over at least the Phanerozoic (Berner, 2006), and possibly for most of Earth history (Zhang and Zindler, 1993; Tolstikhin and Marty, 1998). In contrast, geochemistry (Mitchell et al., 2010; Busigny et al., 2011; Barry and Hilton, 2016), other models (Hart, 1978; Stueken et al., 2016), and physical proxies (Som et al., 2012; Kavanagh and Goldblatt, 2015; Som et al., 2016) directly contradict the steady-state hypothesis. The later proxies are consistent with movement of N between different reservoirs of the Earth and significant changes in the mass of the atmosphere over time. Additional thermodynamic calculations argue that the evolution of mantle redox 2 and Eh-pH state at subduction zones directly affects N2 outgassing, and therefore the distribution of N in the Earth through time (Mikhail and Sverjensky, 2014). Either the distribution of N among the main reservoirs of the Earth (atmosphere, mantle, continental crust) has been in steady-state over Earth history or it has been more dynamic. A difficulty in assessing the validity of steady-state and dynamic interpretations of N distribution over Earth history is reconstructing N concentrations in geologic reservoirs in the past. The analysis of glacial tills presented herein suggests an increase in continental N through time, providing a temporal constraint on one of the three major N reservoirs of the Earth system. Nitrogen in glacial tills We analyzed a series of tills from Gaschnig et al. (2016) for N concentration and N isotopes. These till samples consisted of predominantly fine-grained matrix material, and come from formations as old as 2.9 billion years old to formations as young as 0.3 billion years old. We have also included a younger till, Till-4, which is a standard provided by the Geological Survey of Canada. Nitrogen concentrations are low in glacial tills during the Archean and earliest Palaeoproterozoic, moder- ate and variable during the Neoproterozoic, and moderate-high and less variable during the Phanerozoic (Fig. 1, Table 1, Supplemental Information). We define "low" as less than average granite, 54 ppm (Johnson and Goldblatt, 2015), "high" as approaching average upper crust sedimentary rocks, > 400 ppm, and "moderate" as in between. Performing Student's t-test (Student, 1908) indicates that both the mean, shown with one standard deviation, Neoproterozoic (250±180 ppm) and Phanerozoic (380±50 ppm) concentrations are significantly different from the mean of the Archean and earliest Palaeoprotero- zoic (66±100 ppm) samples. There appears to be a secular increase in N content in the continental crust through time. Table 1: Proportion of till samples in each age group that have high (> 400 ppm), low (< 54 ppm), and moderate (in between) N. Age Archean Palaeoproterozoic Neoproterozoic Phanerozoic % low % moderate % high 100 75 10 0 0 0 30 50 0 25 60 50 In contrast, mean (plus one standard deviation) δ15N values remain constant within error for all samples, with a value of 3.5±1.4(cid:104) for the Archean and earliest Palaeoproterozoic, 4.9±4.0(cid:104) for Neoprotero- 3 zoic, and 4.9±2.6(cid:104) for the Phanerozoic (Fig. 2). These three populations are not significantly different using Student's t-test. Such isotopic consistency implies either no biologic fractionation during weather- ing or consistent biologic involvement in glacial weathering through time. The increase in N concentration through time does not appear to be the result of progressive alteration. There is no correlation between N concentration and δ15N, the chemical index of alteration (CIA), or Cs/Zr (see Supplemental information). If N was being lost due to weathering or volatilization, low N samples should have high δ15N and CIA values. If N were behaving as a fluid-mobile element like Cs, there would be a correlation between N and Cs/Zr, with Zr being a non-fluid mobile element. Such lack of correlation indicates that changes in N concentration are not explained by progressive alteration through time. Two of the low-N samples from the Neoproterozoic may not be fully representative of general, contempo- raneously formed, upper crust. One results from erosion of 1.1 Ga Grenville-associated units (Konnarock Formation) and a second is heavily influenced by erosion of bimodal volcanism (Pocatello Formation) (Gaschnig et al., 2016). We suggest Grenvillian rocks may not be representative of the average upper crust, as they typically expose deeper crust from within an orogenic belt. Clasts in the Konnarock For- mation are primarily middle to lower crustal granites (Rankin, 1993). Globally, granites average 54 ppm N (Johnson and Goldblatt, 2015), much lower than sedimentary or metasedimentary rocks. Though more sparsely measured, volcanic rocks tend to have low N as well, around 0.1 to 10 ppm (Johnson and Gold- blatt, 2015), owing to the high volatility of N during the eruption of oxidized magma (Libourel et al., 2003). Tills that sample only igneous rocks may be biased towards low N. Additionally, while there is correlation between N concentration and Rb and K for low-N samples, there is not for moderate and high-N samples (Supplemental Information). Nitrogen is commonly found as NH+ 4 in geologic samples and substitutes for K in silicates (Honma and Itihara, 1981; Hall, 1999). Many studies have observed correlation between N, K, and Rb in metasediments (Bebout and Fogel, 1992; Busigny et al., 2003). Thus, low-N samples suggest incorporation of metasedimentary N into the crust via recycling of N into the mantle at subduction zones (Marty, 1995; Goldblatt et al., 2009; Busigny et al., 2011; Mikhail and Sverjensky, 2014; Barry and Hilton, 2016). Higher N samples imply an additional, or more efficient, transfer mechanism. There appears to be a relationship between the present continent of sample outcrop and N concentra- tions (Fig. 1). African samples appear to increase in the Palaeoproterozoic and remain high during the Neoproterozoic and Phanerozoic. In contrast, samples from North America are low-moderate into the Neoproterozoic with the most recent sample showing high N concentrations. The single sample from South America and both samples from Asia have moderate to high N. While the strongest control on N concentration appears to be age, it is possible that different continents have a different N history due to differences in their growth history (Supplemental Information). It is also possible that this apparent 4 Figure 1: Nitrogen concentration in glacial tills through time. Means of triplicate analyses of each sample, with standard deviation, are shown with shapes representing modern continent of exposure. Black lines and coloured boxes show mean and standard deviation of Archean-Palaeoproterozoic, Neoproterozoic, and Phanerozoic samples. Low-N samples from units that have eroded primarily igneous terranes in North America are noted, and discussed in the text. 5 Age (Ga)00.511.522.53N (ppm)0100200300400500AfricaAsiaNorth AmericaSouth AmericaKonnarockPocatelloFigure 2: Nitrogen isotope values ((cid:104)) in glacial tills through time. Averages (black lines) for each time period (Archean-Palaeoproterozoic, Neoproterozoic, and Phanerozoic) are equivalent within error (one standard deviation, coloured boxes), indicating no change in the isotopic character of the continental crust through time. 6 Age (Ga)00.511.522.5315N051015AfricaAsiaNorth AmericaSouth AmericaδKonnarockrelationship between present-day geography and N concentration is simply an artifact of a small number of samples. Implications for atmospheric evolution How, then, did N accumulate? The isotopic signature is consistent through the record, and is most similar to either modern average marine NO− 3 or sedimentary N (+5 to +7(cid:104)). The till record is distinct from both the modern atmosphere (0(cid:104)) and the best estimate for the MORB-source mantle value of -5(cid:104), (Marty, 1995). The parsimonious explanation would be incorporation of biologically processed N into the crust with time. Such processing implies N-fixing, thus this N is ultimately of atmospheric origin. The mechanisms which transfer N into the continental crust through time are speculative, but have im- portant implications for models of N distribution through time. The most concentrated reservoirs of continental N are in sedimentary and metasedimentary rocks (concentrations 400-500 ppm), with con- centrations much higher than igneous rocks (e.g., 54 ppm in granites, 0.1-10s ppm in basalts, Johnson and Goldblatt, 2015). One likely mechanism of transfer, then, is burial of biologically-processed N at continental margins followed by accretion. An additional mechanism could be input of N at subduction zones. Sparse N concentration and isotopic data suggests that granitic N content has increased through time (Supplemental Information). In addition, granitic samples show an increase in δ15N values through time, consistent with enhanced incorporation of biologically processed N. The exact timing of incorporation of N is also difficult to determine. There are no glacial deposits from the Mesoproterozoic, rendering this till-based approach ill-suited to this time period. Interestingly, there are two high-N (>200 ppm) samples from the Palaeoproterozoic. Gaschnig et al. (2016) note a distinct change in the composition of tills between the Archean and Palaeoproterozoic glaciations, reflecting a transition from greenstone/komatiite dominated Archean crust to more felsic crust in the Proterozoic. This trend is perhaps mirrored in some of the N analyses, with more felsic crust having higher N concentration. Regardless of the timing of the increase of crustal N, we can compare upper crust N budget from the till proxy to previous work. Johnson and Goldblatt (2015) suggest 150 ± 22 ppm N in the upper crust, while Rudnick and Gao (2014) suggest 83 ppm. We use a total continental crust mass of 2.28 × 1022 kg (Laske et al., 2013), with the upper crust being 53% of the total (Wedepohl, 1995). The Rudnick and Gao (2014) estimate of 83 ppm N yields 0.1 × 1018 kg N (0.25 PAN) in the upper crust and 150 ppm from Johnson and Goldblatt (2015) suggests 0.5 PAN. Based on exposed and buried outcrop area, the upper continental crust is 28% Phanerozoic, 31% Neoproterozoic, 16% Mesoproterozoic, 15% Palaeoproterozoic, and 10% Archean (Goodwin, 1991, 1996). Given this crust distribution, and assuming the Mesoproterozoic has 7 the same N concentration as the Archean/Palaeoproterozoic, our work suggests an upper crust N concen- tration of 210 ppm, equivalent to 2.5 × 1018 kg N, or 0.63 PAN (Table 2). Importantly, the N content of the lower crust is poorly constrained, but could be a significant N reservoir as well. Johnson and Goldblatt (2015) suggest 17 ppm N in the lower crust, which would result in a total continental crust N concentration of 120 ppm and a N mass of 2.7 × 1018 kg. Table 2: Distribution of upper continental crust ages after Goodwin (1991, 1996). We assume that tills accurately sample crust of each age, and that the Mesoproterozoic has the same N concentration as the Archean/Palaeoproterozoic. Age Phanerozoic Neoproterozoic Mesoproterozoic Palaeoproterozoic Archean % crust N (ppm) 28 31 16 15 10 380 250 66 66 66 Total upper crust [N] = 210 ppm mass = 2.5 × 1018 kg N The trend of increased N concentration in the continental crust over time is consistent with non-steady state behavoiur of N through Earth history. Specifically, the till record is consistent with net atmospheric drawdown through time (Goldblatt et al., 2009; Busigny et al., 2011; Barry and Hilton, 2016). While the tills provide a constraint on the evolution of one of the three major N reservoirs (continental crust), determining the exact evolution of the other two (mantle and atmosphere) requires more analyses. We cannot necessarily rule out modern or lower pN2 at specific points in Earth history (e.g., Som et al., 2012, 2016; Marty et al., 2013) but the till data is most consistent with higher atmospheric mass in the past. The balance of mantle outgassing at mid-ocean ridges and arcs to in-gassing at subduction zones is an important, and unconstrained, parameter, over Earth history. Strong net mantle outgassing would be required to have non-decreasing atmospheric N through time. Acknowledgments The authors thank Richard Gaschnig and Roberta Rudnick for providing glacial till samples as well as Dante Canil for the initial suggestion to use glacial tills as a crust composition proxy. We also thank Natasha Drage at the University of Victoria for assistance with sample compilation. Andy Schauer at the University of Washington assisted in isotopic analyses. Funding was provided in an NSERC Discovery Grant to CG. We thank Sami Mikhail and an anonymous reviewer for constructive feedback, as well as Helen Williams for editorial support. 8 Supplementary Information Sample description and collection All samples analyzed were collected by Gaschnig et al. (2016). These samples were collected specifi- cally to assess changes in the composition of the upper continental crust through time. Large ice sheets typically erode a wide variety of rock types, thus till samples should represent an average upper crustal composition. The following is a summary of their collection and sample preparation techniques, but please see the original paper for more detail. The sampling strategy focused on collecting fine-grained material. This was achieved by collecting mas- sive diamictite and some drop-stone bearing argillite. In both cases, the fine-grained matrix was crushed in an alumina jaw crusher, clasts larger than 5mm were removed, and the remaining sample crushed to a fine powder using an alumina swing mill. Excepting the Palaeoproterozoic Pecors, Neoproterozoic Blasskranz, and Ordovician Pakhuis formations, all samples are given as composites of each stratigraphic unit. That is, individual crushed samples were homogenized to give a representative average mixture for each formation. Gaschnig et al. (2016) determined major element compositions using x-ray fluorescence and trace ele- ment composition using laser-ablation ICP-MS techniques. We use their values directly, including chem- ical index of alteration (CIA), which is calculated as Al2O3 / (Al2O3 + CaO + K2O + Na2O). Note that they corrected CaO to remove any influence of carbonates and apatite. Detailed N analytical methods All N measurements were done at the University of Washington's IsoLab, following the procedure out- lined in Stueken (2013). Briefly, between 10-100 mg of sample powder was weighed into a 9×5 mm Sn capsule. Samples and standards were analyzed on a Thermo-Finnigan MAT 253 coupled to a Costech Elemental Analyzer. Standards used were two glutamic acids (GA-1, GA-2), and two internal standards (dried salmon and organic-rich McRae shale). Samples were flash-combusted at 1000 ◦C in a combustion column packed with cobaltous oxide (combustion aid) and silvered cobaltous oxide (sulphur scrubber). Combustion products passed over a reduced copper column at 650 ◦C to convert all N to N2 and absorb excess O2. Lastly, sample gas passed through a magnesium perchlorate trap to absorb water and a 3 m gas chromatography column to separate N2 from O2. All analyses were quantified using IsoDat software. 9 Errors reported for individual are one standard deviation based on triplicate analysis of each sample. The mean and one standard deviation are shown for each age group (Archean/Palaeoproterozoic, Neoprotero- zoic, Phanerozoic) are simply calculated from all samples that fall within each age window. Lower N concentrations generally result in greater uncertainty in isotopic measurements due to smaller amounts of N released during analysis. Thus, isotopic uncertainties for the low N samples, most of the are the Archean/Palaeoproterozoic and some Neoproterozoic, are generally higher (Table S-1). In addition, some error may have been introduced due to not preparing samples in a vacuum. It is possible that some at- mospheric N2 adhered to the powder, though any contamination is suspected to be small due to distinctly non-zero (i.e., non-atmospheric) δ15N values and lack of correlation between N concentration and δ15N. If atmospheric contamination was a major issue, we would expect for samples with high N to have low δ15N values, which is not observed. This interpretation implies that δ15N values in tills are non-zero ini- tially, which is consistent with observed δ15N values from a wider variety of continental rocks (Johnson and Goldblatt, 2015). Constraining post-depositional alteration Crucial to our presented interpretation is demonstrating that the N concentrations have not been altered through time. That is, the lower concentrations observed in older rocks are not simply the result of progressive N loss through time. There are a number of approaches to assess this possibility. Firstly, we observe no correlation between δ15N values and N concentration (Fig. S-1). If progressive N loss was occurring, the expected trend would be higher δ15N values associated with lower N concentrations, due to the preferential loss of 14N during diagenesis. Secondly, comparisons with other elemental compositions indicate lack of alteration through time. As discussed in detail in (Gaschnig et al., 2016), a first-pass approach is to use the Chemical Index of Alteration (CIA), which is defined as: Al2O3/(Al2O3 + CaO* + Na2O + K2O) ×100 (Nesbitt and Young, 1982). This index supposes that alteration of feldspars to clay minerals during chemical alteration and weathering will increase the CIA. There is no observed correlation between N concentration and CIA in these samples (Fig. S-2). As discussed in Gaschnig et al. (2016), till samples with a high CIA likely inherited this signal from a weathered source rock, rather than having experienced extensive chemical weathering themselves. An additional comparison can be made using Large Ion Lithophile (LILE) and High Field Strength Ele- ments (HFSE). Both these groups are incompatible, but in general LILE are fluid-mobile and HFSE are not. We use Cs to represent LILE and Zr to represent HFSE. Neither Zr nor Cs abundances in the crust have shown secular changes through time (Gaschnig et al., 2016). If progressive N loss via aqueous al- 10 Table S-1: Nitrogen concentration and stable isotopic data. Samples analyzed are from Gaschnig et al. (2016) and sam- ple names herein are those used in the original publication. Age is in Ga, N concentration is in ppm and δ15N is in per- mil. Continent indicates continent where sample was col- lected: AF - Africa, NA - North America, AS - Asia, SA - South America. Continent Age N (ppm) Stratigraphic unit Mozaan Group Mozaan Group Mozaan Group Afrikander Formation Afrikander Formation Afrikander Formation Promise Formation West Rand Group Witwatersrand Promise Formation West Rand Group Witwatersrand Promise Formation West Rand Group Witwatersrand Coronation Formation West Rand Group Witwatersrand Coronation Formation West Rand Group Witwatersrand Coronation Formation West Rand Group Witwatersrand Bottle Creek Formation Bottle Creek Formation Bottle Creek Formation Gowganda Formation Gowganda Formation Gowganda Formation Bruce Formation Bruce Formation Bruce Formation Ramsay Lake Formation Ramsay Lake Formation Ramsay Lake Formation Gowganda Formation Gowganda Formation Gowganda Formation Makganyene Formation 11 AF AF AF AF AF AF AF AF AF AF AF AF NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA AF δ15N 4.3 2.9 5.6 2.9 5.8 2.9 3.1 2.9 2.8 2.9 4.8 2.9 4.2 2.9 4.1 2.9 4.5 2.9 2.0 2.9 2.5 2.9 -0.9 2.9 2.8 2.4 0.9 2.4 2.7 2.4 3.4 2.4 0.5 2.4 2.1 2.4 -0.9 2.4 3.6 2.4 6.6 2.4 3.1 2.4 3.6 2.4 3.0 2.4 3.4 2.4 0.5 2.4 2.1 2.4 2.3 4.9 Continued on next page 24 31 28 9 11 16 22 21 28 33 35 30 10 42 48 11 10 13 19 12 30 14 20 14 11 10 13 15 Table S-1 -- continued from previous page Continent Age N (ppm) Stratigraphic unit Makganyene Formation Makganyene Formation Transvaal/Griqualand Timeball Hill Formation Timeball Hill Formation Timeball Hill Formation Duitschland Formation Duitschland Formation Duitschland Formation Konnarock Formation Konnarock Formation Konnarock Formation Gucheng Formation near bottom of unit Gucheng Formation Gucheng Formation Nantuo Formation lower part of unit Nantuo Formation middle part of unit Nantuo Formation top of unit Pocatello Formation upper diamictite Pocatello Formation upper diamictite Pocatello Formation upper diamictite Blaubeker Formation Blaubeker Formation Blaubeker Formation Kaigas Formation Kaigas Formation Kaigas Formation Numees Formation Numees Formation Numees Formation Ghaub Formation Ghaub Formation Ghaub Formation Chuos Formation Chuos Formation Chuos Formation 12 AF AF AF AF AF AF AF AF NA NA NA AS AS AS AS AS AS NA NA NA AF AF AF AF AF AF AF AF AF AF AF AF AF AF AF δ15N 4.8 2.3 2.7 2.3 5.2 2.2 4.5 2.2 5.5 2.2 5.8 2.4 5.3 2.4 5.3 2.4 12.2 0.7 19.3 0.7 16.4 0.7 3.3 0.7 3.4 0.7 3.1 0.7 4.3 0.64 3.2 0.64 3.8 0.64 4.6 0.7 4.1 0.7 5.4 0.7 4.0 0.7 4.7 0.7 4.4 0.7 3.8 0.75 3.7 0.75 3.3 0.75 6.4 0.6 7.8 0.6 8.5 0.6 -0.4 0.64 2.2 0.64 1.2 0.64 2.8 0.7 3.0 0.7 0.7 2.5 Continued on next page 15 18 295 306 305 230 239 220 54 36 45 524 545 542 539 519 530 99 78 78 106 86 74 315 353 193 96 89 100 163 463 320 306 467 397 Stratigraphic unit Gaskiers Formation Gaskiers Formation Gaskiers Formation Bolivia Bolivia Bolivia DwykaEast Group DwykaEast Group DwykaEast Group DwykaWest DwykaWest DwykaWest Till4 Table S-1 -- continued from previous page Continent Age N (ppm) NA NA NA SA SA SA AF AF AF AF AF AF NA 0.58 0.58 0.58 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.001 83 109 103 321 357 314 373 461 386 340 372 332 440 δ15N 1.3 2.6 2.1 3.2 2.6 2.1 2.1 3.3 2.8 8.0 8.0 8.0 4.3 teration through time was the sole cause for the trend in increased N through time, samples that have low N would have a correspondingly low Cs/Zr. While samples with the very lowest N and Cs/Zr are from the Archean/Palaeoproterozoic (Fig. S-3), there are also a number of younger samples with low Cs/Zr and high N.Thus, we suggest that post-depositional aqueous alteration alone cannot explain the trend of decreased N concentrations back in time, though we cannot rule out some alteration for the lowest N samples. Mineral hosts for N We do not know the exact mineral host of N in these samples. Typically, NH + is the most common ge- 4 ological species of N, though there may be small amounts of organic N as well. In detail, some samples that have high N also have a high Al2O3/SiO2 ratio (Fig. S-5), which is indicative of a weathering envi- ronment rich in feldspars and clays compared to quartz (Gaschnig et al., 2016, and references therein). Thus, for many samples, K-bearing phases are a likely host for N as NH + 4 , though the lack of correlation between N and Rb (which also substitutes for K), indicates that this simple relationship may not be true for all samples (Fig. S-4). Palaeoproterozoic samples from South Africa exemplify this relationship (Fig. S-5). Other high N sam- ples, such as those from North America, have a similar Al2O3/SiO2 ratio as higher N samples from Africa 13 Figure S-1: Nitrogen isotope values plotted against N concentration. Lack of correlation between isotopes and concentrations from samples of all ages suggests that there has not been N loss during diagenesis. Nitrogen loss tends to result in samples with low N concentration having high δ15N values, which is not observed. 14 N (ppm)0100200300400500600/15N0246810121416Archean/PaleoproterozoicNeoproterozoicPhanerozoicAfricaAsiaNorth AmericaSouth AmericaFigure S-2: Nitrogen concentration plotted against Chemical Index of Alteration (CIA). See text for details, but lack of correlation suggests that aqueous alteration alone cannot explain the increase in N concentration through time. 15 CIA5060708090100N (ppm)0100200300400500600Archean/PaleoproterozoicNeoproterozoicPhanerozoicAfricaAsiaNorth AmericaSouth AmericaFigure S-3: Nitrogen concentration plotted against Cs/Zr. See text for details, but lack of correlation suggests that aqueous alteration alone cannot explain the increase in N concentration through time, with the possible exception of the lowest N samples. There is no change in the Cs/Zr ratio through time, suggesting that geographic and temporal evolution of Cs and Zr does not explain variation seen in a. 16 Cs/Zr 00.050.1N (ppm)0100200300400500600a.Archean/PaleoproterozoicNeoproterozoicPhanerozoicAge (Ga)0123Cs/Zr00.010.020.030.040.050.060.070.080.09b.AfricaAsiaNorth AmericaSouth AmericaFigure S-4: Nitrogen concentration plotted against Rb and K2O. We note that for low N samples there is a positive correlation, but for moderate and high N samples, no correlation is observed. 17 Rb (ppm)0100200300N (ppm)0100200300400500600K2O (wt. %)0246N (ppm)0100200300400500600Archean/PaleoproterozoicNeoproterozoicPhanerozoicAfricaAsiaNorth AmericaSouth AmericaFigure S-5: Nitrogen concentration plotted against Al2O3/SiO2. and South America, indicating clays/feldspars may not be the main host for N. Another mineral may be the host in these settings. Nitrogen concentrations by continent As mentioned in the main text, there is an apparent relationship between the current continent of each sample and N concentration. That is, samples from Africa, South America, and Asia are generally high while samples from North American are generally low. The strongest correlation both regionally and globally is age, but another possibility that could explain some of the geographic control is a different history of continental growth and assemblage. Zircon ages from Africa and North America from the 18 Al2O3/SiO20.120.140.160.180.20.220.240.260.280.3N (ppm)0100200300400500600Archean/PaleoproterozoicNeoproterozoicPhanerozoicAfricaAsiaNorth AmericaSouth AmericaFigure S-6: Histograms of detrital zircons from Africa, South America, North America, and Asia through time. African zircons have two peaks at 2.1 and 0.75 Ga, while South America has the peak at 2.1 Ga, and perhaps a peak from 0 to 0.5 Ga. Both North America and Asia have a single dominant peak, at 1.2 Ga and 0-0.5 Ga, respectively. It is possible that peaks in ages correspond to periods of continental growth and, by extension, periods of N-sequestration in the crust. 19 compilation of (Belousova et al., 2010) show two peaks and one peak in ages, respectively (Figs. S-6). If these zircon ages peaks correspond with periods of enhanced continental growth, it would suggest that Africa grew somewhat earlier than North America, and correspondingly biologically processed N was incorporated during this phase of continental growth. A major continental growth period occurred later in North America, around 1.2 Ga, with an increase in N in till samples not seen till the Phanerozoic. There would be some lag time between continental growth and erosion by glaciers, thus these periods of growth represent a maximum hypothesized age of N incorporation for each continent. A reasonable test of this pulsed N addition hypothesis would be the analysis of N concentration in felsic intrusive igneous rocks, which are more temporally associated with the continental-growth phases. Though there are few till samples from Asia and South America, distinct zircon age peak distributions indicate possible different continental growth histories, with correspondingly distinct N histories. It is also possible that paleogeography could exert a control over timing of N incorporation. However, if the source of N to the continental crust is biologically processed, subduction zone transported material, both the manner of biologic processing and geometry and extent of subduction zones should have a greater effect. Perhaps, though, if continental subduction zones are located nearer to more productive areas, more N could be buried with organic matter and processed in a subduction zone. Indeed, areas with high productivity near Central America have more N in sediments than low-productivity areas in the eastern Pacific (Elkins et al., 2006; Mitchell et al., 2010). In the modern ocean, areas with high levels of N-fixing (driving atmospheric N2 drawdown) are mostly identified in the tropics. Thus, it is possible that high N-fixing or high productivity areas overlying subduction zones could enhance incorporation of N into continental crust. The distribution of N-fixing regions throughout geologic time are poorly constrained; we are not aware of any data speaking to this directly. In addition, the temperature/pH/redox of subduction zones seems to exert strong control over the fate of subducted N (Busigny et al., 2011; Mikhail and Sverjensky, 2014), and might be more important than paleogeography. Granitic N Nitrogen concentration and isotopic data from the compilation of Johnson and Goldblatt (2015) are shown in Fig. S-7. These are data from both whole rock and mineral specific N analyses, scaled up to whole rock values. The rocks analysed are from the British Isles (Hall, 1987, 1988, 1993; Cooper and Bradley, 1990; Bebout et al., 1999), southern Europe (Hall et al., 1991; Hall, 1999), the Canadian shield (Jia and Kerrich, 2000), Iran (Ahadnejad et al., 2011), Finland (Itihara and Suwa, 1985), and the United States (Hoering, 1955). While data are sparse, especially isotopic data, they are consistent with an increase in crustal N 20 Figure S-7: Nitrogen concentration and δ15N values of granitic rocks through time. These data are consistent with an increase in N content of continental crust through time, with increase in isotopic values consistent with a biologically processed source for said N. 21 Age (Ga)0123N (ppm)050100150200250Age (Ga)0123δ15N-404812through time. Samples from the Archean and Palaeoproterozoic have a mean N concentration of 16 ppm, while those from the Phanerozoic average 43 ppm N. These means are statistically different as shown by Student's t-test (Student, 1908). Isotopic data may increase from depleted, mantle-like values towards more enriched, biologic/sedimentary values through time. Analyses are heavily biased towards Europe, and especially the British Isles. Additional geographic coverage, in addition to temporal coverage, would greatly help with future interpretations. References Ahadnejad, V., Hirt, A. M., Valizadeh, M.-V., and Bokani, S. J. (2011). The ammonium content in the Malayer igneous and metamorphic rocks (Sanandaj-Sirjan Zone, Western Iran). Geologica Carpath- ica, 62(2):171 -- 180. Barry, P. and Hilton, D. (2016). Release of subducted sedimentary nitrogen throughout Earth's mantle. Geochemical Perspectives Letters, 2:148 -- 159. Bebout, G., Cooper, D., Bradley, A. D., and Sadofsky, S. J. (1999). Nitrogen-isotope record of fluid- rock interactions in the Skiddaw Auerole and granite, English Lake District. American Mineralogist, 84:1495 -- 1505. Bebout, G. and Fogel, M. (1992). Nitrogen-isotope compositions of metasedimentary rocks in the Catalina Schist, California: implications for metamorphic devolatilization history. Geochimica et Cos- mochimica Acta, 56(7):2839 -- 2849. Belousova, E., Kostitsyn, Y., Griffin, W. L., Begg, G. C., O'Reilly, S. Y., and Pearson, N. J. (2010). The growth of the continental crust: constraints from zircon Hf-isotope data. Lithos, 119(3):457 -- 466. Berner, R. A. (2006). Geological nitrogen cycle and atmospheric N2 over phanerozoic time. Geology, 34(5):413 -- 415. Boyd, S. (2001). Nitrogen in future biosphere studies. Chemical Geology, 176(1):1 -- 30. Breneman, A. (1889). The Fixation of Atmospheric Nitrogen. (Concluded from Issue 1). Journal of the American Chemical Society, 11(2):31 -- 48. Busigny, V., Cartigny, P., and Philippot, P. (2011). Nitrogen isotopes in ophiolitic metagabbros: A re-evaluation of modern nitrogen fluxes in subduction zones and implication for the early earth atmo- sphere. Geochimica et Cosmochimica Acta, 75:7502 -- 7521. Busigny, V., Cartigny, P., Philippot, P., Ader, M., and Javoy, M. (2003). Massive recycling of nitrogen and other fluid-mobile elements (K, Rb, Cs, H) in a cold slab environment: evidence from HP to UHP oceanic metasediments of the Schistes Lustr´es nappe (western Alps, Europe). Earth and Planetary Science Letters, 215(1):27 -- 42. 22 Byrne, B. and Goldblatt, C. (2015). Diminished greenhouse warming from Archean methane due to solar absorption lines. Climate of the Past, 11:559 -- 570. Canil, D. and Lacourse, T. (2011). An estimate for the bulk composition of juvenile upper continental crust derived from glacial till in the North American Cordillera. Chemical Geology, 284(3):229 -- 239. Cartigny, P. and Marty, B. (2013). Nitrogen isotopes and mantle geodynamics: The emergence of life and the atmosphere-crust-mantle connection. Elements, pages 359 -- 366. Cooper, D. and Bradley, A. (1990). The ammonium content of granites in the English Lake District. Geological Magazine, 127(06):579 -- 586. Elkins, L., Fischer, T., Hilton, D., Sharp, Z., McKnight, S., and Walker, J. (2006). Tracing nitrogen in volcanic and geothermal volatiles from the Nicaraguan volcanic front. Geochimica et Cosmochimica Acta, 70(20):5215 -- 5235. Fuelner, G. (2012). The faint young sun problem. Reviews of Geophysics, 50:1 -- 29. Gaschnig, R. M., Rudnick, R. L., McDonough, W. F., Kaufman, A. J., Valley, J. W., Hu, Z., Gao, S., and Beck, M. L. (2016). Compositional evolution of the upper continental crust through time, as constrained by ancient glacial diamictites. Geochimica et Cosmochimica Acta, 186:316 -- 343. Goldblatt, C., Claire, M., Lenton, T., Matthews, A., Watson, A., and Zahnle, K. (2009). Nitrogen- enhanced greenhouse warming on early Earth. Nature Geoscience, 2(12):891 -- 896. Goldschmidt, V. (1933). Grundlagen der quantitativen Geochemie. Fortschr. Mineral. Krist. Petrog, 17(1):12. Goodwin, A. M. (1991). Precambrian Geology: the Dynamic Evolution of the Continental Crust. Aca- demic Press London. Goodwin, A. M. (1996). Principles of Precambrian geology. Academic Press. Gruber, N. and Galloway, J. (2008). An Earth-system perspective of the global nitrogen cycle. Nature, 451(7176):293 -- 296. Hall, A. (1987). The ammonium content of Caledonian granites. Journal of the Geological Society, 144(4):671 -- 674. Hall, A. (1988). The distribution of ammonium in granites from South-West England. Journal of the Geological Society, 145(1):37 -- 41. Hall, A. (1993). Application of the indophenol blue method to the determination of ammonium in silicate rocks and minerals. Applied geochemistry, 8(1):101 -- 105. Hall, A. (1999). Ammonium in granites and its petrogenetic significance. Earth-Science Reviews, 45(3):145 -- 165. 23 Hall, A., Bencini, A., and Poli, G. (1991). Magmatic and hydrothermal ammonium in granites of the Tuscan magmatic province, Italy. Geochimica et Cosmochimica Acta, 55(12):3657 -- 3664. Hart, M. H. (1978). The evolution of the atmosphere of the earth. Icarus, 33(1):23 -- 39. Hoering, T. (1955). Variations of nitrogen-15 abundance in naturally occurring substances. Science, 122(3182):1233 -- 1234. Honma, H. and Itihara, Y. (1981). Distribution of ammonium in minerals of metamorphic and granitic rocks. Geochimica et Cosmochimica Acta, 45(6):983 -- 988. Itihara, Y. and Suwa, K. (1985). Ammonium contents of biotites from Precambrian rocks in Finland: The 4 as a possible chemical fossil. Geochimica et Cosmochimica Acta, 49(1):145 -- 151. significance of NH+ Jia, Y. and Kerrich, R. (2000). Giant quartz vein systems in accretionary orogenic belts: the evidence for a metamorphic fluid origin from δ15N δ13C studies. Earth and Planetary Science Letters, 184(1):211 -- 224. Johnson, B. W. and Goldblatt, C. (2015). The Nitrogen Budget of Earth. Earth Science Reviews. Kavanagh, L. and Goldblatt, C. (2015). Using raindrops to constrain past atmospheric density. Earth and Planetary Science Letters, 413:51 -- 58. Klingler, J., Mancinelli, R., and White, M. (1989). Biological nitrogen fixation under primordial martian partial pressures of dinitrogen. Advances in Space Research, 9(6):173 -- 176. Laske, G., Masters, G., Ma, Z., and Pasyanos, M. (2013). Update on CRUST1.0 - A 1-degree Global Model of Earth's Crust. Geophysical Research Abstracts, 15(Abstract EGU2013-2658). Libourel, G., Marty, B., and Humbert, F. (2003). Nitrogen solubility in basaltic melt. Part I. Effect of oxygen fugacity. Geochimica et Cosmochimica Acta, 67(21):4123 -- 4135. Marty, B. (1995). Nitrogen content of the mantle inferred from N2 -- Ar correlation in oceanic basalts. Nature, 377(6547):326 -- 329. Marty, B., Zimmermann, L., Pujol, M., Burgess, R., and Philippot, P. (2013). Nitrogen isotopic compo- sition and density of the Archean atmosphere. Science, 342:101 -- 104. Mikhail, S. and Sverjensky, D. A. (2014). Nitrogen speciation in upper mantle fluids and the origin of Earth's nitrogen-rich atmosphere. Nature Geoscience, 7:816 -- 819. Mitchell, E. C., Fischer, T. P., Hilton, D. R., Hauri, E. H., Shaw, A. M., de Moor, J. M., Sharp, Z. D., and Kazahaya, K. (2010). Nitrogen sources and recycling at subduction zones: Insights from the Izu-Bonin-Mariana arc. Geochemistry, Geophysics, Geosystems, 11(2). Nesbitt, H. and Young, G. (1982). Early Proterozoic climates and plate motions inferred from major element chemistry of lutites. Nature, 299(5885):715 -- 717. 24 Rankin, D. W. (1993). The volcanogenic Mount Rogers Formation and the overlying glaciogenic Kon- narock Formation; two late Proterozoic units in southwestern Virginia. Technical report, USGPO; US Geological Survey, Map Distribution. Rudnick, R. and Gao, S. (2014). Composition of the Continental Crust. Treatise on Geochemistry, 4:1 -- 69. Sagan, C. and Mullen, G. (1972). Earth and Mars: Evolution of atmospheres and surface temperatures. Science, 177(4043):52 -- 56. Sheldon, N. D. (2006). Precambrian paleosols and atmospheric CO2 levels. Precambrian Research, 147:148 -- 155. Som, S. M., Buick, R., Hagadorn, J. W., Blake, T. S., Perreault, J. M., Harnmeijer, J. P., and Catling, D. C. (2016). Earth's air pressure 2.7 billion years ago constrained to less than half of modern levels. Nature Geoscience, 9:448 -- 451. Som, S. M., Catling, D. C., Harnmeijer, J. P., Polivka, P. M., and Buick, R. (2012). Air density 2.7 billion years ago limited to less than twice modern levels by fossil raindrop imprints. Nature, 484(7394):359 -- 362. Student (1908). The probable error of a mean. Biometrika, pages 1 -- 25. Stueken, E., Kipp, M., Koehler, M., Schwieterman, E., Johnson, B. W., and Buick, R. (2016). Modeling pN2 through geologic time: Implications for atmospheric biosignatures. Astrobiology, 16(12):949 -- 963. Stueken, E. E. (2013). A test of the nitrogen-limitation hypothesis for retarded eukaryote radiation: nitro- gen isotopes across a Mesoproterozoic basinal profile. Geochimica et Cosmochimica Acta, 120:121 -- 139. Tolstikhin, I. and Marty, B. (1998). The evolution of terrestrial volatiles: a view from helium, neon, argon and nitrogen isotope modelling. Chemical Geology, 147(1-2):27 -- 52. Wedepohl, H. K. (1995). The composition of the continental crust. Geochimica et Cosmochimica Acta, 59(7):1217 -- 1232. Wordsworth, R. and Pierrehumbert, R. (2013). Hydrogen-nitrogen greenhouse warming in earth's early atmosphere. Science, 339(6115):64 -- 67. Zerkle, A. and Mikhail, S. (2017). The geobiological nitrogen cycle: From microbes to the mantle. Geobiology, 15:343 -- 352. Zhang, Y. and Zindler, A. (1993). Distribution and evolution of carbon and nitrogen in Earth. Earth and Planetary Science Letters, 117(3):331 -- 345. 25
1503.09184	1	1503	2015-03-31T19:48:58	Transits and starspots in the WASP-6 planetary system	[ "astro-ph.EP" ]	We present updates to \textsc{prism}, a photometric transit-starspot model, and \textsc{gemc}, a hybrid optimisation code combining MCMC and a genetic algorithm. We then present high-precision photometry of four transits in the WASP-6 planetary system, two of which contain a starspot anomaly. All four transits were modelled using \textsc{prism} and \textsc{gemc}, and the physical properties of the system calculated. We find the mass and radius of the host star to be $0.836\pm 0.063\,{\rm M}_\odot$ and $0.864\pm0.024\,{\rm R}_\odot$, respectively. For the planet we find a mass of $0.485\pm 0.027\,{\rm M}_{\rm Jup}$, a radius of $1.230\pm0.035\,{\rm R}_{\rm Jup}$ and a density of $0.244\pm0.014\,\rho_{\rm Jup}$. These values are consistent with those found in the literature. In the likely hypothesis that the two spot anomalies are caused by the same starspot or starspot complex, we measure the stars rotation period and velocity to be $23.80 \pm 0.15$\,d and $1.78 \pm 0.20$\,km\,s$^{-1}$, respectively, at a co-latitude of 75.8$^\circ$. We find that the sky-projected angle between the stellar spin axis and the planetary orbital axis is $\lambda = 7.2^{\circ} \pm 3.7^{\circ}$, indicating axial alignment. Our results are consistent with and more precise than published spectroscopic measurements of the Rossiter-McLaughlin effect. These results suggest that WASP-6\,b formed at a much greater distance from its host star and suffered orbital decay through tidal interactions with the protoplanetary disc.	astro-ph.EP	astro-ph	Mon. Not. R. Astron. Soc. 000, 000 -- 000 (0000) Printed 13 August 2019 (MN LATEX style file v2.2) Transits and starspots in the WASP-6 planetary system Jeremy Tregloan-Reed 1,2⋆, John Southworth 2, M. Burgdorf 3, S. Calchi Novati 4,5,6, M. Dominik 7†, F. Finet 8,9, U. G. Jørgensen 10, G. Maier 11, L. Mancini 12, S. Proft 11, D. Ricci 13,14,15, C. Snodgrass 16, V. Bozza 5,17, P. Browne 7, P. Dodds 7, T. Gerner 11, K. Harpsøe 10, T. C. Hinse 18,19, M. Hundertmark 7, N. Kains 20, E. Kerins 21, C. Liebig 7, M. T. Penny 22, S. Rahvar 23, K. Sahu 20, G. Scarpetta 6,5,17, S. Schafer 24, F. Schonebeck 11, J. Skottfelt 10, J. Surdej 8, 1 NASA Ames Research Center, Moffett Field, CA 94035, USA 2 Astrophysics Group, Keele University, Staffordshire, ST5 5BG, UK 3 Universitat Hamburg Meteorologisches Institut Bundesstrasse 55 20146 Hamburg, Germany 4 NASA Exoplanet Science Institute, MS 100-22, California Institute of Technology, Pasadena, CA 91125, US 5 Dipartimento di Fisica "E.R. Caianiello", Universit`a di Salerno, Via Giovanni Paolo II 132, 84084, Fisciano (SA), Italy 6 Istituto Internazionale per gli Alti Studi Scientifici (IIASS), 84019 Vietri Sul Mare (SA), Italy 7 SUPA, University of St Andrews, School of Physics & Astronomy, North Haugh, St Andrews, KY16 9SS, UK 8 Institut d'Astrophysique et de G´eophysique, Universit´e de Li`ege, 4000 Li`ege, Belgium 9 Aryabhatta Research Institute of Observational Sciences (ARIES), Manora Peak, Nainital-263 129, Uttarakhand, India 10 Niels Bohr Institute & Centre for Star and Planet Formation, University of Copenhagen, Østervoldgade 5, 1350 Copenhagen K, Denmark 11 Astronomisches Rechen-Institut, Zentrum fur Astronomie, Universitat Heidelberg, Monchhofstrasse 12-14, 69120 Heidelberg, Germany 12 Max Planck Institute for Astronomy, Konigstuhl 17, 69117 Heidelberg, Germany 13 Observatorio Astron´omico Nacional, Instituto de Astronom´ıa -- Universidad Nacional Aut´onoma de M´exico, Ap. P. 877, Ensenada, BC 22860, Mexico 14 Instituto de Astrof´ısica de Canarias, E-38205 La Laguna, Tenerife, Spain 15 Universidad de La Laguna, Departmento de Astrof´ısica, E-38206 La Laguna, Tenerife, Spain 16 Planetary and Space Sciences, Department of Physical Sciences, The Open University, Milton Keynes, MK7 6AA, UK 17 Istituto Nazionale di Fisica Nucleare, Sezione di Napoli, 80126 Napoli, Italy 18 Korea Astronomy and Space Science Institute, Daejeon 305-348, Republic of Korea 19 Armagh Observatory, College Hill, Armagh BT61 9DG, UK 20 Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218, USA 21 Jodrell Bank Centre for Astrophysics, University of Manchester, Oxford Road, Manchester M13 9PL, UK 22 Department of Astronomy, Ohio State University, 140 W. 18th Ave., Columbus, OH 43210, USA 23 Department of Physics, Sharif University of Technology, P. O. Box 11155-9161 Tehran, Iran 24 Institut fur Astrophysik, Georg-August-Universitat Gottingen, Friedrich-Hund-Platz 1, 37077 Gottingen, Germany 13 August 2019 ABSTRACT We present updates to PRISM, a photometric transit-starspot model, and GEMC, a hybrid op- timisation code combining MCMC and a genetic algorithm. We then present high-precision photometry of four transits in the WASP-6 planetary system, two of which contain a starspot anomaly. All four transits were modelled using PRISM and GEMC, and the physical properties of the system calculated. We find the mass and radius of the host star to be 0.836 ± 0.063 M⊙ and 0.864 ± 0.024 R⊙, respectively. For the planet we find a mass of 0.485 ± 0.027 MJup, a radius of 1.230 ± 0.035 RJup and a density of 0.244 ± 0.014 ρJup. These values are con- sistent with those found in the literature. In the likely hypothesis that the two spot anomalies are caused by the same starspot or starspot complex, we measure the stars rotation period and velocity to be 23.80 ± 0.15 d and 1.78 ± 0.20 km s−1, respectively, at a co-latitude of 75.8◦. We find that the sky-projected angle between the stellar spin axis and the planetary orbital axis is λ = 7.2◦ ± 3.7◦, indicating axial alignment. Our results are consistent with and more precise than published spectroscopic measurements of the Rossiter-McLaughlin effect. These results suggest that WASP-6 b formed at a much greater distance from its host star and suffered orbital decay through tidal interactions with the protoplanetary disc. Key words: planetary systems -- stars: fundamental parameters -- stars: spots -- stars: individual: WASP-6 -- techniques: photometric c(cid:13) 0000 RAS ⋆ Email: [email protected] † Royal Society University Research Fellow 2 Tregloan-Reed et al. 1 INTRODUCTION At present1 a total of 1890 planets outside of our own solar sys- tem are listed in the authoritative catalogue of Schneider et al. (2011). Of these approximately two thirds have been discovered from ground-based (e.g. SuperWasp: Pollacco et al. 2006; HAT: Bakos et al. 2004) or space-based (CoRoT: Baglin et al. 2006; Ke- pler: Borucki et al. 2010) transit surveys, and later confirmed by use of the radial velocity technique (Butler et al. 1996, 1999; Queloz et al. 2000). Many more candidate exoplanets have been listed in the literature, mainly from the Kepler satellite survey which has also detected several Earth-size planets in the habitable zone (HZ) of their parent star, indicating new worlds with mass and size simi- lar to our own Earth (Borucki et al. 2012; Borucki et al. 2013). During a planetary transit, the planet follows a path (called the transit chord) across the surface of the stellar disc and can be used to probe changes in brightness on the stellar surface (Silva 2003). Starspots have different temperatures to the surrounding photosphere, so emit a different amount of flux. Because photom- etry measures the change in intensity as a function of time, the oc- cultation of a starspot by the planet causes an anomaly in the light curve (Silva 2003). The anomaly is either an increase or decrease in the amount of light received from the star. If the starspot is a cool spot then the amount of light will increase when the planet crosses the starspot (Rabus et al. 2009; Pont et al. 2007; Winn et al. 2010b). If the starspot is a hot spot (e.g. a facula) then the amount of light will reduce when the planet occults the spot. At present, when a light curve of a transiting exoplanet is observed to have a starspot anomaly, the transit and the spot are generally modelled separately (e.g D´esert et al. 2011; Maciejew- ski et al. 2011; Nutzman et al. 2011; Sanchis-Ojeda et al. 2011). First, a transit model is fitted to the datapoints not affected by the starspot anomaly. Then the spot-affected residuals versus the best- fitting model are modelled using a Gaussian function (e.g. Sanchis- Ojeda et al. 2011; Sanchis-Ojeda & Winn 2011). This method ne- glects the fact that the starspot affects the entire transit shape and not just the section where the planet crosses the spot (Ballerini et al. 2012). Carter et al. (2011) use the idea that a starspot on the stellar disc will affect the transit depth to explain the observed changes in transit depth for GJ 1214. This is due to the change in the star's brightness in its long-term light curve due to starspots rotating on and off the stellar disc. The transit depth is not the only property of a transit light curve that the starspot affects: it also affects the determination of the measured stellar mean density, stellar radius, orbital inclination and limb darkening (LD) coefficients (Ballerini et al. 2012). The LD coefficients depend on wavelength: because a starspot has a dif- ferent temperature compared to the surrounding photosphere, it has a different spectral energy distribution and thus different LD coeffi- cients. Therefore the application of a LD law with a single set of co- efficients to the entire stellar surface causes a bias in the modelling process (Ballerini et al. 2012). The difference in LD coefficients between the spot and the photosphere can be as much as 30% in the UV. The effects on the measured stellar radius and orbital incli- nation of the system are artifacts from errors in the measured plane- tary radius, which is derived from the transit depth. A change in the measured planetary radius must be compensated for by a change in the measured stellar radius or semimajor axis in order to retain the same transit duration. Starspots can also affect the measured transit midpoint (Sanchis-Ojeda et al. 2011; Barros et al. 2013) and create 1 (http://exoplanet.eu) accessed on 2015/02/20 false positives in transit timing measurements. Sanchis-Ojeda et al. (2011) calculated that a starspot anomaly in a transit of WASP-4 with an amplitude of 0.3 to 0.5 mmag could produce a timing noise of five to ten seconds. 1.1 Introducing WASP-6 The transiting planetary system WASP-6 was discovered by Gillon et al. (2009) using photometry from the WASP-South telescope. They determined an orbital period of P = 3.361 days for the planet WASP-6 b. Dedicated photometric observations were then performed in the i′ band using the 2-m Faulkes Telescope South (FTS) and in a broad V +R band using the RISE instrument (Steele et al. 2008) on the 2-m Liverpool Telescope (LT). Radial velocity (RV) measurements were obtained using two spectrographs: CORALIE on the 1.2-m Euler telescope (Baranne et al. 1996; Queloz et al. 2000) and HARPS on the ESO 3.6- m telescope (Mayor et al. 2003). Gillon et al. (2009) determined the stellar mass and radius to be M⋆ = 0.88+0.05 −0.08 M⊙ and R⋆ = 0.870+0.025 −0.036 R⊙, respectively. They found the planetary mass and radius to be Mp = 0.503+0.019 −0.038 MJup and Rp = 1.224+0.051 −0.052 RJup. They also determined a value for the projected stellar rotational velocity of v sin I = 1.4 ± 1.0 km s−1 from mea- surements of line widths in the HARPS spectra with a macrotur- bulence (vmac) value of 2 km s−1. They noted that if a value of vmac = 0 km s−1 is used then v sin I = 3.0 ± 0.5 km s−1, while if vmac became slightly larger than 2 km s−1 then v sin I would drop to zero. From their RVs Gillon et al. (2009) measured the Rossiter- McLaughlin (RM) effect. They found that the system is in align- ment with a sky-projected spin orbit alignment, λ = 11+14 −18 deg. The spectral analysis of 11 WASP host stars by Doyle et al. (2013) included WASP-6 A. Doyle et al. (2013) derived new val- ues for the stellar mass and radius of M⋆ = 0.87 ± 0.06 M⊙ and R⋆ = 0.77 ± 0.07 R⊙, in agreement with those of Gillon et al. (2009). Doyle et al. (2013) determined vmac = 1.4 ± 0.3 km s−1 and v sin I = 2.4 ± 0.5 km s−1, and an effective temperature of Teff = 5375 ± 65 K. An optical transmission spectrum for WASP-6 has been con- structed using multi-object differential spectrophotometry with the IMACS spectrograph on the Magellan Baade telescope (Jord´an et al. 2013). The observations comprised 91 spectra covering 480 -- 860 nm. The analysis yielded a mostly featureless transmission spectrum with evidence of atmospheric hazes and condensates. Most recently, Nikolov et al. (2014) used the Hubble Space Tele- scope to perform transmission spectroscopy of WASP-6, and found a haze in the atmosphere of WASP-6 b. They also determined a ro- tational modulation of Prot = 23.6 ± 0.5 d for WASP-6 A. 2 UPDATES TO PRISM & GEMC A code written in IDL2 called PRISM (Planetary Retrospective Inte- grated Star-spot Model) was developed to model a starspot anomaly in transit light curves of WASP-19 (see Tregloan-Reed et al. 2013). PRISM uses a pixellation approach to represent the star and planet on a two-dimensional array in Cartesian coordinates. This makes it possible to model the transit, LD and starspots on the stellar disc 2 The acronym IDL stands for Interactive Data Language and is a trademark of Exelis Visual Information Solutions. For further details see http://www.exelisvis.com/ProductsServices/IDL.aspx. c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 simultaneously. LD was implemented using the standard quadratic law. PRISM uses the ten parameters given in Table 1 to model the system, where the fractional stellar and planetary radii are defined as the absolute radii scaled by the semimajor axis (r⋆,p = R⋆,p/a). A new optimisation algorithm called GEMC (Genetic Evolu- tion Markov Chain) was also created alongside PRISM to help im- prove the efficiency of finding a global solution in a rugged param- eter space compared to conventional MCMC algorithms (Tregloan- Reed et al. 2013). GEMC is a hybrid between an MCMC and a ge- netic algorithm3 and is based on the Differential Evolution Markov Chain (DE-MC) put forward by Ter Braak (2006). During the 'burn-in' stage GEMC runs N chains in parallel and for every gen- eration each chain is perturbed in a vector towards the current best- fitting chain. Once the burn-in stage has been completed GEMC switches to a conventional MCMC algorithm (each chain used in the burn-in begins independent MCMC runs) to determine the pa- rameter uncertainties (see Tregloan-Reed et al. 2013). Since the development of both PRISM and GEMC other authors have used the codes to not only help ascertain the photometric pa- rameters of a transiting system but to also derive the parameters of the starspots observed in transit light curves (e.g. Mancini et al. 2013, 2014; Mohler-Fischer et al. 2013). B´eky et al. (2014) used PRISM to help calibrate their semi-analytic transit-starspot model SPOTROD. Before using both PRISM and GEMC in modelling the WASP- 6 system, it was decided to make a few improvements4. The orig- inal version of PRISM assumed a circular orbit, as most transiting planets either have a circular orbit or lack a measurement of the or- bital eccentricity. However, Gillon et al. (2009) found that the orbit of WASP-6 b has a small orbital eccentricity of e = 0.054+0.018 −0.015, with an argument of periastron ω = 97.4+6.9 −13.2 degrees. As a con- sequence PRISM was extended to allow for eccentric orbits. e and ω have been set to roam within the physically bounded ranges of 0 6 e 6 1 and 0◦ 6 ω 6 360◦. A Gaussian prior is used to help constrain the parameter values close to the expected values found in the literature. The logic behind using a Gaussian prior stems from the fact that it is not possible to ascertain these values from pho- tometry alone (due to only observing a small fraction of the orbit) unless an occultation is observed (Kipping et al. 2012). Because we have the knowledge of where the values of e and ω should lie and that they have an effect on the other system parameters (in par- ticular i and r⋆), it is imperative to examine every potential solu- tion selected from a Gaussian probability distribution of e and ω to accurately estimate the uncertainties in all of the other system parameters. It was shown by both Silva-Valio et al. (2010) and Mohler- Fischer et al. (2013) that in some cases there can be more than one starspot anomaly in a single transit light curve. While PRISM was originally designed to model multiple starspots, the static coding of GEMC made it only possible to fit for either a single starspot or a spot free stellar surface. To facilitate further work GEMC was modi- fied to fit for multiple starspots. This was accomplished by allowing the initial reading of the input file to be dynamic, so GEMC can de- termine the number of starspots to be fitted based on the number of parameters used. This can be done by adding multiple spot pa- rameter ranges in the input file. It is possible to fix the position of 3 A genetic algorithm mimics biological processes by spawning successive generations of solutions based on breeding and mutation operators from the previous generation. 4 The new versions of both PRISM and GEMC are available from http://www.astro.keele.ac.uk/∼jtr c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 Transits and starspots of WASP-6 3 a starspot and therefore assign starspots to sections of the stellar disc where they will not be occulted by the planet, thus allowing investigations of the effects of unocculted starspot on transit light curves. PRISM was designed to use the pixellation approach, and to maintain numerical resolution was hard-coded to set the planetary radius at 50 pixels. The host star's radius in pixels was scaled ac- cordingly based on the input parameters. The new version now al- lows users to set the size of the planetary radius in pixels. This makes it possible to reduce the amount of time required to com- plete each model iteration, at the cost of numerical resolution (see Section 2.1 for more details). In tests using a planet radius set at 15 pixels it took PRISM and GEMC approximately 13 s to model a single generation of 256 solutions using synthetic data, which equates to approximately 0.05 seconds per iteration. For compari- son, a planetary radius of 50 pixels results in approximately 0.47 s per iteration.5 To increase the efficiency of determining the parameter uncer- tainties, the MCMC component of GEMC was replaced with DE- MC (Ter Braak 2006). DE-MC combines the genetic algorithm dif- ferential evolution (DE) (Price & Storn 1997; Storn & Price 1997) with MCMC. The combination of DE and MCMC is used to solve a problem in MCMC by determining the orientation and the scale of the step sizes. Adaptive directional sampling in MCMC does solve the orientation problem, but not the scale (Ter Braak 2006). DE- MC works by creating a population of MCMC chains whose start- ing points are initialised from overdispersed states and instead of letting the chains run independently and checking for convergence (e.g. Gelman & Rubin 1992) they are instead run in parallel and learn from each other. The perturbation steps taken by each chain are given by Eq. 1. Assuming a d-dimensional parameter space and using N chains then the population X is a N × d matrix, with the chains labelled x1, x2, . . . xN . Therefore the proposal vector xp is generated by: xp = xi + γ (xR1 − xR2) + e where xi is the current ith chain, γ is the scale factor calculated from γ = 2.4/√2d (Ter Braak 2006), xR1 and xR2 are two ran- domly selected chains and e is drawn from a symmetric distribu- tion with a small variance compared to that of the target. xp is then tested for fitness and if accepted it is used as the next step in xi. (1) After the 'burn in' stage of a MCMC chain, determining the required step size to allow a 20 -- 25% acceptance rate can be diffi- cult. For a transit light curve altering the orbital inclination, i, by 0.05% should only cause a small increase in χ2 but a 0.05% alter- ation in the transit midpoint, T0, could cause a large increase in χ2. DE-MC overcomes the problem with the scale of the step sizes by using the clustering of the chains around the global solution after the 'burn in': the difference vector between two randomly selected chains will contain the individual scale for each parameter (e.g. 0.05% for i and 0.00001% for T0). Ter Braak (2006) argues that DE-MC is a single N-chain that is simply a single random walk Markov Chain in a N × d dimensional space. The use of DE-MC in the exoplanet community is increas- ing, especially for models involving a large number of parameters. For example, models of transiting circumbinary planets can con- tain over 30 parameters (e.g. Doyle et al. 2011; Orosz et al. 2012; Welsh et al. 2012; Schwamb et al. 2013). To accurately estimate the parameter uncertainties the MCMC component of GEMC required 5 These tests were performed on a 2.4 GHz quad-core laptop. 4 Tregloan-Reed et al. Table 1. Original and recovered parameters from a simulated transit light curve using either 15 or 50 pixels for the planetary radius, plus the interval within which the best fit was searched for using GEMC. Parameter Symbol Original value Search interval Recovered value rp = 50 pixels Recovered value rp = 15 pixels Radius ratio Sum of fractional radii Linear LD coefficient Quadratic LD coefficient Orbital Inclination (degrees) Transit epoch (Phase) Longitude of spot (degrees) Co-latitude of spot (degrees) Spot angular radius (degrees) Spot contrast rp/r⋆ rs + rp u1 u2 i T0 θ φ rspot ρspot 0.15 0.25 0.3 0.2 85.0 0.015 30.0 65.0 12.0 0.8 0.05 to 0.30 0.10 to 0.50 0.0 to 1.0 0.0 to 1.0 70.0 to 90.0 -0.50 to 0.50 -90.0 to +90.0 0.0 to 90.0 0.0 to 30.0 0.0 to 1.0 0.1496 ± 0.0013 0.2486 ± 0.0024 0.291 ± 0.104 0.192 ± 0.042 85.16 ± 0.46 0.1498 ± 0.0011 0.2512 ± 0.0026 0.281 ± 0.114 0.189 ± 0.039 85.29 ± 0.44 0.01494 ± 0.00011 0.01502 ± 0.00010 30.50 ± 1.17 64.51 ± 5.83 12.73 ± 2.00 0.797 ± 0.057 30.47 ± 1.21 64.17 ± 5.55 12.33 ± 1.87 0.781 ± 0.061 106 function iterations (10 chains each of 105 steps). The DE-MC component requires approximately 2× 105 function iterations (128 chains each of 1500 steps (e.g. Welsh et al. 2012)). This equates to a five-fold reduction in the amount of computing time required to fit a transit light curve. When using a set of synthetic transit data it took GEMC approximately 5.4 days to fit the data using a planet radius of 50 pixels coupled with the MCMC component. The use of a planet radius of 15 pixels combined with the DE-MC algorithm resulted in GEMC taking only 2.7 hours to fit the same data. 2.1 Forward simulation of synthetic data The modifications to PRISM and GEMC were validated by mod- elling simulated transit data containing a starspot anomaly. For this test PRISM was used to create multiple simulated transits with a range of parameters. Noise was then added to the light curves so that the rms scatter between the original simulated light curves and the light curves with added noise was ≈ 500 ppm. Other levels of noise where also used in similar tests. This was to approximate a realistic level of noise found in transit light curves observed using the defocused photometry technique. Error bars were then assigned to each datapoint to give the original noise-free model a reduced chi squared value of χ2 ν = 1. Once a simulated transit light curve had been created, GEMC and PRISM were used in an attempt to recover the initial input pa- rameters. Different values for the planetary pixel radius were also used to test for numerical resolution. Table 1 shows the results for one of the tests using both rp = 50 and rp = 15 pixels, while Fig. 1 shows the simulated transit light curve together with the original and recovered models for the same test using rp = 50 pixels. From studying both Table 1 and Fig. 1, it can be seen that the recovered parameter values agree with the original values within their 1σ uncertainties. Interestingly, the rms scatter of the recov- ered model was found to be 499 ppm while the rms scatter of the original model is 511 ppm. This showed that GEMC not only ex- plored the large parameter search space but also scanned the local area around the global solution to find the best possible fit6 to the simulated data. This result is expected, and a testament to an opti- misation algorithm designed to find the lowest achievable χ2 ν (the recovered solution in this case had a χ2 ν = 0.94) in a given param- eter space. Similar results were found on all the simulation tests 6 This best fit is in fact a phantom solution generated by the addition of noise. Figure 1. Recovered and original models to simulated transit data created by PRISM and recovered by GEMC and PRISM. The residuals are shown at the bottom. The model was calculated with rp = 50 pixels. and show that both GEMC and PRISM are capable of accurately and precisely determining the properties of transit light curves. The recovered parameter values from setting rp = 50 and rp = 15 pixels also agree within their 1σ uncertainties (see Ta- ble 1). The scale of the 1σ uncertainties for when rp = 15 are com- parable in scale to that of the 1σ uncertainties for when rp = 50. This indicates that using a smaller number of pixels for the plane- tary radius (this reduction depends on the number of datapoints and the overall scale of the system being modelled) has little effect on the numerical resolution of the determined parameters or their as- sociated uncertainties. However, using a smaller number of pixels for the planetary radius does affect the smoothness of the plotted best-fit model. It is therefore advisable that, once the best-fitting parameters have been found, GEMC is used again with the param- eters fixed at the best-fitting values and with rp = 50 to calculate a smooth best-fitting model. These tests showed that it is possible to obtain precise results and correctly estimated parameter uncer- tainties, whilst, using a planetary pixel radius of less than 50. There are, though, some values which should not be used. For example in tests using rp = 5 the parameter uncertainties were heavily un- c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 derestimated, due to numerical noise in the model. By making the planet only 10 pixels across, the numerical resolution decreases to the point where adverse effects can be seen in the results and un- certainties. 3 OBSERVATIONS AND DATA REDUCTION Four transits of WASP-6 were observed on 2009/06/26, 2009/08/02, 2009/08/29 and 2010/07/31 by the MiNDSTEp con- sortium (Dominik et al. 2010) using the Danish 1.54 m telescope at ESO's La Silla observatory in Chile. The instrument used was the DFOSC imager, operated with a Bessell R filter. In this setup the CCD covers a field of view of (13.7′)2 with a pixel scale of 0.39′′ pixel−1. The images were unbinned but windowed for faster readout, resulting in a dead time between consecutive im- ages of between 22 and 35 s. The exposure times were 80 -- 120 s. The Moon's brightness and distance to the target star is given in Table 2. The telescope was defocused and autoguiding was main- tained through all observations. The amount of defocus applied caused the resulting PSFs to have a diameter of 86 pixels for the night of 2009/06/26, 32 pixels for the night of 2009/08/02, 44 pixels for the night of 2009/08/29 and 37 pixels for the night of 2010/07/31. We reduced the data in an identical fashion to Southworth et al. (2009a,b). In short, aperture photometry was performed with an IDL implementation of DAOPHOT (Stetson 1987), and the aper- ture sizes were adjusted to obtain the best results (see Table 2). A first order polynomial was then fitted to the outside-transit data whilst simultaneously optimising the weights of the comparison stars. The resulting data have scatters ranging from 0.591 to 1.215 mmag per point versus a transit fit using PRISM. The timestamps from the fits files were converted to BJD/TDB. An observing log is given in Table 2 and the final light curves are plotted in Fig.2. 4 DATA ANALYSIS All four transits were modelled using PRISM and GEMC. To do this a large parameter search space was selected to allow the global best fit solution to be found. As discussed in Tregloan-Reed et al. (2013), the ability of GEMC to find the global minimum in a short amount of computing time meant that it was possible to search a large area of parameter space to avoid the possibility of missing the best solution. The parameter search ranges used in analysing the WASP-6 datasets are given in Table 3. We modelled the two datasets containing a starspot anomaly independently, in order to obtain two sets of starspot parameters. This helps the investigation of whether the two anomalies are due to the same starspot (see Sec- tion 5). The separate models of the four datasets of WASP-6 have pa- rameters which are within 1σ of each other (Table 3). Ballerini et al. (2012) noted that starspots can affect the LD coefficients by up to 10% in the R band. This is not seen in the WASP-6 data, unlike in the transit data of WASP-19 (Tregloan-Reed et al. 2013). The scatter around the weighted mean is χ2 ν = 0.149 for the linear co- efficient and 0.355 for the quadratic coefficient. The error bars on the LD coefficients are too large to allow the effects of starspots to be detected. This is due to the lower quality of the data compared to WASP-19 (Tregloan-Reed et al. 2013). The combined best-fit LD coefficients are also in agreement within their 1σ uncertainties with c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 Transits and starspots of WASP-6 5 Figure 2. The four light curves of WASP-6 presented in this work, in the order presented in Table 2. Times are given relative to the midpoint of each transit. the theoretically predicted values for WASP-6 A of u1 = 0.4125 and u2 = 0.2773 (Claret 2000). 4.1 Photometric results The final photometric parameters for the WASP-6 system are given in Table 4 and are weighted means together with their 1σ uncertain- ties of the results from the four individual fits. Fig. 3 compares the light curves to the best-fitting models, including the residuals. The available times of mid-transit for WASP-6 were collected from the literature (Gillon et al. 2009; Dragomir et al. 2011; Sada et al. 2012; Jord´an et al. 2013; Nikolov et al. 2014). All timings were converted to the BJD/TDB timescale and used to obtain an improved orbital ephemeris: T0 = BJD/TDB 2 454 425.02180(11) + 3.36100208(31) × E where E represents the cycle count with respect to the reference 6 Tregloan-Reed et al. Table 2. Log of the observations presented for WASP-6. Nobs is the number of observations. 'Moon illum.' and 'Moon dist.' are the fractional illumination of the Moon, and its angular distance from WASP-6 in degrees, at the midpoint of the transit. Date Start time 2009/06/26 2009/08/02 2009/08/29 2010/07/31 (UT) 06:33 04:18 02:32 03:51 End time Nobs (UT) Exposure time (s) Filter Airmass Moon Moon illum. dist. Aperture sizes (px) Scatter (mmag) 10:43 10:31 07:47 10:20 91 175 129 193 120 90 -- 120 120 80 Bessell R 1.32 → 1.05 Bessell R 1.28 → 1.44 Bessell R 1.28 → 1.20 Bessell R 1.45 → 1.34 0.271 0.934 0.750 0.686 160.5 59.6 63.8 42.4 65, 90, 110 27, 40, 70 28, 40, 60 25, 35, 55 1.215 0.939 0.598 0.591 Table 3. Derived photometric parameters from each light curve, plus the interval within which the best fit was searched for using GEMC. Parameter Search interval 2009/06/26 2009/08/02 2009/08/29 2010/07/31 Radius ratio Sum of fractional radii Linear LD coefficient Quadratic LD coefficient Orbital Inclination (degrees) Transit epoch (BJD/TDB) Longitude of spot (degrees) Co-latitude of Spot (degrees) Spot angular radius (degrees) Spot contrast 0.05 to 0.30 0.10 to 0.50 0.0 to 1.0 0.0 to 1.0 70.0 to 90.0 ±0.5 in phase -90 to +90 0.0 to 90.0 0.0 to 30.0 0.0 to 1.0 0.1443 ± 0.0055 0.1102 ± 0.0060 0.366 ± 0.119 0.245 ± 0.191 88.47 ± 0.99 0.1444 ± 0.0043 0.1109 ± 0.0048 0.397 ± 0.116 0.325 ± 0.222 88.55 ± 0.85 0.1474 ± 0.0017 0.1115 ± 0.0025 0.368 ± 0.077 0.186 ± 0.123 88.33 ± 0.48 0.1454 ± 0.0021 0.1114 ± 0.0023 0.402 ± 0.067 0.192 ± 0.134 88.36 ± 0.53 2455009.83622 ± 0.00021 2455046.80720 ± 0.00015 2455073.69529 ± 0.00013 2455409.79541 ± 0.00010 -26.15 ± 1.52 78.76 ± 1.58 12.25 ± 1.40 0.649 ± 0.187 21.30 ± 0.99 72.77 ± 1.12 12.17 ± 0.81 0.798 ± 0.082 Table 4. Combined system and spot parameters for WASP-6. The system parameters are the weighted means from all four data sets. The spot angular size and contrast are the weighted means from the two transits containing a starspot anomaly. Table 5. Times of minimum light of WASP-6 and their residuals versus the ephemeris derived in this work. References: (1) Gillon et al. (2009); (2) This work; (3) Dragomir et al. (2011); (4) Jord´an et al. (2013); (5) Sada et al. (2012); (6) Nikolov et al. (2014) Parameter Radius ratio Sum of fractional radii Linear LD coefficient Quadratic LD coefficient Orbital Inclination (degrees) Spot angular radius (degrees) Spot contrast Stellar rotation period (d) Projected spin orbit alignment (degrees) Symbol rp/r⋆ rs + rp u1 u2 i rspot ρspot Prot λ Value 0.1463 ± 0.0012 0.1113 ± 0.0015 0.386 ± 0.043 0.214 ± 0.077 88.38 ± 0.31 12.19 ± 0.70 0.774 ± 0.075 23.80 ± 0.15 7.2 ± 3.7 epoch and the bracketed quantities represent the uncertainty in the final two digits of the preceding number. Fig. 4 and Table 5 show the residuals of these times against the ephemeris. The results show no evidence for transit timing variations. Initially we used the quoted mid-transit time from Gillon et al. (2009), but found that this value disagreed with the other 10 mid- transit times at the 2.2σ level. This may be because the value found by Gillon et al. (2009) was derived by simultaneously fitting the original WASP data plus two incomplete transits from RISE and a single complete transit from the FTS. We therefore used the same approach as Nikolov et al. (2014) and fitted (using PRISM) the archival FTS light curve to determine the mid-transit time. The value found using just the FTS data is in better agreement (0.6σ) with the other ten mid-transit times. Therefore it was decided to use the mid-transit time from the FTS light curve in our analysis, not just due to the better agreement but also due to the fact that it comes directly from a light curve covering a full transit. Time of minimum (BJD/TDB − 2400000) 54425.02167 ± 0.00022 55009.83622 ± 0.00021 55046.80720 ± 0.00015 55073.69529 ± 0.00013 55409.79541 ± 0.00010 55446.76621 ± 0.00058 55473.65438 ± 0.00016 55846.72540 ± 0.00045 56088.71800 ± 0.00013 56095.43973 ± 0.00017 56132.41081 ± 0.00010 Cycle no. 0.0 174.0 185.0 193.0 293.0 304.0 312.0 423.0 495.0 497.0 508.0 Residual (BJD) -0.00013 0.00006 0.00001 0.00008 -0.00000 -0.00023 -0.00007 -0.00028 0.00017 -0.00011 -0.00005 Reference 1 2 2 2 2 3 4 5 6 6 6 4.2 Physical properties of the WASP-6 system With the photometric properties of WASP-6 measured the physi- cal characteristics could be determined. The analysis followed the method of Southworth (2009), which uses the parameters measured from the light curves and spectra, plus tabulated predictions of the- oretical models. We adopted the values of i, rp/r⋆ and r⋆ + rp from Table 4, the orbital velocity amplitude K⋆ = 74.3+1.7 −1.4 m s−1 and eccentricity e = 0.054+0.018 −0.015 from Gillon et al. (2009), and the stellar effective temperature Teff = 5375 ± 65 K and metal abundance (cid:2) Fe An initial value of the velocity amplitude of the planet, Kp, was used to calculate the physical properties of the system using standard formulae and the physical constants listed by Southworth H (cid:3) = −0.15 ± 0.09 from Doyle et al. (2013). c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 Transits and starspots of WASP-6 7 Table 6. Physical properties of the WASP-6 system. Where two errorbars are given, the first is the statistical uncertainty and the second is the system- atic uncertainty. Parameter MA ( M⊙) RA ( R⊙) log gA (cgs) ρA ( ρ⊙) Mb ( MJup) Rb ( RJup) gb ( m s−2) ρb ( ρJup) eq (K) T ′ Θ a (AU) Age (Gyr) Value 0.836 ± 0.063 ± 0.024 0.864 ± 0.024 ± 0.008 4.487 ± 0.017 ± 0.004 1.296 ± 0.053 0.485 ± 0.027 ± 0.009 1.230 ± 0.035 ± 0.012 7.96 ± 0.30 0.244 ± 0.014 ± 0.002 1184 ± 16 0.0390 ± 0.0014 ± 0.0004 0.0414 ± 0.0010 ± 0.0004 9.0 +8.0 −12.7 +4.0 −9.0 depend on the theoretical models. The final statistical errorbar for each parameter is the largest of the individual ones from the solu- tions using each of the five different stellar models. The systematic errorbar is the largest difference between the mean and the individ- ual values of the parameter from the five solutions. 5 STARSPOT ANOMALIES Two of the light curves, from 2009/08/02 and 2009/08/29, contain apparent starspot anomalies (see Fig. 2). Due to a 27 day gap be- tween the two light curves it is not possible to conclusively demon- strate that the anomalies are due to the same spot. But if so, the stellar rotation period and sky-projected spin orbit alignment can be calculated and compared to the values found by Gillon et al. (2009), Doyle et al. (2013) and Nikolov et al. (2014). This will al- low an indirect check on whether the two spot anomalies are due to the same starspot. Firstly we consider whether the spot could last for a 27 day period. On the Sun a spot's lifetime T is proportional to its size A0 following the Gnevyshev-Waldmeier (G-W) Relation (Gnevyshev 1938; Waldmeier 1955). A0 = W T (2) where A0 is measured in MSH (micro-Solar hemispheres) and T is in days. Petrovay & van Driel-Gesztelyi (1997) state that W = 10.89 ± 0.18 MSH day−1. Henwood et al. (2010) showed that large sunspots also followed the Gnevyshev-Waldmeier rela- tionship. If the same relationship is applied to starspots then a min- imum lifetime of 30 days requires a minimum size of 327 MSH, or an angular radius of just greater than 1◦. Bradshaw & Hartigan (2014) argues that the standard solar G-W relation overestimates the lifetime of a starspot. Bradshaw & Hartigan (2014) uses tur- bulent magnetic diffusivity at supergranule size scales to calculate the magnetic diffusivity which in turn allows W in the G-W rela- tion to be re-calculated. Depending on the turbulent scale length being used and to have a minimum lifetime of 30 days requires a angular radius of 3◦ to 9◦ (see fig. 1 Bradshaw & Hartigan 2014). The sizes of the starspot anomalies in the WASP-6 light curves are greater than 10◦, so we conclude that a single spot can last suffi- ciently long to cause both anomalies, irrespective of the turbulent scale length used. Figure 3. Transit light curves and the best-fitting models of WASP-6. The residuals are displayed at the base of the figure. (2011). The mass and (cid:2) Fe H (cid:3) of the star were then used to obtain the expected Teff and radius, by interpolation within a set of tabulated predictions from theoretical stellar models. Kp was iteratively re- fined until the best agreement was found between the observed and expected Teff, and the measured r⋆ and expected R⋆ a . This was per- formed for ages ranging from the zero-age to the terminal-age main sequence, in steps of 0.01 Gyr. The overall best fit was found, yield- ing estimates of the system parameters and the evolutionary age of the star. This procedure was performed separately using five different sets of stellar theoretical models (see Southworth 2010), and the spread of values for each output parameter was used to assign a systematic error. Statistical errors were propagated using a pertur- bation algorithm (see Southworth 2010). The final results of this process are in reasonable agreement with themselves and with published results for WASP-6. The fi- nal physical properties are given in Table 6 and incorporate sepa- rate statistical and systematic errorbars for those parameters which c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 8 Tregloan-Reed et al. Figure 4. Residuals of the available times of mid-transit versus the orbital ephemeris found for WASP-6. The four timings from this work are the cluster of three points between the cycle numbers 170 -- 200 and the point close to cycle 290. 5.1 Starspot anomalies results The results from modelling the two spot anomalies suggest that they are due to the same spot rotating around the surface of the star, as the spot sizes and contrasts are in good agreement and the lifetime of a spot this size is much greater than the time interval between the two spotted transits. Fig. 5 is a representation of the stellar disc, the spot and the transit chord for the two nights of ob- servations. By assuming that the two spot anomalies are indeed caused by the same spot, it is straightforward to calculate the sky-projected spin orbit alignment of the system. We find a value of λ = 7.2◦ ± 3.7◦ from the measured positions of the starspot during the two transits. It is also possible to calculate the rotational period of the star, using the spot positions and an estimate of the number of stellar ro- tations which occurred between the two transits (see Tregloan-Reed et al. 2013; Mancini et al. 2014). Due to the 27 day gap between the light curves the star could have rotated N full rotations plus 47.5◦ ± 2.5◦. If N = 0 then this would imply that WASP-6 has a rotation period of approximately 200 days, which is extremely long for a main sequence G-star. If N = 1 then the spot has travelled 407.5◦ ± 2.5◦ between the transits, giving a rotational period of Prot = 23.80 ± 0.15 d at a co-latitude of 75.8◦. This is in excel- lent agreement with the measurement of Prot = 23.6 ± 0.5 d from Nikolov et al. (2014). Combining this with the stellar radius (see Table 6), the latitudinal rotational velocity of the star was calcu- lated to be v(75.8◦ ) = 1.78±0.20 km s−1. This is also in agreement with v sin I from both Gillon et al. (2009) and Doyle et al. (2013). If N = 2 then the spot has travelled 767.5◦ ± 2.5◦, giving a rota- tional period of Prot = 12.63 ± 0.15 d at a co-latitude of 75.8◦ (or v(75.8◦ ) = 3.36 ± 0.20 km s−1). This agrees with the v sin I from Gillon et al. (2009) and Doyle et al. (2013), but not with the Prot from Nikolov et al. (2014). The agreement with Gillon et al. (2009) and Doyle et al. (2013) is due to the fact that any value of v that is found to be greater than v sin I can be considered to agree based on the nature of sin I. We conclude that the N = 1 case is much more likely than the two alternatives discussed above. 5.2 Degeneracy of the stellar rotation period Whilst there is no clear photometric signal in the SuperWASP light curve of WASP-6, Nikolov et al. (2014) were able to mea- sure a rotation period of Prot = 23.6 ± 0.5 d from photometry of higher precision, however, none of the STIS observations de- tected a starspot anomaly indicating that starspots on WASP-6 A are either rare or of low contrast. This is also supported by the Figure 5. Representation of the stellar disc, starspot, transit chord and equa- tor for the two datasets of WASP-6 containing spot anomalies. The axis of stellar rotation lies in the plane of the page and in the case of λ = 0◦ points upwards. upper limit of the photometric variability of about 1% (Nikolov et al. 2014). There are also two measurements of v sin I from Gillon et al. (2009) (v sin I = 1.4 ± 1.0 km s−1) and Doyle et al. (2013) (v sin I = 2.4 ± 0.5 km s−1). Both v sin I measurements agree with the v found when combining Prot and R⋆ at a co- latitude of 75.8◦ to give either v(75.8◦ ) = 1.78 ± 0.20 km s−1 or v(75.8◦ ) = 3.36 ± 0.20 km s−1. The problem that arises from checking measurements of v against v sin I is that due to the sin I projection factor any value for v that is found to be greater than v sin I can be considered to agree. A second unknown is the amount of differential rotation that is experienced by WASP-6 A. In the absence of any differential rotation the single full rotation value of Prot = 23.80± 0.15 d would lead to an equatorial rotational ve- locity of v = 1.84± 0.20 km s−1. This result agrees again with the v sin I value from both Gillon et al. (2009) and Doyle et al. (2013). Our results from PRISM do show though that the two starspot posi- tions are only approximately 10◦ from the stellar equator. As such the effect from differential rotation would be small, so any large divergence of v from v sin I would imply that I ≪ 90◦. WASP-6 A has Teff = 5375 ± 65 K (Doyle et al. 2013) so is a cool star (Teff < 6250 K). The trend seen between host star Teffs and projected orbital obliquity (see Fig. 6) suggests that the orbital rotation axis of WASP-6 b should be aligned with the stellar rotation axis of WASP-6 A. For this to be true then I would have to be ≈ 90◦, and thus sin I ≈ 1. If this is the case then the value v(75.8◦ ) = 3.36 ± 0.20 km s−1 no longer agrees with the v sin I from either Gillon et al. (2009) or Doyle et al. (2013). This supports the supposition that the rotation period of WASP-6 A is Prot = 23.80 ± 0.15 d. Brown (2014) calculated the stellar rotation period of WASP-6 A to be Prot = 27.1+3.6 −3.8 d from Gaussian distribution c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 sampling of v sin I, i and RA, which is further evidence for the conclusion that the rotation period of WASP-6 A is Prot = 23.80± 0.15 d. 6 DISCUSSION AND CONCLUSIONS We have determined the physical properties of the WASP-6 plane- tary system (Table 6) based on four new high-precision transit light curves, finding values which are consistent with and more precise than those in the literature. We find the mass and radius of the host star to be 0.836 ± 0.063 M⊙ and 0.864 ± 0.024 R⊙, respectively. For the planet we find a mass of 0.485 ± 0.027 MJup, a radius of 1.230± 0.035 RJup and a density of 0.244± 0.014 ρJup. These re- sults also serve as a secondary check for the accuracy of the PRISM and GEMC codes. By studying the individual results for each of the four transits (see Table 3) it can be seen that the system parame- ters from each light curve agree within their 1σ uncertainties. This shows that PRISM can retrieve reliable photometric properties from transit light curves containing starspot anomalies. The four transits of WASP-6 were modelled using PRISM and GEMC. Two of the transits contained a starspot anomaly but are separated by 27 days. Whilst it is not possible to prove that the two spot anomalies are caused by the same starspot, the available evi- dence strongly favours this scenario. The results from PRISM show that the angular size and contrast of the starspot in both light curves agree to within 0.05σ and 0.73σ, respectively. As with WASP-19 (see Tregloan-Reed et al. 2013), only part of the starspot(s) is on the transit chord (Fig. 5). Because the light curve only holds infor- mation on what is happening inside the transit chord then a likely scenario is that the planet is passing over a band of smaller starspots which form an active region on WASP-6. In this active region, there could be a number of starspots each with sizes much less than 1◦ and therefore lifetimes shorter than 30 days (see Section 5). Future observations may allow changes to be seen in the overall contrast from the starspot region. In either case as a whole the region would remain a similar size and shape over a 27 day period. In the case of a single large starspot, rspot (Table 4) and R⋆ (Table 6) can be combined to find the starspot radius. We find Rspot = 127902 ± 11102 km, which equates to approximately 4.5% of the visible stellar surface. This value is similar to starspots found on other G-type stars (Strassmeier 2009). If the two starspot anomalies are assumed to be generated by the planet crossing the same starspot then it is possible to calculate the latitudinal rotation period of WASP-6. It was found that either Prot = 23.80± 0.15 d or Prot = 12.63± 0.15 d at a co-latitude of 75.8◦. These calculations assumed that WASP-6 had made either one or two full rotations prior to the difference seen in the light curves. Even without knowing the number of full rotations that WASP-6 completed between the two spotted light curves, if the starspot anomalies are due to the same spot then the sky-projected spin orbit alignment λ of the system can be measured. We find λ = 7.2◦ ± 3.7◦. This result agrees with, and is more precise than, the previous measurement of λ using the RM effect (λ = 11◦ +14 −18; Gillon et al. 2009). λ gives the lower boundary of the true spin-orbit angle, ψ. As stated by Fabrycky & Winn (2009), finding a small value for λ can be interpreted in different ways. Either ψ lies close to λ and the system is aligned, or ψ lies far from λ and the system is not aligned. As discussed in Section 5.1 because the spot is close to the stellar equator then it could be assumed that the change in v at the equator due to differential rotation would be small. Coupled c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 Transits and starspots of WASP-6 9 Figure 6. λ against Teff for 83 transiting planets from TEPCat including WASP-19 and WASP-6. The green and red datapoints are WASP-6 (left) and WASP-19 (right). The green datapoints represent values from the lit- erature (WASP-6: Gillon et al. 2009; WASP-19: Hellier et al. 2011) and the red datapoints represent the values found from this work and Tregloan- Reed et al. (2013). The trend in the data suggests that cool host stars harbour aligned systems. with the uncertainties measured in v sin I from both Gillon et al. (2009) and Doyle et al. (2013) it is plausible that sin I ≈ 1 and therefore ψ ≈ 7◦ if Prot = 23.80 ± 0.15 d. As a consequence we have two different scenarios: an aligned system with a slowly rotat- ing star or a misaligned system with a rapidly rotating star. Taking into account the Teff of WASP-6 A and the statistical trend seen in misaligned systems it is more probable that the WASP-6 system is in fact aligned, suggesting ψ ≈ 7◦ and Prot = 23.80 ± 0.15 d. It would be desirable to observe consecutive transits of WASP-6 in an attempt to definitively identify multiple planetary crossings of a single starspot and to precisely determine Prot, λ and potentially ψ of WASP-6. If the starspot anomalies are due to the same starspot, λ = 7.2◦ ± 3.7◦ and there is no direct evidence for a spin-orbit mis- alignment in the WASP-6 system. With potentially a low obliquity and a cool host star, WASP-6 seems to follow the idea put forward by Winn et al. (2010a) that planetary systems with cool stars will have a low obliquity. It also lends weight to the idea that WASP-6 b formed at a much greater distance from its host star and suffered or- bital decay through tidal interactions with the protoplanetary disc (i.e. either Type I or Type II disc-migration, Ward 1997). At present there are 83 transiting planets with published λ 10 Tregloan-Reed et al. values7. The λ values for WASP-6 (this work) and WASP-19 (Tregloan-Reed et al. 2013) were updated and a plot of λ against Teff was created (see Fig. 6). To remove any ambiguity in the plot due to negative values of λ, we plot its absolute value. It can be seen that a large proportion (75 %) of cool stars (Teff < 6250 K) are in aligned systems, while the majority (56 %) of hot host stars have misaligned systems. This trend supports Winn et al. (2010a) in that cool stars with hot Jupiters will have low obliquities. This trend can also be explained by the time required for the system to align. Hot stars will have thinner convective zones and will therefore take longer to align the photosphere with the planetary orbit. Because of this, by examining λ of hot stars a greater proportion will have misaligned systems compared to cool stars where the alignment process is much shorter and so will have a higher proportion of aligned systems. Cool stars also live longer so the ones that are ob- served are on average older. They have therefore had more time for tidal effects to work (Triaud 2011). By determining λ and ψ of the planetary system it is possible to begin to understand the primary process in the dynamical evolu- tion of the system. The RM effect can be used to ascertain a value for λ. One limitation of this method though is from an excess RV jitter (stellar activity e.g. starspots). Therefore, the use of the RM effect either requires magnetically quiet stars or the transit chord of the planet to bypass any active latitudes on the stellar disc. The opposite is true when using starspot anomalies in light curves to de- termine λ. Due to this the two different methods complement each other in probing the dominant process in the dynamical evolution of transiting planets. It should be noted that in both the cases of WASP-19 and WASP-6 (see Fig. 6) the measured uncertainty in λ is much smaller than measured using the RM effect. This indicates that the starspot method to measure λ is superior to the RM effect in terms of reduced uncertainty in measuring λ. However, as was shown in observing WASP-50 (see Tregloan-Reed & Southworth 2013), the starspot method does not always work in terms of ob- taining transit light curves affected by a starspot anomaly. The RM effect does have a high success rate in measuring a value of λ but rarely achieves a similar precision. 7 ACKNOWLEDGEMENTS We like to thank the anonymous referee for the helpful comments on the manuscript. The operation of the Danish 1.54 m telescope at ESOs La Silla observatory is financed by a grant to UGJ from The Danish Council for Independent Research (FNU). Research at the Armagh Observatory is funded by the Department of Culture, Arts & Leisure (DCAL). JTR acknowledges financial support from STFC in the form of a Ph.D. Studentship (the majority of this work) and also acknowledges financial support from ORAU (Oak Ridge Associated Universities) and NASA in the form of a Post-Doctoral Programme (NPP) Fellowship. JS acknowledges financial support from STFC in the form of an Advanced Fellowship. DR acknowl- edges financial support from the Spanish Ministry of Economy and Competitiveness (MINECO) under the 2011 Severo Ochoa Pro- gram MINECO SEV-2011-0187. FF, DR (boursier FRIA) and J Surdej acknowledge support from the Communaut´e franc¸aise de 7 All measured λ and Teff values of the known planetary systems were obtained from the 2014/10/20 version of the TEPCat catalogue (Southworth 2011). (http://www.astro.keele.ac.uk/∼jkt/tepcat/) Belgique -- Actions de recherche concert´ees -- Acad´emie Wallonie -- Europe. REFERENCES Baglin, A., et al., 2006, in 36th COSPAR Scientific Assembly, vol. 36 of COSPAR Meeting, p. 3749 Bakos, G., Noyes, R. W., Kov´acs, G., Stanek, K. Z., Sasselov, D. D., Domsa, I., 2004, The Publications of the Astronomical Society of the Pacific, 116, 266 Ballerini, P., Micela, G., Lanza, A. F., Pagano, I., 2012, A&A, 539, A140 Baranne, A., et al., 1996, A&AS, 119, 373 Barros, S. C. C., Bou´e, G., Gibson, N. P., Pollacco, D. L., San- terne, A., Keenan, F. P., Skillen, I., Street, R. A., 2013, MNRAS, 430, 3032 B´eky, B., Kipping, D. M., Holman, M. J., 2014, MNRAS, 442, 3686 Borucki, W. J., et al., 2010, Science, 327, 977 Borucki, W. J., et al., 2012, ApJ, 745, 120 Borucki, W. J., et al., 2013, Science, 340, 587 Bradshaw, S. J., Hartigan, P., 2014, ApJ, 795, 79 Brown, D. J. A., 2014, MNRAS, 442, 1844 Butler, R. P., Marcy, G. W., Williams, E., McCarthy, C., Dosanjh, P., Vogt, S. S., 1996, The Publications of the Astronomical Soci- ety of the Pacific, 108, 500 Butler, R. P., Marcy, G. W., Fischer, D. A., Brown, T. M., Contos, A. R., Korzennik, S. G., Nisenson, P., Noyes, R. W., 1999, ApJ, 526, 916 Carter, J. A., Winn, J. N., Holman, M. J., Fabrycky, D., Berta, Z. K., Burke, C. J., Nutzman, P., 2011, ApJ, 730, 82 Claret, A., 2000, A&A, 363, 1081 D´esert, J.-M., et al., 2011, ApJS, 197, 14 Dominik, M., et al., 2010, Astronomische Nachrichten, 331, 671 Doyle, A. P., et al., 2013, MNRAS, 428, 3164 Doyle, L. R., et al., 2011, Science, 333, 1602 Dragomir, D., et al., 2011, ApJ, 142, 115 Fabrycky, D. C., Winn, J. N., 2009, ApJ, 696, 1230 Gelman, A., Rubin, R., 1992, Statistical Science, 7, 457 Gillon, M., et al., 2009, A&A, 501, 785 Gnevyshev, M. N., 1938, Izvestiya Glavnoj Astronomicheskoj Observatorii v Pulkove, 16, 36 Hellier, C., Anderson, D. R., Collier-Cameron, A., Miller, G. R. M., Queloz, D., Smalley, B., Southworth, J., Triaud, A. H. M. J., 2011, ApJ, 730, L31 Henwood, R., Chapman, S. C., Willis, D. M., 2010, Solar Physics, 262, 299 Jord´an, A., et al., 2013, ApJ, 778, 184 Kipping, D. M., Dunn, W. R., Jasinski, J. M., Manthri, V. P., 2012, MNRAS, 421, 1166 Maciejewski, G., Raetz, S., Nettelmann, N., Seeliger, M., Adam, C., Nowak, G., Neuhauser, R., 2011, A&A, 535, A7 Mancini, L., et al., 2013, MNRAS, 436, 2 Mancini, L., et al., 2014, MNRAS, 443, 2391 Mayor, M., et al., 2003, The Messenger, 114, 20 Mohler-Fischer, M., et al., 2013, A&A, 558, A55 Nikolov, N., et al., 2014, ArXiv: 1411.4567 Nutzman, P. A., Fabrycky, D. C., Fortney, J. J., 2011, ApJL, 740, L10 Orosz, J. A., et al., 2012, Science, 337, 1511 c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000 Transits and starspots of WASP-6 11 Petrovay, K., van Driel-Gesztelyi, L., 1997, Solar Physics, 176, 249 Pollacco, D. L., et al., 2006, The Publications of the Astronomical Society of the Pacific, 118, 1407 Pont, F., et al., 2007, A&A, 476, 1347 Price, K., Storn, R., 1997, Dr Dobb's Journal, 264, 18 Queloz, D., et al., 2000, A&A, 354, 99 Rabus, M., et al., 2009, A&A, 494, 391 Sada, P. V., et al., 2012, The Publications of the Astronomical Society of the Pacific, 124, pp. 212 Sanchis-Ojeda, R., Winn, J. N., 2011, ApJ, 743, 61 Sanchis-Ojeda, R., Winn, J. N., Holman, M. J., Carter, J. A., Osip, D. J., Fuentes, C. I., 2011, ApJ, 733, 127 Schneider, J., Dedieu, C., Le Sidaner, P., Savalle, R., Zolotukhin, I., 2011, A&A, 532, A79 Schwamb, M. E., et al., 2013, ApJ, 768, 127 Silva, A. V. R., 2003, ApJ, 585, L147 Silva-Valio, A., Lanza, A. F., Alonso, R., Barge, P., 2010, A&A, 510, A25 Southworth, J., 2009, MNRAS, 394, 272 Southworth, J., 2010, MNRAS, 408, 1689 Southworth, J., 2011, MNRAS, 417, 2166 Southworth, J., et al., 2009a, MNRAS, 396, 1023 Southworth, J., et al., 2009b, MNRAS, 399, 287 Steele, I. A., Bates, S. D., Gibson, N., Keenan, F., Meaburn, J., Mottram, C. J., Pollacco, D., Todd, I., 2008, in Society of Photo- Optical Instrumentation Engineers (SPIE) Conference Series, vol. 7014 of Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, p. 6 Stetson, P. B., 1987, PASP, 99, 191 Storn, R., Price, K., 1997, Journal of Global Optimisation, 11, 341 Strassmeier, K. G., 2009, A&A Rev, 17, 251 Ter Braak, C. J. F., 2006, Statistics and Computing, 16, 239 Tregloan-Reed, J., Southworth, J., 2013, MNRAS, 431, 966 Tregloan-Reed, J., Southworth, J., Tappert, C., 2013, MNRAS, 428, 3671 Triaud, A. H. M. J., 2011, A&A, 534, L6 Waldmeier, M., 1955, Ergebnisse und Probleme der Sonnen- forschung., Leipzig, Geest & Portig, 1955. 2. erweiterte Aufl. Ward, W. R., 1997, icarus, 126, 261 Welsh, W. F., et al., 2012, Nature, 481, 475 Winn, J. N., Fabrycky, D., Albrecht, S., Johnson, J. A., 2010a, ApJ, 718, L145 Winn, J. N., et al., 2010b, ApJ, 723, L223 c(cid:13) 0000 RAS, MNRAS 000, 000 -- 000
1004.3171	1	1004	2010-04-19T12:16:46	Capture of dark matter by the Solar System. Simple estimates	[ "astro-ph.EP", "astro-ph.GA", "hep-ph" ]	We consider the capture of galactic dark matter by the Solar System, due to the gravitational three-body interaction of the Sun, a planet, and a dark matter particle. Simple estimates are presented for the capture cross-section, as well as for density and velocity distribution of captured dark matter particles close to the Earth.	astro-ph.EP	astro-ph	Capture of dark matter by the Solar System. Simple estimates I.B. Khriplovich1 Budker Institute of Nuclear Physics 630090 Novosibirsk, Russia, Abstract We consider the capture of galactic dark matter by the Solar System, due to the gravitational three-body interaction of the Sun, a planet, and a dark matter particle. Simple estimates are presented for the capture cross-section, as well as for density and velocity distribution of captured dark matter particles close to the Earth. 1 Introduction The density of dark matter (dm) in our Galaxy is (see, e.g., [1]) ρg ≃ 4 · 10−25 g/cm3 . (1) However, only upper limits on the level of 10−19 g/cm3 (see [2, 3]) are known for the density of dark matter particles (dmp) in the Solar System (SS). Besides, even these limits are derived under the quite strong assumption that the distribution of dm density in the SS is spherically- symmetric with respect to the Sun. Meanwhile, information on dm density in SS is very important, in particular for the experiments aimed at the detection of dark matter. The capture of dark matter by the SS was addressed previously in [4 -- 8]. In particular, in [8] the total mass of the captured dark matter was estimated analytically. In the present note the analytical estimates are given for the capture cross-section, as well as for the density and velocity distribution of captured dm close to the Earth. Of course, a particle cannot be captured by the Sun alone. The interaction with a planet is necessary for it, i.e. this is essentially a three-body (the Sun, planet and dmp) problem. Obvi- ously, the capture is dominated by the particles with orbits close to parabolic ones with respect to the Sun; besides, the distances between their perihelia and the Sun should be comparable with the radius of the planet orbit rp . Just the trajectories of these particles are most sensitive to the attractive perturbation by the planet. The capture can be effectively described by the so-called restricted three-body problem (see, e.g., [9]). In this approach the interaction between two heavy bodies (the Sun and a planet in our case) is treated exactly. As exactly is treated the motion of the third, light body (a dmp in our case) in the gravitational field of the two heavy ones. One neglects however the [email protected] 1 back reaction of a light particle upon the motion of the two heavy bodies. Obviously, this approximation is fully legitimate for our purpose. Still, the restricted three-body problem is rather complicated, its solution requires both subtle analytical treatment and serious numerical calculations (see, for instance, [10]). Under certain conditions the dynamics of light particle becomes chaotic. The "chaotic" effects are extremely important for the problem. However their quantitative investigation is quite com- plicated and remains beyond the scope of the present note. We confine here instead to simple estimates which could be also of a methodological interest by themselves. On the other hand, thus derived results for the total mass and density of the captured dark matter constitute at least an upper limit for their true value. As to the velocity distribution of dmp's given here, together with the mentioned result for the dark matter density, it could be possibly of some practical interest for planning the experimental searches for dm. 2 Total mass of dark matter captured by the Earth The Solar System is immersed in the halo of dark matter and moves together with it around the center of our Galaxy. To simplify the estimates, we assume that the Sun is at rest with respect to the halo. The dark matter particles in the halo are assumed also to have in the reference frame, comoving with the halo, the Maxwell distribution (see [11]): f (v) dv = r 54 π v2dv u3 exp(cid:18)− 3 2 v2 u2(cid:19) , (2) with the local rms velocity u ≃ 220 km/s. Let us note that the velocities v discussed in this section are the asymptotic ones, they refer to large distances from the Sun, so that their values start at v = 0 and formally extend to ∞. The amount of dm captured by the SS can be found by means of simple estimates2. The total mass captured by the Sun (its mass is M) together with a planet with the mass mp, during the lifetime of the SS, can be written as follows: T ≃ 4.5 · 109 years ≃ 1017 s µp = ρg T < v σ > ; (3) (4) here σ is the capture cross-section. The product σv is averaged over distribution (2); with all typical velocities in the SS much smaller than u, this distribution simplifies to f (v) dv = r 54 π v2dv u3 . (5) To estimate the average value < v σ >, we resort to dimensional arguments, supplemented by two rather obvious physical requirements: the masses mp and M of the two heavy components of our restricted three-body problem should enter the result symmetrically, and the mass of the dmp should not enter the result at all in virtue of the equivalence principle. Thus, we arrive at < v σ > ∼ √54π k2 mp M u3 , (6) 2These estimates were given previously in [8]. Here we repeat them, as well as results (8), (9), (19) (see below), since they are essential for the present discussions. 2 or Z ∞ 0 dv v3 σ ∼ π k2 mp M ; (7) here k is the Newton gravitation constant; an extra power of π, inserted into these expressions, is perhaps inherent in σ. Since the capture would be impossible if the planet were not bound to the Sun, it is only natural that the result is proportional to the corresponding effective "coupling constant" k mp M. One more power of k corresponds to the gravitational interaction of the dark matter particle. The final estimate for the captured mass is For the Earth it constitutes µp ∼ ρgT√54π k2 mp M/u3 . µE ∼ 4 · 1018 g . 3 Capture cross-section (8) (9) By the same dimensional reasons (and in the complete correspondence with formula (7)), the total capture cross-section for the Earth should look as follows: σ ∼ π k2 mE M /v4 , (10) where mE is the mass of the Earth, and v is some velocity which can be estimated as follows. It is natural to assume that the capture of dm particles occurs when they are close to the Earth, i.e. at the distances ∼ rE from the Sun. As natural are the following assumptions: 1) the initial velocities of the captured dmp's exceed only slightly the parabolic one vpar (v2 par = 2kM/rE); 2) their final velocities are only slightly less than vpar. To our accuracy, here we omit the factor of 2 in the definition of v2 E = kM/rE (vE = 30 km/s is the velocity of the Earth). Thus, the capture cross-section is par, and thus put v2 ∼ v2 This formula can be also conveniently rewritten as σ ∼ π k2 mE M /v4 E . σ ∼ π r2 E (mE/M) . (11) (12) Let us note here that the impact parameter corresponding to formula (12), i.e. the typical distance between a dmp and the Earth crucial for the capture, is rimp ∼ rE (mE/M)1/2 ≪ rE . (13) In fact, this impact parameter corresponds to the distance at which the attraction to the Earth equals the attraction to the Sun, i.e. where km/r2 > kM/r2 E (r ≪ rE) . (14) . Up to now, in all relevant formulae, (7), (10), (11), we dealt with the capture cross-section averaged over the directions of the dmp velocity v. However, this cross-section depends essen- tially on the mutual orientation of v and vE. Certainly, it is maximum when these velocities are parallel and as close as possible by modulus. Besides, the impact parameter rimp of the 3 collision is much less than the radius rE of the Earth orbit (see (13)), and thus within the distances ≃ rimp both the Earth and dmp trajectories can be treated as rectilinear. Therefore, it looks quite natural to identify v in (10) with the relative velocity v − vE of the dmp and the Earth, i.e. to generalize formula (11) as follows: dσ ∼ k2 mp M (v − vE)4 1 4 dΩ (factor 1/4 is introduced here for correspondence with factor π in (11): (1/4)R dΩ = π). Thus derived total cross-section is σ ∼ 1 4 Z dΩ k2 mp M (v − vE)4 = π k2 mp M E)2 . (v2 − v2 (15) (16) Clearly, it is the particles moving initially with the velocities only slightly above the parabolic one √2 vE = 42 km/s that are captured predominantly, and thus, with v = √2 vE, cross-sections (11) and (16) practically coincide. On the other hand, it follows from (16) that in the vicinity of the Earth the captured particles move with respect to it with the velocities close to (√2 − 1)vp ≃ 12 km/s. 4 Space distribution of captured dark matter The captured dmp's had initial trajectories predominantly close to parabolas focussed at the Sun, and the velocities of these dmp's change only slightly as a result of scattering. Therefore, their trajectories become elongate ellipses with large semimajor axes, still focussed at the Sun. The ratio of their maximum rmax and minimum rmin distances from the Sun is [12] = , rmax rmin 1 + e 1 − e where e is the eccentricity of the trajectory. In our case, as a result of the capture, the eccentricity changes from 1 + ε1 to 1 − ε2, where ε1,2 ≪ 1. This loss of eccentricity is due to the gravitational perturbation by the Earth, and therefore is proportional to mE. On the other hand, rmin is close to the radius rE of the Earth orbit. Thus, for dimensional reasons, we arrive at [8] (17) rmax ∼ rE (M/mE) . (18) Let us note here that the analogous estimate for the case of Jupiter complies qualitatively with the results of corresponding numerical calculations presented in [10]. Obviously, the semimajor axis admp of the trajectory of a captured dmp is on the same order of magnitude as rmax. Then, the time spent by a dmp, with the characteristic velocity close to vE and at the distance from the Sun close to rE, is comparable to the orbital period of the Earth TE = 1 year. Besides, the orbital period T is related to the semimajor axis a as follows [12]: T ∼ a3/2. Thus, we arrive at the following estimate for the orbital period of the captured dmp3: (19) In other words, the relative time spent by a dmp at the distances ∼ rE from the Earth can be estimated as (mE/M)3/2. Moreover, the typical distances from the Earth at which a dmp Tdmp ∼ TE (M/mE)3/2 . 3In the case of the Earth, this orbital period is huge, ∼ 108 years. Still, it is much less than the lifetime of the SS, ∼ 5 · 109 years. 4 can be captured, should be less than the impact parameter rimp ∼ rE (mE/M)1/2 (see (13)). Thus, the relative time spent by a dmp sufficiently close to the Earth to be captured, can be estimated as (mE/M)2. With the impact parameter (13), the corresponding volume V , centered at the Earth and crucial for the capture, can be estimated as 4π 3 r3 imp ∼ V ∼ 4π 3 E (mE/M)3/2 ≪ r3 4π 3 r3 E. (20) Let us combine formula (9) for the total captured mass with the effective volume (20) occupied by this mass and with the estimate (mE/M)2 for the relative time spent by a dmp within the impact parameter (13) with respect to the Earth. In this way we arrive at the following estimate for the density of dark matter, captured by the SS, in the vicinity of the Earth: ρE ∼ 5 · 10−25 g/cm3 . (21) This estimate practically coincides with the value (1) for the galactic dm density. In fact, the result (21) for the density of the captured dm, as well as the estimates (8) and (9) for its total mass, should be considered as upper limits only, since we have neglected therein the inverse process, that of the ejection of the captured dm from the SS. The characteristic time of the inverse process is not exactly clear now. Therefore, it cannot be excluded that it is comparable to, or even larger than, the lifetime T of the SS [8]. In this case our estimates are valid. If this is the case indeed, then the dm around the Earth consists essentially of two compo- nents with comparable densities. In line with the common component with the typical velocity around u ∼ 220 km/s, there is one more, with the velocity relative to the Earth >∼12 km/s. Acknowledgements. I am grateful to V.V. Sokolov for useful discussions. The work was supported by the Russian Foundation for Basic Research through Grant No. 08-02-00960-a. References [1] G. Bertone, D. Hooper, and J. Silk, Phys. Rep. 405, 279 (2005). [2] I.B. Khriplovich and E.V. Pitjeva, Int. J. Mod. Phys. D 15, 615 (2006). [3] I.B. Khriplovich, Int. J. Mod. Phys. D 16, 1475 (2007). [4] A. Gould and S.M.K. Alam, Astrophys. J. 549, 72 (2001). [5] J. Lundberg and J. Edsjo, Phys. Rev. D 69 (2004) 123505. [6] A.H.G. Peter and S. Tremaine, Proc. of Sc. IDM 2008; arXiv:0806.2133 (astro-ph) [7] A.H.G. Peter, Phys. Rev. D 79, 1003531, 1003532, 1003533 (2009). [8] I.B. Khriplovich and D.L. Shepelyansky, Int. J. Mod. Phys. D 18, 1903 (2009). [9] V. Szebehely, Theory of Orbits, Academic Press, N.Y. (1964). [10] T.Y. Petrosky, Phys. Lett. A 117, 328 (1986). [11] M.S. Alenazi and P. Gondolo, Phys. Rev. D 74, 083518 (2006). [12] L.D. Landau and E.M. Lifshitz, Mechanics, Nauka, Moscow (1988), §15. 5
0902.1640	1	0902	2009-02-10T12:21:35	A Study of the Accuracy of Mass-Radius Relationships for Silicate-Rich and Ice-Rich Planets up to 100 Earth Masses	[ "astro-ph.EP" ]	A mass-radius relationship is proposed for solid planets and solid cores ranging from 1 to 100 Earth-mass planets. It relies on the assumption that solid spheres are composed of iron and silicates, around which a variable amount of water is added. The M-R law has been set up assuming that the planetary composition is similar to the averaged composition for silicates and iron obtained from the major elements ratio of 94 stars hosting exoplanets. Except on Earth for which a tremendous amount of data is available, the composition of silicate mantles and metallic cores cannot be constrained. Similarly, thermal profiles are poorly known. In this work, the effect of compositional parameters and thermal profiles on radii estimates is quantified. It will be demonstrated that uncertainties related to composition and temperature are of second order compared to the effect of the water amount. The Super-Earths family includes four classes of planets: iron-rich, silicate-rich, water-rich, or with a thick atmosphere. For a given mass, the planetary radius increases significantly from the ironrich to the atmospheric-rich planet. Even if some overlaps are likely, M-R measurements could be accurate enough to ascertain the discovery of an earth-like planet .The present work describes how the amount of water can be assessed from M-R measurements. Such an estimate depends on several assumptions including i) the accuracy of the internal structure model and ii) the accuracy of mass and radius measurements. It is shown that if the mass and the radius are perfectly known, the standard deviation on the amount of water is about 4.5 %. This value increases rapidly with the radius uncertainty but does not strongly depend on the mass uncertainty.	astro-ph.EP	astro-ph	A study of the accuracy of Mass-Radius relationships for silicate-rich and ice-rich planets up to 100 Earth masses. Authors: Grasset O. 1, Schneider J. 2, Sotin C. 3 Address : 1. Laboratoire de Planétologie et Géodynamique, UMR-CNRS 6112, Université de Nantes, France. 2. Laboratoire Univers et Théories, UMR-CNRS 8102, Observatoire de Paris , France 3. Jet Propulsion Laboratory, California Technology Institute, Pasadena, CA. 1 Abstract A mass-radius relationship is proposed for solid planets and solid cores ranging from 1 to 100 Earth-mass planets. It relies on the assumption that solid spheres are composed of iron and silicates, around which a variable amount of water is added. The M-R law has been set up assuming that the planetary composition is similar to the averaged composition for silicates and iron obtained from the major elements ratio of 94 stars hosting exoplanets. Except on Earth for which a tremendous amount of data is available, the composition of silicate mantles and metallic cores cannot be constrained. Similarly, thermal profiles are poorly known. In this work, the effect of compositional parameters and thermal profiles on radii estimates is quantified. It will be demonstrated that uncertainties related to composition and temperature are of second order compared to the effect of the water amount. The Super-Earths family includes four classes of planets: iron-rich, silicate-rich, water-rich, or with a thick atmosphere. For a given mass, the planetary radius increases significantly from the iron- rich to the atmospheric-rich planet. Even if some overlaps are likely, M-R measurements could be accurate enough to ascertain the discovery of an earth-like planet .The present work describes how the amount of water can be assessed from M-R measurements. Such an estimate depends on several assumptions including i) the accuracy of the internal structure model and ii) the accuracy of mass and radius measurements. It is shown that if the mass and the radius are perfectly known, the standard deviation on the amount of water is about 4.5 %. This value increases rapidly with the radius uncertainty but does not strongly depend on the mass uncertainty. 2 I. Introduction Observing other planetary systems opens the perspective of discovering new types of planets which are unknown in our Solar System. The recent discoveries of exoplanets with masses below 20 Earth masses (Santos et al. 2004; McArthur et al. 2004; Rivera et al. 2005; Beaulieu et al. 2006; Lovis et al., 2006; Udry et al. 2007; Bennett et al., 2008; Mayor, 2008) confirms the existence of « Super-Earths », a word that has been used by exoplanetologists for a few years (e.g. Melnick et al. 2001). In agreement with Seager et al. (2007), the Super-Earths family considered in this work includes not only terrestrial planets such as Mars, Venus and Earth (Sotin et al, 2007), water-rich planets which are similar to giant icy moons (Léger et al., 2004), and iron-rich planets similar to Mercury (Valencia et al., 2006), but also “mini-Neptunes”. The abundance of low mass planets is not yet known. So far, over the 304 exoplanets which have been discovered, more than 90 % are giant planets, and only 26 are below 30 Earth masses. But this observed distribution, which is biased by observational limitations, is evolving very rapidly. First, low mass planets are preferentially in multi-planet systems (10 out of the 13 planets with masses below 20 MEarth are in multi-planet systems). It indicates that planetary systems have generally several low mass planets. Second, preliminary results from the HARPS spectrograph suggest that more than 30% of stars host Super-Earths (Mayor, 2008). Finally, planets around 5 Earth masses can now be detected as shown by the discoveries of Gliese 581c (Udry et al. 2007), OGLE-2005-BLG-390Lb (Beaulieu et al., 2006), MOA-2007-BLG-192 (Bennett, 2008), and HD40307a-c (Mayor, 2008). Planets can be either gaseous or solid. Some overlap should exist between these categories and we argue that large planets, several tens of Earth masses, are not necessarily gaseous planets. The presence (or absence) of an atmosphere depends indeed on the interplay between three factors: primordial gas accretion (Benz, 2006), outgassing (Musselwhite and Drake, 2000), and atmosphere erosion by the stellar wind (Erkaev et al., 2007). Zuckerman et al (1995) have shown that stellar winds may lead to the inhibition of giant planet formation by rapid gas depletion. Since the structure of a giant planet consists of a solid core surrounded by a large primordial atmosphere of H2 and He, a 3 strong depletion can result in a very massive naked core which bears many similarities with solid planets discussed in the present paper. Recently, Baraffe et al. (2008) have discovered such a 200 Earth-mass planet within HAT-2 b. Previous work by Zapolski and Salpeter (1969) and Fortney et al. (2007) have proposed mass- radius relationships in order to distinguish between gaseous and solid planets over a very large mass range. Several models have also been developed for solid bodies of moderate masses (Léger et al., 2004; Valencia et al., 2006; Sotin et al., 2007; Seager et al, 2007; Adams et al., 2008). These studies use robust equations of state (EOS’s) which allow the calculation of densities over a very large range of pressure and temperature. Although different assumptions for the planet composition are used, the results are generally in good agreement. A power-law relation of the form R/R0=a.(M/M0)b can be found for each family with a power exponent which decreases from 0.3 for low-mass exoplanets down to 0.2 for large-mass exoplanets (above 70 Earth masses), in agreement with the pioneering work of Zapolski and Salpeter (1969). The parameter (a) strongly depends on the planet composition, but its dependence is difficult to quantify because the models differ in many ways : silicate composition, thermal profile, metallic core mass fraction, choice of equation of state, and range of investigated masses. Both the thermal structure and the precise composition will not be known in the near future since these parameters are still debated for the Earth. In this work, the effects of these different inputs are quantified in order to investigate the possibility for M-R relationships to distinguish between water-poor (like Earth) and water-rich (Ocean-planets) solid planets. It illustrates how accurate will be the determination of the composition of a solid planet in the near future. This study focuses on silicate and water rich planets. By analogy with the solar system, a metallic iron core is envisaged if the amount of iron is sufficient. One important aspect of these planets is that they are potentially « habitable », in the sense of bearing a sufficient amount of liquid water at the surface (if permitted by the planetary surface temperature). It is anticipated that terrestrial concepts such as oceans and continents, volcanic and tectonic activity and even vegetal cover (« forests ») are applicable to these planets. Planets having a mass lower than one Earth' mass, have probably lost their 4 liquid (and vapour) water and are thus not habitable, loosing their exobiological interest (Huang 1960). Interestingly, Earth may lie at the bottom of an « habitability sequence » since on average, habitable exoplanets are probably larger than the Earth. In addition, planets larger and more massive than the Earth are easier to detect, whatever the detection method: radial velocity, astrometry, transits, microlensing or direct detection (Selsis et al., 2007; Schneider 1999). These are the two reasons leading us to investigate Super-Earths rather than « Mini-Earths ». It is assumed that the structure of a super-Earth is similar to that of the Earth and other planets and icy satellites of the solar system: a core composed of an iron-rich alloy, overlaid by a silicate mantle, and a water-rich outer layer. The bulk composition is then fixed by the amount of water, and the atomic ratios Fe/Si and Mg/Si (see section II). Stars which host exoplanets, have Fe/Si and Mg/Si ratios which can differ significantly from solar ones (Beirao et al., 2005; Gilli et al., 2006). It is shown that the present model is compatible with all these compositions, which means that the presence of silicate-rich planets with or without water is very likely around these stars. In addition, the variations of Fe/Si and Mg/Si ratios are investigated and their implications on the M-R relationships are quantified. Section II describes the chemical composition of planetary interiors and the modelling. A reference model assumes specific composition and physical parameters (Table 1). In section III, M-R relationships are inferred using the reference model. Section III also describes the influence of compositional and temperature variations on the scaling law. In a last part, results are compared with those of previous studies. In addition, it is shown that future exploration programs may allow for a rough characterization of the solid part of discovered exoplanets if the atmospheric contribution can be removed. 5 II. Internal structure of super-Earths II.1. Composition of planets Massive planets are assumed to be fully differentiated and composed of an iron rich core, a thick silicate layer, and possibly a water-rich layer which draws the transition from Earth-like to icy- rich planets (Figure 1). Both silicate and water layers are split into two sublayers in order to take into account both high and low pressure mineralogy. This slight complexity avoids an underestimate of the radii. It is not fully required for very massive planets where low pressure phase layers are very thin but is important for planets below 10 Earth masses (ME). As an example, the upper silicate layer composed of low-pressure phases on Earth occupies 35 % of the total volume of the mantle. In the solar system, silicate mantles and iron cores are composed of mineral phases which mainly combine eight elements: Si (3.548 10-5), Mg (3.802 10-5), Fe (3.467 10-5), O (8.51 10-4), Ca (2.291 10-6), Al (2.951 10-6), Ni (1.778 10-6), and S (1.862 10-5). Numbers indicated in parenthesis are the atomic abundance of each species relative to hydrogen in the solar atmosphere (Cox et al., 1999). The elementary abundance of a planet is supposed to be the stellar abundance, as it is the case for the meteorites when their composition is compared to that of the Sun’s photosphere (Cox et al., 1999; Lodders et al., 2003). Sotin et al (2007) have shown that a planet can be described with great accuracy using only the four elements (Si, Mg, Fe, O) and adding (Ca, Al, Ni, S) to their closest major element (Ni behaves like Fe, most of the sulphur is present in the iron core, Al is equally divided between Mg and Si in order to account for charge conservation, and Ca is added to Mg). In addition, hydrogen must be added for introducing the water layer above the silicates. The (Fe/Mg) ratio is supposed constant throughout the mantle (Sotin et al., 2007) and identical in each phase. Its value is fixed by the Mg# parameter which is defined by: Mg =# Mg Mg +    Fe    silicates (1) 6 Using the simplified composition proposed by Sotin et al. (2007), planetary composition is fixed with only four parameters : [Fe/Si], [Mg/Si], Mg#, and the amount of water in the planet. Solar values of [Fe/Si] and [Mg/Si] are 0.977 (0.986) and 1.072 (1.131) respectively, with the values in parenthesis being the ratio once Ca, Al, Ni, and S are added to their closest major elements. Throughout the paper, these corrections, which are relevant for silicate material, are imposed. Distribution of elements within each layer is not straightforward. The water compound must be mainly put in the two upper layers which are considered as low and high pressure icy mantles (Figure 1). But a small amount of water could be incorporated in hydrated silicates at moderate pressures. On Earth, the amount of water which is trapped in the upper mantle is very small and still debated (Manning, 2006). Tajika (1998) suggests a total amount roughly equal to the water abundance in the oceans while Karato (2004) argues that a large amount of water, up to 10 times the ocean content, can be present in Earth's mantle. Even with this very high estimate, the water which is trapped on Earth’s mantle represents less than 0.23 % of the planetary mass. Such a small relative amount leads us to neglect the amount of water present in the silicates. Distribution of chemical compounds (Si, Mg, Fe, O) between the silicate mantle and the iron core is not fully known and depends on the processes involved in the differentiation of the core. Whereas magnesium can only be incorporated in silicates, the other (Fe, O, Si) elements can be located either in silicates or in the metallic core, with Fe going preferentially in the core and O and Si in the silicates. Following Sotin et al. (2007), we propose a simplified mineralogy for the silicate mantle. The lower mantle is composed of perovskite ([Mg,Fe]SiO3) and magnesowüstite ([Mg,Fe]O). The upper mantle is modelled with two phases: olivine ([Mg,Fe]2SiO4) and pyroxene ([Mg,Fe]2Si2O6). The whole mantle is assumed chemically homogeneous and the limit between upper and lower mantle is only due to the pressure controlled transformation of olivine to perovskite at 23 GPa. Although iron is the major component of the metallic core, a large amount can be trapped in the mantle (for example, Mg# ~ 0.7 on Mars). Si and O are preferentially in the silicate matrix but can also mix with iron in the metallic core. Silicium has been proposed first for explaining the low density of 7 the Earth’s core (Birch, 1964) and several studies suggest that more than 7 %wt of Si is trapped in the core (e.g. Allègre et al., 1995; Javoy, 1995). Oxygen is not soluble with iron at ambient condition, but at high pressure and temperature it can form iron oxide FeO. Ohtani et al. (1984) have shown that 8.7 %wt of oxygen can mix with iron at 20 GPa and 2500 °C. Sulphur can be incorporated in the core as FeS. Sulphur abundance cannot exceed 3 %wt in the Earth’s core (Dreibus and Palme, 1996; Allègre et al., 1995; Javoy, 1995), but much larger values (around 14 %wt) are expected on Mars (Sohl and Spohn; 1997). Since it is not possible to constrain the amount of volatiles which are trapped in the metallic core of exoplanets, the model described in section II, assumes that the core is made of pure iron (Table 1). This assumption allows us to compare the accuracy of several equation of states under extreme conditions. The effect of volatiles on M-R relationships will be assessed in section III. It will be shown that it influences the size of the core but only slightly the size of the planet. First, the diversity of planetary compositions is investigated by looking at the chemical abundances of 94 stars with planetary companion (Fig. 2). The data come from Beirao et al. (2005) and Gilli et al. (2006). Measured ratio are corrected by adding the contribution Ca, Al, and Ni to their closest major element. Stars have Mg/Si ratio that vary between 0.55 and 1.85 and Fe/Si ratio between 0.5 and 1.3. The population is distributed around Fe/Si=1.01 and Mg/Si = 1.34 (white dot). While Fe/Si is similar to the solar value (0.99), Mg/Si is significantly larger (1.13). It is worth noting that all these data points are within the validity domain of the planetary model proposed above. Ranges of Fe/Si and Mg/Si ratio for specific internal structure without hydrosphere are plotted depending on the weight percentage of the core and the Mg#. With the silicate mineralogy described above, the Mg/Si ratio is allowed to vary within the range [0.6 - 2.0]. Therefore, all the planets which can be derived from the known compositions of the stars are compatible with our petrological model. There is no upper limit on the Fe/Si ratio since large values can be accounted for by the relative size of the core. For example, Mercury which has a Fe/Si ratio around 8, has a relatively large core An increase in the Mg# parameter allows for higher and broader values of Mg/Si. Distribution of Fe/Si around 1.0 is compatible with planets possessing a metallic core (20-30 % wt). For Mg#=0.8, almost 90% of the star 8 compositions are compatible with a silicate-rich planet, while the remaining data on the right with high Mg/Si values require higher Mg# to be explained. In this paper, we assume that there is always a metallic core. Nonetheless, if both Fe/Si and Mg/Si are low, the core is not required if Mg# is very low (bottom left part of stellar compositions in Figure 2). We have chosen not to consider this extreme case because it corresponds to a very small amount of data. In addition, solutions are not unique and both low Fe/Si and Mg/Si ratio can be explained with moderate Mg# (~0.8) and small metallic cores. As a reference (Table 1), we propose to take [Fe/Si]=1.01 and [Mg/Si] = 1.34, well centred within the proposed distribution of Gilli et al. (2006). The Mg number is fixed to 0.8 which means that silicates are mostly forsteritic and that the iron is mainly located into the core. This hypothesis is very realistic because iron is known to mix with magnesium within silicates. Its value is around 0.9 on Earth and 0.7 on Mars. II.2. A brief description of the model The model used in this paper is mainly based on the previous version described in details in Sotin et al. (2007). Input parameters are described in Table 1. For a given composition and a planetary mass, internal structure of a planet is computed using the following set of equations: 2   ( ) dxx (2) 2 x ρ i R i + 1 = = 5  ( ) ∑ ∫ M r dr r 4 ρ π =  i R i  i 1 = ( ) ( ) ( ( ) ) rTrP r f , ρ  i i EOS i R ( ) ( ) ( ) ( )  ∫ rP dxxgx RP i 1 + ρ + i i i i 1 + r  2 RRg G 4 π    r ( ( ) ) ∫ rg i +   i i r r 2 R   i  γ  ( ) ( ) )  rg dr rT exp  i i  Φ  ( RT i ∫ r = R i + 1 = 1 + 9 with r the radius, Ri the position of the interfaces between layers i-1 and i, ρ the density, P the pressure, T the temperature, g the acceleration of gravity, and γ and Φ the Grüneisen and the seismic parameters, respectively (Poirier, 2000). This set of equations is solved iteratively by adapting the size of the core in order to fit the imposed Fe/Si and Mg/Si ratio. Details of computation and parameter values can be found in Sotin et al. (2007). This model has been validated using solar system bodies (including large icy moons) for both Earth-like and water-rich planets (Léger et al., 2004; Sotin et al., 2007). Sotin et al. (2007) proposed mass-radius relationships for Earth-like and icy planets up to 10 Earth masses. The present work expands the previous work of Sotin et al. (2007) to masses up to 100 Earth mass and investigates the precision of scaling laws for future interpretations of (M-R) observations. The most important improvement is related to the equations of state (EOS) and is detailed below. II.3. Equations of state In the set of equations (2), the second line is the relation between temperature, pressure, and density, for each material. As shown in Seager et al. (2007), two different domains can be distinguished. At low pressures (< 200 GPa), different (P-T-ρ) relations are used: Mie-Grüneisen- Debye (MGD), Vinet, and Birch-Mürnaghan formulations. The parameters of these EOS are well constrained by experimental data. In addition, these equations partly take into account the complex chemistry of the metallic core and the silicate mantle because EOS have been constrained on a large amount of minerals (Seager et al., 2007; Sotin et al., 2007; Valencia et al., 2006, 2007a, Léger et al., 2004). Vinet and MGD formulations are quite robust when extrapolated to pressures up to 1 TPa. At very high pressures (> 10 TPa), first principles equation of state are commonly used such as the one derived from the Thomas–Fermi–Dirac (TFD) theory, in which the atomic structure of the solid is not considered, and the ab initio quantum mechanical theory for specific minerals (e.g. Poirier, 10 2000). Salpeter and Zapolsky (1967) improved the original TFD formulation in order to take into account some effects related to the interaction between electrons. This approach has been used recently by Fortney et al. (2007) for studying M-R relationship up to 1000 ME. The problem is mainly in the range [200 GPa – 10 TPa] where no EOS is available. Although the Rankine – Hugoniot EOS well describes shock-wave experiments, the extrapolation necessary to provide isothermal EOS is difficult. Alternatively, Seager et al. (2007) extrapolate low pressure EOS up to the pressure where it crosses the TFD. For ice, a more sophisticated interpolation process is described. A third option, chosen in the present paper, is to use the ANEOS code for filling up the gap between low (P<200 GPa) and high (P>10 TPa) pressure domains. ANEOS was initially set up for shock physics studies and provides analytic equation of state for many materials (Thompson, 1990). The density of water, MgSiO3, and iron at 300K is represented as a function of pressure below 100 TPa (Figure 3). For each material, the three EOS (MGD, ANEOS, TFD) are compared. Several tests at high temperatures up to 6000 K (in the solid domain) have been conducted with the same conclusions. The grey domain represents the pressure range in which the compound can be observed for planets less massive than 100 ME. TFD curves are derived from Seager et al. (2007). MGD data are issued from the previous work of Sotin et al. (2007). First, it appears that ANEOS and TFD are very close above 1 TPa except for pure iron where it is close only above 10 TPa. For pure iron, the MGD formulation can be extrapolated up to 40 TPa, which is a surprisingly high value. ANEOS provides iron densities in very good agreement with both MGD estimates at moderate pressure and TFD theory above 10 TPa. Thus, it has been chosen for describing the entire metallic cores. In the silicate layer, the MGD extrapolation is close to both ANEOS and TFD theory up to 100 TPa. This result differs from the tests described in Seager et al. (2007) where strong differences occur above 28 TPa. This difference between the two studies is explained by the parameters which are used in the MGD formulation. Several sets of parameters can fit the experimental constraints at moderate pressures up to 200 GPa. But depending on the chosen set, extrapolation to higher pressures may differ significantly. One important feature of the MGD formulation (or equivalently the Vinet and the Birch- 11 Mürnaghan theories) is that the chemistry of the silicates, specifically the iron amount, can be taken into account in order to calculate the bulk density of the silicate mantle. The difference is small at moderate pressure and negligible above 200 GPa (see for example figure 2 in Seager et al., 2007). Thus, the MGD formulation is justified for low-mass planets where the density of the silicate mantle, which is under moderate pressures, is influenced by its iron content. In this range which is below several tens of GPa, both our formulation and the one proposed by Seager et al. (2007) are in perfect agreement. In the transition zone between 110 GPa and 10 TPa, and above 10 TPa, the ANEOS formulation has been chosen. It is close to our MGD estimates up to 10 TPa. Above this value, ANEOS and TFD provide the same densities. Compared to previous studies, the chosen MGD formulation for silicates under low pressure is comparable with the EOS used by Valencia et al. (2006), and Seager et al. (2007), and in agreement with high pressure experiments data. In the transition zone between 200 GPa and 2 TPa and at very large pressures, density estimates of the pure forsterite (without any iron in the silicates) using ANEOS are consistent with the values proposed by both Seager et al. (2007) and Fortney et al. (2007). In the silicate mantle, the transition of perovskite into post-perovskite above 120 GPa (Shim, 2008) is not considered. Post-perovskite may be the most abundant phase within silicate mantles of Super-Earths, but due to the lack of experimental data on the equation of state, the perovskite phase has been preferred. We argue that this simplification does not change strongly M-R relationships because density measurements of post-perovskite and perovskite in the P-T conditions of the Earth mantle differ by less than 1.2 % (Murakami et al., 2004). On the other hand, post-perovskite may be able to incorporate more iron than perovskite, which could imply a slight increase of the planetary radii at large masses. But in absence of robust experimental constraints, we prefer not to investigate this effect. For ices at pressures below 45 GPa, MGD provides an accurate estimate of the ice density (Hemley et al., 1987). In the moderate pressure range, there is a gradual transformation of ice VII to ice X and MGD, ANEOS, and TFD provide estimates of the ice density which differ significantly. In fact, ANEOS densities are close to the proposed EOS from Belonoshko and Saxena (1991), which has 12 been set up using molecular dynamics in the super-critical domain. Above 8 TPa, ANEOS and TFD provide similar density estimates. Thus, we choose the ANEOS formulation for both the transitional domain and the very large pressure domain. In the transitional domain between 45 GPa and 8 TPa, Seager et al. (2007) prefer using a regular transition from MGD to ANEOS which is based on ab initio simulations, but the density estimates are actually very similar. Without any experimental data in this pressure range, it is impossible to know which approach is better than the other. But in any case, the fact that M-R scaling laws for water-rich planets are very similar in both studies (see figure 9 and section IV.2) indicates that this uncertainty on the EOS of ices in the transitional domain is not critical for setting up M-R relationships. III. Mass - radius relationships for super-Earths and icy worlds III.1 The reference case Planetary radii have been calculated for a range of masses and amount of water ice (Xw) in the reference case (Table 1 and Figure 4) . The first result is that radii are only multiplied by 3 when the mass is increased from 1 to 100 ME, whereas it should be 1001/3 = 4.6 if the sphere was homogeneous and incompressible. The exponent decreases as mass increases resulting in a significant flattening of the curves above 20 ME. For a given mass, the transition from a dry silicate-rich planet to an icy world induces an increase of the radius between 30 % (100 ME) and 40 % (1 ME). All this information can be gathered into one empirical equation which links the mass to the radius using parameters which depend on the amount of water: log    R R E  =  log   ) ( + + γβα   M M E +   ε  M M E    2     log    M M E    (3) 13 with ME the mass of the Earth, RE=6430 km the radius of a one Earth mass planet in the reference case (but not the Earth’s radius which is 6371 km), α, β, γ, and ε parameters which depend on the amount of water Xw according to a power 2 law: ξ = 2 ∑ i Xξ 0i = i w (4) where ξ symbolizes either α, β, γ, or ε (Table 2). Equations (3-4) with parameters of Table 2 describe M-R relationships estimated from the model in the range 1-100 ME for any planet ranging from Earth- like to pure ice planets with an accuracy better than 1 % (Figure 4). III.2. Effect of composition parameters Since stars with orbiting planets have different amounts of major elements, we have investigated the effect of the elementary composition on the core size and the radius of the silicate mantle. Variations of both core and silicate radii are calculated for a large range of composition distributed randomly about the reference values (Table 1). Masses of 1, 10, 50, and 100 ME have been tested and it is worth noting that results are similar, whatever the mass is. As expected, increasing the Fe/Si ratio results in a decrease of the planetary radius and an increase of the core radius (top panel in Figure 5). For a ratio Fe/Si twice larger, the radius decreases of 1.5 % only but the inner metallic core is increased by 20 %. The effect of Mg/Si is illustrated in the second plot. It is quite small and opposite to the Fe/Si ratio. For Mg/Si twice larger, the silicate mantle radius increases by 2.5 % and the inner core radius decreases by 10 %. Finally, the influence of the amount of volatiles which may be trapped within the metallic core, has been investigated. If a large amount of volatiles is incorporated within the metallic core, its density decreases and its size increases. If 40 % of volatiles are incorporated, the size of the metallic core increases by 25 % while the radius of the silicate mantle is increased by only 1 % (third panel in Figure 5). The bottom diagram represents the influence of the iron content within the 14 silicates (Mg#). The radius of the silicate layer is almost insensitive to this parameter which strongly affects the size of the metallic core since it controls the amount of iron which is incorporated within the silicates. If the amount of iron is increased within silicates (lower Mg#), it implies logically a significant reduction of the metallic core. The main result is that compositional effects in silicates and metallic alloys influence only slightly the global size of the dry planet (iron core + silicate layer). From left to right, the planetary radius is always within the range ± 2 % of the reference value. On the contrary, radius of the core is very sensitive to the composition of silicates and alloys. Variations as large as 10 % can be expected depending on the amount of iron and volatiles within the planet. But the most important parameter is the Mg# which implies variations of the core size as large as 30 %. III.3. Effect of temperature variations The effect of temperature variations is plotted in figure 6 for a 10 Earth mass planet with Xw= 0.01 %wt. Ninety thermal profiles have been produced randomly within a plausible range by varying both temperature drops at the interfaces and adiabatic slopes within the different layers (Table 1). The warmest thermal profile is actually not realistic since silicate mantle is almost fully liquid. Similarly, the coldest case is close to the case of complete cooling, for which no heat sources (e.g. gravitational energy, tidal heating, or radiogenic heating) remain within the planet. It may be possible, but only for small and very old bodies. Core radius is plotted versus planetary radius at the bottom of figure 6. It can be seen that the effect of temperature profiles is very small for the two radii since the largest variations from the reference case are lower than 1.5 %. Similar tests have been conducted in the same way for different masses leading to the same conclusions. 15 III.4. Determination of the water amount in a solid planet : model uncertainties The two previous sections demonstrate that the radius of an Earth-like planet is not strongly affected by reasonable compositional variations and temperature differences. In fact, the radius varies by less than 5 % from the smallest possible case to the largest one, whatever the mass of the planet is. If the mass and the radius of the solid part of a planet are precisely measured and if the atmospheric contribution can be removed, the amount of water can be determined. In order to estimate the accuracy of this water amount, 530 simulations have been performed. For each simulation, the different parameters are randomly fixed in the ranges indicated in the third column of Table 1. Distribution of the different parameters is plotted in figure 7. In each case, the planet is described by its mass Mp, its global composition (Xw, Fe/Si, Mg/Si, Mg#) and a randomly fixed temperature profile. The model computes the whole internal structure and the planetary radius Rp. Since the characteristics of the planet are not those of the reference case, the computed radius Rp differs from the theoretical radius computed with equations (3-4). Alternatively, if one assumes that (Mp, Rp) are the measured mass and radius of a given planet, it is possible using equations (3-4) to estimate iteratively the amount of water of this planet wX . Using the 530 simulations plotted in figure 7, and assuming that the (Mp, Rp) * measurements are exact, the distribution of the differences Xw- wX can be obtained. It follows a * gaussian distribution with an average value almost equal to 0 and a standard deviation mod = xσ 045.0 (figure 8). Thus, we argue that the amount of water of a Super-Earth can be determined with a standard deviation of 4.5 % as long as its mass and radius are known exactly. This standard deviation is the result of the different unknowns in the internal structure in term of composition and temperature. It is a very low value, which illustrates well the fact that silicate composition and temperature profiles are of second order importance relative to the amount of water for M-R relationships. 16 IV. Discussion A description of the mass-radius relationship for super-Earths has been developed in the mass range [1 – 100 MEarth]. The accuracy of the model in terms of compositional and thermal uncertainties has been investigated. In this section, the relevance of this study for future super-Earths exploration is discussed. IV.1. Theoretical mass-radius relationships After the pioneering work of Zapolsky and Salpeter (1969), M-R relationships for cold spheres had not risen scientific interest for several decades. But the recent discovery of low-mass exoplanets and the future possibility of discovering super-Earths have motivated the search for more precise M-R relationships. Léger et al. (2004), and Sotin et al. (2007) have proposed M-R relations for Earth-like and water-rich planets at low masses (< 10 MEarth). Valencia et al. (2006) chose to compare Mercury- like planets with Earth-like planets. Fortney et al (2007) provided M-R relationships for masses ranging from 0.01 Earth masses to 10 Jupiter masses. In previous studies, it has been shown that for planets less than 10 Earth masses, a simple power law relation can be found (Valencia et al., 2006; 2007b; Sotin et al., 2007). Compared to equation (3) and to the results from Seager et al. (2007), these previous works neglect the influence of the mass on the exponential factor (ε = γ = 0) because they are dealing with planets of low masses. But taking into account larger masses induces a significant decrease of the power exponent. A flattening of the M-R curves is expected at large masses because compression effects become more and more important (Zapolsky and Salpeter, 1969; Seager et al., 2007; Fortney et al., 2007). Without going to very large masses (around 1000 ME) where the materials begin to behave like a Fermi gas (Fortney et al., 2007), the compression effect is actually noticeable at low masses. A possible way to take this effect into account is to use a constant exponent for the power law (1/3) and add a correction term which increases with mass (Seager et al., 2007). But we prefer the formulation proposed in equation (3) because it illustrates well how the power exponent decreases when mass increases. 17 IV.2. Influence of planetary composition on M-R relationships Zapolski and Salpeter (1969) studied planetary M-R relations assuming homogeneous spheres. In the recent work of Seager et al. (2007), mass-radius relationships are investigated for cold planets made primarily of iron, silicates, water, and carbon compounds over a very wide range of masses. Léger et al. (2004), Valencia et al. (2006, 2007b), and Sotin et al. (2007) used complex chemistry for silicates and introduced thermal effects. The effect of large amounts of water on M-R relationships was first investigated by Léger et al. (2004). In figure 9, the different estimates of M-R relationships in the range 1-100 ME have been plotted. The main result is that all these studies are in good agreement. The slight differences can be explained by the fact that models use different temperature fields, different modal compositions for silicates, and different equations of state. In fact, figure 9 illustrates well the fact that precise compositions and/or thermal profile choices are of second order importance in M-R relations. Nonetheless, two significant discrepancies can be noted: in the pure water case for very large masses, and in the pure silicate case at low masses. In the case of icy planets, Fortney et al. (2007) proposed an M-R relationship which is significantly above the others. They used an empirical equation of state corrected of thermal effects along the Uranus/Neptune adiabat (Guillot, 2005), while in Seager et al (2007) and in this work, a combination of low and high pressure equation of state is preferred. Seager et al. (2007) and the present work use equations of state which are not empirical, especially at low pressure where the density strongly influences the radius. Both studies are in very good agreement with each other. Although Fortney et al (2007) applied the EOS over a much wider range of masses, we suggest that their estimate of radii in the pure water case and in the [ 1 – 100 MEarth] domain is overestimated. In the case of silicates under 10 Earth masses, the M-R curve from Seager et al. (2007) is above the others. In particular, it is above the Earth point (1 Earth radius for 1 Earth mass). In fact, this curve 18 is the M-R relationship for a pure forsterite planet, that is without iron. In the studies of Valencia et al. (2007b), Sotin et al. (2007) and this work, an iron core is always considered, which reduces significantly the radius of low mass planets compared to the ideal silicate case. In addition, our model includes some iron in the silicates (Mg#=0.8 in the reference case), which makes the silicates denser than pure forsterite. It is worth noting that these two points become negligible above 20 Earth masses because the results from Seager et al. (2007) are very similar to those of Fortney et al. (2007) and ours. In figure 2, it has been shown that the compositions of the stars which host exoplanets are compatible with planets composed of an iron core, a silicate mantle and eventually an hydrosphere and an atmosphere. Silicates are made of pyroxene and olivine or eventually pyroxene and quartz for very low Mg/Si ratio. But one must consider that planets may not be necessarily composed of a mixture of iron, silicates, water and He-H gas. In particular, carbon rich planets can be envisaged in some stellar environments. To consider such new families is certainly relevant, especially from the astrobiological point of view, but it is worth noting that it will be very hard to distinguish this specific family from Super-Earths. First of all, the carbon either in its graphitic form or mixed with silicon, has density similar to a mix of silicates and water as is shown in Figure 9 using Zapolky and Salpeter data (1969) for pure carbon. If it mixes with water forming hydrates and or clathrate hydrates, densities are almost equal to water densities, at least at low pressures where these phases are known to exist. Seager et al. (2007) have indeed shown that carbon- rich planets are located within the Super-Earth domain. That is why these specific planets have not been investigated in this work. IV.3. How to distinguish the different families in the future? For a given mass, the size of a gaseous planet is roughly three times larger than a planet from the Super-Earths family. In this class of planets, the radius of a Mercury-like planet is about 80 % that of an Earth-like planet (Figure 9). Radii also strongly depend on the amount of ices. In fact, it could be possible to characterize Super-Earths composition (Mercury-like, Earth-like, water-rich, or mini- 19 Neptune) from M-R measurements. Nonetheless, solutions are not unique. As an example, it must be noted that a silicate-rich planet surrounded by a very thick atmosphere could provide the same mass and radius than an ice-rich planet without atmosphere (Adams et al., 2008). The existence of a thick atmosphere can be explained either by an important outgassing during the planet history or by an erosion of a massive primordial atmosphere mostly composed of hydrogen and helium. The last case is very likely since a primordial atmosphere can be eroded by a strong stellar wind (Khodachenko et al., 2007). In the first case, the atmosphere might be composed mostly of N2 and CO2. But outgassing processes do not produce enough atmosphere to give a significant difference to the Earth-like case. According to solar-system examples (Venus, Earth, Titan), outgassing is limited in the sense that even for larger planets where more silicates are available, the amount of volatiles which can be outgassed through geological processes is very limited. For transiting planets, both mass and radius can be determined. Mass measurement is provided by radial velocity and, in the near future, by astrometry. The precision on the planet mass derived from radial velocity measurements is dominated by the precision on the amplitude K of the radial velocity measurement. In the case of the 4.7 Earth mass planet HD 40307 b, it has been shown that a typical 5 % accuracy can be achieved with a 3.6 m telescope (Mayor et al. 2008). With high accuracy radial velocity spectrograph projects such as Codex at the E-ELT (Pasquini et al., 2005) it will be significantly better. The radius is inferred from the relative decrease of the stellar flux during the transit. The precision of the planet radius depends on the precision of the transit depth, itself depending on the photon noise for stable stellar flux, and on the precision of the parent star radius. The latter is the dominant factor as illustrated by the case of HD 189733 b: the precision of the stellar radius is 1.5 % while the precision on the transit depth (measured by the Hubble Space Telescope) is 0.27 % (Pont et al. 2007). We infer that for a planet with 2 Earth radius (1/5 RJup), the precision on the depth is 5 * 0.27 % = 1.35 %, leading to a radius estimate with 35.1 2 + 5.1 2 = 2 % accuracy. In summary, mass and radius of transiting planets can be determined with an accuracy of roughly 5 % 20 and less than 3 %, respectively. But only 0.5 % of planets at 1 UA from a Solar type star make such transits. In these very specific cases, the classification of the planet will be straightforward. For non transiting planets, only the mass M (resp. the product M sin i) can be estimated from astrometry (resp. radial velocity) observations. This information cannot provide any constraint on the nature of a Super-Earth (Mercury-like, Earth-like, water-rich, or mini-Neptune) as long as the radius is not estimated. This measurement will probably be obtained from « mutual events » for planets having a sufficiently large companion (Cabrera and Schneider, 2007), a situation which is very frequent in the solar system. But if mutual events are expected to occur in some cases, on must keep in mind that direct measurements of planetary radii will not be available in the general case. In the mid-term the most significant progress in exoplanetology will most likely come from the physicochemical characterization of exoplanets using spectro-polarimetric observations. It will then be possible to get two independent observations for each planet: the planetary mass (by radial velocity), and the flux. To obtain radius estimate from the flux will be difficult because many parameters are involved in the flux- radius relationships. The radius could be computed from the thermal flux through the relation: F th ( ) ( )λ a TR 4 4 2 πσλ = s (5) with σ the Boltzmann constant, Ts the surface temperature and a(λ) an absorption factor reflecting the effective absorption by molecular species at wavelength λ. This expression does not take into account the possibility that the planet is surrounded by rings or possesses a companion, two factors which significantly contributes to the flux (e.g. Schneider, 2003). Radius could also be inferred from the reflected flux, but it relies on an estimate of the albedo: 21 F refl ( P, λ ) = F * 2  ( ) λ   R a    ( P,A λ 4 ) Φ (6) where P is the state of polarization of the observed reflected light, Φ an orbital phase factor, and A a factor combining the planet reflectance and molecular absorption features. Here again this relation does not hold in case of rings and companions surrounding the planet (Arnold and Schneider, 2004; Cabrera and Schneider, 2007). Equations (5) and (6) illustrate that the thermal and reflected flux are proportional to the planet radius to the square. It is out of the goal of this work to discuss the real and complex modelling of these fluxes. The point is that in the general case, one can expect to get a rough estimate of the planet radius from the flux measurement, but most certainly with large uncertainties because of the many parameters involved in the relationships between the flux and the radius. In the following discussion, we assume that the contribution of the atmosphere can be removed. In fact the detectability of the spectral signatures observable in transmission, depends on the molecular weight of the main atmospheric component, the planetary gravity field and the atmospheric temperature. As the temperature can be, in first approximation, derived from orbital characteristics, we are certainly able to distinguish a rocky/icy planet wrapped into an atmosphere from a gaseous one. Moreover, using the secondary transit technique in the visible spectral range, albedo can be measured (Rowe et al., 2006). If the mass of the planet is known and if the reflected flux can be measured together with the planetary albedo, the radius can be inferred by combining equations (3) and (5-6). If the radius estimate derived from the M-R relation differs significantly from the radius measured with the planet flux (thermal or reflected), it can only be due to the presence of a thick atmosphere. In the last case, M-R relationships allow the estimation of the size of the solid core by subtracting to the measured radius the theoretical radius derived from equation (3). If for a planet from the Super-Earth family, both mass and radius are measured, and if the atmospheric contribution can be removed, then its nature is deduced from internal structure modelling. 22 As an example, the following discussion illustrates the degree of precision that can be achieved for the estimate of the water amount. The standard deviation of this estimate xσ is related to the standard deviation of ( ( ) 2 mod σ σ x x = ) 2 the model ( )2 meas σ x + mod xσ , and the precision of the M-R measurements : . In this relation, the second term has been computed in section III.4. The last term is estimated iteratively from equations (3-4) for any planetary mass. In the ideal case where both M and R have been measured exactly, the amount of water in the solid (or liquid) part of a Super- Earth will be known with a precision better than 4.5 % (figure 8; lower left corner of figure 10). In the general case, radius estimate of the solid part of the planet will be poorly constrained (especially if it is deduced from the removal of a thick atmosphere). But it will still be possible to get a rough estimate of the amount of water within a planet. Results are plotted in figure 10 in a σR/R versus σM/M diagram. It appears that radius estimates must be precise for having a good description of the planetary composition but that mass estimates can be less accurate, especially if the radius measurement is inaccurate. This is illustrated by the important flattening of the curves. For example, it can be seen that for a relative radius uncertainty of 3 %, the amount of water can be determined with a precision between 10 % and 12.5 % depending on the mass uncertainty. These results, in good agreement with the previous works from Seager et al. (2007) and Valencia et al. (2007b), clearly illustrate the importance of having good radius measurements for characterization purposes. It is only below 2% of relative accuracy on radius measurement (which is probably not achievable in near future except for transiting planets), that the accuracy of mass measurement will provide more constraints on the composition of the planet. V. Conclusion In this paper, internal structure modelling of planetary interiors has been used to investigate the mass – radius relations for Super-Earths planets. A general law has been proposed which describes the 23 mass-radius relation with an accuracy of 1 % in a nominal case from 1 to 100 Earth masses and for any amount of water ice in the planet. Numerical results are in good agreements with previous works and fill the link between the studies devoted to planets with very low masses and those developed for very large masses. In a second part, the uncertainties on radius estimates due to model approximation have been investigated. Compositional variations in silicates and in iron alloys change the relative size of the metallic core but do not influence strongly the planetary radius (less than 2 %). Similarly, very large thermal variations within the planetary layers do not affect significantly the size of the planet (1.5 %). A combination of these two factors indicates that numerical modelling provides radii estimate of dry silicate – rich planets with a precision better than 5 %. Since it will be very hard to provide any constraints on the silicate composition and on the thermal profiles, one cannot expect to have a more precise constraint from numerical modelling. The M-R relation proposed in this work will allow the characterization of the main composition of the solid cores of exoplanets as long as both mass and radii can be measured, and if the contribution of the atmosphere can be removed. This drastic conditions are not yet fulfilled but the different approaches that will allow such determinations in the future have been discussed in this work. In addition, a diagram showing the standard deviation on the water amount estimate as a function of planetary mass and radius uncertainties has been proposed. VI. Acknowledgments The authors want to thank G. Tinetti for her valuable comments on detection perspectives and an anonymous reviewer for her (his) constructive remarks, Part of this work was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under contract with NASA. 24 References : Adams , Seager S., Elkins-Tantor. 2008, Astrophys. J., 673, (2), 1160 Allègre C.J., Poirier J.P., Humler E., Hofmann A.W. 1995, 134: 515. Arnold L. and Schneider J. 2004, Astron. Astroph., 420, 1153. Baraffe, I.; Chabrier, G.; Barman, T. 2008, Astron. & Astrophys., 482 (1), 315. Beaulieu, J.-P.; Bennett, D. P.; Fouqué, P.; Williams, A.; Dominik, M.; Jørgensen, U. G.; Kubas, D.; Cassan, A.; Coutures, C.; Greenhill, J.; Hill, K.; Menzies, J.; Sackett, P. D.; Albrow, M.; Brillant, S.; Caldwell, J. A. R.; Calitz, J. J.; Cook, K. H.; Corrales, E.; Desort, M.; Dieters, S.; Dominis, D.; Donatowicz, J.; Hoffman, M.; Kane, S.; Marquette, J.-B.; Martin, R.; Meintjes, P.; Pollard, K.; Sahu, K.; Vinter, C.; Wambsganss, J.; Woller, K.; Horne, K.; Steele, I.; Bramich, D. M.; Burgdorf, M.; Snodgrass, C.; Bode, M.; Udalski, A.; Szymański, M. K.; Kubiak, M.; Wi(cid:31)ckowski, T.; Pietrzyński, G.; Soszyński, I.; Szewczyk, O.; Wyrzykowski, Ł.; Paczyński, B.; Abe, F.; Bond, I. A.; Britton, T. R.; Gilmore, A. C.; Hearnshaw, J. B.; Itow, Y.; Kamiya, K.; Kilmartin, P. M.; Korpela, A. V.; Masuda, K.; Matsubara, Y.; Motomura, M.; Muraki, Y.; Nakamura, S.; Okada, C.; Ohnishi, K.; Rattenbury, N. J.; Sako, T.; Sato, S.; Sasaki, M.; Sekiguchi, T.; Sullivan, D. J.; Tristram, P. J.; Yock, P. C. M.; Yoshioka, T. 2006, Nature, 439 (7075), 437. Beirao P., Santos N.C., Isralian G., Mayor M. 2005, Astron. & Astrophys., 438, 251. Belonoshko A. and Saxena S.K. 1991, Geochem. Cosmoch. Acta, 381. Bennett, D. P.; Bond, I. A.; Udalski, A.; Sumi, T.; Abe, F.; Fukui, A.; Furusawa, K.; Hearnshaw, J. B.; Holderness, S.; Itow, Y.; Kamiya, K.; Korpela, A. V.; Kilmartin, P. M.; Lin, W.; Ling, C. H.; Masuda, K.; Matsubara, Y.; Miyake, N.; Muraki, Y.; Nagaya, M.; Okumura, T.; Ohnishi, K.; Perrott, Y. C.; Rattenbury, N. J.; Sako, T.; Saito, To.; Sato, S.; Skuljan, L.; Sullivan, D. J.; Sweatman, W. L.; Tristram, P. J.; Yock, P. C. M.; Kubiak, M.; Szymanski, M. K.; Pietrzynski, G.; Soszynski, I.; Szewczyk, O.; Wyrzykowski, L.; Ulaczyk, K.; Batista, V.; Beaulieu, J. P.; Brillant, S.; Cassan, A.; Fouque, P.; Kervella, P.; Kubas, D.; Marquette, J. 2008, Astrophys. J., 2008arXiv0806.0025B. Benz W. 2006, Meteoritics & Planetary Science, 41, Supplement, Proceedings of 69th Annual Meeting of the Meteoritical Society, 5393. Birch F. 1964, J. Geophys. Res. 69: 4377. Cabrera J. and Schneider J. 2007, Astron. & Astrophys., 464, 1133. Cox et al., 1999, Allen's astrophysical quantities, 4th edition, Arthur N. Cox Editor, Springer – Verlag. Dreibus G., Palme H. 1996, Geochimica et Cosmochimica Acta, 60, (7), 1125. Erkaev N., Lammer H., Kulikov Y., Langmayr D., Selsis F., Jaritz G. & Biemat H. 2007, Astron. & Astrophys., 472 , 329 Fortney J.J., Marley M.S., Barnes J.W. 2007, Astroph. J., Volume 659 (2), 1661. Gilli G., Israelian G., Ecuvillon A., Santos N.C., Mayor M. 2006, Astron. Astroph., 449 (2), 723. Hemley R.J., Jephcoat A.P., Mao H.K., Zha C.S., Finger L.W., Cox D.E. 1987,. Nature, 330, 737. Huang S. 1960, Publ. Astron. Soc. Pacific, 72, 489. Javoy M. 1995, Geophys. Res. Lett., 22(16): 2219. Karato S. 2004, American Geophysical Union, Fall Meeting 2004, abstract #T44B-01. Khodachenko M., Ribas I., Lammer H., et al. 2007, Astrobiology, 29, 167. Léger A., Selsis F., Sotin C., Guillot T., Despois D., Mawet D., Ollivier M., Labèque A., Valette C., Brachet F., Chazelas B., Lammer H. 2004, Icarus, 169, 499. Lodders, K. 2003, Astrophys. J., 591, 1220 Lovis, C., M.Mayor, Pepe, F., Alibert, Y., Benz, W., Bouchy, F., Correia, A. C. M., Laskar, J., Mordasini, C., Queloz, D., Santos, N. C., Udry, S., Bertaux, J., & Sivan, J. 2006, Nature, 441, 305. Manning C.E. 2006, Geochimica et Cosmochimica Acta, 70 (18), 388. Mayor, M. 2008, Super-Earths workshop, Nantes, June 2008. 25 Mayor M., Udry S., Lovis C., Pepe F., Queloz D., Benz W., Bertaux J.-L., Bouchy F., Mordasini C., & Segransan D., 2008. Astron. & Astrophys. submitted . McArthur, Barbara E.; Endl, Michael; Cochran, William D.; Benedict, G. Fritz; Fischer, Debra A.; Marcy, Geoffrey W.; Butler, R. Paul; Naef, Dominique; Mayor, Michel; Queloz, Diedre; Udry, S.; Harrison, Thomas E. 2004, Astrophys. J.,614 (1), 81. Melnick et al. 2001, 199th AAS Meeting, #09.10 Murakami M., Hirose K., Kawamura K., Sata N., Ohishi Y. 2004, Science, 304, 855. Musselwhite D. and Drake M. 2000,. Icarus, 148, 160 Ohtani E., Ringwood A.E., and Hibberson W. 1984, Earth Planet. Sci. Lett., 71: 94. Pasquini L, Cristiani S., Dekker H., Haehnelt M., Molaro P., Pepe F., Avila G., Delabre B., D'Odorico V.,Vanzella, E. and 12 coauthors 2005. The ESO Messenger no 122 p. 10 Poirier J.-P.. 2000, Cambridge Univ. Press, 2nd ed., Cambridge, UK. Pont F., Gilliland R., Moutou C., Charbonneau D., Bouchy F., Brown T., Mayor M., Queloz D., Santos N. & Udry S. 2007. Astron. & Astrophys., 467, 1347 Rivera, E. J.; Lissauer, J. J.; Butler, R. P.; Marcy, G. W.; Vogt, S. S.; Fischer, D. A.; Brown, T. M.; Laughlin, G.; Henry, G. W. 2005, Bull. Am. Astron. Soc., 37, 1487. Rowe, Jason F.; Matthews, Jaymie M.; Seager, Sara; Kuschnig, Rainer; Guenther, David B.; Moffat, Anthony F. J.; Rucinski, Slavek M.; Sasselov, Dimitar; Walker, Gordon A. H.; Weiss, Werner W. 2006, Astrophys. J., 646 (2), 1241. Salpeter E.E., and Zapolsky H.S. 1967, Physical Review II, 158, 876. Santos, N. C.; Bouchy, F.; Mayor, M.; Pepe, F.; Queloz, D.; Udry, S.; Lovis, C.; Bazot, M.; Benz, W.; Bertaux, J.-L.; Lo Curto, G.; Delfosse, X.; Mordasini, C.; Naef, D.; Sivan, J.-P.; Vauclair, S. 2004, Astron. & Astrophys., 426 (9), 19. Schneider J. 1999, C. R. Acad. Sci. Paris 327, 621. Schneider J. 2003. in Proceedings of Towards other Earths: Darwin/TPF and the search for extrasolar terrestrial planets , ESA SP-539, p. 205 Seager S., Kuchner M., Hier-Majumder C. A., Militzer B. 2007, Astrophys. J., 669 (2), 1279. Selsis, F.; Chazelas, B.; Bordé, P.; Ollivier, M.; Brachet, F.; Decaudin, M.; Bouchy, F.; Ehrenreich, D.; Griessmeier, J.-M.; Lammer, H.; Sotin, C.; Grasset, O.; Moutou, C.; Barge, P.; Deleuil, M.; Mawet, D.; Despois, D.; Kasting, J. F.; Léger, 2007, Icarus, 191 (2), 453. Shim S.H. 2008, Ann. Rev. Earth Planet. Sci. (36), 569. Sohl F., Spohn T. 1997, J. Geophys. Res., 102 (E1), 1613. Sotin C., Grasset O., Mocquet A. 2007, Icarus, 191 (1), 337. Tajika E. 1998, Geophys. Res. Lett.25(21), 3991. Thompson S.L. 1990, Sandia Natl. Lab. Doc., SAND89-2951. Udry, S., Bonfils, X., Delfosse, X., Forveille, T., Mayor, M., Perrier, C., Bouchy, F., Lovis, C., Pepe, F., Queloz, D., & Bertaux, J.-L. 2007, Astron. & Astrophys., 469 (3), L43. Valencia D., O’Connel R.J., Sasselov D. 2006, Icarus, 181, 545. Valencia D., Sasselov D., O’Connell R.J. 2007a, Astrophys. J. 656, 545. Valencia D., Sasselov D.; O'Connell R., Astrophys. J., 2007b, 665 (2), 1413-1420. Zapolsky H.S., Salpeter E.E. 1969, Astrophys. J., 158, 809. Zuckerman B., Forveille T. & Kastner J. 1995, Nature, 373, 494. 26 Figure captions: Figure 1: Internal structure of planets: A fully differentiated planet is composed of a metallic core, surrounded by a silicate mantle and possibly an icy mantle. Both silicates and water ices can be split into two sublayers corresponding to low and high pressure phases. The low pressure silicate layer influences the global size of the planet if the amount of water is small and the planetary mass is below 10 ME. For a large amount of water, the pressure at the base of the icy mantle is above the pressure transition from olivine to perovskite. Figure 2: Distribution of Fe/Si and Mg/Si for stars which hosts exoplanets (black diamonds). Measured ratio are corrected by adding the contribution of Ca, Al, and Ni to their closest major element. The distribution is located around Fe/Si=1.01 and Mg/Si = 1.34 (white dot). The solar value is indicated by the white square. Three domains have also been plotted depending on the size of the core (1 %, 20 %, and 40 % respectively). On each domain, variations of the Mg# parameter are indicated (small numbers on dashed lines). It appears that the petrological model used for describing both metallic core and silicate mantles allows to investigate the whole range of composition of stellar systems which are known to host exoplanets. Figure 3: Density variation at 300 K as a function of pressure for pure iron (a), forsterite MgSiO3 (b), and pure water (c). TFD curves are from Seager et al. (2007). ANEOS curves are obtained using the ANEOS package (Thompson, 1990). Mïe-Gruneisen-Debye estimate are derived from Sotin et al. (2007). In addition, data from Belonoshko and Saxena (1991) have been added in the supercritical domain for the water compound (dashed line). For each material, the grey domain indicates the plausible pressure range relevant to planetary interiors below 100 ME. Figure 4: M-R relationships below 100 ME in the reference case. Numbers indicate the amount of water in %wt (Xw). Black diamonds: Model results from equation (2). Black lines: M-R relationships using equations (3-4) with parameters from Table 2. Uranus, Neptune and GJ436 B are indicated by white dots. 27 Figure 5: Effect of composition on radii in the case of 10 ME. These curves are almost insensitive to planetary mass. R0 and Rc0 are the planetary and the metallic core radii respectively in the reference case (eq. 3-4) for a dry planet (Xw=0). For 10 ME, R0=12321 km and R c0= 5483 km. The radius of the planet is almost insensitive to the amount of volatiles in the core and the Mg#, but varies slightly (± 2%) with (Mg/Si) and (Fe/Si). Core radius depends strongly on the compositional parameters, and especially on Mg#. Figure 6: Top: The investigated thermal domain. Ninety thermal profiles have been generated randomly within the grey area by changing the adiabatic slope in each layer and the temperature drop through the interfaces. Bottom: Core and planetary radii for each thermal case assuming a 10 Earth mass planet. It shows that thermal constraints are of second order importance for describing M-R relation of massive exoplanets. Results do not depend on the mass of the planet. The reference radii R0 and Rc0 are defined as in figure 5. Figure 7: Top: M-R distribution of the 530 simulated Earth-like planets. The points are randomly distributed within the M-R domain described by equations (3-4). Bottom: Compositional variations of the simulated planets as a function of their mass. Compositions are fixed by the four independent parameters Mg#, Fe/Si, Mg/Si and Xw, which are allowed to vary within the ranges defined in Table 1. In addition, temperature profiles are varied randomly for each planet with respect to the standard deviations proposed in table 1 for both adiabatic slopes and temperature drops through the interfaces (not plotted). Figure 8: Top: Comparison between the amount of water Xw imposed in the model and the computed value wX assuming that the planet is in the reference case. Data are obtained using the 530 * simulated cases plotted in figure 7. Bottom: Gaussian distribution of the Xw- wX differences for * mod = xσ 045.0 . It illustrates the fact that, if radius and mass of a Super-Earth are exactly known, its amount of water can be estimated with an accuracy of 4.5 %. 28 Figure 9: Mass-radius relationships for solid planets composed of iron, silicates, and ices (shaded areas). The amount of water varies from 0 % to 100 % from bottom to top of the white domain limited by the silicates curves at the bottom and the pure ices curve at the top (see figure 4). Results from previous studies and for several pure compounds (H, He, C, Fe, Mg) are also plotted for a purpose of comparison. Position of GJ436b and solar gaseous planets are indicated by white circles. Figure 10: Accuracy of the water amount measurement using M-R relationships in the case of Super-Earths . Small numbers indicate the value of the standard deviation σXw/Xw depending on the precision on mass and radius measurements. The bottom left corner, (at 4.5 %), corresponds to the standard deviation related to the internal structure modelling (choice of thermal profiles, equation of states, compositions of silicate and iron alloy phases). 29 Allowed ranges ± 0.4 ± 0.6 unchanged unchanged ± 500 K ± 2 % ± 0.2 unchanged ± 200 K ± 2 % ± 0.2 unchanged ± 500 K unchanged unchanged unchanged ± 100 K ± 2 % unchanged unchanged ± 200 K ± 2 % Planet description Global composition Fe/Si Mg/Si %wt of water (Xw) Metallic core % volatiles EOS Upper Temp. drop Temperature profile Lower silicate mantle Mg# EOS Upper Temperature drop Temperature profile Upper silicate mantle Mg# EOS Upper Temperature drop Temperature profile Icy mantle % volatiles EOS Upper Temp. drop Temperature profile Liquid water % volatiles EOS Surface temperature Temperature profile Reference Case 1.01 1.34 0.1 - 80 none ANEOS 800 K adiabatic 0.8 ANEOS above 100 GPa MGD below 100 GPa 300 K adiabatic 0.8 3rd order Birch-Mürnaghan 1300 K adiabatic none MGD+ ANEOS+ transitional regime 100 K adiabatic none 3rd order Birch-Mürnaghan 300 K adiabatic Table 1: Input parameters and main characteristics of the reference case. It corresponds to a planet which possesses a global composition equal to the averaged ratio of the stars studied by Beirao et al. (2005) and Gilli et al. (2006). It is an “extreme case” because the metallic core is composed of pure iron (no volatiles). The third column indicates the ranges in which the different parameters have been randomly distributed for studying the accuracy of the model in section III. ξ2 ξ1 ξ0 α -1.704 10-5 β -9.015 10-7 γ 5.397 10-8 ε -3.166 10-10 5.714 10-3 -2.116 10-4 -4.221 10-6 3.085 10-8 1.010 2.859 10-1 -5.518 10-4 2.096 10-6 Table 2: Parameters used in equation 4. These values provide a description of M-R relations in the mass range [1-100 ME] for any amount of ices (from the Earth-like case to the pure water world). Metallic core Silicate mantle Water ices Figure 1 40 0.6 0.4 0.4 20 1 2 1,8 1,6 1,4 1,2 1 0,8 0,6 0,4 0,2 i S / e F 0.8 0.8 0.6 0.6 0.8 0 0,4 0,6 0,8 1 1,2 40 20 1 1.0 1.0 1.0 1,6 1,8 2 1,4 Mg/S i Figure 2 Mie-Grüneisen-Debye Thomas-Fermi-Dirac ANEOS a - iron 100000 D ensit y (k g/m 3 ) 10000 4000 100000 D ensit y (k g/m 3 ) b - forsterite 10000 2000 100000 D ensit y (k g/m 3 ) c - water 10000 a i n m o p e r c ritic a l d S u 1000 1 10 100 1000 P ressur e (GP a) 10000 100000 Figure 3 E R / R 4 3 2 1 1 100 GJ436b 80 U N 60 40 20 0 10 M/M E 100 Figure 4 1.01 0 R / R 1.00 1.01 0 R / R 1.00 0.99 1.02 1.01 1.00 0.99 1.01 1.00 0.99 0.98 0.97 0 R / R 0 R / R Core Planet 0.80 1.00 (F e/S i)* 1.20 1.40 1.00 1.20 (Mg/S i)* 1.40 1.60 %mol.F eS 0 0.1 0.2 0.3 Mg # 0.8 0.7 1.10 1.00 R c / R c 0.90 0 0.80 1.60 1.06 1.02 R c / R c 0.98 0 0.94 1.80 1.20 1.10 R c / R c 1.00 0 0.90 0.4 1.10 1.00 R c 0.90 / R c 0 0.80 0.9 0.70 1.0 Figure 5 10 000 ) K ( p m e T 5000 0 0 1.02 1.01 1.00 0 c R / c R 0.99 0.99 0.2 0.4 R/R 0 0.6 0.8 1.0 1.00 R/R 0 1.01 Figure 6 1 10 M/ME 100 R/RE ) t w % ( t n u o m a r e t a W i S / e F i S / g M 4 3 2 1 80 60 40 20 0 1.4 1.2 1.0 0.8 0.6 1.8 1.4 1.0 0.6 1.0 0.9 0.8 # g M 0.7 0.6 1 10 M/ME 100 Figure 7: Grasset et al. 1 0,8 0,6 0,4 0,2 0 W X 0 0,2 0,4 X * W 0,6 0,8 1 160 120 b N 80 40 0 -3 -2 -1 2 1 0 (XW -X * )W smo d x 3 Figure 8: Grasset et al. 10 9 8 7 6 5 4 3 2 1 R/RE S - 0.25 He 0.75 H 2 He GJ ice C Zapolski and Salpeter (1969) Fortney et al. (2007) Seager et al. (2007) Valencia et al. (2007b) This work H2 U N M g e t s a s ili c Fe Figure 9 1 10 M/ME 100 0.10 0.08 0.06 0.04 0.02 R / R s 0. 325 0. 30 0. 275 0. 25 0. 225 0. 20 0. 175 0. 15 0. 125 0. 10 0. 075 0. 05 0.00 0.00 0.02 0.04 0.06 sM/M 0.08 0.10 bbbb
1811.00935	1	1811	2018-11-02T15:29:30	The HST PanCET Program: Hints of Na I & Evidence of a Cloudy Atmosphere for the Inflated Hot Jupiter WASP-52b	[ "astro-ph.EP" ]	We present an optical to near-infrared transmission spectrum of the inflated hot Jupiter WASP-52b using three transit observations from the Space Telescope Imaging Spectrograph (STIS) mounted on the Hubble Space Telescope, combined with Spitzer/Infrared Array Camera (IRAC) photometry at 3.6 microns and 4.5 microns. Since WASP-52 is a moderately active (log(Lx/Lbol) = -4.7) star, we correct the transit light curves for the effect of stellar activity using ground-based photometric monitoring data from the All-Sky Automated Survey for Supernovae (ASAS-SN) and Tennessee State University's Automatic Imaging Telescope (AIT). We bin the data in 38 spectrophotometric light curves from 0.29 to 4.5 microns and measure the transit depths to a median precision of 90 ppm. We compare the transmission spectrum to a grid of forward atmospheric models and find that our results are consistent with a cloudy spectrum and evidence of sodium at 2.3-sigma confidence, but no observable evidence of potassium absorption even in the narrowest spectroscopic channel. We find that the optical transmission spectrum of WASP-52b is similar to that of the well-studied inflated hot Jupiter HAT-P-1b, which has comparable surface gravity, equilibrium temperature, mass, radius, and stellar irradiation levels. At longer wavelengths, however, the best fitting models for WASP-52b and HAT-P-1b predict quite dissimilar properties, which could be confirmed with observations at wavelengths longer than ~1 micron. The identification of planets with common atmospheric properties and similar system parameters will be insightful for comparative atmospheric studies with the James Webb Space Telescope.	astro-ph.EP	astro-ph	DRAFT VERSION NOVEMBER 5, 2018 Typeset using LATEX twocolumn style in AASTeX61 THE HST PanCET PROGRAM: HINTS OF Na I & EVIDENCE OF A CLOUDY ATMOSPHERE FOR THE INFLATED HOT JUPITER WASP-52b GREGORY W. HENRY,4 JORGE SANZ-FORCADA,5 MICHAEL H. WILLIAMSON,4 THOMAS M. EVANS,2 HANNAH R. WAKEFORD,6 GIOVANNI BRUNO,6, 7 GILDA E. BALLESTER,8 KEVIN B. STEVENSON,6 NIKOLE K. LEWIS,6, 9 JOANNA K. BARSTOW,10 MUNAZZA K. ALAM,1 , ∗ NIKOLAY NIKOLOV,2 MERCEDES LÓPEZ-MORALES,1 DAVID K. SING,2, 3 JAYESH M. GOYAL,2 VINCENT BOURRIER,11 LARS A. BUCHHAVE,12 DAVID EHRENREICH,11 AND ANTONIO GARCÍA MUÑOZ13 1Department of Astronomy, Harvard-Smithsonian Center for Astrophysics, Cambridge, MA 02138, USA 2Astrophysics Group, School of Physics and Astronomy, University of Exeter, Exeter EX4 4QL, UK 3Department of Earth and Planetary Sciences, Johns Hopkins University, Baltimore, MD 21218, USA 4Center of Excellence in Information Systems, Tennessee State University, Nashville, TN 37209, USA 5Centro de Astrobiología (CSIC-INTA), ESAC Campus, Villanueva de la Cañada, Madrid, Spain 6Space Telescope Science Institute, Baltimore, MD 21218, USA 7INAF-Osservatorio Astrofisico di Catania, via S. Sofia, 78, 95123 Catania, Italy 8Lunar and Planetary Laboratory, University of Arizona, Tucson, AZ 85721, USA 9Department of Astronomy and Carl Sagan Institute, Cornell University, Ithaca, NY 14853, USA 10Department of Physics and Astronomy, University College London, London WC1E 6BT, UK 11Observatoire de l'Université de Genève, Sauverny, Switzerland 12DTU Space, National Space Institute, Technical University of Denmark, Lyngby, Denmark 13Zentrum für Astronomie und Astrophysik, Technische Universität Berlin, Berlin, Germany ABSTRACT We present an optical to near-infrared transmission spectrum of the inflated hot Jupiter WASP-52b using three transit ob- servations from the Space Telescope Imaging Spectrograph (STIS) mounted on the Hubble Space Telescope, combined with Spitzer/Infrared Array Camera (IRAC) photometry at 3.6 µm and 4.5 µm. Since WASP-52 is a moderately active (log(Lx/Lbol) = −4.7) star, we correct the transit light curves for the effect of stellar activity using ground-based photometric monitoring data from the All-Sky Automated Survey for Supernovae (ASAS-SN) and Tennessee State University's Automatic Imaging Telescope (AIT). We bin the data in 38 spectrophotometric light curves from 0.29 to 4.5 µm and measure the transit depths to a median precision of 90 ppm. We compare the transmission spectrum to a grid of forward atmospheric models and find that our results are consistent with a cloudy spectrum and evidence of sodium at 2.3σ confidence, but no observable evidence of potassium absorp- tion even in the narrowest spectroscopic channel. We find that the optical transmission spectrum of WASP-52b is similar to that of the well-studied inflated hot Jupiter HAT-P-1b, which has comparable surface gravity, equilibrium temperature, mass, radius, and stellar irradiation levels. At longer wavelengths, however, the best fitting models for WASP-52b and HAT-P-1b predict quite dissimilar properties, which could be confirmed with observations at wavelengths longer than ∼1 µm. The identification of plan- ets with common atmospheric properties and similar system parameters will be insightful for comparative atmospheric studies with the James Webb Space Telescope. Keywords: planets and satellites: atmospheres -- planets and satellites: composition -- planets and satellites: individual (WASP-52b) 1. INTRODUCTION Corresponding author: Munazza Alam [email protected] ∗ National Science Foundation Graduate Research Fellow Transiting exoplanets offer unprecedented opportunities for the detection and detailed characterization of planets be- yond the Solar System (Struve 1952). From transit observa- tions, we can make inferences about the formation and evo- lutionary histories of these planets, their bulk compositions, and their atmospheres (Winn 2010). Atmospheric studies can 8 1 0 2 v o N 2 . ] P E h p - o r t s a [ 1 v 5 3 9 0 0 . 1 1 8 1 : v i X r a 2 ALAM ET AL. be performed using transmission spectroscopy to constrain atmospheric structure and chemical composition (e.g., Char- bonneau et al. 2002; Vidal-Madjar et al. 2003; Wakeford et al. 2017), secondary eclipses to measure temperature and ther- mal structure (e.g., Deming et al. 2005; Charbonneau et al. 2008; Sing & López-Morales 2009; Evans et al. 2017), and orbital phase curves to probe atmospheric circulation (e.g., Knutson et al. 2007; Stevenson et al. 2017). The subject of this paper is transmission spectroscopy (Seager & Sasselov 2000; Brown 2001). During transit, light from the host star passes through the atmosphere of the planet. At wavelengths where absorption by atoms, molecules, and aerosols takes place, the planet blocks slightly more stellar flux, resulting in variations in the apparent radius of the planet as a function of wavelength. These variations in planetary radius reveal the composition of the planetary atmosphere. The gaseous atmospheres of short-period hot Jupiters are most accessible to such observations because of their large scale heights and short orbital periods. Narrow peaks of H I (1215 Å) in the UV, Na I (5893 Å) and K I (7665 Å) in the optical, H2O (1.4 µm; 1.9 µm), CO (2.3 µm; 4.7 µm), CO2 (2 µm; 15 µm), and CH4 (2.2 µm; 7.5 µm) in the near-infrared, and scattering by molecular hydrogen (H2) are expected to be prominent features in clear hot Jupiter atmospheres (Vidal- Madjar et al. 2003; Seager & Sasselov 2000; Sudarsky et al. 2003; Burrows et al. 2010; Fortney et al. 2010). Charbon- neau et al. (2002) detected the first exoplanet atmosphere for the hot Jupiter HD 209458b, which was later confirmed to have absorption from Na I, H2 Rayleigh scattering, and possi- ble TiO/VO absorption (Sing et al. 2008; Snellen et al. 2008; Lecavelier Des Etangs et al. 2008; Désert et al. 2008; Sing et al. 2008). The first near-UV to IR (0.3-8.0 µm) transmission spectrum of the hot Jupiter HD 189733b (Pont et al. 2013) revealed clouds/hazes consistent with Rayleigh scattering by small condensate particles in addition to narrow peaks of Na I and K I, H2O absorption, and an escaping H atmosphere (Grillmair et al. 2008; Sing et al. 2009; Lecavelier Des Etangs et al. 2010; Bourrier et al. 2013). To date, a diversity of hot Jupiters with a continuum of clear to cloudy atmospheres (Sing et al. 2016) has been de- tected, with no apparent correlation between the observed spectra and other system parameters. Space-based atmo- spheric studies have yielded detections of Na I and K I (e.g., Charbonneau et al. 2002; Nikolov et al. 2014), elucidated the presence of thick atmospheric cloud decks (e.g., Sing et al. 2015), and have provided water abundance constraints (e.g., Kreidberg et al. 2014; Wakeford et al. 2018). From ground- based transmission spectral surveys of hot Jupiters, we have detected Na I (e.g., Nikolov et al. 2016, 2018; Wyttenbach et al. 2015, 2017), K I (e.g., Sing et al. 2011; Nikolov et al. 2016), cloudy/hazy atmospheres (e.g., Jordán et al. 2013; Mallonn et al. 2016; Huitson et al. 2017), and Rayleigh scat- tering slopes (e.g., Gibson et al. 2017). Here we present results for WASP-52b (Hébrard et al. 2013) from the Hubble Space Telescope (HST) Panchromatic Comparative Exoplanetology Treasury (PanCET) program (GO 14767; PIs Sing & López-Morales). The scientific goals of PanCET are to provide a uniform, statistically compelling ultraviolet (UV) through infrared (IR) study of clouds/hazes and chemical composition in exoplanet atmospheres, and as- semble a UVOIR legacy sample of exoplanet transmission spectra that will be well-suited for follow-up with the James Webb Space Telescope (JWST). WASP-52b is a 0.46 MJup and 1.27 RJup inflated hot Jupiter (Teq = 1300 K) orbiting a moderately active (log(R(cid:48) HK) = −4.4 ± 0.2, Hébrard et al. 2013; log(Lx/Lbol) = −4.7, Sec- tion 3.2) K2V star with a period of 1.75 days (Hébrard et al. 2013). This planet, at a spectroscopic parallax distance of 175.7±1.3 pc (Gaia Collaboration et al. 2018), is a favorable target for atmospheric studies via transmission spectroscopy due to its large scale height (H = 700 km) and deep transit (δ = 0.028). Based on its surface gravity (log(g) = 2.87 dex) and equilibrium temperature (Teq = 1315 K), WASP-52b is predicted to have a predominantly cloudy atmosphere with muted spectral features (Stevenson 2016). However, two re- cent ground-based atmospheric analyses of this target claim discrepant conclusions. Louden et al. (2017) cite an opti- cally thick cloud deck to explain an observed flat transmis- sion spectrum, but note that their results are inconsistent with deeper transit depths at longer wavelengths (Kirk et al. 2016). Conversely, Chen et al. (2017) report a cloudy atmosphere with a noticeable Na I detection and a weaker detection of K I absorption. In this work, we measure WASP-52b's transmission spec- trum over the ∼0.29−4.5 µm wavelength range by combin- ing HST/STIS and Spitzer/IRAC observations. An outline of the paper is as follows. In Section 2, we describe the ob- servations and data reduction techniques. In Section 3, we detail the stellar activity correction. The light curve fits and measurement of the transmission spectrum are described in Section 4. We compare our results to previously published measurements and to a grid of forward atmospheric models in Section 5, and present an interpretation of the transmission spectrum in Section 6. We summarize the paper in Section 7. 2. OBSERVATIONS & DATA REDUCTION 2.1. Observations We obtained time series spectroscopy during three tran- sits of WASP-52b with HST/STIS on UT 2016 November 01 and UT 2016 November 29 using the G430L grating (2892- 5700 Å) and UT 2017 May 11 using the G750L grating (5240-10270 Å). Two additional transits were observed with Spitzer/IRAC as part of GO program 13038 (PI Stevenson) WASP-52b 3 Table 1. Transit Observations of WASP-52b Obs Date Visit Number Telescope/Instrument Grating/Grism Number of Images Exposure Time (sec) UT 2016 Nov 01 52 UT 2016 Nov 29 53 UT 2017 May 11 54 − UT 2016 Oct 18 − UT 2018 Mar 22 a Central Wavelength: 4300 Å b Central Wavelength: 7751 Å HST/STIS HST/STIS HST/STIS Spitzer/IRAC Spitzer/IRAC G430La G430La G750Lb [3.6 µm] [4.5 µm] 37 37 37 28800 29300 253 253 253 1.92 1.98 in the 3.6 µm channel on UT 2016 October 18 and in the 4.5 µm channel on UT 2018 March 22. Table 1 summarizes the transit observations and instrument settings for each visit. 2.1.1. HST/STIS The G430L and G750L STIS data have resolving power R∼500 and each consist of 37 stellar spectra taken over four, consecutive 96-minute orbits. The visits were scheduled to include the transit event in the third orbit, providing an out- of-transit baseline time series before and after the transit as well as good coverage between second and third contact. We used a 128 pixel wide subarray and exposure times of 253 seconds to reduce the readout times between exposures. To minimize slit losses, the data were taken with the 52 x 2 arcsec2 slit. 2.1.2. Spitzer/IRAC Two transits of WASP-52b were observed on UT 2016 Oc- tober 18 and UT 2018 March 22 with the Spitzer space tele- scope (Werner et al. 2004) using the 3.6 µm and 4.5 µm IRAC channels (Fazio et al. 2004). Although these obser- vations cover the complete phase curve of the planet, for this work we only use a 6-hour portion of the phase curve cen- tered on the transit event with enough out-of-transit baseline flux to allow for accurate analysis. Each IRAC exposure had an effective integration time of ∼2 seconds, resulting in ∼30,000 images for the portion of the phase curve cor- responding to the transit. 2.2. Data Reduction 2.2.1. HST/STIS We reduced (bias-, dark- and flat-corrected) the raw 2D G430L and G750L spectra using the CALSTIS1 pipeline (version 3.4) and the relevant calibration frames. Follow- ing the procedure detailed in Nikolov et al. (2014), we used median-combined difference images to identify and correct 1 http://www.stsci.edu/hst/stis/software/ analyzing/calibration/pipe_soft_hist/intro.html for cosmic ray events and bad pixels flagged by CALSTIS. We extracted 1D spectra from the calibrated .flt science files using IRAF's APALL task. To identify the most appro- priate aperture, we extracted light curves of aperture widths ranging from 6 to 18 pixels with a step size of 1. We defined the best aperture for each grating according to the lowest pho- tometric dispersion in the out-of-transit baseline flux. Based on this criterion, we used an aperture size of 13 pixels in our analysis. For each exposure, we computed the mid-exposure time in MJD. Aperture extractions with no background subtraction min- imize the out-of-transit standard deviation of the white light curves (Sing et al. 2015). Although the STIS 2D spectra are known to show negligible background sky contribution (Sing et al. 2011, 2013; Huitson et al. 2012, 2013; Nikolov et al. 2015; Gibson et al. 2017), we assessed the potential bias of the sky background on the light curves by obtaining time series spectroscopy both with and without background sub- traction. Comparing the light curves, we find that both data sets are fully consistent. To obtain a wavelength solution, we used the x1d files from CALSTIS to re-sample all of the extracted spectra and cross-correlate them to a common rest frame. The cross-correlation measures the shift, and the spectra are re-sampled to align them and remove sub-pixel drifts in the dispersion direction. These drifts can be associ- ated with the different locations of the spacecraft on its orbit around the Earth (e.g., Huitson et al. 2013). 2.2.2. Spitzer/IRAC We analyzed the IRAC photometry following the method- ology of Nikolov et al. (2015) and Sing et al. (2015, 2016). We started our analysis using the Basic Calibrated Data (.bcd) files and converted the images from flux in mega- Jansky per steradian (MJy sr−1) to photon counts (i.e., elec- trons) by multiplying each image by the gain and exposure time and then dividing by the flux conversion factor. Follow- ing the reduction procedure outlined in Knutson et al. (2012) and Todorov et al. (2013), we filtered for outliers (hot or lower pixels in the data) by following each pixel in time. We scanned the images in two passes: first by removing outliers 4 ALAM ET AL. Table 2. Summary of Photometric Observations for WASP- 52b Observing Nobs Date Range Sigma Seasonal Mean Season (HJD - 2,450,000) (mag) (mag) AIT 2014−2015 2015−2016 2016−2017 61 48 27 56943−57066 57293−57418 57705−57785 0.14288 0.15069 0.16062 0.01270 0.00892 0.01335 ASAS-SN 2013−2014 2014−2015 2015−2016 2016−2017 154 175 209 202 56638−57031 57145−57385 57507−57749 57878−58115 0.16692 0.15601 0.11866 0.11769 0.01162 0.01558 0.01195 0.01167 ≥8σ away from the median value of each frame compared to the 10 surrounding images and then by removing outliers above the 4σ level following the same procedure. The total fraction of corrected pixels was 0.05%. We estimated and subtracted the sky background for each image using an iterative 3σ clipping procedure in which we excluded all pixels associated with the stellar point spread function (PSF), background stars, or hot pixels. In the last iteration, we created a histogram from the remaining pixels and determined the sky background based on a Gaussian fit to the distribution of remaining pixels. To locate the center of the PSF for each image, we used the flux-weighted centroid- ing method with a 5-pixel radius circular region centered on the approximate position of the star. The variation of the x and y positions of the PSF on the detectors were measured to be 0.19 and 0.24 pixels, respectively. We extracted photometric points following fixed and time- variable photometry. In the fixed approach, we used circular apertures ranging in radius from 4 to 8 pixels in increments of 0.5. For the time-variable photometry, the aperture size was scaled by the value of the noise pixel parameter (the normal- ized effective background area of the IRAC point response function), which depends on the full-width-at-half-maximum (FWHM) of the stellar PSF squared (Mighell 2005; Knutson et al. 2012; Lewis et al. 2013; Nikolov et al. 2015). We identi- fied the best results from both photometric methods by com- paring the light curve residual dispersion, as well as the white and red noise components measured with the wavelet tech- nique detailed in Carter, & Winn (2009). The time-variable method resulted in the lowest white and random red noise correlated with data points co-added in time. 3. STELLAR ACTIVITY CORRECTION Since WASP-52 is a moderately active star (log(R(cid:48) HK) = −4.4±0.2, Hébrard et al. 2013), we used ground-based activ- ity monitoring data to track stellar activity levels during the epochs of our transits. Before fitting the transit light curves, we corrected the baseline flux levels for the effect of stellar variability using a quasi-periodic Gaussian process regres- sion model and corrected for the effect of unocculted stel- lar spots following the prescription of Huitson et al. (2013). Table 2 summarizes the photometric monitoring campaigns, and Table 3 includes the average flux correction for our tran- sit observations. 3.1. Stellar Variability Monitoring We acquired 135 out-of-transit R-band images over the 2014-2015, 2015-2016, and 2016-2017 observing seasons with Tennessee State University's 14-inch Celestron Auto- matic Imaging Telescope (AIT). These observations, how- ever, do not include the epochs of the Spitzer and ground- based (Chen et al. 2017; Louden et al. 2017) transit observa- tions. We therefore used 740 out-of-transit V -band images from Ohio State University's All-Sky Automated Survey for Supernovae2 (ASAS-SN) program (Shappee et al. 2014; Kochanek et al. 2017) for more detailed activity monitor- ing coverage. The ASAS-SN observations were taken over the 2013−2014, 2014−2015, 2015−2016, and 2016−2017 ob- serving seasons. The number of observations, date range, mean magnitude, and standard deviation of the data taken in each observing season are included in Table 2. 3.2. Estimating Activity Levels To quantify the level of activity in WASP-52, we used ob- servations from the Advanced CCD Imaging Spectrometer (ACIS) on the Chandra X-ray telescope. These data were taken from the Chandra public archive (GO 15728; PI Wolk). Standard CIAO3 tasks were applied to reduce the data. The software ISIS (Houck, & Denicola 2000) was used to fit the spectrum with a metallicity fixed to photospheric values ([Fe/H]=0.03) and the ISM absorption assumed to be NH = 4 × 20 cm−3 (based on target location and distance), consistent with the fit. The resulting 1-T fit has log T (K)= 6.54+1.06 −0.29 and log EM (cm−3) = 51.14+0.33 −0.63. We measured an X-ray luminos- ity of Lx = (3.5 ± 1.0) ×1028 erg s−1 in the 0.12−2.48 keV band (Sanz-Forcada et al., in prep.). The light curve shows variability, but poor statistics hamper a detailed analysis. The corresponding log(Lx/Lbol) = −4.7 indicates that WASP-52 is a moderately active star. Further details will be included in Sanz-Forcada et al. (in prep.) 2 https://asas-sn.osu.edu/ 3 CIAO (Chandra Interactive Analysis of Observations) 4.9 and CALDB 4.7.4 versions were used in the analysis (Fruscione et al. 2006). WASP-52b 5 3.3. Modeling the Variability Monitoring Data We initially adopted a simple sinusoidal model of period P = 11.8 ± 3.3 days based on the Hébrard et al. (2013) ro- tational period, but found that this approach did not accu- rately model the variations in amplitude and period of the AIT and ASAS-SN variability monitoring data over all ob- serving seasons. Discrepancies between the sinusoidal model and the photometric monitoring data are likely due to differ- ent spot configurations at different observed epochs (Dang et al. 2018), suggesting that a quasi-periodic model would more accurately fit the data. We therefore jointly modeled the AIT and ASAS-SN ground-based stellar activity data using a Gaussian process (GP) regression, a framework which has been shown to ac- curately disentangle stellar activity signals from planetary signals in radial velocity data (e.g., Aigrain et al. 2012; Hay- wood et al. 2014) and photometry (e.g., Pont et al. 2013; Aigrain et al. 2016; Angus et al. 2018). We ran a GP optimization routine (Ambikasaran et al. 2015) using the George package for Python. We used a three component kernel to model the quasi-periodicity of the ground-based activity monitoring data, irregularities in the amplitude of the ground-based photometry, and stellar noise. See Ap- pendix A.1 for further details regarding the functional form of our chosen kernel. We use a gradient based optimiza- tion routine to find the best-fit hyperparameters and set the 11.8±3.3 day rotation period from Hébrard et al. (2013) as a uniform prior with an uncertainty three times larger than the literature value. Figure 1 shows the ground-based variabil- ity monitoring data overplotted with the Gaussian process model for all seasons modeled jointly and for each observing season separately. The GP regression model with a quasi-periodic kernel re- produces the observed flux variations well for epochs during which we have stellar activity data, but has large uncertain- ties outside of a given season. The model accurately predicts the stellar activity behavior for each observing season when fit separately, but this model prediction has a significantly larger 1σ uncertainty when fitting the data from all seasons together. We therefore modeled the data for each observing season separately, and used the amplitude of the photometric variation given by the GP model for each epoch to correct the transmission spectrum for the effects of stellar activity (see Section 3.4). 3.4. Correcting for Unocculted Spots The temperature difference between the stellar photo- sphere and the spotted region introduces a slope in the planet spectrum (e.g., Kreidberg 2017), so it is necessary to correct for this effect. Using the results of the GP regression model described in Section 3.3, we followed the prescription of Huitson et al. (2013) to correct the spectrophotometric light Table 3. Stellar flux values (normal- ized by the non-spotted stellar flux) for each transit observation & aver- age flux correction for each band- pass Instrument fnorm Error ∆ f STIS (visit 52) STIS (visit 53) STIS (visit 54) IRAC 0.978 0.955 0.972 0.959 0.009 0.016 0.016 0.009 0.039 0.039 0.022 0.006 curves for the effect of unocculted stellar spots of a fixed size. This method involves estimating the variability ampli- tude for a broadband wavelength value (determined by the photometric filter of the activity monitoring data), which can then be used as an anchor for the wavelength-dependent flux correction on the HST and Spitzer data. We first converted the ASAS-SN and AIT photometric variability monitoring data to relative flux. Excluding in- transit measurements, we estimated the non-spotted stellar flux to be F(cid:63) = max(F) + kσ, where F is the variability mon- itoring data, σ is the dispersion of the photometric measure- ments, and k is a factor fixed to unity (Aigrain et al. 2012; see also Appendix A.2). The variability monitoring data was then normalized to the non-spotted stellar flux to estimate the amount of dimming. Table 3 gives the dimming values for each transit observation. We also computed the amplitude of the spot corrections at the variability monitoring wavelength ∆ f0 = 1 − fnorm, where fnorm is the mean of the normalized flux array. To derive the wavelength-dependent flux correction, we used a stellar flux model (Te f f = 5000 K, log(Z) = −1.5, log(g) = 4.5) and a spot model (Te f f = 4750 K, log(Z) = −1.5, log(g) = 4.5) from the Kurucz (1993) 1D ATLAS grid4 of stellar at- mospheric models. The stellar flux model was chosen based on the effective temperature, metallicity, and surface gravity of WASP-52 given in Hébrard et al. (2013). The spot model is the same as the stellar model but 250 K cooler (Berdyug- ina 2005). To select this spot model, we tested different cold spots ranging from 3500−4750 K (corresponding to temper- ature differences between the spot and stellar photosphere from 1500 K to 250 K). Figure 2 shows that spots at colder temperatures exhibit less flux dimming at longer wavelengths and a weaker slope in the optical. To correct for the effect of unocculted spots, we therefore use a spot model at 4750 K for 4 http://kurucz.harvard.edu/ 6 ALAM ET AL. which cold spots give the strongest slope to account for the maximum possible contribution from star spots in the data. We interpolated the stellar model to the spot model grid and computed the wavelength-dependent correction factor derived in Sing et al. (2011): (cid:17)(cid:30)(cid:16) (cid:17) (cid:16) f (λ,T ) = 1 − Fλ,Tspot Fλ,Tstar 1 − Fλo,Tspot Fλo,Tstar (1) where Fλ,Tspot is the stellar model flux at temperature Tspot and at the wavelength of the transit observations, Fλ,Tstar is the stellar model flux at the wavelength of the transit observa- tions and at temperature Tstar, Fλo,Tspot is the stellar model flux at temperature Tspot at the reference wavelength of the activ- ity monitoring data, and Fλo,Tstar is the stellar model flux at temperature Tstar at the activity monitoring reference wave- length. The final flux dimming correction was then calcu- lated as ∆ f = ∆ f0 × f (λ,T ). We computed the average flux correction over each band- pass (see Table 3) and applied the correction to each spec- trophotometric light curve using: ycorrected = y + ∆ f (1 − ∆ f ) yoot (2) where ycorrected is the corrected light curve flux, y is the orig- inal (uncorrected) light curve, and yoot is the mean of the out of transit exposures. We then fit analytic transit light curve models (Mandel & Agol 2002) to the stellar activity corrected light curves, as detailed in Section 4. 4. LIGHT CURVE FITS We extracted the broadband transmission spectrum and fit the spectrophotometric light curves following the meth- ods described in Sing et al. (2011, 2013) and Nikolov et al. (2014), and they are briefly summarized here. To simulta- neously fit for the transit and systematic effects, we mod- eled each of the STIS and Spitzer transit light curves with a two-component function consisting of a transit model multi- plied by a systematics model. We adopted the complete an- alytic transit models of Mandel & Agol (2002), which are parametrized by the mid-transit time T0, orbital period P, inclination i, normalized planet semi-major axis a/R(cid:63), and planet-to-star radius ratio Rp/R(cid:63). 4.1. STIS White Light Curves We produced the STIS broadband (wavelength-integrated) light curves by summing the time series over the complete wavelength range of the bandpass (2892−5700 Å for the G430L grating; 5240−10270 Å for the G750L grating). The white light curves for each of the STIS visits are shown in Figure 3. Photometric uncertainties were derived based on pure photon statistics. The raw white light curves exhibited instrumental systematics related to the orbital motion of the spacecraft (Gilliland et al. 1999; Brown 2001). In particular, the HST focus is known to experience significant variations on the spacecraft orbital time-scale resulting from thermal expansion/contraction during the spacecraft's day/night or- bital cycle. As in past studies (e.g., Huitson et al. 2013; Sing et al. 2013; Nikolov et al. 2014), we accounted for and detrended these instrumental systematic effects by fitting a fourth-order polynomial to the flux dependence on HST orbital phase. We excluded the first orbit and the first exposure of each subse- quent orbit in accordance with common practice, since these data have unique, complex systematics (see Figure 3) and were taken while the telescope was thermally relaxing into its new pointing position. We applied orbit-to-orbit flux cor- rections to the STIS data by fitting a polynomial of the space- craft orbital phase (φt), drift of the spectra on the detector (x and y), the shift of each stellar spectrum cross-correlated with the first spectrum of the time series (ω), and time (t). We then generated systematics models spanning all possi- ble combinations of detrending variables (see Appendix B.1 and Table B1 for details), and performed separate fits using each systematics model included in the two-component func- tion. For each model, we fixed P, i, and a/R(cid:63) to the values given in Hébrard et al. (2013), assumed zero eccentricity, and fit for T0, Rp/R(cid:63), stellar baseline flux, and instrument sys- tematic trends. We determined the best fit parameters of the two-component model using a Levenberg-Marquardt least- squares fitting routine (Markwardt 2009) to derive the system parameters and calculated the Akaike Information Criterion (AIC; Akaike 1974) for each function. The results of these fits are shown in Table 4. The STIS stellar spectra and raw and detrended white light curves for each visit are shown in Figure 3. We marginalized over the entire set of functions following the framework outlined in Gibson (2014). Marginalization over multiple systematics models assumes equally weighted priors for each model tested. Table B2 shows the results for each systematics model using the STIS white light curves. The reduced chi-squared χr from the fits can be used as a proxy for the photon noise level of a given data set (Nikolov et al. 2014). We selected the systematics model for use in de- trending the STIS white light curves based on the lowest AIC value, since the light curves show no significant correlation to the quantified systematic parameters (Nikolov et al. 2014). 4.2. STIS Spectrophotometric Light Curves We produced the spectrophotometric light curves by divid- ing the spectra into 17 − 400 Å width bins and integrating the flux from each bandpass. The width of the bins is determined primarily by the need to achieve a given photometric preci- sion to be sensitive to features in the planet's transmission spectrum that are comparable in amplitude to an atmospheric WASP-52b 7 Table 4. Derived System Parameters for WASP-52b Hébrard et al. (2013) 1.7497798 ± 0.0000012 Chen et al. (2017) Louden et al. (2017) This worka 1.7497798 (fixed) 1.74978089 ± 0.00000013 1.749779800 (fixed) 85.17 ± 0.13 0.0 (fixed) 7.22 ± 0.07 0.1639 ± 0.0005 Period P [days] Orbital inclination i [◦] Orbital eccentricity e Scaled semi-major axis a/R(cid:63) 85.35 ± 0.20 0.0 (fixed) 7.38 ± 0.10 85.06 ± 0.27 0.0 (fixed) 7.14 ± 0.12 85.33 ± 0.22 0.0 (fixed) 7.23 ± 0.12 Radius ratio Rp/R(cid:63) 0.1741 ± 0.0063 a The values reported here are the weighted mean of fitted system parameters from the Spitzer observations. 0.1646 ± 0.0020 0.1608 ± 0.0018 scale height. The smaller bin sizes were chosen to be cen- tered at specific absorption features, such as Na I at 5893 Å and K I at 7665 Å. We modeled the systematic errors using two methods. In the first approach, we independently fit each of the binned light curves with the same family of transit+systematics mod- els (see Appendix B.1) as the broadband light curve. The only differences are that we fixed the mid-transit time T0 and the scaled semi-major axis a/R(cid:63) to the white light curve best fit values. We also fixed the limb darkening coefficients (computed following the procedure outlined in Section 4.4) to the derived theoretical values. In the second approach, we performed a common mode correction to remove color- independent systematic trends from each spectral bin. Then, we fit the common mode corrected spectroscopic light curves by fitting for residuals with a parametrized model of six fewer free parameters (c1 −c4, T0, and a/R(cid:63)) and marginalizing over the entire set of functions defined in Appendix B.1. The common mode trends are computed by dividing the raw flux of the white light curve in each grating by the best fit ana- lytic transit model. We applied the common mode correction by dividing each binned light curve by the derived common mode flux. Removing common mode trends is known to re- duce the amplitude of the observed HST breathing system- atics, since these trends are similar in wavelength across the detector (Sing et al. 2013; Nikolov et al. 2014). The common mode corrected light curves are shown in Figure 3. Both methods produced similar results (i.e., consistent baseline in Rp/R(cid:63)). Since the common mode correction pro- duces lower dispersion in the spectrophotometric light curves and smaller Rp/R(cid:63) uncertainties (Nikolov et al. 2015), we re- port the common mode corrected results for the fitted Rp/R(cid:63) (before and after applying the correction for unocculted spots discussed in Section 3.4) and non-linear limb darkening co- efficients for each spectroscopic channel in Table 5. The raw and detrended STIS spectrophotometric light curves for the G430L and G750L gratings are shown in Figures 4, 5, and 6. 4.3. IRAC Light Curves (2011, 2013) and Nikolov et al. (2014). To correct for flux variations from intrapixel sensitivity, we fit a polynomial to the stellar centroid position (Reach et al. 2005; Charbonneau et al. 2005, 2008; Knutson et al. 2008). This technique is ef- fective on short timescales (<10 hours) and for small (<0.2 pixels) variations in the stellar centroid position (Lewis et al. 2013). We corrected for systematic effects using a model given by the linear combination: f (t) = a0 + a1x + a2x2 + a3y + a4y2 + a5xy + a6t (3) where f (t) is the stellar flux as a function of time, the coef- ficients a0 through a6 are free fitting parameters, x and y are the detector positions of the stellar centroid, and t is time. We generate all possible model combinations of Equation 3, which we marginalize over using the Gibson (2014) proce- dure as detailed in Section 4.1 and Appendix B.1. Using the WASP-52 system parameters from Hébrard et al. (2013) as priors, we jointly fit for all parameters and find that our re- sults (see Table 4) agree within 1σ with the STIS white light curve analysis. To measure Rp/R(cid:63) for the transmission spec- trum, we fixed the orbital period P, normalized planet semi- major axis a/R(cid:63), inclination i, and central transit time T0 to the best fit values from the joint fit. The limb darkening co- efficients were also fixed to their theoretical values based on 3D stellar atmosphere models (see Section 4.4). The mea- sured planetary radius and limb darkening coefficients are included in Table 5. Figures 7 and 8 show the raw and de- trended Spitzer transit light curves. 4.4. Limb Darkening Models We modeled the limb darkening of WASP-52b using the four parameter non-linear limb darkening law (Claret 2000) given by: 4(cid:88) n=1 I(µ) I(1) = 1 − cn(1 − µn/2) (4) We modeled the 3.6 µm and 4.5 µm IRAC transit photom- etry in accordance with the methods outlined in Sing et al. where I(1) is the intensity at the center of the stellar disk, cn (n = 1−4) are the limb darkening coefficients, and µ = cos(θ), 8 ALAM ET AL. where θ is the angle between the normal to the stellar surface and the line of sight. To derive the stellar limb darkening coefficients, we fol- lowed the procedure described in Sing (2010) and initially used values for the four limb darkening coefficients based on 1D ATLAS theoretical stellar models (Kurucz 1993). We then derived the limb darkening coefficients from 3D stellar models (Magic et al. 2015) and compared to the 1D results to eliminate the known wavelength-dependent degeneracy of limb darkening with transit depth (Sing et al. 2008). This ap- proach reduces the number of free parameters in the fit (typi- cally four parameters per grating), but may cause an underes- timation of errors in the derived spectrum. The derived non- linear 3D limb darkening coefficients for each spectrophoto- metric light curve are shown in Table 5. 5. RESULTS The broadband STIS+Spitzer transmission spectrum for WASP-52b corrected for stellar activity and compared to the- oretical forward atmospheric models (Goyal et al. 2018) and past transmission spectrum measurements (Chen et al. 2017; Louden et al. 2017) is shown in Figure 9. The transmission spectrum shows evidence of Na I absorption (5893 Å) at 2.3σ confidence and no observable detection of K I absorption. To visualize the amplitude of the spot corrections, we also compare the raw transmission spectrum (before applying the stellar activity correction) and the spot corrected spectrum in Figure 10. 5.1. Constraints on Na I & K I We inspect the presence and significance of the Na I and K I features in the WASP-52b transmission spectrum using a grid of spectrophotometric channels ranging in width from 30−255 Å in steps of 15 Å, centered on the Na I (5893 Å) and K I (7665 Å) resonance doublets. The minimum bin size (30 Å) is defined to include both doublet lines. If a planetary atmospheric signal is present, this binning scheme should demonstrate a gradual decay in the measured transit depth for larger bin sizes. The results of this analysis are shown in Figure 11. For the Na I feature, we note a gradual decrease in the mea- sured transit depth for bins 30−100 Å wide. For wider bins, the signal is largely washed out and the transit depth remains unchanged within the uncertainties. This trend is expected when observing a narrow absorption peak and is consistent with the presence of a cloud deck. By measuring the dif- ference in transit depth between the narrowest (30 Å) bin and the flat transit depth baseline, we detect the core of the Na I doublet at 2.3σ confidence. In the case of K I, we do not see evidence of absorption even in the narrowest spectro- scopic channel, suggesting that this feature is either masked by a thick cloud deck or not as abundant in the atmosphere of WASP-52b. For further discussion, see Section 6.1. 5.2. Fits to Forward Atmospheric Models We fit the combined STIS+Spitzer transmission spectrum to the publicly available grid of forward model transmis- sion spectra (Goyal et al. 2018) produced using the ATMO 1D radiative-convective equilibrium model (Amundsen et al. 2014; Tremblin et al. 2015, 2016; Drummond et al. 2016). The models are generated for the parameters (e.g., mass, ra- dius, gravity, etc.) of WASP-52b. The grid5 includes 3,920 model transmission spectra of WASP-52b for five temperatures (1015 K, 1165 K, 1315 K, 1465 K, 1615 K), seven metallicities (0.005, 0.1, 1, 10, 50, 100, 200× solar), seven C/O ratios (0.15, 0.35, 0.56, 0.70, 0.75, 1.0, and 1.5), four values of the haziness parameter αhaze (1, 10, 150, and 1100), and four values of the cloudiness parameter αcloud (0, 0.06, 0.2, and 1). The parameter αhaze is a proxy for the haze enhancement factor of small scattering aerosol particles suspended in the atmosphere, where αhaze=1 indicates no haze and αhaze=1100 indicates thick hazes. The cloudiness parameter αcloud gives the strength of gray scatter- ing due to H2 at 350 nm, with αcloud=0 corresponding to no clouds and αcloud=1 corresponding to a thick cloud deck. See Goyal et al. (2018) and references therein for further details. The transmission spectra are computed assuming isothermal pressure−temperature (P−T ) profiles and condensation with- out rainout (local condensation). We computed the mean model prediction for the wave- length range of each spectroscopic channel (see Table 5), and performed a least-squares fitting of the band-averaged model to the spectrum. For the fitting procedure, we allowed the vertical offset in Rp/R(cid:63) between the spectrum and model to vary while holding all other parameters fixed in order to pre- serve the model shape. The number of degrees of freedom for each model is n − m, where n is the number of data points and m is the number of fitted parameters. Since n = 38 and m = 1, the number of degrees of freedom for each model is con- stant. From the fits, we computed the χ2 statistic to quantify our model selection. Figure 9 shows the best fit model, representative clear and hazy models, and a flat model compared to the observed transmission spectrum. The best fitting model (χ2=39.3) is cloudy (αcloud=1.0) and slightly hazy (αhaze=10) with a 2.3σ signature of Na I absorption, a temperature of T = 1315 K, solar metallicity ([M/H]=0.0), and slightly super-solar C/O (C/O=0.70). The selected clear model (χ2=49.5) has a lower temperature (T = 1015 K) and no clouds (αcloud=0.00) or hazes (αhaze=1). The representative hazy model (χ2=43.3) is similar to the clear model, but with extreme haziness 5 https://bd-server.astro.ex.ac.uk/exoplanets/ WASP-52/ WASP-52b 9 Table 5. Broadband transmission spectrum results for WASP-52b for the STIS G430L & G750L and Spitzer IRAC data λ (Å) (Rp/R∗)uncorr (Rp/R∗)corr c1 c2 c3 c4 2900−3700 3700−3950 3950−4113 4113−4250 4250−4400 4400−4500 4500−4600 4600−4700 4700−4800 4800−4900 4900−5000 5000−5100 5100−5200 5200−5300 5300−5400 5400−5500 5500−5600 5600−5700 5700−5800 5800−5878 5878−5913 5913−6070 6070−6200 6200−6300 6300−6450 6450−6600 6600−6800 6800−7000 7000−7200 7200−7450 7450−7645 7645−7720 7720−8100 8100−8485 8485−8985 8985−10300 36000 45000 0.16957 ± 0.00385 0.16684 ± 0.00201 0.16703 ± 0.00157 0.16871 ± 0.00102 0.16672 ± 0.00116 0.16618 ± 0.00112 0.16685 ± 0.00113 0.16827 ± 0.00089 0.16625 ± 0.00120 0.16630 ± 0.00088 0.16557 ± 0.00119 0.16806 ± 0.00085 0.16888 ± 0.00120 0.16769 ± 0.00074 0.16589 ± 0.00101 0.16802 ± 0.00091 0.16665 ± 0.00083 0.16621 ± 0.00086 0.16782 ± 0.00172 0.16556 ± 0.00195 0.17087 ± 0.00209 0.16643 ± 0.00112 0.16515 ± 0.00105 0.16611 ± 0.00160 0.16590 ± 0.00095 0.16371 ± 0.00087 0.16633 ± 0.00103 0.16419 ± 0.00170 0.16491 ± 0.00091 0.16448 ± 0.00068 0.16656 ± 0.00126 0.16113 ± 0.00312 0.16805 ± 0.00121 0.16538 ± 0.00081 0.16572 ± 0.00096 0.16607 ± 0.00089 0.16305 ± 0.00050 0.16390 ± 0.00110 0.15421 ± 0.00677 0.15781 ± 0.00296 0.16681 ± 0.00296 0.16457 ± 0.00159 0.16394 ± 0.00148 0.16101 ± 0.00142 0.16358 ± 0.00149 0.16583 ± 0.00106 0.16211 ± 0.00187 0.16290 ± 0.00115 0.16079 ± 0.00156 0.16463 ± 0.00105 0.16623 ± 0.00175 0.16541 ± 0.00103 0.16330 ± 0.00114 0.16416 ± 0.00132 0.16279 ± 0.00108 0.16157 ± 0.00154 0.16566 ± 0.00170 0.16346 ± 0.00192 0.16858 ± 0.00206 0.16434 ± 0.00110 0.16310 ± 0.00105 0.16407 ± 0.00157 0.16388 ± 0.00094 0.16179 ± 0.00086 0.16439 ± 0.00102 0.16231 ± 0.00168 0.16307 ± 0.00090 0.16268 ± 0.00067 0.16479 ± 0.00126 0.15943 ± 0.00309 0.16632 ± 0.00121 0.16372 ± 0.00080 0.16413 ± 0.00095 0.16459 ± 0.00088 0.16305 ± 0.00050 0.16390 ± 0.00110 0.4371 0.7648 0.4778 0.4831 0.6151 0.4691 0.4777 0.5977 0.4411 0.5159 0.4399 0.5409 0.5605 0.5381 0.5732 0.5975 0.6370 0.5662 0.5828 0.6127 0.6511 0.6208 0.6192 0.6439 0.6508 0.6569 0.6561 0.6494 0.6824 0.6914 0.6972 0.7151 0.6976 0.7008 0.7212 0.7134 0.4935 0.5344 -0.7679 -1.1573 -0.6459 -0.6750 -0.8347 -0.4858 -0.4094 -0.6932 -0.2389 -0.3136 -0.1314 -0.3989 -0.4363 -0.2913 -0.3581 -0.4018 -0.4886 -0.2476 -0.2962 -0.3245 -0.5383 -0.3332 -0.3316 -0.3583 -0.3875 -0.3766 -0.3747 -0.3485 -0.4389 -0.4691 -0.4678 -0.5133 -0.4689 -0.4971 -0.5332 -0.5378 -0.2505 -0.5777 1.6319 1.7190 1.6564 1.7025 1.7798 1.5912 1.5029 1.6924 1.1548 1.2538 1.0138 1.1899 1.0910 1.0934 1.1751 1.1388 1.2206 0.9413 1.0423 0.9821 1.2452 0.9989 0.9650 0.9653 0.9927 0.9577 0.9214 0.8652 0.9298 0.9457 0.9219 0.9350 0.8838 0.8888 0.8718 0.8618 0.1831 0.5534 -0.3645 -0.4157 -0.5511 -0.5879 -0.6519 -0.6570 -0.6494 -0.6811 -0.4505 -0.5560 -0.4335 -0.4566 -0.3757 -0.4736 -0.5264 -0.4758 -0.5147 -0.4076 -0.4802 -0.4278 -0.5387 -0.4502 -0.4310 -0.4321 -0.4467 -0.4418 -0.4106 -0.3871 -0.4051 -0.4158 -0.4080 -0.4061 -0.3862 -0.3849 -0.3765 -0.3737 -0.0638 -0.1991 (αhaze=1100). The flat model (χ2=52.2) represents a feature- less (gray) spectrum. The χ2 contour plot for the model grid fits is shown in Fig- ure 12. The grid provides constraints on the atmospheric and physical parameters of WASP-52b, with the 1σ confidence region favoring high cloudiness, slight haziness, solar metal- licity, slightly super-solar C/O ratio (>0.56) and an equilib- rium temperature of 1315 K. 5.3. Comparison with Previous Results In addition to the combined STIS+Spitzer transmission spectrum we report here, there are two ground-based op- tical transmission spectrum measurements for WASP-52b. Most recently, Louden et al. (2017) used spectroscopy be- tween 4000 and 8750 Å for two transit observations from the ACAM instrument mounted on the William Herschel Tele- scope (WHT). Chen et al. (2017) observed one transit with the Gran Telescopio Canarias's (GTC) OSIRIS instrument in 10 ALAM ET AL. Table 6. System parameters and best fitting model param- eters for WASP-52b & HAT-P-1b (Nikolov et al. 2014) Spectral Type Stellar mass M(cid:63) (M(cid:12)) Stellar radius R(cid:63) (R(cid:12)) Surface gravity log(g(cid:63)) (cgs) Stellar Temperature Te f f (K) Metallicity [Fe/H] Stellar irradiation I (erg/cm2s) log(R(cid:48) HK) Mp (MJ) Rp (RJ) ρ (ρJ) WASP-52b HAT-P-1b K2V 0.87 ± 0.03 0.79 ± 0.02 4.58 ± 0.014 5000 ± 100 0.03 ± 0.12 6.5±0.4 × 108 −4.4± 0.2 0.434 ± 0.024 1.253 ± 0.027 0.206 ± 0.009 G0V 1.15 ± 0.05 1.17 ± 0.03 4.36 ± 0.01 5980 ± 50 0.13 ± 0.01 7.0±0.4 × 108 −4.98± 0.1 0.525 ± 0.019 1.319 ± 0.019 0.213 ± 0.010 Best fitting ATMO models Teq (K) Fe/H C/O αcloud αhaze 1315 0.0 0.70 1.0 10 1322 1.0 0.15 0.20 10 the 5220−9030 Å wavelength range. We show these trans- mission spectra compared to our STIS+Spitzer results in Fig- ure 9. With the WHT/ACAM observations, Louden et al. (2017) modeled spot-crossing events via Gaussian processes, adopted a harmonic analysis of ground-based photometric monitoring, and found varying levels of activity over time with evidence of differential rotation. These results reveal a flat transmission spectrum attributed to an optically thick cloud deck. The GTC/OSIRIS observations indicate a cloudy atmosphere with a ∼3σ detection of Na I and a weaker detec- tion of K I. Calculations of the integrated absorption depth for the Na I and K I signals suggest an inverted temperature structure for the upper atmosphere of WASP-52b (Chen et al. 2017). Our Rp/R(cid:63) baseline is consistent within 1σ with the ground-based transmission spectrum from WHT/ACAM (Louden et al. 2017). The most significant difference is that the WHT spectrum does not show any variation in Rp/R(cid:63) around 5893 Å. If there is a weak signal of Na I absorption from the planet, the resolution of the spectrum, comprised of equally sized spectrophotometric bins of width 250 Å, may be washing it out as illustrated in Figure 11. The baseline of our spectrum matches less well with the ground-based GTC/OSIRIS transmission spectrum (Chen et al. 2017). We note a constant offset in the absolute mea- sured transit depths of our STIS+Spitzer spectrum and the GTC measurement, with a difference in the Rp/R(cid:63) baseline of ∼3σ. The authors attribute their shallower transit depth measurement compared to previous studies (e.g., Hébrard et al. 2013; Kirk et al. 2016; Mancini et al. 2017) to the effects of stellar activity. We find evidence of a Na I signal that is consistent with the GTC detection within 1σ but our spec- trum shows no evidence of K I absorption, contrary to the Chen et al. (2017) result. This discrepancy could be due to the different methods used to correct for the effects of stellar activity (see Section 3 and c.f. Chen et al. 2017). 5.4. Comparison to HAT-P-1b We compared WASP-52b to the well-studied inflated hot Jupiter HAT-P-1b (c.f., Nikolov et al. 2014), since both planets have comparable system parameters and atmospheric properties. HAT-P-1b and WASP-52b have overlapping sur- face gravity, equilibrium temperature, mass, radius, and stel- lar irradiation, and the transmission spectra of both planets are marginally flat with evidence of Na I absorption but no observable K I absorption. Table 6 compares the stellar and planetary parameters for WASP-52 (Hébrard et al. 2013; Ta- ble 4) and HAT-P-1 (Nikolov et al. 2014). HAT-P-1b has a precise transmission spectrum measure- ment from HST/STIS (Nikolov et al. 2014), which we use to compare the atmospheric properties of both planets. For this comparison, we reconstructed the HST/STIS transmis- sion spectrum for WASP-52b using the same binning scheme of Nikolov et al. (2014) for HAT-P-1b and fit the light curves for these bins based on the methods outlined in Section 4. Figure 13 shows the HST/STIS transmission spectra of both planets with identical binning. Based on this comparison, the spectra of both planets are identical within the uncertainties with an average 1σ difference. As reported in Nikolov et al. (2014), the best fit model for HAT-P-1b is a hazy spectrum with Na I absorption and an extra optical absorber to account for the observed absorption enhancement at wavelengths longer than ∼0.85 µm. To di- rectly compare this interpretation to the WASP-52b results reported here, we fit the HAT-P-1b spectrum (Nikolov et al. 2014) to the open source ATMO grid of forward models gen- erated for the parameters of HAT-P-1b (see Appendix C and Figure C.1 for details). The best fit model parameters for HAT-P-1b and WASP-52b are shown in Table 6. Since WASP-52b and HAT-P-1b have similar optical trans- mission spectra (∼0.29−1 µm), we compare the atmospheric properties of these planets in the near-infrared to ascertain if these planets would be good candidates for comparative at- mospheric analyses with JWST. Figure 14 shows the best fit models for both planets from 0.29−5 µm. Beyond 1 µm, we find that the best fitting models for HAT-P-1b and WASP-52b do not agree with each other as they do in the optical. For WASP-52b 11 further discussion regarding potential reasons for this near- infrared discrepancy, see Section 6.2. 6. DISCUSSION 6.1. Interpreting Alkali Detections in the Presence of Clouds As reported in Section 5.1, we find hints of Na I absorp- tion but no evidence of K I in the transmission spectrum of WASP-52b. A similar trend has been observed for HD 189733b (Pont et al. 2013), HAT-P-1b (Nikolov et al. 2014), and WASP-17b (Sing et al. 2016). The reverse trend (i.e., the presence of a K I signal but no Na I) has been observed in several planets, including WASP-31b (Sing et al. 2015) and HAT-P-12b (Sing et al. 2016). We note, however, that the majority of current sodium and potassium detections in exo- planet atmospheres are low significance and a non-detection of K I can only be interpreted as an upper limit to the abun- dance of that element in the atmospheric layers probed by our observations, considering their uncertainties. Based on the available data and their uncertainties, we estimate an abun- dance ratio of ln[Na/K] = 8.32+6.51 −6.03. This value is consistent with solar abundances but has large uncertainties and is not well-constrained. These results are consistent with our forward model anal- ysis, which favors the presence of clouds and slight hazes that mute spectroscopic features in the planet's atmosphere. The best fitting models give temperatures ranging from T ∼ 1000−1300 K. Compared to the planet's equilibrium temper- ature (Teq = 1315 K; Hébrard et al. 2013), these temperature estimates may be driven by the gradient of the Rayleigh scat- tering slope. Significant sodium condensation in the form of Na2S is expected at lower temperatures (∼1000 K) and potassium condensation in the form of KCl at even lower temperatures (∼600 K) (Marley et al. 2013). Tenuous Na2S and KCl clouds could form at shallow pressures in WASP- 52b's atmosphere, since the equilibrium temperature does not represent the full range of temperatures in a planet's atmo- sphere. The cloud deck in WASP-52b's atmosphere could therefore be comprised of species other than sodium or potas- sium compounds, such as silicates. Regardless of the composition of these clouds, they are likely masking the K I feature and the wings of the Na I res- onance core. We do not resolve the broad wings of the Na I line (Figure 11), which may suggest the presence of an ex- tra absorber or scatterer in the atmosphere that is obscuring or masking the atmospheric Na I and K I absorption features (Seager & Sasselov 2000; Nikolov et al. 2014). If a K I signal is truly lacking in the transmission spec- trum of WASP-52b, an alternative explanation could be at- tributed to an underabundance of this element in the plane- tary atmosphere. If the cloud deck is comprised of sodium or potassium compounds, however, the gas phase abundances of these species would not reflect primordial abundances. Since K I is a weaker spectroscopic feature, it may be present in the planet's atmosphere but not detectable with the precision of the STIS data. Higher resolution, higher precision observa- tions are necessary to confirm this idea. 6.2. Contextualizing WASP-52b Although the atmospheres of planets studied thus far ap- pear diverse (Sing et al. 2016), we have not yet been able to identify any clear correlations between planetary atmo- spheric properties and other system parameters. Compar- ing the transmission spectra of planets with similar system parameters may therefore prove insightful in searching for common atmospheric characteristics. WASP-52b is a good target for such comparisons, since the well-studied inflated hot Jupiter HAT-P-1b has comparable system parameters and atmospheric properties. We compare the observed transmission spectrum of WASP-52b presented here to that of the well-studied in- flated hot Jupiter HAT-P-1b. These two planets have similar system parameters (see Table 6), although WASP-52 is ∼0.9 times more active that HAT-P-1b (Nikolov et al. 2014). Ad- ditionally, the optical transmission spectra of both planets are marginally flat with evidence of Na I but no observable evidence of K I. We compare their transmission spectra with identical binning schemes in Figure 13, and find that the spectra of both planets in the optical (∼0.29−1 µm) are iden- tical within the uncertainties with an average 1σ difference. Extending this comparison to near-infrared wavelengths, however, reveals that their transmission spectra differ con- siderably beyond 1 µm. The best fit ATMO model for WASP- 52b shows muted H2O and CH4 spectral features compared to the best fitting HAT-P-1b model, which could indicate that WASP-52b has a higher aerosol layer. Near-infrared HST/WFC3 observations of HAT-P-1b reveal H2O absorp- tion at >5σ confidence (Wakeford et al. 2013), and the deep H2O feature shown in the best fit ATMO model matches these data. The best fit atmospheric model for WASP-52b shows a weaker (but still observable) H2O feature at 1.4 µm, and evidence of H2O absorption at 1.4 µm for WASP-52b has recently been shown in Bruno et al. (2018). The activity lev- els of the host stars may also contribute to the divergence of the transmission spectra for these two planets at near- infrared wavelengths, although observations of the stellar UV fluxes are necessary to confirm this hypothesis. Comparative near-infrared observations with JWST can confirm the atmo- spheric similarities of these planets at shallower atmospheric layers compared to those probed by STIS. 7. SUMMARY Our key results are summarized as follows: 12 ALAM ET AL. • We present an optical to near-infrared transmission spectrum of WASP-52b (measured to a median pre- cision of 90 ppm) from ∼0.29 − 5.0 µm using tran- sit observations from HST/STIS and Spitzer/IRAC. We correct for the effects of stellar activity and fit the ob- served transmission spectrum to a grid of forward at- mospheric models. • Based on these fits (Figure 9), we find that our trans- mission spectrum measurement best matches a mod- erately cloudy atmospheric model with an equilibrium temperature of 1315 K, a thick cloud deck (αcloud = 1.00), a slight Rayleigh scattering slope in the blue (αhaze = 10), and hints of a 2.3σ Na I signal at 5893 Å. Within the precision of our observations, we do not detect K I absorption. • We compare the observed transmission spectra of HAT-P-1b and WASP-52b, two planetary systems with similar stellar and planetary parameters (Table 6). By constructing optical HST/STIS transmission spectra with similar binning schemes (Figure 13), we find that the spectra of these two planets are identical within the uncertainties at optical wavelengths but differ in the near-infrared (Figure 14) based on our best fit models. • The difference in the transmission spectra of WASP- 52b and HAT-P-1b from ∼1.0 − 5.0 µm may be caused by the presence of an extra optical absorber in the at- mosphere of HAT-P-1b (Nikolov et al. 2014) or uncer- tainties in the best fitting models, which are isothermal and therefore cannot accurately capture cloud forma- tion. • Comparative atmospheric observations with JWST for WASP-52b and HAT-P-1b will be key to understand- ing planets with similar system parameters and over- lapping atmospheric properties. In a forthcoming paper, we aim to combine the STIS+Spitzer transmission spectrum presented here with existing near- infrared HST/WFC3 observations (Bruno et al. 2018). Us- ing the full optical to near-infrared transmission spectrum, we will retrieve the planet's atmospheric properties (Bruno, Alam, et al., in prep.) to better constrain the atmospheric structure and chemical composition of this inflated hot Jupiter as well as precisely estimate the Na I and K I abun- dances in the planet's atmosphere. Such an analysis will indicate if comparative planetology and comparative atmo- spheric studies of WASP-52b with future JWST observations will prove insightful. The authors thank the anonymous referee for helpful com- ments that greatly improved this manuscript. This paper makes use of observations from the NASA/ESA Hubble Space Telescope, obtained at the Space Telescope Science In- stitute, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS 5-26555. These observations are associated with program GO 14767. The research leading to these results has received funding from the European Research Council under the Eu- ropean Union's Seventh Framework Programme (FP7/2007- 2013)/ERC grant agreement number 336792. We are thank- ful to Raphaelle Haywood, James Kirk, Chani Nava, and Ian Weaver for useful discussions. MKA acknowledges support by the National Science Foundation through a Graduate Research Fellowship. GWH and MHW acknowledge support from Tennessee State Uni- versity and the State of Tennessee through its Centers of Ex- cellence program. JSF acknowledges funding by the Spanish MINECO grant AYA2016- 79425-C3-2-P. JKB acknowl- edges support from the Royal Astronomical Society. VB and DE have received funding from the European Research Council (ERC) under the European Unions Horizon 2020 research and innovation program (project Four Aces; grant agreement no. 724427). REFERENCES Struve, O. 1952, The Observatory, 72, 199 Markwardt, C. B. 2009, Astronomical Data Analysis Software and Systems XVIII, 411, 251 Seager, S., & Sasselov, D. D. 2000, ApJ, 537, 916 Brown, T. M. 2001, ApJ, 553, 1006 Stevenson, K. B., Désert, J.-M., Line, M. R., et al. 2014, Science, Fraine, J., Deming, D., Benneke, B., et al. 2014, Nature, 513, 526 Luszcz-Cook, S. H., & de Pater, I. 2013, Icarus, 222, 379 Kirk, J., Wheatley, P. J., Louden, T., et al. 2016, MNRAS, 463, 2922 Todorov, K. O., Deming, D., Knutson, H. A., et al. 2013, ApJ, 770, 102 346, 838 Louden, T., Wheatley, P. J., Irwin, P. G. J., Kirk, J., & Skillen, I. Kreidberg, L., Bean, J. L., Désert, J.-M., et al. 2014, Nature, 505, 2017, MNRAS, 470, 742 69 Sing, D. K., Fortney, J. J., Nikolov, N., et al. 2016, Nature, 529, 59 Pont, F., Sing, D. K., Gibson, N. P., et al. 2013, MNRAS, 432, 2917 Chen, G., Pallé, E., Nortmann, L., et al. 2017, A&A, 600, L11 Hébrard, G., Collier Cameron, A., Brown, D. J. A., et al. 2013, A&A, 549, A134 WASP-52b 13 Nikolov, N., Sing, D. K., Pont, F., et al. 2014, MNRAS, 437, 46 Sing, D. K., Pont, F., Aigrain, S., et al. 2011, MNRAS, 416, 1443 Mandel, K., & Agol, E. 2002, ApJL, 580, L171 Charbonneau, D., Brown, T. M., Noyes, R. W., & Gilliland, R. L. 2002, ApJ, 568, 377 Ambikasaran, S., Foreman-Mackey, D., Greengard, L., Hogg, D. W., & O'Neil, M. 2015, IEEE Transactions on Pattern Analysis and Machine Intelligence, 38. Evans, T. M., Sing, D. K., Kataria, T., et al. 2017, Nature, 548, 58. Nikolov, N., Sing, D. K., Burrows, A. S., et al. 2015, MNRAS, Vidal-Madjar, A., Lecavelier des Etangs, A., Désert, J.-M., et al. 447, 463. 2003, Nature, 422, 143 Sing, D. K., Lecavelier des Etangs, A., Fortney, J. J., et al. 2013, Deming, D., Seager, S., Richardson, L. J., & Harrington, J. 2005, MNRAS, 436, 2956. Nature, 434, 740 Charbonneau, D., Knutson, H. A., Barman, T., et al. 2008, ApJ, 686, 1341 Sing, D. K., & López-Morales, M. 2009, A&A, 493, L31 Knutson, H. A., Charbonneau, D., Allen, L. E., et al. 2007, Nature, 447, 183 Sing, D. K., Vidal-Madjar, A., Désert, J.-M., Lecavelier des Etangs, A., & Ballester, G. 2008, ApJ, 686, 658 Désert, J.-M., Vidal-Madjar, A., Lecavelier Des Etangs, A., et al. 2008, A&A, 492, 585 Lecavelier Des Etangs, A., Vidal-Madjar, A., Désert, J.-M., & Sing, D. 2008, A&A, 485, 865 Sing, D. K., Vidal-Madjar, A., Lecavelier des Etangs, A., et al. 2008, ApJ, 686, 667 Snellen, I. A. G., Albrecht, S., de Mooij, E. J. W., & Le Poole, R. S. 2008, A&A, 487, 357 Sudarsky, D., Burrows, A., & Hubeny, I. 2003, ApJ, 588, 1121 Burrows, A., Rauscher, E., Spiegel, D. S., & Menou, K. 2010, ApJ, 719, 341 Lecavelier Des Etangs, A., Ehrenreich, D., Vidal-Madjar, A., et al. 2010, A&A, 514, A72. Fortney, J. J., Shabram, M., Showman, A. P., et al. 2010, ApJ, 709, 1396. Sing, D. K., Désert, J.-M., Lecavelier Des Etangs, A., et al. 2009, A&A, 505, 891. Fischer, P. D., Knutson, H. A., Sing, D. K., et al. 2016, ApJ, 827, 19. Sing, D. K., Wakeford, H. R., Showman, A. P., et al. 2015, MNRAS, 446, 2428. Jordán, A., Espinoza, N., Rabus, M., et al. 2013, ApJ, 778, 184. Huitson, C. M., Sing, D. K., Pont, F., et al. 2013, MNRAS, 434, 3252. Grillmair, C. J., Burrows, A., Charbonneau, D., et al. 2008, Nature, 456, 767. Huitson, C. M., Désert, J.-M., Bean, J. L., et al. 2017, AJ, 154, 95. Gibson, N. P., Nikolov, N., Sing, D. K., et al. 2017, MNRAS, 467, 4591. Nikolov, N., Sing, D. K., Gibson, N. P., et al. 2016, ApJ, 832, 191. Mallonn, M., Bernt, I., Herrero, E., et al. 2016, MNRAS, 463, 604. Aigrain, S., Pont, F., & Zucker, S. 2012, MNRAS, 419, 3147. Kurucz, R. L. 1993, VizieR Online Data Catalog , VI/39. Sing, D. K. 2010, A&A, 510, A21. Fazio, G. G., Hora, J. L., Allen, L. E., et al. 2004, The Astrophysical Journal Supplement Series, 154, 10. Werner, M. W., Roellig, T. L., Low, F. J., et al. 2004, The Astrophysical Journal Supplement Series, 154, 1. Sedaghati, E., Boffin, H. M. J., MacDonald, R. J., et al. 2017, Nature, 549, 238 Angus, R., Morton, T., Aigrain, S., et al. 2018, MNRAS, 474, 2094. Haywood, R. D. 2015, Ph.D. Thesis. Rajpaul, V., Aigrain, S., Osborne, M. A., et al. 2015, MNRAS, 452, 2269. Gibson, N. P. 2014, MNRAS, 445, 3401. Knutson, H. A., Lewis, N., Fortney, J. J., et al. 2012, ApJ, 754, 22. Eastman, J., Siverd, R., & Gaudi, B. S. 2010, Publications of the Astronomical Society of the Pacific, 122, 935. Carter, J. A., & Winn, J. N. 2009, ApJ, 704, 51. Knutson, H. A., Charbonneau, D., Allen, L. E., et al. 2008, ApJ, 673, 526. Reach, W. T., Megeath, S. T., Cohen, M., et al. 2005, Publications of the Astronomical Society of the Pacific, 117, 978. Mighell, K. J. 2005, MNRAS, 361, 861. Charbonneau, D., Allen, L. E., Megeath, S. T., et al. 2005, ApJ, 626, 523. Claret, A. 2000, A&A, 363, 1081. Gilliland, R. L., Goudfrooij, P., & Kimble, R. A. 1999, Publications of the Astronomical Society of the Pacific, 111, 1009. Akaike, H. 1974, IEEE Transactions on Automatic Control, 19, 716. Goyal, J. M., Mayne, N., Sing, D. K., et al. 2018, MNRAS, 474, 5158. Wakeford, H. R., Sing, D. K., Deming, D., et al. 2018, AJ, 155, 29. Kochanek, C. S., Shappee, B. J., Stanek, K. Z., et al. 2017, Publications of the Astronomical Society of the Pacific, 129, 104502. Drummond, B., Tremblin, P., Baraffe, I., et al. 2016, A&A, 594, A69. Tremblin, P., Amundsen, D. S., Chabrier, G., et al. 2016, ApJ, 817, L19. Stevenson, K. B. 2016, ApJ, 817, L16. Tremblin, P., Amundsen, D. S., Mourier, P., et al. 2015, ApJ, 804, L17. 14 ALAM ET AL. Shappee, B. J., Prieto, J. L., Grupe, D., et al. 2014, ApJ, 788, 48. Amundsen, D. S., Baraffe, I., Tremblin, P., et al. 2014, A&A, 564, A59. Lewis, N. K., Knutson, H. A., Showman, A. P., et al. 2013, ApJ, 766, 95. Howarth, I. D. 2011, MNRAS, 413, 1515. Fortney, J. J., Marley, M. S., Saumon, D., et al. 2008, ApJ, 683, 1104. Sing, D. K., Désert, J.-M., Fortney, J. J., et al. 2011, A&A, 527, A73. Berdyugina, S. V. 2005, Living Reviews in Solar Physics, 2, 8. Kreidberg, L. 2017, Handbook of Exoplanets, Edited by Hans J. Deeg and Juan Antonio Belmonte. Springer Living Reference Work, ISBN: 978-3-319-30648-3, 2017, id.100, 100 Bruno, G., Lewis, N. K., Stevenson, K. B., et al. 2018, AJ, 156, 124. Hubbard, W. B., Fortney, J. J., Lunine, J. I., et al. 2001, ApJ, 560, Wyttenbach, A., Lovis, C., Ehrenreich, D., et al. 2017, A&A, 602, 413. Dang, L., Cowan, N. B., Schwartz, J. C., et al. 2018, Nature Astronomy, 2, 220. Nikolov, N., Sing, D. K., Fortney, J. J., et al. 2018, Nature, 557, 526 Wakeford, H. R., Sing, D. K., Kataria, T., et al. 2017, Science, 356, 628. Mancini, L., Southworth, J., Raia, G., et al. 2017, MNRAS, 465, 843. Rackham, B., Espinoza, N., Apai, D., et al. 2017, ApJ, 834, 151. Aigrain, S., Parviainen, H., & Pope, B. J. S. 2016, MNRAS, 459, 2408. Heng, K. 2016, ApJ, 826, L16. Magic, Z., Chiavassa, A., Collet, R., et al. 2015, A&A, 573, A90. Haywood, R. D., Collier Cameron, A., Queloz, D., et al. 2014, MNRAS, 443, 2517. Mandell, A. M., Haynes, K., Sinukoff, E., et al. 2013, ApJ, 779, 128. Wakeford, H. R., Sing, D. K., Deming, D., et al. 2013, MNRAS, 435, 3481. A36. Stevenson, K. B., Line, M. R., Bean, J. L., et al. 2017, AJ, 153, 68. Wyttenbach, A., Ehrenreich, D., Lovis, C., et al. 2015, A&A, 577, A62. Bourrier, V., Lecavelier des Etangs, A., Dupuy, H., et al. 2013, A&A, 551, A63. Marley, M. S., Ackerman, A. S., Cuzzi, J. N., et al. 2013, Comparative Climatology of Terrestrial Planets, 367. Huitson, C. M., Sing, D. K., Vidal-Madjar, A., et al. 2012, MNRAS, 422, 2477. Winn, J. N. 2010, ArXiv e-prints , arXiv:1001.2010. Fruscione, A., McDowell, J. C., Allen, G. E., et al. 2006, Society of Photo-optical Instrumentation Engineers (SPIE) Conference Series, 62701V. Houck, J. C., & Denicola, L. A. 2000, Astronomical Data Analysis Software and Systems IX, 591. Gaia Collaboration, Brown, A. G. A., Vallenari, A., et al. 2018, A&A, 616, A1 WASP-52b 15 Figure 1. Ground-based photometric observations of WASP-52 from ASAS-SN (black points) and AIT (red triangles) during the 2013−2014 (middle left), 2014−2015 (middle right), 2015−2016 (bottom left), and 2016−2017 (bottom right) observing seasons. The data are flux relative to the average brightness of comparison stars. The Gaussian process regression model (red) and 1σ uncertainty (gray) fit to the combined ASAS-SN and AIT data are overplotted for the full dataset (top panel) and each observing season separately. The dashed vertical lines indicate the WHT/ACAM (orange), GTC/OSIRIS (light green), HST/STIS G430L (blue), HST/STIS G750L (red), and Spitzer/IRAC (dark green) transit epochs. Time (HJD - 2400000)Time (HJD - 2400000)Time (HJD - 2400000)16 ALAM ET AL. Figure 2. Theoretical dimming of unocculted spots in the wavelength range of the HST/STIS and Spitzer/IRAC observations for stellar flux models ranging in temperature from 3500 − 4750 K (corresponding to temperature differences between the spot and stellar photosphere from 1500 K to 250 K). The dimming ∆ f (λ,T ) is derived by multiplying the wavelength-dependent flux correction by the non-spotted stellar flux (Section 3.4). Top panel: Spot dimming for the full STIS+Spitzer wavelength range (∼0.29-4.5 µm). Bottom panel: Zoom-in of the STIS wavelength range only (∼0.29-1.0 µm). 123450.010.020.030.040.050.06 f(,T)4750 K4500 K4250 K4000 K3750 K3500 K0.30.40.50.60.70.80.91.0Wavelength (m)0.020.040.06 f(,T)WASP-52b 17 Figure 3. HST/STIS stellar spectra of WASP-52b and the corresponding white light curves for visits 52 (left column), 53 (middle column), and 54 (right column). Top panel: Example stellar spectra taken with the G430L (blue) and G750L (red) grating. The vertical gray dashed lines indicate the wavelength range used to produce the white light curves. Second panel: The raw white light curves for each visit. Third panel: The raw and detrended light curves (excluding the first orbit and the first exposure of each subsequent orbit) with the best-fit model overplotted. The common mode correction (gray squares) is the transit+systematics model divided by the best-fit transit model. Bottom panel: Transit fit residuals with error bars. 18 ALAM ET AL. Figure 4. HST/STIS G430L observations of WASP-52b visit 52, excluding the first orbit and the first exposure of each subsequent orbit. Raw (left panel) and detrended (middle panel) light curves are shown for each wavelength bin, and are offset vertically by an arbitrary constant for clarity. Observed minus computed residuals with error bars are shown in the right panel. 2021.001.051.101.151.201.251.301.35Normalized Flux21012Time from Mid-Transit (hours)2900-3700 Å3700-3950 Å3950-4113 Å4113-4250 Å4250-4400 Å4400-4500 Å4500-4600 Å4600-4700 Å4700-4800 Å4800-4900 Å4900-5000 Å5000-5100 Å5100-5200 Å5200-5300 Å5300-5400 Å5400-5500 Å5500-5600 Å5600-5700 Å297 ppm183 ppm118 ppm93 ppm100 ppm108 ppm118 ppm95 ppm112 ppm91 ppm130 ppm93 ppm104 ppm70 ppm143 ppm89 ppm85 ppm76 ppm505505505505505505505505505505O - C (ppm)505505505505505505505202505WASP-52b 19 Figure 5. Same as Figure 4, but for visit 53. 2021.001.051.101.151.201.251.301.35Normalized Flux21012Time from Mid-Transit (hours)2900-3700 Å3700-3950 Å3950-4113 Å4113-4250 Å4250-4400 Å4400-4500 Å4500-4600 Å4600-4700 Å4700-4800 Å4800-4900 Å4900-5000 Å5000-5100 Å5100-5200 Å5200-5300 Å5300-5400 Å5400-5500 Å5500-5600 Å5600-5700 Å305 ppm226 ppm191 ppm112 ppm83 ppm91 ppm107 ppm82 ppm84 ppm87 ppm108 ppm71 ppm102 ppm72 ppm84 ppm98 ppm71 ppm116 ppm505505505505505505505505505505O - C (ppm)50550550550550550550520250520 ALAM ET AL. Figure 6. HST/STIS G750L observations of WASP-52b visit 54, excluding the first orbit and the first exposure of each subsequent orbit. Raw (left panel) and detrended (middle panel) light curves are shown for each wavelength bin, and are offset vertically by an arbitrary constant for clarity. Observed minus computed residuals with error bars are shown in the right panel. 2021.001.051.101.151.201.251.301.35Normalized Flux21012Time from Mid-Transit (hours)5700-5800 Å5800-5878 Å5878-5913 Å5913-6070 Å6070-6200 Å6200-6300 Å6300-6450 Å6450-6600 Å6600-6800 Å6800-7000 Å7000-7200 Å7200-7450 Å7450-7645 Å7645-7720 Å7720-8100 Å8100-8485 Å8485-8985 Å8985-10300 Å97 ppm96 ppm135 ppm61 ppm56 ppm70 ppm60 ppm47 ppm52 ppm74 ppm54 ppm42 ppm69 ppm88 ppm55 ppm55 ppm67 ppm97 ppm2.50.02.52.50.02.52.50.02.52.50.02.52.50.02.52.50.02.52.50.02.52.50.02.52.50.02.52.50.02.5O - C (ppm)2.50.02.52.50.02.52.50.02.52.50.02.52.50.02.52.50.02.55052022.50.02.5WASP-52b 21 Figure 7. Spitzer/IRAC 3.6 µm transit light curve. Top panel: Raw flux (black points) and best-fit transit model (red points). Middle panel: Detrended light curve (gray points) and best-fit transit model (green line) overlaid with the binned light curve (magenta points; 45 bins of 224 data points each). Bottom panel: observed minus computed residuals (black points) of the raw light curve. 210120.960.981.001.02Normalized Raw Flux210120.960.970.980.991.001.011.021.03Normalized Flux21012Time from Mid-Transit (hours)2.50.02.5O - C (×102)22 ALAM ET AL. Figure 8. Spitzer/IRAC 4.5 µm transit light curve. Top panel: Raw flux (black points) and best-fit transit model (red points). The small gap at the end of the transit ingress corresponds to the delay between subsequent data readouts. Middle panel: Detrended light curve (gray points) and best-fit transit model (green line) overlaid with the binned light curve (magenta points; 43 bins of 239 data points each). Bottom panel: observed minus computed residuals (black points) of the raw light curve. 210120.960.981.001.021.04Normalized Raw Flux210120.960.970.980.991.001.011.021.03Normalized Flux3210123Time from Mid-Transit (hours)2.50.02.5O - C (×102)WASP-52b 23 Figure 9. Top panel: Stellar activity corrected transmission spectrum for WASP-52b from HST/STIS and Spitzer/IRAC (black circles). Ground- based optical transmission spectrum measurements from Louden et al. 2017 (blue diamonds) and Chen et al. 2017 (salmon rectangles) are included for comparison. We show a subset of the best-fit theoretical atmospheric models (lines), and find that the observed transmission spectrum is consistent with evidence of Na I at 2.3σ confidence and a cloudy atmosphere with no TiO. Models are smoothed by a constant. The average Rp/R(cid:63) baseline of the transmission spectrum (dashed black line) is shown for reference. Bottom panel: Same as above, but zoomed in to the STIS wavelength range (∼0.29-1.0 µm). 1.000.30.40.50.60.70.80.92.03.04.05.0Wavelength (m)0.1550.1580.1600.1630.1650.1670.1700.1720.175Rp/Rthis workLouden et. al (2017)Chen et. al (2017)best fit model (T_eff=1315,_haze=0010,_cloud=1), 2=39.323, BIC=4.938hazy (T_eff=1015,_haze=1100,_cloud=0), 2=43.339, BIC=8.633clear (T_eff=1315,_haze=0010,_cloud=0), 2=49.474, BIC=13.664featureless spectrum (gray atmosphere), 2=52.162, BIC=15.6721012Scale Heights (km)1.000.300.400.500.600.700.800.90Wavelength (m)0.1560.1580.1600.1620.1640.1660.1680.1700.172Rp/R21012Scale Heights (km)24 ALAM ET AL. Figure 10. Comparison of the raw (gray diamonds) and spot corrected (black circles) transmission spectra for WASP-52b. The best fitting (solid blue line) and gray atmosphere (dashed black line) models from Figure 9 are shown for reference. 1.000.30.40.50.60.70.80.92.03.04.05.0Wavelength (m)0.1550.1580.1600.1630.1650.1670.1700.1720.175Rp/Rspot corrected spectrumraw (uncorrected) spectrumbest fit model (T_eff=1615,_haze=10,_cloud=1), 2=42.71, BIC=8.078featureless spectrum (gray atmosphere), 2=52.162, BIC=15.67521012Scale Heights (km)WASP-52b 25 Figure 11. Top panel: WASP-52b transmission spectrum (black points) zoomed in to the central wavelength of Na I at 5893 Å (left) and K I at 7665 Å (right) with the best fit (blue), hazy (green) and clear (red) models overplotted. Bottom panel: Absorption depth of Na I at 5893 Å (left) and K I at 7665 Å (right) for spectroscopic channels ranging in size from 30-255 Å . The average Rp/R(cid:63) baseline of the transmission spectrum (dashed black line) is shown for reference. 0.5800.5850.5900.5950.600Wavelength (m)0.1620.1640.1660.1680.1700.172Rp/R0.740.750.760.770.780.790.80Wavelength (m)0.1560.1580.1600.1620.1640.1660.16850100150200250Bin Width (Å)0.1620.1640.1660.1680.1700.1720.174Rp/RNa I (5893 Å)50100150200250Bin Width (Å)0.1450.1500.1550.1600.1650.1700.175K I (7665 Å)26 ALAM ET AL. Figure 12. χ2 map for WASP-52b for the no rainout ATMO model grid. The cloud, haze, and metallicity axes are log-scaled. The contours show all combinations of the grid parameters, and the colors indicate the confidence intervals corresponding to the colorbar to the right. The white regions correspond to parameter space on the grid that are not feasible given current observations at 4σ confidence and can therefore be easily ruled out. 101513151615-2-1012Metallicity1σ2σ3σ4σ1015131516150.20.611.4C/O-ratio1015131516151101001000Haze101513151615Temperature00.21Clouds-2-10120.20.611.4-2-10121101001000-2-1012Metallicity00.210.20.611.411010010000.20.611.4C/O-ratio00.211101001000Haze00.21WASP-52b 27 Figure 13. Top panel: Comparison of the observed STIS transmission spectra of WASP-52b (black circles) and HAT-P-1b (purple squares; Nikolov et al. 2014) for identical wavelength bins. The HAT-P-1b spectrum is offset by an arbitrary constant such that the first spectroscopic bin is anchored to the first Rp/R(cid:63) value of the WASP-52b spectrum. Bottom panel: Difference between the observed WASP-52b spectrum and the offset HAT-P-1b spectrum. The spectra of both planets are identical within the uncertainties (except for one channel at ∼0.7 µm), with an average 1σ difference. The horizontal black dashed line indicates the baseline offset between the two spectra. 1.000.30.40.50.60.70.80.90.14250.14500.14750.15000.15250.15500.15750.16000.1625Rp/RWASP-52b (this work)HAT-P-1b (Nikolov et al. 2014)1.000.30.40.50.60.70.80.9Wavelength (m)0.0400.0450.050(RpR)W52 - (RpR)H128 ALAM ET AL. Figure 14. Comparison of the best fit models for HAT-P-1b (purple) and WASP-52b (black) from ∼0.29 to 5 µm. The spectra are identical within the uncertainties at optical wavelengths but differ in the near-infrared. 1.000.30.40.50.60.70.80.92.03.04.05.0Wavelength (m)0.1200.1220.1240.1260.1280.1300.132Rp/R + ConstantHAT-P-1bWASP-52bWASP-52b APPENDIX A. STELLAR ACTIVITY CORRECTION A.1. Kernel for Gaussian Process Regression Model 29 We used a Gaussian process (GP) regression to model the ground-based stellar activity monitoring data (Pont et al. 2013; Haywood 2015; Aigrain et al. 2016; Angus et al. 2018). In our GP analysis, we used a three component kernel of the form K = k1 + k2 + k3. The k1 term models the flexible (quasi-) periodicity of the ground-based activity monitoring data. It is a squared exponential kernel multiplied by an exponential sine squared kernel (k1 = A2[k(r2)× k(xi,x j)]), where k(r2) = er2/2 (A1) represents the squared exponential kernel for periodic variations in the time series data parametrized by r. The exponential sine squared kernel is given by k(xi,x j) = exp where Γ represents the scale of the correlations and P is the period of the oscillations. The second term, k2, represents the irregularities in amplitude and period. It is a rational quadratic kernel of the form: (cid:16) − Γsin2(cid:104) π k2(r2) = A2(cid:16) P 1 − r2 2α xi − x j(cid:105)(cid:17) (cid:17) (A2) (A3) for amplitude A, Gamma distribution parameter α, and length scale r. The k3 term incorporates stellar noise in the GP model and is a squared exponential kernel (of the form shown in Equation A1) added to a white kernel of the form k3(xi,x j) = cδi j for constant c and diagonal value δi j. A.2. Deriving the Wavelength-Dependent Flux Correction As described in Section 3, we account for the effects of stellar activity and unocculted starspots on the transmission spectrum by using ground-based photometry and by deriving the wavelength-dependent flux correction ∆ f (λ,T ) (Sing et al. 2011; see also Section 3.4, Equation 1). This correction depends on the parameter k, which provides an assumption for the non-spotted flux on the stellar surface (Aigrain et al. 2012). We fixed k to unity based on Aigrain et al. (2012), which found this value appropriate to use for active stars. Physically, k=1 corresponds to a spot contribution that the viewer never sees (or always sees) that is about the same as the contribution of spots that come into and out of view. The value of this parameter is based on several assumptions; however, it is, in actuality, very ill-constrained and we warn the reader of the difficulty in selecting a value of k. To explore the potential effect of the choice of k on the final transmission spectrum, we tested different values of this parameter ranging from k=0−1 in steps of 0.2 (Figure 15). We find that larger values of k correspond (linearly) to a higher derived flux correction ∆ f (λ,T ). When applying the activity correction to the binned light curves, we find that the light curves are re-scaled to a higher Rp/R(cid:63) baseline if a larger correction is applied. Thus, the parameter k is not wavelength-dependent and therefore does not affect the shape of the slope of the transmission spectrum. Rather, varying the value of k simply shifts the Rp/R(cid:63) baseline of the transmission spectrum vertically. B. STIS LIGHT CURVES B.1. STIS White Light Curve Systematics Models As described in Section 4.1, we detrended instrument systematic effects in the light curves by fitting a fourth-order polynomial to the flux dependence on HST orbital phase. Each model represents a unique linear combination of the detrending variables: orbital phase (φt), drift of the spectra on the detector (x and y), the shift (ω) of each stellar spectrum cross-correlated with the first spectrum of the time series, and time t. The variables x and y are the trace slope and an offset in the cross dispersion direction, respectively. The parameter ω is measured by cross-correlating a reference spectrum with the remaining spectra and is measured prior to re-sampling the spectra. For both the G430L and G750L STIS observations, we used the 25 systematics models listed in Table B1 to detrend the light curves. After performing separate fits for each model, we marginalized over the entire set of systematics models assuming equally weighted priors to select which systematics model to use. Table B2 summarizes the selection of systematics models based on the STIS white light curves. We selected the model for detrending based on the lowest Aikake Information Criterion (AIC) value. 30 ALAM ET AL. Figure 15. Wavelength-dependent flux correction for a fixed spot temperature (4750 K) and different values of the parameter k, a proxy for the non-spotted fraction of the stellar surface. For the activity correction described in Section 3.4, we assume k=1 (i.e., the spot contribution that the viewer never sees (or always sees) is about the same as the contribution of spots that come into and out of view). B.2. STIS Spectrophotometric Light Curves The broadband STIS+Spitzer transmission spectrum for WASP-52b reported in Table 5 gives the weighted mean of the two STIS G430L observations (visits 52 and 53). In Tables B3 and B4 below, we report the spectrophotometric light curves for each G430L visit. To produce the raw transmission spectrum from the spot corrected spectrum, the reader may simply reverse the correction given in Table 3 on each transit observation separately. C. HAT-P-1b FORWARD MODEL FITS Figure C.1 shows the HST/STIS transmission spectrum of the well-studied inflated hot Jupiter HAT-P-1b from Nikolov et al. (2014) compared to the best fitting theoretical forward model from the ATMO grid generated for the parameters of HAT-P-1b6 in addition to representative clear and cloudy models. We performed these fits using the procedure described in Section 5.2. We find that the best fit model has T = 1322 K with a strong Na I signal and is slightly cloudy (αcloud=0.20) and hazy (αhaze=10). 6 https://bd-server.astro.ex.ac.uk/exoplanets/HAT-P-01/ 12345Wavelength (m)0.000.010.020.030.040.050.06 f(,T)k = 0.0k = 0.2k = 0.4k = 0.6k = 0.8k = 1.0WASP-52b 31 Figure C.1. HST/STIS transmission spectrum of HAT-P-1b (black points; Nikolov et al. 2014) compared to a subset of the best-fit theoretical atmospheric models (lines) for the no rainout ATMO grid. For reference, the dashed black line shows the average Rp/R(cid:63) baseline of the transmission spectrum. Table B1. White Light Curve Systematics Models Model G430L models 1 2 3 4 5 6 7 8 9 10 11 12 13 14 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + t t + t + ω + x2 t + t + x + y2 t + t + x2 + y t + t + ω t + t + x t + t + y t + t + ω + ω3 + x t + t + ω + y t + t + ω + x + y t + t + ω + ω2 t + t + ω + ω2 + ω3 t + t + x + x2 + y t + t + ω + x + x2 + x3 Table B1 continued 1.000.300.400.500.600.700.800.90Wavelength (m)0.1160.1180.1200.1220.124Rp/Rbest fit model (T_eff=1322,_haze=0010,_cloud=0.20), 2=18.906, BIC=6.845cloudy (T_eff=1322,_haze=0010,_cloud=1.00), 2=19.912, BIC=7.57clear (T_eff=1322,_haze=0010,_cloud=0.00), 2=19.243, BIC=7.092featureless spectrum (gray atmosphere), 2=37.774, BIC=16.53532 ALAM ET AL. Table B1 (continued) Model 15 16 17 18 19 20 21 22 23 24 25 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + t + ω + y + y2 t + t + ω + y + y2 + y3 t + t + ω + ω2 + ω3 + x t + t + x + x2 t + t + x + x2 + x3 t + t + y + y2 t + t + y + y2 + y3 t + t + ω + ω2 + x t + t + ω + ω2 + x + y t + t + ω + ω2 + x + x2 + y + y2 t + t + ω2 + ω3 + x + x2 + x3 + y + y2 + y3 G750L models φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 φt + φ2 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ3 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + φ4 t + t t + t + ω + x2 t + t + x + y2 t + t + x2 + y t + t + ω t + t + x t + t + y t + t + ω + x t + t + ω + y t + t + ω + x + y t + t + ω + ω2 t + t + ω t + t + x + x2 + y t + t + ω + x + x2 + x3 t + t + ω + y + y2 t + t + ω + y + y2 + y3 t + t + ω + ω2 + x t + t + x + x2 t + t + x + x2 + x3 t + t + y + y2 t + t + y + y2 + y3 t + t + ω + ω2 + x t + t + ω + ω2 + x + y t + t + ω + ω2 + x + x2 + y + y2 t + t + ω + ω2 + x + x3 + y2 + y3 Table B2. Systematics model selection for the STIS white light curves Model χ2 BIC n d.o.f i (degrees) a/R(cid:63) Rp/R(cid:63) visit 52 1 46.55 72.92 27 19 85.35 7.60 0.1677 Table B2 continued WASP-52b Table B2 (continued) 33 Model χ2 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 visit 53 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 43.79 34.76 30.09 45.82 36.48 30.09 33.54 29.88 29.39 43.40 42.23 29.28 32.63 25.93 24.98 33.51 34.97 34.32 26.00 25.04 33.52 29.38 25.48 24.59 46.30 42.60 43.78 45.11 43.80 46.04 46.09 43.27 43.70 43.30 43.61 43.54 44.95 33.46 41.64 41.60 43.25 45.27 40.70 43.61 43.33 43.29 43.26 38.82 BIC 76.74 67.72 63.05 75.48 66.15 59.76 69.80 62.84 65.64 76.36 78.48 65.54 72.18 62.18 64.53 73.06 67.93 70.57 58.96 61.30 69.78 68.93 71.62 77.32 72.66 75.56 76.73 78.06 73.46 75.70 75.75 79.52 76.66 79.55 76.57 79.79 81.21 73.01 77.90 81.15 82.80 78.23 76.95 76.57 79.59 79.54 82.81 84.96 n 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 d.o.f i (degrees) a/R(cid:63) Rp/R(cid:63) 17 17 17 18 18 18 16 17 16 17 16 16 15 16 15 15 17 16 17 16 16 15 13 11 19 17 17 17 18 18 18 16 17 16 17 16 16 15 16 15 15 17 16 17 16 16 15 13 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 0.1676 0.1688 0.1691 0.1676 0.1685 0.1691 0.1683 0.1690 0.1690 0.1675 0.1673 0.1693 0.1687 0.1696 0.1696 0.1683 0.1689 0.1689 0.1696 0.1695 0.1683 0.1690 0.1696 0.1697 0.1656 0.1655 0.1654 0.1658 0.1655 0.1657 0.1657 0.1656 0.1655 0.1656 0.1660 0.1659 0.1658 0.1659 0.1653 0.1653 0.1658 0.1657 0.1659 0.1654 0.1653 0.1658 0.1657 0.1658 Table B2 continued 34 ALAM ET AL. Table B2 (continued) Model χ2 BIC 25 26.00 78.73 visit 54 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 58.65 50.54 45.06 54.79 51.13 48.52 56.89 38.76 46.54 35.87 49.75 51.13 46.63 33.56 46.18 45.65 38.75 47.33 41.75 56.75 56.24 38.75 35.87 30.89 26.00 81.73 80.20 74.72 84.45 77.50 74.89 83.26 68.42 76.20 68.83 79.41 77.50 79.59 69.82 79.14 81.90 71.71 76.99 74.70 86.41 89.20 71.71 72.12 73.73 68.84 n 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 d.o.f i (degrees) a/R(cid:63) Rp/R(cid:63) 11 20 18 18 18 19 19 19 18 18 17 18 19 17 16 17 16 17 18 17 18 17 17 16 14 14 85.35 7.60 0.1689 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 85.35 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 7.60 0.1681 0.1684 0.1671 0.1685 0.1682 0.1673 0.1681 0.1673 0.1681 0.1674 0.1681 0.1682 0.1678 0.1669 0.1681 0.1681 0.1673 0.1677 0.1672 0.1680 0.1681 0.1673 0.1674 0.1670 0.1675 WASP-52b 35 Table B3. STIS G430L (visit 52) transmission spectrum results for WASP-52b λ (Å) (Rp/R∗)uncorr (Rp/R∗)corr c1 c2 c3 c4 2900−3700 3700−3950 3950−4113 4113−4250 4250−4400 4400−4500 4500−4600 4600−4700 4700−4800 4800−4900 4900−5000 5000−5100 5100−5200 5200−5300 5300−5400 5400−5500 5500−5600 5600−5700 0.15814 ± 0.00712 0.16152 ± 0.00285 0.16997 ± 0.00300 0.16770 ± 0.00162 0.16683 ± 0.00151 0.16366 ± 0.00144 0.16610 ± 0.00151 0.16833 ± 0.00108 0.16453 ± 0.00191 0.16528 ± 0.00117 0.16317 ± 0.00159 0.16720 ± 0.00107 0.16898 ± 0.00178 0.16782 ± 0.00104 0.16556 ± 0.00116 0.16641 ± 0.00133 0.16496 ± 0.00110 0.16367 ± 0.00156 0.17008 ± 0.00444 0.16801 ± 0.00294 0.16293 ± 0.00181 0.16633 ± 0.00128 0.16379 ± 0.00175 0.16739 ± 0.00175 0.16536 ± 0.00170 0.16574 ± 0.00158 0.16496 ± 0.00150 0.16532 ± 0.00134 0.16626 ± 0.00176 0.16696 ± 0.00135 0.16612 ± 0.00159 0.16525 ± 0.00103 0.16470 ± 0.00200 0.16719 ± 0.00122 0.16675 ± 0.00124 0.16527 ± 0.00101 0.4371 0.7648 0.4778 0.4831 0.6151 0.4691 0.4777 0.5977 0.4411 0.5159 0.4399 0.5409 0.5605 0.5381 0.5732 0.5975 0.6370 0.5662 -0.7679 -1.1573 -0.6459 -0.6750 -0.8347 -0.4858 -0.4094 -0.6932 -0.2389 -0.3136 -0.1314 -0.3989 -0.4363 -0.2913 -0.3581 -0.4018 -0.4886 -0.2476 1.6319 1.7190 1.6564 1.7025 1.7798 1.5912 1.5029 1.6924 1.1548 1.2538 1.0138 1.1899 1.0910 1.0934 1.1751 1.1388 1.2206 0.9413 -0.3645 -0.4157 -0.5511 -0.5879 -0.6519 -0.6570 -0.6494 -0.6811 -0.4505 -0.5560 -0.4335 -0.4566 -0.3757 -0.4736 -0.5264 -0.4758 -0.5147 -0.4076 Table B4. STIS G430L (visit 53) transmission spectrum results for WASP-52b λ (Å) (Rp/R∗)uncorr (Rp/R∗)corr c1 c2 c3 c4 2900−3700 3700−3950 3950−4113 4113−4250 4250−4400 4400−4500 4500−4600 4600−4700 4700−4800 4800−4900 4900−5000 5000−5100 5100−5200 5200−5300 5300−5400 5400−5500 5500−5600 5600−5700 0.15814 ± 0.00712 0.16152 ± 0.00285 0.16997 ± 0.00300 0.16770 ± 0.00162 0.16683 ± 0.00151 0.16366 ± 0.00144 0.16610 ± 0.00151 0.16833 ± 0.00108 0.16453 ± 0.00191 0.16528 ± 0.00117 0.16317 ± 0.00159 0.16720 ± 0.00107 0.16898 ± 0.00178 0.16782 ± 0.00104 0.16556 ± 0.00116 0.16641 ± 0.00133 0.16496 ± 0.00110 0.16367 ± 0.00156 0.15421 ± 0.00677 0.15781 ± 0.00296 0.16681 ± 0.00296 0.16457 ± 0.00159 0.16394 ± 0.00148 0.16101 ± 0.00142 0.16358 ± 0.00149 0.16583 ± 0.00106 0.16211 ± 0.00187 0.16290 ± 0.00115 0.16079 ± 0.00156 0.16463 ± 0.00105 0.16623 ± 0.00175 0.16541 ± 0.00103 0.16330 ± 0.00114 0.16416 ± 0.00132 0.16279 ± 0.00108 0.16157 ± 0.00154 0.4371 0.7648 0.4778 0.4831 0.6151 0.4691 0.4777 0.5977 0.4411 0.5159 0.4399 0.5409 0.5605 0.5381 0.5732 0.5975 0.6370 0.5662 -0.7679 -1.1573 -0.6459 -0.6750 -0.8347 -0.4858 -0.4094 -0.6932 -0.2389 -0.3136 -0.1314 -0.3989 -0.4363 -0.2913 -0.3581 -0.4018 -0.4886 -0.2476 1.6319 1.7190 1.6564 1.7025 1.7798 1.5912 1.5029 1.6924 1.1548 1.2538 1.0138 1.1899 1.0910 1.0934 1.1751 1.1388 1.2206 0.9413 -0.3645 -0.4157 -0.5511 -0.5879 -0.6519 -0.6570 -0.6494 -0.6811 -0.4505 -0.5560 -0.4335 -0.4566 -0.3757 -0.4736 -0.5264 -0.4758 -0.5147 -0.4076
1108.4493	2	1108	2011-10-05T06:31:26	Further Observations of the Tilted Planet XO-3: A New Determination of Spin-Orbit Misalignment, and Limits on Differential Rotation	[ "astro-ph.EP" ]	We report on observations of the Rossiter-McLaughlin (RM) effect for the XO-3 exoplanetary system. The RM effect for the system was previously measured by two different groups, but their results were statistically inconsistent. To obtain a decisive result we observed two full transits of XO-3b with the Subaru 8.2-m telescope. By modeling these data with a new and more accurate analytic formula for the RM effect, we find the projected spin-orbit angle to be \lambda=37.3 deg \pm 3.0 deg, in good agreement with the previous finding by Winn et al. (2009). In addition, an offset of ~22 m/s was observed between the two transit datasets. This offset could be a signal of a third body in the XO-3 system, a possibility that should be checked with future observations. We also attempt to search for a possible signature of the stellar differential rotation in the RM data for the first time, and put weak upper limits on the differential rotation parameters.	astro-ph.EP	astro-ph	PASJ: Publ. Astron. Soc. Japan , 1 -- ??, c(cid:13) 2018. Astronomical Society of Japan. 1 1 0 2 t c O 5 . ] P E h p - o r t s a [ 2 v 3 9 4 4 . 8 0 1 1 : v i X r a Further Observations of the Tilted Planet XO-3: A New Determination of Spin-Orbit Misalignment, and Limits on Differential Rotation∗ Teruyuki Hirano,1,2 Norio Narita,3 Bun'ei Sato,4 Joshua N. Winn,2 Wako Aoki,3 Motohide Tamura,3 Atsushi Taruya1 and Yasushi Suto1,5 [email protected] 1 Department of Physics, The University of Tokyo, Tokyo, 113-0033, Japan 2 Department of Physics, and Kavli Institute for Astrophysics and Space Research, Massachusetts Institute of Technology, Cambridge, MA 02139, USA 3 National Astronomical Observatory of Japan, 2-21-1 Osawa, Mitaka, Tokyo, 181-8588, Japan 4 Department of Earth and Planetary Sciences, Tokyo Institute of Technology, 2-12-1 Ookayama, Meguro-ku, Tokyo 152-8551 5Department of Astrophysical Sciences, Princeton University, Princeton, NJ 08544, USA (Received 2011 August 7; accepted 2011 September 29) Abstract We report on observations of the Rossiter-McLaughlin (RM) effect for the XO-3 exoplanetary system. The RM effect for the system was previously measured by two different groups, but their results were statistically inconsistent. To obtain a decisive result we observed two full transits of XO-3b with the Subaru 8.2-m telescope. By modeling these data with a new and more accurate analytic formula for the RM effect, we find the projected spin-orbit angle to be λ = 37.3◦ ± 3.0◦, in good agreement with the previous finding by Winn et al. (2009). In addition, an offset of ∼ 22 m s−1 was observed between the two transit datasets. This offset could be a signal of a third body in the XO-3 system, a possibility that should be checked with future observations. We also attempt to search for a possible signature of the stellar differential rotation in the RM data for the first time, and put weak upper limits on the differential rotation parameters. Key words: stars: planetary systems: individual (XO-3) -- stars: rotation -- techniques: radial velocities -- techniques: spectroscopic -- 1. Introduction Since the discovery of the first transiting planet, many groups have been studying the stellar obliquities (spin- orbit angles) of planet-hosting stars through measure- ments of the Rossiter-McLaughlin (RM) effect. The RM effect is a distortion of stellar spectral lines that occurs during transits, originating from the partial occultation of the rotating stellar surface. It is often manifested as a pat- tern of anomalous radial velocities (RVs) during a plane- tary transit (Queloz et al. 2000; Ohta et al. 2005; Winn et al. 2005; Narita et al. 2007; Triaud et al. 2010). By modeling the RM effect, one can determine the angle λ between the sky projections of the stellar rotational axis and the orbital axis. The statistics of the spin-orbit angle λ should provide a clue to the formation and evolution of close-in giant planets (hot-Jupiters and hot-Neptunes). Since 2008, many transiting systems with significant spin- orbit misalignments have been reported (e.g. H´ebrard et al. 2008; Narita et al. 2009b; Pont et al. 2010). This has attracted much attention to the importance of dy- namical mechanisms for producing close-in planets, as well as tidal evolution of planets and their host stars (Fabrycky & Tremaine 2007; Wu et al. 2007; Nagasawa et * Based on data collected at Subaru Telescope, which is operated by the National Astronomical Observatory of Japan. al. 2008; Chatterjee et al. 2008; Triaud et al. 2010; Winn et al. 2010). In this paper, we present the measurement of the RM effect for the XO-3 system. The XO-3 system was discov- ered by Johns-Krull et al. (2008). Photometric follow-ups by Winn et al. (2008) allowed the system parameters to be refined. The large mass of the planet (Mp = 11.79 ± 0.59MJup) and its eccentric orbit (e = 0.260 ± 0.017) at- tracted further interest in this system. H´ebrard et al. (2008) detected the RM effect with the SOPHIE instrument on the 1.93 m telescope at Haute Provence Observatory (OHP). They found λ = 70◦ ± 15◦, suggesting a significant spin-orbit misalignment for the first time among the known planetary systems. On the other hand, Winn et al. (2009) independently mea- sured the RM effect with the High Resolution Echelle Spectrometer (HIRES) installed on the Keck I telescope, and found λ = 37.3◦ ± 3.7◦, which differs by more than 2 σ from the former result. The reason for the discrepancy was unclear, but it may indicate the presence of unknown systematic errors in one of the datasets, or even in both. It is equally possible that the discrepancy should be ascribed to the different tech- niques adopted in modeling the RM effect. H´ebrard et al. (2008) and Winn et al. (2009) both used analytic formu- lae to compute the anomalous RVs, but the former group used a formula that was based on a calculation of the first 2 moment of the distorted line profile, while the latter group used a formula that was calibrated by numerical analysis of simulated RM spectra. Recently Hirano et al. (2011) presented a new and more accurate analytic formula for the RM effect, showing in particular that the RM velocity anomaly depends on many factors such as the rotational velocity of the star, the macroturbulent velocity, and even the instrumental profile (IP) of the spectrograph, not all of which were considered in the previous literatures. Specifically, for rapidly rotating stars like XO-3, the velocity anomaly calculated by Hirano et al. (2011) dif- fers strongly from the simpler, previous analytic descrip- tions based on the first-moment approach (Ohta et al. 2005; Gim´enez 2006). When the incorrect relation is used between the RM velocity anomaly and the position of the planet, the results for λ may be biased. In order to resolve the disagreement, and obtain a de- cisive result for the angle λ with fewer systematic errors, we observed another two full transits of XO-3b with the High Dispersion Spectrograph (HDS) on the Subaru 8.2- m telescope. We also applied the new analytic formula by Hirano et al. (2011) to model the RM effect with greater accuracy. We find that the best-fit value for λ based on our new measurements is very close to that reported by Winn et al. (2009). We describe the detail of the observation in Section 2. The data analysis procedure and the derived parameters are presented in Section 3. Section 4 discusses the compar- ison with the previous results, and considers the possible effect of the stellar differential rotation. 2. Observations We observed two complete transits of XO-3b with Subaru/HDS on November 29, 2009 and February 4, 2010 (UT). We also obtained several out-of-transit spectra on each of those two nights as well as on January 15, 2010 (UT). The out-of-transit spectra were obtained in order to help establish the Keplerian orbital parameters of the system. We adopted a typical exposure time as 600-750 seconds, and chose the slit width as 0.4′′, corresponding to the spectral resolution of ∼ 90,000. We used the Iodine cell for precise RV calibration. We reduced the images to one-dimensional (1D) spectra using standard IRAF procedures. The typical signal-to- noise ratio was ∼100 per pixel in the 1D spectra. We then processed the reduced spectra with the RV analysis routines for Subaru/HDS developed by Sato et al. (2002). Table 1 gives the resulting RVs (corrected for the motion of the Earth) and the associated errors, which are com- puted from the dispersion of RVs that were determined from individual 4 A segments of the spectrum (Sato et al. 2002). We obtained a typical RV precision of 11-14 m s−1. 3. Analysis and Results 3.1. Fit to the Subaru RV data alone We determined the projected spin-orbit angle λ in sev- eral steps. First, in order to provide an independent de- Table 1. Radial velocities measured with Subaru/HDS. [Vol. , Time [BJD (TDB)] Relative RV [m s−1] Error [m s−1] 2455164.703174 2455164.711814 2455164.719564 2455164.727304 2455164.735054 2455164.742804 2455164.750544 2455164.758275 2455164.766005 2455164.773745 2455164.781485 2455164.789205 2455164.796945 2455164.804675 2455164.812405 2455164.820135 2455164.827875 2455164.835605 2455164.839555 2455211.720046 2455211.729516 2455211.738975 2455231.709440 2455231.718219 2455231.725949 2455231.733688 2455231.741418 2455231.749157 2455231.756887 2455231.764636 2455231.772366 2455231.780095 2455231.787834 2455231.795574 2455231.803313 2455231.811063 2455231.818802 2455231.826532 2455231.834271 2455231.842001 2455231.849740 2455231.857469 523.9 514.4 497.3 457.9 404.1 358.9 333.2 277.8 232.3 198.3 195.7 149.4 111.1 112.2 109.7 130.9 157.7 124.0 166.4 766.1 774.0 844.1 492.8 524.8 506.8 480.2 427.6 444.2 416.1 356.7 283.6 308.1 220.1 183.6 179.2 132.0 127.6 70.5 72.9 122.6 139.6 110.5 13.4 13.5 13.4 13.1 11.5 13.5 12.8 13.0 12.7 12.5 13.1 12.9 12.8 12.2 13.0 11.9 12.7 12.4 21.8 14.5 13.7 14.2 14.0 13.3 13.9 13.3 13.5 13.1 12.5 13.8 14.9 14.2 13.0 12.2 13.4 12.8 12.8 12.3 12.0 13.2 13.1 13.2 termination, we use only the transit data from our new Subaru observations. Since those data alone are insuffi- cient to determine all the Keplerian orbital parameters of the system, we also use the out-of-transit RV data points from OHP/SOPHIE (H´ebrard et al. 2008). This essen- tially provides an independent determination of λ since we do not use the in-transit RV data from OHP/SOPHIE. Our model for the RVs is similar in some respects to the previous analyses by Narita et al. (2009a) and Narita et al. (2010). Each RV data set (Subaru/HDS and OHP/SOPHIE) is modeled as Vmodel = K[cos(f + ) + e cos()] + ∆vRM + γoffset, (1) 2500 2000 1500 1000 500 0 -500 -1000 ] 1 - s m [ V R ] 1 - s m [ l i a u d s e R 100 50 0 -50 -100 160 2000 1500 1000 500 0 -500 -1000 100 50 0 -50 -100 ] 1 - s m [ V R ] 1 - s m [ l i a u d s e R 3 Subaru out-of-transit best-fit model 170 180 200 190 BJD - 2455000 210 220 230 OHP Subaru best-fit model -0.4 -0.2 0 0.2 0.4 Orbital Phase (Upper) New RV data outside of transits, obtained Fig. 1. with Subaru/HDS. (Lower) The orbit of XO-3b based on the measurements with Subaru/HDS (blue), and the previously published RVs obtained with OHP/SOPHIE (black). For this figure, a linear RV trend ( γ) was fitted to the data and then subtracted. For each of the figures above, the best-fit model is shown as a red curve and the RV residuals from the best-fit model are plotted at the bottom. No. ] where K is the orbital RV semi-amplitude, f is the true anomaly, e is the orbital eccentricity, is the angle be- tween the direction of the pericenter and the line of sight, and finally γ is a constant offset for the data from a given spectrograph. The RM velocity anomaly ∆vRM is modeled with Equation (16) of Hirano et al. (2011). In order to compute ∆vRM, we adopt the following values for the basic spectro- scopic parameters; the macroturbulence dispersion ζ = 6.0 km s−1, the Gaussian dispersion (including the instrumen- tal profile) β = 3.0 km s−1, and the Lorentzian dispersion γ = 1.0 km s−1. These values are taken from Gray (2005) and from the comparison with the numerical simulations by Hirano et al. (2011). Also, we assume the quadratic limb darkening law with u1 = 0.32 and u2 = 0.36 following Claret (2004). We fit the two RV data sets (Subaru and OHP) by min- imizing obs − V (i) model σ(i) χ2 =Xi " V (i) #2 , (2) where V (i) obs is the observed RV value labeled by i while V (i) model corresponds to Equation (1). The uncertainty for each RV point is expressed by σ(i). Since we do not have any new photometric observations of the transit, we fix the photometrically measured parameters to be Rp/Rs = 0.09057, a/Rs = 7.07, and io = 84.2◦ from the refined parameter set by Winn et al. (2008). The remain- ing parameters are K, e, , γoffset (for each data set), the rotational velocity of the star v sin is, and the spin- orbit angle λ. We allow all the parameters to vary freely to minimize χ2, using the AMOEBA algorithm. We add the stellar jitter of σjitter = 13.4 m s−1 in quadrature to the RV uncertainties in Table 1 so that the reduced χ2 in the global RV fitting becomes unity (after adding an ad- ditional parameter to allow for an offset between the two Subaru transits, as explained below). This jitter is ac- counted for in estimating the uncertainty for the system parameters in Table 2. By fitting the Subaru/HDS data along with the out-of- transit OHP/SOPHIE data, we find the spin-orbit angle to be λ = 36.7◦ ± 3.0◦. This is in agreement with the pre- vious finding by Winn et al. (2009), and in disagreement with the previous finding by H´ebrard et al. (2008). The reduced chi-squared is χ2 = 1.14. Interestingly, when we plot the residuals between the Subaru/HDS data and the best-fit model, we find a small negative trend as a function of time over the 67-day span of the observations. To show this, we plot our new out-of-transit RV data as a function of BJD in the upper panel of Figure 1, along with the best-fit curve (red). The residuals from the best-fit curve are shown at the bottom. This trend cannot be corrobo- rated or refuted by the previously published observations; the RV precision obtained by Johns-Krull et al. (2008) and H´ebrard et al. (2008) was insufficient, and the precise RV measurements of Winn et al. (2009) did not cover a sufficiently long observation period. 4 Table 2. The best-fit parameter sets. Parameter K [m s−1] e ω [◦] v sin is [km s−1] λ [◦] γ [m s−1 day−1] −0.322 ± 0.088 χ2 (A) Subaru 1499.5 ± 9.9 0.2859+0.0028 −0.0027 347.4 ± 1.4 17.0 ± 1.2 37.3 ± 3.0 0.91 (B) Subaru + Keck 1494.0 ± 9.5 0.2883 ± 0.0025 346.1+1.2 −1.1 18.4 ± 0.8 37.4 ± 2.2 −0.320 ± 0.088 1.00 (fixed) This RV trend might indicate a possible additional body in the XO-3 system, but it is obviously premature to con- clude so only with the 3 epochs of data. Future obser- vations are needed. For the present purpose, in order to account for the offset in the overall RV between the dif- ferent transit epochs, we introduced an additional model parameter γ, representing a constant radial acceleration. We then refitted the Subaru RV data. The results from this fit are given in column (A) in Table 2. The uncer- tainty for each parameter is derived by the criteria that ∆χ2 becomes unity. The inclusion of the constant ac- celeration improves the reduced chi-squared significantly ( χ2 = 0.91 from 1.14 in the absence of γ) and the best-fit RV acceleration is γ = −0.322 ± 0.088 m s−1 day−1, indi- cating a 3.6σ detection. Since the two transit observations are separated by 67 days, the RV offset between the two transits is estimated as ∼ 22 m s−1. The resultant RVs as a function of the orbital phase are shown in the lower panel of Figure 1. 3.2. Joint fit to the Subaru and Keck data Now that we have seen that our new results by Subaru/HDS support the previous RM measurement by Keck/HIRES (Winn et al. 2009), we would like to try to combine the two independent measurements (Subaru and Keck) and carry out a joint analysis in order to derive the parameters with greater precision. We fit all of the transit data from Subaru/HDS and Keck/HIRES, and also the out-of-transit data from OHP/SOPHIE. We allow for a constant RV acceleration γ, as before. We estimate the best-fit values for K, e, , v sin is, λ, and γ as in Section 3.1. The results are summarized in the column (B) of Table 2. Most of the values are very close to the best-fit values in case (A). The projected rotation rate of v sinis = 18.4 ± 0.8 km s−1 is in good agreement with the spectroscopically measured value (v sin is = 18.54 ± 0.17 km s−1, Johns-Krull et al. 2008). The resulting phase-folded RV anomalies during transits are plotted in Figure 2, in which the Keplerian motion and the linear RV trend are subtracted from the data. The RV data taken by Subaru/HDS are indicated in blue for the first transit and purple for the second tran- sit, and those by Keck/HIRES are shown in green. The red solid curve is the best-fit curve based on the analytic formula of Hirano et al. (2011). 100 50 0 -50 -100 -150 ] 1 - s m [ v ∆ ] 1 - s m [ l i a u d s e R -200 80 60 40 20 0 -20 -40 -60 -80 -0.03 [Vol. , Keck Subaru 1st visit Subaru 2nd visit -0.02 -0.01 0 0.01 0.02 0.03 Orbital Phase Fig. 2. RV data spanning the transit, after subtracting the orbital contributions to the velocity variation, and also a lin- ear function of time. The plotted data includes the new Subaru/HDS data (blue for the fist transit on UT 2009 Nov. 29 and purple for the second transit on UT 2010 Feb. 4) and the previously published Keck/HIRES data taken on UT 2009 Feb. 2 (green). The RV residuals are plotted at the bottom. 4. Discussion and Summary We have investigated the RM effect for the XO-3 sys- tem, which was the first confirmed system with a signifi- cant spin-orbit misalignment (H´ebrard et al. 2008). The new spectroscopic measurements including two full tran- sits taken by Subaru/HDS and the new analysis method using the analytic formula for the RM effect by Hirano et al. (2011), found the spin-orbit angle of λ = 37.3◦ ± 3.0◦, supporting the result by Winn et al. (2009) based on the measurement with Keck/HIRES. The joint analysis of all the RV data sets covering three transits with Subaru/HDS and Keck/HIRES have shown that the projected stellar spin velocity estimated by the RM analysis well agrees with the spectroscopically measured value. Our analysis also detected an RV trend, or at least RV offsets, among the three epochs of the Subaru/HDS obser- vations. The cause of the extra RV variation is not clear. It is possibly an indication of a third body in the sys- tem: a stellar companion (binary), or an additional mas- sive planet. Nevertheless it should be noted that this star is known to have a high "RV jitter" of around 15 m s−1, and the precise physical causes and timescales of the jitter are not known. It is possible for starspots or other surface inhomogeneities being carried around by stellar rotation to produce a systematic offset in RV observations con- ducted on a single night. Since the rotational velocity of the star is large for a planet-hosting star, even a rela- tively small spot could cause an apparent RV anomaly in No. ] a similar manner as the RM effect. For example, the RV acceleration of 22 m s−1 could be caused by a very dark spot whose size is only 0.002 of the total stellar disk area. The best way to investigate these possibilities is with ad- ditional measurements of the out-of-transit RV variation, with a precision better than 15 m s−1. As for the results for λ, we would like to understand the reason for the discrepancy between the OHP/SOPHIE re- sults, and the Subaru/HDS + Keck/HIRES results. To this end we try several additional tests. As we have pointed out, H´ebrard et al. (2008) employed the analytic formula based on the first moment of the distorted line profiles to describe the RM effect (Ohta et al. 2005). For rapidly rotating stars, however, the RM velocity anomaly computed from the first moment significantly deviates from that based on the cross-correlation method (Hirano et al. 2010). Therefore, we reanalyze the OHP data using the new analytic formula by Hirano et al. (2011) to see if the original estimate for the spin-orbit angle λ was biased. Instead of fixing the stellar spin velocity v sin is as done by H´ebrard et al. (2008), we allow it to be a free param- eter, and fit all the OHP RV data (H´ebrard et al. 2008). The resulting spin-orbit angle is λ = 58.8◦ ± 8.9◦ and the stellar spin velocity of vsinis = 15.9±2.6 km s−1. The cen- tral value for λ approaches our new results (37.4◦ ± 2.2◦), but they still disagree with each other with > 2σ. This shows that a biased model played only a minor role in the discrepancy. The major reason seems to have been sys- tematic effects in the OHP/SOPHIE dataset, perhaps due to the short-term or long-term instrumental systematics (instability) for fainter objects as reported by Husnoo et al. (2011). Incidentally, with only a small modification, the ana- lytic formula by Hirano et al. (2011) can also be used to calculate the RM velocity anomaly in the presence of dif- ferential rotation (DR). The detection of DR would be of great interest for understanding the convective/rotational dynamics of the host star. Furthermore, it allows a possi- bility to break the degeneracy between the projected and the real three-dimensional spin-orbit misalignment angle by inferring the inclination angle of the stellar spin axis with respect to the line of sight (is), an angle that is or- dinarily not measurable with the RM observations (if the star is a solid rotator). Since XO-3 has a comparably large vsinis, our new data may provide a good opportunity to search for the signa- ture of DR, or at least to put constraints on the degree of DR quantitatively. To model DR, we introduce two major parameters: the stellar inclination is and the coefficient of DR, α. The stellar angular velocity Ω as a function of the latitude l on the stellar surface is written as Ω(l) = Ωeq(1 − αsin2 l), where Ωeq is the angular velocity at the equator (Reiners 2003b). We step through a two-dimensional grid in α and cosis, and for each grid point we fit the RVs with the six parameters listed in Table 2. We compute the resulting χ2 at each point (α, cos is). We note here that the DR of our Sun is well described by α ≃ 0.2. This also seems to be a typical value of other stars based on the spectral 5 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 s i s o c ∆χ2= 1.0 0 -0.2 -0.15 -0.1 -0.05 ∆χ2= 0.0 0.05 0.1 0.15 0.2 0 α Fig. 3. Contour plot of ∆χ2 in the space of the DR parame- ters α and is. The confidence region where ∆χ2 ≤ 1.0 is sur- rounded by the red solid curve. We also show the confidence boundary of ∆χ2 = 2.30 by the black dashed line, which de- termines the 1σ region in a two-dimensional parameter space. line analysis of (Reiners & Schmitt 2003a), although those authors also point out that some stars may have "anti- solar" like differential rotations in which α < 0. Thus, our grid extends from −0.2 ≤ α ≤ 0.2 and 0 ≤ cos is ≤ 0.95. The case where cos is > 0.95 is very unlikely because the star would need to be rotating unrealistically rapidly to give the observed value of v sin is. Figure 3 shows contours of ∆χ2 ≡ χ2 − χ2 min in the (α, cos is) plain. The location of the best-fit model (defining the condition ∆χ2 = 0.0) is plotted with a black cross. This figure shows that with the current RV data, we are only able to provide fairly weak constraints on the param- eters. We are able to rule out the far upper left and right corners of this parameter space, corresponding to Solar- like DR viewed at low inclinations. We can rule out much stronger levels of DR (α > ∼ 0.5) regardless of orientation, but such strong levels of differential rotation are unlikely in any case. The non-detection of DR may be ascribed to the large stellar jitter of the host star (15 m s−1). This is often typical of relatively hot and rapidly rotating stars such as XO-3. The best cases for studying DR through the RM effect would be somewhat cooler stars that are still moderately rapid rotators (≈ 5-10 km s−1), for which a greater signal-to-noise ratio can be obtained. We acknowledge the support for our Subaru HDS observations by Akito Tajitsu, a support scientist for the Subaru HDS. The data analysis was in part car- ried out on common use data analysis computer system at the Astronomy Data Center, ADC, of the National Astronomical Observatory of Japan. T.H. is supported by Japan Society for Promotion of Science (JSPS) Fellowship for Research (DC1: 22-5935). N.N. acknowledges a support by NINS Program for Cross-Disciplinary Study. J.N.W. acknowledges support from the NASA Origins program (NNX11AG85G) as well as the Keck PI Data 6 [Vol. , Analysis Fund. M.T. is supported by the Ministry of Education, Science, Sports and Culture, Grant-in-Aid for Specially Promoted Research, 22000005. Y.S. grate- fully acknowledges support from the Global Collaborative Research Fund (GCRF) "A World-wide Investigation of Other Worlds" grant and the Global Scholars Program of Princeton University. Finally, we wish to express spe- cial thanks to the referee, Fr´ed´eric Pont, for his helpful comments on this paper. References Claret A. 2004, A&A, 428, 1001 Chatterjee, A., Sarkar, A., Barat, P., Mukherjee, P., & Gayathri, N. 2008, ApJ, 686, 580 Fabrycky, D., & Tremaine, S. 2007, ApJ, 669, 1298 Gim´enez, A. 2006, ApJ, 650, 408 Gray, D. F. 2005, The Observation and Analysis of Stellar Photospheres (3rd ed.; Cambridge University Press.) H´ebrard, G., et al. 2008, A&A, 488, 763 Hirano, T., et al. 2010, ApJ, 709, 458 Hirano, T., et al. 2011, ApJ, in press Husnoo, N., et al. 2011, MNRAS, 413, 2500 Johns-Krull, C. M., et al. 2008, ApJ, 677, 657 Nagasawa, M., Ida, S., & Bessho, T. 2008, ApJ, 678, 498 Narita, N., et al. 2007, PASJ, 59, 763 Narita, N., et al. 2009a, PASJ, 61, 991 Narita, N., Sato, B., Hirano, T., & Tamura, M. 2009b, PASJ, 61, L35 Narita, N., et al. 2010, PASJ, 62, L61 Ohta, Y., Taruya, A., & Suto Y. 2005, ApJ, 622, 1118 Pont, F., et al. 2010, MNRAS, 402, L1 Queloz, D., Eggenberger, A., Mayor, M., Perrier, C., Beuzit, J. L., Naef, D., Sivan, J. P., & Udry, S. 2000, A&A, 359, L13 Reiners, A., & Schmitt, J. H. M. M. 2003, A&A, 398, 647 Reiners, A. 2003, A&A, 408, 707 Sato, B., Kambe, E., Takeda, Y., Izumiura, H., & Ando, H. 2002, PASJ, 54, 873 Triaud, A. H. M. J., et al. 2010, A&A, 524, A25 Winn, J. N., et al. 2005, ApJ, 631, 1215 Winn, J. N., et al. 2008, ApJ, 683, 1076 Winn, J. N., et al. 2009, ApJ, 700, 302 Winn, J. N., Fabrycky, D., Albrecht, S., & Johnson, J. A. 2010a, ApJ, 718, L145 Wu, Y., Murray, N. W., & Ramsahai, J. M. 2007, ApJ, 670, 820
1801.02814	2	1801	2018-01-10T14:00:36	Breakthrough Listen Observations of 1I/'Oumuamua with the GBT	[ "astro-ph.EP", "astro-ph.IM" ]	We have conducted a search for radio emission consistent with an artificial source targeting 1I/'Oumuamua with the Robert C. Byrd Green Bank Telescope (GBT) between 1.1 and 11.6 GHz. We searched the data for narrowband signals and found none. Given the close proximity to this interstellar object, we can place limits to putative transmitters with extremely low power (0.08 W).	astro-ph.EP	astro-ph	Breakthrough Listen Observations of 1I/′Oumuamua with the GBT J. Emilio Enriquez,1, 2 Andrew Siemion,1, 2, 3 T. Joseph W. Lazio,4 Matt Lebofsky,1 David H. E. MacMahon,1 Ryan S. Park,4 Steve Croft,1 David DeBoer,1 Nectaria Gizani,1, 5 Vishal Gajjar,6 Greg Hellbourg,1 Howard Isaacson,1 and Danny C. Price1, 7 1Department of Astronomy, University of California, Berkeley 2Department of Astrophysics/IMAPP,Radboud University, Nijmegen, Netherlands 3SETI Institute, Mountain View, California 4Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California 5School of Science and Technology, Hellenic Open University, Greece 6Space Sciences Laboratory, University of California, Berkeley 7Centre for Astrophysics & Supercomputing, Swinburne University of Technology, Australia Keywords: asteroids – extraterrestrial intelligence INTRODUCTION. On 2017 October 18, the Pan-STARRS collaboration discovered an object within our solar system that appeared to be on a hyperbolic orbit (Williams 2017). Subsequent observations suggest this object (now designated 1I/2017 U1 or 1I/′Oumuamua) is of interstellar origin, lacks a coma, and that its shape is highly elongated relative to other known asteroids (Meech et al. 2017). It has long been suggested that advanced extraterrestrial civilizations, should they exist, could conceivably send probes to other stars either for exploration or communication purposes (Bracewell 1960; Freitas 1980; Rose & Wright 2004; Gertz 2016). Interstellar probes would likely be equipped with communication technology that could potentially be operating in the radio band. The Breakthrough Listen (BL) program (Worden et al. 2017) conducted an observing campaign targeting 1I/′Oumuamua with the Robert C. Byrd Green Bank Telescope (GBT), with a goal of detecting, or placing lim- its on, radio emission consistent with a technological source. Because 1I/′Oumuamua is much closer to Earth than typical stellar targets, we have the opportunity to carry out a search for extremely weak transmitters: unprecedented in any other SETI experiment. Here we present a description of the observing campaign, and preliminary results. OBSERVATIONS In December 2017, the BL Team conducted an initial 8-hour observation campaign using the GBT with four different receivers: L-band (1.1–1.9 GHz), S-band (1.73–2.6 GHz), C-band (4–8 GHz) and X-band (8–11.6 GHz). Observations were conducted using the Breakthrough Listen back-end (MacMahon et al. 2017) to collect and analyze the data. The 8-hour block was divided into two hours per receiver. Considering the suggested rotation of 1I/′Oumuamua, we subdivided the two hours per receiver into half-hour blocks, one for each phase quadrant of the rotation. Each half hour was further subdivided into six 5-minute observations in an ABACAD configuration as described in Isaacson et al. (2017) and Enriquez et al. (2017). We set the time of the first observation as t0 (2017 December 13 at 21:53:22 UTC; 58100.9121 MJD), and used a rotational period of 8.1 ±0.02 hrs from Bolin et al. (2017) for the phase calculation. Over two weeks, we covered the full rotational phase of 1I/′Oumuamua with each receiver. Table 1 shows the MJD dates relating receiver band coverage to quadrant phase. Corresponding author: J. Emilio Enriquez [email protected] 2 Table 1. Observation MJD dates for a given quadrant/band combination. The quadrants Q1, Q2, Q3 and Q4 are defined in phase space as 0.-0.25, 0.25-0.5, 0.5-0.75, and 0.75-1.0 respectively. L-band S-band C-band X-band Q1 58110.0323 58100.9121 58106.0037 58100.9791 Q2 58101.0313 58106.0570 58107.1261 58117.9094 Q3 58107.1768 58109.9187 58101.09403 58109.8753 Q4 58109.9994 58105.8901 58109.9749 58105.9473 RESULTS AND DISCUSSION We conducted a search of the data for narrow-band (3Hz resolution) drifting sinusoids as described by Enriquez et al. (2017), over a drift rate range of ±2 Hz/s. This range includes any acceleration in the geocentric and barycentric frames, as well as accommodating a transmitter located anywhere on the body itself. Our preliminary results show no narrow band radio emission from the direction of 1I/′Oumuamua at any rotational phase. The object was at ∼ 2 AU when observed, and given an approximate SEFD of 20 Jy, with a 300 s observation, and a 5σ threshold, these observations were sensitive to a hypothetical transmitter with an EIRP of ∼ 0.08 W (∼3,000 times weaker than the Dawn spacecraft communication down-link.). Based on the possibility that 1I/′Oumuamua could be in fact a dormant comet with delayed outgassing, we also searched for any indication of hydroxyl emission at the four transitions between 1612 MHz and 1720 MHz. We searched the L-band data taken during quadrant Q2 by stacking the three 5-min observations. No emission was detected, confirming previous observations during closer approach that the nature of the object is consistent with an asteroid- like composition (Park 2017 in prep). All data collected by BL will be publicly available. A subset of the observations described here can be downloaded from the Breakthrough Listen Public Web Archive. Funding for BL is provided by the Breakthrough Prize Foundation. Part of this research was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration. Bolin, B. T., Weaver, H. A., Fernandez, Y. R., et al. 2017, Park, R., Pisano, D., Lazio, T., Chodas, P., & Naidu, S. REFERENCES ArXiv e-prints, arXiv:1711.04927 Bracewell, R. N. 1960, Nature, 186, 670 Enriquez, J. E., Siemion, A., Foster, G., et al. 2017, ApJ, 849, 104 Freitas, R. A. J. 1980, British Interplanetary Society, 33, 95 Gertz, J. 2016, arXiv.org, 1609.04635v1 Isaacson, H., Siemion, A. P. V., Marcy, G. W., et al. 2017, PASP, 129, 054501 MacMahon, D., et al. 2017, PASP; in Press Meech, K. J., Weryk, R., Micheli, M., et al. 2017, Nature, 552, 378 EP 2017, submitted to AJ Rose, C., & Wright, G. 2004, Nature, 431, 47 Williams, G. 2017, MPEC 2017-U181: COMET C/2017 U1 (PANSTARRS), Minor Planet Electronic Circular, MPEC Worden, S. P., Drew, J., Siemion, A., et al. 2017, Acta Astronautica, 139, 98
1111.1007	1	1111	2011-11-03T22:34:59	The Pan-Pacific Planet Search. I. A Giant Planet Orbiting 7 CMa	[ "astro-ph.EP" ]	We introduce the Pan-Pacific Planet Search, a survey of 170 metal-rich Southern hemisphere subgiants using the 3.9m Anglo-Australian Telescope. We report the first discovery from this program, a giant planet orbiting 7 CMa (HD 47205) with a period of 763+/-17 days, eccentricity e=0.14+/-0.06, and m sin i=2.6+/-0.6 M_jup. The host star is a K giant with a mass of 1.5+/-0.3 M_sun and metallicity [Fe/H]=0.21+/-0.10. The mass and period of 7 CMa b are typical of planets which have been found to orbit intermediate-mass stars (M*>1.3 M_sun). Hipparcos photometry shows this star to be stable to 0.0004 mag on the radial-velocity period, giving confidence that this signal can be attributed to reflex motion caused by an orbiting planet.	astro-ph.EP	astro-ph	The Pan-Pacific Planet Search. I. A Giant Planet Orbiting 7 CMa Robert A. Wittenmyer1, Michael Endl2, Liang Wang3, John Asher Johnson4, C.G. Tinney1, S.J. O'Toole5 [email protected] ABSTRACT We introduce the Pan-Pacific Planet Search, a survey of 170 metal-rich South- ern hemisphere subgiants using the 3.9m Anglo-Australian Telescope. We re- port the first discovery from this program, a giant planet orbiting 7 CMa (HD 47205) with a period of 763±17 days, eccentricity e = 0.14±0.06, and m sin i=2.6±0.6 MJup. The host star is a K giant with a mass of 1.5±0.3 M⊙and metallicity [Fe/H]=0.21±0.10. The mass and period of 7 CMa b are typical of planets which have been found to orbit intermediate-mass stars (M∗ > 1.3M⊙). Hipparcos photometry shows this star to be stable to 0.0004 mag on the radial- velocity period, giving confidence that this signal can be attributed to reflex motion caused by an orbiting planet. Subject headings: planetary systems -- techniques: radial velocities -- stars: indi- vidual (HD 47205) 1. Introduction Nearly 20 years of concerted radial-velocity monitoring of solar-type main-sequence stars has unveiled a fascinating diversity of planets and planetary system configurations. From the many hundreds of planets now characterised, the observational evidence is mounting 1Department of Astrophysics, School of Physics, University of NSW, 2052, Australia 2McDonald Observatory, University of Texas at Austin, 1 University Station C1400, Austin, TX 78712, USA 3Key Laboratory of Optical Astronomy, National Astronomical Observatories, Chinese Academy of Sci- ences, A20 Datun Road, Chaoyang District, Beijing 100012, China 4Department of Astrophysics, California Institute of Technology, MC 249-17, Pasadena, CA 91125, USA 5Australian Astronomical Observatory, PO Box 296, Epping, 1710, Australia -- 2 -- for several interesting relationships between the properties of planets and their host stars. Among these are: (1) giant planet occurrence is positively correlated with stellar metallicity (Fischer & Valenti 2005) and mass (Johnson et al. 2010a; Endl et al. 2006), (2) short-period "Super-Earths" with m sin i <10 M⊕ are about an order of magnitude more common than close-in giant planets (Howard et al. 2010; Wittenmyer et al. 2011), and (3) planet mass is positively correlated with host star mass (Bowler et al. 2010). Most of the stars which have been targeted by radial-velocity surveys have masses which fall in the range 0.7-1.3 M⊙ (Johnson 2007; Valenti & Fischer 2005). This is a consequence of the technical requirements of Doppler exoplanetary detection, which demand that stars be cool enough to present an abundance of spectral lines, and rotate slowly enough that their absorption lines are not significantly broadened by rotation. Stars of lower mass (e.g. M dwarfs) are intrinsically faint in the optical, making the acquisition of high signal-to-noise spectra extremely expensive in telescope time (Endl et al. 2006). Main sequence stars of higher mass have few usable absorption lines (due to their high temperatures), and also tend to be fast rotators (v sin i > 50 km s−1; Galland et al. 2005) due to their youth. In addition, the shorter main-sequence lifetimes of higher-mass stars means that they will preferentially be observed at younger ages. Stars earlier than about F7 also have much shallower convection zones, and so do not experience the magnetic braking which slows the rotation of later-type (lower-mass) stars. As a result, only the most massive planets can be detected orbiting A and F dwarfs. It is only recently that a significant number of planetary systems have been discovered orbiting intermediate-mass stars (M∗ > 1.3M⊙). These stars have proven to be a fertile hunting ground for interesting planetary systems, such as the 4:3 mean-motion resonant planets orbiting HD 200964 (Johnson et al. 2011). Now, some headway is beginning to be made in addressing the crucial question of how planet formation depends on stellar mass (Bowler et al. 2010; Johnson et al. 2010a; Sato et al. 2010). A number of surveys are seeking to expand our knowledge of planetary systems orbiting stars more massive than the Sun, e.g. Setiawan et al. (2003); Hatzes et al. (2005); Sato et al. (2005); Johnson et al. (2006b); Dollinger et al. (2007); Niedzielski et al. (2009). These surveys are exploiting the advantage wrought by stellar evolution: as stars evolve off the main sequence into subgiants and giants, their atmospheres expand and cool, making precision Doppler velocity measurements possible due to an abundance of narrow spectral lines. The well-known planet-metallicity correlation (Gonzalez 1999; Fischer & Valenti 2005), whereby main-sequence stars with higher metal content are more likely to host planets, has come under some scrutiny. Analysis of the metallicities of planet-hosting giant stars by Schuler et al. (2005) showed that the giant star planet hosts were significantly more metal- poor than their main-sequence counterparts. Pasquini et al. (2007) have also argued that the planet-metallicity correlation does not apply for evolved stars. They propose that this is evi- -- 3 -- dence for a "pollution" scenario, in which main-sequence stars hosting planets appear metal- rich because they have accreted material from the protoplanetary disk (Murray & Chaboyer 2002). When a star evolves off the main sequence, the convective zone increases in size by about a factor of 35 (Pasquini et al. 2007). If the high metallicities observed in planet hosts are due to pollution, this expansion of the convective zone will significantly dilute the extent of that pollution, and the subgiant's photosphere would return to its "birth" metal- licity. Hence, one would not expect a significant correlation between metallicity and planet frequency for subgiants. However, the importance of planet pollution was downplayed even by the authors who proposed it, as they felt it should play only a minor role in shaping the planet-metallicity cor- relation seen in dwarf stars. Further, Valenti & Fischer (2008), Johnson et al. 2010a, Takeda et al. (2007) and others found no evidence of a decreasing planet-metallicity correlation among F dwarfs, subgiants or K giants. The sample of K giants studied by Pasquini et al. (2007) had a limited metallicity range, with [Fe/H]< +0.2. Examination of the form of the planet-metallicity correlation of Fischer & Valenti (2005) and Johnson et al. (2010a) shows that, for small numbers of stars, the correlation over this metallicity range would look ap- proximately flat. One way to test this is to search for planets around a sample of evolved stars that are unambiguously metal rich. Sandage et al. (2003) pointed out that stars on the red-edge of the subgiant branch represent such a population. Here, we introduce a new Southern Hemi- sphere survey, the "Pan-Pacific Planet Search," which uses the 3.9-metre Anglo-Australian Telescope (AAT) to search for planets among these evolved, metal-rich stars to test for evidence of planet pollution. In this paper, we present the first result from this new survey: the detection of a 2.6 MJup planet orbiting the nearby evolved star HD 47205. In Section 2, we introduce the Pan- Pacific Planet Search and give a complete target list for the survey. Section 3 describes the observational data and gives the stellar parameters for 7 CMa. In Section 4, we detail the orbit-fitting process and present the planetary parameters. Finally, in Section 5 we discuss the further implications of this discovery. 2. The Pan-Pacific Planet Search 2.1. Survey Strategy and Target Selection The Pan-Pacific Planet Search (PPPS) originated as a Southern hemisphere extension of the established Lick & Keck Observatory survey for planets orbiting Northern "retired A -- 4 -- stars" (Johnson et al. 2006b, 2007, 2010a). This program is using the 3.9m Anglo-Australian Telescope (AAT) to observe a metal-rich sample of Southern Hemisphere subgiants. We have selected 170 Southern stars with the following criteria: 1.0 < (B −V ) < 1.2, 1.8 < MV < 3.0, and V < 8.0. By requiring (B − V ) > 1, we extend the red limit of the Johnson et al. (2006b) survey to the colours that stellar models indicate will be dominated by metal-rich subgiants (Girardi et al. 2002). This aims to deliver improved planetary detection statistics at [Fe/H]>0.0. In light of the observed positive correlation between stellar metallicity and planet occurrence, this should also deliver a roughly equivalent number of planetary detec- tions to that obtained at Lick and Keck, though for metal-rich hosts. At the same time, by requiring MV > 1.8, we exclude giant-branch stars, as these have significant intrinsic velocity noise ("jitter") due to random convective motion and pulsations (Saar et al. 1998; Wright 2005) -- typically about 20 m s−1 (Hekker et al. 2006). Our target list includes about 30 stars from the Lick survey; this overlap will serve as a check on the systematics between the two telescopes. Together, the three telescopes are observing more than 600 stars. The complete PPPS target list is given in Table 1. Observing time is scheduled such that each target should receive 4 -- 6 observations per year. This strategy would appear to reduce the probability of detecting shorter-period plan- ets (P <∼ 50 days), which require more densely-sampled observations in continuous blocks of time (O'Toole et al. 2009a,b; Vogt et al. 2010). However, we note that the same scheduling has been used in the Lick and Keck survey, which has detected the P = 6.5 day planet or- biting HD 102956 (Johnson et al. 2010c). By employing this strategy, we are able to target more stars with a fixed amount of observing time, which should increase the probability of detecting the types of planets which are known to orbit nearly 20% of these types of stars (Johnson et al. 2010a). 2.2. Observations and Data Reduction PPPS Doppler measurements are made with the UCLES echelle spectrograph (Diego et al. 1991) at the 3.9-metre Anglo-Australian Telescope (AAT). UCLES achieves a resolution of 45,000 with a 1-arcsecond slit. An iodine absorption cell provides wavelength calibration from 5000 to 6200 A. The spectrograph point-spread function and wavelength calibration are derived from the iodine absorption lines embedded on every pixel of the spectrum by the cell (Valenti et al. 1995; Butler et al. 1996). The result is a precision Doppler velocity estimate for each epoch, along with an internal uncertainty estimate, which includes the effects of photon-counting uncertainties, residual errors in the spectrograph PSF model, and variation in the underlying spectrum between the iodine-free template, and epoch spectra -- 5 -- observed through the iodine cell. The photon-weighted mid-time of each exposure is deter- mined by an exposure meter. All velocities are measured relative to the zero-point defined by the template observation. Velocities are obtained using the Austral code as first dis- cussed in Endl et al. (2000). Austral is a proven Doppler code which has been used by the McDonald Observatory planet search programs for nearly 10 years (e.g. Endl et al. 2004, 2006; Wittenmyer et al. 2009). Observations for the PPPS began at the Anglo-Australian Telescope in 2009 February. Since its inception, the program has received 20 nights per year, of which approximately 50% have resulted in usable data. We aim for a signal-to-noise (S/N) of 100 at 5500 A per spectral pixel each epoch, resulting in exposure times ranging from 100 s up to a maximum of 20 minutes. We have observed 7 CMa on 21 epochs, and an iodine-free template observation was obtained on 2010 Jan 30. Since 7 CMa is an extremely bright star, exposure times ranged from 100 to 500 s, with a resulting S/N of ∼200-300 per pixel each epoch. The data span a total of 917 days, and have a mean internal velocity uncertainty of 6.5 m s−1. 3. Stellar Parameters of 7 CMa 7 CMa (=HD 47205, HIP 31592) is one of the brightest stars in the PPPS survey (V = 3.95). In addition, it is accessible from most sites in both hemispheres (RA: 06 36 41.038, Dec: -19 15 21.17), and so it has been well-studied. Table 2 summarises the physical parameters of this star. We have used our iodine-free template spectrum to derive spectroscopic stellar parameters, using methods described fully in Wang et al. (2011). In brief, 7 CMa is an evolved, somewhat metal-rich ([Fe/H]∼ 0.2), intermediate-mass star (1.52 M⊙) with a low level of activity. Hipparcos observations indicate that it is photometrically stable, with a median Hipparcos magnitude of 4.1200±0.0004 (van Leeuwen 2007; Perryman et al. 1997). 4. Orbit Fitting and Planetary Parameters The AAT data show a root-mean-square (RMS) scatter of 29.5 m s−1 about the mean velocity. Visual investigation of the data after two years of observation revealed a clear sinusoidal trend. Due to the relative paucity of data points (N = 21) compared to typical radial-velocity planet detections (N >∼ 40), the traditional periodogram approach does not produce reliable estimates of statistical significance. Rather, since the periodic signal is read- ily apparent by eye, we used a genetic algorithm (Charbonneau 1995) to determine Keplerian -- 6 -- orbital parameters. Those parameters were then used as initial inputs for a standard least- squares fitting routine. Our previous experience with genetic algorithms (Cochran et al. 2007; Tinney et al. 2011) has shown that the solution "evolves" quite rapidly toward a sharp χ2 minimum when brought to bear on data containing a real and coherent Keplerian signal. Indeed, with an allowed period range of 600-800 days, the algorithm converged on a solution with a period of 769 days and a small eccentricity e = 0.23. We then used the GaussFit code (Jefferys et al. 1987) to obtain a Keplerian model fit for the planet. For the final orbit fitting, we added 5 m s−1 of jitter in quadrature to the internal uncertainties of the data shown in Table 4. The jitter estimate of 5 m s−1 is derived from Johnson et al. (2010a). In that work, 382 velocity measurements from 72 stable stars in the Lick and Keck survey of "retired A stars" were used to make an empirical jitter estimate which was then applied to the seven planet-host stars described therein. It would be ideal to estimate jitter for PPPS stars using the same methodology, but at this time, we have insufficient data on radial-velocity stable stars to make a statistically meaningful estimate. This is due to the short time baseline (2.5 yr) and limited available data on a smaller number of stars in the PPPS as compared to the well-established Lick & Keck survey. However, we consider the 5m s−1 jitter estimate to be a reasonable approximation for PPPS targets due to the similarity in physical prop- erties to the Johnson et al. (2010a) sample. We also note that there is substantial ( >∼ 50%) uncertainty in the estimation of radial-velocity jitter (Wright 2005). Using a stellar mass of 1.52±0.30 M⊙, we estimate the minimum mass m sin i to be 2.6±0.6 MJup. The fit is shown in Figure 1 and the planetary parameters are given in Table 3. The residuals of the fit show no evidence for additional signals (Figure 3). As a further test of the veracity of the planet fit, we used the "scrambled velocity" approach of Marcy et al. (2005). This technique serves to test the null hypothesis that the observed velocity variation is attributable to noise. For this test, we scramble the velocities amongst the observation epochs, creating 5000 shuffled data sets. Then, we perform the same least-squares Keplerian orbit-fitting on the shuffled data and log the resulting best-fit χ2. The results of these trials are shown in Figure 2 -- not one of the scrambled data sets achieved a better χ2 than the planet fit to our original data. We thus conclude that there is a less than 0.02% probability that the detected signal arose by chance from noise. 5. Discussion 5.1. Testing the Planet Hypothesis Since many K giants have intrinsic radial-velocity variations with periods of hundreds of days (Hekker et al. 2008), it is prudent for us to further examine the planet hypothesis for -- 7 -- 7 CMa to ensure that the observed velocity variations are not associated with known activity and rotational cycles. The first and simplest test is to combine the available estimates of the star's radius and v sin i minimum rotational velocity to obtain a maximum rotation period. Using the values for these quantities given in Table 2, this yields a maximum Prot = 116 days (for v sin i=1.0 km s−1, Massarotti et al. 2008). Unfortunately, neither estimate of v sin i has an uncertainty, but if we apply a typical uncertainty of 1 km s−1, then maximum rotation periods shorter than 776 days fall within the 1σ range. For a spotted star, the rotation period can be deduced from photometry. A peri- odogram of the Hipparcos photometry (after removing one outlier which was more than 1 magnitude discrepant) is shown in Figure 4. Two peaks are evident at periods of 12.2 and 103.7 days. We estimate the false-alarm probability using the bootstrap randomization method (Kurster et al. 1997). The bootstrap method randomly shuffles the observations while keeping the times of observation fixed. The periodogram of this shuffled data set is then computed and its highest peak recorded. From 10,000 such realizations, we find a false- alarm probability of 2.5% for the peak at 12.2 days, and 3.2% for that at 103.7 days. At the 763-day period of the candidate planet, the bootstrap false-alarm probability is 98.7%. The amplitude of the photometric variations for either of the two marginally significant periods is 20±6 × 10−4 mag. In any case, if these small photometric variations are due to the star's rotation, their periodicities are clearly well-separated from that of the candidate planet. One can argue that the absence of significant variations in the Hipparcos photometry on the 763-day period is not a complete refutation of the starspot hypothesis. Stars have activity cycles, and so there is the possibility that the activity of 7 CMa was at a minimum during the Hipparcos observations (20 years before the radial-velocity observations), but is now at a maximum which, if the rotation period were as long as ∼763 days, could mimic the signal of an orbiting planet. Line bisector analysis is a fairly common technique used by some planet-search programs (e.g. HARPS) to make sure a signal is not due to stellar activity. Such analysis has the advantage of being contemporaneous with the velocity measurements. We note that for the radial-velocity programs using AAT/UCLES, spectra are of relatively low resolution (R = 45, 000) and nearly all of the usable spectral range is superimposed with iodine lines. We have computed the bisector velocity spans for 8 strong unblended lines redward of the iodine region. The results are shown in Figure 5; each point represents the mean bisector velocity span of 8 lines, and its uncertainty is the standard deviation about the mean value. While the uncertainties are large, it is evident from Figure 5 that the bisector velocity spans are uncorrelated with the radial velocities. Furthermore, the right panel of Figure 5 shows a periodogram of the bisector velocity spans, which also indicates no periodicity near the 763-day period of the planet. These independent lines of evidence thus lead us to conclude that the radial-velocity variations observed in 7 CMa are attributible not to an intrinsic stellar process, but to an orbiting giant planet. -- 8 -- 5.2. Conclusions Using the Anglo-Australian Telescope, we have begun a Southern hemisphere search for planets orbiting evolved, intermediate-mass stars. Our Pan-Pacific Planet Search team members are based in Australia, China, and the US; we now report the first planet detection from our ongoing survey. There is an emerging trend that planets orbiting intermediate-mass stars tend to have higher masses and longer periods that planets orbiting solar-mass stars (Bowler et al. 2010; Johnson et al. 2010b). With a period of 2.1 years and a minimum mass of 2.6 MJup, 7 CMa b is quite similar to other planets known to orbit intermediate-mass stars. Figure 6 shows the mass and period distribution of all radial-velocity detected planets known to orbit stars with M∗ > 1.3M⊙, with 7 CMa b plotted as a large filled triangle. Given the abundance of planets with P >∼ 2 yr known to orbit these types of stars, we anticipate that 7 CMa b is the first of many planet detections to come from the PPPS. We gratefully acknowledge the UK and Australian government support of the Anglo- Australian Telescope through their PPARC, STFC and DIISR funding; STFC grant PP/C000552/1, ARC Grant DP0774000 and travel support from the Australian Astronomical Observatory. RW is grateful to the Chinese Academy of Sciences for the support of his stay in Beijing. RW is supported by a UNSW Vice-Chancellor's Fellowship. We thank the ATAC for the generous allocation of telescope time which facilitated this detection. This research has made use of NASA's Astrophysics Data System (ADS), and the SIMBAD database, operated at CDS, Strasbourg, France. This research has made use of the Exoplanet Orbit Database and the Exoplanet Data Explorer at exoplanets.org. Bowler, B. P., et al. 2010, ApJ, 709, 396 REFERENCES Butler, R. P., Marcy, G. W., Williams, E., McCarthy, C., Dosanjh, P., & Vogt, S. S. 1996, PASP, 108, 500 Charbonneau, P. 1995, ApJS, 101, 309 Cochran, W. D., Endl, M., Wittenmyer, R. A., & Bean, J. L. 2007, ApJ, 665, 1407 da Silva, L., et al. 2006, A&A, 458, 609 -- 9 -- Diego, F., Charalambous, A., Fish, A. C., & Walker, D. D. 1990, Proc. Soc. Photo-Opt. Instr. Eng., 1235, 562 Dollinger, M. P., Hatzes, A. P., Pasquini, L., et al. 2007, A&A, 472, 649 Endl, M., Cochran, W. D., Kurster, M., Paulson, D. B., Wittenmyer, R. A., MacQueen, P. J., & Tull, R. G. 2006, ApJ, 649, 436 Endl, M., Hatzes, A. P., Cochran, W. D., McArthur, B., Allende Prieto, C., Paulson, D. B., Guenther, E., & Bedalov, A. 2004, ApJ, 611, 1121 Endl, M., Kurster, M., & Els, S. 2000, A&A, 362, 585 Fischer, D. A., & Valenti, J. 2005, ApJ, 622, 1102 Galland, F., Lagrange, A.-M., Udry, S., Chelli, A., Pepe, F., Queloz, D., Beuzit, J.-L., & Mayor, M. 2005, A&A, 443, 337 Girardi, L., Bertelli, G., Bressan, A., Chiosi, C., Groenewegen, M. A. T., Marigo, P., Salas- nich, B., & Weiss, A. 2002, A&A, 391, 195 Gonzalez, G. 1999, MNRAS, 308, 447 Gray, R. O., Corbally, C. J., Garrison, R. F., McFadden, M. T., Bubar, E. J., McGahee, C. E., O'Donoghue, A. A., & Knox, E. R. 2006, AJ, 132, 161 Hatzes, A. P., Guenther, E. W., Endl, M., Cochran, W. D., Dollinger, M. P., & Bedalov, A. 2005, A&A, 437, 743 Hekker, S., & Mel´endez, J. 2007, A&A, 475, 1003 Hekker, S., Reffert, S., Quirrenbach, A., Mitchell, D. S., Fischer, D. A., Marcy, G. W., & Butler, R. P. 2006, A&A, 454, 943 Hekker, S., Snellen, I. A. G., Aerts, C., Quirrenbach, A., Reffert, S., & Mitchell, D. S. 2008, A&A, 480, 215 Howard, A. W., et al. 2010, Science, 330, 653 Jefferys, W. H., Fitzpatrick, M. J., & McArthur, B. E. 1987, Celestial Mechanics, 41, 39 Johnson, J. A., et al. 2011, AJ, 141, 16 -- 10 -- Johnson, J. A., Aller, K. M., Howard, A. W., & Crepp, J. R. 2010a, PASP, 122, 905 Johnson, J. A., et al. 2010c, ApJ, 721, L153 Johnson, J. A., Howard, A. W., Bowler, B. P., Henry, G. W., Marcy, G. W., Wright, J. T., Fischer, D. A., & Isaacson, H. 2010b, PASP, 122, 701 Johnson, J. A., et al. 2006, ApJ, 647, 600 Johnson, J. A., Marcy, G. W., Fischer, D. A., Henry, G. W., Wright, J. T., Isaacson, H., & McCarthy, C. 2006, ApJ, 652, 1724 Johnson, J. A. 2007, arXiv:0710.2904 Johnson, J. A., et al. 2007, ApJ, 665, 785 Kurster, M., Schmitt, J. H. M. M., Cutispoto, G., & Dennerl, K. 1997, A&A, 320, 831 Marcy, G. W., Butler, R. P., Vogt, S. S., Fischer, D. A., Henry, G. W., Laughlin, G., Wright, J. T., & Johnson, J. A. 2005, ApJ, 619, 570 Massarotti, A., Latham, D. W., Stefanik, R. P., & Fogel, J. 2008, AJ, 135, 209 Murray, N., & Chaboyer, B. 2002, ApJ, 566, 442 Niedzielski, A., Go´zdziewski, K., Wolszczan, A., et al. 2009, ApJ, 693, 276 O'Toole, S., et al. 2009, ApJ, 697, 1263 O'Toole, S., et al. 2009, ApJ, 701, 1732 Pasquini, L., Dollinger, M. P., Weiss, A., Girardi, L., Chavero, C., Hatzes, A. P., da Silva, L., & Setiawan, J. 2007, A&A, 473, 979 Perryman, M. A. C., et al. 1997, A&A, 323, L49 Saar, S. H., Butler, R. P., & Marcy, G. W. 1998, ApJ, 498, L153 Sandage, A., Lubin, L. M., & VandenBerg, D. A. 2003, PASP, 115, 1187 Sato, B., Kambe, E., Takeda, Y., Izumiura, H., Masuda, S., & Ando, H. 2005, PASJ, 57, 97 Sato, B., et al. 2010, PASJ, 62, 1063 Schuler, S. C., Kim, J. H., Tinker, M. C., Jr., et al. 2005, ApJ, 632, L131 -- 11 -- Setiawan, J., Pasquini, L., da Silva, L., von der Luhe, O., & Hatzes, A. 2003, A&A, 397, 1151 Takeda, G., Ford, E. B., Sills, A., Rasio, F. A., Fischer, D. A., & Valenti, J. A. 2007, ApJS, 168, 297 Tinney, C. G., Wittenmyer, R. A., Butler, R. P., Jones, H. R. A., O'Toole, S. J., Bailey, J. A., Carter, B. D., & Horner, J. 2011, ApJ, 732, 31 Valenti, J. A., & Fischer, D. A. 2005, ApJS, 159, 141 Valenti, J. A., & Fischer, D. A. 2008, Physica Scripta Volume T, 130, 014003 Valenti, J. A., Butler, R. P. & Marcy, G. W. 1995, PASP, 107, 966. van Leeuwen, F. 2007, A&A, 474, 653 Vogt, S. S., et al. 2010, ApJ, 708, 1366 Wang, L., Liu, Y., Zhao, G., & Sato, B. 2011, PASJ, accepted. Wittenmyer, R. A., Tinney, C. G., Butler, R. P., O'Toole, S. J., Jones, H. R. A., Carter, B. D., Bailey, J., & Horner, J. 2011, ApJ, 738, 81 Wright, J. T. 2005, PASP, 117, 657 Wright, J. T., et al. 2011, PASP, 123, 412 This preprint was prepared with the AAS LATEX macros v5.2. -- 12 -- Table 1. Pan-Pacific Planet Search Target List Star RA Dec V 224910 749 1817 2643 4145 5676 5877 6037 7931 9218 9925 10731 11343 11653 12974 13471 13652 14805 14791 15414 19810 20035 20924 24316 25069 28901 29399 31860 34851 33844 37763 39281 40409 43429 46262 47141 47205 51268 58540 59663 67644 72467 76321 76437 76920 00 01 44.93 00 11 38.06 00 22 21.22 00 29 54.99 00 43 50.09 00 58 12.43 00 59 19.25 01 01 38.60 01 18 32.20 01 30 13.84 01 35 43.11 01 43 28.25 01 50 06.22 01 53 00.51 02 06 56.07 02 10 54.48 02 12 34.49 02 20 43.08 02 22 07.00 02 26 12.48 03 10 51.47 03 11 57.63 03 21 58.79 03 51 32.65 03 58 52.42 04 32 06.76 04 33 34.10 04 57 46.42 05 12 02.21 05 12 36.08 05 31 52.66 05 48 34.16 05 54 05.90 06 15 17.71 06 20 53.87 06 36 05.42 06 36 41.00 06 53 33.56 07 22 57.03 07 25 09.11 08 06 20.28 08 32 01.89 08 54 57.73 08 55 01.65 08 55 16.78 -16 31 54.2 -49 39 21.2 -50 59 33.4 -32 16 23.8 -12 00 40.7 -25 52 35.8 -58 24 17.2 -16 15 55.3 -28 43 58.7 -28 51 55.5 -53 11 59.8 -56 14 04.1 -54 27 53.5 -52 41 30.1 -01 49 25.2 -32 03 42.9 -26 19 20.7 -62 32 45.8 -36 06 23.8 -62 55 05.2 -11 07 29.4 -39 21 57.1 -15 27 31.4 -17 09 58.8 -05 28 10.3 -28 48 22.0 -62 49 25.1 -34 53 32.3 -75 21 37.6 -14 57 04.3 -76 20 30.0 -53 40 34.1 -63 05 27.7 -18 28 37.2 -79 04 00.5 -24 51 57.8 -19 15 20.6 -54 52 59.3 -55 34 38.8 -70 24 13.9 -54 02 45.9 -29 22 15.1 -15 46 45.9 -34 08 35.0 -67 15 55.9 7.83 7.91 6.68 8.15 6.01 7.89 7.78 6.47 7.89 7.96 7.82 7.97 7.88 7.91 7.49 7.65 7.92 7.68 7.87 7.92 7.22 6.98 7.26 7.71 5.85 7.42 5.79 7.60 7.85 7.29 5.18 7.85 4.65 5.99 7.31 7.45 3.95 7.97 6.89 7.75 7.97 7.59 7.10 7.15 7.83 -- 13 -- Table 1 -- Continued Star RA Dec V 80275 81410 84070 85128 85035 87089 86950 HIP50638 94386 98516 98579 100939 104358 104704 104819 105096 108991 109866 110238 114899 115066 115202 121056 121156 121930 124087 125774 126105 130048 131182 132396 133166 133670 134443 134692 135760 136295 136905 137115 137164 136135 138061 138716 138973 142132 09 17 46.62 09 24 49.04 09 41 13.31 09 43 01.74 09 48 47.03 10 00 34.00 10 01 37.61 10 20 33.31 10 53 32.86 11 19 47.64 11 20 19.05 11 36 48.17 12 01 00.72 12 03 22.27 12 04 11.05 12 06 01.36 12 31 38.47 12 38 49.98 12 40 59.85 13 14 26.28 13 15 04.35 13 15 58.58 13 53 52.27 13 54 16.75 13 59 46.00 14 11 45.85 14 21 49.23 14 24 00.33 14 46 13.51 14 52 21.07 15 00 00.35 15 04 36.20 15 06 27.10 15 11 31.92 15 14 59.90 15 18 17.43 15 20 41.44 15 23 26.06 15 24 57.58 15 27 45.75 15 27 47.09 15 29 59.95 15 34 10.52 15 36 08.23 15 54 24.55 -35 41 23.8 -23 49 34.4 -46 22 55.1 -79 35 30.2 -19 18 48.6 -61 45 31.3 -17 19 58.8 -23 38 25.4 -15 26 44.5 -28 11 19.6 -28 19 56.1 -37 02 20.5 -26 28 47.2 -55 19 17.0 -22 22 15.6 -54 15 28.1 -30 58 54.8 -62 01 54.1 -31 44 15.9 -54 57 43.6 -30 10 53.0 -19 56 34.2 -35 18 51.1 -28 34 09.9 -50 13 40.5 -19 01 03.2 -10 40 00.5 -19 48 02.8 -07 47 48.9 -11 28 22.7 -36 01 49.7 -43 53 50.4 -22 01 54.1 -45 16 44.7 -66 53 36.5 -41 25 13.0 -25 59 24.0 -06 36 36.7 -22 02 37.1 -63 01 14.1 -79 18 22.4 -12 46 35.6 -10 03 50.3 -21 44 46.6 -41 10 22.3 7.70 7.35 7.88 7.30 7.02 7.93 7.47 7.54 6.34 7.06 6.68 7.94 7.76 7.49 7.93 7.03 6.73 7.76 7.70 7.99 7.83 5.21 6.17 6.05 7.58 7.74 7.99 7.32 7.14 7.95 6.94 7.92 6.13 7.38 7.91 7.05 7.11 7.29 7.65 7.44 7.61 7.78 4.61 7.72 7.70 -- 14 -- Table 1 -- Continued Star RA Dec V 142384 143561 144073 145428 148760 153438 153937 154250 155233 154556 159743 162030 166309 166476 170707 170286 173902 175905 176002 175304 177897 176794 181342 181809 188981 191067 196676 199809 200073 201931 204073 204057 204203 205577 205972 205478 208431 208791 214573 215005 216640 216643 218266 219553 222076 15 55 56.46 16 02 45.22 16 05 01.36 16 11 51.34 16 31 22.87 17 00 29.72 17 06 11.92 17 06 40.99 17 11 04.37 17 12 19.85 17 37 01.69 17 49 57.49 18 11 15.84 18 14 20.10 18 33 33.37 18 35 02.98 18 49 17.14 18 57 35.98 19 00 01.36 19 03 29.12 19 08 42.79 19 10 44.78 19 21 04.26 19 22 40.30 19 58 56.37 20 08 01.75 20 39 05.86 21 00 19.00 21 02 27.05 21 14 16.67 21 26 25.07 21 26 27.06 21 27 29.20 21 36 43.65 21 39 15.19 21 41 28.47 21 56 47.43 21 59 01.31 22 40 07.11 22 42 45.92 22 54 45.60 22 55 11.14 23 07 11.97 23 16 49.69 23 38 08.10 -40 47 11.0 -42 30 25.5 -26 56 52.0 -25 53 00.3 -26 32 15.2 -21 27 41.3 -60 25 14.8 -48 00 43.5 -20 39 15.2 -70 43 15.2 -18 59 30.9 -24 12 25.1 -29 38 22.4 -58 42 20.9 -50 12 41.5 -71 19 26.1 -34 44 56.0 -00 31 34.6 -43 20 49.7 -76 06 54.8 -45 04 34.3 -76 24 15.7 -23 37 10.2 -20 38 33.6 -30 32 17.7 -00 40 40.9 -04 55 46.2 -27 20 35.9 -38 31 50.0 -45 46 57.2 -12 05 42.0 -15 14 42.9 -20 12 45.2 -21 30 09.8 -13 53 41.0 -77 23 22.1 -28 49 03.3 -11 17 03.4 -49 35 53.2 -37 20 43.7 -16 16 18.3 -46 40 43.9 -45 50 33.2 -21 12 10.7 -70 54 12.3 7.41 7.97 7.60 7.73 6.07 7.35 7.43 7.96 6.81 6.21 7.45 7.02 7.61 7.81 7.75 7.72 6.59 7.66 7.92 7.75 7.74 6.94 7.55 6.72 6.27 5.97 6.46 7.93 5.93 6.89 6.70 7.97 7.84 7.93 7.25 3.73 7.91 7.79 7.37 7.93 5.53 7.53 7.92 7.25 7.47 -- 15 -- Table 1 -- Continued Star RA Dec V 222768 223301 223860 23 43 32.71 23 48 28.19 23 53 13.59 -22 54 07.7 -11 30 31.4 -11 00 52.6 7.81 7.60 7.66 Table 2. Stellar Parameters for 7 CMa Parameter Value Reference (m s−1) Spec. Type MV B − V Mass (M⊙) Radius (R⊙) Luminosity (L⊙) Distance (pc) V sin i (km s−1) a SMW [F e/H] vmicro (km s−1) Tef f (K) log g K1 III 2.46±0.03 1.037±0.041 1.52±0.30 1.32±0.12 2.3±0.1 11.3±0.3 19.75±0.09 1.15 1.0 0.132 0.21±0.10 0.18±0.1 0.21 1.32±0.10 1.30 1.45 1.0 4792±100 4744±70 4830 4799 3.25±0.10 3.11±0.07 3.40 3.05 Gray et al. (2006) da Silva et al. (2006) van Leeuwen (2007) This work da Silva et al. (2006) This work This work van Leeuwen (2007) Hekker & Mel´endez (2007) Massarotti et al. (2008) Gray et al. (2006) This work da Silva et al. (2006) Hekker & Mel´endez (2007) This work da Silva et al. (2006) Hekker & Mel´endez (2007) Gray et al. (2006) This work da Silva et al. (2006) Hekker & Mel´endez (2007) Gray et al. (2006) This work da Silva et al. (2006) Hekker & Mel´endez (2007) Gray et al. (2006) aMount Wilson S-index -- 16 -- Table 3. 7 CMa Planetary Parameters Parameter Estimate Period (days) Eccentricity ω (degrees) K (m s−1) T0 (JD-2400000) M sin i (MJup) a (AU) RMS of fit (m s−1) N 763±17 0.14±0.06 12±41 44.9±4.0 55520±89 2.6±0.6 1.9±0.1 7.5 21 Table 4. AAT Radial Velocities for 7 CMa JD-2400000 Velocity (m s−1) Uncertainty (m s−1) 54866.09800 54866.10107 54866.94000 54867.91576 54869.08575 54871.03478 55140.18702 55140.19229 55227.06602 55317.85529 55317.85839 55317.86143 55495.13719 55525.22369 55526.21028 55581.09317 55601.00002 55670.87749 55706.84304 55783.30394 55783.31112 24.11 17.47 1.93 -5.81 12.59 5.22 -42.28 -36.90 -37.51 -4.99 -9.87 -3.89 51.81 50.62 31.35 38.45 14.37 -10.70 -12.95 -43.20 -39.91 5.73 5.86 5.81 6.09 5.72 6.13 5.56 6.07 5.82 5.40 5.33 5.31 11.84 5.68 5.66 5.83 6.01 13.99 5.68 5.85 6.72 -- 17 -- Fig. 1. -- Radial-velocity data and Keplerian orbit fit for 7 CMa. The RMS scatter about the fit is 7.5 m s−1, consistent with the mean uncertainty of 8.3 m s−1 (including 5 m s−1 of jitter added in quadrature). -- 18 -- Fig. 2. -- Results of 4999 trials in which the velocity data for 7 CMa were scrambled amongst the observation epochs. The reduced χ2 of the original data is shown as a dashed vertical line. None of the scrambled data sets resulted in a better χ2, indicating a less than 0.02% chance that the observed variations are due to noise. -- 19 -- Fig. 3. -- Periodogram of the residuals to the Keplerian orbit fit; no further signals are evident. -- 20 -- Fig. 4. -- Periodogram of Hipparcos photometry for 7 CMa (N = 168). The two highest peaks are at periods of 12.2 and 103.7 days. The vertical dashed line indicates the 763-day period of the planet; there is no significant periodicity in the photometry near this period. -- 21 -- Fig. 5. -- Left panel: Bisector velocity span versus radial velocity for the 21 observations for 7 CMa. A correlation would indicate that the observed radial-velocity variations were due to an intrinsic stellar process rather than an orbiting planet; no correlation is evident. Right panel: Periodogram of the bisector velocity spans. The vertical dashed line indicates the 763-day period of the planet; there is no significant periodicity near this period. -- 22 -- Fig. 6. -- Mass-period plot for 83 radial-velocity detected planets orbiting stars with M∗ > 1.3 M⊙; planet data are from the Exoplanet Orbit Database (Wright et al. 2011). 7 CMa b is marked as a large filled triangle. Its parameters are consistent with other planets orbiting intermediate-mass stars.
1805.02261	1	1805	2018-05-06T18:34:02	First direct detection of a polarized companion outside of a resolved circumbinary disk around CS Cha	[ "astro-ph.EP" ]	In the present study we aim to investigate the circumstellar environment of the spectroscopic binary T Tauri star CS Cha. From unresolved mid- to far-infrared photometry it is predicted that CS Cha hosts a disk with a large cavity. In addition, SED modeling suggests significant dust settling, pointing towards an evolved disk that may show signs of ongoing or completed planet formation. We observed CS Cha with the high contrast imager VLT/SPHERE in polarimetric differential imaging mode to resolve the circumbinary disk in near infrared scattered light. These observations were followed-up by VLT/NACO L-band observations and complemented by archival VLT/NACO K-band and HST/WFPC2 I-band data. We resolve the compact circumbinary disk around CS Cha for the first time in scattered light. We find a smooth, low inclination disk with an outer radius of $\sim$55 au (at 165 pc). We do not detect the inner cavity but find an upper limit for the cavity size of $\sim$15 au. Furthermore, we find a faint co-moving companion with a projected separation of 210 au from the central binary outside of the circumbinary disk. The companion is detected in polarized light and shows an extreme degree of polarization (13.7$\pm$0.4 \% in J-band). The companion's J- and H-band magnitudes are compatible with masses of a few M$_\mathrm{Jup}$. However, K-, L- and I-band data draw this conclusion into question. We explore with radiative transfer modeling whether an unresolved circum-companion disk can be responsible for the high polarization and complex photometry. We find that the set of observations is best explained by a heavily extincted low mass ($\sim 20 \mathrm{M}_\mathrm{Jup}$) brown dwarf or high mass planet with an unresolved disk and dust envelope.	astro-ph.EP	astro-ph	Astronomy & Astrophysics manuscript no. CSCha-aa_revised May 8, 2018 c(cid:13)ESO 2018 First direct detection of a polarized companion outside of a resolved circumbinary disk around CS Cha(cid:63) C. Ginski1, 2, M. Benisty3, 4, R. G. van Holstein1, A. Juhász5, T. O. B. Schmidt6, G. Chauvin3, 4, J. de Boer1, M. Wilby1, C. F. Manara7, P. Delorme4, F. Ménard4, P. Pinilla8, T. Birnstiel9, M. Flock10, C. Keller1, M. Kenworthy1, J. Milli4, 11, J. Olofsson12, 13, L. Pérez14, F. Snik1, and N. Vogt12 8 1 0 2 y a M 6 . ] P E h p - o r t s a [ 1 v 1 6 2 2 0 . 5 0 8 1 : v i X r a 1 Leiden Observatory, Leiden University, P.O. Box 9513, 2300 RA Leiden, The Netherlands e-mail: [email protected] 2 Anton Pannekoek Institute for Astronomy, University of Amsterdam, Science Park 904, 1098 XH Amsterdam, The Netherlands 3 Unidad Mixta Internacional Franco-Chilena de Astronomia, CNRS/INSU UMI 3386 and Departamento de Astronomia, Universi- dad de Chile, Casilla 36-D, Santiago, Chile 4 Univ. Grenoble Alpes, CNRS, IPAG, F-38000 Grenoble, France 5 Institute of Astronomy, University of Cambridge, Madingley Road, Cambridge, CB3 0HA, UK 6 LESIA, Observatoire de Paris, PSL Research University, CNRS, Sorbonne Universités, UPMC Univ. Paris 06, Univ. Paris Diderot, Sorbonne Paris Cité, 5 place Jules Janssen, 92195 Meudon, France 7 Scientific Support Office, Directorate of Science, European Space Research and Technology Centre (ESA/ESTEC), Keplerlaan 1, 2201 AZ Noordwijk, The Netherlands 8 Department of Astronomy/Steward Observatory, The University of Arizona, 933 North Cherry Avenue, Tucson, AZ 85721, USA 9 University Observatory, Faculty of Physics, Ludwig-Maximilians-Universität München, Scheinerstr. 1, 81679 Munich, Germany 10 Max-Planck-Institut für Astronomie, Königstuhl 17, 69117, Heidelberg, Germany 11 European Southern Observatory (ESO), Alonso de Córdova 3107, Vitacura, Casilla 19001, Santiago, Chile 12 Instituto de Física y Astronomía, Facultad de Ciencias, Universidad de Valparaíso, Av. Gran Bretaña 1111, Valparaíso, Chile 13 Núcleo Milenio Formación Planetaria - NPF, Universidad de Valparaíso, Av. Gran Bretaña 1111, Valparaíso, Chile 14 Universidad de Chile, Departamento de Astronoma, Camino El Observatorio 1515, Las Condes, Santiago, Chile Received December, 2017; accepted April, 2018 ABSTRACT Aims. To understand planet formation it is necessary to study the birth environment of planetary systems. Resolved imaging of young planet forming disks allows us to study this environment in great detail and find signs of planet-disk interaction, as well as disk evolution. In the present study we aim to investigate the circumstellar environment of the spectroscopic binary T Tauri star CS Cha. From unresolved mid- to far-infrared photometry it is predicted that CS Cha hosts a disk with a large cavity. In addition, SED modeling suggests significant dust settling, pointing towards an evolved disk that may show signs of ongoing or completed planet formation. Methods. We observed CS Cha with the high contrast imager VLT/SPHERE in polarimetric differential imaging mode to resolve the circumbinary disk in near infrared scattered light. These observations were followed-up by VLT/NACO L-band observations and complemented by archival VLT/NACO K-band and HST/WFPC2 I-band data. Results. We resolve the compact circumbinary disk around CS Cha for the first time in scattered light. We find a smooth, low inclination disk with an outer radius of ∼55 au (at 165 pc). We do not detect the inner cavity but find an upper limit for the cavity size of ∼15 au. Furthermore, we find a faint co-moving companion with a projected separation of 210 au from the central binary outside of the circumbinary disk. The companion is detected in polarized light and shows an extreme degree of polarization (13.7±0.4 % in J-band). The companion's J- and H-band magnitudes are compatible with masses of a few MJup. However, K-, L- and I-band data draw this conclusion into question. We explore with radiative transfer modeling whether an unresolved circum-companion disk can be responsible for the high polarization and complex photometry. We find that the set of observations is best explained by a heavily extincted low mass (∼ 20 MJup) brown dwarf or high mass planet with an unresolved disk and dust envelope. Key words. Stars: individual: CS Cha – planetary systems: protoplanetary disks – planetary systems: planet-disk interactions – Techniques: polarimetric 1. Introduction In the past few years high contrast and high resolution observa- tions across a large wavelength range have revealed a variety of distinct features in planet forming disks. Multiple ringed systems were uncovered such as HL Tau (ALMA Partnership et al. 2015), (cid:63) Based on observations performed with VLT/SPHERE under pro- gram ID 098.C-0760(B) and 099.C-0891(B) and VLT/NACO under program ID 298.C-5054(B) and 076.C-0292(A) HD 97048 (Walsh et al. 2016, Ginski et al. 2016, van der Plas et al. 2017) or TW Hya (Andrews et al. 2016, van Boekel et al. 2017). Other systems like MWC 758 (Grady et al. 2013, Benisty et al. 2015), HD 100453 (Wagner et al. 2015, Benisty et al. 2017) or Elias 2-27 (Pérez et al. 2016) show huge spiral arms or vari- able shadows (HD 135344 B, Stolker et al. 2017b). It is still un- clear whether these features in general or in part are linked to ongoing planet formation, or rather to other processes within the disks. In addition to ever more detailed images of circumstellar Article number, page 1 of 19 A&A proofs: manuscript no. CSCha-aa_revised disks, a growing number of giant planets at wide orbital sep- arations (typically >100 au) are discovered (e.g. HD 106906 b, Bailey et al. 2014; HD 203030 b, Metchev & Hillenbrand 2006; CVSO 30 c, Schmidt et al. 2016). These objects are of partic- ular interest to understand planet formation mechanisms, since they are the youngest planets that we have discovered and we can study their atmospheres in great detail via resolved spec- troscopy. Yet these objects are also particularly puzzling, be- cause typical planet formation mechanisms such as core accre- tion should take much longer than 100 Myr at these distances (Pollack et al. 1996), while the typical dissipation timescale of gas rich disks is at least an order of magnitude shorter (Haisch et al. 2001). Clearly, detailed characterization of other, younger, systems is required to refine the current paradigm and to under- stand whether the observed disk structures are linked to planet formation. In this work we concentrate on a previously unre- solved disk around a nearby T Tau object. CS Cha is a young (2±2 Myr, Luhman et al. 2008) classical T Tauri object of spectral type K2Ve (Appenzeller 1977, Manara et al. 2014), located in the Chamaeleon I association at a dis- tance of 165±30 pc (combined estimate from Whittet et al. 1997, Bertout et al. 1999 following Schmidt et al. 2008).1 Guenther et al. (2007) found that CS Cha is likely a single lined spectro- scopic binary with a minimum mass of the secondary compo- nent of 0.1 M(cid:12) and a minimum orbital period of 2482 d (∼4 au semi-major axis, assuming a system mass of 1 M(cid:12)). In a later study by Nguyen et al. (2012) the binary nature of CS Cha was confirmed. They could fit the broadened spectral lines with two Gaussian profiles, making the system potentially a double lined spectroscopic binary. They found a flux ratio of the two compo- nents of 1.0±0.4. CS Cha is well known to feature a large infrared excess in its spectral energy distribution (SED) with a pronounced dip at 10 µm (see e.g. Gauvin & Strom 1992). The lack of emission at this wavelength regime was attributed to a large cavity by several studies (Gauvin & Strom 1992, Espaillat et al. 2007, Kim et al. 2009, Espaillat et al. 2011, Ribas et al. 2016), indicating that the system might be in a transition stage from a young gas-rich disk to a debris disk. The radius of the cavity has been a subject of intense modeling using unresolved photometric measurements. Espaillat et al. (2007, 2011) find rather large cavity radii between 38 au and 43 au, while a more recent study by Ribas et al. (2016) based on Herschel data estimates a smaller radius of 18+6−5au. The most likely explanation is that the disk cavity is caused entirely by the stellar binary companion, since the cavity size is within a factor of a few of the binary semi-major axis. ALMA band 3 observations by Dunham et al. (2016) did not resolve the disk with a beam size of 2.7×1.9 arcsec, limiting the outer extent of the disk to radii smaller than 169 au for the population of mm- sized dust grains. Radiative transfer modeling of the unresolved photometry by Es- paillat et al. (2007) suggested that significant dust settling and large dust grains (5 µm) are needed to fit the SED in the far infrared and mm wavelength ranges. This hints at an advanced stage of dust evolution. Pascucci et al. (2014) note that they re- solve circumstellar structure around CS Cha with 3.3 cm ATCA observations outside of 30 arcsec. Since it can be excluded that this emission stems from the disk itself, they conclude that it is 1 We note that in a recent study by Voirin et al. (2017), the distance to the Cha I cloud was estimated to be slightly larger at 179 pc. This is well covered by our uncertainties and we prefer to use the smaller dis- tance for better comparability with previous studies until a direct dis- tance measurement for CS Cha by Gaia becomes available. Article number, page 2 of 19 likely a jet, which is launched from the disk at a position angle of ∼ 162◦. We used the SPHERE (Spectro-Polarimetric High-contrast Ex- oplanet REsearch, Beuzit et al. 2008) extreme adaptive optics imager to study the circumstellar environment of CS Cha in po- larized near infrared light. Our goals were to resolve the disk cavity for the first time and to study potential features of dust evolution or planet disk interaction such as rings/gaps and spiral arms. In addition to our SPHERE observations we used archival high contrast data to strengthen our conclusions. 2. Observations and data reduction 2.1. The initial SPHERE polarimetric observations CS Cha was first observed with SPHERE/IRDIS (Infra-Red Dual Imaging and Spectrograph, Dohlen et al. 2008) in Differential Polarization Imaging mode (DPI, Langlois et al. 2014) in J-band on February 17th 2017 as part of our ongoing program to un- derstand dust evolution in transition disks via the distribution of small dust particles. Conditions during the night were excellent with clear sky and an average seeing in the optical of 0.6 arcsec and a coherence time of ∼5 ms. The (unresolved) central binary was placed behind a corona- graph with a diameter of 185 mas (Martinez et al. 2009, Carbillet et al. 2011). We used an integration time of 96 s in individual ex- posures and one exposure per half wave plate (HWP) position. A total of 11 polarimetric cycles were recorded with a combined integration time of 70.4 min. In addition to the science data, we recorded star center frames at the beginning and end of the se- quence as well as flux calibration frames and sky frames. For the star center frames a symmetrical waffle pattern is induced on the deformeable mirror that produces 4 satellites spots in the image. These spots can be used to accurately determine the position of the source behind the coronagraph (Langlois et al. 2013). For the flux frames the central source was offset from the coronagraph and a total of 10 images were taken with an individual exposure time of 2 s and a neutral density filter in place in order to prevent saturation. The data reduction follows generally the description given in Ginski et al. (2016), with the main difference being the instru- mental polarization and cross-talk correction. We give a short summary here. Our reduction approach uses the double differ- ence method (Kuhn et al. 2001, Apai et al. 2004). For this pur- pose we first subtract the ordinary and the extraordinary beam to create individual Q+, Q−, U+ and U− images, corresponding to HWP positions of 0◦,45◦, 22.5◦ and 67.5◦. We then subtract Q− and Q+ (and U− and U+) to remove the instrumental polariza- tion downstream from the HWP within SPHERE. This is done on a cycle by cycle basis before all resulting images are median combined to obtain the Stokes Q and U images. We also create a total intensity image, i.e. Stokes I, from our data. This is done by adding in all cases ordinary and extraordinary beams and then median combining all resulting images over all polarimetric cy- cles. The Stokes Q and U images still contain residual instru- mental polarization mainly induced by the VLT/UT3 mirror 3 and SPHERE mirror 4. To most accurately determine the angles and degree of linear polarization, it is necessary to correct for the instrumental polarization and cross-talk. For this purpose we used the detailed Mueller matrix model and correction method of van Holstein et al., in prep. (including telescope mirrors, instru- ment common path and IRDIS itself). This model was calibrated using an unpolarized standard star as well as the SPHERE/IRDIS internal (polarized) calibration light source and was validated C. Ginski et al.: First direct detection of a polarized companion outside of a resolved circumbinary disk around CS Cha with polarimetric observations of the TW Hya disk. The correc- tion was performed on each individual double difference image taking the rotation angles of all optical components from the image headers into account. The instrument polarization model was successfully applied in several recent studies of circumstel- lar disks imaged with SPHERE, such as the cases of T Cha (Pohl et al. 2017), DZ Cha (Canovas et al. 2018) and TWA 7 (Olofsson et al. 2018), as well as for the observation of substellar compan- ion polarization in the case of the HR 8799 system (van Holstein et al. 2017). Finally we used the Stokes Q and U images to compute the ra- dial Stokes parameters Qφ and Uφ (see Schmid et al. 2006). The Qφ image contains all azimuthally polarized flux as positive sig- nal and radially polarized flux as negative signal. Uφ contains all flux with polarization angles 45◦ offset from radial or azimuthal directions. In the case of single scattering by a central source, we expect all signal to be contained in Qφ and thus Uφ can be used as a convenient noise estimate. This is typically a valid assumption for disks seen under low inclination (Canovas et al. 2015). We show our final reduced polarimetric images in Fig. 1. We show the total intensity image in Fig 2. In our polarimetric images we clearly detected a compact, low inclination circumstellar disk in scattered light around CS Cha. Furthermore, we detected in our total intensity images a faint companion candidate approximately 1.3 arcsec to the West of CS Cha. After inspection of the polarized intensity images at the companion position, it became apparent that the companion is also detected in polarized light. We show the final polarized in- tensity image including circumstellar disk and companion over- laid with the angle of linear polarization in Fig 3. Since the com- panion was detected in polarized light as well as in total intensity we can calculate its degree of linear polarization. We discuss this in detail in section 5. 2.2. Archival NACO imaging data The CS Cha system was previously observed with VLT/NACO (Lenzen et al. 2003, Rousset et al. 2003) as part of a stellar and sub-stellar multiplicity survey among young Chamaeleon mem- bers (see Vogt et al. 2012 for results of that survey). Observa- tions were carried out on February 17th 2006, i.e. exactly 11 years before our new SPHERE observations. The data was taken in standard jitter mode in the Ks-band. Integration time for each individual exposure was 1 s, and 35 exposures were taken and co-added per jitter position. The total integration time of the data set was 11.7 min. We used ESO-Eclipse for the standard data reduction of the NACO data. This consisted of flat-fielding and bad pixel mask- ing, as well as sky subtraction. The individual reduced images were then registered with respect to the central source and me- dian combined. In addition to standard data reduction, we removed the radial symmetric part of the stellar PSF by subtracting a 180◦ rotated version of the image from itself. This was done in order to high- light faint companions at close angular separations and enable an accurate photometric and astrometric measurement without influence of residual stellar flux. The final reduced images are shown in Fig. 2. We re-detected the faint companion candi- date first seen in our SPHERE observations in the final reduced NACO image. 2.3. Archival HST/WFPC2 observations In addition to the NACO archival observations, CS Cha was also observed with the Hubble Space Telescope's Wide Field and Planetary Camera 2 (HST/WFPC2, Trauger & WFPC2 Science Team 1994) on February 18th 1998. CS Cha was centered in the Planetary Camera sub-aperture of WFPC2 with an effective pixel scale of 46 mas/pixel. The observations consisted of two expo- sures each in the F606W and the F814W filters, i.e. the WFPC2 equivalents of R and I-band. Exposure times for the F606W filter were 8 s for the first exposure and 100 s for the second exposure with gain settings of 14 and 7 respectively. In the F814W filter the exposure times were 7 s and 80 s with the same gain setup. The innermost 2 pixels of the primary PSF as well as additional pixels along the central pixel readout column were saturated in all exposures. The data was reduced using the standard archival HST/WFPC2 pipeline. To increase the detectability of faint point sources around the primary star we subtracted a scaled reference star PSF from the long exposure images. As reference star we used the K5 star HD 17637, which was imaged for that purpose in the same pro- gram as the science data. As noted by Krist et al. (2000), the two main factors in achieving a good PSF subtraction result are a similar spectral type of the reference star and science target, as well as the placement of the reference star on the detector. Due to the under-sampling of the HST PSF, it is important to use a reference star that was imaged as close in detector position as possible to the science target. From the multiple images that were taken of HD 17637 we thus chose the one with the smallest angular separation from the position of the science target in both filters. Since the reference star and CS Cha both had a saturated PSF core, we could not use the PSF peak for scaling of the refer- ence star PSF. We instead used an annulus along the unsaturated flanks of the PSF to compute the scaling factor. After subtraction of the reference star, we detected a faint point source at the expected companion candidate position in the F814W images with a signal-to-noise ratio of 5.0. The com- panion candidate was hidden under one of the bright diffraction spikes of the primary PSF. We show the subtracted and non- subtracted image in Fig. 2. In the F606W data set we could not find a significant detection at the companion candidate position. 2.4. NACO L-band follow-up observations To image CS Cha in the thermal infrared, we used again VLT/NACO. The observations were acquired on April 28th 2017, using the angular differential imaging (ADI) mode of NACO with the L' filter and the L27 camera following the strat- egy described by Chauvin et al. (2012). The NACO detector cube mode was in addition used for frame selection with exposure time of 0.2 s. A classical dithering sequence was used with the repetition of five offset positions to properly remove the sky con- tribution. In the end, the typical observing sequence represented a total of 57 cubes of 100 frames each, i.e a total integration time of 19 min for an observing sequence of 45 min on target. Two sequences of non saturated PSFs were acquired using a neu- tral density filter at the beginning and the end of each observing sequence to monitor the image quality. These data also served for the calibration of the relative photometric and astrometric measurements. The reduction of the ADI saturated dithered dat- acubes was performed with the dedicated pipeline developed at the Institut de Planétologie et d'Astrophysique de Grenoble (IPAG; Chauvin et al. 2012) providing various flavors of ADI algorithms. At the separation of the candidate, the background Article number, page 3 of 19 A&A proofs: manuscript no. CSCha-aa_revised noise is the main source of limitation. Spatial filtering and sim- ple derotation or classical ADI are therefore sufficient to process the ADI data. After final data reduction we did not detect the companion can- didate in our NACO Lp-band data. 2.5. SPHERE polarimetric follow-up observations Our initial SPHERE observations were followed-up on June 18th 2017 with SPHERE/IRDIS DPI observations in H-band with the goal to obtain H-band photometry of the companion candidate and to confirm its detection in polarized light. The conditions during the observations were overall poor. Even though the see- ing in the optical was on average only 0.7 arcsec, the coherence time was very short, on the order of 2 ms on average during the observations. This lead to a much poorer AO correction com- pared to the previous J-band observations. The setup of the June observations was similar to the first set of observations in February. The central binary was again placed behind the 185 mas coronagraph. We used a slightly shorter in- dividual exposure time of 64 s due to the unstable weather con- ditions. We recorded a total of 7 polarimetric cycles with a com- bined integration time of 29.9 minutes. Data reduction was performed analogously to the previous J- band data set. We found that the circumstellar disk and the companion candidate were again detected in polarized light. We could also detect the companion in our stacked total intensity images. Final reduced images are shown in Fig. 1 and Fig. 2. 3. Astrometric confirmation of the companion Since the companion was detected in our new SPHERE data as well as the archival NACO and HST data, we were able to to test if the companion is co-moving with CS Cha. To ensure minimal contamination by the central stars' flux in the SPHERE and NACO images, we subtracted a 180◦ rotated version of the images from the original (see Fig. 2). For the HST F814W im- age we used the reference star subtracted image to determine the companion position. In the case of the SPHERE and NACO images we used IDL starfinder (Diolaiti et al. 2000) to fit a reference PSF to the companion and extract its position in detector coordinates. As reference PSF we used the unsaturated stellar primary in the NACO image and the dedicated flux frames taken for the SPHERE images. Since the separation of the companion is smaller than the average isoplanatic angle at Paranal (see e.g. Martin et al. 2000), no significant distortions of the compan- ion PSF compared to the primary star PSF are expected. For the HST PSF under-sampling and residuals from the diffrac- tion spike made PSF fitting problematic. Instead we used ESO- MIDAS (Banse et al. 1983) to fit a two dimensional Gaussian to the companion position. Measurements were repeated several times with different input parameters in terms of measuring box size and starting position to ensure that the fit converged well. We used the average value of all measurements as the final ex- tracted companion position. We used the individual fitting uncer- tainties of the Gaussian fit as the uncertainty of the companion's detector position. We ensured that this uncertainty was signif- icantly larger than the standard deviation of multiple repeated measurements with different initial parameters. To extract the stellar position, we used different approaches for the SPHERE, NACO and HST data. For the NACO data no coro- nagraph was used, so we used the same approach to extract the Article number, page 4 of 19 stellar position as was used for the companion position. How- ever, for SPHERE no direct measurement was possible since the central source is obscured by the coronagraph. Instead we used the center calibration frames to determine stellar position, as de- scribed in Langlois et al. (2013). Since we had multiple center frames taken at the beginning and end of the sequence, we used the deviation between the recovered positions as the uncertainty of the central source position measurement. For the HST image the primary star was heavily saturated with significant column bleeding, making a fit to the remaining stellar PSF difficult. In- stead we fit linear functions to the positions of the diffraction spikes and used their intersection as stellar center position. To translate the recovered detector position for the central bi- nary and the companion into on-sky separation and position an- gle, our observations required an astrometric calibration. For the archival NACO data several binary stars were imaged as astro- metric calibrators during the same night as the science data as part of the original program. The results of these astrometric cal- ibrations (including also potential orbital motion of the binary calibrators), are given by Vogt et al. (2012). For the SPHERE data, calibrators are regularly imaged during the ongoing SPHERE GTO survey. Primary calibrators are stel- lar clusters such as 47 Tuc, Θ Ori B and NGC 6380. The results of these astrometric calibrations are given in Maire et al. (2016). In addition, detailed solutions for the geometric distortions were calculated by these authors. The instrument has proven to be ex- tremely stable within their given uncertainties. We thus utilize their results for the broad band J and H filter to calibrate our data. We also use their distortion solution to correct geometric distor- tions in our SPHERE image. For the true north of the J-band data we use the more recent measurement published in Chauvin et al. (2017), done within a few days of our observations. The uncertainties for the SPHERE data include those of the detector coordinates of central source and companion as well as the cali- bration uncertainty and the uncertainty of the distortion solution. Lastly, for the HST data we used the astrometric calibration pro- vided in the image header. We list final results in table 1. After we extracted the astrometry in all epochs, we measured the proper motion of the companion relative to CS Cha. The fi- nal results are shown in Fig. 4. We show three different diagrams, since the proper motion of CS Cha is given with slightly different values in the NOMAD (Zacharias et al. 2004) and SPM4 (Girard et al. 2011) catalogs, as well as by Smart & Nicastro (2014). In all three cases we can clearly reject the background hypothesis with 7.1 to 8.7 σ in separation and with 4.4 to 8.5 σ in position angle. Within the given uncertainties we observe no significant relative motion in separation over our ∼19 yr baseline. However we observe relative motion in position angle, which is consis- tent with a circular face-on or low inclination orbit, i.e. with a similar inclination as is observed for the resolved circumbi- nary disk. Within the given error bars the companion is thus co-moving with the primary stars. This is a very strong indi- cation that the companion is gravitationally bound to CS Cha. In particular it is extremely unlikely that the companion is a blended extragalactic source, since such a source would have to move at very high velocity and would need to be by-chance aligned in proper motion and close to CS Cha. The probabil- ity for a blended galactic source might be slightly higher. To quantify this we used the TRILEGAL (Girardi et al. 2005, 2012) population synthesis model to compute the number of expected galactic sources in close vicinity of CS Cha. As input we gave the galactic coordinates of CS Cha as well as the J-band magni- tude of the companion as limiting magnitude. Following Lillo- Box et al. (2012) the number of expected objects can then be C. Ginski et al.: First direct detection of a polarized companion outside of a resolved circumbinary disk around CS Cha Fig. 1: 1st row: Reduced SPHERE DPI J-band Qφ and Uφ as well as intensity image. North is up and East to the left. 2nd Row: The same for our H-band observations. Color scale (linear) and stretch are the same for all Qφ and Uφ images. We did not correct for the 1/r2 drop-off in stellar irradiation. The grey hatched disk overplotted on the images shows the size of the utilize coronagraph. translated into a probability to find a background object at a cer- tain separation. Using this approach we find that the chance of a faint blended galactic source within 1.3 arcsec of CS Cha is 0.4 %, i.e. improbable at the 2.9 σ level. Such a source would then still need to be by-chance aligned in proper motion with CS Cha making this scenario even less likely. One last concern might be that the companion could be a blended local source within the Cha I cloud but several pc behind CS Cha. For exam- ple Cambresy et al. (1998) found a number of very faint and highly embedded YSOs in Cha I. To test the likelihood of a by chance aligned local source in Cha I we checked the dispersion of proper motions of known members. As input we used the cat- alog by Teixeira et al. (2000), which contains 29 such members, including CS Cha. We find that the dispersion in proper motion is quite high with ∼18 mas/yr in right ascension and ∼73 mas/yr in declination. In contrast we find that the companion shows no significant deviation from the proper motion of CS Cha in right ascension and only 3.6±1.8 mas/yr in declination, which can be well explained by orbital motion as mentioned earlier. In Fig. B.1, in the appendix, we furthermore show that the recov- ered colors of the companion do not match the YSO colors in Cha I by Cambresy et al. (1998). We can thus firmly exclude a blended local object as well. We overall conclude that the companion is in all likelihood grav- itationally bound to CS Cha. We explore the orbital motion of the companion in detail in section 7.1. 4. Photometric measurements and detection limits 4.1. SPHERE and NACO photometry Article number, page 5 of 19 0.50.00.5J−bandQφUφ0.50.00.50.50.00.5QφH−band0.50.00.5Uφ∆Dec[arcsec]∆RA[arcsec]A&A proofs: manuscript no. CSCha-aa_revised Fig. 2: 1st row: SPHERE J and H-band intensity images as well as the NACO Ks-band image and the WFPC2 F814W image of CS Cha. North is up and East to the left. In all images the position of the faint companion candidate is marked by a white dashed circle. 2nd Row: The same images as above but subtracted with a 180◦ rotated version of themselves to remove the bright stellar halo. In the case of WFPC2 we subtracted a reference star scaled to the flux of CS Cha to remove the bright stellar PSF and especially the bright diffraction spike on top of the companion position. In the WFPC2 images we removed the central columns containing the PSF peak since they were heavily saturated. Table 1: Astrometric meassurements and calibrations of all observation epochs. Instrument Epoch 1998.1339 WFPC2 2006.1311 NACO 2017.1311 2017.4617 SPHERE/IRDIS SPHERE/IRDIS Pixel Scale [mas/pixel] 45.52±0.01 13.24±0.18 12.263±0.009 12.251±0.009 True North [deg] 31.69±0.005 0.18±1.24 -1.71±0.06 -1.75±0.11 Separation [arcsec] 1.314±0.039 1.299±0.018 1.314±0.002 1.319±0.001 Position Angle [deg] 258.26±1.21 260.30±1.24 261.41±0.12 261.40±0.23 companion was detected and derived upper limits for the non- detections. In our SPHERE J and H-band epochs, as well as in the NACO Ks-band epoch, the companion was well detected but was still located close enough to CS Cha so that the background at the companion position is dominated by the bright stellar halo. We did not image PSF reference stars (neither was a PSF refer- ence available for the archival NACO data), thus we assumed that the low frequency structure of the stellar halo is approxi- mately radial symmetric. To remove this radial symmetric halo, we subtracted 180◦ rotated versions of the images from them- selves. The results are shown in Fig. 2. While strong signal re- mains within ∼0.5 arcsec of CS Cha, the companion position ap- pears free of strong residuals. After this initial background subtraction we utilized IDL starfinder to perform PSF fitting photometry to measure rel- ative brightness between the companion and CS Cha (the latter in the unsubtracted images). We used the flux calibration frames for the SPHERE observations to obtain an unsaturated reference PSF. For the NACO Ks-band image we used CS Cha itself as reference PSF since it was not saturated during the science se- quence. Once a PSF fitting result was obtained we subtracted the companion from the data to check for strong residuals at the companion position. The results are given in table 2 as differ- Fig. 3: Polarized intensity image of CS Cha and its companion in J-band, after instrumental polarization correction. The circumbi- nary disk as well as the companion are well detected in polar- ized light. We overlayed the angle of the linear polarized light with light blue bars. The companion deviates by ∼20◦ from an azimuthal polarization w.r.t. CS Cha indicating that it is intrinsi- cally polarized and does not just scatter stellar light. To understand the nature of the faint companion, we per- formed photometric measurements in all bands in which the Article number, page 6 of 19 1.51.00.50.00.51.01.5WFPC2/F814WISPHERE/BB−JISPHERE/BB−HI1.51.00.50.00.51.01.5NACO/KsI1.51.00.50.00.51.01.51.51.00.50.00.51.01.5WFPC2/F814WI−reference1.51.00.50.00.51.01.5SPHERE/BB−JI−1801.51.00.50.00.51.01.5SPHERE/BB−HI−1801.51.00.50.00.51.01.51.51.00.50.00.51.01.5NACO/KsI−180∆Dec[arcsec]∆RA[arcsec]0.60.60.10.10.40.40.90.91.41.4∆RA[arcsec]0.50.50.00.00.50.5∆Dec[arcsec]C. Ginski et al.: First direct detection of a polarized companion outside of a resolved circumbinary disk around CS Cha (a) PM from NOMAD catalogue (b) PM from SPM4 catalogue (c) PM from Smart et al. 2014 Fig. 4: Proper motion diagrams of the companion relative to CS Cha. The wobbled gray lines are the area in which a non-moving background object would be expected. The "wobbles" are due to the parallactic displacement of such an object visible during the Earths revolution around the Sun. The dashed lines mark the area in which a co-moving companion would be located. The dashed lines take potential orbital motion into account assuming an inclination (circular face-on for the position angle and circular edge-on for the separation) and total system mass (1 M(cid:12), i.e. this assumes that the mass of the companion is small compared to CS Cha). In all three diagrams the companion is co-moving with CS Cha and thus in all likelihood gravitationally bound. We note that we see a small differential motion in position angle across our 19 year observation baseline, which is consistent with a circular face-on (or close to face-on) orbit. ential magnitudes. To convert the differential J, H and Ks mag- nitudes to apparent magnitudes of the companion we used the corresponding 2MASS (Cutri et al. 2003) magnitudes of CS Cha as calibration. We then also list absolute magnitudes for which we assumed a distance of 165±30 pc. For conversion to physical fluxes we used a HST/STIS Vega spectrum as well as the filter curves of SPHERE and NACO. We note that we find a clear systematic uncertainty in the SPHERE H-band observations, induced by the poor observing conditions. In particular the coherence time of the atmosphere degraded during the sequence, with longer values at the start than at the end of the sequence. However, our flux calibration frames were only taken at the end of the sequence, i.e. in the worst ob- serving conditions, and thus have lower Strehl than the previous science images in the sequence. Thus using them for the flux measurement of the companion during the whole sequence over- predicts the companion flux. To estimate this systematic effect we sub-divided the science sequence into four equally long bins, which we reduced individually in order to detect the companion in each bin. We then measured the relative loss of signal in the companion due to changing weather conditions between all bins. We found a deviation of 0.46 mag between the first and the last bin. We consider this as an additional error term for the lower limit (since we know the direction of the effect) of the compan- Article number, page 7 of 19 1995.02000.02005.02010.02015.02020.0time[year]256258260262264266268270272positionangle[deg]VLT/SPHEREHST/WFPC2VLT/NACO1995.02000.02005.02010.02015.02020.0time[year]1.21.31.41.51.61.71.8separation[arcsec]1995.02000.02005.02010.02015.02020.0time[year]258260262264266positionangle[deg]VLT/SPHEREHST/WFPC2VLT/NACO1995.02000.02005.02010.02015.02020.0time[year]1.21.31.41.51.61.71.8separation[arcsec]1995.02000.02005.02010.02015.02020.0time[year]256258260262264266268270positionangle[deg]VLT/SPHEREHST/WFPC2VLT/NACO1995.02000.02005.02010.02015.02020.0time[year]1.21.31.41.51.61.71.8separation[arcsec]A&A proofs: manuscript no. CSCha-aa_revised ion flux in H-band. As mentioned earlier the companion was not detected in our NACO L-band observation. We thus evaluated the detection limit of our observation. The detection performances reached by our observation were estimated by computing 2D detection limit maps, at 5σ in terms of Lp contrast with respect to the primary. We computed the pixel-to-pixel noise within a sliding box of 1.5 × 1.5 FWHM. The detection limits were then derived by taking the ADI flux loss using fake planet injection and the transmission of the neutral-density filter into account, and were normalized by the unsaturated PSF flux. Our final detection limits map is shown in Fig. 5 and the computed detection limit at the companion po- sition is given in table 2. We use the WISE (Cutri et al. 2012) W1 magnitude as close proxy for the L-band magnitude of the primary star to convert contrast limits to apparent and absolute magnitude limits. 4.2. WFPC2 photometry and detection limits To estimate the brightness of the companion in the WFPC2 F814W filter, several analysis steps were necessary. The primary star was saturated in the long exposure in which we detected the companion. To enable a relative measurement of the companion brightness we thus first determined the brightness of the primary star in the exposure. For this purpose we used TinyTim (Krist et al. 2011), a program to generate HST point spread functions based on the instrument setup, target spectral type, time of obser- vations and position on the detector. We created a matching PSF for the WFPC2 F814W observations and then fitted this theoret- ical PSF to the unsaturated flanks of the CS Cha PSF by applica- tion of a scaling factor. We then used this scaled theoretical PSF for the relative brightness measurement with the companion. The photometry of the companion in the WFPC2 image is chal- lenging since it is contaminated by the bright diffraction spike of the primary star. Even after subtraction of a reference star, residuals of this diffraction spike are still visible around the com- panion position. Due to the low S/N of the detection and the under-sampling of the HST PSF, we decided against PSF fitting photometry in this case and instead applied aperture photome- try. For this purpose we measured the flux of the companion in a 3×3 pixel box centered on the brightest pixel of the companion PSF. We then estimated the local background by measurements with the same box 3 pixels moved in radial direction towards and away from the the central star along the diffraction spike. The average of both measurements was then subtracted from the companion measurement. To estimate the uncertainty of the background measurement we computed the standard deviation in the background apertures and multiplied it by the surface area of the aperture. In addition to the uncertainty of the background we took into account the read noise of the WFPC2 planetary camera for a gain setting of 7e−/DN. We used a read noise of 5 e−/pix. Overall the measurement is strongly dominated by the uncertainty of the background, which is a factor 4 higher than the estimated read noise. We give our result for the relative bright- ness measurement in magnitudes in table 2. To convert this relative measurement to an apparent magnitude of the companion, we determined the Vega magnitude of CS Cha in the F814W filter. For this purpose we calculated the total flux of CS Cha in the F814W filter using the filter curve and the spec- tral energy distribution of CS Cha, shown in Fig. 6. We then con- verted this to a Vega magnitude by comparison with the flux of Vega in the same filter. We give apparent and absolute magni- tudes for the companion also in table 2, along with the physical flux of the companion in the F814W filter. Article number, page 8 of 19 Fig. 5: Detection limit map derived from our NACO Lp-band ob- servations. Detection limits are given in relative contrast to the the primary star. We mark the expected position of the compan- ion with a black, dashed circle. The companion was not detected in these observations and should thus exhibit a contrast larger than 8.2 mag relative to CS Cha. In the F606W filter the companion was not detected. We thus es- timated detection limits at the companion position. For this pur- pose we measured the standard deviation at the companion po- sition in a 3×3 pixel aperture. We again used TinyTim to create an unsaturated reference PSF scaled to the primary star bright- ness on the detector. Given the noise at the companion position and using the primary star as reference we then computed the limiting magnitudes for a 5 σ detection. The result is given in table 2. 5. Polarization of the companion Since we detected the companion both in the total intensity and in the polarized intensity images in our SPHERE/IRDIS J and H- band epochs, we could calculate the degree of linear polarization of the companion. For this purpose we used aperture photometry in both images in each band. We used aperture photometry over PSF fitting photometry due to the slight change in the compan- ion PSF during the double difference steps of the DPI reduction. We checked in J-band that PSF fitting photometry and aperture photometry give consistent results in the total intensity image. We found that the results were consistent within 0.01 mag. We used an aperture radius of 3 pix in J-band, which corresponds to the full width at half maximum of 2.77 pix as measured by fitting a Moffat profile to the stellar PSF. In H-band we used a value of 4 pix due to the poorer observing conditions. As in the PSF fit- ting photometry, we first subtracted the radial symmetric bright stellar halo from the intensity images by rotating them 180◦ and subtracting them from themselves. We then estimated the local background with two sub-apertures in each band. In the J-band case, the companion is slightly contaminated by a stellar diffrac- tion spike. We thus used two sub-apertures in radial direction along this spike. In H-band we used two azimuthal sub-apertures at the same separation from the central star and offset by a few degrees from the companion position. The measurement in polarized intensity was performed in the same way as in the intensity image. However, the measurements 2.01.51.00.50.00.51.01.52.0∆RA[arcsec]2.01.51.00.50.00.51.01.52.0∆Dec[arcsec]NACO/LP−band012345678910∆magC. Ginski et al.: First direct detection of a polarized companion outside of a resolved circumbinary disk around CS Cha were actually performed in the Stokes Q and U images rather than in the combined polarized intensity image, since all signal becomes positive in this image and thus even background noise might give a spurious polarization signal. We find a degree of linear polarization in J-band of 13.7±0.4 % and an angle of linear polarization of 153.0◦±0.8◦. Our H-band results are consistent with these measurements. We find a degree of linear polarization of 14.1±1.4 % and an angle of linear po- larization of 154.0◦±2.9◦. The uncertainties in the H-band mea- surements are higher due to the poorer signal-to-noise compared to the J-band data. In both cases the error bars are strongly domi- nated by measurement uncertainties (due to photon, speckle and background noise), while the instrument model allows for a fac- tor of ∼10 higher accuracy. We now need to investigate if the polarization of the compan- ion is intrinsic to the object, i.e. either due to scattered light from the primary stars or a central object within the compan- ion itself, or if it is caused by interstellar dust between Earth and the CS Cha system. This is of particular importance since CS Cha is indeed located within close proximity or behind the Cha I dark cloud. Detailed optical polarization measurements of the region have been performed by Covino et al. (1997). In Fig. 7 we show their optical polarization map of the Cha I cloud region and superimpose the position of CS Cha. Using the nine stars in their study located closest to the position of CS Cha, we find an average maximum polarization degree of 6.7±1.7 % at a peak wavelength of 0.65 µm. The average angle of polarization in the I-band that they measure for the same stars is 132.08◦±12.9◦. Using Serkowski's empirical law (Serkowski et al. 1975) we can extrapolate the expected polarization degree in the J and H-band. (cid:104)−K ln2(λ/λmax) (cid:105) p(λ) = p(λmax) exp (1) Wherein p(λ) is the polarization degree at the wavelength λ. The factor K was empirically determined to be dependent on the peak wavelength of the polarization degree by Wilking et al. (1982). K = 1.86 λmax − 0.1 (2) Using these relations with the average values from Covino et al. (1997) we find that the degree of linear polarization could lie between 5.5 % and 3.3 % for the J-band and between 3.4 % and 2.0 % for the H-band. These degrees of polarization are sig- nificantly lower than what we find for the companion giving a first indication that the polarization of the companion is indeed intrinsic and not caused by interstellar dust. To test this more rigorously we measured the degree of linear polarization of the unresolved primary stars. For this purpose we used an annulus at the bright speckle halo that marks the adaptive optics correction radius and that contains only stellar light. We find that the primary stars have a degree of linear po- larization of 0.57±0.28 % in J-band with an angle of polariza- tion of 133.1◦±8.2◦. For the H-band we find similar values of 0.34±0.02 % and 141.1◦±5.4◦ for the degree and angle of lin- ear polarization respectively. Both of these values are consis- tent with a previous measurement in the optical of CS Cha by Yudin (2000) who find a degree of linear polarization of 0.7 % (but did not provide uncertainties). The degree of polarization that we find for the primary stars is much lower than suggested by the optical data by Covino et al. (1997) in combination with Serkowski's law. It could be that the average value we assumed is not a good proxy for the cloud density at the position of CS Cha Article number, page 9 of 19 Fig. 6: Spectral energy distribution of CS Cha (blue dots) and its companion (red dots and triangles). Pointing down triangles denote upper limits. Spectral flux densities were computed from broad band photometry using a Vega spectrum and the broad band filter curves. All values for the companion are given in ta- ble 2 Fig. 7: Reproduction of Fig. 2 from Covino et al. (1997). Shown is a visual polarization map of the Cha I dark cloud using 33 stars they observed. Average angle and degree of polarization is indicated by the solid line vector field. We show the position of CS Cha in this map to indicate the expected degree of linear polarization introduced by the Cha I dark cloud in the optical. 100101λ[µm]10-1710-1610-1510-1410-1310-12λFλ[W/m2]WFPC2/F606WWFPC2/F814WSPHERE/BB−JSPHERE/BB−HNACO/KsNACO/L0CS ChaAMES-DUSTY 2MJupAMES-DUSTY 5MJupAMES-DUSTY 20MJupCS Cha b/BCS ChaA&A proofs: manuscript no. CSCha-aa_revised or that CS Cha is located slightly in front of the cloud. In any case, the low degree of stellar polarization strongly suggests that the high degree of polarization found for the companion is intrin- sic to the object and not caused by interstellar dust if we assume both objects are located at the same distance as suggest by their common proper motion. This conclusion is additionally supported by the disagreement between the angle of polarization of the companion and that of stellar sources. In both bands the angle of linear polarization of the stellar binary is within 1 σ consistent with the average polar- ization angle in the region as determined by Covino et al. (1997). In our data we find that the companion polarization angle devi- ates by ∼ 22◦ (2.6σ) from the stellar polarization angle in J-band and by ∼ 15◦ (2.4σ) in H-band. Thus it seems again plausible that the cause for the polarization of the stellar binary and the companion are different and that the companion polarization is not caused by interstellar dust. Given the angle of polarization of the companion, we can finally try to understand which is the dominating source of illumina- tion, assuming polarization by single scattering of light. If the companion is primarily illuminated by the central stellar binary we would expect its angle of linear polarization to be azimuthal with respect to the binary position. The expected angle of linear polarization for azimuthally scattered light at the companion po- sition is 171.7◦±0.1◦. Comparing this to the more accurate angle of linear polarization in J-band, we find a significant deviation of 18.7◦±0.8◦. We can thus conclude that the origin of the po- larized light is not (entirely) single-scattered light emitted by the primary stars. It is of course still possible that the linear polar- ization that we measure is a superposition of scattered stellar and companion emission. However, given the angle of polariza- tion we can already conclude that the companion object contains a central source massive enough that we can detect its' emission. Polarization can give us important information about the struc- ture of the atmosphere of low-mass objects, as well as their direct environment. Polarization has indeed been measured for field brown dwarfs previously (see e.g. Ménard et al. 2002, Zapatero Osorio et al. 2005 and Miles-Páez et al. 2013), but was not de- tected so far for companions to nearby stars (see e.g. Jensen- Clem et al. 2016, van Holstein et al. 2017). This is to the best of our knowledge the first time a faint and thus likely low-mass companion to a nearby star was detected in polarized light and its degree of polarization measured. We discuss the implications for the object in detail in section 7. 6. The circumbinary disk around CS Cha 6.1. Position angle and inclination As visible in Fig. 1, we resolve for the first time a small disk around the central stellar binary in the CS Cha system. The disk appears compact, smooth and close to face-on. From our scat- tered light images we can extract the orientation of the disk. For this purpose we measure the disk diameter in radial disk profiles with orientations between 0◦ and 360◦ in steps of 2◦. The re- sulting disk diameter versus disk orientation data was fitted with the corresponding value for an ellipse. The disk diameter was de- fined in our radial profiles as the separation between the two out- ermost points at which the disk flux reaches a certain threshold. To determine this threshold we measured the standard deviation of the background outside of the disk signal and set the threshold to a multiple of this standard deviation. In practice we found that there is a small dependency on the threshold value and the re- covered disk orientation. We thus used multiples between 5 and Article number, page 10 of 19 100 in steps of 2 and considered the recovered median values for disk inclination and position angle, and the standard deviation between these values, as the uncertainty of our measurement. Assuming a radial symmetric disk that only appears elliptical due to its relative inclination towards us, we find a inclination of 24.2◦±3.1◦ and a position angle of 75.6◦±2.2◦ from our J-band observation. This disk position angle is well consistent with the position angle of the suspected jet emission of ∼ 162◦ detected by Pascucci et al. (2014), since the jet position angle should be offset by 90◦ from the disk major axis. The H-band observa- tion has much lower signal-to-noise than the J-band observation and suffers from convolution with a rather distorted PSF (see Fig. A.1). We find an inclination of 34.9◦±10.6◦ and a position angle of 86.1◦±2.2◦ for this data set. The ∼10◦ larger position an- gle can be explained by the elongated PSF shape and orientation of this observation. We thus consider the J-band measurements as final values for inclination and position angle. 6.2. Inner and outer radius To measure the outer radius of the disk we considered a radial profile along the major axis as determined in the previous sec- tion. We then computed the radial extent at which the disk signal is for the first time 5σ above the image background value. We again used the J-band images, due their higher quality. We found an outer radius of scattered light of 337 mas, i.e. 55.6 au at a dis- tance of 165 pc. This is consistent with the upper limit of 169 au given by Dunham et al. (2016) from their unresolved ALMA ob- servations. Note that we are only tracing small dust at the disk surface, so that it is possible that the disk has a larger size but is partially self shadowed. Another possibility is that the disk outer extent is larger, but that it is below the noise floor in our images due to the 1/r2 drop-off of the stellar irradiation. We show an azimuthally averaged radial profile of the disk in Fig. 8. In this profile a decline in brightness inside of ∼115 mas is visible. To investigate if this is a tentative detection of a cav- ity, we compared the radial disk profile with a model profile of the coronagraph attenuation. The NIR APLC coronagraph normalization profile was calculated based on IRDIS DB_H23 dual-band imaging observations of the 0.6" diameter disk of Ceres, performed on the 14th of December 2016. This was car- ried out in the N_ALC_YJH_S coronagraph imaging mode and the Ceres disk was nodded off-center by 490 mas to provide a non-coronagraphic reference. This was used to produce a 2D attenuation profile of the coronagraph for an extended, inco- herent source. Monochromatic Fourier modeling of the three- plane APLC coronagraph was also performed, using the APO1 SPHERE amplitude apodiser, ACL2 (185 mas diameter) focal- plane mask and NIR Lyot stop including dead actuator masks (Guerri et al. 2011, Sauvage et al. 2016). This model confirmed that the observed Ceres attenuation profile is nearly diffraction- limited and azimuthally symmetric. The radial profile outside of 85 mas is dominated by direct throughput of the target, while that inside 85 mas is dominated by internally scattered light in the instrument (for full results see Wilby et al., in prep.). The close agreement between the forward model and observed data allows the H23-band profile to be extrapolated to J-band via an equivalent model at 1.26 µm. This was then used to correct the radial CS Cha profile for coronagraph attenuation. As visible in Fig. 8, after the correction with the coronagraph throughput profile, no significant decline in flux is visible outside the coronagraphic mask. We can thus put an upper limit on the size of the inner cavity of the CS Cha disk of 15.3 au (92.5 mas C. Ginski et al.: First direct detection of a polarized companion outside of a resolved circumbinary disk around CS Cha Table 2: Photometric measurements of the companion. The apparent magnitudes in J, H and Ks-band were calculated using the closest 2MASS magnitude of CS Cha as calibration. The apparent magnitude in the HST F814W filter was computed using the theoretical Vega magnitude of CS Cha in this band given its SED. The absolute magnitudes were computed from the apparent magnitudes assuming a distance of 165±30 pc. We give the central wavelength and spectral width of all filters along with the measurements. Spectral flux densities were computed using the filter curves of the instruments as well as a Vega spectrum taken with HST/STIS. Instrument HST/WFPC2 HST/WFPC2 SPHERE SPHERE NACO NACO Filter λc [µm] F606W 0.5997 F814W 0.8012 1.245 BB-J BB-H 1.625 2.18 Ks Lp 3.80 ∆λ [µm] 0.1502 0.1539 0.240 0.290 0.35 0.62 ∆mag >8.9 9.81±0.48 10.05±0.21 9.20+0.61−0.15 9.21±0.16 >8.2 >20.4 app. magnitude 19.71±0.48 19.16±0.21 17.65+0.62−0.16 17.40±0.16 >16.4 >14.3 abs. magnitude 13.62±0.62 13.07±0.45 11.56+0.74−0.43 11.32±0.43 >10.3 < 2.03 · 10−16 Fλ [W·m−2 · µm−1] (1.37 ± 0.76) · 10−16 (6.31 ± 1.39) · 10−17 (2.54+0.41−1.95) · 10−16 (3.44 ± 0.56) · 10−17 < 1.35 · 10−17 at 165 pc) from the scattered light imaging (tracing small dust grains). 7. The nature of the companion To understand the nature of this new companion, we compare its SED to known substellar objects in Chamaeleon as well as theoretical model atmospheres. We then use the astrometry over a 19 yr base line to determine if it is possible to constrain the companion mass from the orbital motion. Finally we use our own radiative transfer models to explain the photometry and degree of linear polarization of the companion. 7.1. A planetary mass object on a wide orbit? Photometry In Fig. 6 we show the measured SED of the companion, and compared it with theoretical models of low mass substellar objects calculated with the Phoenix (Helling et al. 2008) atmosphere code using the AMES-Dusty models (Chabrier et al. 2000, Allard et al. 2001) for an age of 2 Myr as input. We tentatively explored a mass range between 2 MJup and 20 MJup. The closest fit is achieved with a 5 MJup planet corresponding to an effective temperature of 1700 K. However, it is clearly visible that even this best fit does not properly explain the measured photometry of the companion. While the J-H color may be explained by such an object (taking some red-ward shift due to extinction by circum-planetary material into account), the model clearly over-predicts the flux in K and L-band by an order of magnitude. A significantly lower mass object of 2 MJup corresponding to an effective temperature of 1100 K could explain the K-band photometry, but still significantly over-predicts the L-band flux and is not compatible with the flux in the shorter wavelengths. Generally there is no model that can explain all photometric data points and upper limits. This is a strong indication that we are looking at an object that is either significantly more complex than a "naked" planetary photosphere, or that the object (or the primary) is for some reason strongly variable. Variability is indeed possible since all observation epochs have been taken months and sometimes decades apart. One explanation for the peculiar shape of the SED could be a companion with a small (unresolved) surrounding disk. There are in fact two comparably faint objects in Chamaeleon known around which a circumplanetary disk is expected. One of them is the wide direct imaging companion to the young T Tau star CT Cha (Schmidt et al. 2008). CT Cha b shows Pa β emission in the J-band (Schmidt et al. 2008), as well as strong H α emission in the R-band (Wu et al. 2015), both strong indicators for ongoing accretion of material on the companion. The companion mass is estimated to be 9-35 MJup with a temperature range between 2500 K and 2700 K (Schmidt et al. 2008, Bonnefoy et al. 2014, Wu et al. 2015). We show the near infrared spectrum of CT Cha b along with optical photometry in Fig. 9. We overplot the photometry of the companion to CS Cha for comparison. While R, I and H-band photometry are comparable in both objects, the J and K-band fluxes of CT Cha b are significantly larger than for the CS Cha companion. A sec- ond comparison object is the recently discovered free floating planetary mass object Cha J11110675-7636030 (Esplin et al. 2017), for which we also show available photometry in Fig. 9. Assuming an age range of 1-3 Myr and using a variety of planet evolutionary models, Esplin et al. (2017) find a mass range of 3-6 MJup for this object. They note that the mid-IR photometry suggests the existence of excess emission best explained by circum-planetary material. The object shows J-K colors similar to the CS Cha companion, and is also consistent with the L-band non-detection. The H-band photometry of both objects, on the other hand, differs significantly. For both comparison objects CT Cha b and Cha J11110675-7636030 we have no information on the geometry of the surrounding circum-planetary material. In particular we do not know the inclination of these inferred disks. It is possible that the companion around CS Cha is indeed more massive than both these objects, but is strongly extincted by a very inclined circum-companion disk. This scenario is indeed also supported by the high degree of linear polarization that we find for the companion. We thus explore several models with circum-companion material in section 7.2. Astrometry Since we have an observational baseline of ∼19 yr, we at- tempted to fit the orbital motion of the companion around the primary stars. For this purpose we used the Least-Squares Monte-Carlo (LSMC) approach as described in Ginski et al. (2013). We generated 107 random orbit solutions from uniform priors and then used these as starting points for a least squares minimization with the Levenberg-Marquardt algorithm. In contrast to Ginski et al. (2013), we did not assume a system mass but left it as a free parameter. To limit the large parameter space we constrained the semi-major axis to values between 0.5 arcsec and 3.0 arcsec. This seems justified given the current position of the companion at ∼1.3 arcsec and the fact that we see no significant change in separation between astrometric epochs. In addition, we limited the total system mass to values Article number, page 11 of 19 A&A proofs: manuscript no. CSCha-aa_revised (a) (b) Fig. 8: Left: Azimuthal average of the polarized intensity profile of the circumbinary disk around CS Cha in J-band (red squares). The profile was measured in the Qφ image, while the estimated uncertainties were determined in the Uφ image. We indicate the radius of the coronagraphic mask with the black dotted line. In addition, we show the throughput curve of the utilized coronagraph as discussed in section 6.2 (green solid line). Finally we show the azimuthal disk profile corrected by the coronagraph throughput (blue diamonds). Right: Azimuthal average of the polarized intensity profile of the circumbinary disk around CS Cha in J-band (blue solid line) and H-band (red dash-dotted line). Angular separations were converted to projected separations using the distance of 165 pc. are large compared to the SPHERE measurements, the fits are strongly dominated by the latter. We find that the current astrometric epochs do not allow for constraint of the mass of the companion, since we find valid orbital solutions for the full range of input masses. However, we can make a few observations about the system architecture. If the companion is indeed a Jovian planet or brown dwarf, then we can conclude that it must be on an eccentric orbit with the lower limit of the eccentricity between 0.2 and 0.26 depending on the central stars' masses. In fact the total system mass should be above 1.4 M(cid:12) to allow for circular orbits. In this case the companion would be a low mass star with a mass between 0.4 M(cid:12) and 0.5 M(cid:12). Independent of the mass, we find an upper limit for the eccentricity of 0.8. This upper limit is, however, introduced by our artificial cut off of the semi-major axis at 3 arcsec. If we allow for larger semi-major axes, then we find even more eccentric orbits. This correlation between semi-major axis and eccentricity is indeed common for orbits which are not well covered with observations (e.g. Ginski et al. 2014). Overall we find a peak of the eccentricity at ∼0.6. The vast majority of these eccentric orbits exhibits a face-on inclination. It is interesting to investigate if co-planar orbits with the resolved circumbinary disk are possible since this could give an indication of the formation history. We find that such co-planar orbits indeed exist. However, regardless of the total system mass there are no circular (e = 0) co-planar orbits recovered. Overall the distribution of the total mass and eccentricity closely match the non-coplanar case. In Fig. 11 we show the three best fitting orbit solutions that were recovered by our LSMC fit as well as the best fitting solutions for a circular, co-planar and low mass (companion mass below 0.03 M(cid:12)) orbit. The respective orbital elements are given in table 3. The best fitting orbits are not co-planar and exhibit eccentricities between 0.41 and 0.63. Since most of these orbits are seen face-on there would be a significant misalignment be- tween the inclination of the resolved circumbinary disk and the Fig. 9: Spectral energy distribution of the CS Cha companion (red dots and triangles). Pointing down triangles denote up- per limits. We show the known substellar companion CT Cha b (Schmidt et al. 2008, Wu et al. 2015) as well as the free floating planetary mass object in Chamaeleon Cha J11110675-7636030 (Esplin et al. 2017) as comparison. between 0.9 M(cid:12) and 2.0 M(cid:12). The lower end of this mass interval is determined by the lower limit of the combined mass of the central binary star, i.e. in this case the companion mass would be small compared to the primary mass in the planet or brown dwarf regime. The upper end is given by twice the upper limit of the central binary mass, i.e. in this case the companion would have roughly one solar mass. We do not expect the companion to be more massive than the primary stars, since the resolved circumbinary disk would otherwise likely be truncated to an even smaller outer radius. In Fig. 10 we show the resulting semi-major axis, inclination and mass versus eccentricity distributions of the 1% best fitting orbits. Since the uncertainties of the NACO and HST epochs Article number, page 12 of 19 100150200250separation[mas]0.00.51.01.5PolarizedIntensity[arbitraryunits]CoronagraphIWAcoronagraph profiledisk profilecorrected profile101520253035404550separation[au]0.00.20.40.60.81.01.2PolarizedIntensity[arbitraryunits]CoronagraphIWAJ-bandH-band0.51.03.0λ[µm]10-1710-1610-15λFλ[W/m2]CS Cha b/BCT Cha b, Schmidt et al. 2008CT Cha b, Wu et al. 2015J11110675 Esplin et al. 2017C. Ginski et al.: First direct detection of a polarized companion outside of a resolved circumbinary disk around CS Cha Table 3: Orbit elements of the three best-fitting orbits shown in Fig. 11, as well as the best fitting circular (c.), co-planar (c.-p.) and low mass (l.m.) orbits. 1 a ["] 1.82 m [M(cid:12)] 1.66 e 0.49 P [yr] 4048.4 i [deg] Ω [deg] 200.0 ω [deg] 337.0 T0 [yr] 1651.9 0 2 0 2.65 1.28 0.64 8076.2 185.4 3.3 1668.1 3 0 1.43 1.84 0.41 2667.0 190.0 327.7 1602.1 c. 1.46 1.96 0 2678.6 45.2 110.1 275.7 3011.5 c.-p. 1.48 1.96 0.45 2731.6 21.6 75.3 83.3 4363.7 l.m. 2.14 1.02 0.56 6579.6 184.0 0 0 1593.6 orbital plane, as well as a misalignment with a putative highly inclined circum-companion disk. Such spin-orbit and spin-spin misalignments in multiple systems are indeed predicted by hydrodynamic simulations of stellar formation in clusters (see e.g. Bate 2012, 2018) and were more recently observed in multiple systems with ALMA (see the case of IRAS 43, Brinch et al. 2016). The total system masses for the best fitting orbits lie between 1.28 M(cid:12) and 1.84 M(cid:12), which puts the companion in the low stellar mass regime. However, we stress that lower (e.g. planetary) masses for the companion can not be ruled out with the existing astrometry. One example for an orbital solution that fits the astrometry and requires only a companion mass below 0.03 M(cid:12) is shown in Fig. 11. In general these best fitting orbits may still change significantly, with the availability of new high precision astrometric epochs in the future. Thus while the recovered distributions of orbital elements are meaningful, we caution to over interpret these specific orbit solutions. 7.2. Detection of a circum-companion disk? To test whether the peculiar measurements of the companion can be explained by a substellar object surrounded by a disk, we aimed to model its photometric (SED) and polarimetric (de- gree of polarization) properties using the radiative transfer code RADMC3D. In all our models, we consider astronomical sili- cates (Weingartner & Draine 2001), and use a single dust size for simplicity. We again consider the AMES-DUSTY atmosphere models as input for the central object. We run 3 different fami- lies of model. (a) Our first model includes a disk around a substellar com- panion. The circumplanetary disk extends up to 2 au. This max- imum outer radius was inferred from the fact that we do not re- solve the companion. Its density is described as: (cid:32) −z2 (cid:33) ρdisk(r, z) = √ Σ(r) 2πHp(r) exp 2Hp(r)2 (3) where Σ(r) is the surface density and Hp(r) is the pressure scale height. Both quantities are described as power laws with radius, with exponents ζ (flaring) and p. The model parameters are given in Table 4. The model has a complex parameter space, which we explored qualitatively by varying the mass and thus the lumi- nosity of the central object, the grain size, the disk mass and inclination. To produce a significant level of polarization (above 5%), the disk must be strongly inclined (seen close to edge-on), in turn extinguishing the thermal emission from the companion and the innermost disk radii. To match the SED at high incli- nations we thus must increase the mass of the central object. An increase in disk mass has an effect similar to an increase of the inclination on the SED of the central object. Finally the grain size can be varied to modulate the polarization efficiency. We considered grain sizes of 0.5 µm, 1 µm and 2 µm. We show a sketch of the model, along with the best fitting results for a 5 MJup and a 20 MJup companion in the left column of Fig. 12. For the lower mass we require a low disk inclination in order to get enough flux of the companion in the near-infrared. However, this model under-predicts the I-band flux and over-predicts the L-band flux, in addition to not revealing significant polarization. For the higher mass we find a much better fit. We can increase the disk inclination to much higher values, which match the J and H-band polarization well. Furthermore, the resulting SED is a close fit to all photometric measurements of the compan- ion, excluding the H-band. We have also investigated whether we can derive an upper mass limit for the companion by plac- ing a 72 MJup companion in the center of the disk. We found that even for high inclinations, such a model severely over-predicts the K and L-band flux. Note that in this model, we do not consider the binary as a source of irradiation. The only way that the companion would scatter a significant amount of light from the central binary, at that dis- tance, would be if it stands outside of the plane of the circumbi- nary disk. Otherwise the light from the central binary would be blocked by the disk. We therefore tested the same model, but with an additional irradiation source (the central binary) at 214 au, and after placing the companion and disk outside of the plane of the circumbinary disk. We find that the scattered light signal from the central binary alone is between 2-3 orders of magnitude fainter than our measurements (see Fig. C.1 for com- parison). It thus only has a marginal influence on our modeling results and was ignored for simplicity. (b) We then changed our models to test a different geometry, and considered an envelope of dust grains surrounding the com- panion, as this should enhance the amount of scattered light. The density structure is given by:  √ (cid:113) r2 + z2 R2 out + z2 out q ρenv(r, z) = ρ0 (4) where ρ0 is the density of the envelope at its outer radius. We chose the mass of the envelope so that its optical depth would be the same as in the disk model. Since in this model we do not have a disk to modulate the flux of the companion, we only tested models for a 5 MJup central object, which provided the closest match to the companion SED. The results are shown in the middle column of Fig. 12. We find a similar match to the SED as in the disk model for a low mass companion, but our models underestimate the degree of polarization with values at most in the order of a few % (∼7% for micron-size dust grains). (c) Our final model is a combination of the previous two. We consider a companion surrounded by a disk plus an additional envelope. For this model, we consider a 20 MJup companion since it provided the best fit for the disk-only model. The density at each (r,z) is taken as the maximum between ρdisk and ρenv. Note that the mass of the envelope is negligible compared to that of the disk (∼0.4%). This configuration allows us to obtain a large degree of polarization, by increasing the amount of scattered light with the envelope, while reducing the total intensity from the central object with an inclined disk. In the right column of Fig. 12, we show the results for three models with different grain sizes. Note that we also varied the inclination of the disk in order to obtain a good fit of the data. Article number, page 13 of 19 A&A proofs: manuscript no. CSCha-aa_revised (a) (b) (c) Fig. 10: Left: Semi-major axis versus eccentricity distribution of all recovered orbit solutions for the companion following the LSMC approach. Shown are the 1% best fitting orbits. Middle: Same as left, but for eccentricity versus inclination of the orbital plane. Right: Same as left, but for eccentricity versus total system mass. grains also lead to a dramatic over prediction of the degree of polarization in the near infrared. Larger grains than 1 µm do not significantly contribute to the degree of polarization in the J and H-bands. Given these results it is conceivable that a more complex grain size distribution (instead of a single grain size) including grain sizes between 0.1 µm and 1 µm may be able to reproduce the degree of polarization as well as the photometry. However, we would like to point out that the parameter space is complex and degenerate between multiple parameters, such as companion mass, disk inclination and dust grain size. Thus we do not claim that the disk plus dust envelope model with the given parameters is the only model that can reproduce our measurements. Additional measurements are needed before an attempt is made to constrain the nature of the companion to CS Cha further. An observation with SPHERE/ZIMPOL to detect the companion in optical polarized light could help to constrain the dust grain sizes as well as the presence of a dust envelope. An ALMA observation on the other hand may constrain the mm-dust mass at the companion position and thus indirectly the mass of the companion itself. From the angle of polarization we can deduce the geometry of such a system. The angle of polarization will mostly be determined by the region of the unresolved disk from which we receive the largest amount of polarized light. In the disk only model, this is the earth-facing forward scattering side of the disk, and in the the disk+envelope model these are the "poles" of the circular envelope away from the disk. In both cases we would thus expect that the angle of polarization is aligned or closely aligned with the position angle of the circum-companion disk. We note that this scenario would change in the presence of an outflow which dominates as the source of scattered light. In such a case we would expect the angle of polarization to be perpendicular to the disk plane (Tamura & Sato 1989). However, we have not modelled such a scenario. We have in general, not included this geometrical consideration in our models since the degree of polarization and the photometry are independent of the disk position angle. 8. Summary and conclusions We observed the CS Cha system for the first time in high res- olution polarimetry with SPHERE/IRDIS in J and H-band. We resolved a circumbinary disk with an outer extent of 55.6 au in Fig. 11: Best fitting orbits to the current astrometry of the com- panion as recovered by our LSMC fit. The inset in the upper left is zoomed-in on the data points. Although none of our models fit both the photometry and the level of polarization perfectly, we find that a disk composed of 1 µm sized dust grains and a high inclination of 80◦ is consistent with the observed photometry. This model still under predicts the polarization in the J-band by a factor of ∼1.7. Our model using smaller grains on the other hand will fit the HST and SPHERE J-band photometry slightly better, while it misses the SPHERE H-band and NACO L-band measurements. Smaller Article number, page 14 of 19 0.00.10.20.30.40.50.60.70.8e1.01.52.02.5a[arcsec]0.000.250.500.751.001.251.501.752.00log10(N)0.00.10.20.30.40.50.60.70.8e0102030405060i[deg]0.00.40.81.21.62.02.42.83.2log10(N)0.00.10.20.30.40.50.60.70.8e1.01.21.41.61.82.0systemmass[Mfl]0.00.20.40.60.81.01.21.41.6log10(N)21012∆RA[arcsec]1012345∆Dec[arcsec]best fit 1best fit 2best fit 3circularco-planarlow mass1.321.260.320.280.240.200.16C. Ginski et al.: First direct detection of a polarized companion outside of a resolved circumbinary disk around CS Cha Fig. 12: 1st row: Sketches of the model families described in section 7.2. We show in all cases a cross-section. 2nd row: Photometry of the companion along with the model photometry for different model parameters. The legend gives information about the assumed companion mass, the size of the considered dust grains, as well as the circumplanetary disk inclination (in models a) and c)). 3rd row: Same as the second row, but for degree of linear polarization. Colors and line styles represent the same models as in the previous row. scattered light. The disk cavity predicted by previous studies was not detected due to the limited inner working angle of the coronagraph. The upper limit for the radius of the disk cavity is 15.3 au, consistent with previous models by Ribas et al. (2016) using unresolved Herschel data. We find that the disk has an in- clination of 24.2◦±3.1◦. Outside of the disk at a projected separation of 214 au we find a faint companion with an extreme degree of linear polarization. To our knowledge this is the first faint and likely very low mass companion to a nearby star that has been discovered in polar- ized light. With HST and NACO archival data we show with high confidence that the companion is co-moving with the pri- mary stars, and is thus bound to the system, placing it at the same distance and age as CS Cha. The complex photometry of the companion could not be explained with current atmosphere models. If just J and H-band were considered, a 5 MJup mass may be inferred. However, this does not fit the photometry in other bands, in particular the non-detection in L-band. Furthermore, a "naked" substellar companion is expected to have a low intrinsic polarization. Stolker et al. (2017a) showed recently that the ex- pected degree of linear polarization from such a companion due to rotational oblateness or patchy cloud covers should not exceed 3 % and is typically lower2. Thus we suggest that we are looking at a companion with a surrounding disk or dust-envelope. We 2 See also the previous work by Sengupta & Krishan (2001), who find a similar range for the degree of linear polarization. Article number, page 15 of 19 a)b)c)10-2010-1910-1810-1710-1610-1510-14λFλ[W/m2]5MJup/0.1µm/30◦20MJup/1µm/82◦5MJup/0.1µm5MJup/1.0µm10-2010-1910-1810-1710-1610-1510-1420MJup/0.1µm/77◦20MJup/1.0µm/80◦20MJup/2.0µm/75◦0.51.03.001020304050p[%]0.51.03.0λ[µm]0.51.03.001020304050A&A proofs: manuscript no. CSCha-aa_revised Table 4: Radiative transfer model parameters. Model Parameters Mcomp [MJup] Teff [K] Rcomp [R(cid:12)] Rin [au] Rout [au] Mdisk [M(cid:12)] Menv [M(cid:12)] ρ0 [g/cm3] Hp(Rout)/Rout ζ p q (a) Disk (b) Envelope 5/20 1580/2500 0.17/0.25 0.003 1.9 10−7 2 - - 0.18 0.25 -1 - 1580 0.17 0.003 8.5 10−9 1 10−16 5 2 - - - - -1 (c) Disk & Envelope 2 20 2500 0.25 0.003 1.9 10−7 2.3 10−10 5 10−17 0.18 0.25 -1 -1 explored the wide parameter space for such a model with the ra- diative transfer code RADMC3D. We find that we can explain the companion SED and polarization reasonably well, with ei- ther a highly inclined disk around a 20 MJup object or with a disk and additional dust envelope around an object of the same mass. This puts the companion clearly in the substellar regime: either a very low mass brown dwarf or a high mass planet. From our orbit fit to the available astrometry over a time base line of 19 yr, we can conclude that the orbit of the companion is likely eccentric with a minimum eccentricity of 0.3. This gives some indication of how the companion may have formed. For an in-situ formation, either by core-accretion or by gravitational collapse in the outer circumbinary disk, one would not expect an eccentric orbit. Also the strong misalignment of the circumbi- nary and the circum-companion disk do not fit these scenarios. However, the eccentricity may be explained by dynamical inter- action with the unresolved stellar binary. The two systems could be caught in Kozai-Lidov type resonances effectively exchang- ing relative inclination and eccentricity (see e.g. Takeda & Rasio 2005). Another possibility for an eccentric orbit would be the formation at close separations in the circumbinary disk and a subsequent dynamical scattering event in which again the cen- tral binary may have played a role. However, in such a scenario one would expect that the companion lost the surrounding disk and that some sign of perturbation would be visible in the cir- cumbinary disk. Both do not seem to be the case. While for typical planet formation scenarios the location and ec- centricity of the orbit of the CS Cha companion are problematic, this is less so for a more star-like formation by collapse in the molecular cloud in which also the CS Cha binary formed. In such a case the misaligned disks around the companion and the stellar sources would also not be problematic as many such examples are known, most prominently the HK Tau system (Stapelfeldt et al. 1998) with a similar configuration as the CS Cha system. To better constrain the mass and properties of the companion and its surrounding disk, additional observational data is neces- sary, in particular, ALMA observations will allow to detect the amount of mm-sized dust around the companion, and likely, re- veal its true nature. Additional SPHERE/ZIMPOL observations that would help to determine the grain size distribution and also potentially if the disk or disk plus envelope scenario explains the system configuration best. Only few other systems are known that harbor a sub-stellar com- panion with a disk around it, such as the FW Tau (Kraus et al. 2014, 2015) system or the 1SWASP J140747.93-394542.6 sys- Article number, page 16 of 19 tem (Mamajek et al. 2012, Kenworthy et al. 2015). The former confirmed by ALMA observations, while the latter was detected in transit. However, the CS Cha system is the only systems in which a circumplanetary disk is likely present as well as a re- solved circumstellar disk. It is also to the best of our knowl- edge the first circumplanetary disk directly detected around a sub-stellar companion in polarized light, constraining its geom- etry. Once the system is well understood it might be considered a benchmark system for planet and brown dwarf formation sce- narios. Acknowledgements. We thank an anonymous referee for significantly improv- ing our original manuscript. We acknowledge I. Pascucci, M. Min, M. Hogerhi- jde, C. Dominik and G. Muro-Arena for interesting discussions. MB acknowl- edges funding from ANR of France under contract number ANR-16-CE31-0013 (Planet Forming Disks). AJ acknowledges the support by the DISCSIM project, grant agreement 341137 funded by the European Research Council under ERC- 2013-ADG. JO acknowledges support from the Universidad de Valparaíso and from ICM Núcleo Milenio de Formación Planetaria, NPF. The research of FS leading to these results has received funding from the European Research Coun- cil under ERC Starting Grant agreement 678194 (FALCONER). SPHERE is an instrument designed and built by a consortium consisting of IPAG (Greno- ble, France), MPIA (Heidelberg, Germany), LAM (Marseille, France), LESIA (Paris, France), Laboratoire Lagrange (Nice, France), INAF - Osservatorio di Padova (Italy), Observatoire de Geneve (Switzerland), ETH Zurich (Switzer- land), NOVA (Netherlands), ONERA (France) and ASTRON (Netherlands) in collaboration with ESO. SPHERE was funded by ESO, with additional contri- butions from CNRS (France), MPIA (Germany), INAF (Italy), FINES (Switzer- land) and NOVA (Netherlands). SPHERE also received funding from the Eu- ropean Commission Sixth and Seventh Framework Programmes as part of the Optical Infrared Coordination Network for Astronomy (OPTICON) under grant number RII3-Ct-2004-001566 for FP6 (2004-2008), grant number 226604 for FP7 (2009-2012) and grant number 312430 for FP7 (2013-2016). This research has made use of the SIMBAD database as well as the VizieR catalogue ac- cess tool, operated at CDS, Strasbourg, France. This research has made use of NASA's Astrophysics Data System Bibliographic Services. Finally CG would like to thank Donna Keeley for language editing of the manuscript. References Allard, F., Hauschildt, P. H., Alexander, D. R., Tamanai, A., & Schweitzer, A. 2001, ApJ, 556, 357 ALMA Partnership, Brogan, C. L., Pérez, L. M., et al. 2015, ApJ, 808, L3 Andrews, S. M., Wilner, D. J., Zhu, Z., et al. 2016, ApJ, 820, L40 Apai, D., Pascucci, I., Wang, H., et al. 2004, in IAU Symposium, Vol. 221, Star Formation at High Angular Resolution, ed. M. G. Burton, R. Jayawardhana, & T. L. Bourke, 307 Appenzeller, I. 1977, A&A, 61, 21 Bailey, V., Meshkat, T., Reiter, M., et al. 2014, ApJ, 780, L4 Banse, K., Crane, P., Grosbol, P., et al. 1983, The Messenger, 31, 26 Bate, M. R. 2012, MNRAS, 419, 3115 Bate, M. R. 2018, MNRAS, 475, 5618 Benisty, M., Juhasz, A., Boccaletti, A., et al. 2015, A&A, 578, L6 Benisty, M., Stolker, T., Pohl, A., et al. 2017, A&A, 597, A42 Bertout, C., Robichon, N., & Arenou, F. 1999, A&A, 352, 574 Beuzit, J.-L., Feldt, M., Dohlen, K., et al. 2008, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 7014, Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, 18 Bonnefoy, M., Chauvin, G., Lagrange, A.-M., et al. 2014, A&A, 562, A127 Brinch, C., Jørgensen, J. K., Hogerheijde, M. R., Nelson, R. P., & Gressel, O. 2016, ApJ, 830, L16 Cambresy, L., Copet, E., Epchtein, N., et al. 1998, A&A, 338, 977 Canovas, H., Ménard, F., de Boer, J., et al. 2015, A&A, 582, L7 Canovas, H., Montesinos, B., Schreiber, M. R., et al. 2018, A&A, 610, A13 Carbillet, M., Bendjoya, P., Abe, L., et al. 2011, Experimental Astronomy, 30, 39 Chabrier, G., Baraffe, I., Allard, F., & Hauschildt, P. 2000, ApJ, 542, 464 Chauvin, G., Desidera, S., Lagrange, A.-M., et al. 2017, A&A, 605, L9 Chauvin, G., Faherty, J., Boccaletti, A., et al. 2012, A&A, 548, A33 Covino, E., Palazzi, E., Penprase, B. E., Schwarz, H. E., & Terranegra, L. 1997, A&AS, 122, 95 Catalog, 2246 Cutri, R. M., Skrutskie, M. F., van Dyk, S., et al. 2003, VizieR Online Data Cutri, R. M., Wright, E. L., Conrow, T., et al. 2012, Explanatory Supplement to the WISE All-Sky Data Release Products, Tech. rep. C. Ginski et al.: First direct detection of a polarized companion outside of a resolved circumbinary disk around CS Cha Diolaiti, E., Bendinelli, O., Bonaccini, D., et al. 2000, A&AS, 147, 335 Dohlen, K., Langlois, M., Saisse, M., et al. 2008, in Proc. SPIE, Vol. 7014, Ground-based and Airborne Instrumentation for Astronomy II, 70143L Dunham, M. M., Offner, S. S. R., Pineda, J. E., et al. 2016, ApJ, 823, 160 Espaillat, C., Calvet, N., D'Alessio, P., et al. 2007, ApJ, 664, L111 Espaillat, C., Furlan, E., D'Alessio, P., et al. 2011, ApJ, 728, 49 Esplin, T. L., Luhman, K. L., Faherty, J. K., Mamajek, E. E., & Bochanski, J. J. Gauvin, L. S. & Strom, K. M. 1992, ApJ, 385, 217 Ginski, C., Neuhäuser, R., Mugrauer, M., Schmidt, T. O. B., & Adam, C. 2013, 2017, AJ, 154, 46 MNRAS, 434, 671 Ginski, C., Schmidt, T. O. B., Mugrauer, M., et al. 2014, MNRAS, 444, 2280 Ginski, C., Stolker, T., Pinilla, P., et al. 2016, A&A, 595, A112 Girard, T. M., van Altena, W. F., Zacharias, N., et al. 2011, AJ, 142, 15 Girardi, L., Barbieri, M., Groenewegen, M. A. T., et al. 2012, Astrophysics and Space Science Proceedings, 26, 165 Girardi, L., Groenewegen, M. A. T., Hatziminaoglou, E., & da Costa, L. 2005, Grady, C. A., Muto, T., Hashimoto, J., et al. 2013, ApJ, 762, 48 Guenther, E. W., Esposito, M., Mundt, R., et al. 2007, A&A, 467, 1147 Guerri, G., Daban, J.-B., Robbe-Dubois, S., et al. 2011, Experimental Astron- A&A, 436, 895 omy, 30, 59 Haisch, Jr., K. E., Lada, E. A., & Lada, C. J. 2001, ApJ, 553, L153 Helling, C., Dehn, M., Woitke, P., & Hauschildt, P. H. 2008, ApJ, 675, L105 Jensen-Clem, R., Millar-Blanchaer, M., Mawet, D., et al. 2016, ApJ, 820, 111 Kenworthy, M. A., Lacour, S., Kraus, A., et al. 2015, MNRAS, 446, 411 Kim, K. H., Watson, D. M., Manoj, P., et al. 2009, ApJ, 700, 1017 Kraus, A. L., Andrews, S. M., Bowler, B. P., et al. 2015, ApJ, 798, L23 Kraus, A. L., Ireland, M. J., Cieza, L. A., et al. 2014, ApJ, 781, 20 Krist, J. E., Hook, R. N., & Stoehr, F. 2011, in Proc. SPIE, Vol. 8127, Optical Modeling and Performance Predictions V, 81270J Krist, J. E., Stapelfeldt, K. R., Ménard, F., Padgett, D. L., & Burrows, C. J. 2000, ApJ, 538, 793 Kuhn, J. R., Potter, D., & Parise, B. 2001, ApJ, 553, L189 Langlois, M., Dohlen, K., Vigan, A., et al. 2014, in Proc. SPIE, Vol. 9147, Ground-based and Airborne Instrumentation for Astronomy V, 91471R Langlois, M., Vigan, A., Moutou, C., et al. 2013, in Proceedings of the Third AO4ELT Conference, ed. S. Esposito & L. Fini, 63 Lenzen, R., Hartung, M., Brandner, W., et al. 2003, in Proc. SPIE, Vol. 4841, Instrument Design and Performance for Optical/Infrared Ground-based Tele- scopes, ed. M. Iye & A. F. M. Moorwood, 944–952 Lillo-Box, J., Barrado, D., & Bouy, H. 2012, A&A, 546, A10 Luhman, K. L., Allen, L. E., Allen, P. R., et al. 2008, ApJ, 675, 1375 Maire, A.-L., Langlois, M., Dohlen, K., et al. 2016, in Proc. SPIE, Vol. 9908, Ground-based and Airborne Instrumentation for Astronomy VI, 990834 Mamajek, E. E., Quillen, A. C., Pecaut, M. J., et al. 2012, AJ, 143, 72 Manara, C. F., Testi, L., Natta, A., et al. 2014, A&A, 568, A18 Martin, F., Conan, R., Tokovinin, A., et al. 2000, A&AS, 144, 39 Martinez, P., Dorrer, C., Aller Carpentier, E., et al. 2009, A&A, 495, 363 Ménard, F., Delfosse, X., & Monin, J.-L. 2002, A&A, 396, L35 Metchev, S. A. & Hillenbrand, L. A. 2006, ApJ, 651, 1166 Miles-Páez, P. A., Zapatero Osorio, M. R., Pallé, E., & Peña Ramírez, K. 2013, Nguyen, D. C., Brandeker, A., van Kerkwijk, M. H., & Jayawardhana, R. 2012, Olofsson, J., van Holstein, R. G., Boccaletti, A., et al. 2018, ArXiv e-prints A&A, 556, A125 ApJ, 745, 119 [arXiv:1804.01929] Pascucci, I., Ricci, L., Gorti, U., et al. 2014, ApJ, 795, 1 Pérez, L. M., Carpenter, J. M., Andrews, S. M., et al. 2016, Science, 353, 1519 Pohl, A., Sissa, E., Langlois, M., et al. 2017, A&A, 605, A34 Pollack, J. B., Hubickyj, O., Bodenheimer, P., et al. 1996, Icarus, 124, 62 Ribas, Á., Bouy, H., Merín, B., et al. 2016, MNRAS, 458, 1029 Rousset, G., Lacombe, F., Puget, P., et al. 2003, in Proc. SPIE, Vol. 4839, Adap- tive Optical System Technologies II, ed. P. L. Wizinowich & D. Bonaccini, 140–149 Sauvage, J.-F., Fusco, T., Petit, C., et al. 2016, Journal of Astronomical Tele- scopes, Instruments, and Systems, 2, 025003 Schmid, H. M., Joos, F., & Tschan, D. 2006, A&A, 452, 657 Schmidt, T. O. B., Neuhäuser, R., Briceño, C., et al. 2016, A&A, 593, A75 Schmidt, T. O. B., Neuhäuser, R., Seifahrt, A., et al. 2008, A&A, 491, 311 Sengupta, S. & Krishan, V. 2001, ApJ, 561, L123 Serkowski, K., Mathewson, D. S., & Ford, V. L. 1975, ApJ, 196, 261 Smart, R. L. & Nicastro, L. 2014, A&A, 570, A87 Stapelfeldt, K. R., Krist, J. E., Ménard, F., et al. 1998, ApJ, 502, L65 Stolker, T., Min, M., Stam, D. M., et al. 2017a, ArXiv e-prints [arXiv:1706.09427] Stolker, T., Sitko, M., Lazareff, B., et al. 2017b, ApJ, 849, 143 Takeda, G. & Rasio, F. A. 2005, ApJ, 627, 1001 Tamura, M. & Sato, S. 1989, AJ, 98, 1368 Teixeira, R., Ducourant, C., Sartori, M. J., et al. 2000, A&A, 361, 1143 Trauger, J. T. & WFPC2 Science Team. 1994, in Bulletin of the American Astro- nomical Society, Vol. 26, American Astronomical Society Meeting Abstracts #184, 892 van Boekel, R., Henning, T., Menu, J., et al. 2017, ApJ, 837, 132 van der Plas, G., Wright, C. M., Ménard, F., et al. 2017, A&A, 597, A32 van Holstein, R. G., Snik, F., Girard, J. H., et al. 2017, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 10400, Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, 1040015 Vogt, N., Schmidt, T. O. B., Neuhäuser, R., et al. 2012, A&A, 546, A63 Voirin, J., Manara, C. F., & Prusti, T. 2017, ArXiv e-prints [arXiv:1710.04528] Wagner, K., Apai, D., Kasper, M., & Robberto, M. 2015, ApJ, 813, L2 Walsh, C., Juhász, A., Meeus, G., et al. 2016, ApJ, 831, 200 Weingartner, J. C. & Draine, B. T. 2001, ApJ, 548, 296 Whittet, D. C. B., Prusti, T., Franco, G. A. P., et al. 1997, A&A, 327, 1194 Wilking, B. A., Lebofsky, M. J., & Rieke, G. H. 1982, AJ, 87, 695 Wu, Y.-L., Close, L. M., Males, J. R., et al. 2015, ApJ, 801, 4 Yudin, R. V. 2000, A&AS, 144, 285 Zacharias, N., Monet, D. G., Levine, S. E., et al. 2004, in Bulletin of the Ameri- can Astronomical Society, Vol. 36, American Astronomical Society Meeting Abstracts, 1418 Zapatero Osorio, M. R., Caballero, J. A., & Béjar, V. J. S. 2005, ApJ, 621, 445 Article number, page 17 of 19 A&A proofs: manuscript no. CSCha-aa_revised Appendix A: Stellar PSF in the SPHERE polarimetric images Fig. B.1: Color-color plot of YSOs in Cha from Cambresy et al. (1998). For comparison we added the companion to CS Cha in the plot as blue cross (marking the uncertainties of the photome- try). The companions colors are much bluer in I-J than expected for a YSO. In J-K they are still very blue compared to most objects. Overall we can conclude that the companion does not match YSO colors well and is in all likelihood no embedded background YSO. Fig. A.1: Left Column: Total intensity images derived from our polarimetric observations with SPHERE in J and H-band. The circumstellar disk is not visible in these images since flux is com- pletely dominated by the bright stellar PSF. The visible "cavity" in the image center is caused by the SPHERE/IRDIS corona- graph. The stellar PSF is strongly elongated in the South-East to North-West direction in the H-band observations. Additionally the star is not well centered behind the coronagraph, as is vis- ible in the illumination pattern behind the (slightly transparent) coronagraph. Right Column: Qφ images from Figure 1 divided by the total intensity images in the left column. Note that this is not equivalent with polarization degree, since the intensity im- ages are dominated by the stellar PSF rather than disk signal. The visible bright lobes are aligned with the semi-minor axis of the stellar PSF, i.e. they are caused by a drop off in stellar flux and not by the polarized phase function of the disk scattered light. It is visible that the assymmetric scattered light signal in the H-band is less apparent in this image, indicating that it origi- nates from the asymmetric flux in the stellar PSF which was not well centered behind the coronagraph, rather than from an actual astrophysical asymmetry in scattered light from the disk. Appendix B: Color comparison of the companion with YSOs in Cha Appendix C: Model of a circum-companion disk irradiated by the central binary star Article number, page 18 of 19 0.50.00.5J−bandIQφ/I0.50.00.50.50.00.5IH−band0.50.00.5Qφ/I∆Dec[arcsec]∆RA[arcsec]1.01.52.02.53.03.54.04.55.05.5J−K[mag]0.00.51.01.52.02.53.03.54.04.5I−J[mag]YSO in Cha from Cambresy et al. 1998C. Ginski et al.: First direct detection of a polarized companion outside of a resolved circumbinary disk around CS Cha Fig. C.1: Photometry and degree of linear polarization of the companion assuming the central object is of low enough mass that its thermal emission is negligible. This very low mass ob- ject is still surrounded by a circum-companion disk. Thus the light received from the companion is entirely scattered light from the primary stars. We used different grain sizes and inclinations of the circum-companion disk, indicated by line style and color. We find that, while we can explain the degree of linear polar- ization with such a model, the received flux is several order of magnitude below our measurements. Article number, page 19 of 19 0.51.03.0λ[µm]10-2010-1910-1810-1710-1610-1510-14λFλ[W/m2]0.1µm,i=0◦0.1µm,i=45◦0.1µm,i=67◦1.0µm,i=0◦1.0µm,i=45◦1.0µm,i=67◦0.51.03.0λ[µm]020406080100p[%]
1911.02814	1	1911	2019-11-07T09:14:18	NGTS-8b and NGTS-9b: two non-inflated hot-Jupiters	[ "astro-ph.EP", "astro-ph.SR" ]	We report the discovery, by the Next Generation Transit Survey (NGTS), of two hot-Jupiters NGTS-8b and NGTS-9b. These orbit a V = 13.68 K0V star (Teff = 5241 +/- 50 K) with a period of 2.49970 days, and a V = 12.80 F8V star (Teff = 6330 +/- 130 K) in 4.43527 days, respectively. The transits were independently verified by follow-up photometric observations with the SAAO 1.0-m and Euler telescopes, and we report on the planetary parameters using HARPS, FEROS and CORALIE radial velocities. NGTS-8b has a mass, 0.93 +0.04 -0.03 MJ and a radius, 1.09 +/- 0.03 RJ similar to Jupiter, resulting in a density of 0.89 +0.08 -0.07 g cm-3. This is in contrast to NGTS-9b, which has a mass of 2.90 +/- 0.17 MJ and a radius of 1.07 +/- 0.06 RJ , resulting in a much greater density of 2.93 +0.53 -0.49 g cm-3. Statistically, the planetary parameters put both objects in the regime where they would be expected to exhibit larger than predicted radii. However, we find that their radii are in agreement with predictions by theoretical non-inflated models.	astro-ph.EP	astro-ph	MNRAS 000, 1 -- 12 (2019) Preprint 11 November 2019 Compiled using MNRAS LATEX style file v3.0 NGTS-8b and NGTS-9b: two non-inflated hot-Jupiters Jean C. Costes,12(cid:63) Christopher A. Watson,12 Claudia Belardi,5 Ian P. Braker,5 Matthew R. Burleigh,5 Sarah L. Casewell,5 Philipp Eigmuller,6 Maximil- ian N. Gunther,3,15,16 James A. G. Jackman,1,2 Louise D. Nielsen,4 Mar- itza G. Soto,13 Oliver Turner,4 David R. Anderson,1,2 Daniel Bayliss,1,2 Fran¸cois Bouchy,4 Joshua T. Briegal,3 Edward M. Bryant,1,2 Juan Cabrera,6 Alexander Chaushev,10 Szilard Csizmadia,6 Anders Erikson,6 Samuel Gill,1,2 Edward Gillen,3 Michael R. Goad,5 Matthew J. Hooton,12 James S. Jenkins,8,9 James McCormac,1,2 Maximiliano Moyano,14 Didier Queloz,3 Heike Rauer,6,10,11 Liam Raynard,5 Alexis M. Smith,6 Andrew P. G. Thompson,12 Rosanna H. Tilbrook,5 Stephane Udry,4 Jose I. Vines,8 Richard G. West,1,2 Peter J. Wheatley1,2 1Centre for Exoplanets and Habitability, University of Warwick, Gibbet Hill Road, Coventry CV4 7AL, UK 2Dept. of Physics, University of Warwick, Gibbet Hill Road, Coventry CV4 7AL, UK 3Astrophysics Group, Cavendish Laboratory, J.J. Thomson Avenue, Cambridge CB3 0HE, UK 4Observatoire de Gen`eve, Universit´e de Gen`eve, 51 Ch. des Maillettes, 1290 Sauverny, Switzerland 5Department of Physics and Astronomy, University of Leicester, University Road, Leicester LE1 7RH, UK 6Institute of Planetary Research, German Aerospace Center, Rutherfordstrasse 2, 12489 Berlin, Germany 7Institute of Astronomy, Cambridge University, Madingley Road, Cambridge CB3 0HA, UK 8Departamento de Astronomia, Universidad de Chile, Casilla 36-D, Santiago, Chile 9 Centro de Astrof´ısica y Tecnolog´ıas Afines (CATA), Casilla 36-D, Santiago, Chile. 10Center for Astronomy and Astrophysics, TU Berlin, Hardenbergstr. 36, D-10623 Berlin, Germany 11Institute of Geological Sciences, FU Berlin, Malteserstr. 74-100, D-12249 Berlin, Germany 12Astrophysics Research Centre, School of Mathematics and Physics, Queen's University Belfast, BT7 1NN Belfast, UK 13School of Physics and Astronomy, Queen Mary University of London, 327 Mile End Road, E1 4NS, UK 14Instituto de Astronom´ıa, Universidad Cat´olica del Norte, Angamos 0610, 1270709 Antofagasta, Chile 15Department of Physics, and Kavli Institute for Astrophysics and Space Research, Massachusetts Institute of Technology, Cambridge, MA 02139, USA 16Juan Carlos Torres Fellow S. Accepted 2019 Nov 04. Received 2019 Oct 08; in original form 2019 Aug 05 ABSTRACT We report the discovery, by the Next Generation Transit Survey (NGTS), of two hot-Jupiters NGTS-8b and NGTS-9b. These orbit a V = 13.68 K0V star (Te f f = 5241 ± 50 K) with a period of 2.49970 days, and a V = 12.80 F8V star (Te f f = 6330 ± 130 K) in 4.43527 days, respectively. The transits were independently verified by follow-up photometric observations with the SAAO 1.0-m and Euler telescopes, and we report on the planetary parameters using HARPS, FEROS and CORALIE radial velocities. NGTS-8b has a mass, 0.93 +0.04 −0.03 MJ and a radius, 1.09 ± 0.03 RJ similar to Jupiter, resulting in a density of 0.89 +0.08 −0.07 g cm−3. This is in contrast to NGTS-9b, which has a mass of 2.90 ± 0.17 MJ and a radius of 1.07 ± 0.06 RJ , resulting in a much greater density of 2.93+0.53 −0.49 g cm−3. Statistically, the planetary parameters put both objects in the regime where they would be ex- pected to exhibit larger than predicted radii. However, we find that their radii are in agreement with predictions by theoretical non-inflated models. Key words: techniques: photometric, stars: individual: NGTS-8 and NGTS-9 plan- etary systems, planets and satellites: detection 9 1 0 2 v o N 7 . ] P E h p - o r t s a [ 1 v 4 1 8 2 0 . 1 1 9 1 : v i X r a © 2019 The Authors 2 J. C. Costes et al. 1 INTRODUCTION Hot-Jupiters are giant gas exoplanets similar to Jupiter, but with a shorter orbital period, inferior to 10 days. While rare, these planets are the easiest to detect from ground-based surveys due to their relatively deep transits (∼ 1%), their large radial velocity (RV) signals, and their short periods, which make hot-Jupiters important targets in order to un- derstand the structure, composition and evolution of plane- tary systems. From the currently observed population of exoplanets with known radii, masses and orbital distances, the evo- lution of planetary radii has been modelled (e.g. Fortney et al. 2007; Baraffe et al. 2008). These models, where the effects of stellar irradiation and heavy element cores are in- cluded, agree with observations at low stellar irradiation. However, the observed radii of highly irradiated gas giants are discrepant with theoretical expectations. For instance, at fluxes greater than 2×105 W m−2 (Miller & Fortney 2011; Demory & Seager 2011) the gas giants are increasingly found with anomalously large radii. This is the case for the hot- Jupiters WASP-17 b, WASP-121 b and Kepler-435 b, which all have measured radii R > 1.8 RJ (Anderson et al. 2011; Al- menara et al. 2015; Delrez et al. 2016). A number of possi- ble mechanisms have been postulated to explain these in- flated planetary radii including kinetic heating (Guillot & Showman 2002), enhanced atmospheric opacities (Burrows et al. 2007), double diffusive convection (Chabrier & Baraffe 2007), Ohmic heating through magnetohydrodynamic ef- fects (Batygin & Stevenson 2010; Perna et al. 2010; Wu & Lithwick 2012; Ginzburg & Sari 2016), tidal dissipation (Bodenheimer et al. 2001; Bodenheimer et al. 2003; Arras & Socrates 2010; Jermyn et al. 2017) and vertical advection of potential temperature (Youdin & Mitchell 2010; Tremblin et al. 2017). However, the exact mechanisms responsible are still, as yet, unidentified and the problem remains unsolved. In order to perform robust statistical studies of hot-Jupiter radii and constrain the dominant 'inflation' mechanisms at work (e.g. as done by Sestovic et al. 2018) we need to in- crease the sample of planets spanning a range of planetary masses, radii, stellar irradiation levels, as well as planetary system ages. In this paper we present two hot-Jupiters that appear to be non-inflated, despite being highly irradiated with an incident flux greater than 2 × 105 W m−2 (like many inflated planets). In §2, the NGTS discovery data is described. §3 explains the photometric follow-up campaigns and §4 re- ports the mass determination via RV monitoring from spec- troscopy. §5 details the analysis of the stellar parameters, presents the stellar activity and its relation with the stellar rotation and shows the global modelling process to charac- terize the planets. §6 presents an investigation regarding the incident flux, the planetary mass and the radius. Finally we finish with our conclusions in §7. 2 DISCOVERY PHOTOMETRY FROM NGTS The Next Generation Transit Survey (NGTS), operating since early 2016, is a wide-field transit survey located at ESO's Paranal Observatory in Chile, whose primary goal is to discover Neptune-sized or bigger exoplanets. NGTS has a fully robotized array of twelve 20 cm Newtonian tele- scopes, and each telescope is equipped with 2K×2K e2V deep-depleted Andor IKon-L CCD cameras with 13.5 µm pixels and an instantaneous field of view of 8 deg2. For a description of this facility and its capabilities, optimised for detecting planets, we refer the reader to Wheatley et al. (2018). NGTS has already detected 4 hot-Jupiters: NGTS- 1b (Bayliss et al. 2018), NGTS-2b (Raynard et al. 2018), NGTS-3Ab (Gunther et al. 2018) and NGTS-6b (Vines et al. 2019). Here we report the latest hot-Jupiter discoveries from NGTS: NGTS-8b and NGTS-9b. NGTS-8 was observed using a single NGTS camera (#811) over a 227 night baseline between the 21st of April 2016 and the 3rd of December 2016. NGTS-9 was also ob- served using a single camera (#806) over a 234 night base- line between the 8th of October 2016 and the 29th of May 2017. A total of 177 799 and 167 933 images were obtained, respectively, each with an exposure time of 10 s. These data were taken using the custom NGTS filter (520 -- 890 nm) (Wheatley et al. 2018) and the telescope was auto-guided using an improved version of the DONUTS auto-guiding al- gorithm (McCormac et al. 2013). The data were reduced and aperture photometry was extracted using the CASU- Tools1 photometry package. A total of 177120 and 166043 valid data-points were extracted from the raw images and then de-trended for nightly trends, such as atmospheric ex- tinction, using our implementation of the SysRem algorithm (Tamuz et al. 2005). Both datasets were searched for transit-like signals us- ing orion, an optimized implementation of the box-least- squares (BLS) fitting algorithm (Collier Cameron et al. 2006). A 1.6% deep transit signal was detected at a pe- riod of 2.49970 days for the K0V star, NGTS-8, and a 0.6% deep transit signal at 4.43527 days for the F8V star, NGTS- 9. These periods were distinguished from other aliases using the photometry and spectroscopy follow-up -- see Section 3 and Section 4 for details. The detrended NGTS data for the two stars, phase-folded on the planetary orbital periods, are shown in Figure 1 and Figure 2. A sample of the NGTS re- duced photometric measurements are presented in Table 1 and Table 2, with the full data available electronically from the journal. The NGTS data were searched for signs that would in- dicate that the planetary candidates were false positives. No evidence for a secondary eclipse or out-of-transit varia- tions indicating an eclipsing binary system were identified in the NGTS light curves of the two stars. However, for both sources, some stars were found to be in close proximity to our targets. Using Gaia we confirmed that these nearby stars did not appreciably dilute the light from NGTS-8 or NGTS-9, and also confirmed that NGTS-8 and NGTS-9 were not gi- ants stars -- see Section 5.1.1 for details. Based on the NGTS detection, NGTS-8 and NGTS-9 were followed-up with fur- ther photometry and spectroscopy to confirm the planetary nature of the system and to measure the planetary param- eters, which we report on in the next section. 1 http://casu.ast.cam.ac.uk/surveys-projects/ software-release MNRAS 000, 1 -- 12 (2019) NGTS-8b and NGTS-9b 3 Figure 1. From top to bottom: Figure 1.a represents the NGTS discovery light curve of NGTS-8b with residuals, phase-folded on the orbital period and zoomed on the transit. Figure 1.b represnts the ingress and mid-transit of NGTS-8b observed with SAAO, with residuals. Figure 1.c represents the ingress and mid transit of NGTS-8b observed with Euler, with residuals. For all the plots, the blue data points are binned every 7 min to aid visualisation. The red lines show 20 light curve models generated from randomly drawn posterior samples of the allesfitter fit. Figure 2. From top to bottom: Figure 2.a represents the NGTS discovery light curve of NGTS-9b with residuals, phase-folded on the orbital period and zoomed on the transit. Figure 2.b rep- resents the full transit of NGTS-9b observed with SAAO, with residuals. This image uses night 20181221, where ingress and mid- transit can be seen, and night 20190103, with mid-transit and egress. For both figures, the blue data points are binned every 10 min to aid visualisation. Figure 2.c represents the full transit of NGTS-9b observed with Euler, with residuals. The data was taken on a 100sec cadence, shown in blue. For all the plots, the red lines show 20 light curve models generated from randomly drawn posterior samples of the allesfitter fit. MNRAS 000, 1 -- 12 (2019) 4 J. C. Costes et al. Table 1. A sample of the photometric data of NGTS-8 from NGTS, SAAO and Eulercam. The full dataset is available elec- tronically from the journal. Time Relative (BJD-2450000) flux Flux error Filter Instrument 7499.8655 7499.8657 7499.8658 7499.8660 7499.8661 7499.8663 7499.8664 7499.8666 7499.8667 7499.8669 ... 1.0179 1.0320 1.0016 0.9964 0.9911 0.9995 0.9892 1.0054 0.9910 0.9901 ... 0.0221 NGTS 0.0221 NGTS 0.0220 NGTS 0.0220 NGTS 0.0220 NGTS 0.0220 NGTS 0.0220 NGTS 0.0220 NGTS 0.0219 NGTS 0.0219 NGTS ... ... NGTS NGTS NGTS NGTS NGTS NGTS NGTS NGTS NGTS NGTS ... Table 2. A sample of the photometric data of NGTS-9 from NGTS, SAAO and Eulercam. The full dataset is available elec- tronically from the journal. Time Relative (BJD-2450000) flux Flux error Filter Instrument 7669.8612 7669.8614 7669.8617 7669.8618 7669.8620 7669.8621 7669.8623 7669.8624 7669.8626 7669.8627 ... 0.9871 1.0027 0.9919 0.9848 1.0107 1.0014 0.9956 1.0167 1.0076 1.0000 ... 0.0087 NGTS 0.0087 NGTS 0.0087 NGTS 0.0087 NGTS 0.0087 NGTS 0.0087 NGTS 0.0087 NGTS 0.0087 NGTS 0.0087 NGTS 0.0087 NGTS ... ... NGTS NGTS NGTS NGTS NGTS NGTS NGTS NGTS NGTS NGTS ... 3 FOLLOW-UP PHOTOMETRY 3.1 SAAO photometry Follow-up photometry of NGTS-8 was obtained with the 1.0 m Elizabeth telescope at the South African Astronomi- cal Observatory (SAAO) on 2017 July 17 and 2017 July 18, utilising the frame-transfer CCD Sutherland High-speed Op- tical Camera "SHOC'n'awe" (Coppejans et al. 2013, SHOC). With a pixel scale of 0.167 arcsec/pixel, the SHOC cam- eras on the 1 m telescope were binned 4 × 4 pixels in the X and Y directions. The field of view of its instruments is 2.85(cid:48)×2.85(cid:48). This allow to observe simultaneously the target and a comparison star of similar brightness for differential photometry. The data, obtained using a z(cid:48) filter with an ex- posure of 30 s, were bias and flat field corrected. This was performed in python using the standard procedure with the CCDPROC package (Craig et al. 2015). Then, using the 'SEP' package (Barbary 2016), the aperture photometry of both the target and the comparison star were extracted. Fi- nally, the sky background was measured and substracted using the SEP background map. We also obtained two follow-up light curves of NGTS-9 on 2018 December 21 and 2019 January 30, with the same telescope and instrument set-up as described above. This time the data were obtained with an I filter and an ex- posure time of 20 s. The data were reduced with the local SAAO SHOC pipeline, which is driven by python scripts running iraf tasks (pyfits and pyraf), and incorporating the usual bias and flat-field calibrations. Aperture photom- etry was performed using the Starlink package autopho- tom. For the first observation of NGTS-9 we used a 4 pixel radius aperture that maximised the signal/noise, and the background was measured in an annulus surrounding this aperture with inner and outer radii of 7 and 9 pixels, respec- tively. Two comparison stars were then used to perform dif- ferential photometry on the target. The 2019 January 30 ob- servation was obtained in slightly poorer seeing conditions, and we therefore utilised a 6 pixel aperture, a correspond- ingly larger background annulus, and only one comparison star for differential photometry. The transits of these two exoplanets observed from SAAO are shown in Figures 1.b and 2.b. Regarding NGTS-8, only a partial transit was observed with SAAO. For NGTS- 9, 2 nights were combined: night 20181221, where ingress and mid-transit were seen, and night 20190103, where mid- transit and egress were seen. While only partial transits were observed, these SAAO data were able to confirm the tran- sits and the consistency of the transit depths and were used to revise the orbital ephemerides for subsequent follow-up observations. In particular, the 2017 July observations of NGTS-8 were helpful in confirming the orbital period and ruling out aliases of similar power in the original NGTS data. 3.2 Eulercam We also observed both objects with Eulercam (Lendl et al. 2012) on the 1.2 m Euler Telescope at La Silla Observatory. NGTS-8 was observed on the 21st of August 2017, 502 expo- sures were acquired using the Cousins-I filter, an exposure time of 12 s and a defocus of 0.05 mm. NGTS-9 was observed on the 12th of January 2019. We acquired 134 images using the Gunn-R filter, a 100 s exposure time and no defocus. For both target, their data were reduce using the standard procedure of bias subtraction and flat field correction. The aperture photometry as well as x- and y-position, FWHM, airmass and sky background of the target star were extracted using the PyRAF implementation of the phot routine. The comparison stars and the photometric aperture radius were carefully chosen in order to reduce the RMS in the scatter out of transit. Using both follow-up photometry for the 2 stars, we can conclude that the nearby stars did not blend with the two targets. The Euler data for the two stars are shown in Figures 1.c and 2.c. Regarding NGTS-8, only a partial transit was observed with Euler. Concerning NGTS-9, the full transit was observed, with some systematics that were removed using a Gaussian process. Again, these data confirm the transits as we see the ingress or the full transit around the predicted times. 4 SPECTROSCOPY 4.1 NGTS-8 NGTS-8 was observed with the HARPS spectrograph (Mayor et al. 2003) on the ESO 3.6 m telescope at La Silla Observatory, Chile, between the 5th of August 2017 and the MNRAS 000, 1 -- 12 (2019) NGTS-8b and NGTS-9b 5 Table 3. A summary of the follow-up photometry of NGTS-8 and NGTS-9. Night Instrument Target Nimages 20170717 20170718 20170821 20181221 20190103 20190112 Shoc'n'awe NGTS-8b Shoc'n'awe NGTS-8b Eulercam NGTS-8b Shoc'n'awe NGTS-9b Shoc'n'awe NGTS-9b Eulercam NGTS-9b 676 606 502 550 648 134 Exptime (seconds) 30 60 12 20 20 100 Binning (X×Y) 4 × 4 4 × 4 1 × 1 4 × 4 4 × 4 1 × 1 Filter Comment z' z' IC I I RG shown in Figure 1.b no transit observed shown in Figure 1.c shown in Figure 2.b shown in Figure 2.b shown in Figure 2.c 28th of October 2017 under programmes 099.C-0303 and 0100.C-0474. We used the high-efficiency mode, EGGS, due to the faintness of the host star and large expected RV ampli- tude. The exposure times for each spectrum varied between 1800 and 1200 s resulting in a signal-to-noise (SNR), mea- sured around 550 nm, of 10-15 per exposure. The standard HARPS data reduction software (DRS) was used to mea- sure the RVs of NGTS-8 at each epoch. This was done via cross-correlation with a K0 binary mask. Three additional spectra were obtained with FEROS (Kaufer & Pasquini 1998), mounted on the MPG 2.2 m tele- scope at La Silla Observatory, Chile, on the 20th and 21st of August 2017. All spectra were obtained with an expo- sure time of 1800 s, and the data were reduced using the FEROS routine of the CERES pipeline (Brahm et al. 2017). CERES also performed a radial velocity extraction, by cross- correlating the spectra with a G2 binary mask. The resulting SNR of the spectra, taken around 550 nm per resolution el- ement, was around 45. The RVs from both HARPS and FEROS are listed in Table 4, along with their associated error, FWHM and bi- sector span. While not presented in this table, the error on the BIS and on the FWHM were calculated using the same standard treatment as done previously (West et al. 2018). The errors for the BIS and for the FWHM are set to equal twice the error and 2.3548 times the error on the RV, re- spectively. The RV measurements of NGTS-8, shown phase folded in Figure 3, are in phase with the period detected by orion, with a semi-amplitude of K = 149.95 ± 3.56 m s−1. This indicates a transiting planet with the mass of a hot- Jupiter. No evidence of a correlation between the RVs and the measured bisector spans or FWHMs were found, with a Spearman correlation of -0.05 and -0.21, respectively. Thus, the RV signal does not originate from cool stellar spots or a blended eclipsing binary (Queloz et al. 2001b). 4.2 NGTS-9 NGTS-9 was observed with the CORALIE spectrograph (Queloz et al. 2001a) on the 1.2 m Euler telescope at La Silla Observatory, Chile, between the 24th of December 2017 and the 5th of April 2018. Exposure times of either 1800 or 2700 s were used depending on seeing and general observing con- ditions at the time. RVs were calculated with a G2 binary mask using the standard data reduction pipelines. Initial analysis, shown phase folded in Figure 4, confirm the plan- etary nature of the candidate with a mass of a hot-Jupiter. The RV variations are in phase with the period detected by orion with a semi-amplitude of K = 293.44 ± 15.08 m s−1. Correlations between RVs and bisector span and FWHM MNRAS 000, 1 -- 12 (2019) Figure 3. Phase folded radial velocity data in km s−1 and resid- uals from HARPS, in blue, and FEROS, in orange, for NGTS-8. The red lines show 50 light curve models generated from randomly drawn posterior samples of the allesfitter fit. were also checked but no evidence for any such correlations was found, with a Spearman correlation of -0.05 and -0.15, respectively. The CORALIE RVs for NGTS-9 are listed in Table 5, along with their associated error, FWHM and bi- sector span. 5 ANALYSIS 5.1 Stellar Properties 5.1.1 Gaia To obtain astrometric information for NGTS-8 and NGTS-9 we crossmatched both sources with Gaia DR2. To check the quality of the astrometric solutions we calculated the unit weight error (UWE) and then renormalised UWE (RUWE). We find that both sources pass the filters recommended by the Gaia team (RUWE < 1.4, see Lindegren et al. 2018, for a discussion on the recommended UWE filters). Along with this, the two targets also have zero astrometric noise, giving us confidence that they are both single sources without evi- dence of unresolved binarity. They also pass the photometric 6 J. C. Costes et al. Table 4. HARPS and FEROS radial velocities for NGTS-8. JDB (-2400000) RV (km s−1) RV error (km s−1) FWHM (km s−1) BIS (km s−1) 57970.7400 57979.7893 57980.7591 57986.8002 57986.8147 57987.8182 57993.6958 57994.6689 57998.7907 58023.6064 58025.5985 58026.7251 58052.5954 58054.6266 15.4538 15.7466 15.4692 15.3768 15.4527 15.1913 15.5568 15.7472 15.5861 15.5259 15.4713 15.7423 15.6232 15.7407 0.0141 0.0084 0.0096 0.0147 0.0147 0.0108 0.0089 0.0093 0.0109 0.0082 0.0057 0.0091 0.0107 0.0073 7.0583 6.9506 6.9007 10.4776 10.4080 10.5177 6.9108 6.9062 6.9314 6.9028 6.9379 6.8863 6.9310 6.9014 -0.1398 -0.0254 -0.0021 0.0000 0.0360 -0.0620 0.0092 -0.0238 -0.0221 -0.0128 -0.0243 -0.0141 -0.0558 -0.0074 Exptime Instrument (s) 1800 1800 1800 1800 1800 1800 1800 1800 1200 1800 1800 1800 1200 1200 HARPS HARPS HARPS FEROS FEROS FEROS HARPS HARPS HARPS HARPS HARPS HARPS HARPS HARPS Table 5. CORALIE radial velocities for NGTS-9. JDB (-2400000) RV (km s−1) RV error (km s−1) FWHM (km s−1) BIS (km s−1) 58111.8084 58113.7093 58118.6806 58129.8234 58169.7427 58172.5402 58195.5527 58201.7069 58202.7184 58207.5924 58208.5990 58209.5025 58210.4972 58211.5079 58212.5487 58214.4983 35.4749 36.0010 36.0550 35.5313 35.4951 36.0431 35.6563 35.6593 36.0204 36.0663 35.7431 35.5571 35.7017 36.0788 36.0007 35.5283 0.0369 0.0411 0.0418 0.0702 0.0448 0.0497 0.0450 0.0735 0.0382 0.0410 0.0561 0.0513 0.0624 0.0566 0.0556 0.0796 11.7616 11.9519 12.0847 11.8283 12.0594 12.0437 11.8628 12.1195 11.9054 11.7546 11.8260 12.0882 12.1724 11.8893 12.0946 12.3596 -0.2434 0.0337 0.0198 0.0679 0.0809 -0.0176 -0.0829 0.0072 -0.0995 -0.0344 -0.0614 0.0690 0.0087 0.2033 -0.1138 0.1707 Exptime Instrument (s) 2700 2700 2700 1800 2700 2700 2700 1800 2700 2700 2700 2700 1800 1800 1800 1800 CORALIE CORALIE CORALIE CORALIE CORALIE CORALIE CORALIE CORALIE CORALIE CORALIE CORALIE CORALIE CORALIE CORALIE CORALIE CORALIE filters specified by Arenou et al. (2018) to identify blended stars. With the Gaia information for each source, we calcu- late the absolute magnitude and plot their positions on the Hertzsprung-Russell diagram in Figure 5, where we can see that both NGTS-8 and NGTS-9 lie in the region expected for single main sequence stars. 5.1.2 SPECIES The stacked spectra for both targets were also analyzed us- ing SPECIES (Soto & Jenkins 2018), a python tool to derive stellar parameters in an automated fashion from high reso- lution echelle spectra. By measuring the equivalent widths (EWs) for a list of irons lines, and by using the ATLAS9 model atmospheres (Castelli & Kurucz 2004), SPECIES first solves the radiative transfer equation using MOOG (Sneden 1973). From an iterative process, SPECIES derives then the atmospheric parameters (Teff, log g, [Fe/H]) of our target. By interpolation through a grid of MIST isochrones (Dotter 2016), the mass and radius are estimated, using a Bayesian approach. This method delivers an estimate of the age of the system as well. However, due to the fact that solar-type stars spend most of their lives in the evolutionary stage and because the dependence of their effective temperature and luminosity with the age of the system is weak, the estima- tion of the age of the system is very unconstrained for main sequence stars. Finally, SPECIES derives the rotational and macroturbulent velocities from the stellar temperature and by line-fitting to a set of five absorption lines. Parameters found by SPECIES for both targets are displayed in Table 6 and Table 7. 5.2 Stellar Activity and Rotation on NGTS-8 In addition to modeling the stellar parameters, we also at- tempted to search for stellar activity and rotation signals. As mentioned earlier, the out-of-eclipse light curves of both targets show no appreciable variability, and there are no correlations with the measured RVs and the bisector or the FWHM for the two targets -- see Section 4. Nonetheless, de- termining the stellar rotation period, along with knowledge of the stellar radius and v sin i∗, can enable the inclination angle of the stellar rotation axis to be constrained. This can MNRAS 000, 1 -- 12 (2019) NGTS-8b and NGTS-9b 7 Figure 5. HR diagram using Gaia DR2 absolute magnitude. The K0V star, NGTS-8, is shown in red and the F8V, NGTS-9 in light blue. star of the same spectral type as NGTS-8 and artificially broadening it by different v sin i∗ amounts (using a Gray ro- tational broadening profile). The projected rotational broad- ening of NGTS-8 was then measured using an optimal- subtraction technique in which the broadened template spec- tra were multiplied by a constant and then subtracted from the NGTS-8 spectrum. This is done after correcting for ra- dial velocity shifts and re-interpolating to a constant veloc- ity scale. The value of the rotational broadening is then the one that minimises the scatter in the residual spectrum af- ter performing the optimal subtraction. For our template spectrum, we used α Cen B, which has a spectral type very close to NGTS-8 and a low rotation rate. We constructed the template spectrum by stacking archival HARPS spectra taken over 1 night when α Cen B was known to be inactive. The result of this analysis yielded a value consistent with that found by SPECIES. Adopting the firm upper-limit on the stellar rotation pe- riod from the v sin i∗ measurement would ordinarily lead to a much higher log R(cid:48) H K level than the one observed. While no definitive answer can explain this difference, some sys- tems hosting hot-Jupiters are known to have suppressed Ca II H & K re-emission (e.g. WASP-12 -- Fossati et al. 2013), leading to a lower measured value of the log R(cid:48) H K . However, these systems generally contain hot-Jupiters very close to filling their Roche lobes, which is not the case for our planet, NGTS-8b. We therefore conclude that the most likely expla- nation of the discrepancy is that we have caught NGTS-8 in an extended low-activity state. 5.3 Global Modelling Analysis of the different photometric and spectroscopic data was performed on NGTS-8 and NGTS-9 data using alles- Figure 4. Phase folded radial velocity data in km s−1 and resid- uals from CORALIE for NGTS-9. The red lines show 50 light curve models generated from randomly drawn posterior samples of the allesfitter fit. enable misaligned star-planet systems to be identified (Wat- son et al. 2010). In order to put constraints on the stellar rotation pe- riod, two methods were used. The first one consists of us- ing the activity of the star. Using the formulae described in Lovis et al. (2011), the log R(cid:48) H K was measured for each individual HARPS spectrum. Since the log R(cid:48) H K of NGTS- 9, an F8V star, was not measurable, we will only focus on the K0V star, NGTS-8, in this section. The calculated value of the log R(cid:48) H K varied between −4.643 and −5.053 in the in- dividual spectra due to a low signal-to-noise ratio in the blue band, with a mean value of −4.783. In order to increase the precision of the measured data, a stacked spectrum of all the spectra was created. Figure 6 shows this spectrum, zoomed on the H (3933.664 A) and K (3968.470 A) bands, represented with dashed lines. Using this spectrum, we found a value of −4.817 ± 0.110 for the log R(cid:48) H K . Finally, using the relation from Noyes et al. (1984), the stellar rotation period of NGTS-8 was derived to be 37.7 ± 4.1 days. Our second approach used the v sin i∗ of the star mea- sured from SPECIES and assumed that the stellar inclination angle, i∗, was 90◦ (i.e. sin(i∗) = 1). This then enables an up- per limit to be placed on the stellar rotation period when the stellar radius is known (e.g. Watson et al. 2010). An upper limit of 13.92±2.64 days was found, which is discrepant with our previous result by almost 5 sigma and would rule out the long rotation period inferred from the log R(cid:48) H K measure- ment. Even adopting the most extreme individual log R(cid:48) H K measurement implies a stellar rotation period greater than 26 days. Given this discrepancy, we decided to verify the v sin i∗ value measured by SPECIES with another technique. We did this by taking a stellar spectrum of a slowly rotating MNRAS 000, 1 -- 12 (2019) 8 J. C. Costes et al. Table 6. Stellar Properties for NGTS-8. Table 7. Stellar Properties for NGTS-9. Property Value Source Property Value Source Astrometric Properties R.A. Dec 2MASS I.D. Gaia source I.D. µR.A. (mas y−1) µDec. (mas y−1) parallax (mas) 21h55m54.s2 −14◦04(cid:48)05.(cid:48)(cid:48)85 6840435777723109888 21555419-1404062 21.363 ± 0.047 −10.194 ± 0.048 2.3027 ± 0.0299 Photometric Properties V (mag) B (mag) g (mag) r (mag) i (mag) G (mag) GRP (mag) GBP (mag) J (mag) H (mag) K (mag) W1 (mag) W2 (mag) W3 (mag) 13.68 ± 0.06 14.59 ± 0.03 14.12 ± 0.03 13.43 ± 0.06 13.21 ± 0.06 13.4954 ± 0.0003 12.8780 ± 0.0006 13.9606 ± 0.0015 12.14 ± 0.02 11.75 ± 0.02 11.64 ± 0.02 11.59 ± 0.02 11.62 ± 0.02 11.86 ± 0.38 Derived Properties Spectral type Teff (K) [Fe/H] v sin i∗ (km s−1) vmac (km s−1) log g Ms (M(cid:12)) Rs (R(cid:12)) Age (Gyrs) Distance (pc) K0V 5241 ± 50 0.24 ± 0.09 3.56 ± 0.67 1.49 ± 0.64 4.41 ± 0.03 0.89 +0.05 −0.04 0.98 ± 0.02 12.48 +3.23 −3.68 434.273 ± 5.639 2MASS 2MASS 2MASS Gaia DR2 Gaia DR2 Gaia DR2 Gaia DR2 APASS APASS APASS APASS APASS Gaia DR2 Gaia DR2 Gaia DR2 2MASS 2MASS 2MASS WISE WISE WISE Gaia DR2 SPECIES SPECIES SPECIES SPECIES SPECIES SPECIES SPECIES SPECIES Gaia DR2 Astrometric Properties R.A. Dec 2MASS I.D. Gaia source I.D. µR.A. (mas y−1) µDec. (mas y−1) parallax (mas) 09h27m41.s0 −19◦20(cid:48)50.(cid:48)(cid:48)33 5678340222972504832 09274096-1920515 −6.078 ± 0.057 1.723 ± 0.063 1.6136 ± 0.0416 Photometric Properties V (mag) B (mag) g (mag) r (mag) i (mag) G (mag) GRP (mag) GBP (mag) J (mag) H (mag) K (mag) W1 (mag) W2 (mag) W3 (mag) 12.80 ± 0.02 13.36 ± 0.04 13.03 ± 0.05 12.65 ± 0.02 12.55 ± 0.07 12.6547 ± 0.0002 12.2157 ± 0.0013 12.9503 ± 0.0015 11.71 ± 0.03 11.49 ± 0.02 11.45 ± 0.02 11.39 ± 0.02 11.42 ± 0.02 11.58 ± 0.20 Derived Properties Spectral type Teff (K) [Fe/H] v sin i∗ (km s−1) vmac (km s−1) log g Ms (M(cid:12)) Rs (R(cid:12)) Age (Gyrs) Distance (pc) F8V 6330 ± 130 0.31 ± 0.15 6.38 ± 1.05 5.47 ± 1.05 4.37 ± 0.20 1.34 ± 0.05 1.38 ± 0.04 0.96 ± 0.60 619.732 ± 15.977 2MASS 2MASS 2MASS Gaia DR2 Gaia DR2 Gaia DR2 Gaia DR2 APASS APASS APASS APASS APASS Gaia DR2 Gaia DR2 Gaia DR2 2MASS 2MASS 2MASS WISE WISE WISE Gaia DR2 SPECIES SPECIES SPECIES SPECIES SPECIES SPECIES SPECIES SPECIES Gaia DR2 2MASS (Skrutskie et al. 2006); APASS (Henden & Munari 2014); WISE (Wright et al. 2010); Gaia DR2 (Gaia Collaboration et al. 2018) 2MASS (Skrutskie et al. 2006); APASS (Henden & Munari 2014); WISE (Wright et al. 2010); Gaia DR2 (Gaia Collaboration et al. 2018) fitter (Gunther & Daylan 2019, and in prep.). allesfit- ter is a user-friendly and publicly available software package for modeling data from photometric and RV instruments. Its generative model can account for multi-star systems, stellar flares, star spots and multiple exoplanets. For this, it constructs an inference framework that unites the ver- satile packages ellc (light curve and RV models; Maxted 2016), aflare (flare model; Davenport et al. 2014), dynesty (nested samplingl; Speagle 2019), emcee (MCMC sampling; Foreman-Mackey et al. 2013), and celerite (GP models; Foreman-Mackey et al. 2017). allesfitter is accesible at https://github.com/MNGuenther/allesfitter. For NGTS-8 and NGTS-9, the Nested Sampling ap- proach (see Skilling 2004) was used, which enables simul- taneous fitting of the transit light curves and radial velocity data. In particular, we fit for the following astrophysical pa- rameters: a planet's orbital period P, the transit epoch TC , the radius ratio Rp/R(cid:63), the sum of radii over the semi-major axis (Rp + R(cid:63))/a, the cosine of the inclination cos i, the eccen- e sin ω tricity and argument of periastron parameterized as e cos ω, and the RV semi-amplitude K. For the transit and √ √ We find that NGTS-8b has a mass of 0.93 +0.04 light curve modeling, a quadratic limb-darkening law was adopted parameterized after Kipping (2013) as u1 and u2. Systematic trends in the transit light curves were modeled by a Gaussian process with Matern 3/2 kernels parameter- ized by the GP's amplitude ln ρ and time scale ln σ. For both planets, all photometric data were used for the fits as well as all spectroscopic data, with instrumental offsets taken into account, relevant for NGTS-8 where HARPS and FEROS data were combined for the modelling. −0.03 MJ and a radius 1.09 ± 0.03 RJ , while NGTS-9b has a mass of 2.90 ± 0.17 MJ and a radius 1.07 ± 0.06 RJ . The results of the fits for the two planets are summarized in Tables 8 and 9 and shown in earlier plots. Figure 1 shows (in red) 20 light curve models generated from randomly drawn posterior samples of the allesfitter fit to the NGTS, SAAO and Euler light curves, respectively, for NGTS-8. In the same way, Figure 2 shows the photometric data of NGTS, SAAO and Euler, respectively, for NGTS-9, with (in red) 20 light curve mod- els generated from randomly drawn posterior samples of the allesfitter fit. For the RV data, Figure 3 shows the mod- MNRAS 000, 1 -- 12 (2019) NGTS-8b and NGTS-9b 9 Figure 7. TESS light curve of NGTS-9b with residuals. The blue data points are binned every 7 min to aid visualisation. The red lines show 20 light curve models generated from randomly drawn posterior samples of the allesfitter fit. Table 8. Planetary properties for NGTS-8b using allesfitter. Property Value P (days) TC (BJD) T14 (hours) a/R∗ R/R∗ K (m s−1) e i (degrees) Mp (MJ ) Rp (RJ ) ρp (g cm−3) a (AU) Teq (K) 2.49970 ± 0.00001 2457500.17830 ± 0.00072 2.61 ± 0.06 7.60 ± 0.18 0.114 ± 0.002 149.95 ± 3.56 0.010 +0.014 −0.010 86.9 ± 0.5 0.93 +0.04 −0.03 1.09 ± 0.03 0.89 +0.08 −0.07 0.035 ± 0.001 1345 ± 19 Table 9. Planetary properties for NGTS-9b using allesfitter. Property Value P (days) TC (BJD) T14 (hours) a/R∗ R/R∗ K (m s−1) e i (degrees) Mp (MJ ) Rp (RJ ) ρp (g cm−3) a (AU) Teq (K) 4.43527 ± 0.00002 2457671.81086 ± 0.00265 2.05 ± 0.07 9.06 ± 0.31 0.080 ± 0.004 293.44 ± 15.08 0.060+0.076 −0.052 84.1 ± 0.4 2.90 ± 0.17 1.07 ± 0.06 2.93+0.53 −0.49 0.058+0.003 −0.002 1448 ± 36 Figure 6. Stacked spectrum of NGTS-8 from HARPS, zoomed on the H (3933.664 A) and K (3968.470 A) bands, represented with dashed lines. elling of HARPS, blue points, and FEROS, orange points, for NGTS-8 and Figure 4 shows the modelling of the CORALIE data for NGTS-9. In order to check our results, we also performed another analysis of the photometric data from NGTS and available spectroscopic data on NGTS-8 and NGTS-9 using the EX- Oplanet traNsits and rAdIal veLocity fittER (EXO-NAILER -- Espinoza et al. 2016). Using the Markov chain Monte Carlo (MCMC) with a total of 250 walkers for 20000 jumps and 5000 burn-in steps, the modelling was done assuming pure white-noise for the inputted light curves. A logarithmic limb- darkening law was adopted with limb-darkening coefficients taken from Claret et al. (2013), and sampled according to Es- pinoza & Jord´an (2015). The results from this second anal- ysis all agreed, within the error bars, with those found from allesfitter. 5.4 TESS During the preparation of this manuscript TESS photome- try was released for NGTS-9, which was observed in Sector 8. In response to this data release, we re-analysed, using all available data, NGTS-9 with allesfitter. The TESS photometric data is presented in Figure 7 with the mod- els generated from the fit. We confirmed that using TESS data in our modelling of NGTS-9 did not change or improve the values obtained, and thus the TESS data was not taken into account in the analysis presented in this work. The fact that TESS does not improve the results can be explained by the magnitude of NGTS-9, V = 12.80 ± 0.02. At these magnitudes we have found that NGTS and TESS perform similarly (Wheatley et al. 2018). No TESS data is available for NGTS-8. MNRAS 000, 1 -- 12 (2019) 10 J. C. Costes et al. 6 DISCUSSION As outlined in the introduction, at incident fluxes greater than 2×105 W m−2 (Miller & Fortney 2011; Demory & Seager 2011), hot-Jupiters are increasingly found with radii that are significantly larger than theoretically predicted (Anderson et al. 2011; Delrez et al. 2016; Almenara et al. 2015). Using Gaia DR2 measurements for the stellar luminosity and the orbital parameters listed in Tables 6 and 8 for NGTS-8b and listed in Tables 7 and 9 for NGTS-9b, we calculated the flux received by both planets to be greater than this limit (6.85 ± 0.45 × 105 W m−2 and 9.92 ± 1.09 × 105 W m−2 for NGTS-8b and NGTS-9b, respectively). Thus, the stellar irradiation levels received by both of these planets puts them firmly in the regime where we might expect them to exhibit larger than predicted planetary radii. Sestovic et al. (2018) conducted a statistical investiga- tion on hot-Jupiter radii and found that above a threshold in incident flux (2 × 105 W m−2 Miller & Fortney 2011; De- mory & Seager 2011), the observed radius follow the thermal evolution models (Miller & Fortney 2011; Thorngren et al. 2016) with the addition of an inflation parameter, ∆R. This observed radius 'inflation' is dependent on both the inci- dent stellar flux and the mass of the planet. They proposed a flux-mass-radius relationship that has distinct forms for 4 different planetary mass regimes: below 0.37 MJ , between 0.37 -- 0.98 MJ , between 0.98 -- 2.50 MJ and over 2.50 MJ . Using these relationships, we calculated what would be the expected radius inflation (∆R) values for our two planets. For NGTS-8b, its mass lies on the edge of two regimes in Sestovic et al. (2018) (Mp < 0.98 MJ and Mp > 0.98 MJ ), we thus de- termined predicted ∆Rs from both relationships of 0.24±0.02 and 0.02 ± 0.01 RJ , respectively. While the first value would suggest a highly inflated radius, the second value however suggests almost no inflation. Concerning NGTS-9b, the pre- dicted radius inflation, ∆R, is 0.18± 0.01 RJ . Thus, from the work of Sestovic et al. (2018), these two planets would be expected to exhibit planetary radii larger than predicted. We finally compared the observed planetary radii of NGTS-8b and NGTS-9b to the mass-radius models of Baraffe et al. (2008) and Fortney et al. (2007) who present, assuming a solar-type star, tables of planetary radii as a function of core mass, mass of the planet, orbital separation and age of the system. Since neither of the host stars of the planets presented here are solar-like, we had to renormalise the orbital separation in order to keep the same incident flux. In this scenario, the distance from their host star would be equal to 0.044 AU and 0.038 AU for NGTS-8b and NGTS-9b, respectively. As one can see in Tables 10 and 11, both models seem consistent and correctly predict the measured radius of NGTS-8b, 1.09 ± 0.03 RJ , and NGTS-9b, 1.07 ± 0.06 RJ , using the described parameters. To conclude, even if both planets are in a regime where we expect planets to exhibit larger than predicted radii, our two planets are non-inflated hot-Jupiters. This could be due to the planets being enriched with heavy elements, yielding a more compact structure and thus a smaller radius, like HD 149026b (Sato et al. 2005). Figure 8. Shows NGTS-8b and NGTS-9b in the planetary radius vs planetary mass plot, in regards with all exoplanets from the NASA Exoplanet Archive with radius uncertainty below 10% or mass uncertainty below 50% and with an incident flux received by the planets greater than 2 × 105 W m−2. The background and the dotted black lines represents the point density per grid element. 7 CONCLUSIONS We have presented the latest discovery by the Next Gen- eration Transit Survey (NGTS) of two non-inflated hot- Jupiters: NGTS-8b and NGTS-9b. NGTS, SAAO and Eu- ler photometric data and spectroscopic data from HARPS, FEROS and CORALIE were used to confirm the detection of these two planets. By combining some of these data, an anal- ysis of the transiting planets was performed using allesfit- ter and confirmed with EXO-NAILER. From this model, both planets have orbits consistent with being circular, as ex- pected for such short period hot-Jupiters. The characteristic of the planets were calculated such as: NGTS-8b with a mass −0.03 MJ and a radius 1.09 ± 0.03 RJ , and NGTS-9b of 0.93 +0.04 with a mass of 2.90 ± 0.17 MJ and a radius of 1.07 ± 0.06 RJ . Figure 8 shows these discoveries in comparison to known planets with radius higher than 0.4 RJ . A study of the rotational period of the K0V star, NGTS- 8, was performed using different models and despite a signif- icant discrepancy that we assume is due to an extended low activity of the star, we measured an upper limit of 13.92±2.64 days. While its host is considerably fainter, our analysis also suggests that its planet, NGTS-8b, could have similar prop- erties to HD 189733b, one of the best studied hot-Jupiters. Further observations of NGTS-8b will allow direct compar- isons to be drawn between these two hot-Jupiters. The up- coming launch of JWST will enable high-precision obser- vations of NGTS-8b's full-phase curve, which would be of particular interest due to the efficient dayside to nightside heat recirculation that HD 189733b exhibits relative to other hot-Jupiters (Knutson et al. 2007; Schwartz et al. 2017). Concerning NGTS-9b, the planet is highly irradiated, with an incident flux around 9.59 ± 0.74 × 105 W m−2, yet non-inflated. This radius could be due to the planet being extremely enriched with heavy elements, explaining its den- sity, 2.93+0.53 with similar masses, as shown in Figure 8. −0.49 g cm−3, one of the highest compare to planets MNRAS 000, 1 -- 12 (2019) NGTS-8b and NGTS-9b 11 Table 10. A summary of the mass-radius model for NGTS-8b with radius 1.09 ± 0.03 RJ . Model Mass of the planet (MJ ) Orbital Age of the separation (AU) system (Gyrs) Mass fraction of heavy material Core mass (%) Radius (RJ ) Baraffe et al. (2008) Fortney et al. (2007) 1 1 0.045 0.045 8.93 - 10.00 0.02 - 0.1 4.5 0 - 25 1.025 - 1.074 1.050 - 1.107 Table 11. A summary of the mass-radius model for NGTS-9b with radius 1.07 ± 0.06 RJ . Model Mass of the planet (MJ ) Orbital Age of the separation (AU) system (Gyrs) Mass fraction of heavy material Core mass (%) Radius (RJ ) Baraffe et al. (2008) Fortney et al. (2007) 2 - 5 2.44 0.045 0.045 0.30 - 1.78 0.30 - 1 0.02 - 0.1 0 - 100 1.063 - 1.177 1.065 - 1.199 ACKNOWLEDGEMENTS Based on data collected under the NGTS project at the ESO La Silla Paranal Observatory. The NGTS facility is operated by the consortium institutes with support from the UK Science and Technology Facilities Council (STFC) project ST/M001962/1. This paper uses observations made at the South African Astronomical Observatory (SAAO). We thank Marissa Kotze (SAAO) for developing the SHOC camera data reduction pipeline. The contributions at the University of Warwick by PJW, RGW, DLP, DJA, BTG and TL have been supported by STFC through consoli- dated grants ST/L000733/1 and ST/P000495/1. Contribu- tions at the University of Geneva by DB, FB, BC, LM, and SU were carried out within the framework of the Na- tional Centre for Competence in Research "PlanetS" sup- ported by the Swiss National Science Foundation (SNSF). The contributions at the University of Leicester by MRG and MRB have been supported by STFC through consol- idated grant ST/N000757/1. CAW acknowledges support from the STFC grant ST/P000312/1. EG gratefully ac- knowledges support from Winton Philanthropies in the form of a Winton Exoplanet Fellowship. JSJ acknowledges sup- port by Fondecyt grant 1161218 and partial support by CATA-Basal (PB06, CONICYT). DJA gratefully acknowl- edges support from the STFC via an Ernest Rutherford Fellowship (ST/R00384X/1). PE and HR acknowledge the support of the DFG priority program SPP 1992 "Explor- ing the Diversity of Extrasolar Planets" (RA 714/13-1). LD acknowledges support from the Gruber Foundation Fellow- ship. MNG acknowledges support from MIT's Kavli Insti- tute as a Torres postdoctoral fellow. The research leading to these results has received funding from the European Re- search Council under the FP/2007-2013 ERC Grant Agree- ment number 336480 and from the ARC grant for Concerted Research Actions, financed by the Wallonia-Brussels Feder- ation. This work was also partially supported by a grant from the Simons Foundation (PI Queloz, ID 327127). This work has made use of data from the European Space Agency (ESA) mission Gaia (https://www.cosmos.esa.int/gaia), processed by the Gaia Data Processing and Analysis Con- sortium (DPAC, https://www.cosmos.esa.int/web/gaia/ dpac/consortium). Funding for the DPAC has been pro- vided by national institutions, in particular the institutions participating in the Gaia Multilateral Agreement. PyRAF is a product of the Space Telescope Science Institute, which is MNRAS 000, 1 -- 12 (2019) operated by AURA for NASA. This research has made use of the NASA Exoplanet Archive, which is operated by the California Institute of Technology, under contract with the National Aeronautics and Space Administration under the Exoplanet Exploration Program. REFERENCES Almenara J., et al., 2015, Astronomy & Astrophysics, 575, A71 Anderson D., et al., 2011, Monthly Notices of the Royal Astro- nomical Society, 416, 2108 Arenou F., et al., 2018, A&A, 616, A17 Arras P., Socrates A., 2010, The Astrophysical Journal, 714, 1 Baraffe I., Chabrier G., Barman T., 2008, A&A, 482, 315 Barbary K., 2016, SEP: Source Extractor as a library, The Jour- nal of Open Source Software, doi:10.21105/joss.00058 Batygin K., Stevenson D. J., 2010, The Astrophysical Journal Letters, 714, L238 Bayliss D., et al., 2018, MNRAS, 475, 4467 Bodenheimer P., Lin D., Mardling R., 2001, The Astrophysical Journal, 548, 466 Bodenheimer P., Laughlin G., Lin D. N., 2003, The Astrophysical Journal, 592, 555 Brahm R., Jord´an A., Espinoza N., 2017, Publications of the As- tronomical Society of the Pacific, 129, 034002 Burrows A., Hubeny I., Budaj J., Hubbard W., 2007, The Astro- physical Journal, 661, 502 Castelli F., Kurucz R. L., 2004, ArXiv Astrophysics e-prints, Chabrier G., Baraffe I., 2007, The Astrophysical Journal Letters, 661, L81 Claret A., Hauschildt P. H., Witte S., 2013, A&A, 552, A16 Collier Cameron A., et al., 2006, MNRAS, 373, 799 Coppejans R., et al., 2013, Publications of the Astronomical So- ciety of the Pacific, 125, 976 Craig M. W., et al., 2015, ccdproc: CCD data reduction software, Astrophysics Source Code Library (ascl:1510.007) Davenport J. R. A., et al., 2014, ApJ, 797, 122 Delrez L., et al., 2016, Monthly Notices of the Royal Astronomical Society, 458, 4025 Demory B.-O., Seager S., 2011, ApJS, 197, 12 Dotter A., 2016, ApJS, 222, 8 Espinoza N., Jord´an A., 2015, MNRAS, 450, 1879 Espinoza N., et al., 2016, The Astrophysical Journal, 830, 43 Foreman-Mackey D., Hogg D. W., Lang D., Goodman J., 2013, PASP, 125, 306 Foreman-Mackey D., Agol E., Ambikasaran S., Angus R., 2017, celerite: Scalable 1D Gaussian Processes in C++, Python, and Julia, Astrophysics Source Code Library (ascl:1709.008) 12 J. C. Costes et al. Fortney J. J., Marley M. S., Barnes J. W., 2007, The Astrophys- ical Journal, 659, 1661 Fossati L., Ayres T. R., Haswell C. A., Bohlender D., Kochukhov O., Floer L., 2013, ApJ, 766, L20 Gaia Collaboration Brown A. G. A., Vallenari A., Prusti T., de Bruijne J. H. J., Babusiaux C., Bailer-Jones C. A. L., 2018, preprint, (arXiv:1804.09365) Ginzburg S., Sari R., 2016, The Astrophysical Journal, 819, 116 Guillot T., Showman A. P., 2002, Astronomy & Astrophysics, 385, 156 Gunther M. N., Daylan T., 2019, allesfitter: Flexible star and exoplanet inference from photometry and radial velocity (ascl:1903.003) Gunther M. N., et al., 2018, MNRAS, 478, 4720 Henden A., Munari U., 2014, Contributions of the Astronomical Observatory Skalnate Pleso, 43, 518 Jermyn A. S., Tout C. A., Ogilvie G. I., 2017, Monthly Notices of the Royal Astronomical Society, 469, 1768 Kaufer A., Pasquini L., 1998, in D'Odorico S., ed., Proc. SPIEVol. 3355, Optical Astronomical Instrumentation. pp 844 -- 854, doi:10.1117/12.316798 Kipping D. M., 2013, MNRAS, 435, 2152 Knutson H. A., et al., 2007, Nature, 447, 183 Lendl M., et al., 2012, A&A, 544, A72 Lindegren L., et al., 2018, A&A, 616, A2 Lovis C., et al., 2011, arXiv e-prints, p. arXiv:1107.5325 Maxted P. F. L., 2016, A&A, 591, A111 Mayor M., et al., 2003, The Messenger, 114, 20 McCormac J., Pollacco D., Skillen I., Faedi F., Todd I., Watson C. A., 2013, PASP, 125, 548 Miller N., Fortney J. J., 2011, The Astrophysical Journal Letters, 736, L29 Noyes R. W., Hartmann L. W., Baliunas S. L., Duncan D. K., Vaughan A. H., 1984, ApJ, 279, 763 Perna R., Menou K., Rauscher E., 2010, The Astrophysical Jour- nal, 724, 313 Queloz D., et al., 2001a, The Messenger, 105, 1 Queloz D., et al., 2001b, A&A, 379, 279 Raynard L., et al., 2018, MNRAS, 481, 4960 Sato B., et al., 2005, The Astrophysical Journal, 633, 465 Schwartz J. C., Kashner Z., Jovmir D., Cowan N. B., 2017, ApJ, 850, 154 Sestovic M., Demory B.-O., Queloz D., 2018, A&A, 616, A76 Skilling J., 2004, in AIP Conference Proceedings. pp 395 -- 405 Skrutskie M. F., et al., 2006, AJ, 131, 1163 Sneden C. A., 1973, PhD thesis, THE UNIVERSITY OF TEXAS AT AUSTIN. Soto M. G., Jenkins J. S., 2018, A&A, 615, A76 Speagle J. S., 2019, arXiv e-prints, p. 1904.02180 Tamuz O., Mazeh T., Zucker S., 2005, MNRAS, 356, 1466 Thorngren D. P., Fortney J. J., Murray-Clay R. A., Lopez E. D., 2016, ApJ, 831, 64 Tremblin P., et al., 2017, The Astrophysical Journal, 841, 30 Vines J. I., et al., 2019, arXiv e-prints, p. arXiv:1904.07997 Watson C., Littlefair S., Cameron A. C., Dhillon V., Simpson E., 2010, Monthly Notices of the Royal Astronomical Society, 408, 1606 West R. G., et al., 2018, arXiv e-prints, p. arXiv:1809.00678 Wheatley P. J., et al., 2018, MNRAS, 475, 4476 Wright E. L., et al., 2010, AJ, 140, 1868 Wu Y., Lithwick Y., 2012, The Astrophysical Journal, 763, 13 Youdin A. N., Mitchell J. L., 2010, The Astrophysical Journal, 721, 1113 This paper has been typeset from a TEX/LATEX file prepared by the author. MNRAS 000, 1 -- 12 (2019)
1308.0607	1	1308	2013-08-02T20:00:34	New developments for modern celestial mechanics. I. General coplanar three-body systems. Application to exoplanets	[ "astro-ph.EP" ]	Modern applications of celestial mechanics include the study of closely packed systems of exoplanets, circumbinary planetary systems, binary-binary interactions in star clusters, and the dynamics of stars near the galactic centre. While developments have historically been guided by the architecture of the Solar System, the need for more general formulations with as few restrictions on the parameters as possible is obvious. Here we present clear and concise generalisations of two classic expansions of the three-body disturbing function, simplifying considerably their original form and making them accessible to the non-specialist. Governing the interaction between the inner and outer orbits of a hierarchical triple, the disturbing function in its general form is the conduit for energy and angular momentum exchange and as such, governs the secular and resonant evolution of the system and its stability characteristics. Focusing here on coplanar systems, the first expansion is one in the ratio of inner to outer semimajor axes and is valid for all eccentricities, while the second is an expansion in eccentricity and is valid for all semimajor axis ratios [...]. Our generalizations make both formulations valid for arbitrary mass ratios. [...]. We demonstrate the equivalence of the new expansions, identifying the role of the spherical harmonic order m in both and its physical significance in the three-body problem, and introducing the concept of principal resonances. Several examples of the accessibility of both expansions are given including resonance widths and the secular rates of change of the elements. Results in their final form are gathered together at the end of the paper for the reader mainly interested in their application, including a guide for the choice of expansion.	astro-ph.EP	astro-ph	Mon. Not. R. Astron. Soc. 000, 1–45 (2013) Printed 12 December 2013 (MN LATEX style file v2.2) New developments for modern celestial mechanics. I. General coplanar three-body systems. Application to exoplanets Rosemary A. Mardling12⋆ 1School of Mathematical Sciences, Monash University, Victoria, 3800, Australia 2Astronomy Department of the University of Geneva, Geneva Observatory, 51 Ch des Mail lettes, CH1290 Versoix Accepted ... Received ...; in original form ... ABSTRACT Modern applications of celestial mechanics include the study of closely packed sys- tems of exoplanets, circumbinary planetary systems, binary-binary interactions in star clusters, and the dynamics of stars near the galactic centre. While developments have historically been guided by the architecture of the Solar System, the need for more general formulations with as few restrictions on the parameters as possible is obvi- ous. Here we present clear and concise generalisations of two classic expansions of the three-body disturbing function, simplifying considerably their original form and making them accessible to the non-specialist. Governing the interaction between the inner and outer orbits of a hierarchical triple, the disturbing function in its general form is the conduit for energy and angular momentum exchange and as such, governs the secular and resonant evolution of the system and its stability characteristics. Focusing here on coplanar systems, the first expansion is one in the ratio of inner to outer semima jor axes and is valid for all eccentricities, while the second is an expansion in eccentricity and is valid for all semima jor axis ratios, except for systems in which the orbits cross (this restriction also applies to the first expansion). Our generalizations make both formulations valid for arbitrary mass ratios. The classic versions of these appropriate to the restricted three- body problem are known as Kaula’s expansion and the literal expansion respectively. We demonstrate the equivalence of the new expansions, identifying the role of the spherical harmonic order m in both and its physical significance in the three-body problem, and introducing the concept of principal resonances. Several examples of the accessibility of both expansions are given including res- onance widths and the secular rates of change of the elements. Results in their final form are gathered together at the end of the paper for the reader mainly interested in their application, including a guide for the choice of expansion. Key words: celestial mechanics – methods: analytical – chaos – planets and satellites: dynamical evolution and stability 1 INTRODUCTION The recent explosion in the study of exoplanets is bringing together specialists from diverse backgrounds. On the observational side, spectroscopic and photometric techniques for binary star observations have been refined to the point where radial velocity ⋆ E-mail: [email protected] c(cid:13) 2013 RAS 2 Mard ling and eclipse contrast measurements at the 50 cm s−1 and 10−4 level respectively are possible (with the HARPS spectrograph and the Kepler satellite; Pepe et al. 2011; Dumusque et al. 2012; Borucki et al. 2011; Batalha et al. 2012), while the field is breathing new life into observational techniques ranging from direct imaging (Marois et al. 2008, 2010; Kalas et al. 2008; Lagrange et al. 2009), to infrared and even x-ray photometry and spectroscopy (Richardson et al. 2006; Fortney & Marley 2007; Pillitteri et al. 2010). As a direct consequence of the ability to make high-precision estimates of the masses and radii of transiting planets, as well as to observe their atmospheres directly (Seager & Deming 2010, and references therein), geo- physical and atmospheric scientists are finding a place in this booming field, with new centres for planet characterization and habitability springing up around the world. On the theoretical side, the discovery of 51 Peg b in its 4.6 day orbit (Mayor & Queloz 1995) immediately reinvigo- rated existing theories of planet formation, especially the question of the role of planet migration which until then (and still) had specialists wondering why the giant planets in the Solar System apparently migrated so little (Lin & Papaloizou 1979; Goldreich & Tremaine 1979; Lin et al. 1996). The discovery of significantly eccentric exoplanets (Holman et al. 1997; Naef et al. 2001) and the detection of misaligned and even retrograde planetary orbits (Winn et al. 2010; Triaud et al. 2010) revived ideas about the secular evolution of inclined systems (Kozai 1962; Kiseleva et al. 1998; Fabrycky & Tremaine 2007), as well as dynamical interactions resulting in scattering during the formation process (Rasio & Ford 1996). The discovery of dynamically packed multiplanet systems (Lovis et al. 2011; Lissauer et al. 2011) reminds us of the age-old quest to under- stand the stability of the Solar System (Laskar 1996), and indeed as Poincar´e well appreciated (Poincar´e 1892), the quest to understand stability in the “simpler” three-body problem (Wisdom 1980; Robutel 1995; Mardling 2008, 2013). Large surveys have yielded a rich harvest of planets by now (Udry & Santos 2007), enabling a comparison with the results of Monte Carlo-type simulations of the planet-formation process (Mordasini et al. 2009). It is now becoming clear that the planet mass distribution continues to rise towards lower masses with a hint of a deficit at 30M⊕ (Mayor et al. 2011), and with almost no ob jects in the range 25 − 45MJ at the high-mass end of the distribution (Sahlmann et al. 2011). The latter is the so-called brown dwarf desert and it supports the idea that at least two distinct mechanisms operate for the formation of planets and stars (Udry & Santos 2007). Until very recently, the notion that planets might form in a circumbinary disk was purely theoretical,1 with no clear concensus on whether or not the strong fluctuating gravitational field of the binary pair would prevent their formation (Meschiari 2012a,b; Paardekooper et al. 2012; Pelupessy & Portegies Zwart 2012). The Kepler survey has revealed that Nature does, indeed, accomplish this feat (Doyle et al. 2011; Welsh et al. 2012; Orosz et al. 2012a,b), with planet-binary period ratios almost as low as they can be stability-wise (the current range is 5.6 for Kepler 16 (Doyle et al. 2011) to 10.4 for Kepler 34 (Welsh et al. 2012)). With our understanding of the growth of planetesimals in a relatively laminar environment still in its infancy (Meisner et al. 2012; Okuzumi et al. 2012; Takeuchi & Ida 2012), the very fact that planets exist in such “hostile” environments puts strong constraints on how and where planetesimals form. At every step of the way, some knowledge of celestial mechanics and the dynamics of small-N systems is essential. A short list might include using stability arguments to place limits on the masses in a multi-planetary system observed spec- trocopically (Mayor et al. 2009); including dynamical constraints in orbit-fitting algorithms (Lovis et al. 2011; Laskar et al. 2012); modelling the planet-planet interaction in a resonant system to infer the presence of a low-mass planet (Rivera et al. 2005; Correia et al. 2010); using transit timing variations (TTVs) to infer the presence of unseen companions (Torres et al. 2011; Ballard et al. 2011; Nesvorn´y et al. 2012) or to estimate the masses of planets in a multi-transiting system (where no spectroscopic data is available: Holman et al. 2010; Lissauer et al. 2011; Steffen et al. 2012; Fabrycky et al. 2012); under- standing the origin of resonant and near resonant systems (Papaloizou 2011; Lithwick & Wu 2012; Batygin & Morbidelli 2012, Mardling & Udry in preparation); deciphering the complex light curves of multi-transiting systems (Lissauer et al. 2011) and eclipsing binaries with circumbinary planets (Orosz et al. 2012a); inferring the true orbit of a planet observed astrometrically (McArthur et al. 2010); understanding the influence of stellar and planetary tides on a system which includes a short-period planet (Wu & Goldreich 2002; Mardling 2007, 2010); inferring information about the internal structure of such a planet (Wu & Goldreich 2002; Batygin et al. 2009; Mardling 2010). Most of these examples involve orbit-orbit interactions in sys- tems with arbitrary mass ratios, eccentricities and inclinations, and as such many researchers resort to expensive (time-wise) direct integrations. Moreover, numerical studies offer shrouded insight into parameter dependence and physical processes. The availability of a generalised and easy-to-use disturbing function therefore seems timely; this is the focus of the present paper. 1 Unless one considers the HD 202060 system to be in this category; it is composed of an inner pair with masses 1.13M⊙ and 0.017M⊙ , and an outer planet with mass 2.44MJ (Correia et al. 2005). c(cid:13) 2013 RAS, MNRAS 000, 1–45 General coplanar systems 3 1.1 The disturbing function As the interaction potential for a hierarchically arranged triple system of bodies, the disturbing function (la fonction per- turbatrice of Le Verrier (1855)) is developed as a Fourier series expansion whose harmonic angles are linear combinations of the various orbital phase and orientation angles in the problem (except for the inclinations), and whose coefficients are functions of the masses, semima jor axes, eccentricities and inclinations. With the Solar System as the only planetary system known until 1992 (Wolszczan 1992), many expansions of the disturbing function from Laplace to Le Verrier (1855) and on (see Murray & Dermott (2000) for other references) rely on the smallness of eccentricities, inclinations and mass ratios, with ranges for Solar System planets being 0.007 − 0.093, 0o − 3.4o and 3.2 × 10−7 − 10−3 respectively (not including Mercury and Pluto). Moreover, since the period ratios between adjacent pairs of planets are not very large, varying between 1.5 and 2.5 (including the mean period of the asteroids), such expansions tend to place no restriction on the ratio of semima jor axes except that the orbits should not cross. These include the literal expansions which are written in terms of Laplace coefficients (Murray & Dermott 2000, and Appendix C of this paper). They are often presented in somewhat long and unwieldy form (for example, the paper by Le Verrier (1855) is 84 pages long), resulting in the reluctance of non-specialists to use them. Moreover, such presentations often obscure the dependence of the disturbing function on the various system parameters, making their use somewhat doubtful in an era where Morse’s law still holds. Many elegant Hamiltonian formulations exist, for example, Laskar & Robutel (1995) which uses Poincar´e canonical heliocentric variables to study the planetary three-body problem. The Hamiltonian form is used, for example, when one wishes to exploit the integrability of the underlying subsystems in the case that they are weakly interacting. The aim of Laskar & Robutel (1995) was to use this form to study the stability of the planetary three-body problem (Robutel 1995) in the framework of the famous KAM theorem (Kolmogorov 1954; Arnol’d 1963; Moser 1962). While the Laskar & Robutel (1995) formulation does not make any assumptions about the mass ratios, their dependence is not explicit in the resulting expressions and it is hard to get a feel for the dependence on the eccentricities and inclinations. Moreover, it has the drawback that except for the inner-most orbit, ellipses described by the orbital elements are not tangential to the actual motion (they are not “osculating”) because the dominant body is used as the origin. In contrast, Jacobi coordinates which refer the motion of a body to the centre of mass of the sub-system to its interior ensure that this is true. Jacobi coordinates are used here. The study of systems with more substantial eccentricities and inclinations, for example comet orbits perturbed by the giant planets, or stellar triples and quaduples etc., has largely relied on numerical integrations and continues to do so since the discovery of significantly eccentric exoplanets. On the analytical side, Legendre expansions in the ratio of semima jor axes generally use orbit averaging to produce a secular disturbing function which is suitable for the study of systems in which resonance does not play a role. Examples include the formulations of Innanen et al. (1997), Ford et al. (2000), Laskar & Boue (2010) and Naoz et al. (2011). Each of these uses a Hamiltonian approach, each is valid for arbitrary eccentricities and inclinations, and all but Innanen et al. (1997) are valid for arbitrary mass ratios. Innanen et al. (1997) includes only quadrupole terms (l = 2), Ford et al. (2000) and Naoz et al. (2011) include quadrupole and octopole terms (l = 2, 3), while Laskar & Boue (2010) includes terms up to l = 5 for inclined systems and l = 10 for coplanar systems. Naoz et al. (2011) demonstrate the importance of including the full mass dependence, especially for systems for which significant angular momentum is transferred between the inner and outer orbits. While secular expansions give significant insight into the long-term evolution of arbitrary configurations, care should be taken to determine that such an expansion is approriate for any given system. The orbit averaging technique effectively involves discarding all terms which depend on the fast-varying mean longitudes. However, even if resonance doesn’t play a role, that is, if no ma jor harmonic angles are librating, the influence of some non-secular harmonics on the secular evolution can be significant. In effect, all harmonics in a Fourier expansion of the disturbing function force all other harmonics at some level (Mardling 2013; hereafter Paper II in this series). On the other hand, the idea behind the averaging technique is that this influence is only short term in nature, and has no effect on the long-term evolution (see Wisdom 1982 for an excellent historical and insightful discussion of the use of the averaging principle).2 In fact, as with all forced nonlinear oscillatory systems (see, for example, Neyfeh & Mook 1979), the characteristic frequencies are modulated by forcing at some level, and for secularly varying triple configurations this becomes more pronounced the closer to the stability boundary the system is. A demonstration of this is given in Giuppone et al. (2011) who study the secular variation of a test particle orbiting the primary in a binary star system. In particular they demonstrate that predictions from the usual “first-order” secular theory consistently underestimate the frequency and overestimate the equilibrium value of the forced eccentricity, with relative errors 2 The averaging principle relies on adiabatic invariance in the system; see Landau & Lifshitz (1969) for an thorough discussion of this concept. c(cid:13) 2013 RAS, MNRAS 000, 1–45 4 Mard ling as large as 80% and 40% respectively when the ratio of binary to test particle semima jor axes is 10. Note that they take a binary mass ratio of 0.25 and a binary eccentricity of 0.36. Using a technique called Hori’s averaging process (Hori 1966), they go on to calculate the “second-order” corrections to the secular frequency and amplitude, reducing the errors to a few percent. While the authors do not identify the nature of the expansion parameter, (and claim that their expressions are too complicated to write down), in effect they are using the neglected harmonics to force the secular system, and in so doing obtain corrected frequencies. In fact, Hori’s averaging process is simply a version of the better-known Lindstedt-Poincar´e method for correcting the frequencies of a forced nonlinear oscillator (see Neyfeh 1973 for some simple applications including the Duffing equation). Here we present two new formulations of the hierarchical three-body problem which are accessible to anyone interested in the short and long-term evolution of small-N systems, be they stellar or planetary systems or a mixture of both (for example, circumbinary planets). The two formulations are valid for arbitrary mass ratios, and are distinguished by their choice of expansion parameter and hence their range of validity in those parameters. For closely packed systems, our generalisation of the literal expansion3 (Le Verrier 1855; Murray & Dermott 2000, and references therein) with its lack of constraint on the period ratio (except that the orbits should not cross) and its use of the eccentricities as expansion parameters is appropriate, while more widely spaced systems are best studied with the spherical harmonic expansion, a generalisation of the work of Kaula (1962) (also see Murray & Dermott 2000), which exploits the properties of spherical harmonics and which is valid for all eccentricities. Inclined systems will be studied in Paper III in this series. Throughout the paper we refer to “moderate mass ratio systems”. Our formal definition of such a system is one whose stability characteristics are govered by the interaction of N : 1 resonances (Paper II). In practice, however, this corresponds roughly to systems for which both mass ratios m2/(m1 + m2 ) and m3/(m1 + m2 + m3 ) are greater than around 0.05, where m1 > m2 . The paper is arranged as follows: 5 2. Spherical harmonic expansion 2.1. Derivation 2.2. Practical application: dominant terms. 2.3. Resonance widths and stability. 2.3.1. Libration frequency. 2.4. The secular disturbing function in the spherical harmonic expansion. 3. Literal expansion 12 3.1. Derivation 3.2. Eccentricity dependence. 3.2.1. Power series representations of Hansen coefficients and the choice of expansion order. 3.3. Dependence on the mass and semima jor axis ratios. 3.3.1. Summary of leading terms in α. 3.3.2. Coefficients when m2 /m1 → 0. 3.4. The spherical harmonic order m and principal resonances. 3.4.1. “Zeeman splitting” of resonances. 3.5. A second-order resonance. 3.6. First-order resonances. 3.7. Resonance widths using the literal expansion. 3.7.1 Libration frequency. 3.7.2. Widths of first-order resonances. 3.8. The secular disturbing function in the literal expansion. 4. Comparison of formulations to leading order in eccentricities 5. Equivalence of formulations 6. Comparisons with classic expansions 7. Conclusion and highlights of new results 22 24 25 27 8. Quick Reference 27 8.1. Harmonic coefficients for the semima jor axis expansion. 8.1.1. Secular disturbing function to octopole order. 8.1.2. Dom- inant non-secular terms. 8.1.3. Widths and libration frequencies of [N : 1](2) resonances. 8.2. Harmonic coefficients for the eccentricity expansion. 8.2.1. The secular disturbing function to second order in the eccentricities. 8.2.2. Widths and libration frequencies of [n′ : n](m) resonances. Appendices 32 A. Spherical harmonics. B. Hansen coefficients. C. Laplace coefficients. D. Lagrange’s planetary equations for the variation of the elements. E. The mean longitude at epoch. F. Notation. 3 The origin of the use of the word “literal” in this context is unclear, but one should take it to indicate that the dependence on the ratio of semima jor axes is via Laplace coefficients. c(cid:13) 2013 RAS, MNRAS 000, 1–45 General coplanar systems 5 k (b) m 3 m 2 θo θi C12ϕ i ϕ o C123 R m 3 i m 2 (a) r C12 m 1 Figure 1. (a): Jacobi coordinates r and R. C12 and C123 refer to the centre of mass of bodies 1 and 2, and of the whole system respectively. A hierarchical triple behaves like two weakly interacting binaries, with the inner binary composed of bodies 1 and 2, and the outer binary composed of body 3 plus a body of mass m1 + m2 situated at C12 . Bodies 1 and 2 are labelled such that m2 6 m1 . (b): Spherical polar angles associated with r (θi , ϕi ) and R (θo , ϕo ). The origin corresponds to the centre of mass of bodies 1 and 2. 2 SPHERICAL HARMONIC EXPANSION 2.1 Derivation Both expansions presented in this paper make use of three-body Jacobi or hierarchical coordinates and their associated osculating orbital elements to describe the dynamics of the system (eg. Murray & Dermott 2000). Illustrated in Figure 1(a), these are used for systems for which the motion of two of the bodies is predominantly Keplerian about their common centre of mass (the “inner orbit”), with the third body executing predominantly Keplerian motion about the centre of mass of the inner pair (the “outer orbit”). Note that the word “hierarchical” need not imply that the orbits are necessarily well-spaced, but rather, that the orbits retain their identities for at least several outer periastron passages, although the osculating orbital elements may vary dramatically from orbit to orbit if the system is unstable. With this broad definition, even systems involving the exchange of the outer body with one of the inner pair can be considered as hierarchical. Note also that one of the many advantages of using these coordinates is that the osculating semima jor axes are constant on average when resonance does not play a role. When resonance does play a role, it is then easy to define the energy exchanged between the orbits. Using Jacobi coordinates, the equations of motion for the inner and outer orbits are ∂R ∂ r Gm1m2 r2 µir + r = (1) and R = ∂R ∂R µo R + Gm12m3 R2 where r is the position of body 2 relative to body 1, R is the position of body 3 relative to the centre of mass of the inner pair, m1 and m2 are the masses of the bodies forming the inner orbit, m3 is the mass of the outer body, m12 = m1 + m2 , µi = m1m2/m12 and µo = m12m3/m123 are the inner and outer reduced masses with m123 = m1 + m2 + m3 , r and R are unit vectors in the r and R directions respectively, r = r, R = R and Gm12m3 Gm1m3 Gm2m3 R = − + + R R − β1 r R − β2 r is the disturbing function or interaction energy, with β1 = m1/m12 and β2 = −m2/m12 .4 In general we will use the subscripts i and o to represent quantities associated with the inner and outer orbits respectively. The notation ∂ /∂ r refers to the gradient with respect to the spherical polar coordinates (β1 r, θi , ϕi ) associated with the position of body 2 relative to the centre of mass of bodies 1 and 2, and similarly for ∂ /∂R for the position of body 3 with spherical polar coordinates (R, θo , ϕo ) relative to (2) (3) 4 Note that the use of Jacobi coordinates and the fact that the disturbing function has the units of energy and not energy per unit mass means that it is only necessary to define a single disturbing function. In constrast, in the classic expansions one distinguishes between an “internal” and “external” disturbing function, each composed of a “direct” component as well as individual “indirect” components (Murray & Dermott 2000). Similarly, there is no distinction between “internal” and “external” resonances when the single general disturbing function is used. c(cid:13) 2013 RAS, MNRAS 000, 1–45 6 Mard ling the same origin; see Figure 1(b). The equations of motion (1) and (2) written in this form clearly demonstrate the perturbed Keplerian nature of a hierarchical triple system when R and its gradients are small,5 which is clearly the case when r ≪ R. Note that the total energy is given by E = Ei + Eo − R, where (4) and Gm1m2 Gm12m3 2 µo R · R − Eo = 1 Ei = 1 2 µi r · r − R r are the instantaneous binding energies of the inner and outer orbits respectively. A common way to expand terms of the form of the last two in (3) is in terms of Legendre polynomials. In this case we have for s = 1, 2 ∞ R (cid:19)l Xl=0 (cid:18) βs r 1 R − βs r where cos ψ = r · R and Pl is a Legendre polynomial. A disadvantage of this is that the angles associated with each individual orbit do not appear explicitly. The use of spherical harmonics overcomes this problem as follows.6 Using the addition theorem for spherical harmonics (eg. Jackson 1975) one can write Pl (cos ψ), 1 R (5) (6) = . (7) (8) (9) Pl (cos ψ) = 4π 2l + 1 R = Gµim3 Ylm (θi , φi )Y ∗ lm (θo , ϕo ), l Xm=−l where Ylm is a spherical harmonic of degree l and order m with Y ∗ lm its complex conjugate. Using (6) and (7), the disturbing function (3) becomes ∞ l 2l + 1 (cid:19) Ml (cid:18) r l Rl+1 (cid:19) Ylm (θi , ϕi )Y ∗ Xm=−l (cid:18) 4π Xl=2 lm (θo , ϕo ), where the mass factor Ml is given by 1 + (−1)lml−1 ml−1 1 − β l−1 = β l−1 2 Ml = 2 ml−1 12 Note that M2 = 1 for any masses, while for equal masses Ml = 0 when l is odd. In this paper we focus on coplanar systems for which we take θi = θo = π/2, ϕi = fi + i and ϕo = fo + o , where fi and fo are the true anomolies of the inner and outer orbits respectively with i and o the corresponding longitudes of periastron. With the definition of the spherical harmonic used in Jackson (1975), Ylm (θ, ϕ) = s 2l + 1 4π where P m is an associated Legendre function, the disturbing function becomes l ∞ l lmMl eim(i −o ) (cid:16)r l eimfi (cid:17) (cid:18) e−imfo Rl+1 (cid:19) , Xl=2 Xm=−l,2 2 c2 1 P m l (cos θ) eimϕ , R = Gµim3 (l − m)! (l + m)! (10) (11) where c2 lm = [Ylm (π/2, 0)]2 = c2 l −m . 8π 2l + 1 A closed-form expression for these constants is given in Appendix A with specific values given in Table B2. Note that the sum over m is in steps of two because Ylm (θ, ϕ) ∝ cos θ when l − m is odd so that Ylm (π/2, ϕ) = 0 in this case. For stable systems including those near the stability boundary, the expressions in the last two pairs of brackets in (11) associated with the inner and outer orbits are nearly periodic in their respective orbital periods, and therefore can be expressed in terms of Fourier series of the inner and outer mean anomalies, Mi = νi t + Mi (0) and Mo = νo t + Mo (0) respectively, with Mi (0) and Mo (0) their values at t = 0 and νi and νo the associated orbital frequencies (mean motions). Thus (12) 5 Relative to the binding energy of the individual orbits and the other terms in (1) and (2) respectively. 6 The advantage of using spherical harmonics becomes particularly apparent when inclined systems are considered; see Paper III of this series. c(cid:13) 2013 RAS, MNRAS 000, 1–45 1 + ei cos fi (cid:19)l i (cid:18) 1 − e2 r l eimfi = al i eimfi = al i ∞ Xn=−∞ X l,m n (ei )einMi and General coplanar systems 7 (13) e−imfo Rl+1 = a−(l+1) o ∞ 1 + eo cos fo (cid:19)−(l+1) (cid:18) 1 − e2 Xn′=−∞ o where ai and ao are the inner and outer semima jor axes, ei and eo are the corresponding eccentricities, the Fourier coefficients e−imfo = a−(l+1) o (eo )e−in′Mo , X −(l+1),m n′ (14) ) ) and (15) (eo ) = X l,m n (ei ) = X −(l+1),m n′ (r/ai )l eimfi e−inMi dMi = O(em−n i 2π Z 2π 1 0 2π Z 2π e−imfo 1 (R/ao )l+1 ein′Mo dMo = O(em−n′ o 0 are called Hansen coefficients (Hughes 1981) and we have indicated the order of the leading terms (see Appendix B for graphical representations, closed-form expressions and approximations, Mathematica programs and in particular, Appendix B2 for general expansions which demonstrate the form of the leading terms). Note that since the real part of the integrands of (15) and (16) are even and the imaginary parts are odd, the integrals are real so that7 n i∗ hX l,m and i∗ hX −(l+1),m n′ justifying the notation used in (16). The disturbing function (11) can then be expressed as −n = X l,m = X l,−m n = X −(l+1),−m −n′ = X −(l+1),m n′ (17) (18) (16) , = (20) (19) R = (eo ) eiφmnn′ (eo ) cos φmnn′ Gµim3 ao Gµim3 ao n (ei )X −(l+1),m lmMl αl X l,m 2 c2 1 n′ n (ei )X −(l+1),m lm Ml αl X l,m ζm c2 n′ ∞ l ∞ Xn′=−∞ Xm=−l,2 Xn=−∞ ∞ ∞ l Xn′ =−∞ Xn=−∞ Xm=mmin ,2 ∞ Xl=2 ∞ Xl=2 where α = ai/ao , φmnn′ = nMi − n′Mo + m(i − o ) is a harmonic angle, ζm = ( 1/2, m = 0 and mmin = ( 0, 1, otherwise 1, In going from (19) to (20) we have used the properties (17) and (18) and have grouped together terms with the same value of m (thus the factor 1/2 in the definition of ζm ). Writing the harmonic angle in terms of longitudes only (in anticipation of employing Lagrange’s planetary equations for the rates of change of the elements), (21) becomes φmnn′ = nλi − n′λo + (m − n)i − (m − n′ )o , where λi = Mi + i and λo = Mo + o are the inner and outer mean longitudes respectively. Note that the harmonic angle should be invariant to a rotation of the coordinate axes. Since such a rotation changes all longitudes by the same amount, their coefficients should add up to zero thereby satisfying the d’Alembert relation (Murray & Dermott 2000), which indeed (23) does. Expression (20) for the disturbing function may be compared with that derived by Kaula (1962) for the case where m2 is a test particle (see also Murray & Dermott (2000), p232). Note that the Kaula expression is valid for arbitrary inclinations; this case will be considered Paper III in this series. l even l odd. (21) (22) (23) 0 f (x)dx while odd functions have the property that R a −a f (x)dx = 2 R a 7 Recall that even functions have the property that R a −a f (x)dx = 0. c(cid:13) 2013 RAS, MNRAS 000, 1–45 8 Mard ling Defining the coefficient of cos φmnn′ as Rmnn′ , it is desirable to change the order of summation of l and m so that m no longer depends on l (ie., it becomes a free index independent of any other index). The simplest way to see how this works is to write out the first few terms, grouping them appropriately. Thus ∞ Xl=2 l Xm=mmin ,2 Tlm = [T20 + T22 ] + [T31 + T33 ] + [T40 + T42 + T44 ] + [T51 + T53 + T55 ] + . . . = [T20 + T40 + . . .] + [T31 + T51 + . . .] + [T22 + T42 + . . .] + [T33 + T53 + . . .] + . . . ∞ ∞ Xl=lmin ,2 Xm=0 Tlm , = 2, m = 0 3, m = 1 m, m > 2 where lmin =   so that the disturbing function (20) becomes ∞ ∞ ∞ Xn′ =−∞ Xn=−∞ Xm=0 Rmnn′ cos φmnn′ R = with (24) (25) (26) (eo ) = Gµim3 Rp (eo ). Gµim3 ao Rmnn′ = n (ei )Z−(l+1),m lm Ml ρl X l,m ζm c2 n′ n (ei )X −(l+1),m lm Ml αl X l,m ζm c2 n′ ∞ Xl=lmin ,2 ∞ Xl=lmin ,2 Here ρ = ai/Rp with Rp = ao (1 − eo ) the outer periastron distance, and we will refer to (eo ) = (1 − eo )l+1X −(l+1),m Z−(l+1),m (eo ) n′ n′ as a modified Hansen coefficient. The form (28) is especially useful for systems with high outer eccentricity since X −(l+1),m (eo ) n′ is singular at eo = 1 while Z−(l+1),m (eo ) is not. Note that the summation over l in (27) is in steps of 2. Moreover, note that n′ there are only three independent indices associated with each harmonic for a coplanar system, although usually an additional one is included erroneously (see discussion in Section 6). This makes sense because there are three independent frequencies in the problem, that is, the two orbital frequencies and the rate of change of the relative orientation of the orbits. We will refer to the quantity Rmnn′ as the harmonic coefficient associated with the harmonic angle φmnn′ . Moreover, the classical nomenclature for terms associated with l = 2 and l = 3 is “quadrupole” and “octopole” (or “octupole”) respectively. (27) (28) (29) 2.2 Practical application: dominant terms The spherical harmonic expansion is significantly simpler to use than the literal expansion when the accuracy required can be achieved with only one or two values of l, and hence is recommended for use in preference to the latter except for the very closest systems (period-ratio-wise). In Section 4 we compare the two expansions to leading order in the eccentricities with the aim of determining the minimum period ratio for which the spherical harmonic expansion is acceptably accurate when only the two lowest values of l are included. Figures 2 and 3 suggest that this minimum is around 2 (panel (a) of Figure 2), although the expansion is still reasonably accurate for a period ratio as low as 1.5 (panel (a) of Figure 3). The question then arises: which harmonics in the triple-infinite series (26) should one include for a given application? How does one know whether or not resonant harmonics play a role? If they don’t, is it only necessary to include the secular terms in (26), that is, terms which do not depend on the mean longitudes (those with n = n′ = 0; see Section 2.4)? What about non-resonant, non-secular harmonics? A few general comments can be offered here, however, in general the answers depend on the questions being asked, on the timescales of interest and of course on the configuration itself. Timescales on which point-mass three-body systems which are coplanar and non-orbit crossing vary can generally be arranged according to the following hierarchy: Pi < Tp 6 Po < Plib < Psec ≷ τstab (30) c(cid:13) 2013 RAS, MNRAS 000, 1–45 General coplanar systems 9 where Pi and Po are the inner and outer orbital periods, Tp ≡ (1 − eo )3/2Po is the “time of periastron passage” of the outer body, a timescale of interest when the outer orbit is significantly eccentric and the orbit-orbit interaction is effective only around outer periastron, Plib = 2π/ωlib is the libration period in the case that the system is in (or near) resonance, with ωlib given by (44), Psec is the period on which the eccentricities vary secularly, and τstab is the dynamical stability timescale in the case that the system is unstable to the escape of one of the bodies (Lagrange instability). If one is interested in studying short-period variations on timescales up to a few times Po , non-secular harmonics whose coefficients are zeroth and/or first order in ei are generally included, independent of the value of the inner eccentricity. However, the selection from amongst such harmonics depends on the value of the outer eccentricity, and these are not necessarily those which are low-order in eo except when eo is small. Inspection of Figure B2 shows that for significant values of eo , harmonics spanning a wide range of values of n′ have similar amplitudes so that in principle, many harmonics should be included in such cases. One can avoid this by using overlap integrals, one for each value of n (and l and m); this technique will be discussed in a future paper in this series. While the eccentricities and semima jor axes of resonant or near resonant systems vary on the timescale of the orbital periods, these variations tend to accumulate on the libration timescale and it is the resonant harmonics which govern the behaviour. For stable systems with significant outer eccentricities it is usually adequate to include only one term in the analysis, that term being the [N : 1](2) harmonic with N ≃ σ ≡ νi /νo as discussed in Section 2.3 and Paper II. Here and later the notation [n′ : n](m) refers to the harmonic term associated with the angle φmnn′ and coefficient Rmnn′ (see Section 3.4). Unstable coplanar systems are also governed by resonant harmonics, but it is their interaction with “neighbouring” non- resonant terms which result in the chaotic behviour of the system. In this case it is usually sufficient to include only the resonant harmonic [N : 1](2) and its neighbour [N + 1 : 1](2) (Section 2.3), although one often needs to take into account the forced and secular variation of the eccentricities (Paper II). For stable systems one is often interested in the long-term secular variation of the elements, in which case it is usually sufficient only to include the secular [0 : 0](m) harmonics, that is, those which do not depend on the mean longitudes. In addition, it is normally only necessary to include the first two of these, that is, m = 0 and m = 1. However, for stable systems which are relatively close period-ratio-wise, it may be necessary to include the forcing effect of some non-secular harmonics; this is discussed in Section 1.1. On the sub ject of the secular variation of the elements, it is worth mentioning here that while the purely secular coplanar three-body system governed by the single angle φ100 = i − o is integrable and therefore not admitting of chaotic solutions, non-coplanar secular three-body systems as well as coplanar (and non-coplanar) higher-order (four-body etc) systems are governed by two or more independent angles and hence do admit chaotic solutions (see, for example, Laskar 1988 for a numerical study of the long-term secular evolution of the Solar System). 2.3 Resonance widths and stability Amongst systems with moderate mass ratios, stable systems tend to have significant period ratios because strong mutual interactions between the bodies tend to destabilize closer systems. The stability of a system can be studied using the heuris- tic resonance overlap stability criterion which involves calculating the widths of resonances (Chirikov 1979; Wisdom 1980; Mardling 2008; Paper II). For systems with moderate mass ratios, it is the [n′ : 1](2) resonances which govern the exchange of energy between the orbits and hence it is these which are responsible for the stability of the system. In this section we summarize the derivation of a simple expression for the widths of these resonances, the full derivation of which can be found in Paper II where a thorough study of resonance and stability in hierarchical systems with moderate mass ratios is presented. The general harmonic angle has the form φmnn′ = nλi − n′ λo + (m − n)i − (m − n′ )o , and unless n′ /n is sufficiently close to the period ratio νi /νo , this angle will circulate, that is, it will pass through all values [0, 2π ] because there is no commensurability between the rates of change of the individual angles making up φmnn′ . In fact, were there no (nonlinear) coupling between the inner and outer orbits, φmnn′ would circulate no matter how close n′ /n was to νi/νo (except if φmnn′ = 0 precisely). But because the orbits are able to exchange energy, the period ratio changes slightly each outer orbit allowing for the possibility of libration of φmnn′ when the energy is coherently transferred (ie, when conjunction occurs at almost the same place in the orbit; see, for example, Peale (1976) for a general discussion). In that case, φmnn′ will oscillate between two values such that H cos φmnn′ dφmnn′ 6= 0, where the integral is taken over one libration cycle. Then we refer to the harmonic angle in question as a resonance angle and say that the system is in resonance.8 The period of libration may be tens to hundreds or 8 Note that it is possible for the harmonic angle to librate and for the system to be not in resonance; formally the latter requires the c(cid:13) 2013 RAS, MNRAS 000, 1–45 10 Mard ling νi (t′ ) dt′ + ǫi , even thousands of outer orbital periods, depending on the system parameters; an expression for this is given below in terms of “distance” from exact commensurability in dimensionless units of period ratio. In order to study resonant behaviour, we use the pendulum model for resonance (see, for example, Chirikov 1979; Wisdom 1980; Murray & Dermott 2000) which involves deriving a pendulum-like differential equation for φmnn′ .9 To do this, we need to take into account the the dependence of the orbital frequency on time. Following Brouwer & Clements (1961) p285, the mean longitude is defined in terms of the orbital frequency such that λi = Mi + i = Z t νi (t′ ) dt′ + Mi (T0 ) + i = Z t T0 T0 where ǫi is the mean longitude at epoch t = T0 .10 With a similar expression for λo , the rate of change of a harmonic angle is then φmnn′ = nνi − n′ νo + n ǫi − n′ ǫo + (m − n) i − (m − n′ ) o . Except for systems with very small eccentricities, we have in general that i ≪ νo and o ≪ νo (see Section 3.4.1 for an example which illustrates this). Moreover, for all systems, ǫi ≪ νo and ǫo ≪ νo . Since some eccentricity is always induced11 and this is only small when the mass ratios are very small, for systems with moderate mass ratios it is a reasonable approximation to take φmnn′ ≃ nνi − n′ νo . Now consider the [n′ : n](m) = [N : 1](2) harmonic, where N is an integer close to the period ratio νi/νo . Libration of the angle φ21N ≡ φN will occur when φN = νi − N νo ≃ 0. Thus one can ask: how close to exact commensurability should the system be for this angle to librate? This is equivalent to asking for the width of the resonance. We can get a good answer to this question by showing that φN satisfies approximately a pendulum equation of the form φN = −ω 2 N sin φN , where ω 2 N depends on the parameters of the system. Note that the [N : 1](2) resonance librates about φN = 0 as we show N is known, the range of values of φN for which φN librates is determined from the equation for the pendulum below. Once ω 2 separatrix, that is, 2 (cid:19) , φN = ±2 ωN cos (cid:18) φN so that libration occurs if φN < 2 ωN when φN = 0. In order to determine ωN , we start by writing φN = νi − N νo = νo (cid:18) νi ao (cid:19) , νo (cid:18)σ νo (cid:19) = − ai 3 ao νo νi ai − N νi − N νo 2 where σ = νi/νo is the period ratio, and we have used Kepler’s third law to replace νi/νi by − 3 2 ai /ai and similarly for the outer orbit. The rates of change of the semima jor axes are given by Lagrange’s planetary equation (D4). For the latter, consider a reduced disturbing function which contains only the [n′ : 1](2) harmonics truncated at l = 2; one might call this the quadrupole contribution to the disturbing function, although it does not contain terms with m = 0 or n 6= 1. Referring to this as Rq , we have from (26) and (27) that ∞ 3 Gµim3 Xn′ =−∞ X −3,2 α2X 2,2 1 (ei ) n′ ao 4 (eo ) cos φn′ , (36) (37) (33) (34) Rq = (31) (32) (35) (38) existence of a hyperbolic point in the phase space ( φmnn′ , φmnn′ ) and this may not be the case when the eccentricities are very small (Delisle et al. 2012) which is never the case when at least one of the mass ratios is significant (except when the period ratio is very large). 9 Another model used to study resonance is the second fundamental model of resonance of Henrard & Lemaitre (1983). The associated Hamiltonian was designed specifically for the study of resonance capture, although it is possible to study this phenomenon without the Hamiltonian formalism using the pendulum model (Mardling & Udry in preparation). 10 See Appendix D for a discussion of the this orbital element. 11 An expression for the induced eccentricity is given in Paper II. c(cid:13) 2013 RAS, MNRAS 000, 1–45 General coplanar systems 11 9 4 φN = −ω 2 N sin φN = where φn′ = λi − n′ λo + i + (n′ − 2)o (39) and we have put c22 = 3/4 and M2 = 1. Now suppose that the harmonic angle with n′ = N librates and for now, assume that it is unaffected by all the harmonics contributing to Rq except itself. Retaining only the [N : 1](2) harmonic in (38), (37) together with Lagrange’s planetary equation (D4) gives 12 (cid:19)# X 2,2 o "(cid:18) m3 m123 (cid:19)2/3 (cid:18) m1m2 m123 (cid:19) + N 2/3 (cid:18) m12 1 (ei )X −3,2 ν 2 N (eo ) sin φN , m2 where we have replaced σ by N and used Kepler’s third law to replace α by (m12 /m123 )1/3N −2/3 . If we further replace the Hansen coefficients by the approximations given in Table B1 and (B5), we obtain o ( 9H22√2π "(cid:18) m3 o )3/4N 3/2 e−N ξ(eo )) sin φN , 12 (cid:19)# (cid:18) ei m123 (cid:19)2/3 (cid:18) m1m2 m123 (cid:19) + N 2/3 (cid:18) m12 o (cid:19) (1 − 13 φN = −ν 2 i )(1 − e2 24 e2 m2 e2 where ξ (eo ) = Cosh−1 (1/eo ) − √1 − e2 o and from Table B1 H22 = 0.71, giving us an expression for ωN and hence the range φN for which φN librates. Moreover, we see that φN does indeed librate around φN = 0 due to the fact that of values of X 2,2 1 (ei ) < 0 for all 0 < ei 6 1. A more practical definition of the resonance width is in terms of the “distance” from exact commensurability in di- mensionless units of period ratio. Rewriting (34) and incorporating the libration condition, we have that libration occurs when φN = νo (σ − N ) < 2 ωN so that the width of the [N : 1](2) resonance is approximately 12 (cid:19)#1/2 e1/2 eo ! (1 − 13 (2π )1/4 "(cid:18) m3 m123 (cid:19)2/3 (cid:18) m1m2 6H1/2 m123 (cid:19) + N 2/3 (cid:18) m12 o )3/8N 3/4 e−N ξ(eo )/2 , (43) i )1/2 (1 − e2 24 e2 22 i m2 where limeo→0 ∆σN is infinite for N = 1, finite for N = 2 and zero for N > 3. Notice the steep dependence on the quantity N ξ (eo ); since ξ (eo ) is a monotonically decreasing function of eo , the widths of high-N resonances are only signficant when eo is also high. Notice also that ∆σN = 0 when ei = 0; this implies that systems with circular inner orbits are always stable which is most certainly not the case. In fact one needs to know how much eccentricity is induced dynamically to calculate the true resonance width, and moreover one needs to know the maximum inner eccentricity the system acquires during a secular cycle to study its stability. This is thoroughly addressed in Paper II in which stability maps are plotted which clearly demonstrate the success of (43) as a predictor of instability using the concept of resonance overlap. Simple algorithms are also provided for determining the stability of any moderate-mass ratio hierarchical triple. Note that our definition of the resonance width does not involve the usual concept of “internal” and “external” resonance (Murray & Dermott 2000), in the same way that the present formulation does not involve separate internal and external disturbing functions. ∆σN = 2 ωN /νo = (40) (41) (42) 2.3.1 Libration frequency While the libration frequency of a pendulum depends on the amplitude, for small amplitudes it is independent of amplitude and is given approximately by ωN . Thus the libration frequency of the angle φN is ωN = νo∆σN /2. (44) For example, for an equal mass system with ei = 0.1, eo = 0.5 and σ = 20, the libration period is 1000 outer orbital periods, while increasing eo to 0.6 decreases this to only 66 outer orbital periods (with the same factor increase in the resonance width). 2.4 The secular disturbing function in the spherical harmonic expansion Keeping in mind the caveats discussed in the Introduction, one can use the averaging principle to eliminate fast-varying terms from the disturbing function (26), a process involving integrating over the two mean longitudes individually (as if they were independent) for an orbital period of each. In practice this is achieved simply by retaining only the n = n′ = 0 terms in (26). Using the notation R for the averaged disturbing function, we obtain c(cid:13) 2013 RAS, MNRAS 000, 1–45 12 Mard ling R = ∞ Xm=0 where Rm cos [m(i − o )] , (45) Rm = Gµim3 ao R = Gµim3 ao (ei )X −(l+1),m 0 (eo ). ζm c2 lm Ml αl X l,m 0 ∞ Xl=lmin ,2 n (ei ) and X −(l+1),m Closed-form expressions exist for X l,m (eo ) when n = n′ = 0; these are given in Appendix B, with some n′ explicit forms given in Table B2. Expanding to octopole order, the disturbing function (45) becomes " 1 cos(i − o )# . ao (cid:19)3 (cid:18) m1 − m2 ao (cid:19)2 4 e2 2 e2 m12 (cid:19) ei eo (1 + 3 1 + 3 i ) 16 (cid:18) ai 4 (cid:18) ai 15 i o )3/2 − (1 − e2 (1 − e2 o )5/2 For coplanar secular systems, only the rates of change of the eccentricities and longitudes of the periastra are of interest. From Lagrange’s planetary equations (Appendix D), these are i (cid:1)1/2 i ) (cid:0)1 − e2 4 e2 ao (cid:19)4 eo (1 + 3 m12 (cid:19) (cid:18) ai m12 (cid:19) (cid:18) m1 − m2 16 (cid:18) m3 15 sin(i − o ), = −νi o )5/2 (1 − e2 i (cid:1)1/2 i (cid:1)1/2 m12 (cid:19) " 3 i ) (cid:0)1 − e2 4 e2 ao (cid:19)3 (cid:0)1 − e2 ao (cid:19)4 eo (1 + 9 = νi (cid:18) m3 m12 (cid:19) (cid:18) ai 16 (cid:18) m1 − m2 4 (cid:18) ai 15 o )3/2 − o )5/2 ei (1 − e2 (1 − e2 ao (cid:19)3 ei (1 + 3 4 e2 i ) m12 (cid:19) (cid:18) ai 12 (cid:19) (cid:18) m1 − m2 16 (cid:18) m1m2 15 = νo sin(i − o ), m2 o )2 (1 − e2 o )3 cos(i − o )# . 12 (cid:19) " 3 ao (cid:19)2 (cid:0)1 + 3 ao (cid:19)3 eo (cid:0)1 + 4e2 2 e2 o (cid:1) i (cid:1) = νo (cid:18) m1m2 m12 (cid:19) (cid:18) ai 16 (cid:18) m1 − m2 4 (cid:18) ai ei 15 o )2 − m2 (1 − e2 (1 − e2 Thus, for example, it is clear that for systems with m1 = m2 , there is no secular variation in the eccentricities at this level of approximation and consequently, the inner and outer rates of apsidal motion are constant. cos(i − o )# , 4 e2 (1 + 3 i ) (50) dei dt di dt deo dt do dt (51) (46) (47) (48) (49) 3 LITERAL EXPANSION 3.1 Derivation The original literal expansions (for example, Le Verrier 1855) were especially devised to study planetary orbits in the Solar System, and in particular, to take advantage of the small planet-to-star mass ratios, small eccentricities and inclinations, while putting essentially no restrictions on the ratio of semima jor axes except that they should not cross. Our aims here are to generalise the formulation so that no assumptions about the mass ratios are made, and to present the formulation in a clear and concise way which makes it easy to use to any order in the eccentricities and for any appropriate application. Again, for this paper we consider coplanar configurations only, and for clarity of presentation, we will repeat the definition of some quantities already defined in Section 2. We start by writing down the disturbing function in a form which is useful for the coming analysis: = Gm12m3 R = − R Gµim3 ao Gm2m3 Gm1m3 + + R − β1 r R − β2 r (cid:20)β−1 R − β2r (cid:21) , 2 (cid:16) ao ao ao R (cid:17) + β−1 1 β−1 R − β1 r − β−1 1 2 where again, β1 = m1/m12 and β2 = −m2/m12 . Now consider the second and third term in (52) and write ao ao 1 ao = , = R − βs r R p1 − 2xs cos ψ + x2 pR2 − 2βs r · R + (βs r)2 s with s being 1 or 2, xs = βs r/R and cos ψ = r · R so that for coplanar systems, ψ = fi + i − fo − o , (52) (53) (54) c(cid:13) 2013 RAS, MNRAS 000, 1–45 General coplanar systems 13 where again, fi and fo , and i and o are the inner and outer true anomalies and longitudes of periastron respectively. The literal expansion involves (1): a Taylor series expansion about the circular state xs = αs ≡ βsα with small parameter ǫ related to the eccentricities, followed by (2): a Fourier expansion in the angle ψ ; this step introduces the Laplace coefficients, then (3): binomial expansions of powers of ǫ; (4): Fourier series in the mean anomalies, and finally (5): combining the three contributions to the disturbing function in a simple expression. This procedure is set out in Murray & Dermott (2000) for the case of the restricted problem; here we present a significantly more compact formulation for the general problem. Now let s (cid:3)−1/2 g (xs , ψ) = (cid:2)1 − 2xs cos ψ + x2 Writing ao (cid:19) (r/ai ) = βs (cid:18) ai βs r xs = R (R/ao ) = αs (cid:18) 1 − e2 1 + ei cos fi (cid:19) (cid:18) 1 + eo cos fo i 1 − e2 o ≡ αs (1 + ǫ) where (cid:19) (56) (55) . (57) ǫ = (r/ai )/(R/ao ) − 1 is first-order in the eccentricities, the first two steps described above are s (cid:3)−1/2 (cid:2)1 − 2xs cos ψ + x2 = = = = . . . Step 1 (αs ǫ)j j ! (αs ǫ)j j ! 2 b(m) 1 1/2 (αs ) = = g (αs + αs ǫ, ψ) ∞ (αs ǫ)j ∂ j g s (cid:12)(cid:12)(cid:12)(cid:12)xs=αs Xj=0 ∂ xj j ! ∞ ∂ j Xj=0 [g (αs , ψ)] ∂αj s ∞ s " ∞ 1/2 (αs )eimψ # ∂ j 2 b(m) Xj=0 Xm=−∞ 1 ∂αj ∞ ∞ 2 B(j,m) Xj=0 Xm=−∞ (αs ) ǫj eim(fi −fo ) , eim(i −o ) 1 1/2 where we have used (54) for ψ in the last step. The Fourier coefficient in Step 2 is a Laplace coefficient defined by 2π Z 2π e−imψ 1 p1 − 2αs cos ψ + α2 0 s (Murray & Dermott 2000), with the factor 1/2 (as opposed to the subscript 1/2) introduced to obtain the standard definition of b(m) 1/2 . Note that since the real part of the integrand is even and the imaginary part is odd, the integral is real12 so that 1/2 (αs )i∗ (αs ) = hb(m) The function in (58) involving the j th derivative of the Laplace coefficient is αj dj B(j,m) s 1/2 dαj j ! s General properties of Laplace coefficients and their derivatives are given in Appendix C. Note also that we have used m for the Fourier summation index because in fact it corresponds to the spherical harmonic order m (see Section 5 where the equivalence of the two formulations is demonstrated). Referring to (53) and introducing the factor ao /R, the next step in the procedure is a binomial expansion of ǫj , so that ∞ ∞ (R/ao ) − 1(cid:21)j (αs ) (cid:20) (r/ai ) (cid:16) ao R (cid:17) eim(fi −fo ) s (cid:3)−1/2 Xj=0 Xm=−∞ (ao /R) (cid:2)1 − 2xs cos ψ + x2 2 B(j,m) eim(i −o ) 1 1/2 = b(m) 1/2 (αs ). b(m) 1/2 (αs ). (60) (61) b(−m) 1/2 . . . Step 2 (58) (59) = (62) (αs ) = dψ 12 See footnote on page 7. c(cid:13) 2013 RAS, MNRAS 000, 1–45 14 Mard ling eimfi = (63) (64) e−imfo = = ∞ Xj=0 X −(k+1),m n′ X k,m n (ei ) einMi (eo ) e−in′Mo , ∞ Xm=−∞ 2 B(j,m) eim(i −o ) 1 1/2 (αs ) j k!(−1)j−k h(r/ai )k eimfi i (cid:20) e−imfo Xk=0 j (R/ao )k+1 (cid:21) , . . . Step 3 where we have gathered together in the square brackets quantities associated with the inner and outer orbits in preparation for the next step. As functions of the eccentricities and the sine and cosine of the true anomalies, these terms can be expanded in Fourier series with period 2π such that ∞ 1 + ei cos fi (cid:21)k 1 − e2 (r/ai )k eimfi = (cid:20) Xn=−∞ i and ∞ (cid:21)k+1 e−imfo (R/ao )k+1 = (cid:20) 1 + eo cos fo Xn=−∞ 1 − e2 o (eo ) are again Hansen coefficients given by (15) and (16) (recall that X −(k+1),−m (ei ) and X −(k+1),m where X k,m n −n′ n′ Substituting (64) and (65) into (63) and gathering together the angles, we obtain for s = 1, 2 1/2 (αs ) " j ∞ ∞ ∞ ∞ k!X k,m (eo )# eiφmnn′ (−1)j−k j 2 B(j,m) Xj=0 Xn′ =−∞ Xn=−∞ Xk=0 Xm=−∞ 1 n Xn′ =−∞ " ∞ ∞ ∞ ∞ mnn′ (ei , eo )# cos φmnn′ , (αs ) F (j ) ζm B(j,m) Xj=0 Xn=−∞ Xm=0 1/2 j k!X k,m (−1)j−k j Xk=0 n φmnn′ = nMi − n′Mo + m(i − o ) = nλi − n′ λo + (m − n)i − (m − n′ )o is again a harmonic angle, and ζm is 1/2 when m = 0 and 1 otherwise. As with the spherical harmonic formulation, in going from (66) to (67) we have paired together terms with positive and negative values of m to make the expression manifestly real. ao R − βs r F (j ) mnn′ (ei , eo ) = (ei ) X −(k+1),m n′ (ei ) X −(k+1),m n′ = X −(k+1),m n′ . . . Step 4 (66) (65) ). (67) (68) (69) = = where (eo ), The final step involves writing down the full literal expansion for the disturbing function. Substituting (67) into (52), one obtains Ajm (α; β2 ) F (j ) mnn′ (ei , eo ) cos φmnn′ = R = Gµim3 ao ∞ ∞ ∞ ∞ Xj=0 Xn′ =−∞ Xn=−∞ Xm=0 ∞ ∞ ∞ Xm=0 Xn=−∞ Xn′ =−∞ Rmnn′ cos φmnn′ , where the harmonic coefficient associated with the angle φmnn′ is ∞ Xj=0 Ajm (α; β2 ) F (j ) mnn′ (ei , eo ), Gµim3 ao Rmnn′ = . . . Step 5 (70) (71) with 1/2 (α2 )i Ajm (α; β2 ) = ζm hβ−1 2 B(j,m) 1 B(j,m) 1/2 (α1 ) − β−1 for all j , m except when j = m = 0 in which case 1/2 (α2 )i + β−1 2 hβ−1 2 b(0) 1 b(0) 1 β−1 1/2 (α1 ) − β−1 A00 (α; β2 ) = 1 2 . Recall here that −β2 = m2/m12 = 1 − β1 , α = a1/a2 and αs = βsα, s = 1, 2. Note that the order of the expansion is given by the number of terms included in the summation in (71), that is, it is given by the maximum value of j . The equivalence of the the literal and spherical harmonic formulations is demonstrated in Section 5. (73) (72) c(cid:13) 2013 RAS, MNRAS 000, 1–45 General coplanar systems 15 In contrast to the classical literal expansion which is valid for m2/m1 ≪ 1 (see Section 3.3.2) and involves Laplace coefficients as functions of the ratio of semima jor axes α, (70) is valid for any mass ratios and expresses the disturbing function in terms of Laplace coefficients whose arguments are αs = βsα, that is, the ratio of semima jor axes scaled by the mass ratios ms /m12 , s = 1, 2. By calculating the harmonic coefficients for a second-order resonance as well as those for general first-order resonances and for the secular harmonics, and also by calculating resonance widths, we demonstrate in Sections 3.5, 3.6, 3.7 and 3.8, the ease with which this form of the literal expansion can be used to any required order in eccentricity. 3.2 Eccentricity dependence The dependence of the disturbing function on the eccentricity is via F (j ) mnn′ (ei , eo ), defined in (68). Although each term in em−n′ ), the leading order of F (j ) the finite summation over k is O(em−n mnn′ (ei , eo ) will be either j or j + 1 when j > o i m − n + m − n′ ≡ η . This results from the fact that from (62), F (j ) mnn′ (ei , eo ) ∝ ǫj with ǫ = O[max(ei , eo )], and also that the Hansen coefficients which make up F (j ) mnn′ (ei , eo ) are either odd or even functions of the eccentricity (so their power series expansions have only odd or even powers). In particular, em−n′ O(em−n mnn′ (ei , eo ) =  ), o i F (j ) em−n′ ), . . . , O(ej−m−n′ ej−2−m−n ), O(em−n+2 O(em−n ej−m−n ), j − η even o o o i i i em−n′ ), . . . , O(ej+1−m−n′ ej−1−m−n ), O(em−n+2 O(em−n ej+1−m−n  j − η odd o o o i i i 423 (ei , eo ), F (2) i eo ) as it is for F (1) Thus for example, the leading-order term in an expansion of F (0) 423 (ei , eo ) is O(e2 423 (ei , eo ) and F (3) 423 (ei , eo ), while the leading-order terms of both of F (4) 423 (ei , eo ) and F (5) 423 (ei , eo ) are O(e2 i e3 o ) and O(e4 i eo ). As a consequence, if one requires an expansion of the disturbing function which is correct to order jmax in the eccentricities, then one should include terms up to and including j = jmax in (71), and moreover, expand each of F (j ) mnn′ (ei , eo ), j 6 jmax , to order jmax in the eccentricities. Similarly, one should only include harmonics which are such that m − n + m − n′ 6 jmax . Conversely, if one is particularly interested in a term whose harmonic angle is φmnn′ , one should include terms at least up to j = m − n + m − n′ in order to obtain a non-zero coefficient for such a term. For practical applications, it is usually most efficient to evaluate the functions F (j ) mnn′ (ei , eo ) using Mathematica (Sec- tion B4) or similar for the particular range of values of m, n, n′ , j of interest, summing over the various Hansen coefficients which contribute. However, it is of considerable interest to examine the functional dependence of the power-series representa- tions of the Hansen coefficients, not only on the eccentricity, but also on the associated indices n, j and m. j 6 η j > η , j > η , ), (74) 3.2.1 Power series representations of Hansen coefficients and the choice of expansion order Power series expansions for X j,m n (e) are given in Appendix B2 for arbitrary j and n, and for m = n ± p, p = 0, 1, 2, 3, 4, with the choice of values for m being guided by the use of Hansen coefficients in the study of resonance, since the order of a resonance is m − n + m − n′ (see Section 3.5). Two features in particular emerge from these general series expansions. First one sees that their leading terms are indeed proportional to em−n , and secondly that the coefficients of the leading terms contain contributions which are O(j p ) and O(np ). This means that X j,m n (e) is in fact O{max[(ne)m−n , (j e)m−n ]}, which in turn has implications for the radius of convergence of the series (the range of values of each of the eccentricities for which the series converges), and the order at which one should truncate the series for given eccentricities. This should be kept in mind when using these expansions, especially for systems with period ratio close to one. The effect is evident when one compares, for example, the first-order Hansen coefficients in Figures (B4) and (B5). In both of these figures, numerically evaluated integrals (solid curves) are compared with their fourth-order correct series approximations (dashed curves). For example, the series representation of X 0,2 1 (ei ) = O(ei ) in panel (a) of Figure (B4) approximates the actual function well for <∼ 0.8, while that of X 0,7 0 6 ei 6 (ei ) = O(6ei ) in panel (a) of Figure (B5) is only accurate for 0 6 ei <∼ 0.1 ≃ 0.8/6 = 0.13. While it may be tempting to avoid the series expansions of Hansen coefficients with high values of n, one should remember that any particular harmonic coefficient is only accurate to order jmax > m−n+ m−n′ , even if individual Hansen coefficients (and hence F (j ) mnn′ (ei , eo )) are evaluated accurately. On the other hand, if computational efficiency is required, it is best to calculate individual series expansions of F (j ) mnn′ (ei , eo ) to adequate order in the eccentricities. A Mathematica program is provided in Appendix B4 for this purpose. While we do not attempt a formal convergence analysis here, the most straightforward way to gain confidence in any particular expansion is to compare its predictions with direct numerical integration of the equations of motion. c(cid:13) 2013 RAS, MNRAS 000, 1–45 16 Mard ling 3.3 Dependence on the mass and semima jor axis ratios = ζm E (j,m) 0 Ajm (α; β2 ) = ζm ∞ Xp=0 = ζm E (j,m) 0 The coefficient of any particular harmonic term in the Fourier expanded disturbing function (70) is given by (71), and this itself may be expressed as an infinite series of terms in increasing orders of eccentricity. Each term contributing to a coefficient depends on the mass ratio m2/m1 and the semima jor axis ratio α though the factor Ajm (α; β2 ). To obtain an idea of this dependence, we can use the expansion (C12) for B(j,m) 1/2 (αi ) in (72). Noting that the leading term in this expansion depends on whether j 6 m or j > m, we have in the first case that (cid:2)βm+2p−1 (cid:3) αm+2p − βm+2p−1 E (j,m) p 2 1 + (−1)mmm−1 (cid:20) mm−1 (cid:21) αm + . . . 2 1 mm−1 12 (cid:2)1 + (m − 1)(m2 /m1 ) + . . . + (−1)m (m2/m1 )m−1 + . . .(cid:3) αm + . . . , is given by (C15). However, the leading term is zero when m = 1 in which case (cid:20) m1 − m2 (cid:21) α3 + . . . m12 [1 − 2(m2 /m1 ) + . . .] α3 + . . . , while for m = j = 0 we have from (73) that (cid:20) m3 1 + m3 (cid:21) α4 + . . . 2 E (0,0) 2 E (0,0) α2 + 1 2 A00 (α; β2 ) = 1 2 1 m3 12 4 α2 + 9 64 (1 − 3(m2 /m1 ) + . . .) α4 + . . . = 1 The fact that there are no monopole or dipole terms (ie., no power of α less than 2) is consistent with the spherical harmonic expansion (27). When j > m, Aj1 (α; β2 ) = E (j,1) 1 = E (j,1) 1 where E (j,m) 0 j 6 m, m > 2, (75) j = 0, 1, (76) (77) ∞ Xp=p∗ ζm E (j,m) p∗ − βm+2p−1 (cid:2)βm+2p−1 E (j,m) (cid:3) αm+2p Ajm (α; β2 ) = ζm p 2 1 1 + (−1)jmj−1 (cid:20) mj−1 (cid:21) αj + . . . , =  2 mj−1 12 1 + (−1)j+1mj (cid:20) mj (cid:21) αj+1 + . . . , ζm E (j+1,m) 2  p∗ mj 12 where p∗ = ⌊(j − m + 1)/2⌋ with ⌊ ⌋ denoting the nearest lowest integer and E (j,m) is given by (C16). Note that since the p∗ leading terms of the harmonic coefficient Rmnn′ are of order m − n + m − n′ in eccentricity, they will be such that 0 6 j 6 m (see example in the next Section). j > m, j − m even, j > m, j − m odd. (78) 3.3.1 Summary of leading terms in α Ajm (α; β2 ) = We can summarize the above as follows. For Ajm we have  O(αm ), m > 2, j 6 m O(αj ), j − m even j > m, O(αj+1 ), j − m odd j > m, O(α2 ), j = 0 m = 0,  O(α3 ), m = 1, j = 0, 1, so that the harmonic coefficients are such that Rmnn′ =  O(α2 ), m = 0 O(α3 ), m = 1 O(αm ), m > 2.  (79) (80) c(cid:13) 2013 RAS, MNRAS 000, 1–45 General coplanar systems 17 3.3.2 Coefficients when m2/m1 → 0 The standard literal expansion is derived assuming that one or other of the mass ratios m2/m1 and m3 /m1 is zero (Murray & Dermott 2000). Putting β1 = 1 and β2 = 0 in (72) and (73), we have Ajm = ζmB(j,m) 1/2 (α) for all j , m except when m = 1, j = 0, 1 and when j = m = 0. In these cases, from (C2) and (C12) we have lim β2→0 (81) A01 = b(1) 1/2 (α) − α, db(1) 1/2 (α) dα A11 = α − α, lim β2→0 lim β2→0 and (82) (83) lim β2→0 2 b(0) A00 = 1 1/2 (α) − 1 (recall that b(0) 1/2 (0) = 2; see (C2) and (C3)). Using these approximations makes the disturbing function zeroth-order correct in the mass ratio m2 /m1 (times the factor m2 m3 when it has dimensions of energy as in the formulations presented here), so that the rates of change of the elements are first-order in m3/m1 for the inner elements, and first-order in m2/m1 for the outer elements (see Sections 2.4 and 3.8). (84) 3.4 The spherical harmonic order m and principal resonances For coplanar systems there are three labels, m, n and n′ , associated with each harmonic. In turn, each label is associated with an independent frequency of the system: n and n′ are associated with the inner and outer orbital frequencies respectively, while m is assocated with the difference in the rates of apsidal advance i − o ; see the definition of the harmonic angle (69). In Section 5 we demonstrate the equivalence of the spherical harmonic and literal expansions, where the index m in the literal expansion is shown to correspond to the spherical harmonic order m. This plays an important role in many physical systems, and the three-body problem is no exception. For example, we will show in a future paper in this series that one can define the concept of “modes of oscillation of a binary” which are excited in the presence of a triple companion, in analogy with the modes of oscillation of a star which are excited in the presence of a binary companion. The spherical harmonic order m acts as an azimuthal mode number in the formalism, with the analogy between the two physical problems revealing a rich vein of exploration. First note that m distinguishes resonant states with the same values of n and n′ . In general, a harmonic coefficient is em−n′ O(em−n ), and since o i m − n + m − n′ = ( n′ − n, 2m − n − n′ , the order in eccentricity is minimized at n′ − n when n 6 m 6 n′ . Using the notation [n′ : n](m) introduced in Section 2.2 to emphasise the association of the harmonic angle φmnn′ with the n′ : n resonance of spherical harmonic order m, we refer to the [n′ : n](m) resonances, n 6 m 6 n′ , as the principal resonances or principal harmonics of the n′ : n resonance. For example, the two principal 2 : 1 resonances are [2 : 1](1) and [2 : 1](2), with harmonic angles φ112 = λi − 2λo + o and φ212 = λi − 2λo + i respectively and with harmonic coefficients R112 = O(eo ) and R212 = O(ei ). Similarly, there are four principal 5 : 2 resonances, each third-order in eccentricity, namely [5 : 2](2), [5 : 2](3), [5 : 2](4) and [5 : 2](5), with resonance angles φ225 = 2λi − 5λo + 3o , φ325 = 2λi − 5λo + i + 2o , φ425 = 2λi − 5λo + 2i + o , and φ525 = 2λi − 5λo + 3i , and harmonic coefficients R225 = O(e3 i eo ) and R525 = O(e3 o ), R425 = O(e2 o ), R325 = O(ei e2 i ). In general there are n′ − n + 1 principle harmonics associated with the n′ : n resonance. The 2 : 1 resonance is referred to as a first-order resonance because the minimum order in eccentricity of either of the principal harmonic coefficients is first order. Using the nomenclature introduced here, we can be more definite and say that in general, a resonance is pth-order if the principal resonances are pth-order in eccentricity. It is sometimes desirable to express the “largeness” or otherwise of the values of n and n′ , especially for first-order resonances. The author is aware that the term resonance degree is occasionally used for this purpose, however, in the context of spherical harmonics the words “degree” and “ order” are associated with the indices l and m respectively. In hindsight this is unfortunate because m could have been used for this purpose had the word “order” not already refered to the value of n 6 m 6 n′ , otherwise, (85) c(cid:13) 2013 RAS, MNRAS 000, 1–45 18 Mard ling n′ − n. Moreover, one correctly refers to a polynomial’s degree rather than order when describing its highest power (although the latter is often used), and had degree been adopted for describing the order in eccentricity of a resonance, all would be consistent. But history takes precedence for words in common use, and the 7 : 6 resonance continues to be a seventh-degree first-order resonance. Finally recall from Section 3.3.1 that Rmnn′ = O(αm ), m > 2, while R0nn′ = O(α2 ) and R1nn′ = O(α3 ). This implies that in general, unless eo ≪ ei , it is the principal resonance with m = n which tends to make the largest contribution to the disturbing function, except when n = 1 in which case the m = 2 harmonic tends to make the largest contribution. 3.4.1 “Zeeman splitting” of resonances Just as a magnetic field introduces fine stucture to atomic energy levels (Zeeman splitting), apsidal advance of the inner and outer orbits introduces fine structure in the positions of the centres of resonances relative to exact commensurability. In both cases it is the spherical harmonic order m which labels the associated frequencies and moreover physically, it is the introduction of one or more distinguished directions (magnetic field or third body) which breaks the otherwise symmetric state of the system. The slow rotation of the system about these directions introduces new (generally low) frequencies, splitting the otherwise degenerate state. To get an idea of the magnitude of this effect, consider a two-planet system near the 2 : 1 resonance with stellar and planetary masses m∗ , mi and mo , with mi the mass of the inner planet and mi , mo ≪ m∗ . The rate of apsidal advance is given by Lagrange’s planetary equation (D2) which involves a partial derivative with respect to the eccentricity. To obtain a quick estimate of the rates for both orbits which is correct to first-order in the eccentricities, it is simplest to use the leading terms in the spherical harmonic expression (27) for the harmonic coefficients, including in the disturbing function R000 and R100 for the secular contributions (all others are more than second-order in the eccentricities), and R212 and R112 for the quadrupole and octopole resonant contributions corresponding to the two principal resonant angles φ212 and φ112 . The error incurred in using only the leading term in the sum over l is discussed in Section 4. To second-order in the eccentricities and to zeroth-order in mi/m∗ these are 1 4 Gmimo ao α2 (1 + 3 2 e2 2 e2 i )(1 + 3 o ), 15 16 Gmimo ao α3 ei eo , α2 ei R000 = R100 = − Gmimo Gmimo 9 9 and R112 = R212 = − 4 ao 8 ao where we have used the expansions for the Hansen coefficents in Section B2. If all of i − o , φ212 and φ112 librate, the rates of apsidal advance are then νi (cid:18) mo ei (cid:19) cos(i − o ) − (cid:18) 3 m∗ (cid:19) α3 (cid:20)1 − ei (cid:19) cos(λi − 2λo + i )(cid:21) α (cid:18) eo α3 eo , i = (86) (88) (87) 3 4 5 4 and 5 4 3 4 o = νo (cid:18) mi eo (cid:19) cos(i − o ) + (cid:18) 3 m∗ (cid:19) α2 (cid:20)1 − 2eo (cid:19) cos(λi − 2λo + o )(cid:21) . α (cid:18) ei Whether or not a particular angle contributes on average to i and o depends on whether it librates or not, that is, whether or not the average value of its cosine is non-zero. If it does librate, the sign of its contribution will depend on whether it does so around zero or π (or some other angle in some cases). Using (102), one can show that for small eccentricities, when φ212 and φ112 librate, they do so around zero and π respectively. One may then ask whether it is possible for both angles to librate at the same time. It is possible to show that if the harmonic angle φ21N librates, then all other angles of the form φ21n′ , n′ 6= N , must circulate. This is not necessarily true for a set of principal resonances because for any two angles from the set, labeled, say, by m1 and m2 (not to be confused with the masses), (89) φm2 nn′ = φm1nn′ + (m2 − m1 )(i − o ). Thus if one resonance angle librates and in addition, i − o librates, then all other associated principal resonance angles will librate.13 Now, the angle i −o will librate if the eccentricities are small enough (see, for example, Mardling (2007) for a study of the libration and circulation of this angle in the case of secular evolution). If this occurs, then since φ212 − φ112 = i − o , then i − o must librate around −π (because φ212 librates around zero and φ122 librates around π ) so that the average (90) 13 In fact, al l n′ : n resonance angles will librate, not just the principal angles. c(cid:13) 2013 RAS, MNRAS 000, 1–45 value of cos(i − o ) is −1, while the average values of cos φ212 and cos φ112 are 1 and −1 respectively (at exact resonance). From (88) and (89), the average rates of apsidal advance in this case are therefore m∗ (cid:19) α3 (cid:20)1 − (cid:18) 3 − 5 4 αeo νi (cid:18) mo (cid:19)(cid:21) ≃ − 9 4 νo (mo/m∗ )σ−1 e−1 i ei i = (91) 3 4 and General coplanar systems 19 3 4 o = νo (cid:18) mi (cid:19)(cid:21) ≃ − 9 m∗ (cid:19) α2 (cid:20)1 − 4 (cid:18) 6 − 5αeo 1 8 νo (mi/m∗ )σ−4/3 e−1 o , ei where σ is the period ratio and and the approximations hold for small to moderate eccentricities. In such cases, the rates of change of the two 2 : 1 principal resonances are, from (69), φ212 = νo [(σ − 2) − 9 4 (mo/m∗ )σ−1 e−1 i and φ112 = νo [(σ − 2) − 9 8 (mi/m∗ )σ−4/3 e−1 o ]. For two Jupiter-mass planets orbiting a solar-mass star, the positions of exact resonance (that is, the value of σ for which φ212 = 0 or φ112 = 0) are therefore approximately a distance (94) (92) ] (93) δσ212 = 0.0011 e−1 i and δσ112 = 0.0004 e−1 o (95) away from exact commensurability. These can be signficant for small eccentricities, and this should be remembered when deciding whether or not an oberved system is likely to be in resonance (sub ject to the caveat discussed in footnote 2 on page 9). It is interesting to note here that for non-coplanar systems, there are five independent labels including n, n′ and m, and an additional two spherical harmonic m’s which we denote by mi and mo (non-coplanar systems will be studied in Paper III in this series). The harmonic angle becomes φmimom n n′ = nMi − n′Mo + mi ωi − mo ωo + m(Ωi − Ωo ) = nλi − n′λo + (mi − n)i − (mo − n′ )o + (m − mi )Ωi − (m − mo )Ωo , with respectively ωi and ωo , and Ωi and Ωo , the arguments of periastron and the longitudes of the ascending nodes of the inner and outer orbits respectively. Note that for coplanar systems, mi = mo = m. The additional labels reflect the extra frequencies introduced when the problem becomes three dimensional. The three frequencies associated with m, mi and mo are, respectively, the difference in the rates of precession of the orbital planes about the total angular momentum vector, and the rates of change of the inner and outer arguments of periastron. We note also that the additional fine structure introduced when the orbits are not coplanar has its own analogy with Zeeman splitting. Before the latter phenomenon was understood in the context of quantum mechanics, physicists referred to energy level splittings which were accurately predicted by the classical theory of Lorentz as “normal” and those which were not as “anomalous.” Once the quantum mechanical concept of electron spin was introduced, it became clear that the additional source of angular momentum was responsible for the “anomalous” fine structure, with states not involving electron spin remaining “normal.” Suffice to say here that orbital precession introduces “anomalous” fine structure, with “normal” fine structure associated with principal resonances for which mi = mo = m. As well as illustrating the phenomenon of resonance splitting, the fine structure calculation (86) to (95) serves to further demonstrate the ease with which the spherical harmonic expansion can be applied. We now consider some specific applications which make more explicit the power of the literal expansion. (96) 3.5 A second-order resonance Written in the form (70) with (71), it is easy to include terms to any order in the eccentricities and mass ratios. For example, say one wanted to study the 5 : 3 resonance for which n = 3 and n′ = 5. The principal resonance angles are (from (69)), φ335 = 3λi − 5λo + 2o , φ435 = 3λi − 5λo + i + o and φ535 = 3λi − 5λo + 2i , with R335 = O(e2 o ), R435 = O(ei eo ) and R535 = O(e2 i ). If we choose to include terms, say, up to fourth order in the eccentricities, we would include in the summation in (71) terms up to j = 4. Note that for this resonance, all terms will be even-ordered in eccentricity so that, for example, j = 3 terms will actually be fourth-order. For instance, F (3) 2 e2 i e2 535 = 9 o . Thus, to third-order in eccentricity, the terms in the c(cid:13) 2013 RAS, MNRAS 000, 1–45 20 Mard ling R = R = Gµim3 ao disturbing function associated with the 5 : 3 resonance are 4 A23 (cid:1) e2 (cid:8). . . + (cid:0) 67 8 A03 + 9 4 A13 + 1 o cos(3λi − 5λo + 2o ) − (cid:0)18A04 + 9 2 A14 + 1 2 A24 (cid:1) ei eo cos(3λi − 5λo + i + o ) 4 A25 (cid:1) e2 + (cid:0) 75 8 A05 + 9 4 A15 + 1 i cos(3λi − 5λo + 2i ) + . . .(cid:9) . Note for this example, however, that systems which exist stably in the 5 : 3 resonance tend to have very small values of the mass ratios m2/m1 and m3 /m1 (generally of order 10−4 ), in which case the approximations in Section 3.3.2 are reasonable. One then obtains α2 d2 b(3) ao (. . . + 67 dα2 ! e2 1 Gm2m3 1/2 b(3) o cos(3λi − 5λo + 2o ) 1/2 (α) + 4 8 α2 d2 b(4) − 18 b(4) dα2 ! ei eo cos(3λi − 5λo + i + o ) 9 1/2 1/2 (α) + 2 α2 d2 b(5) db(5) i cos(3λi − 5λo + 2i ) + . . .) . dα2 ! e2 + 75 1 1/2 1/2 dα 4 8 On the other hand, one may wish, for example, to estimate the width of any of the 5 : 3 resonances for some arbitrary configuration (which may or may not be stable) in which case the expression (97), valid for arbitrary mass ratios, should be used. As discussed in the previous section, in general there are n′ − n + 1 distinct principal resonances associated with the n′ : n resonance, and each of these has n′ − n + 1 terms contributing to the lowest order in eccentricity. b(5) 1/2 (α) + db(3) 1/2 dα db(4) 1/2 dα (98) (97) 9 4 α 9 4 α + + + 1 2 α 3.6 First-order resonances R = Gµim3 ao Now consider those harmonic terms in the expansion which are first-order in eccentricity, that is, those terms for which j = 0 or 1 and for which m − n + m − n′ = n′ − n = 1 so that m = n or m = n + 1. The relevant terms in the disturbing function are then (cid:8). . . + 1 2 [(2n + 1)A0n + A1n ] eo cos(nλi − (n + 1)λo + o ) − (cid:2)(n + 1)A0n+1 + 1 2 A1n+1 (cid:3) ei cos(nλi − (n + 1)λo + i ) + . . .(cid:9) , and when m2 ≪ m1 , this reduces to db(n) 2 "(2n + 1)b(n) dα − 4α δn1 # eo cos(nλi − (n + 1)λo + o ) ao (. . . + Gm2m3 1 1/2 1/2 (α) + α db(n+1) dα # ei cos(nλi − (n + 1)λo + i ) + . . .) − "(n + 1)b(n+1) α 1/2 1/2 2 which is consistent with Papaloizou (2011). Here the term involving δn1 comes from (82) and (83). R = (α) + (100) (101) (99) 3.7 Resonance widths using the literal expansion The literal expansion is especially suited to the study of systems with period ratios close to unity. Stable systems in this category tend to have small mass ratios and at most modest eccentricities, with the induced (forced) contributions to the latter being of order m3 /m1 and m2/m1 for the inner and outer eccentricities respectively (Paper II). In this Section we derive an expression for the width of a general [n′ : n](m) resonance, however, we will assume that the eccentricities are not so small that the apsidal advance of one or both orbits contributes significantly to the resonance width; this case will be considered elsewhere. Having said this, one should keep in mind the convergence issues discussed in the previous section. Following the analysis in Section 2.3 for the resonance width in the case of the spherical harmonic expansion, including c(cid:13) 2013 RAS, MNRAS 000, 1–45 General coplanar systems 21 mnn′ ) sin φmnn′ Ajm F (j ) the assumptions that the dynamics is dominated by a single harmonic (not always true when the period ratio is close to 1) and that i and o are negligible compared to νo , the librating angle φmnn′ is governed by φmnn′ = n νi − n′ νo ao (cid:19) νo (cid:18)nσ ai ai − n′ ao 3 = − 2 12 (cid:19)# jmax o (3 n2 "α σ 2 (cid:18) m3 n (cid:19)2 (cid:18) m1m2 m12 (cid:19) + (cid:18) n′ Xj=0 = ν 2 m2 ≡ −ω 2 mnn′ sin φmnn′ , (102) where jmax > m − n + m − n′ and is chosen to equal the highest order in eccentricity required. Libration is around φmnn′ = 0 mnn′ > 0 and about φmnn′ = π if ω 2 if ω 2 mnn′ < 0. The angle φmnn′ will librate when φmnn′ = nνo (σ − n′ /n) < 2 ωmnn′ , so that the width of the [n′ : n](m) resonance, that is, the maximum excursion of σ away from n′ /n is given by jmax n (cid:19) (cid:12)(cid:12)(cid:12)(cid:12)(cid:12) mnn′ (cid:12)(cid:12)(cid:12)(cid:12)(cid:12) ∆σmnn′ = 2√3 (cid:18) n′ m12 (cid:19) + (cid:18) m1m2 (cid:20)α (cid:18) m3 12 (cid:19)(cid:21) Ajm F (j ) Xj=0 m2 where we have put σ = n′ /n and α should be replaced by its value at exact commensurability. 3.7.1 Libration frequency (104) (103) 1/2 , As discussed in Section 2.3.1, the libration frequency of a resonant harmonic is given by ωmnn′ = νo∆σmnn′ /2. (105) 3.7.2 Widths of first-order resonances 1/2 The widths of the two principal first-order resonances are then ∆σn n n+1 = 2√3 (cid:18) n + 1 n (cid:19) (cid:12)(cid:12)(cid:12)(cid:12) 12 (cid:19)(cid:21) [(2n + 1)A0n + A1n ] eo (cid:12)(cid:12)(cid:12)(cid:12) m12 (cid:19) + (cid:18) m1m2 2 (cid:20)α (cid:18) m3 1 m2 and ∆σn+1 n n+1 = 2√3 (cid:18) n + 1 n (cid:19) (cid:12)(cid:12)(cid:12)(cid:12) 2 A1n+1 (cid:3) ei (cid:12)(cid:12)(cid:12)(cid:12) m12 (cid:19) + (cid:18) m1m2 12 (cid:19)(cid:21) (cid:2)(n + 1)A0n+1 + 1 (cid:20)α (cid:18) m3 m2 3.8 The secular disturbing function in the literal expansion As for the spherical harmonic expansion (see Section 2.4), the secular disturbing function is obtained by retaining the n = n′ = 0 terms only in (70). Again using the notation R for the averaged disturbing function, we obtain ∞ R = Rm cos [m(i − o )] , Xm=0 where now (107) (106) (108) 1/2 . Gµim3 ao Rm ≡ Rm00 = Ajm (α; β2 ) F (j ) m00 (ei , eo ). ∞ Xj=0 As before, Alm and F (j ) m00 are given by (72) and (68) respectively, but note that unlike for general n, n′ , closed-form expressions (ei ) and X −(k+1),m exist for the Hansen coefficients X k,m (eo ); these are given in Appendix B. However, since the literal 0 0 formulation for the disturbing function involves an expansion in the eccentricities, it is still only correct to order jmax in the eccentricities, where jmax is the highest value of j included in the expansion (see discussion in Section 3.2). Note that since m − n + m − n′ = 2m is even, all terms in the secular expansion are even order in eccentricity (including products of odd powers). (109) c(cid:13) 2013 RAS, MNRAS 000, 1–45 22 Mard ling dei dt Gµim3 ao Recall from Section 2.4 on the spherical harmonic secular disturbing function that to octopole order (ie., including l = 2 with m = 0, 2 and l = 3 with m = 1, 3), only terms with m = 0 and m = 1 are non-zero because X 2,2 0 (eo ) = 0 and X 3,3 0 (eo ) = 0. Thus the only secular harmonic angle appearing in the spherical harmonic development up to octopole level is φ100 = i − o . From the point of view of the literal expansion, the coefficients of the “quadrupole” and “octopole” harmonic angles φ200 = 2(i − o ) and φ300 = 3(i − o ) are O(e2 i e2 o ) and O(e3 i e3 o ) respectively, and are non-zero because values of j in addition to j = 2 and j = 3 contribute to them. The literal planar secular disturbing function to second-order in the eccentricities and correct for any mass ratios is then R = i e2 2 ei eo (A01 − A11 − A21 ) cos(i − o )(cid:3) + O(e2 i + e2 2 (e2 (cid:2)A00 + 1 o )(A10 + A20 ) + 1 o ) so that to first-order in the eccentricities, the rates of change of the eccentricities and longitudes of the periastra are, from Lagrange’s planetary equations (D1) and (D2), 2 νi (cid:18) m3 m12 (cid:19) eo α(A01 − A11 − A21 ) sin(i − o ), = 1 2 νi (cid:18) m3 m12 (cid:19) (cid:20)2α(A10 + A20 ) + (cid:18) eo ei (cid:19) α(A01 − A11 − A21 ) cos(i − o )(cid:21) , = 1 2 νo (cid:18) m1m2 12 (cid:19) ei (A01 − A11 − A21 ) sin(i − o ), = − 1 m2 2 νo (cid:18) m1m2 eo (cid:19) (A01 − A11 − A21 ) cos(i − o )(cid:21) . 12 (cid:19) (cid:20)2(A10 + A20 ) + (cid:18) ei do = 1 m2 dt In the limiting case that m2 /m1 ≪ 1, the disturbing function to second-order in the eccentricities becomes Gm2m3 1/2 (α) cos(i − o )i , h 1 4 ei eo (cid:0)2 − 2αD − α2D2 (cid:1) b(1) 2 b(0) o ) (cid:0)2αD + α2D2 (cid:1) b(0) R = 8 (e2 i + e2 1/2 (α) − 1 + 1 1/2 (α) + 1 ao where D ≡ d/dα. This is consistent with equations (6.164–6.168) in Murray & Dermott (2000) except for the additional term −Gm2m3 /ao which corresponds to their indirect term. Since in the secular case this term is constant, it contributes nothing to the secular dynamics. di dt deo dt (112) (115) (110) (111) (113) (114) 4 COMPARISON OF FORMULATIONS TO LEADING ORDER IN ECCENTRICITIES In this Section we compare the two formulations in terms of the mass parameter β2 = m2/m12 and the ratio of semima jor axes α. The parameter β2 is chosen because it is taken to be zero in the classic literal expansion and is introduced here without restriction. For each harmonic coefficient considered, the eccentricity dependence is factored out to leading order and the resulting functional dependence on β2 is compared. The dependence on α of the resulting expression is therefore exact in the literal case (to leading order in eccentricity), while in the spherical harmonic case it will depend on the number of terms included in the summation over l. We start by defining the function Smnn′ (α, β2 ) such that the harmonic coefficients are given to leading order in the eccentricities by Gµim3 ao Smnn′ (α, β2 )em−n Rmnn′ ≃ i where for the spherical harmonic expansion, em−n′ o , Smnn′ (α, β2 ) = n x−(l+1),m lm Ml αl xl,m ζm c2 n′ lmax Xl=lmin lmax Xl=lmin with lmin given by (25) and lmax = lmin or lmin + 2 (cases I and II), while for the literal expansion, lm [(1 − β2 )l−1 − β l−1 ζm c2 2 n x−(l+1),m ] αl xl,m n′ = Smnn′ (α, β2 ) = (case III). Here m−n+m−n′ Xj=0 Ajm (α; β2 ) f (j ) mnn′ (116) (117) (118) c(cid:13) 2013 RAS, MNRAS 000, 1–45 General coplanar systems 23 Figure 2. Comparison of the dependence of harmonic coefficients on the inner mass ratio for the two expansions for various harmonics. The function Smnn′ (α; β2 ) is given by (117) for the spherical harmonic expansion and by (118) for the literal expansion. One (blue dashed curves) and two (red dashed curves) values of l are included in the spherical harmonic expansion, while the literal expansion (black curves) is exact to the leading order in eccentricity. For an N : 1 harmonic, α = N −2/3 so that (a): α = 0.63, (b): α = 0.48, (c): α = 0.40 and (a): α = 0.22. xl,m n = lim ei→0 n (ei ) e−m−n X l,m i , x−(l+1),m n′ = lim eo→0 X −(l+1),m n′ (eo ) e−m−n′ o and f (j ) mnn′ = lim ei→0 lim eo→0 mnn′ (ei , eo )e−m−n F (j ) i e−m−n o (119) (120) em−n′ , em−n′ (eo ) and F (j ) n (ei ), X −(l+1),m and em−n are the coefficients of em−n in expansions of X l,m mnn′ (ei , eo ) respec- o o i i n′ tively. Figure (2) compares Smnn′ (α, β2 ) for these three cases for various 2 : 1, 3 : 1, 5 : 1 and 10 : 1 principal resonances, while Figure (3) does the same for the two 3 : 2 and 4 : 3 principal resonances (see Section 3.4 for their definition). The biggest errors incurred are associated with β2 = 0, and these decrease to zero for β2 = 0.5. The main conclusion one draws from these comparisons is that even for systems with period ratios as low as 1.5, including only the first two values of l in the spherical harmonic expansion produces quite accurate estimates of the harmonic coefficients. Since the spherical harmonic expansion is the simplest to use of the two expansions, it is recommended for use in preference to the literal expansion except when the period ratio is less than, say, 2. Finally, while one might hope that the error incurred truncating the spherical harmonic expansion at αlmax is O(αlmax+2 ), this is by no means guaranteed. In fact, for the harmonics plotted in Figures (2) and (3), the error appears to be more like O(αlmax ) for β2 = 0, decreasing with increasing β2 . For example, for the [5 : 1](2) harmonic shown in Figure 2(c), α = 5−2/3 = 0.34 so that including only the l = 2 term in the spherical harmonic expansion (blue dashed line associated with m = 2) incurs an error at β2 = 0 of ∆S215 /S215 = O(α2 ) ≃ 0.012. Thus ∆S215 ≃ 0.6, consistent with the figure. c(cid:13) 2013 RAS, MNRAS 000, 1–45 24 Mard ling Figure 3. Similar to Figure 3 for two first-order harmonics. Two values of l are inadequate for such close systems when β2 = m2 /m12 is small. Here (a): α = 0.76 and (b): α = 0.83. 5 EQUIVALENCE OF FORMULATIONS We now demonstrate the equivalence of the spherical harmonic and literal expansions, which amounts to demonstrating the equivalence of the individual coefficients Rmnn′ given by (27) and (71). Doing this will involve changing summation orders as well as a change of variable. Throughout, one should keep in mind that the indices m, n, n′ are fixed. Our aim is to show that ∞ Xl=lmin ,2 n (ei )X −(l+1),m lmMl αl X l,m c2 n′ Ajm (α; β2 ) F (j ) mnn′ (ei , eo ), Gµim3 ao Gµim3 ao Rmnn′ = ∞ Xj=0 (eo ) = (121) (122) 2, m = 0 3, m = 1 m, m > 2. where again, lmin =   Both Ajm and F (j ) mnn′ can be expressed as series given respectively by (72) and (73) with (C12) and (C13), and (68). Noting the form of (122), consider first m > 2. Distinguishing the left and right sides of (121) by R(L) mnn′ and R(R) mnn′ , we have ∞ (−1)l+m (l − m)!(l + m)! Gµim3 2 (cid:17) αlX l,m 22l−1 [((l + m)/2)!((l − m)/2)!]2 (cid:16)β l−1 n (ei )X −(l+1),m R(L) Xl=m,2 1 − β l−1 mnn′ = n′ ao where we have used respectively (A5) and (9) to replace c2 lm and Ml , and ∞ ∞ (cid:3) αm+2p F (j ) Xp=pmin Xj=0 − βm+2p−1 (cid:2)βm+2p−1 mnn′ (ei , eo ) 2 1 ∞ ∞ Gµim3 2(2m + 2p)!(m + 2p)!(2p)! (cid:3) αm+2p F (j ) Xj=0 Xp=pmin 42p+m j !(m + 2p − j )![p!(m + p)!]2 (cid:2)βm+2p−1 − βm+2p−1 mnn′ (ei , eo ), 1 2 ao with pmin = max (cid:0)0, ⌊ 1 2 (j − m + 1)⌋(cid:1) and ⌊ ⌋ denoting the nearest lowest integer. Referring to Figure 4(a), in the next step we change the order of summation of j and p in (124), then make a change of variable for p putting l = m + 2p so that 2p+m ∞ Gµim3 2(2m + 2p)!(m + 2p)!(2p)! R(R) (cid:3) αm+2p F (j ) Xj=0 Xp=0 42p+m j !(m + 2p − j )![p!(m + p)!]2 (cid:2)βm+2p−1 − βm+2p−1 mnn′ = (125) mnn′ (ei , eo ) 1 2 ao ∞ l (l + m)! l! (l − m)! i αl F (j ) 22l−1 j !(l − j )! [((l − m)/2)!((l + m)/2)!]2 hβ l−1 Xl=m,2 Xj=0 1 − β l−1 mnn′ (ei , eo ) 2 R(R) mnn′ = Gµim3 ao Gµim3 ao E (j,m) p (124) (123) = = (eo ), c(cid:13) 2013 RAS, MNRAS 000, 1–45 p (a) j<m j>m = ( j − m p + 1 ) / 2 j = m + p 2 m General coplanar systems 25 (b) k l k=j j=k j l j Figure 4. Changing the order of summation. (a): Equations (124) and (125). Notice how the summation boundary is stepped in the original order (solid red line), and smooth when the order is changed (dashed red line). (b): Equations (128) and (129). = Gµim3 ao ∞ i αl " l lm hβ l−1 Xl=m,2 Xj=0 c2 1 − β l−1 2 ∞ i αl χlm lm hβ l−1 Xl=m,2 c2 1 − β l−1 nn′ . 2 Note that (−1)l+m = 1 because l + m is always even in the coplanar case. Comparing (126) with (121), it remains to show that n (ei )X −(l+1),m the expression in the large square brackets in (126) which we have defined as χlm nn′ in (127) is equal to X l,m (eo ). n′ Using the definition of F (j ) mnn′ (ei , eo ) from (68), we have mnn′ (ei , eo )# F (j ) l! j !(l − j )! Gµim3 ao (126) (127) ≡ (eo ). X k,m n χlm nn′ = (ei )X −(k+1),m n′ j l j ! l! Xk=0 Xj=0 (−1)j−k (j − k)!k! j !(l − j )! We therefore need to reduce the double summation over j and k to a single term. With that aim in mind, the next step involves (ei )X −(k+1),m gathering together the coefficients of each individual product X k,m (eo ) by changing the order of summation of n n′ j and k. Referring to Figure 4(b), we then have  (l − j )!(j − k)!  l (−1)j−k Xj=k   (ei )X −(k+1),m δkl (l!/k!) X k,m n n′ (ei )X −(k+1),m n′ χlm nn′ = X k,m n (129) (128) (eo ) l Xk=0 l Xk=0 n (ei )X −(l+1),m = X l,m n′ = l! k! (eo ) (eo ), (130) thereby verifying the equivalence of the formulations for m > 2. When m = 1, β l−1 1 − β l−1 2 = 0 so that the summation over l starts at l = 3, consistent with (123). When m = 0, there is no contribution from l = 0 because of the additional term in the definition of A00 (see (73) and (77)). Thus the formulations are equivalent for all values of m. 6 COMPARISONS WITH CLASSIC EXPANSIONS Kaula expansion Using our notation and setting the inclinations equal to zero, the Kaula (1962) expression for the disturbing function with m2 = 0 in units of energy per unit mass, is ∞ Xl=2 c(cid:13) 2013 RAS, MNRAS 000, 1–45 RK = Rl , (131) 26 Mard ling Rl = (eo ) cos φlmnn′ , l−m+n (ei ) X −(l+1),l−m Clm X l,l−m l−m+n′ where l is the spherical harmonic degree, Rl is of the form (see his equation (10)) ∞ l ∞ ao (cid:19)l ao (cid:18) ai Gm3 Xn=−∞ Xm=0 Xn′=−∞ Clm (which is different to our clm ) is a constant involving lengthy expressions from Kaula (1961), and the harmonic angle is φlmnn′ = (l − m + n)Mi − (l − m + n′ )Mo + (l − m)(i − o ). Note that in this form, l appears in the harmonic angle which is not the case in our spherical harmonic expansion (see equation (21)). In particular, the harmonic angle has four indices while ours has three, so that all four cannot be independent. Of special importance in our formulation is the simple form of the general harmonic angle and the clear relationship between the indices and the natural frequencies in the problem (see Section 3.4 for a discussion of this point). Moreover, by swapping the order of summation over the spherical harmonic indices l and m (equation (24)), they become effectively decoupled with m taking on the role of independent harmonic label (together with n and n′ ), while l retains the role of expansion index. The new spherical harmonic expansion benefits especially from its general dependence on the mass ratios, as well as its simple and evincing dependence on the eccentricities via power series and asymptotic approximations. For the inner eccentricity dependence, power series expansions of the Hansen coefficients associated with the dominant terms are accurate for most eccentricities less than unity (see Table B1 and Figure B1(a)), while for the outer eccentricity dependence, asymptotic expressions demonstrate explicitly the exponential falloff of the harmonic coefficients with period ratio and eccentricity (equations (B5) and (B7)). (132) (133) Literal expansion Basing their analysis on the work of Le Verrier (1855), a lengthy derivation of the literal expansion of the direct and indirect parts of the disturbing function for the restricted problem is given in Murray & Dermott (2000) to second order in the eccentricities and inclinations, and for either m2 = 0 or m3 = 0. No general expression for the harmonic coefficients is given,14 but rather several tables of the harmonic angles and their coeffiencients up to fourth order in those elements are provided. One of the features of our formulation which simplifies the analysis is the fact that the expression for the general harmonic coefficient (71) involves derivatives of b(m) 1/2 only, rather than b(m) for many values of the half-integer index s (although the s derivatives themselves involve values of s other than s = 1/2 via (C8) and (C9)). This together with simple expressions for the eccentricity functions makes it straightforward to write down any harmonic coefficient (see Appendix B which gives series approximations for Hansen coefficients as well as a short Mathematica program which generates a power series for F (j ) mnn′ (ei , eo ) to the order of the expansion). We have demonstrated in Sections 3.5 and 3.6 the ease with which this can be done. It has previously been assumed that it is not possible to use Jacobi coordinates for a literal expansion of the disturb- ing function without performing an expansion in the mass ratios as well (see, for example, the discussion on page 195 of Laskar & Robutel (1995)). Here we avoid such an expansion by taking advantage of the symmetry in the mass ratios m1 /m12 and m2/m12 in our form of the disturbing function (52). The novelty is in the use of two mass-weighted ratios of the semima jor axes as the arguments of Laplace coefficients. Note that the usual form for the disturbing function (in units of energy per unit mass) is R3 (cid:21) , Ri = Gmi (cid:20) r · R 1 R − r − where mi is m2 ≪ m1 or m3 ≪ m1 , that is, there are two distinct disturbing functions, each with a “direct” and “indirect” term (the first and second terms respectively). (134) 14 A general expression for the coefficients of a hybrid Kaula-literal expansion (Ellis & Murray 2000) is given, with many conditions on the maximum and minimum values of the summation indices. The derivation of this expression is not given, and a promise of such a derivation does not appear to have been met. In addition, there is no discussion of the errors associated with the expansion. Note that there appears to be a typographical error in equation (52) of Ellis & Murray (2000) which has carried over to equation (6.113) of Murray & Dermott (2000): the index j on the Laplace coefficient should be k . c(cid:13) 2013 RAS, MNRAS 000, 1–45 General coplanar systems 27 7 CONCLUSION AND HIGHLIGHTS OF NEW RESULTS The aim of this paper has been to provide new general expansions of the disturbing function which are clear and accessible to anyone contributing to the rapidly expanding field of exoplanets, as well as to many other fields of astrophysics. The expansions are applicable to systems with arbitrary mass ratios, eccentricities and period ratios, making them suitable for the study of any of the diverse stellar and planetary configurations now being discovered in great numbers by surveys such as HARPS and Kepler. The many applications include determining the rates of change of the orbital elements of both secular and resonant systems; calculating the widths of resonances for the purpose of studying libration cycles and their effect on TTVs, or for the purpose of studying the stability characteristics of arbitrary configurations (not just those with small eccentricities and masses); deriving analytical constraints for use in orbit fitting procedures; and calculating the dynamical characteristics of circumbinary planetary systems. Several new results and concepts have been introduced here including (i) Arbitrary dependence of the disturbing function on the mass ratios for both the spherical harmonic and literal expansions; (ii) Simple general expressions for all harmonics to arbitrary order in the ratio of semima jor axes (spherical harmonic expansion) and the eccentricities (literal expansion); (iii) Accurate and simple approximations for Hansen coefficients for 0 6 ei 6 1 and 0 6 eo 6 1 for the dependence of both expansions on eccentricities, including asymptotic expressions (Paper II) for the outer eccentricity which reveal the exponential dependence of the disturbing function on n′ and eo ; (iv) The fact that for a given level of accuracy, the order in eccentricity at which one truncates the series depends on the configuration being studied. For example, given the eccentricities, one requires fewer terms when studying the 2 : 1 resonance than one does for the 7 : 6 resonance, even though they are both first order resonances; (v) The equivalence of the spherical harmonic and literal expansions revealing the role of the spherical harmonic order m in the literal expansion; (vi) The concept of “principal resonances” and the physical importance of the spherical harmonic order m including “Zeeman splitting” of resonances; (vii) Comparison of the two expansions showing that the simpler spherical harmonic expansion can be used for problems with period ratios as low as 2. This work has revealed that the link between the three-body problem and spherical harmonics is more than just a convenient way to label Fourier terms. Via analogy with other physical problems involving spherical harmonics, the analysis presented here has the potential to expose deep symmetries in this rich problem. 8 QUICK REFERENCE This section provides a quick reference to the main results for readers mainly interested in their application. Equation numbers corresponding to the main text are provided. The paper derives two expansions, one in the ratio of semima jor axes (the spherical harmonic expansion), and the other in the eccentricities (the literal expansion). Both are valid for any masses. The choice of which to use depends on the configuration being studied as well as the application, and there is no clear boundary between them. In general one uses the expansion in eccentricity for systems with period ratios less than, say, two or three and for which the eccentricities are small, while the expansion in semima jor axis ratio is best for wider eccentric systems. For both expansions the disturbing function is expressed as a triple Fourier series over the indices m, n and n′ , where the frequencies associated with n and n′ are those of the inner and outer orbits, and the frequency associated with m is the difference in the apsidal motion rates. We write this as ∞ ∞ ∞ Xn′ =−∞ Xn=−∞ Xm=0 where the harmonic angle φmnn′ can be written in terms of the mean anomalies of the inner and outer orbits, Mi and Mo , or the corresponding mean longitudes λi and λo , as well as the longitudes of periastron i and o , so that φmnn′ = nMi − n′Mo + m(i − o ) = nλi − n′ λo + (m − n)i − (m − n′ )o . c(cid:13) 2013 RAS, MNRAS 000, 1–45 Rmnn′ (ei , eo , ai , ao ; m1 , m2 , m3 ) cos φmnn′ , R = (26), (70) (23), (69) 28 Mard ling The harmonic coefficients Rmnn′ depend on the other parameters in the problem, namely the inner and outer eccentricities ei and eo , the semima jor axes ai and ao , and the masses m1 , m2 and m3 , and are given below for the semima jor axis and eccentricity expansions respectively. 8.1 Rmnn′ for the semima jor axis expansion In this case, the harmonic coefficients are given by Gµim3 ao (eo ) = Gµim3 Rp n (ei )Z−(l+1),m lm Ml ρl X l,m ζm c2 n′ (eo ). Rmnn′ = n (ei )X −(l+1),m lm Ml αl X l,m ζm c2 n′ lmax Xl=lmin ,2 lmax Xl=lmin ,2 where α = ai/ao , ρ = ai /Rp with Rp the outer periastron distance, lmin = m for m > 2 and 2 or 3 if m = 0 or 1 respectively, lmax is chosen according to the accuracy required, the notation Pl=lmin ,2 means the summation is in steps of 2, ζm takes on the values 1/2 or 1 according to whether m is zero or not zero respectively, c2 20 = 1/2, c2 22 = 3/4, c2 31 = 3/8, c2 33 = 5/8 with a general expression given by (12), and 1 + (−1)lml−1 ml−1 2 (m1 + m2 )l−1 (27) (28) Ml = . (9) (eo ) = (1 − eo )l+1X −(l+1),m (eo ) are Hansen coefficients and Z−(l+1),m n (ei ) and X −(l+1),m The eccentricity functions X l,m (eo ) n′ n′ n′ is a modified Hansen coefficient. These are defined and discussed in Appendix B. The leading terms in their Taylor expansions ) and O(em−n′ are O(em−n ) respectively. They may be evaluated numerically with as much precision as one requires, by o i closed-form expression when n = n′ = 0, or approximately by series expansion (Sections B2 and B4) or closed-form asymptotic approximation (equation (B5)). 8.1.1 Secular disturbing function to octopole order The secular disturbing function R is given by (27) with n = n′ = 0. To octopole order this is " 1 cos(i − o )# . ao (cid:19)3 (cid:18) m1 − m2 ao (cid:19)2 2 e2 4 e2 m12 (cid:19) ei eo (1 + 3 1 + 3 i ) 16 (cid:18) ai 4 (cid:18) ai 15 Gµim3 i R = o )3/2 − o )5/2 (1 − e2 (1 − e2 ao The secular rates of change of the eccentricities and apsidal longitudes are given by (48) to (51). Note that care should be taken when using secular expansions; see the discussion in Section 1.1. (47) 8.1.2 Dominant non-secular terms The dominant non-secular harmonics for systems well represented by the semima jor axis expansion tend to be those with m = 2 and n = 1. Including only l = 2 in (27) and using Table B1 for X 2,2 1 (ei ) and the asymptotic approximation (B5) for X −3,2 (eo ) (with accuracies indicated in Table B1 and Figure B3 respectively), their coefficients are approximately n′ ao (cid:18) H22√2π (cid:19) α2 (3ei − 13 Gµim3 o )3/4n′3/2 e−n′ ξ(eo ) e−2 i )(1 − e2 8 e3 R21n′ = − o , where ξ (eo ) = Cosh−1 (1/eo ) − √1 − e2 o and H22 = 0.71.15 Note the steep dependence on n′ ξ (eo ). Note also that at this order of the expansion (quadrupole) there is no dependence on the inner mass ratio (apart from the factor µi ) because M2 = 1. Most systems down to a period ratio of around 2 are well approximated by the spherical harmonic expansion with only one or two values of l included (see Section 4). (135) 15 Note the additional scaling factors indicated in Figure B3 for n′ 6 10. c(cid:13) 2013 RAS, MNRAS 000, 1–45 8.1.3 Widths and libration frequencies of [N : 1](2) resonances The spherical harmonic expansion is especially useful for studying the stability properties of eccentric systems with moderate mass ratios (Paper II). Such systems tend to have significant period ratios and hence can be quite accurately truncated at the quadrupole level. For a system with period ratio close to the integer value N , the harmonic angle of interest is the one corresponding to n = 1, n′ = N and m = 2, that is, General coplanar systems 29 φ21N = λi − N λo + i + (N − 2)o . Using the general notation [n′ : n](m) for an n′ : n resonance of spherical harmonic order m, the width of the [N : 1](2) resonance is approximately 12 (cid:19)#1/2 e1/2 (2π )1/4 "(cid:18) m3 eo ! (1 − 13 m123 (cid:19)2/3 (cid:18) m1m2 6H1/2 m123 (cid:19) + N 2/3 (cid:18) m12 i )1/2 (1 − e2 24 e2 o )3/8N 3/4 e−N ξ(eo )/2 . 22 i m2 Note that limeo→0 ∆σN is infinite for N = 1, finite for N = 2 and zero for N > 3. The corresponding libration frequency ωN is ∆σN = (136) (43) ωN = νo∆σN /2, where νo is the outer orbital frequency. 8.2 Rmnn′ for the eccentricity expansion In this case, the harmonic coefficients are given by (44) (71) Rmnn′ = Gµim3 ao (eo ), (ei ) X −(k+1),m n′ F (j ) mnn′ (ei , eo ) = Ajm (α; β2 ) F (j ) mnn′ (ei , eo ), jmax Xj=0 where jmax is the order in eccentricity of the expansion, the eccentricity functions F (j ) mnn′ (ei , eo ) are finite sums of products of Hansen coefficients given by j (−1)j−k j k!X k,m Xk=0 n with the latter defined and discussed in Appendix B, and the Ajm depend on the semima jor axis ratio α and the mass ratios β1 = m1/m12 , β2 = −m2/m12 through Ajm (α; β2 ) = ζm hβ−1 1/2 (α2 )i 1 B(j,m) 2 B(j,m) 1/2 (α1 ) − β−1 for all j , m except when j = m = 0 in which case 2 hβ−1 1/2 (α2 )i + β−1 1 b(0) 2 b(0) 1/2 (α1 ) − β−1 1 β−1 A00 (α; β2 ) = 1 2 . s )b(m) 1/2 (αs ) is a Laplace coefficient defined in (59) and B(j,m) Here αs = βs α, s = 1, 2, b(m) s /j !)(dj /dαj 1/2 (αs ) = (αj 1/2 (αs ). The evaluation of these is discussed in Appendix C, while a Mathematica program for series expansions of F (j ) mnn′ (ei , eo ) is given in Appendix B4. Examples are presented in Sections 3.5 and 3.6 in which terms contributing to the 5 : 3 resonance and general first-order resonances are calculated. (68) (72) (73) 8.2.1 The secular disturbing function to second order in the eccentricities To second-order in the eccentricities the secular disturbing function is Gµim3 R = 2 ei eo (A01 − A11 − A21 ) cos(i − o )(cid:3) + O(e2 2 (e2 i e2 i + e2 o )(A10 + A20 ) + 1 (cid:2)A00 + 1 o ) ao which reduces to (115) in the limit that m2 /m12 → 0. Equations governing the rates of change of the elements are given in (111) to (114). Note that care should be taken when using secular expansions; see the discussion in Section 1.1. (110) c(cid:13) 2013 RAS, MNRAS 000, 1–45 30 Mard ling 8.2.2 Widths and libration frequencies of [n′ : n](m) resonances 1/2 , The widths of the principal harmonics of the n′ : n resonance (those which are lowest order in the eccentricities) are given by n (cid:19) (cid:12)(cid:12)(cid:12)(cid:12)(cid:12) mnn′ (cid:12)(cid:12)(cid:12)(cid:12)(cid:12) jmax ∆σmnn′ = 2√3 (cid:18) n′ m12 (cid:19) + (cid:18) m1m2 (cid:20)α (cid:18) m3 12 (cid:19)(cid:21) Ajm F (j ) Xj=0 m2 where n 6 m 6 n′ . The corresponding libration frequencies are ωmnn′ = νo∆σmnn′ /2. To first-order in the eccentricities, the widths of the two principal first-order resonances are ∆σn n n+1 = 2√3 (cid:18) n + 1 n (cid:19) (cid:12)(cid:12)(cid:12)(cid:12) 2 A1n (cid:3) eo (cid:12)(cid:12)(cid:12)(cid:12) m12 (cid:19) + (cid:18) m1m2 12 (cid:19)(cid:21) (cid:2)(n + 1)A0n + 1 (cid:20)α (cid:18) m3 m2 and ∆σn+1 n n+1 = 2√3 (cid:18) n + 1 n (cid:19) (cid:12)(cid:12)(cid:12)(cid:12) 12 (cid:19)(cid:21) [(2n + 1)A0n+1 + A1n+1 ] ei (cid:12)(cid:12)(cid:12)(cid:12) m12 (cid:19) + (cid:18) m1m2 2 (cid:20)α (cid:18) m3 1 m2 Note that expressions (104) to (107) do not include contributions from i and/or o which can be significant for first-order resonances when the eccentricities are very small. (104) (105) 1/2 . 1/2 (106) (107) ACKNOWLEDGMENTS The author wishes to thank Jaques Laskar for suggesting the nomenclature “harmonic angle” for the general (not necessarily resonant) angle occurring in Fourier expansions of the disturbing function, Stephane Udry for a thorough critical reading of the manuscript which improved its clarity for the non-celestial mechanician, and the anonymous referee whose suggestions resulted in significant improvements to the paper. DEDICATION For SDU. REFERENCES Arnol’d, V. I. 1963, Russian Mathematical Surveys, 18, 9 Ballard, S., Fabrycky, D., Fressin, F., et al. 2011, ApJ, 743, 200 Batalha, N. M., Rowe, J. F., Bryson, S. T., et al. 2012, arXiv:1202.5852 Batygin, K., Bodenheimer, P., & Laughlin, G. 2009, ApJL, 704, L49 Batygin, K., & Morbidelli, A. 2012, arXiv:1204.2791 Borucki, W. J., Koch, D. G., Basri, G., et al. 2011, ApJ, 736, 19 Brouwer, D. & Clements, G. M. 1961, Methods of Celestial Mechanics, Academic Press New York and London Chirikov, B. V. 1979, Physics Reports, 52, 263 Correia, A. C. M., Couetdic, J., Laskar, J., et al. 2010, A&A, 511, A21 Correia, A. C. M., Udry, S., Mayor, M., et al. 2005, A&A, 440, 751 Delisle, J.-B., Laskar, J., Correia, A. C. M., & Bou´e, G. 2012, arXiv:1207.3171 Doyle, L. R., Carter, J. A., Fabrycky, D. C., et al. 2011, Science, 333, 1602 Dumusque, X., Pepe, F., Lovis, C., et al. 2012, Nature, November 8 issue Eggenberger, A., Udry, S., & Mayor, M. 2004, A&A, 417, 353 Ellis, K. M., & Murray, C. D. 2000, Icarus, 147, 129 Fabrycky, D. C., Ford, E. B., Steffen, J. H., et al. 2012, ApJ, 750, 114 Fabrycky, D., & Tremaine, S. 2007, ApJ, 669, 1298 Ford, E. B., Kozinsky, B., & Rasio, F. A. 2000, ApJ, 535, 385 Fortney, J. J., & Marley, M. S. 2007, ApJL, 666, L45 c(cid:13) 2013 RAS, MNRAS 000, 1–45 General coplanar systems 31 Giuppone, C. A., Leiva, A. M., Correa-Otto, J., & Beaug´e, C. 2011, A&A, 530, A103 Goldreich, P., & Tremaine, S. 1979, ApJ, 233, 857 Goroff, D. L. 1993, Hist. Mod. Phys. Astron., Vol. 13; also AIP Press Gradstein, I. S. & Ryzhik, I. M. 1980, “Table of Integrals, Series, and Products,” Academic Press, New York. Henrard, J., & Lemaitre, A. 1983, Celestial Mechanics, 30, 197 Holman, M. J., Fabrycky, D. C., Ragozzine, D., et al. 2010, Science, 330, 51 Holman, M., Touma, J., & Tremaine, S. 1997, Nature, 386, 254 Hori, G. 1966, PASJ, 18, 287 Hughes, S. 1981, Celestial Mechanics, 25, 101 Innanen, K. A., Zheng, J. Q., Mikkola, S., & Valtonen, M. J. 1997, AJ, 113, 1915 Jackson, J. D. 1975, Classical Electrodynamics, New York: Wiley, 1975, 2nd ed. Kalas, P., Graham, J. R., Chiang, E., et al. 2008, Science, 322, 1345 Kaula, W. M. 1961, Geophys. J. Roy. Astr. S., 5, 104 Kaula, W. M. 1962, AJ, 67, 300 Kiseleva, L. G., Eggleton, P. P., & Mikkola, S. 1998, MNRAS, 300, 292 Kolmogorov, A. N. 1954, Dokl. Akad. Nauk. SSSR, 98, 527 Kozai, Y. 1962, AJ, 67, 591 Lagrange, A.-M., Gratadour, D., Chauvin, G., et al. 2009, A&A, 493, L21 Landau, L. D., & Lifshitz, E. M. 1969, Course of Theoretical Physics, Oxford: Pergamon Press Laskar, J. 1988, A&A, 198, 341 Laskar, J. 1996, Celestial Mechanics and Dynamical Astronomy, 64, 115 Laskar, J. 2005, Celestial Mechanics and Dynamical Astronomy, 91, 351 Laskar, J., & Bou´e, G. 2010, A&A, 522, A60 Laskar, J., Bou´e, G., & Correia, A. C. M. 2012, A&A, 538, A105 Laskar, J., & Robutel, P. 1995, Celestial Mechanics and Dynamical Astronomy, 62, 193 Le Verrier, U.-J. 1855, Annales de l’Observatoire de Paris, 1, 258 Lin, D. N. C., Bodenheimer, P., & Richardson, D. C. 1996, Nature, 380, 606 Lin, D. N. C., & Papaloizou, J. 1979, MNRAS, 188, 191 Lissauer, J. J., Fabrycky, D. C., Ford, E. B., et al. 2011, Nature, 470, 53 Lithwick, Y., & Wu, Y. 2012, ApJL, 756, L11 Lovis, C., S´egransan, D., Mayor, M., et al. 2011, A&A, 528, A112 Mardling, R. A. 2007, MNRAS, 382, 1768 Mardling, R. A. 2008, Lecture Notes in Physics, Berlin Springer Verlag, 760, 59 Mardling, R. A. 2010, MNRAS, 407, 1048 Mardling, R. A. 2013, Submitted: Paper II Marois, C., Macintosh, B., Barman, T., et al. 2008, Science, 322, 1348 Marois, C., Zuckerman, B., Konopacky, Q. M., Macintosh, B., & Barman, T. 2010, Nature, 468, 1080 Mayor, M., & Queloz, D. 1995, Nature, 378, 355 Mayor, M., Bonfils, X., Forveille, T., et al. 2009, A&A, 507, 487 Mayor, M., Marmier, M., Lovis, C., et al. 2011, arXiv:1109.2497 McArthur, B. E., Benedict, G. F., Barnes, R., et al. 2010, ApJ, 715, 1203 Meisner, T., Wurm, G., & Teiser, J. 2012, A&A, 544, A138 Meschiari, S. 2012, ApJ, 752, 71 Meschiari, S. 2012, arXiv:1210.7757 Mordasini, C., Alibert, Y., Benz, W., & Naef, D. 2009, A&A, 501, 1161 Moser, J. 1962, Nachr. Akad. Wiss. Gottingen II, Math. Phys. KD, 1, 1 Murray, C. D., & Dermott, S. F. 2000, Solar System Dynamics, Cambridge University Press Naef, D., Latham, D. W., Mayor, M., et al. 2001, A&A, 375, L27 Naoz, S., Farr, W. M., Lithwick, Y., Rasio, F. A., & Teyssandier, J. 2011, arXiv:1107.2414 Nesvorn´y, D., Kipping, D. M., Buchhave, L. A., et al. 2012, Science, 336, 1133 Nayfeh, A. H. 1973, Perturbation Methods, Wiley-Interscience, New York Neyfeh, A. H. & Mook, D. T. 1979, Nonlinear Oscillations, John Wiley & Sons, New York Okuzumi, S., Tanaka, H., Kobayashi, H., & Wada, K. 2012, ApJ, 752, 106 c(cid:13) 2013 RAS, MNRAS 000, 1–45 32 Mard ling Orosz, J. A., Welsh, W. F., Carter, J. A., et al. 2012, Science, 337, 1511 Orosz, J. A., Welsh, W. F., Carter, J. A., et al. 2012, ApJ, 758, 87 Paardekooper, S.-J., Leinhardt, Z. M., Th´ebault, P., & Baruteau, C. 2012, ApJL, 754, L16 Papaloizou, J. C. B. 2011, Celestial Mechanics and Dynamical Astronomy, 111, 83 Peale, S. J. 1976, ARAA, 14, 215 Pelupessy, F. I., & Portegies Zwart, S. 2012, arXiv:1210.4678 Pepe, F., Lovis, C., S´egransan, D., et al. 2011, A&A, 534, A58 Pillitteri, I., Wolk, S. J., Cohen, O., et al. 2010, ApJ, 722, 1216 Poincar´e, H. 1892-1899, Les M´ethode Nouvelle de la M´ecanique C´eleste, Paris: Gauthier-Villars Pollard, H. 1966, Mathematical Introduction to Celestial Mechanics, Prentice-Hall Rasio, F. A., & Ford, E. B. 1996, Science, 274, 954 Richardson, L. J., Harrington, J., Seager, S., & Deming, D. 2006, ApJ, 649, 1043 Rivera, E. J., Lissauer, J. J., Butler, R. P., et al. 2005, ApJ, 634, 625 Robutel, P. 1995, Celestial Mechanics and Dynamical Astronomy, 62, 219 Sahlmann, J., S´egransan, D., Queloz, D., & Udry, S. 2011, IAU Symposium, 276, 117 Seager, S., & Deming, D. 2010, ARAA, 48, 631 Steffen, J. H., Fabrycky, D. C., Ford, E. B., et al. 2012, MNRAS, 421, 2342 Takeuchi, T., & Ida, S. 2012, ApJ, 749, 89 Torres, G., Fressin, F., Batalha, N. M., et al. 2011, ApJ, 727, 24 Triaud, A. H. M. J., Collier Cameron, A., Queloz, D., et al. 2010, A&A, 524, A25 Udry, S., & Santos, N. C. 2007, ARAA, 45, 397 Welsh, W. F., Orosz, J. A., Carter, J. A., et al. 2012, Nature, 481, 475 Winn, J. N., Fabrycky, D., Albrecht, S., & Johnson, J. A. 2010, ApJL, 718, L145 Wisdom, J. 1980, AJ, 85, 1122 Wisdom, J. 1982, AJ, 87, 577 Wolfram Research, Inc., Mathematica, Version 8.0, Champaign, IL (2010). Wolszczan, A., & Frail, D. A. 1992, Nature, 355, 145 Wu, Y., & Goldreich, P. 2002, ApJ, 564, 1024 APPENDIX A: SPHERICAL HARMONICS Using the definition of Ylm (θ, ϕ) (Jackson 1975) Ylm (θ, ϕ) = s 2l + 1 4π with P m l (cos θ) eimϕ (l − m)! (l + m)! Xj=0 l j!(−1)l−j x2j l x2j−l−m , = (−1)m 2l l! P m l (x) = (−1)m 2l l! (1 − x2 )m/2 (2j )! (2j − l − m)! (−1)m (1 − x2 )m/2 dl+m (1 − x2 )m/2 dl+m dxl+m (x2 − 1)l = dxl+m 2l l! (−1)l−j l j! l Xj=⌊(l+m+1)/2⌋ where ⌊ ⌋ denotes the nearest lowest integer, we have Ylm (π/2, f ) = s 2l + 1 l (0) eimf ≡ r 2l + 1 (l − m)! P m 4π (l + m)! 4π where from equation (A2), (l + m)/2! (l + m)! l (0) = (−1)(l+m)/2 l P m 2l l! and zero otherwise, so that clm eimf , l + m even (A1) (A2) (A3) (A4) c(cid:13) 2013 RAS, MNRAS 000, 1–45 n /maxei (X l,m Table B1. Hansen coefficients for n = 1 and n = 2, errors Eei ≡ δX l,m n ) and scale factors General coplanar systems 33 l m X l,m 1 −3ei + 13 i + 5 192 e5 8 e3 i (ei ) 2 2 E0.7 , E0.9 0.006, 0.05 (ei ) X l,m 2 i − 65 i + 23 1 − 5 288 e6 16 e4 2 e2 i E0.7 , E0.9 Hlm 0.002, 0.025 0.71 4 3 2 1 i + 79 48 e5 −4ei − 3e3 i 0.006, 0.025 i + 35 i − 43 36 e6 16 e4 1 + e2 i 0.007, 0.06 1.44 i − 41 i − 37 1 + 2e2 576 e6 64 e4 i i + 23 i − 35 0.0007, 0.014 − 1 576 e7 96 e5 2 ei + e3 i 0.002, 0.006 1.91 c2 lm = (l − m)!(l + m)! 22l−1 [((l + m)/2)! ((l − m)/2)!]2 . Note that the association of non-zero values of P m l (0) with even l + m is consistent with the sum in (11) being in steps of 2. Some values of c2 lm are listed in Table B2. (A5) APPENDIX B: HANSEN COEFFICIENTS The two formulations presented in this paper are distinguished by the expansion parameter; for the spherical harmonic expansion the parameter is the ratio of semima jor axes and it places no restrictions on the two eccentricities, while for the literal expansion the parameters are the eccentricities, with no restriction on the ratio of semima jor axes (except that the orbits should not cross). For the spherical harmonic expansion, we therefore require expressions for the Hansen coefficients which are accurate for all eccentricities, while for the literal expansion, power series representations are appropriate because the expansion is valid only to order jmax in the combined powers of the eccentricities. B1 Hansen coefficients relevant for the spherical harmonic expansion Hansen coefficients are defined such that 2π Z 2π 1 0 (r/ai )l eimfi e−inMi dMi X l,m n (ei ) = and (B1) (B2) (eo ) = X −(l+1),m n′ 2π Z 2π e−imfo 1 (R/ao )l+1 ein′Mo dMo . 0 (eo ) = (1 − eo )l+1X −(l+1),m n (ei ) and Z−(l+1),m show the numerically integrated functions X l,m (eo ) Figures (B1) and (B2) n′ n′ for quadrupole and octopole values of l and m, and for the first 10 positive values of n′ and n (see equations (15) and (16)). The scaling factor (1 − eo )l+1 replaces ao with the outer periastron separation Rp = ao (1 − eo ) in the definition (16) of (eo ), factoring out the singularity at eo = 1. We refer to the Z−(l+1),m X −(l+1),m (eo ) as modified Hansen coefficients. While no n′ n′ closed form expressions for these integrals exist (except for n′ = n = 0; see Section B3), for many applications it is reasonable (CPU-wise) to integrate them numerically. However, simple approximations exist as outlined below, the analytic form of which provides insight into the behaviour of the physical variables which depend on them. In Section B2 we give general power series expansions which are correct to fourth order in the eccentricity. Amongst other things, these expressions demonstrate that the leading terms are such that (eo ) = O(em−n′ n (ei ) = O(em−n X l,m i i consistent with Figures B1 and B2. For most applications for which a spherical harmonic expansion of the disturbing function is appropriate, it suffices to know expressions for the Hansen coefficients associated with the [n′ : 1](2) harmonics for l = 2, 4 (see, for example, Sections 2.2 and 2.3), that is, X 2,2 (eo ) and X −5,2 1 (ei ), X −3,2 1 (ei ), X 4,2 (eo ), and perhaps those associated with the [2n′ + 1 : 2](2) harmonics n′ n′ 2n′ +1 (eo ) and X −5,2 (those half-way between the [n′ : 1](2) harmonics), that is, X 2,2 2 (ei ), X 4,2 2 (ei ), X −3,2 2n′ +1 (eo ). Table B1 gives the first few terms of the series expansions of these Hansen coefficients for which ei is the argument, as well as the error at ei = 0.7 and 0.9, defined such that and X −(l+1),m n′ (B3) ) ), c(cid:13) 2013 RAS, MNRAS 000, 1–45 34 Mard ling n=2 n=3 n=1 n=3 n=2 n=0 n=0 n=−1 n=1 n=1 n=−1 n=0 n=−1 n=1 n=0 Figure B1. Hansen coefficients X l,m n (ei ) for quadrupole (l = 2, m = 0, 2) and octopole (l = 3, m = 1, 3) values of l and m, and for various values of n, with dominant values labeled. Notice the O(em−n ) behaviour for small ei . The black curves are for n = 1, 2, . . . , 10, i the blue curves are for n = −10, −9, . . . , −1 and the pink curves are for n = 0 for which closed-form expressions are given in Table B2. The red dashed curves are for the polynomial approximations X 2,2 i . and X 3,1 1 (ei ) ≃ −3ei + 13 i − 41 8 e3 1 (ei ) ≃ 1 + 2e2 64 e4 i (see Table B1). n ] (cid:12)(cid:12)(cid:12)(cid:12)(cid:12) Eei ≡ (cid:12)(cid:12)(cid:12)(cid:12)(cid:12) δX l,m n (ei ) maxei [X l,m where δX l,m n (ei ) is the difference between the numerically integrated expression and the series approximation at the given n ] is the maximum value of X l,m value of ei , and maxei [X l,m over the interval 0 6 ei 6 1. The expansions are correct to O(e6 i ), n except for X 3,1 2 (ei ) which is correct to O(e8 i ) because the errors are an order of magnitude smaller with the extra term. Note that the error decreases monotonically with decreasing ei for each approximation. The functions X 2,2 1 (ei ) and X 3,1 1 (ei) are compared to the numerically calculated integrals in Figure B1, panels (a) and (d) respectively. A good approximation for modified Hansen coefficients governing the dependence of the disturbing function on the outer eccentricity is given by the expression (Paper II) (1 − e2 o )3/4 e2 o (eo ) ≃ (1 − eo )3 · where ξ (eo ) = Cosh−1 (1/eo ) − p1 − e2 o , (B6) and the constant H22 is an empirical scaling factor given in Table B1, determined by comparing the maxima of Z−3,2 (eo ) 20 and Z−3,2 (eo ) and scaling the latter so that the values of their maxima are the same. All other Z−3,2 (eo ) are then scaled 20 n′ by the same factor, except for n 6 10 which involve additional scale factors listed in panel (a) of Figure B3. Note that 4H22 3√2π n′ 3/2 e−n′ ξ(eo ) ≡ Z−3,2 n′ (eo ), Z−3,2 n′ (B4) (B5) c(cid:13) 2013 RAS, MNRAS 000, 1–45 General coplanar systems 35 n’=2 n’=3 n’=4 n’=1 n’=3 n’=4 n’=2 n’=0 n’=1 n’=2 n’=1 n’=2 n’=0 Figure B2. The modified Hansen coefficients Z −(l+1),m (eo ) = (1 − eo )l+1X −(l+1),m (eo ) for quadrupole (l = 2, m = 0, 2) and octopole n′ n′ (l = 3, m = 1, 3) values of l and m, and for various values of n, with dominant values labeled. The black curves are for n = 1, 2, . . . , 10, the blue curves are for n = −10, −9, . . . , −1 (only visible in panel (d)), and the pink curves are for n = 0 for which closed-form expressions are given in Table B2. limeo→0 e−n′ ξ(eo ) e−2 o = 0, n′ > 3, while for n′ = 2 the limit is e2 /2. However, for small values of eo it may be preferable to use a power series approximation for X −3,2 (eo ). n′ The derivation of (B5) uses the method of steepest decents to evaluate the integrals and is therefore referred to as an asymptotic approximation. They are closely related to overlap integrals which quantify the strength of the interaction between the orbits (Paper II). Scale factors are necessary because approximations made in the analysis rely on the values of eo and σ (and hence n′ ) being high, and accuracy is lost when they are not (although the shape of the curves is preserved). Figure B3 compares numerically evaluated integrals with their asymptotic approximations for selected values of n′ between 4 and 100. The asymptotic approximation for general l, m > 0 and n′ > 2 is o )(3m−l−1)/4 (1 − e2 2m ≃ (1 − eo )l+1 · Hlm√2π n′(l+m−1)/2 e−n′ ξ(eo ) ≡ Z−(l+1),m n′ em (l + m − 1)!! o (eo ) = O(en′−m Note that Z−(l+1),m ), n′ > 2, consistent with (B3). o n′ Z−(l+1),m n′ (eo ). (B7) B2 Expansions up to fourth order in eccentricity for use in the literal expansion Using Mathematica or similar, it is easy to derive general power series expansions for X j,m n (e) valid for any integers j and n and for specific values of m. These series contain either even or odd powers of e only. Writing ν = n sgn(m − n), we have for m = n, m = n ± 1, m = n ± 2, m = n ± 3 and m = n ± 4 correct to O(e4 ), 4 (cid:2)j (j + 1) − 4ν 2 (cid:3) e2 + 1 64 (cid:2)(j + 1)j (j − 1)(j − 2) + ν 2 (16ν 2 − (8j 2 + 9))(cid:3) e4 + . . . , X j,n n (e) = 1 + 1 c(cid:13) 2013 RAS, MNRAS 000, 1–45 (B8) 36 Mard ling n’=4 n’=6 n’=8 (a): scaled by CF.H22 n’ 2 3 4 5 6 7 8 9 10 CF 0.52 0.70 0.79 0.84 0.88 0.90 0.92 0.93 0.95 (b): scaled by H = 0.71 22 n’=12 n’=15 n’=20 n’=30 n’=50 n’=100 Figure B3. Modified Hansen coefficients Z −3,2 (eo ) (black curves) and their asymptotic approximations Z −3,2 (eo ) (equation (B5), red n′ n′ dashed curves). In panel (b) the approximations are scaled by correction factor H22 (already included in (B5)), while in panel (a) they are scaled by the additional correction factors ‘CF’, listed at the right of the panel. (B9) X j,n±4 n X j,n±3 n X j,n±2 n X j,n±1 n 16 (cid:2)−j (j + 2)(j − 1) + ν (8ν 2 + 2ν (2j + 7) − j (2j − 3) + 6)(cid:3) e3 + . . . , 2 [(j + 2) + 2ν ] e + 1 (e) = − 1 8 [(j + 2)(j + 3) + ν (4ν + 4j + 11)] e2 (e) = 1 96 (cid:2)(j + 3)(j + 2)(j − 1)(j − 2) − 2ν (8ν 3 + 34ν 2 + 44ν + 13) + jν (4j 2 − 3j − 47 − 16ν (ν + 3))(cid:3) e4 + . . . , (B10) + 1 48 (cid:2)(j + 4)(j + 3)(j + 2) + 2ν (4ν 2 + 21ν + 31) + 3jν (4ν + 2j + 13)(cid:3) e3 + . . . , (e) = − 1 (B11) 384 (cid:2)(j + 5)(j + 4)(j + 3)(j + 2) + ν (16ν 3 + 136ν 2 + 379ν + 394) 1 (e) = +2jν (4j 2 + 45j + 165 + ν (16ν + 12j + 96))(cid:3) e4 + . . . (B12) n (e) = O(em−n ). Comparison of these approximations with numerically evaluations Notice that these are consistent with X j,m of (15) and (16) are shown in Figures B4 to B6. Note that for the outer eccentricity functions, modified Hansen coefficients are calculated. Comparison of Figure B4 with Figure B5 for the 2 : 1 and 7 : 6 resonances respectively suggests that convergence of the series (B8) to (B12) is slower for the 7 : 6 resonance. Inspection of the dependence of these series on n and j shows that the magnitude of the term proportional to eq is O(n e)q or O(j e)q , whichever is the greatest, although in some cases the errors may cancel to come extent (see, for example, the m = 1 curve in panel (d) of Figure B4 which plots Z−1,1 (eo ) = (1 − eo )X −1,1 (eo ), 2 2 with the series expansion of X −1,1 (eo ) given by (B9)). 2 B3 Closed-form expressions Hughes (1981) has provided closed form expressions for Hansen coefficients with n = 0 and n′ = 0. Those associated with the inner orbit are X l,m 0 (ei ) = 2π Z 2π 1 (r/ai )l eimfi dMi 0 m !F (cid:18) m − l − 1 2 (cid:17)m l + m + 1 i (cid:19) , ei = (cid:16)− ; m + 1; e2 2 where F ( ) is a hypergeometric function given here by (see Gradstein & Ryzhik (1980) for a general definition) m − l 2 , (B13) (B14) c(cid:13) 2013 RAS, MNRAS 000, 1–45 General coplanar systems 37 2:1 resonance m=1 m=2 m=1 m=2 m=2 m=1 m=2 m=1 (ei ) and Z −(j+1),m Figure B4. Some Hansen coefficients associated with the first-order 2 : 1 resonance. Plotted are X j,m (eo ) for m = 1 1 2 (red) and m = 2 (blue), and for j = 0 (panels (a) and (b)) and j = 2 (panels (c) and (d)). Solid curves: “exact” integration, dashed curves: fourth-order correct series expansions. Accuracy is good up to at least ei,o = 0.4 for all Hansen coefficients shown here, this value increasing up to 1 in come cases (for example, Z −1,1 (eo )). These should be compared with series expansions for the first-order 7 : 6 2 resonance which are less accurate given the higher values of n and n′ (see Figure B5). , m − l 2 [(l − m + 1)/2 − k] [(l − m)/2 − k] (m + k + 1)(k + 1) (l−m)/2 j−1 F (cid:18) m − l − 1 i (cid:19) = 1 + Yk=0 Xj=1 ; m + 1; e2 2 with F (−1/2, 0; m + 1; e2 i ) = 1 when m = l. Hansen coefficients associated with the outer orbit are 2π Z 2π 1 X −(l+1),m (eo ) = 0 0 ⌊(l−m−1)/2⌋ = (cid:16) eo 2 (cid:17)m Xj=0 (1 − e2 o )−(2l−1)/2 where ⌊ ⌋ denotes the nearest lowest integer. Quadrupole and octopole eccentricity functions are listed in Table B2. j ! (cid:18) e2 2j + m! 2j + m l − 1 2 (cid:19)j o e−imfo (R/ao )l+1 dMo e2j , , (B15) (B16) (B17) c(cid:13) 2013 RAS, MNRAS 000, 1–45 38 Mard ling 7:6 resonance m=6 m=7 m=6 m=7 m=7 m=6 m=7 m=6 (ei ) and Z −(j+1),m Figure B5. Some Hansen coefficients associated with the first-order 7 : 6 resonance. Plotted are X j,m (eo ) for m = 6 6 7 (red) and m = 7 (blue), and for j = 0 (panels (a) and (b)) and j = 4 (panels (c) and (d)). Solid curves: “exact” integration, dashed curves: fourth-order correct series expansions. While errors in the series expansions (B8) to (B12) are formally of order of the fifth power of the eccentricity, inspection shows that errors are in fact O(n5 e5 ). Putting 65 e5 = 0.1, the expansions for this case are accurate only for ei,o < ∼ 0.1 for a 10% error. Table B2. Secular Hansen coefficients and c2 lm l m X l,m 0 (ei ) X −(l+1),m 0 (eo ) c2 lm 2 3 2 0 3 1 5 2 e2 i 1 + 3 2 e2 i − 35 8 e3 i 0 (1 − e2 o )−3/2 0 − 5 8 ei (4 + 3e2 i ) eo (1 − e2 o )−5/2 3 4 1 2 5 8 3 8 c(cid:13) 2013 RAS, MNRAS 000, 1–45 General coplanar systems 39 4:1 resonance m=1 m=2 m=3 m=4 m=4 m=3 m=2 m=1 m=1 m=3 m=2 m=4 m=4 m=3 m=2 m=1 (ei ) and Z −(j+1),m Figure B6. Some Hansen coefficients associated with the third-order 4 : 1 resonance. Plotted are X j,m (eo ) for m = 1 1 4 (red), m = 2 (blue), m = 3 (purple) and m = 4 (magenta), and for j = 0 (panels (a) and (b)) and j = 3 (panels (c) and (d)). Solid curves: ‘exact’ integration, dashed curves: fourth-order correct series expansions. While approximations are accurate for ei < ∼ 0.5 for n = 1 (panels (a) and (c)), they are accurate only for eo < ∼ 0.15 for n = 4. As in the case of the 7 : 6 resonance in Figure B5, this is because the error in the expansions is O [n5 e5 ]. c(cid:13) 2013 RAS, MNRAS 000, 1–45 40 Mard ling B4 Mathematica programs for Hansen coefficients and F (j ) mnn′ (ei , eo ) and e−inMi dEi ein′Mo dEo , X l,m n (ei ) = r l eimfi e−inMi dMi = It is the experience of the author that most colleagues in Astronomy have access to the software package Mathematica (Wolfram 2010), but many are not familiar with the syntax. For this reason the following short programs are included here. The first two calculate series approximations for the Hansen coefficients 2π Z 2π 2π Z 2π 1 1 r l+1 heifi im 0 0 (cid:2)e−ifo (cid:3)m 2π Z 2π 2π Z 2π e−imfo 1 1 Rl+1 ein′Mo dMo = Rl 0 0 where Ei is the eccentric anomaly with r = 1 − ei cos Ei , Mi = Ei − ei sin Ei , cos fi = (cos Ei − ei )/(1 − ei cos Ei ) and i sin Ei/(1 − ei cos Ei ), and similarly for R, Mo , cos fo and sin fo . The third program calculates F (j ) sin fi = p1 − e2 mnn′ (ei , eo ) according to (68). Recall that if jmax is the order of the literal expansion (ie., the highest combined powers of the eccentricities), then according to Section 3.2 F (j ) mnn′ (ei , eo ) should be expanded to jmax . X −(l+1),m n (eo ) = (B18) (B19) Mathematica program to calculate X 2,2 i + O(e7 192 e5 8 e3 i + 5 1 (ei ) = −3ei + 13 o ) l = 2; m = 2; n = 1; ne = 5; r = 1 - e Cos[EA]; M = EA - e Sin[EA]; cosf = (Cos[EA] - e)/(1 - e Cos[EA]); sinf = Sqrt[1 - e2] Sin[EA]/(1 - e Cos[EA]); expnM = Cos[n M] - I Sin[n M]; Xlmn = Normal[Series[r(l + 1) (cosf + I sinf)m expnM, {e, 0, ne}]]; Xlmnav = Simplify[Integrate[Xlmn/(2 Pi), {EA, 0, 2 Pi}]] Mathematica program to calculate X −3,2 2 o + O(e6 16 e4 2 e2 o + 13 (eo ) = 1 − 5 o ) l = 2; m = 2; n = 2; ne = 4; R = 1 - e Cos[EA]; M = EA - e Sin[EA]; cosf = (Cos[EA] - e)/(1 - e Cos[EA]); sinf = Sqrt[1 - e2] Sin[EA]/(1 - e Cos[EA]); expnM = Cos[n M] + I Sin[n M]; Xlmn = Normal[Series[(cosf - I sinf)m expnM/Rl, {e, 0, ne}]]; Xlmnav = Simplify[Integrate[Xlmn/(2 Pi), {EA, 0, 2 Pi}]] Mathematica program to calculate F (2) 16 ei e3 16 e3 o + O(e6 ) correct to 4th order. Notice that the 435 (ei , eo ) = − 1 i eo − 97 2 ei eo − 71 leading term is of order j = 2, and that all possible combinations of the powers of ei and eo are present given the constraints that only odd powers of ei > m − n = 1 and eo > m − n′ = 1 can appear. Terms of O(e6 ) should be discarded. jmax=4; j = 2; m = 4; c(cid:13) 2013 RAS, MNRAS 000, 1–45 General coplanar systems 41 n = 3; no = 5; nei = Max[Abs[m - n], jmax - Abs[m - no]]; neo = Max[Abs[m - no], jmax - Abs[m - n]]; r = (1 - ei Cos[EA]); M = EA - ei Sin[EA]; cosf = (Cos[EA] - ei)/(1 - ei Cos[EA]); sinf = Sqrt[1 - ei2] Sin[EA]/(1 - ei Cos[EA]); expnM = Cos[n M] - I Sin[n M]; R = (1 - eo Cos[EAo]); Mo = EAo - eo Sin[EAo]; cosfo = (Cos[EAo] - eo)/(1 - eo Cos[EAo]); sinfo = Sqrt[1 - eo2] Sin[EAo]/(1 - eo Cos[EAo]); expnMo = Cos[no Mo] + I Sin[no Mo]; Fjmnnp = Expand[Simplify[Sum[(-1)(j - k) Binomial[j, k] Integrate[ Normal[Series[r(k + 1) (cosf + I sinf)m expnM, {ei, 0, nei}]]/(2 Pi), {EA, 0, 2 Pi}] Integrate[ Normal[Series[(cosfo - I sinfo)m expnMo/R**k , {eo, 0, neo}]]/(2 Pi), {EAo, 0, 2 Pi}], {k, 0, j}]]] APPENDIX C: LAPLACE COEFFICIENTS C1 Series expansion b(m) s (x) = b(m) s The general Laplace coefficient is defined as π Z 2π e−imψ 1 (1 − 2x cos ψ + x2 )s dψ , 0 where s is a postive half integer. A series expansion for this is ∞ Xp=0 where s > 1/2 is a half integer and x2p = O(xm ), (x) = xm C (m,s) p C (m,s) p = 2Γ(s + p)Γ(s + m + p) [Γ(s)]2 p!(m + p)! . Recall that Γ(n) = (n − 1)!, and Γ(n + 1 2 ) = (2n)! 22n n! √π . For s = 1/2 we have = C (m,1/2) p 2(2p)!(2m + 2p)! 24p+2m [p!(m + p)!]2 . By the ratio test, we have that the series (C2) converges as long as x > 1. c(cid:13) 2013 RAS, MNRAS 000, 1–45 (C1) (C2) (C3) (C4) (C5) (C6) 42 Mard ling C2 Derivatives The following formulae are taken from Brouwer & Clements (1961) p502 (and Murray & Dermott (2000) who take their formulae from Brouwer & Clements (1961)). Defining the operator d dx , D = we have (x) = s hb(m−1) s+1 + b(m+1) s+1 − 2 x b(m) Db(m) s s+1 and in general, (x) = s hDn−1 b(m−1) s+1i , s+1 − 2(n − 1)Dn−2 b(m) s+1 + Dn−1 b(m+1) s+1 − 2 x Dn−1 b(m) Dn b(m) s i n > 2. C2.1 Leading terms Recalling the definition from (61) dj 1/2 (x)i , dxj hb(m) B(j,m) 1/2 we have from (C2) that (x) = xj j ! (x) = 1 j ! = xm+2p xm+2p , C (m,1/2) p B(j,m) 1/2 (m + 2p)! (m + 2p − j )! ∞ Xp=pmin ∞ Xp=pmin E (j,m) p where pmin = max (cid:0)0, ⌊ 1 2 (j − m + 1)⌋(cid:1) with ⌊ ⌋ denoting the nearest lowest integer, and j ! C (m,1/2) = m + 2p 2(2m + 2p)!(m + 2p)!(2p)! E (j,m) = 42p+m j !(m + 2p − j )![p!(m + p)!]2 . p p The leading term of B(j,m) 1/2 (x) is therefore (x) =  E (j,m) xm + . . . , 0 E (j,m) xj + . . . , p∗ E (j+1,m) xj+1 + . . . ,  p∗ 2(2m)! 4m j !(m − j )!m! j 6 m j > m, j − m even j > m, j − m odd B(j,m) 1/2 where E (j,m) 0 = , and E (j,m) p∗ = 2(j + m)!((j − m))! 4j [((j − m)/2)!((j + m)/2)!]2 with p∗ = ⌊(j − m + 1)/2⌋. Note that by the ratio test, the series (C12) converges as long as x < 1. (C7) (C8) (C9) (C10) (C11) (C12) (C13) (C14) (C15) (C16) C3 Graphical representations 1/2 (x) for j = 1, both for m = 0, 1, . . . , 20. Notice how, for most values of m, B(1,m) 1/2 (x) and B(j,m) Figure C1 shows b(m) 1/2 1/2 (0) = 0, m > 1 and B(j,m) 1/2 (0) = 2, b(m) (x) when j1 > j2 . Note that b(0) (x) > B(j2 ,m) 1/2 (x). In general, B(j1 ,m) b(m) 1/2 (0) = 0 for all 1/2 1/2 m. (x) > c(cid:13) 2013 RAS, MNRAS 000, 1–45 General coplanar systems 43 Figure C1. b(m) 1/2 (x) and B(j,m) 1/2 (x) for j = 1, both for m = 0, 1, . . . , 20. In particular, the red curves are for m = 0 and the blue for m = 1. APPENDIX D: LAGRANGE’S PLANETARY EQUATIONS FOR THE VARIATION OF THE ELEMENTS Because Lagrange’s equations for the variation of the elements were developed for the restricted three-body problem (the mass of one of the three bodies of interest is negligible compared to the other two so that one orbit is fixed), they are normally given in terms of a disturbing function which has the dimensions of energy per unit mass (Brouwer & Clements 1961; Murray & Dermott 2000). Here the disturbing function has the dimensions of energy, and as a result the same function can be used for the rates of change of the inner and outer orbital elements, that is, there is no need to define separate inner and outer disturbing functions. Note also that Lagrange’s “planetary” equations hold for any mass ratios, and in particular, there is no assumption about the smallness of R. In spite of the fact that in most applications the disturbing function acts to perturb the Keplerian orbits from invariant elliptical motion, the derivation of Lagrange’s equations does not involve a perturbation technique. Rather it uses the method of variation of parameters, those parameters being the orbital elements which are constant when there is no interaction between the orbits (ie, when R = 0), and vary once the orbits are allowed to interact via a non-zero R. For reference, the relevant Lagrange equations for the rates of change of the elements for coplanar systems are ∂R ∂R ε(1 − ε) ∂λ − µν a2 e ∂ ∂R ε µν a2 e ∂ e ∂R ∂ a = − and ∂R 2 da = , (D4) ∂λ µν a dt where ε = √1 − e2 , and with the set {a, e, , ǫ; µ, ν } representing either of the inner orbit {ai , ei , i , ǫi ; µi , νi } or the outer orbit {ao , eo , o , ǫo ; µo , νo }. Note that in this form, λ and ǫ are related through λ = R t 0 ν dt + ǫ rather than the usual λ = ν t + ǫ (see Brouwer & Clements (1961) page 285 for a discussion of this point, in particular that in using this definition one need not consider ν to be a function of a when evaluating ∂R/∂ a). The former definition should always be used when resonance plays a role, however, note that the two definitions are equivalent when it doesn’t (because in that case the semima jor axes are constant). ε µν a2 e ε(1 − ε) µν a2 e ∂R ∂ e 2 µν a de dt d dt dǫ dt , + = − = , (D1) (D2) (D3) APPENDIX E: THE MEAN LONGITUDE AT EPOCH Finally, a note on the mean longitude at epoch, ǫ ≡ M (T0 ) + , where T0 is the time (“epoch”) at which the osculating elements are determined and M (T0 ) is the mean anomaly at that time. For some applications T0 is taken to be the time at periastron passage, while for others it corresponds to some given time, for example, the mid-time of a particular set of observations. While it is clear what it means for to vary, one might reasonably ask what it means for M (T0 ) to vary. c(cid:13) 2013 RAS, MNRAS 000, 1–45 44 Mard ling p,1T > T p,2 e1 < e2 t=Tp,2 t=Tp,1 t=0 0M(T ) δ = ν( ) T − p,2T p,1 Figure E1. Illustration of the effect on the mean anomaly at epoch, M (T0 ) = ν (T0 − Tp ), of increasing the eccentricity while holding the semima jor axis constant. If the perturbation is applied at t = 0 when M = −ν Tp,1 , then although at the instant the force is applied there is no change in the true anomaly f (the true position in the orbit), there is a change in the mean anomaly and hence the mean anomaly at epoch (since T0 is a fixed time). This is given by δM (T0 ) = ν (Tp,1 − Tp,2 ), where Tp,1 and Tp,2 are the times to periastron before and after the perturbation is applied. To answer this question, consider the usual definition of the mean anomaly. This is given in terms of the orbital frequency (mean motion) ν and the time at periastron passage Tp as M = ν (t − Tp ), or in terms of a general epoch T0 , M = ν (t − T0 ) + ν (T0 − Tp ) = ν (t − T0 ) + M (T0 ). (E1) (E2) Either way, M (Tp ) = 0. Referring to Figure E1, imagine at t = 0 a force acts on the system in such a way that only the eccentricity is changed. If the eccentricity increases, the time to periastron passage, Tp , decreases, and since ν remains unchanged and T0 is fixed, the mean anomaly at epoch, M (T0 ) = ν (T0 − Tp ), must increase. The change in the mean anomaly at epoch must therefore be proportional to the change in the time at periastron, that is, δM (T0 ) = ν (Tp,1 − Tp,2 ), (E3) where Tp,1 and Tp,2 are the times to periastron before and after the perturbation is applied. In general, an arbitrary force acting on the system will cause the mean anomaly at epoch to change by an amount (Pollard 1966, page 36) 2 M − rp1 − e2/(e sin f )i (δa/a) + p1 − e2 cot f (δe/e), δM (T0 ) = h 3 that is, only changes to the osculating eccentricity and semima jor axis affect this quantity (as reflected in Lagrange’s planetary equation (D3) for the rate of change of ǫ). (E4) c(cid:13) 2013 RAS, MNRAS 000, 1–45 APPENDIX F: NOTATION General coplanar systems 45 fi , fo Mi , Mo λi , λo i , o ǫi , ǫo ei , eo ai , ao α = ai /ao αs = βsα, s = 1, 2 Pi , Po νi , νo σ = Po/Pi = νi /νo r R m12 = m1 + m2 m123 = m1 + m2 + m3 β1 = m1 /m12 β2 = −m2 /m12 Ml = β l−1 1 − β l−1 2 µi = m1m2 /m12 µo = m12m3 /m123 R R Rmnn′ φmnn′ φN ≡ φ21N [n′ : n](m) ∆σN ∆σmnn′ ωN , ωmnn′ Ylm (θ , ϕ) c2 lm ζm mmin lmin X l,m n (ei ) Z −(l+1),m (eo ) n′ F (j ) mnn′ (ei , eo ) b(m) 1/2 (αs ) Ajm (α; β2 ) inner and outer true anomalies inner and outer mean anomalies inner and outer mean longitudes inner and outer longitudes of periastron inner and outer mean longitudes at epoch inner and outer eccentricities inner and outer semima jor axes inner and outer orbital periods inner and outer orbital frequencies (mean motions) period ratio position of body 2 relative to body 1 position of body 3 relative to the centre of mass of bodies 1 and 2 reduced mass of inner binary reduced mass of outer binary the disturbing function (interaction energy) the orbit-averaged disturbing function harmonic coefficient harmonic angle harmonic associated with the angle φmnn′ width of [N : 1](2) resonance width of [n′ : n](m) resonance libration frequencies a spherical harmonic of degree l and order m coefficients in the semima jor axis expansion 1 2 if m = 0, 1 if m > 1 0 if l is even, 1 if l is odd 2 if m = 0, 3 if m = 1, m if m > 2 a Hansen coefficient a modified Hansen coefficient linear combination of Hansen coefficients a Laplace coefficient coefficient in eccentricity expansion 6 6 7 6 10 7 7 7 13 9 6 9 5 5 5 5 5 5 6 5 5 5 11 8, 14 7, 14 10 9 11 21 10, 21 6 6 7 7 8 7 8 14 13 14 Numbers in the right-hand column refer to the page where the variable is defined c(cid:13) 2013 RAS, MNRAS 000, 1–45
1607.01397	1	1607	2016-07-05T20:00:32	White dwarf pollution by planets in stellar binaries	[ "astro-ph.EP", "astro-ph.SR" ]	Approximately $0.2 \pm 0.2$ of white dwarfs (WDs) show signs of pollution by metals, which is likely due to the accretion of tidally disrupted planetary material. Models invoking planet-planet interactions after WD formation generally cannot explain pollution at cooling times of several Gyr. We consider a scenario in which a planet is perturbed by Lidov-Kozai oscillations induced by a binary companion and exacerbated by stellar mass loss, explaining pollution at long cooling times. Our computed accretion rates are consistent with observations assuming planetary masses between $\sim 0.01$ and $1\,M_\mathrm{Mars}$, although nongravitational effects may already be important for masses $\lesssim 0.3 \, M_\mathrm{Mars}$. The fraction of polluted WDs in our simulations, $\sim 0.05$, is consistent with observations of WDs with intermediate cooling times between $\sim 0.1$ and 1 Gyr. For cooling times $\lesssim 0.1$ Gyr and $\gtrsim 1$ Gyr, our scenario cannot explain the high observed pollution fractions of up to 0.7. Nevertheless, our results motivate searches for companions around polluted WDs.	astro-ph.EP	astro-ph	MNRAS 000, 1 -- 5 (2016) Preprint March 29, 2021 Compiled using MNRAS LATEX style file v3.0 White dwarf pollution by planets in stellar binaries Adrian S. Hamers1 and Simon F. Portegies Zwart1 1Leiden Observatory, Leiden University, PO Box 9513, NL-2300 RA Leiden, The Netherlands Accepted 2016 July 05. Received 2016 June 30; in original form 2016 June 17 ABSTRACT Approximately 0.2 ± 0.2 of white dwarfs (WDs) show signs of pollution by metals, which is likely due to the accretion of tidally disrupted planetary material. Models invoking planet-planet interactions after WD formation generally cannot explain pol- lution at cooling times of several Gyr. We consider a scenario in which a planet is perturbed by Lidov-Kozai oscillations induced by a binary companion and exacer- bated by stellar mass loss, explaining pollution at long cooling times. Our computed accretion rates are consistent with observations assuming planetary masses between ∼ 0.01 and 1 MMars, although nongravitational effects may already be important for masses . 0.3 MMars. The fraction of polluted WDs in our simulations, ∼ 0.05, is con- sistent with observations of WDs with intermediate cooling times between ∼ 0.1 and 1 Gyr. For cooling times . 0.1 Gyr and & 1 Gyr, our scenario cannot explain the high observed pollution fractions of up to 0.7. Nevertheless, our results motivate searches for companions around polluted WDs. Key words: white dwarfs -- stars: chemically peculiar -- planet-star interactions 1 INTRODUCTION The atmospheres of cool white dwarfs (WDs) are expected to consist entirely of hydrogen or helium due to efficient gravitational settling of metals (Schatzman 1945). How- ever, in 0.2 ± 0.2 of white dwarfs (Koester & Wilken 2006; Koester et al. 2014), spectra have revealed emission lines from a large range of metals, suggesting that these 'pol- luted' WDs have recently accreted metal-rich material (see Jura & Young 2014; Veras 2016; Farihi 2016 for reviews). Observations indicate that the pollution rate is approxi- mately independent of cooling time (Koester et al. 2014), requiring a continuous pollution process. (Zuckerman et al. Accretion from the interstellar medium (Dupuis et al. 2003; 1993) has been ruled out Koester & Wilken 2006; Dufour et al. 2007; Jura 2008). WD pollution could instead originate from accreting tidally disrupted rocky planetary material (e.g. Alcock et al. 1986; Aannestad et al. 1993; Debes & Sigurdsson 2002; Jura 2003) with a composition similar to Earth's (see e.g. Jura & Young 2014, and references therein), originating from planetesimals of mass ∼ 1020 kg to planets as massive as Mars (Jura et al. 2009). This is supported by the observation that all WDs with discs are polluted, and by the observed transiting planetesimals in tight orbits around WD 1145+017 (Vanderburg et al. 2015). Polluted WDs are therefore a probe for planetary sys- tems around WDs (see Veras 2016 for a review). Bod- ies in tight orbits are engulfed by the star as it expands along the red giant branch (RGB; Villaver & Livio 2009; c(cid:13) 2016 The Authors Kunitomo et al. 2011; Villaver et al. 2014) and asymptotic giant branch (AGB; Mustill & Villaver 2012) phases. At larger distances, stellar mass loss, tides, interactions with stellar ejecta and nongravitational effects are important. Early after WD formation, dynamical instabilities arising from planet-planet interactions and mass loss could lead to the disruption of planetary material and WD pollution (Debes & Sigurdsson 2002; Bonsor et al. 2011; Debes et al. 2012; Veras et al. 2016). These instabilities typically occur on short time-scales, and cannot explain continued pollution of WDs with cooling times of several Gyr. Bonsor & Veras (2015) proposed a scenario indepen- dent of the WD cooling time, in which the WD planetary system is perturbed by a wide binary companion whose orbit is driven to high eccentricity due to Galactic tides. We investigate a related scenario in which the WD and planet are orbited by a secondary star. We fo- cus on planets with radii & 1000 km, for which non- gravitational effects are not important (e.g. Veras 2016). Mass loss of the primary star triggers adiabatic expan- sion of both the inner (planet's) and outer (secondary's) orbits. The importance of Lidov-Kozai (LK) oscillations (Lidov 1962; Kozai 1962) in the inner orbit then typically increases (Perets & Kratter 2012; Shappee & Thompson 2013; Hamers et al. 2013; Michaely & Perets 2014). Conse- quently, the inner orbit can be driven to high eccentricity for the planet to be tidally disrupted by the WD, polluting the latter. Pollution can be prolonged to several Gyr after the WD formed. 6 1 0 2 l u J 5 . ] P E h p - o r t s a [ 1 v 7 9 3 1 0 . 7 0 6 1 : v i X r a 2 Adrian S. Hamers and Simon F. Portegies Zwart 2 METHODOLOGY 2.1 Algorithm used the secular code dynamics We of Hamers & Portegies Zwart (2016) coupled with the stellar evolution code SeBa (Portegies Zwart & Verbunt 1996; Toonen et al. 2012). In SeBa, we assumed a metallicity of 0.02. Adiabatic mass loss was assumed to compute the dynamical response of the orbits on mass loss. Tidal evolution was modelled with the equilibrium tide model (Eggleton et al. 1998). For the primary star, the tidal dissipation strength was computed using the prescription of (Hurley et al. 2002) with an apsidal motion constant of 0.014, a gyration radius of 0.08, an initial spin period of 10 d and zero obliquity (similar to Fabrycky & Tremaine 2007). The stellar spin period was computed assuming conservation of spin angular momentum. For the planet, we assumed a viscous time-scale of ≈ 1.4 yr (Socrates et al. 2012), an apsidal motion constant of 0.25, a gyration radius of 0.25, an initial spin period of 10 hr and zero obliquity. 102 101 U A / f , 1 a 100 10−1 100 101 a1,i/AU 102 2.2 Initial conditions NMC = 105 systems were generated as follows. The pri- mary mass M⋆ was sampled from a Salpeter distribution (Salpeter 1955) between 1.2 and 6 M⊙. The secondary mass Mc was sampled assuming a linear distribution of q = Mc/M⋆ with 0.1 < q < 1. The mass of the planet, mp, was sampled logarithmically between 0.3 MMars and 1 MJ. The planetary radius was computed using the mass-radius relation of Weiss et al. (2013). According to the latter rela- tion, 0.3 MMars corresponds to ≈ 1000 km. We focused on planets with initial semimajor axes a1 > 1 AU, for which interactions with stellar ejecta can be ne- glected. A linear distribution of a1 was assumed between 1 and 100 AU. The outer orbit semimajor axis a2 was sam- pled assuming a lognormal distribution of the outer orbital period between 10 and 1010 d (Duquennoy & Mayor 1991; Raghavan et al. 2010; Tokovinin 2014). The eccentricities ei were sampled from a Rayleigh distribution with an rms of 0.33 (Raghavan et al. 2010). The orbits were assumed to be randomly orientated. A sampled configuration was rejected if the stability criterion of Holman & Wiegert (1999) was not satisfied. Each system was simulated for 10 Gyr, or until (1) a dynamical instability occurred according to the criterion of Holman & Wiegert (1999), or (2) the planet collided with, or was tidally disrupted by the primary star (assuming a tidal disruption radius rt = ηRp [M⋆/mp]1/3 with η = 2.7 Guillochon et al. 2011). According to SeBa, the fraction of time of 10 Gyr spent during the various evolutionary stages assuming M⋆ = 1.2 M⊙ (M⋆ = 6.0 M⊙) is ≈ 0.56 (≈ 0.007) for the MS, ≈ 0.09 (≈ 0.008) for the giant phases (including core helium burning, i.e. from RGB up to and including AGB), and ≈ 0.35 (≈ 0.985) for the WD phase. Figure 1. Initial versus final a1, showing 5 % of all simulated systems. Refer to Section 3.1 for the meaning of the symbols. Red dashed lines: the maximum radii of the primary star for the lowest and highest masses considered (1.2 and 6 M⊙). Black dashed lines: adiabatic mass loss lines for the mass boundaries. 3 RESULTS 3.1 Overview In Fig. 1, we show initial versus final a1. The various out- comes are distinguished with symbols and colours, as de- scribed below. (i) Black dots in Fig. 1 -- stable planets in expanded or- bits, on lines associated with adiabatic mass loss, a1,f = a1,i (M⋆,MS/M⋆,WD). Given the range of M⋆, this results in a band of systems bounded by the two black dashed adia- batic mass loss lines. (ii) Dark blue filled stars -- pre-WD collisions, on or below a1,f = a1,i. After the main-sequence (MS) phase, tidal dis- sipation becomes more efficient. Possibly coupled with LK cycles, this leads to planetary engulfment. (iii) Light red open stars -- pre-WD tidal disruptions, on a1,f = a1,i. The inner orbit eccentricity is excited by LK cycles during the MS. This leads to tidal disruption in a highly ec- centric orbit because tidal friction in the radiative envelope is very weak. (iv) Green open circles -- post-WD tidal disruptions, within the same band as (i). After the AGB mass loss phase, the de- creased semimajor axis ratio a2/a1 gives rise to extremely high eccentricities and tidal disruption. An example is given in Fig. 2. (v) Blue filled triangles -- dynamically unstable systems (ac- cording to the criterion of Holman & Wiegert 1999), trig- gered by AGB mass loss. In Fig. 3, the fractions of systems corresponding to the outcomes are shown as a function of a1,i (left panel) and a1,f (right panel). The fractions for a1,f > 100 AU are incomplete for outcomes (ii) and (iii). MNRAS 000, 1 -- 5 (2016) 104 102 100 10−2 10−4 U A / r g e d / l e r i 140 120 100 80 60 40 20 0 White dwarf pollution in binaries 3 1012 1011 1010 109 108 107 106 105 ) 1 − s g ( / ⋆ M hmpi = 0.01 MMars hmpi = 0.1 MMars hmpi = MMars hmpi = M⊕ a1 a1(1 − e1) a2 R⋆ rt 500 1000 1500 t/Myr 2000 2500 3000 104 10−3 10−2 10−1 tcool/Gyr 100 101 Figure 2. Example evolution in which the planet is tidally dis- rupted by the star after the latter has evolved to a WD. Top panel: various distances of interest: the planet's semimajor axis a1 (dashed green line) and periapse distance a1(1−e1) (solid green line), the binary orbit semimajor axis a2 (black dashed line), the primary stellar radius R⋆ (red dotted line) and the planetary tidal disruption radius rt (green dotted line). Bottom panel: the incli- nation between the planetary and binary companion orbits. The dashed line shows 90◦. The primary star RGB and AGB phases occur near ≈ 1250 Myr and ≈ 1500 Myr, respectively. During the pre-WD phase, the periapse distance a1(1 − e1) oscillates due to LK cycles, but does not become small enough for strong tidal dissipation, tidal disruption or collision with the primary star. After the AGB phase, the LK eccentricity oscillations increase in amplitude due to the decrease in a2/a1, with a similar mini- mum a1(1 − e1) whereas a1 has increased due to mass loss. At ≈ 2800 Myr, a flip occurs in the orbital orientation from prograde (< 90◦) to retrograde (> 90◦), which is associated with a very high eccentricity and a1(1 − e1) ≈ 10−2 AU, triggering the tidal disruption of the planet. Figure 4. Simulated WD accretion rates as a function of cooling time (solid lines; dashed lines indicate the standard deviation) as- suming various mean planetary masses (indicated in the legend). Black crosses: observational data from Farihi et al. (2009). For small a1,i, the fraction of systems with planets be- ing engulfed during the pre-WD phase is unity, and decreases as a1,i increases. There is a minimum a1,i for which planets can be tidally disrupted after WD formation, or for which a dynamical instability occurs. From Fig. 3, this minimum is a1,i & 5 AU (or a1,f & 10 AU). Beyond the minimum value, the fraction of post-WD tidally disrupted planets (dynami- cally unstable systems) is approximately constant at ∼ 0.03 (∼ 0.01). 100 10−1 10−2 f 10−3 0 20 40 60 a1,i/AU 80 0 100 200 300 a1,f/AU 400 100 10−1 f 10−2 10−3 Figure 3. The fractions of systems corresponding to the out- comes described in Section 3.1 as a function of a1,i (left panel) and a1,f (right panel). MNRAS 000, 1 -- 5 (2016) 3.2 WD pollution -- comparisons to observations Outcome (iv) is expected to result in WD pollution. In Fig. 4, we show WD accretion rates as a function of cool- ing time from the simulations (solid and dashed lines), and observations (crosses, from Farihi et al. 2009). Simulated ac- cretion rates were computed from post-WD tidal disruption events assuming that (1/2) mp is eventually accreted onto the WD (Hills 1988). Disruption rates were found to be in- dependent of planetary mass. Using this result, we assumed a range of mean planetary masses hmpi in Fig. 4. Both simulated and observed accretion rates tend to decrease with cooling time. The bulk of the observations can be explained with hmpi ranging between ∼ 0.01 and 1 MMars. Nongravitational effects may, however, be impor- tant for masses . 0.3 MMars. In Fig. 5, we show the fractions of polluted WDs as a function of cooling time (assuming a binary fraction of 0.5), and including observations from Koester et al. (2014). For cooling times between ∼ 0.1 and 1 Gyr, the fractions from the simulations, ∼ 0.05, are consistent with the observed fractions. The simulations are unable to produce fractions as high as ∼ 0.7 for cooling times of ∼ 0.05 Gyr, or ∼ 0.5 for cooling times of ∼ 2 Gyr. 4 Adrian S. Hamers and Simon F. Portegies Zwart 1.0 0.8 0.6 0.4 0.2 f 10−3 10−2 10−1 tcool/Gyr 100 101 Figure 5. Solid line: the fraction of polluted WDs as a function of cooling time. Black circles and crosses: observed pollution frac- tions from Koester & Wilken (2006) and Koester et al. (2014), respectively. fraction of polluted WDs is consistent with observations of WDs with intermediate cooling times (0.1 Gyr . tcool . 1 Gyr). For short and long cooling times, our scenario can- not explain the high observed pollution fractions of up to 70 per cent. Our scenario may also apply to planetesimals, but further work is needed to incorporate nongravitational effects. ACKNOWLEDGEMENTS We thank the anonymous referee for useful comments. This work was supported by the Netherlands Research Council NWO (grants #639.073.803 [VICI], #614.061.608 [AMUSE] and #612.071.305 [LGM]) and the Netherlands Research School for Astronomy (NOVA). 4 DISCUSSION References Aannestad P. A., Kenyon S. J., Hammond G. L., Sion E. M., 4.1 Approximations in the dynamics 1993, AJ, 105, 1033 In our simulations, the dynamics were modelled using the computationally advantageous secular approach. However, in the 'semisecular' regime of 3 . a2(1 − e2)/a1 . 10 (Antonini & Perets 2012; Antonini et al. 2014), in which the system is still dynamically stable, the approximations made in the secular method break down. In our simulations, ≈ 0.5 of the the tidally disrupted systems have a2(1 − e2) > 10 (at the moment of disruption). For the group in the semisecular regime, we expect that the true eccentricity excitation (i.e. as computed with direct N -body integrations) is at least as effective compared to the secular method, if not higher (see e.g. Fig. 5 of Antonini et al. 2014). Therefore, we do not expect that this strongly affects our conclusions regard- ing WD pollution. Regarding uncertainties associated with the finite order of the expansion in the secular method, we also carried out the population synthesis up and including third-order terms (by default, terms up to and including fifth order were included), and found no statistically distinguish- able results. If a2(1 − e2)/a1 is even smaller, then a short-term dy- namical instability can occur. In our simulations, these con- ditions for dynamical instability are invariably triggered at WD formation (zero cooling ages), and the fraction of sys- tems is lower compared to the 'dynamically stable' tidal disruption systems by a factor of a few (cf. Fig. 3). Such dynamical instabilities can lead to collisions, but also to ejections, most likely of the planet. In the simulations of Perets & Kratter (2012), roughly equal-mass stars were con- sidered, and ≈ 0.01 of the cases led to collisions of objects. Therefore, we do not expect a large contribution to WD pollution from tidal disruptions following a dynamical in- stability at WD formation. 5 CONCLUSIONS We considered a scenario for WD pollution by planets trig- gered by LK oscillations induced by a binary companion. Our computed accretion rates are consistent with observa- tions for planetary masses between ∼ 0.01 and 1 MMars. The Alcock C., Fristrom C. C., Siegelman R., 1986, ApJ, 302, 462 Antonini F., Perets H. B., 2012, ApJ, 757, 27 Antonini F., Murray N., Mikkola S., 2014, ApJ, 781, 45 Bonsor A., Veras D., 2015, MNRAS, 454, 53 Bonsor A., Mustill A. J., Wyatt M. C., 2011, MNRAS, 414, 930 Debes J. H., Sigurdsson S., 2002, ApJ, 572, 556 Debes J. H., Walsh K. J., Stark C., 2012, ApJ, 747, 148 Dufour P., et al., 2007, ApJ, 663, 1291 Dupuis J., Fontaine G., Wesemael F., 1993, ApJS, 87, 345 Duquennoy A., Mayor M., 1991, A&A, 248, 485 Eggleton P. P., Kiseleva L. G., Hut P., 1998, ApJ, 499, 853 Fabrycky D., Tremaine S., 2007, ApJ, 669, 1298 Farihi J., 2016, New Astron. Rev., 71, 9 Farihi J., Jura M., Zuckerman B., 2009, ApJ, 694, 805 Guillochon J., Ramirez-Ruiz E., Lin D., 2011, ApJ, 732, 74 Hamers A. S., Portegies Zwart S. F., 2016, MNRAS, 459, 2827 Hamers A. S., Pols O. R., Claeys J. S. W., Nelemans G., 2013, MNRAS, 430, 2262 Hills J. G., 1988, Nature, 331, 687 Holman M. J., Wiegert P. A., 1999, AJ, 117, 621 Hurley J. R., Tout C. A., Pols O. R., 2002, MNRAS, 329, 897 Jura M., 2003, ApJ, 584, L91 Jura M., 2008, AJ, 135, 1785 Jura Young M., E. D., 2014, Annual Review of Earth and Planetary Sciences, 42, 45 Jura M., Muno M. P., Farihi J., Zuckerman B., 2009, ApJ, 699, 1473 Koester D., Wilken D., 2006, A&A, 453, 1051 Koester D., Gänsicke B. T., Farihi J., 2014, A&A, 566, A34 Kozai Y., 1962, AJ, 67, 591 Kunitomo M., Ikoma M., Sato B., Katsuta Y., Ida S., 2011, ApJ, 737, 66 Lidov M. L., 1962, Planet. Space Sci., 9, 719 Michaely E., Perets H. B., 2014, ApJ, 794, 122 Mustill A. J., Villaver E., 2012, ApJ, 761, 121 Perets H. B., Kratter K. M., 2012, ApJ, 760, 99 Portegies Zwart S. F., Verbunt F., 1996, A&A, 309, 179 Raghavan D., et al., 2010, ApJS, 190, 1 Salpeter E. E., 1955, ApJ, 121, 161 Schatzman E., 1945, Annales d'Astrophysique, 8, 143 Shappee B. J., Thompson T. A., 2013, ApJ, 766, 64 Socrates A., Katz B., Dong S., 2012, preprint, (arXiv:1209.5724) Tokovinin A., 2014, AJ, 147, 86 Toonen S., Nelemans G., Portegies Zwart S., 2012, A&A, 546, A70 MNRAS 000, 1 -- 5 (2016) White dwarf pollution in binaries 5 Vanderburg A., et al., 2015, Nature, 526, 546 Veras D., 2016, Royal Society Open Science, 3, 150571 Veras D., Mustill A. J., Gänsicke B. T., Redfield S., Geor- gakarakos N., Bowler A. B., Lloyd M. J. S., 2016, MNRAS, 458, 3942 Villaver E., Livio M., 2009, ApJ, 705, L81 Villaver E., Livio M., Mustill A. J., Siess L., 2014, ApJ, 794, 3 Weiss L. M., et al., 2013, ApJ, 768, 14 Zuckerman B., Koester D., Reid I. N., Hünsch M., 2003, ApJ, 596, 477 MNRAS 000, 1 -- 5 (2016)
1511.03456	1	1511	2015-11-11T11:14:53	SIDRA: a blind algorithm for signal detection in photometric surveys	[ "astro-ph.EP", "astro-ph.IM" ]	We present the Signal Detection using Random-Forest Algorithm (SIDRA). SIDRA is a detection and classification algorithm based on the Machine Learning technique (Random Forest). The goal of this paper is to show the power of SIDRA for quick and accurate signal detection and classification. We first diagnose the power of the method with simulated light curves and try it on a subset of the Kepler space mission catalogue. We use five classes of simulated light curves (CONSTANT, TRANSIT, VARIABLE, MLENS and EB for constant light curves, transiting exoplanet, variable, microlensing events and eclipsing binaries, respectively) to analyse the power of the method. The algorithm uses four features in order to classify the light curves. The training sample contains 5000 light curves (1000 from each class) and 50000 random light curves for testing. The total SIDRA success ratio is $\geq 90\%$. Furthermore, the success ratio reaches 95 - 100$\%$ for the CONSTANT, VARIABLE, EB, and MLENS classes and 92$\%$ for the TRANSIT class with a decision probability of 60$\%$. Because the TRANSIT class is the one which fails the most, we run a simultaneous fit using SIDRA and a Box Least Square (BLS) based algorithm for searching for transiting exoplanets. As a result, our algorithm detects 7.5$\%$ more planets than a classic BLS algorithm, with better results for lower signal-to-noise light curves. SIDRA succeeds to catch 98$\%$ of the planet candidates in the Kepler sample and fails for 7$\%$ of the false alarms subset. SIDRA promises to be useful for developing a detection algorithm and/or classifier for large photometric surveys such as TESS and PLATO exoplanet future space missions.	astro-ph.EP	astro-ph	Mon. Not. R. Astron. Soc. 000, 1–8 (2014) Printed 12 November 2015 (MN LATEX style file v2.2) SIDRA : a blind algorithm for signal detection in photometric surveys. D. Mislis1 ⋆, E. Bachelet1, K. A. Alsubai1, D. M. Bramich1, N. Parley1 1Qatar Environment and Energy Research Institute (QEERI), HBKU, Qatar Foundation, PO Box 5825, Doha, Qatar Accepted . Received 1; ABSTRACT We present the Signal Detection using Random-Forest Algorithm (SIDRA). SIDRA is a de- tection and classification algorithm based on the Machine Learning technique (Random For- est). The goal of this paper is to show the power of SIDRA for quick and accurate signal detection and classification. We first diagnose the power of the method with simulated light curves and try it on a subset of the Kepler space mission catalogue. We use five classes of simulated light curves (CONSTANT, TRANSIT, VARIABLE, MLENS and EB for constant light curves, transiting exoplanet, variable, microlensing events and eclipsing binaries, re- spectively) to analyse the power of the method. The algorithm uses four features in order to classify the light curves. The training sample contains 5000 light curves (1000 from each class) and 50000 random light curves for testing. The total SIDRA success ratio is > 90%. Furthermore, the success ratio reaches 95 - 100% for the CONSTANT, VARIABLE, EB, and MLENS classes and 92% for the TRANSIT class with a decision probability of 60%. Because the TRANSIT class is the one which fails the most, we run a simultaneous fit using SIDRA and a Box Least Square (BLS) based algorithm for searching for transiting exoplanets. As a result, our algorithm detects 7.5% more planets than a classic BLS algorithm, with better results for lower signal-to-noise light curves. SIDRA succeeds to catch 98% of the planet can- didates in the Kepler sample and fails for 7% of the false alarms subset. SIDRA promises to be useful for developing a detection algorithm and/or classifier for large photometric surveys such as TESS and PLATO exoplanet future space missions. Key words: techniques: photometric - planets and satellites: detection - planets and satellites: fundamental parameters - planetary systems. 1 INTRODUCTION The recent development of wide field photometric surveys opens up a new field of astrophysics. The deployment of both ground based (SuperWASP, HAT, QES) and space based (CoRoT,Kepler) surveys increases dramatically our knowledge about transiting planets. Indeed, the huge amount of collected data leads to a real problem of identifying targets. The OGLE survey, for example, observes more than 300 million stars in the Galactic bulge each night, leading to a sorting problem. This kind of problem is a known as a Big Data problem. Several methods are used to tackle the Big Data problem, and the Machine Learning algorithm is one of them. The Random Tree/Forest algorithm was described in 2001 by L. Breiman (Breiman 2001) as part of Artificial Intelligence and Machine Learning general algorithms. Some teams have already used Machine Learning algorithms for astronomical projects, especially for the large amount of data from the Kepler mission (Hogg et al. ⋆ E-mail: [email protected] 2013). Machine Learning object detection and classification for automated classification of active stars and galaxies is described in Li et al. (2008), using the k-Nearest Neighbours method. Recently, Masci et al. (2014) published an algorithm based on Random Forest for automatic classification of variable stars using the Wide-field Infrared Survey Explorer data with a success ratio from 87.8 to 96.2%. In the same year, OGLE detected a supernova Type Ia event, using real-time detection and Machine Learning automatic classification (Wyrzykowski et al. 2014). Furthermore, a Random Forest algorithm is used by McCauliff et al. (2014) to identify false positives in the Kepler mission data. The exoplanet microlensing surveys such as OGLE and MOA are facing a challenge with real-time photometry and lens detection (Bond et al. 2001). Microlensing detections must be observed by as many teams as possible in order to have a complete phase coverage of the phenomenon. This introduces a need for fast event detection on a huge amount of light curves. In 2002, Kovács et al. (2002) published the Box Least Square (BLS) algorithm for transiting exoplanet detection and since then, there are many different versions of BLS (Foreman-Mackey et al. 2 D. Mislis such as Kepler 2006), HATNet 2015). BLS is a very successful algorithm for almost all of the transiting surveys (Boruki et al. 2010), CoRoT (Moutou et al. 2007) –from space –and SuperWASP (Cameron et al. (Bakos et al. 2011), QES (Alsubai et al. 2013) etc. –from the ground. In principal, in order to detect a transiting signal in a light curve using a BLS algorithm, we have to fit the orbital period of the planet, the centre of the eclipse TC, the duration of the transit and the depth of the transit (Bordé et al. 2007; Bonomo et al. 2012; Cabrera et al. 2012). In this paper, we study a very different approach for signal detection and classification for transiting exoplanets, variable stars and microlensing events by changing the philosophy of signal detection from fitting to blind search using Machine Learning techniques. 2 DESCRIPTION OF THE METHOD AND SIMULATED LIGHT CURVES 2.1 Light–curve simulations In this paper, we used simulated light curves for the training and testing sample in order to perform various tests. We focused on five typical light curve types which can be expected in photometric surveys. These are constant stars (called hereafter CONSTANT), the exoplanet transiting light curves (TRANSIT), the variable stars (VARIABLE), the eclipsing binaries (EB) and the microlensing light curves (MLENS). Each light curve is described by a normal- ized flux as a function of time. We added noise to each light curve with various precisions. The rms is selected from a uniform distri- bution between 1 and 5%. We used a 30-d observing window, with a 30 min sample to simulate the light curves. Some of these types of light curves, such as the TRANSIT sample, require stellar physical properties (stellar mass, stellar radius, effective temperature) given by Kaler (1998). We select main sequence host stars randomly, us- ing a uniform distribution from F0 to M5 spectral type. This range was adopted because it roughly represents 90% of the total stars in the sky (Robin et al. 2004). Table 1, summarizes all the input pa- rameters and ranges we used for the simulated light curves. Fig. 1 shows typical simulated light curves from each class. A more de- tailed description for each light curve class is given below. 2.1.1 CONSTANT This subset of light curves is the most simple, and we can easily create it using pure white noise. The flux fC of a constant light curve is given by fC(t) = 1 + ǫ (1) where ǫ is a random variable set by a normal distribution N(0,rms). The rms is randomly selected to be in the range of 0.01 6 rms 6 0.05. 2.1.2 TRANSIT The host stars spectral type and the physical properties are selected randomly from the main-sequence data set as we describe in Sec- tion 2.1. The planet radius is also chosen randomly to be in the range 0.7 - 2 RJ and the period was chosen between 1 and 15d. The inclination angle i was chosen in the range imin 6 i 6 90 deg to ensure a transit. The imin is the minimum transit angle described by cos imin = R⋆ + RP α (2) where R⋆ is the star radius, RP the planetary radius and α the semi–major axis. Note that we choose a null eccentricity. The tran- sit model is produced using a quadratic limb-darkening law and the adopted flux is given by fT(t) = Pal(t, P, δ, i, d) + ǫ (3) where Pal() is the Pál et al. (2008) transiting analytical model, P is the period, δ is the transit depth and d is the duration of a transit. The limb–darkening coefficients are given by Claret (2004). 2.1.3 VARIABLES For this subset, we selected only stellar types crossing the main sequence and the instability strip of the Hertzsprung–Russel di- agram. This leads to select only F–type stars. We modelled the light curve by using a pure sinusoidal signal, with a flux amplitude 0.01 6 AV 6 0.1 and period P, between 1 and 15d. This way we mainly modelled some common types of variable stars such as RR Lyrae, δ-Scuti etc. Equation 4 gives the flux of our variable subset light curves : fV(t) = 1 + AV · sin(2πt/P) + ǫ. (4) 2.1.4 Eclipsing binaries The eclipsing binaries subset light curves were modelled using the same stellar main sequence characteristics (Section 2.1). The se- lected period is again between 1 and 15d. In order to simulate a full eclipsing binary light curve (primary/secondary eclipse, ellipsoidal variations etc) we used the PHOEBE eclipsing binary analytical model (Prša et al. 2005). PHOEBE requires stellar characteristics for both of the stars, such as stellar temperatures, masses and radii plus the orbital period of the system. We do not describe here the full equation package but the light–curve flux fEB we used is given by equation 5 : fEB(t) = PHOEBE(t, Mi, Ri, Ti, P) + ǫ, (5) where PHOEBE() is the eclipsing binaries model, Mi, Ri and Ti are the stellar mass, radius and temperature respectively for the primary (i = 1) and the secondary (i = 2) star. Finally, P is the orbital period. 2.1.5 Microlensing We produce microlensing light curves as though they have been observed by a survey such as MOA (Bond et al. 2001) or OGLE (Udalski et al. 1992). We select a uniform random value for the time of maximum tO between 0 and 30d. We also select Uo, the minimum impact parameter, from a uniform distribution (Alcock et al. 2000; Sumi 2011). We select the Einstein ring cross- ing time tE from a normal distribution with a mean of 20d and a standard deviation of 5d, which is a rough approximation of the true tE distribution (Sumi 2011). Finally, we select a uniform dis- tribution for the source flux fs and the blend flux fb in the range of 1–10. The adopted flux is fM(t) = fs · A(t) + fb fs + fb + ǫ (6) where A(t) is the microlensing magnification (Paczynski 1986). The SIDRA algorithm 3 2.3 The statistical method SIDRA, basically follows three simple rules : compute them in a fully blind way for all kinds of light curves. • Features must be as general as possible. SIDRA is able to • Features extraction must also be as fast as possible, in order to make the algorithm useful for large and/or real-time surveys (TESS, PLATO, OGLE, ATLAS etc). • Features must show very weak correlation with each other. There is no additional information for highly correlated features. To respect the first and second conditions, we make the choice to derive our features without any model fits and we calculate only the statistical information straight from the light curve. We use four features for SIDRA. • The skewness S : this is a measure of the asymmetry of a distribution, defined as the third standardized moment S = 1 n n Xi=1 (xi − ¯x)3 σ3 , (7) where ¯x is the mean, σ the standard deviation and n the total number of observations. Figure 1. Four random light curves (TRANSIT, VARIABLE, MLENS and EB) and their model (red solid line). Table 1. Input parameters and limits for our simulated light–curves sample. For each value we use uniform random distribution. • The kurtosis K : this is a measure of the flatness of a distribu- tion defined as the fourth standardized moment : rms (ǫ) Period (P) Spectra type Planetary radius (RP) Transit inclination (i) Observing window (t) Time resolution 0.01–0.05 1–15d F0–M5 0.7–2.0 RJ imin–90 deg 30d 30 min K = n Xi=1 (xi − ¯x)4 σ4 (8) • The autocorrelation integral AI : the autocorrelation integral is the sum of the autocorrelation values for all possible delays τ. The autocorrelation versus delay (τ) diagram gives information about periodical patterns of the light curve. For SIDRA, we explore the full observation window and measure the integral given by the au- tocorrelation vector 2.2 Random Forest basics In the Machine Learning domain, the key is to give informative pa- rameters to the algorithm that describe the problem in hand. These parameters, called features, must reflect the intrinsic properties of the different classes. Using these features as inputs, the Random Forest is a three step algorithm, as a typical Machine Learning pro- cedure suggests (train-test-predict). • The first step is the training part of the algorithm. The Random Forest uses the features of each vector of the training sample to build Ntree decision trees which are tuned to fit the output classes (train-step). • After the training process, it is recommended to characterize the performance of the algorithm by using an exercise sample (test- step). • If the user is satisfied with the accuracy of Test-Step, then the Random Forest can be used for any feature of the sample (predict- step). To help in the customization of the Random Forest algorithm, we used various tools. The feature importance vector gives the relative importance of each feature to produce the most accurate estimator. The confusion matrix shows in a simple way how well the algo- rithm performs. Its diagonal values are equal to the success ratio for each of the classes. Also, the i , j element of the confusion matrix give the false positive/negative rates (Masci et al. 2014). n Xτ=1 AI = (cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12) 1 (n − τ) · rms2  n=τ Xi=1 (xi − ¯x) (xi+τ − ¯x) (cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12) (9) • The modified information entropy (ES) - or Shannon Entropy (Shannon et al. 1949) : for each class of light curve, we assume a normal distribution. This is true only for a CONSTANT light curve, but still there is more information for all the other light curve types too. Thus, each xi has a probability based on the Cumulative Dis- tribution Function (CDF). Based on the nature of the survey (exo- planets, variables, microlensing), we can use different CDFs (nor- mal or inversed Gaussian CDF –Fig. 2) in order to cover different light curve cases. For our current SIDRA version we used the nor- mal and the inversed Gaussian CDF (blue and red line –Fig. 2) in combination. The probability P(xi) of each data point xi is given by equation 10: P(xi) = , (10) and the total entropy ES of each light curve is given by equation 11 1 1 + 2 · Xi=1 Z δ2 i=n δ1 2 √π Z 0 xi− ¯x √2 rms· e−t2 dt log2(P(δ))dδ! , ES (x) = − (11) where δ1,2 = xi ± σi and σi is the error of the point xi of the light curve. 4 D. Mislis Figure 2. Two different probability functions. A normal Gaussian CDF (blue dashed line) and the inversed Gaussian CDF (red solid line). Our final ES is calculated by adding two values of ES calculated by the normal and inverse Gaussian CDF. The features we have chosen show weak correlation in the param- eter space. Fig. 3 shows the correlation matrix between all features and classes. VARIABLE, EB and MLENS classes are very well determined. On the other hand there is a confusion between CON- STANT and TRANSIT classes. Because of the low signal-to-noise (S/N) ratio of some light curves, it is impossible to distinguish be- tween constant and transits (pure noise light curves). Fig. 3 –(bot- tom), explains the results of the confusion matrix (Fig. 5). It is clear from Fig. 3 –(top) that in some cases the Random Forest decision is very obvious because the classes are very well sepa- rated, suggesting that maybe we do not need a Random Forest al- gorithm. On the other hand, Random Forest becomes important to distinguish objects where their features are mixed (Fig. 3 –bottom). Also, in this paper we show only some cases. The input classes could be modified by any team, or even increased by adding more different light curves, making the problem much more complicated (distinguish between supernova –microlensing light curves or dif- ferent types of variables for example). 3 PERFORMANCE 3.1 General First we create a ’TRAIN’ sample using the five classes of light curves described in Section 2.1. We use 1000 light curves for each class (5000 in total) and we calculate the S , K, AI and ES fea- tures of each one. We also add a flag (CONSTANT, TRANSIT, VARIABLE, EB, MLENS) for each algorithm decision per class. By training SIDRA we found that the fitting score of the ‘TRAIN’ sample is 90% using 100 trees. That means from the 5,000 light curves, SIDRA could successfully distinguish 4,500 of them. The majority of the remaining 10%, which SIDRA fails, comes from CONSTANT and TRANSIT due to low S/N ratio, as we have de- scribed previously. It is very important for all features to have a good statistical weight in the procedure. The importance of each feature is high enough to be included in the algorithm. After the training procedure, the importance of each feature is shown in Fig. 4. The most important Figure 3. Different features for all classes. Top: the skewness and kurtosis (red and blue, respectively). Bottom: autocorrelation and entropy features (green and magenta, respectively). feature is ES with 40.2%, then AI with 28.5% and skewness and kurtosis with 12.2 and 19.1%, respectively. The AI feature contains high values for high–amplitude light curves such as microlensing and/or variables. On the other hand, skewness and kurtosis include information on the light curve shape, which is different for differ- ent classes. Finally, ES shows values around zero for variables, and constant, with high positive values for microlensing and negative values for planets and binaries. We applied SIDRA to 10000 light curves from all classes as a blind test (50000 light curves in total). We create a confusion matrix us- ing these results. In principal, the algorithm collects all decisions from all different trees. The final decision is made by maximizing the probability from all of the different trees. In the worst case, the minimum decision probability is 0.2 (because we have five classes–1/5), for a five class Random Forest such as SIDRA. It is obvious that the small decision probability decreases the suc- cess ratio of the algorithm because we have to deal with a flip-coin decision. We force SIDRA to take more certain decisions. For our example, we used a decision probability equal to 0.6. The confu- sion matrix shows the results of this test (Fig. 5). The success ratio for each class is 100% for microlensing and variable stars, 97% for The SIDRA algorithm 5 Figure 4. The feature importance statistics. ES value is the most important (40.2%). Figure 6. BLS (blue bars) and SIDRA (white bars) results after simultane- ously searching for different S/N values. features. As an example we can say that it is extremely rare for a transit survey to detect a microlensing event. Most of the transit surveys (if not all) avoid the fields in the Galactic plane. In these fields the probability to detect a microlensing event is close to zero. On the other hand microlensing surveys do not search for transiting planets because of the magnitude range and faint target stars of the field. We plan to give a more detailed analysis for the decision probability based on different algorithms (such as Bayesian, Dempster–Shafer theory and/or Fuzzy Logic) in a future paper. 3.2 A closer look at planets From the tests in Section 3.1, the most confusing classes are CON- STANT and TRANSIT. Fig. 5 suggests that 92% of the transiting light curves can be resolved by SIDRA, but this is not totally true. In Table 1 we select all the host stars between F0 and M5, planetary radius 0.7–2RJ and noise from 1–5%. That means that for 30% of our stellar sample (F0–G0), the transit depth range is from 0.008 to 2.5% for the worst and best case respectively. Most of these planets do not show any signal in the light curve with the noise properties we used and it is impossible to be detected. The real question is how many ‘detectable’ transiting planets SIDRA could flag. In order to judge the algorithm on the transiting light curves sam- ple, we compare SIDRA with a BLS–type algorithm, such as Kovács et al. (2002). We run BLS and SIDRA simultaneously on the same data set. For BLS we select a signal detection threshold at the 1σ level. Even if 1σ is not realistic for a real world survey (we expect signals above 2σ), this threshold is generous for BLS. If we increase the detection threshold, of course we expect much less planets. We compare with SIDRA 0.5 decision probability thresh- old (Fig 6). Both of the algorithms found approximately the same amount of planets. SIDRA detected 85.4% of the sample and BLS 77.9% of the sample (7.5% less planets than SIDRA) even with 1σ detection threshold. Also, Fig. 6 shows that SIDRA is more sensi- tive than BLS for low S/N light curves. Figure 5. The confusion matrix for decision probability of 0.6, where rows and columns refer to input and output light–curve flag, respectively. The color–bar refers to the success ratio. eclipsing binaries (3% planet false alarm), 95% constant (5% planet false alarm) and 92% for transits (8% constant false alarms). If we increase the decision probability (from 0.2 0.6) some light curves in the range of 0.2–0.6 are rejected. At 0.6 decision probability the algorithm rejects 5% of the total sample. It is clear that the success ratio of the algorithm is a function of the decision probability cut and there are no ‘golden’ fixed num- bers for each survey. On the other hand, each survey should define these numbers for their own goals, depending on their targets and 6 D. Mislis Figure 7. The Kepler sample we used for our test. From 0.5 to 50 RE and 5–5000 S/N. The S /N has been calculated in a 400d orbital period window. Plot shows also the SIDRA successful and unsuccessful detections (red and blue dots respectively). 3.3 Real data example - Kepler Mission The next task was to use our algorithm on real data in order to check its success. This section is only a small example with limited amount of data and tests. We plan to present a full Kepler mission data analysis using SIDRA in a future publication. For our tests we focus only on the exoplanets group of light curves using Ke- pler public light curves available from the Kepler archive hosted by the Multi–mission Archive at STScI1. The observations com- prised only from the long-cadence. In order to use these data from a space mission, we modified our training sample. We did not train for variables, microlensing events or eclipsing binaries. We have used Kepler Q1–Q6 KOI data set. We used 2000 light curves flagged as transit candidates (PLANET-SET) (Batalha et al. 2013; Mullally et al. 2015) and 2000 light curves from the Q1–Q6 data set flagged as non-transiting exoplanets (CONSTANT-SET). We select our sample randomly, which means that our sample in- cludes both large and small transiting light curves. Fig 7 shows a planetary radius versus S/N ratio of the Kepler sample we used. Furthermore we create another data set with 1000 light curves flagged as FALSE-SET. These light curves are included in the KOI Kepler catalogue but they are not real planets. We select false alarm light curves which belong to the CONSTANT-SET. Finally, we have to mention that we were using 400d of Q1–Q6 KOI data set. We train the forest using two classes (CONSTANT and TRANSIT) as described in Section 2.1, but we run it for all three data sets. We used rms range from 0.001 to 0.01, period range 1–200d and plan- etary radius of 1–5 RE. The total number of training light curves was 1000 per class. Fig. 8 shows the results of our test. The success ratio of real planets is 98%. The constant success ratio is 100% and the False Alarm success ratio is 93%. The False Alarm ratio is quite important. SIDRA classifies only the 7% of the False Alarm light–curves as planets, which appear in the KOI Kepler list. On the other hand SIDRA seems to miss 2% of real planets. In order 1 http://archive.stsci.edu Figure 8. The confusion matrix of the Kepler data test, where rows and columns refer to input and output light–curve flag, respectively. The color–bar refers to the success ratio. Decision probability is 0.5. to detect planets with higher period and/or smaller S/N, we need many more data than 400d. Also, we did not include any ‘exotic’ light curves in our training sample (Section 3.4). A more accurate analysis is required in a future paper including many more classes other than Constant and Planet data sets. The decision probability of 0.5 contains the 90% of the sample. 3.4 Exotic light curves The Kepler mission data have shown how difficult it is to detect transiting exoplanets around a star with high variability. These kind of light curves are the most important for space missions because with such high photometric accuracy, most of the stars show some kind of real variability. BLS-like algorithms fail to detect these kind of transits because of the algorithm design. BLS assumes that the out-of-transit mean value of all transits is 1 (or 0). This is not true in a variable star light cure with transit. The signal of the variable star dominates the light curve with very strong primary and harmonic periods. Almost all the combined transiting light curves need spe- cial analysis. For a pure blind detection method it is a very difficult problem for any algorithm, including Machine Learning. SIDRA is able to ‘solve’ the problem with a combined anal- ysis method similar to other BLS-like techniques. We simulate multi–period variable plus transiting exoplanet signals using Ke- pler accuracy specifications. Fig. 9–(top) shows an example of our simulated light curve. The strategy is simple. We first run SIDRA on the raw light curve. The algorithm classifies the light curve as a variable with probability 98%. Once we detect a variable signal, we use Lomb–Scargle, FFT or binned polynomial fits in order to re- move strong periodicities (Fig. 9–Bottom). Finally, we run SIDRA once more using the new light curve. We detect a planet with prob- ability of 81%. The SIDRA algorithm 7 We discuss the transiting exoplanet detection power compared with BLS-like algorithms, but we would like to make clear that we do not suggest to replace BLS with SIDRA, even if in our test SIDRA detects 8% more planets and is 1000 times faster. What we sug- gest for a at least an exoplanet survey, is to include both algo- rithms. SIDRA could be a very powerful tool, and could easily detect/classify interesting objects which require further analysis. 4.1 Advantages The algorithm could be easily modified for each team and project and it is as general as possible, solving simultaneously different types of light curves. On the other hand, BLS or high ∆χ2 meth- ods for example, work only for transiting and eclipsing binary light curves. For eclipsing binaries, variables and microlensing light curves, the algorithm is not only detecting the signal correctly but it also mini- mizes the false alarms. In our case of 10000 simulated light curves, false alarms for these objects were eliminated. For transiting light curves, it manages to detect more planets than the classical method of BLS. Furthermore, SIDRA is much faster than BLS. Once we train the network, the classification needs 4 ms for a single light curve of 4300 data points, making SIDRA ideal for huge surveys such as TESS and PLATO transiting exoplanet fu- ture space missions. BLS needs ∼4 s in the same machine for the same light curve. Finally, SIDRA has the ability to become a huge network with al- most all kinds of light curves. We can not only classify variable stars for example, but we can use many more classes in order to identify the type of each variable. On the other hand, we are able to search for non-periodic phenomena such as supernova and flare stars. 4.2 Disadvantages The major disadvantage of the algorithm is that it cannot resolve physical characteristics from the light curves because of its nature. For example we cannot extract the information about the radius of the planet because we do not fit physical parameters but we calcu- late statistical values of each light curve. Of course for the periodic events, the information of the period is hidden in the autocorrela- tion function, but still there is more information which is missing. Finally, the algorithm works with a training light curve set, which means that we have to be very careful on the selection of features and limits in order to maximize the success of the algorithm. 4.3 False Alarms The main problem of every detection and classification algorithm is the false positive and negative alarms. Assuming a 20000 light curve sample, we expect from SIDRA to solve the majority of the light curves correctly. On the other hand there are some limitations. We assume that the 1% of the theoretical sample contains real plan- ets, 15% real variables, 10% real eclipsing binaries and 0% real mi- crolensing events. Table 2 shows the results. From the total number the 95% remain in our sample after the decision probability 0.6 cut. These remaining light curves (5%) are flagged as unknown. In order to deal with the 4% of false alarm and the 5% of unknown light curves, we can decrease step-by-step the decision probability from 0.6. This will increase the SIDRA sample, decreasing the un- known light curves. Also, we can use a typical BLS algorithm. It is Figure 9. Top: a 30d multi–period variable star. Bottom: the same light curve after removing strong periodicities. We run the same experiment using BLS and it was able to detect the planet. We do not claim of course that this technique is new or it does not work with other detection algorithms. We just give an example showing that SIDRA is also able to detect exoplanets hidden in a strong variability. 4 CONCLUSIONS SIDRA is a blind detection and classification algorithm based on Machine Learning–Random Forest technique. This paper is a gen- eral presentation of the algorithm. We used simulated light curves from five different classes. These are constant, transiting, variables, microlensing and eclipsing binary light curves. Assuming a 60% decision probability, the algorithm success ratio is 95-100% for mi- crolensing, eclipsing binaries, variables and constant light curves and 91% for transits for Table 1 input values. Also we test SIDRA with real light curves from the Kepler mission. We detect and clas- sify successfully 100%, 98% and 93% of the constant, transiting and false alarm light curves. Furthermore, we show a simulated ex- ample of SIDRA transit detection around a variable host star. 8 D. Mislis Table 2. Classes statistics assuming a 20000 light curve sample. Classes Total number SIDRA sample (> 0.6) Successful detection False alarms Unknown Constant Planet Variable EB Total 14800 200 3000 2000 20000 14060 190 2850 1900 13357 175 2850 1843 19000 (95%) 18225 (91%) 703 as planets 15 as constant 57 as planets 775 (∼ 4%) 740 10 150 100 1000 (5%) Prša A. & Zwitter T., 2005, ApJ, 628, 426 Robin A. C., Reylé C., Derrière S., Picaud S., 2004, A&A, 416, 157 Shannon C., Weaver W., 1949, The Mathematical Theory of Com- munication. Univ. Illinois Press, Urbanamtc Sumi T., MOA & OGLE Collaboration, 2011, ESS.2.0103 Udalski A., Szymanski M., Kaluzny J., Kubiak M., Mateo M., 1992, Acta Astron., 42, 253 Wyrzykowski L. et al. 2014, Acta Astron., 64, 197 This paper has been typeset from a TEX/ LATEX file prepared by the author. not clear that BLS could solve the false alarms better than SIDRA (Fig. 6). That depends on the S/N of each light curve but we can use BLS as a separate tool. As we mention above, the decision proba- bility is a very important variable and we plan to study it in a future publication. ACKNOWLEDGEMENTS This publication is supported by NPRP grant no. X-019-1-006 from the Qatar National Research Fund (a member of Qatar Foundation). The statements made herein are solely the responsibility of the au- thors. REFERENCES Alcock C. et al. 2000, ApJ, 541, 734 Alsubai K. et al. 2013, Acta Astron., 63, 465 Bakos G., Hartman J., Torres G., Kovàcs G., Noyes R. W., Latham D. W., Sasselov D. D.; Béky B., 2011, EPJ Web. Conf., 11, 01002 Batalha N. et al. 2013, ApJS, 204, 24 Bond I. et al. 2001, MNRAS, 327, 868 Bonomo A. S. et al. 2012, A&A, 547.110 Bordé P., Fressin F., Ollivier M., Léger A., Rouan D., 2007, in Afonso C., Weldrake D., Henning Th., eds, ASP Conf. Ser.- Vol. 366, Transiting Extrasolar Planets Workshop. Astron. Soc. Pac., San Francisco, p. 145 Borucki W. et al. 2010, A&AS, 21510101 Breiman L., 2001, Mach. Learn., 45, 5 Cabrera J., Csizmadia Sz., Erikson A., Rauer H., Kirste S., 2012, A&A, 548, 44 Cameron C. A. et al. 2006, MNRAS, 373, 799 Claret A., 2004, A&A, 428, 1001 Foreman-Mackey D., Montet B., Hogg D., Morton T. D., Wang D., Schölkopf B., 2015, ApJ, 806, 13 Hogg D.W. et al., 2013, Kepler Project Office Call for White Pa- pers: Soliciting Community Input for Alternate Science Investi- gations for the Kepler Spacecraft Kaler J. B., 1998, Mon. Notes Astron. Soc. South. Afr., 57, 89 Kovács G., Zucker S. & Mazeh T., 2002, A&A, 391, 369 Li L., Zhang Y., & Zhao Y., 2008, Sci. China G, 51, 916 McCauliff S., Jenkins J., Catanzarite J., 2014, A&AS, 22412009 Masci F. J., Hoffman D. I., Grillmair C. J., Cutri R. M., 2014, AJ, 148, 21 Moutou C. et al. 2007, in Afonso C., Weldrake D., Henning Th., eds, ASP Conf. Ser.-Vol. 366, Transiting Extrasolar Plan- ets Workshop. Astron. Soc. Pac., San Francisco, p. 127 Mullally F. et al. 2015, ApJS, 217, 31 Paczynski B., 1986, ApJ, 304, 1 Pál A., 2008, MNRAS, 390, 281
1810.13148	1	1810	2018-10-31T08:17:05	A Spatially Resolved AU-scale Inner Disk around DM Tau	[ "astro-ph.EP", "astro-ph.SR" ]	We present Atacama Large Millimeter/submillimeter Array (ALMA) observations of the dust continuum emission at 1.3 mm and 12CO J=2-1 line emission of the transitional disk around DM Tau. DM Tau's disk is thought to possess a dust-free inner cavity inside a few au, from the absence of near-infrared excess on its spectral energy distribution (SED). Previous submillimeter observations were, however, unable to detect the cavity; instead, a dust ring ~20 au in radius was seen. The excellent angular resolution achieved in the new ALMA observations, 43 x 31 mas, allows discovery of a 4 au radius inner dust ring, confirming previous SED modeling results. This inner ring is symmetric in continuum emission, but asymmetric in 12CO emission. The known (outer) dust ring at ~20 au is recovered and shows azimuthal asymmetry with a strong-weak side contrast of ~1.3. The gap between these two rings is depleted by a factor of ~40 in dust emission relative to the outer ring. An extended outer dust disk is revealed, separated from the outer ring by another gap. The location of the inner ring is comparable to that of the main asteroid belt in the solar system. As a disk with a "proto-asteroid belt," the DM Tau system offers valuable clues to disk evolution and planet formation in the terrestrial planet forming region.	astro-ph.EP	astro-ph	Draft version November 1, 2018 Typeset using LATEX preprint style in AASTeX61 A SPATIALLY RESOLVED AU-SCALE INNER DISK AROUND DM TAU Tomoyuki Kudo,1 Jun Hashimoto,2 Takayuki Muto,3 Hauyu Baobab Liu,4 Ruobing Dong,5 Yasuhiro Hasegawa,6 Takashi Tsukagoshi,7 and Mihoko Konishi2 1SubaruTelescope, National Astronomical Observatory of Japan, 650 North A′ohoku Place, Hilo, HI 96720 U.S.A 2Astrobiology Center, National Institutes of Natural Sciences, 2-21-1 Osawa, Mitaka, Tokyo 181-8588, Japan 3Kogakuin University, 2665-1 Nakano, Hachioji, Tokyo 192-0015, Japan 4European Southern Observatory, Karl-Schwarzschild-Str. 2, 85748 Garching, Germany 5Department of Physics & Astronomy, University of Victoria, Victoria, BC, V8P 1A1, Canada 6Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA 91109, U.S.A 7National Astronomical Observatory of Japan, 2-21-1 Osawa, Mitaka, Tokyo 181-8588, Japan (Received; Revised; Accepted) Submitted to ApJL ABSTRACT We present Atacama Large Millimeter/submillimeter Array (ALMA) observations of the dust con- tinuum emission at 1.3 mm and 12CO J = 2 → 1 line emission of the transitional disk around DM Tau. DM Tau's disk is thought to possess a dust-free inner cavity inside a few au, from the absence of near-infrared excess on its spectral energy distribution (SED). Previous submillimeter observations were, however, unable to detect the cavity; instead, a dust ring ∼20 au in radius was seen. The excellent angular resolution achieved in the new ALMA observations, 43×31 mas, allows discovery of a 4 au radius inner dust ring, confirming previous SED modeling results. This inner ring is symmetric in continuum emission, but asymmetric in 12CO emission. The known (outer) dust ring at ∼20 au is recovered and shows azimuthal asymmetry with a strong-weak side contrast of ∼1.3. The gap between these two rings is depleted by a factor of ∼40 in dust emission relative to the outer ring. An extended outer dust disk is revealed, separated from the outer ring by another gap. The location of the inner ring is comparable to that of the main asteroid belt in the solar system. As a disk with a "proto-asteroid belt," the DM Tau system offers valuable clues to disk evolution and planet formation in the terrestrial planet forming region. Keywords: protoplanetary disks -- stars: individual (DM Tau) 8 1 0 2 t c O 1 3 . ] P E h p - o r t s a [ 1 v 8 4 1 3 1 . 0 1 8 1 : v i X r a Corresponding author: Tomoyuki Kudo [email protected] 2 Kudo et al. 1. INTRODUCTION An outstanding problem in planetesimal formation from aggregating dust in protoplanetary disks is radial drift of dust (Weidenschilling. 1977; Nakagawa et al. 1986): particles embedded in a gaseous disk with surface density decreasing outwards feel a headwind, lose angular momentum to the gas, and drift towards the central star. One solution to this problem is a dust trap (Rice et al. 2006; Johansen et al. 2009), in which mm-sized particles are trapped and accumulate at a local gas pressure maximum. To facilitate the formation of planetesimals in protoplanetary disks on the scale of the inner solar system, dust traps at a few au from the star are needed. Such structures can be searched for by high angular resolution observations of mm continuum emission. Many mechanisms have been proposed for producing pressure bumps in disks, such as the edges of gaps opened by planets (Zhu et al. 2012; Dong et al. 2015; note that one planet may produce multiple pressure bumps, Dong et al. 2017a). Magnetohydrodynamic (MHD) effects can also form pressure bumps in disks, generated by zonal flows (e.g., Johansen et al. 2009) or at the boundary of dead zones (e.g., Dzyurkevich et al. 2010). Pressure bumps may form at the locations of snowlines too, due to a change in the activity of the magnetorotational instability (e.g., Kretke & Lin 2007). Dust trapped at radial pressure bumps appears to be annular rings in millimeter continuum observa- tions. Such structures have been found in many objects, such as HL Tau (ALMA Partnership et al. 2015), TW Hya (Andrews et al. 2016; Tsukagoshi et al. 2016), HD 163296 (Isella et al. 2016), and MWC 758 (Dong et al. 2018). Our target, DM Tau (SpT: M1; Kenyon & Hartmann 1995, Teff: 3705 K; Andrews et al. 2011, M∗: 0.53 M⊙; Pi´etu et al. 2007, distance: 145 pc; Gaia Collaboration et al. 2018), has a known transi- tional disk (Espaillat et al. 2014). A central dust cavity ∼3 au in radius has been inferred based on its spectral energy distribution (SED) (Bergin et al. 2004; Calvet et al. 2005, shown in Appendix). Previous SMA sub-millimeter continuum observations were not able to resolve the 3 au cavity due to insufficient angular resolution; instead, a dust ring at 19 au was discovered (Andrews et al. 2011). Modeling of previous low resolution ALMA continuum observations (project ID:2013.1.00198.S; res- olution ∼0.′′4) suggested the presence of another faint dust ring at ∼80 au (Zhang et al. 2016). DM Tau has also been extensively studied in gas emission observations. Bergin et al. (2016a) resolved a C2H emission ring at the edge of the dust continuum disk, and many other molecular species, such as H2CO and CS, have been detected (Loomis et al. 2015; Semenov et al. 2018). 2. OBSERVATIONS DM Tau was observed with ALMA in band 6 in the C43 -- 9 configuration on October 27, 2017, UT as part of the project 2017.1.01460.S, utilizing 47 antennas with the baseline length extending from 135.1 m to 14.9 km. The observations were conducted in five spectral windows: two with bandwidths 117.188 MHz, velocity resolutions ∼0.166 km s−1, and centered at 220.39868 GHz for 13CO (2 → 1) and 219.56035 GHz for C18O (2 → 1); one with bandwidth 117.118 MHz, velocity resolution ∼0.079 km s−1, and centered at 230.53800 GHz for 12CO (2 → 1); and the last two windows for continuum observations with bandwidth 2.0 GHz. The precipitable water vapor was ∼0.5 mm during observations. The total on-source integration time was 66.1 min. The data were calibrated by the Common Astronomy Software Applications (CASA) package (McMullin et al. 2007) version 5.1.1, following the calibration scripts provided by ALMA. We had experimented with self-calibrating the new, high angular resolution ALMA data. However, in these observations, only 16 antennas were AASTEX An Inner Disk around DM Tau 3 located at &2 km baselines. With the 66.1 min on-source integration of our observations, our uv sampling at long baselines is insufficiently redundant. In addition, the continuum flux of our target source is dominated by structures with relatively extended (∼0.2′′) angular scales. As a result, the gain phase self-calibration flagged out over 50% of the &2 km baseline data even when using a per- scan solution interval and combining all spectral windows. This is unfavorable for our major science case of resolving the innermost region of DM Tau. On the other hand, the phase RMS of the <2 km baseline data is low, and the gain phase self-calibration does not help much. Therefore, we decided not performing self-calibration. We utilized the observations of the check (quasar) source J0449+1121 to assess how much our target source can be attenuated due to phase decoherence. The continuum fluxes of J0449+1121 with and without phase self-calibration are 293 and 255 mJy, respectively. This corresponds to a 13% attenuation, which is much smaller than the uncertainty of the dust mass opacity and the dust opacity depth in general. We combined our data with another ALMA dataset (project ID: 2013.1.00498.S; Pinilla et al. 2018) to recover the missing flux (∼70 mJy out of a total of ∼110 mJy; Beckwith et al. 1990) due to the sparseness of short baseline data. The phase centers of long and short baseline data were determined separately by ellipse isophoto fitting at 10 σ RMS noise in the dust continuum images synthesized by CASA with the CLEAN task using a multi-scale multi-frequency deconvolution algorithm (Rau & Cornwell 2011), and were shifted by fixvis in the CASA tools. We compared the amplitudes as a function of uv-distance at less than 200 kλ between our long baseline data and archival short baseline data, and confirmed their consistency. The CLEANed dust continuum image was synthesized with a Briggs weighting of 2.0 to maximize the signal-to-noise (S/N) ratio, providing RMS noise levels of 11 µJy beam−1. The total flux density after combining long and short baseline data is 116.75 ± 0.14 mJy, consistent with previous single dish observations assuming a 10 % uncertainty in absolute flux calibration. The 12CO (2 → 1) line data in both long and short baseline data were extracted by subtracting the continuum in the visibility space with uvcontsub in the CASA tools. The combined line cube was generated by the CLEAN algorithm with a velocity resolution of 0.5 km/s, and was spatially smoothed with a circular Gaussian kernel of 75 mas by imsmooth in CASA for presentation purposes. Though the 13CO and C18O (2 → 1) line data were processed with the same procedure as 12CO, they were not detected significantly. 3. RESULTS Figure 1 shows the 1.3 mm dust continuum image of DM Tau after the CLEANed process. We clearly resolve the dust disk into three components: an inner dust ring, an outer dust ring, and an extended outer disk (Figures 1a to c). The peak flux density at the inner and the outer rings is detected with 24 and 59 σ, respectively. An extended structure beyond r =0.′′4 is also marginally detected with ∼5 σ. To derive the azimuthally averaged radial brightness profile (Figure 1d), we deprojected the dust continuum image in the visibility domain following Zhang et al. (2016). The inclination and the position angle of the disk were derived by fitting an ellipse to the outer ring. The fitting results and derived disk's geometric parameters are shown in Table 1. An inner dust ring at r ∼ 0.′′03 is discovered in our dust continuum images. The ring is spatially resolved into a north and a south blobs (Figure 1c); the north blob is 20 ± 8 % (2.5 σ) brighter than the south one. More data are needed to confirm the apparent asymmetry. The total flux density of the inner ring inside 0.′′06 is 1.33 ± 0.03 mJy. The contamination from possible free-free 4 Kudo et al. Table 1. Best-fit parameters of ellipse fitting in the outer ring R.A. (ICRS) Dec. (ICRS) Radius (′′: au) i (◦) P.A. (◦) 04:33:48.749 [0.001] +18:10:09.64 [0.01] 0.176±0.001: 25.5±0.2 35.2±0.7 157.8±1.0 Note -- Parentheses of R.A. and DEC. indicate 1σ error in arcsecond. The heliocentric distance of the system (145 pc) is used to convert arcsecond into au. emission is less than 8% (3σ level), determined by extrapolating the flux density measured at 3.4 cm (Zapata et al. 2017) to 1.3 mm assuming a spectral index of +0.6. Assuming a distance of 145 pc, a dust opacity per gas mass κν =2.3 cm2 g−1 at 230 GHz (Beckwith & Sargent 1991), a temperature of 100 K, and a gas-to-dust mass ratio of 100, the total mass of the inner disk is measured as 0.04 MJup. Our observations also clearly spatially resolved the outer dust ring and the gap between the two rings at 0.′′18 and r ∼0.′′1, respectively (Figures 1b and 1c). The dust continuum emission is detected with &5 σ in the gap region, suggesting that the gap is not dust free. The outer ring is asymmetric: the brightness contrast between the peak flux density at P.A.=∼270◦ and that at the opposite position is 1.28 ± 0.04. The inner edge of the outer ring is steeper (I(r) ∝ r3.9±0.3) than the outer edge (I(r) ∝ r−2.8±0.1). Because the outer ring is spatially resolved with a radial width of ∼0.′′1 (15 au), the gradient difference is real. The extended structure beyond the outer ring has a nearly flat radial brightness profile (Figure 1d). A possible shallow gap at r ∼0.′′5 can be seen in this structure as well. Figures 2(a) and (b) show the integrated intensity (0th moment) map for 12CO obtained from 1.6 to 11.1 km s−1 with 1σ = 3.5 mJy beam−1 km s−1. The peak emission is 56.5 mJy beam−1 km s−1, located at the north blob around the inner dust ring. The total integrated intensity is 185 ± 7 Jy km s−1 inside 0.′′18. The intensity-weighted velocity (1st moment) map is shown in Figures 2(c) and (d). The center of the CO gas motion nearly coincides with the center of the outer dust ring derived from the ellipse fitting, and might be closer to the north than to the south blob. To check whether the center of the outer ring is consistent with the rotational center of the CO gas, we plot loci of the peak emission of a Keplerian disk around a 0.53 M⊙ star (DM Tau's mass) in the position -- velocity diagram (Figure 2e), finding that the outer ring's center is the center of the gas rotation. 4. MODEL FITTING To test whether the observed asymmetry in the outer dust ring is significant, we performed fitting for dust continuum emission in the visibility domain using a simple analytic disk model, and then subtracted the modeled disk from the data. Our disk model has a simple power-law radial profile with an exponential taper at the outside (e.g., Lynden-Bell & Pringle 1974; Hartmann et al. 1998): I(r) ∝ 2 Xi=1 αi(cid:18) r rci(cid:19)−(q+γi) exp"−(cid:18) r rci(cid:19)2−γi# , AASTEX An Inner Disk around DM Tau 5 ✁✂✄☎✆✄✆ ✝✞✂✄✟ ✆✠✡☛ ☞✞✂✄✟ ✟✠☎✌ ✍☎☎✄✟ ✟✠☎✌ ✸✁✄ ✲✁✂ Figure 1. 1.3 mm dust continuum image around DM Tau. The synthesized beam size of 0.′′043 × 0.′′031 with a position angle of 22.7◦ is shown at the left bottom. The 1σ noise level is 11 µJy beam−1. (a): Overview of the DM Tau disk. (b): Zoomed-up image to the outer ring. The color range is the same as in panel (a). (c): Image focusing on the inner disk region indicated by a dashed white square in panel (b). The white star denotes the center of the outer dust ring determined by ellipse fitting (Table 1). (d): Azimuthally averaged radial profiles in the deprojected image. To minimize effects of the beam elongation and obtain a circular beam, we apply tapering in the CLEAN process of the deprojected image. The beam size is 0.′′051 × 0.′′050 and the noise level is 12 µJy beam−1. The 3σ noise level of the deprojected image is indicated by the dashed line. where αi and rci are a scaling factor and a characteristic scaling radius, respectively. The two global components (i) in the profile are: component 1 for the inner and outer dust rings, and component 2 for the extended outer disk (Figure 3a). The q parameter is introduced to specify the radial dependence of the dust temperature Td, that is Td ∝ r−q. Optically thin emission in the Rayleigh -- Jeans regime scales as Iν ∝ Bν(Td)(1 − e−τ ) ∝ Tdτ, 6 Kudo et al. Figure 2. CLEANed 12CO (2 → 1) line emission maps of the DM Tau disk overlaying the dust continuum contours (14, 19, 34, and 44 σ). Panels (a) and (b) show 0th moment maps, while panels (c) and (d) show 1st moment ones. The 12CO images are spatially smoothed with a circular Gaussian of a 0.′′075 kernel. The white star denotes the center of the outer dust ring determined by ellipse fitting (Table 1). Panel (e) shows the position-velocity diagram along the dashed line centered on the outer ring's center in panel (c), and loci of peak emissions in the Keplerian disk around DM Tau with a mass of M =0.53 M⊙ and systemic velocity of 6.0 km s−1 (Pi´etu et al. 2007). where Bν and τ are the blackbody intensity at frequency ν and the optical depth (τ = κΣ; where κ and Σ denote the opacity and the surface density, respectively). In the radial direction, we have the following components with their scaling factors: 0 for r < rcav δ1 for rcav < r < rgap1 1 for rgap1 < r < rgap2 0 for rgap2 < r, α1 =   ✲ AASTEX An Inner Disk around DM Tau 7 ✁✂✄☎✆✝✞✂ ✁✂✄☎✆✝✞✂ ✟✠✡✡☛☞✌✍✎✏✑ ❆✁✂✄✄☎✆✝✞✟ ✁✆✝s✟✆s✝☎ ✲ (a): Generic surface brightness model. The red and blue solid lines represent the surface Figure 3. brightness profiles of component 1 (the inner and outer rings at rcav < r < rgap1 and rgap1 < r < rgap2 , respectively) and component 2 (the extended outer disk at rgap2 < r) scaled down by the scaling factor αi, respectively. At r < rcav, no dust emission is assumed due to absence of NIR excess in the DM Tau's SED. (b): Normalized surface brightness profile in our best-fit disk model. (c): Real part of the visibilities for the data (black dots) and the best-fit model (red line) in top panel. The bottom panel shows residual visibil- ities between best-fit and observations. (d): Residual image (data−model) overlaying the dust continuum contours (14, 19, 34, and 44 σ). (e): Zoom-in version (see the white dotted box in panel (d)). 0 for r < rgap2 f for rgap2 < r. α2 =  We normalize the total flux in the model (Ftotal) to the observed value. The disk inclination (i) and position angle (P.A.) are fixed as in Table 1. There are 11 free parameters in the model (rcav, rgap1, rgap2, δ1, f , q, γ1, γ2, rc1, rc2, and Ftotal). The depletion factor δ2 at rgap2 (Figure 3a) is measured after the calculations complete. To convert a modeled disk image to complex visibilities with identical uv-coverages of observations, we utilize the public python code vis sample1. The computed visibilities 1 vis sample is publicly available at https://github.com/AstroChem/vis sample or in the Anaconda Cloud at https://anaconda.org/rloomis/vis sample 8 Kudo et al. Table 2. Results of MCMC fitting and its parameter ranges Rcav (au) Rgap1 (au) Rgap2 (au) δ1 δ2 q γ1 γ2 Rc1 (au) Rc2 (au) Flux mJy 3.16+0.22 −0.23 21.00+0.02 −0.02 75.64+0.37 −0.39 0.028+0.001 −0.001 ∼0.356 0.01+0.01 −0.02 1.10+0.05 −0.03 0.01+0.02 −0.01 18.28+0.90 −1.14 124.10+1.05 −1.05 93.30+0.52 −0.46 {0.00-7.25} {14.50-29.00} {72.50-101.50} {1.000-0.001} -- {0.00-1.00} {0.00-2.00} {0.00-2.00} {0.00-29.00} {58.00-145.00} {90.0-105.0} Note -- Parentheses describe parameter ranges in our MCMC calculations. The errors in γi are large due to local maxima in calculations. The depletion factor of δ2 is measured in Figure 3(b). The factor of log f is −1.82+0.05 −0.07 . are deprojected in the uv-plane to calculate their azimuthal averages. For the fitting, we used the Markov Chain Monte Carlo (MCMC) method in the emcee package (Foreman-Mackey et al. 2013). Our calculations used flat priors with the parameter ranges summarized in Table 2. The burn-in phase (from initial conditions to reasonable sampling) employs 500 steps, and we run another 500 steps for convergence, totaling 1000 steps with 100 walkers. The fitting result and the best-fit surface brightness profile are shown in Table 2 and Figure 3(b), respectively. The corner plot with the MCMC posteriors is also shown in Appendix. The depletion factor δ2 at rgap2 is measured as ∼0.356 using the best-fit brightness profile in Figure 3(b). The reduced-χ2 calculated with the observed and modeled visibilities in Figure 3(c) is 2.4. The residual map prepared using the residual visibilities (data−model) is shown in Figures 3(d) and (e). The residual map shows a structure at ∼7 σ level to the west in the outer dust ring, indicating a real azimuthal asymmetry. 5. DISCUSSIONS The most intriguing result in our continuum observations is the detection of the inner dust ring at r ∼4 au and the cavity inside. Combining the SED and measured accretion rate of the system ( M ∼ 6×10−9 M⊙/yr; Manara et al. 2014, only slightly lower than that of typical T Tauri stars; Najita et al. 2015), we now have a more complete picture of its inner region: the cavity inside r ∼4 au has no detectable dust, consistent with the absence of NIR excess on the SED; however the cavity must have a substantial amount of gas in order to sustain a close-to-normal accretion rate. While dust cavities are commonly found in ALMA continuum observations nowdays (e.g. Hashimoto et al. 2015; Isella et al. 2016; Tang et al. 2017; Dong et al. 2017b, see also van der Marel et al. 2018 for a gallery), the inner cavity in the DM Tau disk, together with the cavity at 2.4 au in the TW Hya disk (Andrews et al. 2016; Tsukagoshi et al. 2016), are among the smallest, visible only in long baseline ALMA observations. The origin of the inner cavity is unclear. The measured close-to-normal accretion rate of DM Tau disfavors photoevaporation (e.g., Alexander et al. 2006). A planet can open a gap in gas (e.g., Lin & Papaloizo 1993), in which case the outer edge of the gap (even a shallow on) acts as a "dust filter", trapping ∼millimeter-sized large dust grains and forming a cavity in them (e.g., Rice et al. 2006; Zhu et al. 2012), consistent with our ALMA dust observations. Note that our gas observations are performed with 12CO, to which the disk easily becomes optically thick. We therefore cannot detect a possible gas gap inside the edge of the inner dust cavity. Future observations of optically thinner CO isotopologues are needed to probe the gas surface density structure across the inner cavity (e.g., van der Marel et al. 2016). However, as argued in Zhu et al. (2011), disk -- planet interactions have AASTEX An Inner Disk around DM Tau 9 difficulties in depleting the small (µm-sized) dust in the inner disk, which tend to flow in with the gas. The DM Tau disk is at the extreme -- there is no detectable dust inside the inner cavity based on the SED, while CO emission extends all the way toward the central star. Alternatively, the presence of gas and the absence of small dust inside the cavity may be explained as icy dust being evaporated inside the cavity, fully or partially replenishing the gas. To produce this scenario, the dust that enters the inner cavity has to comprise fully evaporable volatiles (e.g., water), and the gas inside the cavity must be rich in them. If all small icy grains do not evaporate, however, the decreased grain size inside the evaporation front increases the fragmentation efficiency (e.g., Pinilla et al. 2017), thus enriching the cavity with small grains well coupled to the gas, inconsistent with the SED. Overall, both the planet scenario and the grain evaporation scenario have advantage and disadvantage, and additional observations are needed to determine the origin of the inner cavity. The inner ring is located at a strikingly similar distance to the main asteroid belt in the solar system between Mars and Jupiter. The main belt, located beyond the water snowline at 2.7 AU (e.g., Abe et al. 2000), is thought to have profoundly impacted water delivery onto Earth (e.g., Morbidelli et al. 2000). A fraction of the water snowballs in the main belt found their way into the inner solar system and bombarded the early Earth, as their orbits were perturbed by Jupiter. The water snowline in the DM Tau system, with a host star less luminous than the proto-Sun, should be located closer to its host star than that in the solar system (see also Martin & Livio 2013; Notsu et al. 2016). Should terrestrial planets be forming inside its snowline, water delivery from the inner ring onto the inner planets might be plausible. As shown in § 4, an axisymmetric inner dust ring is consistent with the data. On the other hand, 12CO line emission is clearly non-axisymmetric on an ∼au scale -- the peak 12CO emission is located at the north blob (Figure 2), while the rotational center of CO roughly coincides with the stellar location and the center of the inner cavity. A candidate massive planet with Mp ∼3 MJup (COND model at 1 Myr; Baraffe et al. 2003) at a separation of r ∼6 au was detected by Keck sparse aperture masking interferometric observations (Willson et al. 2016). Whether this candidate planet introduces the asymmetry seen in CO emission is yet to be explored. The continuum emission in the gap between the inner and outer rings (∼4 -- 20 au) is substantially suppressed by a factor of ∼40 relative to the outer ring. The gap also has a steep inner edge (Figure 1d). Both features are consistent with predictions of gap opening by planets (Zhu et al. 2011; Dong et al. 2015), but inconsistent with some other gap formation mechanisms such as the secular gravitational instability (SGI; e.g., Takahashi & Inutsuka 2014) and sintering of dust grains (Okuzumi et al. 2016) -- the latter is expected to produce shallow gaps with depletion factors less than 10 (Okuzumi et al. 2016), while the formation timescale of the ring at r ∼10 -- 20 au in the former mechanism is much longer than DM Tau's age. In contrast to the symmetric inner ring, the outer ring has a small azimuthal asymmetry at P.A. ∼270◦, similar to the asymmetries found in a few other disks, but with one of the lowest contrast levels, ∼1.3:1 (cf., ∼130 in Oph IRS 48; van der Marel et al. 2013, ∼30 in HD 142527; Fukagawa et al. 2013). This asymmetry may be caused by a local enhancement in the dust surface density, possibly the remnants of particle trapping in a vortex (Barge et al. 2017), or temperature, if the south-west side is the far side while we observe the inner rim of the outer ring there (see e.g., Muto et al. 2015; Soon et al. 2017, for a discussion of HD 142527). Note that if the outer ring is optically thick and the asymmetry traces dust surface density variations, the small contrast seen in the surface 10 Kudo et al. brightness may not trace the contrast in the dust surface density, which can be much bigger than 1.3. Our observations also reveal an extended outer disk beyond the outer ring at r >0.′′4 in the dust continuum, and a shallow gap inside. The presence of these structures has been previously suggested by Zhang et al. (2016) and Bergin et al. (2016b). Our new ALMA observations reveal exciting new details, and yet raise more questions, in the DM Tau system. With an exo-asteroid belt under formation, the DM Tau disk will continue to offer crucial insights into disk evolution and planet formation in the terrestrial -- planet -- forming region. We thanks an anonymous referee for a helpful review of the manuscript. This paper makes use of the following ALMA data: ADS/JAO.ALMA#2017.1.01460.S and ADS/JAO.ALMA#2013.1.00498.S. ALMA is a partnership of ESO (representing its member states), NSF (USA) and NINS (Japan), together with NRC (Canada), NSC and ASIAA (Taiwan), and KASI (Republic of Korea), in coop- eration with the Republic of Chile. The Joint ALMA Observatory is operated by ESO, AUI/NRAO and NAOJ. This work was supported by NAOJ ALMA Scientific Research Grant Number 2016-02A and JSPS KAKENHI Grant Numbers 17K14258, 17K14244, 26800106, 15H02074 and 17H01103. Y.H is supported by JPL/Caltech under a contract from NASA. Facility: ALMA Software: CASA v5.1.1; (McMullin et al. 2007), vis sample (https://github.com/AstroChem/vis sample), emcee (Foreman-Mackey et al. 2013) AASTEX An Inner Disk around DM Tau 11 In this section, we present the DM Tau's SED and the corner plot with the MCMC posteriors in Figure 5 calculated in § 4. APPENDIX ✲✄ ✶ ✲☎✆ ✁✂ ✲☎☎ ✁✂ ✲☎✝ ✁✂ ✲☎✞ ✁✂ ✲☎✟ ✁✂ ✁✂ ✲☎ ✁✂ ✆ ✁✂ ☎ ✁✂ ✝ ✁✂ ✌✍✎✏✑✏✒✓✔✕ ✖✗✘✙ ✞ ✁✂ ✟ ✁✂ Figure 4. DM Tau's SED ☎ ✄ ✆ ✝ ✞ ✟ ✠ ✡ ☛ ☞ ☛ ☞ Kudo et al. ✁ ✂✄✂☎ ✆✝ ✞✟✞✠ ✞✟✞✡ ✬✭ ✁ ☎✄☎✂✆✝ ✞✟✞✪ ✞✟✞✫ ✬✰ ✁ ✂✄✂☎ ✮✯ ✞✟✞✡ ✞✟✞✠ ❇✣✤✥ ✁ ❅✄☎✸ ✮✯ ✞✟✡✡ ✞✟✡✫ ❇✦✤✧★ ✁ ✵☎✄✂✂✮✯ ✞✟✞✡ ✞✟✞✡ ❇✦✤✧✩ ✁ ✶✷✄✸✹ ✮✯ ✞✟✫✺ ✞✟✫✻ ❑★▲▼◆ ✼✽✾✿✾❀ ✾✿✾❀ ❈❉ ✁ ☎✂ 12 ☛ ☞✌ ☞✍ ✢✣✤✥ ✢✦✤✧★ ✢✦✤✧✩ ✖✗ ✙ ✢✣★ ✢✣✩ ✱✲✳✴ ✎✏✑ ✒✏✓ ✒✏✎ ✑✏✔ ✑✏✕ ✎✏✑ ✒✏✓ ✒✏✎ ✑✏✔ ✑✏✕ ❏✏✜ ✚✏✔ ✕✏✕ ✎✏✛ ✒✏✚ ✎✛ ✎✓ ✎✜ ✎✑ ✒❏ ✛✛ ✛✜ ✔❏ ✔✒ ❏✚ ✑✏✑ ✘✑✏✓ ✘✒✏✎ ✘✒✏✔ ✘✎✏✕ ✑✏✑ ✘✑✏✓ ✘✒✏✎ ✘✒✏✔ ✘✎✏✕ ✎✛ ✎✜ ✒❏ ✒✎ ✓ ✒✜✛ ✒✎✎ ✒✑✕ ✔❏ ❏✑ ✒✑✚ ✒✑✎ ✛✛ ✛✓ ✛✜ ❊ ✁ ❁☎✄❂✵ ✮✯ ✞✟✞✪ ✞✟✞✺ ❇✣★ ✁ ☎❂✄✵❂ ✮✯ ✞✟✻✞ ✠✟✠❃ ❇✣✩ ✁ ☎✵✹✄☎☎✮✯ ✠✟✞✪ ✠✟✞✪ ❋●❍■ ✁ ❄❅✄❅✂ ✮✯ ✞✟✪✡ ✞✟❃❆ ✑✏✓ ✑✏✔ ✒✏✑ ✑✏✕ ✑✏✎ ✑✏✕ ☛ ✒✏✎ ✑✏✔ ☞✌ ✒✏✓ ✎✏✑ ✑✏✕ ✒✏✎ ✑✏✔ ☞✍ ✒✏✓ ✎✏✑ ✒✏✚ ✎✏✛ ✕✏✕ ✚✏✔ ❏✏✜ ✒❏ ✎✑ ✎✜ ✎✓ ✎✛ ❏✚ ✔✒ ✔❏ ✛✜ ✛✛ ✢✣✤✥ ✢✦✤✧★ ✢✦✤✧✩ ✘✎✏✕ ✘✒✏✎ ✘✒✏✔ ✖✗ ✘✑✏✓ ✑✏✑ ✘✎✏✕ ✘✒✏✔ ✘✒✏✎ ✙ ✘✑✏✓ ✑✏✑ ✓ ✒✎ ✒❏ ✎✜ ✎✛ ❏✑ ✔❏ ✢✣★ ✒✎✎ ✒✑✕ ✢✣✩ ✒✜✛ ✛✜ ✛✓ ✛✛ ✒✑✎ ✒✑✚ ✱✲✳✴ Figure 5. Corner plot with the MCMC posterior probability distribution calculated in § 4. The histograms on diagonal are marginal distributions of 11 parameters, provided by the last 500 steps of the 100 walkers' chain. The vertical dashed lines in the histograms represent the median values and the 1 σ confidence intervals of parameters, which are also shown in the titles. The off-diagonal plots show the correlation for corresponding pairs of parameters. AASTEX An Inner Disk around DM Tau 13 REFERENCES Abe, Y., Ohtani, E., Okuchi, T., Righter, K., & Drake, M. 2000, Water in the Early Earth, ed. R. M. Canup, K. Righter, & et al., 413-433 Alexander, R. D., Clarke, C. J., & Pringle, J. E. 2006, MNRAS, 369, 229 Isella, A., Guidi, G., Testi, L., et al. 2016, Phys. Rev. Lett., 117, 251101 Johansen, A., Youdin, A., & Klahr, H. 2009, ApJ, 697, 1269 Kenyon, S. J., & Hartmann, L. 1995, ApJS, 101, ALMA Partnership, Brogan, C. L., P´erez, L. M., 117 et al. 2015, ApJL, 808, L3 Andrews, S. M., Wilner, D. J., Espaillat, C., et al. 2011, ApJ, 732, 42 Andrews, S. M., Wilner, D. J., Zhu, Z., et al. 2016, ApJL, 820, L40 Baraffe, I., Chabrier, G., Barman, T. S., Allard, F., & Hauschildt, P. H. 2003, A&A, 402, 701 Barge, P., Ricci, L., Carilli, C. L., & Previn-Ratnasingam, R. 2017, a, 605, A122 Beckwith, S. V. W., & Sargent, A. I. 1991, ApJ, 381, 250 Beckwith, S. V. W., & Sargent, A. I. 1991, ApJ, 381, 250 Bergin, E., Calvet, N., Sitko, M. L., et al. 2004, ApJL, 614, L133 Bergin, E. A., Du, F., Cleeves, L. I., et al. 2016a, ApJ, 831, 101 -- . 2016b, ApJ, 831, 101 Calvet, N., D'Alessio, P., Watson, D. M., et al. 2005, ApJL, 630, L185 Dong, R., Li, S., Chiang, E., & Li, H. 2017a, ApJ, 843, 127 Dong, R., Zhu, Z., & Whitney, B. 2015, ApJ, 809, 93 Dong, R., van der Marel, N., Hashimoto, J., et al. 2017b, ApJ, 836, 201 Dong, R., Liu, S.-y., Eisner, J., et al. 2018, ApJ, 860, 124 Dzyurkevich, N., Flock, M., Turner, N. J., Klahr, H., & Henning, T. 2010, a, 515, A70 Espaillat, C., Muzerolle, J., Najita, J., et al. 2014, Protostars and Planets VI, 497 Foreman-Mackey, D., Hogg, D. W., Lang, D., & Goodman, J. 2013, PASP, 125, 306 Fukagawa, M., Tsukagoshi, T., Momose, M., et al. 2013, PASJ, 65, L14 Gaia Collaboration, Brown, A. G. A., Vallenari, A., et al. 2018, ArXiv e-prints, arXiv:1804.09365 Hartmann, L., Calvet, N., Gullbring, E., & D'Alessio, P. 1998, ApJ, 495, 385 Kretke, K. A., & Lin, D. N. C. 2007, ApJL, 664, L55 Loomis, R. A., Cleeves, L. I., Oberg, K. I., Guzman, V. V., & Andrews, S. M. 2015, ApJL, 809, L25 Lynden-Bell, D., & Pringle, J. E. 1974, MNRAS, 168, 603 Manara, C. F., Testi, L., Natta, A., et al. 2014, a, 568, A18 Martin, R. G., & Livio, M. 2013, MNRAS, 428, L11 McMullin, J. P., Waters, B., Schiebel, D., Young, W., & Golap, K. 2007, in Astronomical Society of the Pacific Conference Series, Vol. 376, Astronomical Data Analysis Software and Systems XVI, ed. R. A. Shaw, F. Hill, & D. J. Bell, 127 Morbidelli, A., Chambers, J., Lunine, J. I., et al. 2000, Meteoritics and Planetary Science, 35, 1309 Muto, T., Tsukagoshi, T., Momose, M., et al. 2015, PASJ, 67, 122 Najita, J. R., Andrews, S. M., & Muzerolle, J. 2015, MNRAS, 450, 3559 Nakagawa, Y., Sekiya, M., & Hayashi, C. 1986, Icarus, 67, 375 Notsu, S., Nomura, H., Ishimoto, D., et al. 2016, ApJ, 827, 113 Okuzumi, S., Momose, M., Sirono, S.-i., Kobayashi, H., & Tanaka, H. 2016, ApJ, 821, 82 Pi´etu, V., Dutrey, A., & Guilloteau, S. 2007, a, 467, 163 Pinilla, P., Pohl, A., Stammler, S. M., & Birnstiel, T. 2017, ApJ, 845, 68 Pinilla, P., Tazzari, M., Pascucci, I., et al. 2018, ArXiv e-prints, arXiv:1804.07301 Rau, U., & Cornwell, T. J. 2011, a, 532, A71 Rice, W. K. M., Armitage, P. J., Wood, K., & Lodato, G. 2006, MNRAS, 373, 1619 Hashimoto, J., Tsukagoshi, T., Brown, J. M., et Semenov, D., Favre, C., Fedele, D., et al. 2018, al. 2015, ApJ, 799, 43 ArXiv e-prints, arXiv:1806.07707 14 Kudo et al. Soon, K.-L., Hanawa, T., Muto, T., Tsukagoshi, T., & Momose, M. 2017, PASJ, 69, 34 Takahashi, S. Z., & Inutsuka, S.-i. 2014, ApJ, 794, 55 Tang, Y.-W., Guilloteau, S., Dutrey, A., et al. 2017, ApJ, 840, 32 Tsukagoshi, T., Nomura, H., Muto, T., et al. 2016, ApJL, 829, L35 Weidenschilling, S. J. 1977, MNRAS, 180, 57 Willson, M., Kraus, S., Kluska, J., et al. 2016, a, 595, A9 Zapata, L. A., Rodr´ıguez, L. F., & Palau, A. 2017, ApJ, 834, 138 Zhang, K., Bergin, E. A., Blake, G. A., et al. 2016, ApJL, 818, L16 van der Marel, N., Cazzoletti, P., Pinilla, P., & Zhu, Z., Nelson, R. P., Dong, R., Espaillat, C., & Garu, A. 2016, ApJ, 832, 178 van der Marel, N., van Dishoeck, E. F., Bruderer, S., et al. 2013, Science, 340, 1199 van der Marel, N., Williams, J. P., Ansdell, M., et al. 2018, ApJ, 854, 177 Hartmann, L. 2012, ApJ, 755, 6 Zhu, Z., Nelson, R. P., Hartmann, L., Espaillat, C., & Calvet, N. 2011, ApJ, 729, 47
1010.0179	1	1010	2010-10-01T14:58:49	Transiting exoplanets from the CoRoT space mission. XV. CoRoT-15b: a brown dwarf transiting companion	[ "astro-ph.EP" ]	We report the discovery by the CoRoT space mission of a transiting brown dwarf orbiting a F7V star with an orbital period of 3.06 days. CoRoT-15b has a radius of 1.12 +0.30 -0.15 Rjup, a mass of 63.3 +- 4.1 Mjup, and is thus the second transiting companion lying in the theoretical mass domain of brown dwarfs. CoRoT-15b is either very young or inflated compared to standard evolution models, a situation similar to that of M-dwarfs stars orbiting close to solar-type stars. Spectroscopic constraints and an analysis of the lightcurve favors a spin period between 2.9 and 3.1 days for the central star, compatible with a double-synchronisation of the system.	astro-ph.EP	astro-ph	Astronomy&Astrophysicsmanuscript no. CoRoT15accepted June 2, 2018 c(cid:13) ESO 2018 Transiting exoplanets from the CoRoT space mission⋆ XV. CoRoT-15b: a brown dwarf transiting companion Bouchy, F.1,2, Deleuil, M.3, Guillot, T.4, Aigrain, S.5, Carone, L.6, Cochran, W.D.7, Almenara, J.M.8,9, Alonso, R.10, Auvergne, M.11, Baglin, A.11, Barge, P.3, Bonomo, A. S.3, Bord´e, P.12, Csizmadia, Sz.13, De Bondt, K.3, Deeg, H.J.8,9, D´ıaz, R.F.1, Dvorak, R.14, Endl, M.7, Erikson, A.13, Ferraz-Mello, S.15, Fridlund, M.16, Gandolfi, D.16,17, Gazzano, J.C.3, Gibson, N.5, Gillon, M.18, Guenther, E.17, Hatzes, A.17, Havel, M.4, H´ebrard, G.1, Jorda, L.3, L´eger, A.12, Lovis, C.10, Llebaria, A.3, Lammer, H.19, MacQueen, P.J.7, Mazeh, T.20, Moutou, C.3, Ofir, A.20, Ollivier, M.12, Parviainen, H.8,9, Patzold, M.6, Queloz, D.10, Rauer, H.13,21, Rouan, D.11, Santerne, A.3, Schneider, J.22, Tingley, B.8,9, and Wuchterl, G.17 (Affiliations can be found after the references) Received ; accepted ABSTRACT We report the discovery by the CoRoT space mission of a transiting brown dwarf orbiting a F7V star with an orbital period of 3.06 days. CoRoT- 15b has a radius of 1.12+0.30 −0.15 RJup, a mass of 63.3±4.1 MJup, and is thus the second transiting companion lying in the theoretical mass domain of brown dwarfs. CoRoT-15b is either very young or inflated compared to standard evolution models, a situation similar to that of M-dwarfs stars orbiting close to solar-type stars. Spectroscopic constraints and an analysis of the lightcurve favors a spin period between 2.9 and 3.1 days for the central star, compatible with a double-synchronisation of the system. Key words. brown dwarfs - planetary systems - low-mass - techniques: photometry - techniques: radial velocities - techniques: spectroscopic 1. Introduction 2. CoRoT observations The CoRoT space mission (Baglin et al. 2009), in operation since the start of 2007 and extended to the end of 2013, is de- signed to find transiting exoplanets. A natural product of this mission is that any object with size of Jupiter or lower that transits its host star can be detected. This includes stellar and sub-stellar companions such as M-dwarfs and brown dwarfs (BDs). In the mass-radius diagram of transiting companions or- biting solar-type stars, there is up until now only one known brown-dwarf, CoRoT-3b (Deleuil, et al. 2008), located in the gap in mass between planetary and low-mass star companions 1. Determination of the physical properties of such objects are fundamental to understand the link between the population of planets and low-mass stars and to distinguish the formation and evolution processes of the two populations. We report in this paper the discovery of a new transiting brown-dwarf by CoRoT established and characterized thanks to ground-based follow-up observations. CoRoT-15b, with an esti- mated radius of 1.12 RJup and an estimated mass of 63.3 MJup, orbits in 3.06 days an F-type dwarf with solar metallicity. ⋆ The CoRoT space mission, launched on December 27th 2006, has been developed and is operated by CNES, with the contribution of Austria, Belgium, Brazil , ESA (RSSD and Science Programme), Germany and Spain. Observations made with HARPS spectrograph at ESO La Silla Observatory (184.C-0639). 1 After submission of our manuscript, Johnson et al. (2010) reported the discovery of the transiting brown dwarf LHS6343C with a radius of 0.996 ± 0.026 RJup and a mass of 70.6 ± 2.7 MJup. SRa02 was the seventh field observed with CoRoT in the second year after its launch. It corresponds to the third short run and was located towards the so-called galactic anti-center direction. This run started on 2008 October 11 and ended on 2008 November 12, constituting of a total of 31.7 days of almost-continuous ob- servations. More than 30 multi-transiting candidates for planets were found among the 10265 targets of the SRa02 field. About half of them were clearly identified as binaries from light-curve analy- sis and around tenth of high priority planet-size candidates were selected in this short run including SRa02 E1 4106 afterwards called CoRoT-15. The various ID of this target, including coor- dinates and magnitudes are listed in table 1. 3. CoRoT light curve analysis CoRoT-15, with an estimated V magnitude close to 16, is lo- cated at the faint end of the stellar population observed by CoRoT . In this magnitude domain, the signal is not sufficient to split the photometric aperture in three colors and the ex- tracted lightcurve, also called white lightcurve, is monochro- matic (Llebaria & Guterman 2006). The sampling rate was at 512 seconds during the whole run and not oversampled to 32 sec- onds since this candidate was not identified with the alarm mode (Surace et al. 2008). Fig. 1 shows the lightcurve of CoRoT-15 delivered by the N2 data levels pipeline. This lightcurve is quite noisy and is affected by several high energy particles impacts which result in hot pixels as well as possible stellar variability. 0 1 0 2 t c O 1 . ] P E h p - o r t s a [ 1 v 9 7 1 0 . 0 1 0 1 : v i X r a 2 Bouchy, F. et al.: CoRoT-15b: a brown dwarf transiting companion Table 1. IDs, coordinates and magnitudes. SRa02 E1 4106 221686194 0961-0097866 06282781+0611105 06:28:27.82 +06:11:10.47 CoRoT window ID CoRoT ID USNO-B1 ID 2MASS ID RA (J2000) Dec (J2000) B1a B2a R1a R2a Ia Jb Hb Kb a from USNO-B1 catalog b from 2MASS catalog 16.85 16.59 15.47 15.43 14.83 13.801 ± 0.026 13.423 ± 0.037 13.389 ± 0.050 served during a transit event on 2010 January 13 with time-series photometry at the IAC80 telescope, from HJD 2455209.480 to .676. The resultant lightcurve was not sufficiently precise to identify the expected transit on the target, due to photometric errors introduced by the presence of thin cirrus. However, the absence of large brightness variations in the neighboring stars al- lowed us to exclude nearby eclipsing binaries as a source of the signals that were observed by CoRoT . The contamination factor was derived from a measure of the distance and brightness of the nearby stars on a subset of these IAC80 R-filter images ob- tained with the best seeing of that night (1.7 arcsec). Ten nearby stars were identified, with six of them contaminating the CoRoT window aperture with a flux level that amounts to 1.9±0.3% of the main target flux. We checked furthermore that none of the known nearby stars is bright enough to be contaminating eclips- ing binaries. The image of the sky around CoRoT-15 is shown in Fig. 2. 1 arcmin N Fig. 1. Raw light curve of CoRoT-15. A first analysis of the lightcurve (LC), based on a trapezoidal fit to each individual transit, reveals periodic transits with depth of 0.68% and a period of 3.0608±0.0008 days. Since the CoRoT lightcurve is relatively noisy, no meaning- ful limits to either the visible-light albedo of the companion nor to its dayside surface temperature could be established. From the raw lightcurve, we tried to estimate the stellar ro- tation period. We filtered out variations with timescales shorter than 15 data points (∼2.13h) to reduce sensitivity to the satel- lite orbital effects and other short term variations, and we then removed the transits. The computed LS periodogram appears to be quite noisy and affected by low-level discontinuities in the lightcurve. There is tentative evidence of a possible rotational modulation at either 2.9, 3.1 or 6.3 days in the light curve, but the data does not enable us to estimate the period more precisely or to distinguish between these values. 4. Ground-based observations 4.1. Ground-basedPhotometricfollow-up Ground-based photometry was performed with the aim of refin- ing the target ephemeris, to verify that none of its closest con- taminant stars correspond to an eclipsing binary and to deter- mine the contamination from nearby stars inside CoRoT 's pho- tometric aperture mask (Deeg et al. 2009). CoRoT-15 was ob- E Fig. 2. The image of the sky around CoRoT-15 (star in the cen- tre). Left: R-filter image with a resolution of 1.7" taken with the IAC80 telescope. Right: Image taken by CoRoT , at the same scale and orientation. The jagged outline in its centre is the pho- tometric aperture mask; indicated are also CoRoT 's x and y im- age coordinates and positions of nearby stars which are in the Exo-Dat (Deleuil et al 2009) database. 4.2. Radialvelocityfollow-up Radial velocity (RV) observations of CoRoT-15 were performed with the HARPS spectrograph (Mayor et al. 2003) based on the 3.6-m ESO telescope (Chile) as part of the ESO large program 184.C-0639 and with the HIRES spectrograph (Vogt et al. 1994) based on the 10-m Keck-1 telescope as part of NASA's key sci- ence project to support the CoRoT mission. HARPS was used with the observing mode obj AB, with- out simultanenous thorium in order to monitor the Moon back- ground light on the second fiber. The exposure time was set to 1 hour. A set of 9 spectra was recorded between November 24th 2009 and February 21th 2010. We reduced HARPS data and computed RVs with the pipeline based on the cross-correlation techniques (Baranne et al. 1996; Pepe et al. 2002). The signal-to- noise ratio (SNR) per pixel at 550 nm is in the range 3 to 7.8 for this faint target. It corresponds to the faint end in magnitude for HARPS follow-up observations. Radial velocities were obtained by weighted cross-correlation with a numerical G2 mask. HIRES observations were performed with the red cross- disperser and the I2-cell to measure RVs. We used the 0.861" wide slit that leads to a resolving power of R ≈ 45, 000. The Bouchy, F. et al.: CoRoT-15b: a brown dwarf transiting companion 3 Table 2. Radial velocity measurements of CoRoT-15 obtained by HARPS and HIRES. BJD is the Barycentric Julian Date. BJD -2 400 000 RV [km s−1] ±1 σ [km s−1] HARPS 3.6-m ESO 55159.81312 55169.77922 55235.60498 55236.59903 55241.55686 55243.55544 55246.56091 55247.55392 55248.54372 9.944 2.107 2.849 8.057 0.119 -2.422 -2.684 -1.318 10.235 HIRES 10-m Keck 55221.76973 55221.81404 55221.82490 55221.89482 55221.90689 55221.97788 55221.99025 55222.98656 55222.99752 55223.73904 55224.82036 55224.83152 55225.04022 1.084 0.890 0.556 -0.232 -0.492 -1.869 -1.750 -3.193 -3.587 6.527 1.593 1.818 -1.345 0.308 0.315 0.380 0.392 0.338 0.222 0.434 0.411 0.357 0.291 0.189 0.289 0.220 0.216 0.353 0.147 0.340 0.226 0.113 0.374 0.303 0.515 contamination of the HIRES spectra by scattered moon light was significant for this faint target, but the 7" tall decker allowed us to properly correct for the background light. Three spectra without the Iodine cell were obtained on 2010 december 2 and january 9. These 3 spectra were co-added to serve as stellar template for the RV measurements, and to be used for the determination of stellar parameters (see Sect. 4.3). Over a 4-night run from 2010 January 25 to January 28 we have collected 13 spectra of CoRoT-15 with the I2-cell. The average SNR of the spectra with the I2-cell range from 13 to 22 (per pixel) in the iodine region from 500-620 nm. Differential radial velocities were computed using the Austral Doppler code (Endl et al. 2000). Nine RV measurements were made during a transit event on 2010 January 25 but were not sen- sitive enough to detect the signature of the Rossiter-McLaughlin effect expected to have, for this high rotating star, an amplitude of about 100 m s−1. The HARPS and HIRES radial velocities are given in Table 2. The two sets of relative radial velocities were simul- taneously fitted with a Keplerian model, with the epoch and pe- riod of the transit being fixed at the CoRoT value and with an adjusted offset between the two different instruments. No sig- nificant eccentricity was found and we decided to set it to zero. We found a systematic shift in phase using the CoRoT period of P=3.0608 days. It comes from the fact that the quite large uncertainty on the CoRoT period (69 seconds) may introduce after one year a systematic shift of more than 2 hours. When we adjust the period with the RVs fixing the transit epoch as de- termined from CoRoT lightcurve Tt=54753.5570 ± 0.0028, the best solution is obtained for P=3.06039 ± 0.00014 days and a semi-amplitude K=7.376 ± 0.090 km s−1. The dispersion of the residuals is 0.325 km s−1 and the reduced χ2 is 0.90. The joint analysis of the photometric and RV data, presented in Section 5, does not change significantly the results. Figure 3 shows all the radial velocity measurement after subtracting the RV offset and phase folded to the updated orbital period. Fig. 3. Phase-folded radial velocity measurements of CoRoT-15 with HARPS (dark circle) and HIRES (open triangle) 4.3. spectralclassification Three HIRES spectra of CoRoT-15 were acquired without the iodine cell. Each spectrum was set in the barycentric rest frame, cleaned from cosmic rays and from the moon reflected light. The co-addition of these 3 spectra results in a master spectrum cov- ering the wavelength range from 4100 Å to 7800 Å with a SNR per element of resolution in the continuum ranging from 20 at 5300 Å up to 70 at 6820 Å. The 9 co-added HARPS spectra unfortunately did not permit to reach a better SNR. From the analysis of a set of isolated lines, we derived a v sin i of 19 ± 2 km s−1. The spectroscopic analysis was car- ried out using the same methodology as for the previous CoRoT planets and described in details in Bruntt et al. (2010). However, the moderate SNR of the master spectrum of this faint target, combined to the marked rotational broadening of the spectral lines prevented an accurate measurement of the star's photo- spheric parameters. The derived stellar parameters are reported in Table 3. Following Santos et al. (2002) methodology, we also esti- mated the v sin i and an [Fe/H] index from the HARPS cross- correlation average parameters (FWHM and surface). Assuming a B − V of 0.5, we estimated the v sin i of the target to be 16±1 km s−1 an [Fe/H] index close to zero (solar metallicity) in agreement with the spectral analysis. 5. System parameters The time span of the CoRoT light curve is relatively short, and the RV data was collected on year later. Jointly analysing the two datasets therefore yields significantly improved constraints on the period P and time of transit centre Ttr. To do this, we used a Metropolis-Hastings Markov Chain Monte Carlo (MCMC) al- gorithm (see appendix 1 of Tegmark et al. 2004 for a general description of MCMC algorithms and Winn et al. 2008 and ref- erences therein for a detailed description of their application to transits). This has the added advantage of yielding full posterior probability distributions for the fitted parameters, ensuring that the effects the well known degeneracy between the orbital incli- nation i and system scale a/R⋆ (which leads to highly skewed probability distributions for these parameters, as well as for the radius ratio Rc/R⋆) are properly taken into account in the final uncertainties. 4 Bouchy, F. et al.: CoRoT-15b: a brown dwarf transiting companion The light curve was first pre-processed to remove out-of- transit variability as follows. Outliers were identified using an iterative non-linear filter (see Aigrain et al. 2009), and a straight line was fitted to the region around each transit. The effect of contamination reported in section 4.1 was taken into account by subtracting a constant amount of flux equal to 1.9% of the mean flux from the light curve. Each section of the light curve was thus normalised, and a visual check was performed to ensure that no residual discontinuities affected the preprocessed light curve. The photometric uncertainties were then estimated from the out-of-transit scatter in the preprocessed light curve section around each transit. The light curve was modeled using the for- malism of Mandel & Agol (2002). Given the relatively low SNR of the transits, we opted to fix the quadratic limb-darkening pa- rameters ua and ub at the values given by Sing et al. (2010) for the star's effective temperature, gravity and metallicity (adopted values: ua = 0.32, ub = 0.30). The RV data were modelled us- ing a Keplerian orbit with eccentricity fixed at zero, since the data show no evidence for a significant eccentricity. The rela- tive zero-point of the HARPS and HIRES velocities, δV0, was allowed to vary freely. The parameters of the MCMC were thus P, Ttr, Rc/R⋆, a/R⋆, the radial velocity semi amplitude K, the systemic radial velocity Vsys and δV0. After an initial chain of 105 steps to adjust the MCMC step sizes for each parameter, we ran 10 MCMC chains of 105 steps, each with different starting points. The convergence of the chains was checked using the Geldman-Rubin statistic (Geldman & Rubin 1992, Brooks & Geldman 1997). The chains were then combined (after discarding the first 10% of each chain, where the MCMC is settling from its starting point) to produce poste- rior probability distributions for each parameters. We report in Table 3 the median of the probability distribution for each pa- rameter2. To estimate uncertainties for each parameter, we com- puted the range of values which encloses 68.5% of the proba- bility distribution (rejecting 16.25% at each extremum). Our un- certainties thus correspond to 68.5% confidence intervals, just as classical 1-σ uncertainties do for a Gaussian distribution. The best-fit transit model is shown superimposed on the folded light curve in Figure 4. To highlight the correlations between b, Rp/R⋆ and a/R⋆, and explain the rather large resulting uncertainties, we also show the posterior probability distributions and 2-D pro- jections of the combined MCMC chain for these parameters in Figure 5. We used the photospheric parameters from spectral analysis and the stellar density derived from the transit modeling to de- termine the star's fundamental parameters in the (Teff, M1/3 ⋆ /R⋆) space. Using STAREVOL evolution tracks (Palacios, private com.), we find the stellar mass to be M⋆ = 1.32 ± 0.10 M⊙ and the stellar radius R⋆ = 1.46 +0.31 −0.14 R⊙, with an age in the range 1.14 -- 3.35 Gyr. This infers a surface gravity of log g = 4.23 +0.12 −0.20, in good agreement with the spectroscopic value. Calculations using CESAM (Morel & Lebreton 2008, see also Guillot & Havel 2010) confirm these solutions. The age constraints, 1.9±1.7 Gyr, are however extremely weak, and yield 2 The choice of which statistic to report is a somewhat tricky one. When the distributions are (close to) Gaussian, the median, most prob- able and best-fit values coincide. When the distributions are skewed, as in the case of b for example, the median, most probable and best- fit values can differ signficiantly. Whilst the best fit value maximises the merit function for the particular dataset being analysed, it has no special physical meaning. We adopt the value which divides the proba- bility distribution in half, namely the median, as it is arguably the most physically meaningful. Fig. 4. Folded, detrended light curve of CoRoT-15 (top), show- ing the best fit transit model (red solid line) and residuals (bot- tom). Fig. 5. Selected posterior probability distributions and two- dimensional correlations for the transit fit. The panels along the diagonal show single-parameter posterior probability distribu- tions for b, Rp/R⋆ and a/R⋆. The red, blue, and dashed blue vertical lines indicate the location of the best-fit, median and the limits of the 68.5% confidence interval for each parameter. The off-diagonal panels show, for each pair of parameters, scat- ter plots of 1000 points randomly selected from the combined MCMC chain. The density of the points approximates the joint posterior probability distribution. possible pre-main sequence solutions with extremely young ages. We derived for the transiting companion Mc=63.3±4.1 MJup and Rc=1.12+0.30 −0.15 RJup. Bouchy, F. et al.: CoRoT-15b: a brown dwarf transiting companion 5 Table 3. Star and companion parameters. Parameters Transit epoch Ttr [HJD] Orbital period P [days] Transit duration dtr [h] Value 2454753.5608±0.0011 3.06036±0.00003 3.24±0.1 Orbital eccentricity e Semi-amplitude K [ km s−1] Systemic velocity V0 harps [ km s−1] Systemic velocity V0 hires [ km s−1] 0 (fixed) 7.36±0.11 2.23±0.11 1.09±0.07 in that case however, magnetic braking in the central star (e.g. see Barker & Ogilvie 2009) will lead to a loss of angular mo- mentum that will be transferred to the orbit of the companion through tides and lead to orbital decay. We thus propose that close-in massive planets, brown dwarf or M-dwarf can survive when orbiting early or mid F-type dwarfs but that they tend to be engulfed by G-type (or late F-type) dwarfs. In the case of CoRoT-15, we thus expect that the star should be above ∼ 1.25 M⊙ to avoid efficient spin-down, and that the system should be at or close to double-synchronisation (i.e. the spin period of the star should be close to the orbital period of its companion). Radius ratio Rc/R⋆ Scaled semi-major axis a/R⋆ Impact parameter b ⋆ /R⋆ [solar units] M1/3 Stellar density ρ⋆ [g cm−3] Inclination i [deg] Effective temperature Te f f [K] Surface gravity log g [dex] Metallicity [Fe/H] [dex] Rotationl velocity v sin i [ km s−1] Spectral type Star mass [M⊙] Star radius [R⊙] Distance of the system [pc] Orbital semi-major axis a [AU] Companion mass Mc [MJup] Companion radius Rc[RJup] Companion density ρc [g cm−3] Equilibrium temperature T per eq [K] 0.0788+0.0039 −0.0029 6.68+0.49 −1.04 0.38+0.25 −0.26 0.75+0.05 −0.12 0.60+0.13 −0.28 86.7+2.3 −3.2 6350±200 4.3±0.2 0.1±0.2 19±2 F7V 1.32±0.12 1.46+0.31 −0.14 1270±300 0.045+0.014 −0.010 63.3±4.1 1.12+0.30 −0.15 59+37 −32 1740+120 −190 6. Discussion and Conclusion CoRoT-15b is one of the rare transiting companion that lies in the theoretical mass domain of brown dwarfs (13-75 MJup, if one adopts the present IAU convention). Contrary to CoRoT-3b (Deleuil et al. 2008) that is located in the overlapping region be- tween the massive planet and the brown-dwarf domain, CoRoT- 15b is well in the mass domain of BDs. Expanding a bit the mass domain, one can easily include in this ensemble the high mass "planets" (M ≥ 10 MJup) XO-3b (Johns-Krull et al. 2008) and WASP-18b (Hellier et al. 2009), and in the M-dwarf regime, OGLE-TR-122b (Pont et al. 2005a) -123b (Pont et al. 2006) - 106b (Pont et al. 2005b), and HAT-TR-205-013 (Beatty et al. 2007). Interestingly, all these objects are found to orbit F-type stars (see also Deleuil et al. 2008), with one exception: OGLE- TR-122b orbits a G-type dwarf but has a much longer orbital period (7.3 days compared to less than 4.3 days for all other ob- jects). Early- and mid-F-type dwarfs have the particularity of be- ing fast rotators, independently of their age (Nordstrom et al. 1997), a consequence of a small or inexistent outer convective zone, weak stellar winds, and reduced losses of angular momen- tum. The tides raised on a star by its close-in companion (planet, brown dwarf or star) have long been known to pause a threat to its survival (e.g. Patzold & Rauer 2002). This is true when the star's spin is slower than the orbital period of the companion, a common situation for close-in exoplanets. However, massive- enough companions have the possibility of spinning-up the star and may escape engulfment if the total angular momentum of the system is above a critical value (Levrard et al. 2009). Even ] p u J R i [ s u d a R 3.0 2.5 2.0 1.5 1.0 0.5 0.0 Cm Dra A Cm Dra B OGLE-TR-106b OGLE-TR-123b HAT-TR-205-013 XO-3b CoRoT-15b 0.5 Ga WASP-18b CoRoT-3b 5 Ga OGLE-TR-122b brown dwarfs stars 10 100 Mass [MJup] Fig. 6. Masses and radii of eclipsing brown-dwarfs and low- mass stars (circles with error bars, as labeled) compared to the- oretical mass-radius relations (lines). The lines correspond to isochrones of 0.5 (upper orange lines) and 5 Ga (lower blue lines), respectively. The dashed lines are calculated for isolated brown-dwarfs and low-mass stars. The plain lines include the effect of irradiation with Teq = 1800 K. The dash-dotted lines include irradiation and account for a 50% coverage of the photo- sphere with low-temperature spots (see text). A 5 Ga isochrone for isolated brown dwarfs/M stars from Baraffe et al. (2003) is shown for comparison (dotted line). It is interesting to see that given the v sin i and stellar ra- dius determinations, the projected spin period of the central star is P/sini∗ = 3.9+0.8 −1.1 days. An LS periodogram shows the pres- ence of many peaks possibly due to low-level discontinuities in the lightcurve. The most robust peak compatible with the v sin i determination lies between 0.32 and 0.34 cycles/day, and may thus be linked to a stellar spin period between 2.9 and 3.1 days. The CoRoT-15 system thus appears to be indeed close to double- synchronous. Further observations of the system and in particu- lar a precise determination of the stellar spin period would be a powerful mean of understanding the dynamical evolution of this system. Coupled to studies of similar systems, this may also yield strong constraints on the on the tidal dissipation factor in F-type dwarfs. CoRoT-15b is also extremely interesting for its size in com- parison with other objects in this mass range, and of evolution tracks for hydrogen-helium brown dwarfs and stars. Figure 6 shows that it appears inflated compared to standard evolution tracks for these kind of objects (Baraffe et al. 2003), although it may be compatible with a young age if the true size is at the lower-end of the one inferred from our measurements. However, we notice that the same problem arises for OGLE-TR-123b, 6 Bouchy, F. et al.: CoRoT-15b: a brown dwarf transiting companion OGLE-TR-106b and, but to a lesser extent, HAT-TR-205-013. 3 The two other known brown dwarfs with direct radius measure- ments, discovered in the 2MASS J05352184-0546085 eclipsing binary system (Stassun et al. 2006), have very large radii (5.0 and 6.5 RJup) but related to the very young age of the system (∼ 1 Myr), still in the earliest stages of gravitational contraction. In order to examine possible solutions to this puzzle (other than a systematic overestimation of the inferred sizes for the systems known thus far), we calculate evolution tracks using CEPAM (Guillot & Morel 1995), but adding the dominant ther- monuclear reaction cycle, namely the pp-chain (see Burrows & Liebert 1993). The atmospheric boundary condition is adjusted to the Baraffe et al. (2003) evolution tracks, using the analytical solution of Guillot (2010), and values of the thermal and visible mean opacities, κth = 0.04 cm2 g−1 and κv = 0.024 cm2 g−1. The model shows that irradiation effects, although quite significant for Jupiter-like planets have rather small consequences in the brown dwarf regime, and become completely negligible in the stellar regime. We also test the possibility that these inflated sizes may be explained by the presence of cold spots on the brown dwarf, similarly to a mechanism proposed to explain that M- type star in close-in binaries also appear inflated (Chabrier et al. 2007). As shown in Fig. 6, this mechanism works only in com- bination with a young age for the system, and with a rather large fraction (∼50%) of the photosphere covered with spots. An alter- native possibility is that irradiated atmospheres are much more opaque than usually thought, possibly a consequence of photo- chemistry and disequilibrium chemistry. A solar metallicity was assumed for the brown-dwarf models displayed in Fig. 6. We tested the effect of metallicity on the mean molecular mass and the opacity but did not find significant change in the radius. In any case, this shows that CoRoT-15b is a crucial object to understand both the dynamical and physical evolution of gi- ant planets, brown-dwarfs and low-mass stars. Further observa- tions aiming at obtaining more accurate spectra of the star would be highly desirable. Although this is a challenging measurement given the faintness of the target, measurement of secondary tran- sits in the infrared would be extremely interesting because they would inform us on this rare heavily irradiated brown dwarf at- mosphere. Acknowledgements. The authors wish to thank the staff at ESO La Silla Observatory for their support and for their contribution to the success of the HARPS project and operation. HIRES data presented herein were ob- tained at the W.M. Keck Observatory from telescope time allocated to the NASA through the agency's scientific partnership with the California Institute of Technology and the University of California. The French team wish to thank the Programme National de Plan´etologie (PNP) of CNRS/INSU and the French National Research Agency (ANR-08-JCJC-0102-01) for their continu- ous support to our planet search program. The team at IAC acknowledges sup- port by grant ESP2007- 65480-C02-02 of the Spanish Ministerio de Ciencia e Innovaci´on. The German CoRoT Team (TLS and the University of Cologne) acknowledges DLR grants 50OW0204, 50OW0603, and 50QP07011. MG ac- knowledges support from the Belgian Science Policy Office in the form of a Return Grant. SA & NG acknowledge support from STFC standard grand ST/G002266. FB acknowledges the continuous support of PLS-230371. References Aigrain, S., Pont, F., Fressin, F., et al. 2009, A&A, 506, 425 Baglin A., Auvergne M., Barge P., et al., 2009, Transiting Planets, Proceedings of the International Astronomical Union, IAU Symposium, 253, 71 Baraffe, I.; Chabrier, G.; Barman, T.S., et al., 2003, A&A, 402, 701 Baranne, A., Queloz, D., Mayor, M., et al. 1994, A&AS, 119, 373 3 The recent discovery of the transiting brown dwarf LHS6343C by Johnson et al. (2010) also points to an inflated companion. Barker ,A., & Ogilvie, G.I., 2009, in Cosmic Magnetic Fields: From Planets, to Stars and Galaxies, Proceddings of the International Astronomical Union, IAU Symposium, 259, 295 Beatty, T.,G., Fernandez, J.M., Latham, D.W., et al. 2007, ApJ, 663, 573 Brooks, S. P. & Geldman, A., 1997, Journal of Computational and Graphical Statistics, 7, 434 Bruntt, H., Deleuil, M., Fridlund, M., et al., 2010, A&A, 519, 51 Burrows, A. & Liebert, J. 1993, RvMP, 65, 301 Chabrier, G., Gallardo, J., & Baraffe, I. 2007, A&A, 472, L17 Deeg, H., Gillon, M., Shporer, A., et al., 2009, A&A, 506 343 Deleuil, M., Deeg, H., Alonso, R., et al., 2008, A&A, 491, 889 Endl, M., Kurster, M., Els, S., 2000, A&A, 362, 585 Geem, Z. W., Kim, J. H., Loganathan, G. V. 2001, Simulation Vol. 76, 60 Gelman, A. & Rubin, D. B., 1992, Statistical Science, 7, 457 Guillot, T. & Morel, P., 1995, A&AS, 109, 109 Guillot, T. 2010, A&A, in press (arXiv:1006.4702) Guillot & Havel 2010, A&A, submitted Hellier ,C., Anderson, D.R, Cameron, A.C., et al. 2009, Nature, 460, 1098 Johnson, J.A., Apps, K., Gazak, J.Z., et al., 2010, ApJ, submitted Johns-Krull, C.M., McCullough, P.R., Burke, C.J., et al., 2008, ApJ, 677, 657 Levrard, B., Winisdoerffer, C., Chabrier, G., et al. 2009, ApJ, 692, L9 Llebaria, A., & Guterman, P., 2006, in ESA Special Publication, 1306, 293 Mandel, K., & Agol, E. 2002, ApJ 580, L171 Mayor, M., Pepe, F., Queloz, D., et al. 2003, The Messenger, 114, 20 Morel, P., & Lebreton, Y., 2008, Ap&SS, 316, 61 Nordstrom, B., Stefanik, R.P., Latham, D.W., et al., 1997, A&ASS, 126, 21 Patzold, M., & Rauer, H., 2002, ApJ, 568, L117 Pepe, F., Mayor, M., Galland, F., et al., 2002, A&A, 388, 632 Pont, F., Melo, C.H.F., Bouchy, F., et al., 2005, A&A, 433, L21 Pont, F., Bouchy, F., Melo, C., et al., 2005, A&A, 438, 1123 Pont, F., Moutou, C., Bouchy, F., et al. 2006, A&A, 447, 1035 Santos, N.C., Mayor, M., Naef, D., et al., 2002, A&A, 392, 215 Sing, D.K., 2010, A&A, 510, 21 Stassun, K.G., Mathieu, R.D., & Valenti, J.A., 2006, Nature, 440, 311 Surace, C., Alonso, R., Barge, P., et al., 2008, SPIE Conf. Series, 7019, 111 Tegmark, M., Strauss, M.A., Blanton, M.R., et al., 2004, Physical Review D, Vol. 69, Issue 10, 103501 Vogt, S.S., Allen, S.L., Bigelow, B.C., et al. 1994, SPIE, 2198, 362 Winn, J.N., Holman, M.J., Torres, G., et al. 2008, ApJ, 683, 1076 1 Institut d'Astrophysique de Paris, UMR7095 CNRS, Universit´e Pierre & Marie Curie, 98bis Bd Arago, 75014 Paris, France 2 Observatoire de Haute Provence, CNRS/OAMP, 04870 St Michel l'Observatoire, France 3 Laboratoire d'Astrophysique de Marseille, 38 rue Fr´ed´eric Joliot- Curie, 13388 Marseille cedex 13, France 4 Universit´e de Nice-Sophia Antipolis, CNRS UMR 6202, Observatoire de la Cote d'Azur, BP 4229, 06304 Nice Cedex 4, France 5 Department of Physics, Denys Wilkinson Building Keble Road, Oxford, OX1 3RH 6 Rheinisches Institut fur Umweltforschung an der Universitat zu Koln, Aachener Strasse 209, 50931, Germany 7 McDonald Observatory, The University of Texas, Austin, TX 78712, USA Spain 8 Instituto de Astrofosica de Canarias, E-38205 La Laguna, Tenerife, 9 Departamento de Astrof´ısica, Universidad de La Laguna, E-38200 10 Observatoire de l'Universit´e de Gen`eve, 51 chemin des Maillettes, La Laguna, Tenerife, Spain 1290 Sauverny, Switzerland 11 LESIA, UMR 8109 CNRS, Observatoire de Paris, UPMC, Universit´e Paris-Diderot, 5 place J. Janssen, 92195 Meudon, France 12 Institut d'Astrophysique Spatiale, Universit´e Paris XI, F-91405 Orsay, France 13 Institute of Planetary Research, German Aerospace Center, Rutherfordstrasse 2, 12489 Berlin, Germany 14 University of Vienna, Institute of Astronomy, Turkenschanzstr. 17, A-1180 Vienna, Austria 15 IAG, University of Sao Paulo, Brazil 16 Research and Scientific Support Department, ESTEC/ESA, PO Box 299, 2200 AG Noordwijk, The Netherlands Bouchy, F. et al.: CoRoT-15b: a brown dwarf transiting companion 7 17 Thuringer Landessternwarte, Sternwarte 5, Tautenburg 5, D-07778 18 University of Li`ege, All´ee du 6 aout 17, Sart Tilman, Li`ege 1, Tautenburg, Germany Belgium 19 Space Research Institute, Austrian Academy of Science, Schmiedlstr. 6, A-8042 Graz, Austria 20 School of Physics and Astronomy, Raymond and Beverly Sackler Faculty of Exact Sciences, Tel Aviv University, Tel Aviv, Israel 21 Center for Astronomy and Astrophysics, TU Berlin, Hardenbergstr. 36, 10623 Berlin, Germany 22 LUTH, Observatoire de Paris, CNRS, Universit´e Paris Diderot; 5 place Jules Janssen, 92195 Meudon, France
1308.0563	1	1308	2013-08-02T17:49:10	Near-parabolic comets observed in 2006-2010. The individualized approach to 1/a-determination and the new distribution of original and future orbits	[ "astro-ph.EP" ]	Dynamics of a complete sample of 22 small perihelion distance near-parabolic comets discovered in the years 2006 - 2010 is studied. First, osculating orbits are obtained after a careful positional data inspection and processing, including where appropriate, the method of data partitioning for determination of pre- and post-perihelion orbit for tracking then its dynamical evolution. The nongravitational acceleration in the motion is detected for 50 per cent of investigated comets, in a few cases for the first time. Different sets of nongravitational parameters are determined from pre- and post-perihelion data for some of them. The influence of the positional data structure on the possibility of the detection of nongravitational effects and the overall precision of orbit determination is widely discussed. Secondly, both original and future orbits were derived by means of numerical integration of swarms of virtual comets obtained using a Monte Carlo cloning method. This method allows to follow the uncertainties of orbital elements at each step of dynamical evolution. The complete statistics of original and future orbits that includes significantly different uncertainties of 1/a-values is presented, also in the light of our results obtained earlier. Basing on 108 comets examined by us so far, we conclude that only one of them, C/2007 W1 Boattini, seems to be a serious candidate for an interstellar comet. We also found that 53 per cent of 108 near-parabolic comets escaping in the future from the Solar system, and the number of comets leaving the Solar system as so called Oort spike comets is 14 per cent. A new method for cometary orbit quality assessment is also proposed that leads to a better diversification of orbit quality classes for contemporary comets.	astro-ph.EP	astro-ph	Mon. Not. R. Astron. Soc. 000, 1 -- 20 (2013) Printed June 19, 2018 (MN LATEX style file v2.2) Near-parabolic comets observed in 2006 -- 2010. The individualized approach to 1/a-determination and the new distribution of original and future orbits Małgorzata Kr´olikowska1⋆ and Piotr A. Dybczy´nski2† 1Space Research Centre of the Polish Academy of Sciences, Bartycka 18A, 00-716 Warsaw, Poland 2Astronomical Observatory Institute, Faculty of Physics, A. Mickiewicz University, Słoneczna 36, 60-286 Pozna´n, Poland June 19, 2018 ABSTRACT Dynamics of a complete sample of small perihelion distance near-parabolic comets discovered in the years 2006 -- 2010 are studied (i.e. of 22 comets of qosc < 3.1 au). First, osculating orbits are obtained after a very careful positional data inspection and processing, including where appropriate, the method of data partitioning for determination of pre- and post-perihelion orbit for tracking then its dynamical evolution. The nongravita- tional acceleration in the motion is detected for 50 per cent of investigated comets, in a few cases for the first time. Different sets of nongravitational parameters are determined from pre- and post-perihelion data for some of them. The influence of the positional data structure on the possibility of the detection of nongravitational effects and the overall precision of orbit determination is widely discussed. Secondly, both original and future orbits were derived by means of numerical integra- tion of swarms of virtual comets obtained using a Monte Carlo cloning method. This method allows to follow the uncertainties of orbital elements at each step of dynamical evolution. The complete statistics of original and future orbits that includes significantly different uncer- tainties of 1/a-values is presented, also in the light of our results obtained earlier. Basing on 108 comets examined by us so far, we conclude that only one of them, C/2007 W1 Boattini, seems to be a serious candidate for an interstellar comet. We also found that 53 per cent of 108 near-parabolic comets escaping in the future from the Solar system, and the number of comets leaving the Solar system as so called Oort spike comets (i.e. comets suffering very small planetary perturbations) is 14 per cent. A new method for cometary orbit quality assessment is also proposed by means of mod- ifying the original method, introduced by Marsden et al. (1978). This new method leads to a better diversification of orbit quality classes for contemporary comets. Key words: comets: general -- Oort Cloud. 1 INTRODUCTION The origin of comets is a problem discussed for centuries but still not fully understood. An important element of this puzzle is a question of the observed source of near-parabolic comets. There are two important observational facts that should help us to find an answer. First is the almost perfectly spherically symmetric dis- tribution of their perihelia directions, what has lead Hal Levison (1996) to call these comets nearly isotropic comets (NICs). The second is the striking distribution of their original (i.e. before en- tering the planetary zone) orbital energies, typically expressed in ⋆ E-mail: [email protected] † E-mail: [email protected] c(cid:13) 2013 RAS terms of the reciprocal of the original semimajor axis, see for ex- ample Fern´andez 2005, page 105. This distribution, highly concen- trated near zero, was first pointed out by Oort (1950) and used as an evidence, that the Solar system is surrounded by a huge, spherical cloud of comets, now called the Oort Cloud. Oort analysed the sam- ple of only 19 precise original cometary orbits but since that time hundreds of such orbits were obtained and their 1/aori strong con- centration in the interval between zero and 100 × 10−6au−1 is still evident. Since that time more and more authors call comets with the original semimajor axis in this range the Oort spike comets. This term however, is a source of a serious misunderstanding since a very popular and widely repeated opinion that "Comets in the spike come from the Oort cloud" (see for example Fern´andez 2005, page 104) seems to be far-reaching simplification of real- 2 M. Kr´olikowska and P.A. Dybczy´nski ity. As early as in 1978 Marsden et al. (hereafter MSE) formu- lated an opinion, that comets from the Oort spike "are probably making their first passage through the inner part of the solar sys- tem". They stressed (through the use of 'probably' in italic) that this is only a guess or assumption. This word was omitted in the majority of following papers and now is usually and incorrectly postulated that this is an obvious fact that comets having origi- nal barycentric semimajor axis greater than about 10 000 au (or even a few thousand au) are dynamically new. The simplest evi- dence for this to be erratic is the fact, that a significant percentage of future (when leaving planetary zone) semimajor axes of near- parabolic comets are still in the spike but potential observers dur- ing next perihelion passages cannot treat them as making their first visit among planets. The term Solar system transparency i.e. the probability that a near-parabolic comet would pass through the ob- servability zone experiencing infinitesimal planetary perturbations was first proposed and discussed by Dybczy´nski (2004, 2005) and recently also studied by Fouchard et al. 2013. They showed the dependence of this probability on a perihelion distance and esti- mated it to vary from almost zero for smallest perihelion distances, through 20 per cent at q = 5 au up to 70 per cent at q = 10 au. The study of motion of the observed large perihelion distance LPCs through the zone of strong planetary perturbations, often called the Jupiter -- Saturn barrier, was also recently carried by present authors (Dybczy´nski & Kr´olikowska 2011, hereafter Paper 2). In the ob- served sample of LPCs examined by us so far (i.e. 108 LPCs of 1/aori < 10−4 au−1) we estimated the Solar system transparency to be on the level of 14 per cent. In the current paper we will use both terms: 'near-parabolic comets' and 'long period comets', as well as the abbreviation LPCs, treating them as equivalent. There is another strong evidence that not all Oort spike comets make their first visit into the planetary zone. The significant per cent of the previous perihelia obtained from the detailed studies of their past dynamical evolution are placed well inside the planetary zone (e.g. Paper 2, and Kr´olikowska & Dybczy´nski 2010, here- after Paper 1). Formulating his hypothesis, Oort (1950) assumed that near-parabolic comets moving on Keplerian orbits outside the planetary perturbation zone are sometimes (mostly near an aphe- lion) perturbed by passing stars. Since that time, our knowledge on their dynamics has significantly increased, mainly by recognizing the importance of the Galactic perturbations in their motion (see Dones et al. 2004 for a review). Using the first order approxima- tion, one can see that the strength of the Galactic perturbation on a perihelion distance scales with a7/2 (Dones et al. 2004). Basing on 53 observed LPCs with q > 3.0 au we estimated this relation to be D q ∼ a4.06±0.16, see Paper 2 for details. While early estimations of the Galactic disc matter density r lead to the conclusion that for a comet to 'jump over' the Jupiter-Saturn barrier in one orbital period it is sufficient to have the semimajor axis a > 10000 au, the contem- porary value of r = 0.100 M⊙ pc−3 makes this limiting semimajor axis value much larger, typically 20 000 -- 28 000 au (Levison et al. 2001; Dones et al. 2004; Morbidelli 2005). Nowadays it is clear that information about the 1/aori-value is not sufficient to determine the dynamical status of so-called Oort spike comets and the previous perihelion distance must be inspected as we postulate starting from the paper by Dybczy´nski (2001) and what previously was discussed by Yabushita (1989). In his paper, Yabushita showed that only 18 of 48 Oort spike comets discovered up to 1989 are dynamically new (less than 40 per cent). However, to know correctly the dynamical status of 'Oort spike' comet we should follow the cometary orbit to the previous perihe- lion taking into account not only the Galactic disc tide as Yabushita did in his approximation but also account for the Galactic centre term (Fouchard et al. 2005) and for perturbations from passing stars (Fouchard et al. 2011). Even then we might only know whether the previous perihelion passage of an actual comet was inside the inner part of the Solar system or beyond the planetary zone. Starting from the classical paper of MSE it becomes clear that an another important factor, namely nongravitational (NG) effects, should be taken into account in the context of the determination of original inverse semimajor axes of near-parabolic comets. As it was already demonstrated by Kr´olikowska (2001, 2002, 2004) the NG accelerations should be included when determining osculating orbits of LPCs, since they can significantly change their original semimajor axes. This effect was also clearly presented in Paper 1. Therefore, the aim of this paper is twofold. First, we develop our methods of a precise osculating orbit de- termination (hereafter a nominal orbit) for the purpose of previous and next perihelion passage calculations, that will be described in the second part of this investigation (Dybczy´nski & Kr´olikowska 2013a, hereafter Part II). Here, we try to determine an NG orbit for each investigated here LPC (Section 2) and next we discuss these results with the a priori possibilities of NG determinations based on the data structure (Section 4). For a complete sample of near- parabolic comets observed in 2006 -- 2010, and having q < 3.1 au and 1/aori < 150 × 10−6 au−1, we notice the 50 per cent of suc- cess for the detection of NG effects in comet's motion using the positional data. Additional result of our study is a new method of cometary orbit quality assessment that is described in details in Section 3. Secondly, to know the dynamical status in the context of the previous perihelion passage of analysed here comets, we construct a swarm of osculating virtual comets (hereafter VCs) on the basis of the nominal orbit derived previously for each comet. In this part of investigation, we follow each swarm backward and forward to a distance of 250 au from the Sun. Thus, we obtain the original and future orbits for each comet together with the uncertainties of all orbital parameters. The method is described in Section 5. In that section, we focus on two different issues: (i) the change of origi- nal inverse semimajor axis due to incorporating the NG accelera- tion in the osculating orbit determination from the data, and (ii) the statistics of original and future 1/a-distribution of the sample of 108 near-parabolic comets studied by us here and in our previous papers. In the second aspect, we concentrate on a detailed discus- sion of original and future 1/a-distribution as well as the observed planetary perturbation distribution in the context of all so-called Oort spike comets studied by us so far (in Paper 1 & 2). Here, we use the term Oort spike because of its popularity. However, we al- ways have in mind near-parabolic comets, remembering that they consist of two populations of dynamically new and dynamically old comets. We will return to this aspect in Part II. In Part II, we will follow each swarm of barycentric orbits of VCs (taken at 250 au from the Sun) to the previous and next perihelion taking into account the Galactic perturbations and per- turbations of all known stars. Then, we will discuss the observed distribution of Oort spike together with the problem of cometary origin. This paper is in preparation. c(cid:13) 2013 RAS, MNRAS 000, 1 -- 20 Near-parabolic comets observed in 2006 -- 2010 3 Table 1. Description of observational material of 22 LPCs discovered during the period 2006 -- 2010 (columns [4] − [8]) and global characteristics of orbits determined here from the entire data intervals (columns [9] − [10]). Second and third columns show an osculating perihelion distance and perihelion time, respectively. Data distribution relative to a perihelion passage is presented in columns [7] & [8], where 'pre!' ('post!') means that all observations were taken before (after) perihelion passage; 'pre+' ('post+') means that we noticed significantly more pre-perihelion (or post-perihelion) measurements, and additional '+' indicates the drastic dominance of pre-perihelion (or post-perihelion) measurements in both the number and the time interval. Column [10] shows the type of the best model possible to determine from the full interval of data (subscript 'un' informs that for a given comet the GR and even the NG model determined from the entire interval of data is not satisfactory), and column [9] gives the resulting orbital class determined according to Q∗ given in column [12] and recipe given in Section 3; notice that the subscript 'un' in [10] means that due to an inappropriate model the orbital class in a given case is also very preliminary. Column [11] shows our division of investigated comets into four groups (a detailed description is in Section 4.) Data type Orbital Type of Comet Q∗ class model group eq 8 [yyyymmdd] [yyyymmdd -- yyyymmdd] [3] [4] Comet name qosc T Observational arc dates [au] [1] [2] C/2006 HW51 Siding Spring 2.266 2.501 C/2006 K3 McNaught 1.994 C/2006 L2 McNaught 2.431 C/2006 OF2 Broughton 0.171 C/2006 P1 McNaught C/2006 Q1 McNaught 2.764 1.015 C/2006 VZ13 LINEAR 1.212 C/2007 N3 Lulin 2.877 C/2007 O1 LINEAR 3.006 C/2007 Q1 Garradd C/2007 Q3 Siding Spring 2.252 0.850 C/2007 W1 Boattini 1.776 C/2007 W3 LINEAR 1.073 C/2008 A1 McNaught 1.262 C/2008 C1 Chen-Gao C/2008 J6 Hill 2.002 1.202 C/2008 T2 Cardinal 1.422 C/2009 K5 McNaught 2.564 C/2009 O4 Hill 0.405 C/2009 R1 McNaught C/2010 H1 Garradd 2.745 0.482 C/2010 X1 Elenin [11] C B C A B A D A C D A B B B D C C B C B D B Comet was observed to 7 September, however comet started to disintegrating in August. Thus, the data were taken to the end of July. 20060423 -- 20070807 20060522 -- 20080126 20060614 -- 20070707 20060623 -- 20100511 20060807 -- 20070711 20060820 -- 20101017 20061113 -- 20070814 20070711 -- 20110101 20060402 -- 20071113 20070821 -- 20070914 20070825 -- 20110925 20071120 -- 20081217 20071129 -- 20080908 20080110 -- 20100117 20080130 -- 20080528 20080514 -- 20081207 20081001 -- 20090909 20090527 -- 20111028 20090730 -- 20091214 20090720 -- 20100629 20100219 -- 20100702 20101210 -- 201107311 20060929 20070313 20061120 20080915 20070112 20080703 20070810 20090110 20070603 20061211 20091007 20080624 20080602 20080929 20080416 20080410 20090613 20100430 20100101 20100702 20100618 20110910 [8] full pre+ full full full full pre++ full post++ post! full full pre+ full pre++ post! pre+ full pre! pre! full pre! [au] [7] No Data Heliocentric of arc span distance span obs [5] 187 207 408 4917 341 2744 1173 3951 183 43 1368 1703 212 937 815 390 1345 2539 785 792 47 2254 2.87 -- 4.04 3.95 -- 4.13 2.74 -- 3.31 7.88 -- 6.31 2.74 -- 3.34 6.83 -- 7.91 3.84 -- 1.02 6.38 -- 7.83 4.99 -- 2.91 3.88 -- 4.02 7.64 -- 7.24 3.33 -- 2.84 2.89 -- 2.17 3.73 -- 5.82 1.71 -- 1.41 2.04 -- 3.41 3.60 -- 1.78 4.35 -- 6.25 3.04 -- 2.57 5.06 -- 0.41 2.82 -- 2.75 4.22 -- 1.04 [yr] [6] 1.3 1.7 1.1 3.9 0.9 4.2 0.7 3.2 1.6 24d 4.0 1.2 0.8 2.0 0.3 0.6 0.9 2.4 0.4 0.9 0.2 0.6 [9] 1a 1a 1a 1a+ 1b 1a+ 1b 1a+ 1a 3a 1a+ 1a 1b 1a 2a 1b 1b 1a+ 1b 1b 2b 1b [10] GR NG GR NG NG NG NGun NGun GR GR NGun NGun NG NGun GR GR GR GRun GR NG GR GRun 1 [12] 7.5 7.5 7.5 8.5 6.5 9.0 6.5 8.5 7.5 3.5 9.0 7.5 6.5 8.0 6.0 7.0 7.0 8.5 6.5 7.0 5.0 7.0 2 OBSERVATIONS AND OSCULATING ORBIT DETERMINATION We selected all near-parabolic comets discovered during the period 2006 -- 2010 that have small perihelion distances, i.e. qosc < 3.1 au, and 1/aori < 0.000150 au−1. During the same period 23 comets of qosc > 3.1 au and 1/aori < 0.000150 au−1 were detected; five of them were studied in Paper 2 and nine were still observable in November 2012. This means that data sets of these large perihelion comets were incomplete at the moment of this investigation. There- fore, in this study we restricted to complete sample of comets with small perihelion distances. All results presented in this paper are based on positional data retrieved from the IAU Minor Planet Center in August 2012, ex- cept the case of comet C/2010 H1 where we updated the obser- vational data in January 2013 because then four new pre-discovery observations were published at the Web for this comet. Global char- acteristics of the observational material are given in columns 2 -- 8 of Table 1. Most of comets in the investigated sample were ob- served on both orbital legs (compare columns 3, 4 and 8), except of five objects. Two of these comets (C/2007 Q1 and C/2008 J6) were discovered after perihelion passage and three (C/2009 O4, C/2009 R1 and C/2010 X1) were not observed after perihelion passage. The last two comets have passed their perihelia close to the Sun at a distance of 0.40 au and 0.48 au, respectively. Comet C/2010 X1 started to disintegrate about one month before perihe- lion whereas C/2009 R1 was lost after perihelion. We can suspect c(cid:13) 2013 RAS, MNRAS 000, 1 -- 20 that C/2009 R1 also did not survive perihelion passage. One can see that for two other comets, C/2006 VZ13 (qosc = 1.01 au) and C/2008 C1 (qosc = 1.26 au), the observations stopped shortly af- ter perihelion passage at the distance of 1.02 au and 1.41 au from the Sun, respectively. Thus, also in these two cases, especially for C/2006 VZ13 where the last observation was taken when comet was only about 1 au from the Earth, we can make a guess about their possible break-up. Comets passing close to the Sun in their perihelia are of spe- cial interest because we should suspect detectable influence of NG forces on their motion. It means, however, that the orbit de- termination for these LPCs is significantly more complicated than for LPCs with large perihelion distances. The determination of the NG parameters in the motion of LPCs (see Section 2.1) is much more difficult than in the motion of short-period comets mainly due to limited observational mate- rial covering one apparition or even just a half apparition, as we have for six comets mentioned above (compare also columns 3,4 and 8 of Table 1). We discussed earlier in Paper 1 that the appro- priate processing of astrometric data is very important for this pur- pose. In particular, the data weighting is crucial for the orbit fitting not only for comets discovered a long time ago but also for cur- rently observed comets. Thus, we adopted here the same, advanced data treatment as in our previous papers. The detailed procedure of weighting is described in Paper1. In this procedure, each in- dividual set of astrometric data has been processed (selected and 4 M. Kr´olikowska and P.A. Dybczy´nski Table 2. NG parameters derived in orbital solutions based on entire data intervals. First row of each object presents NG model given also in Table 3. Remaining models are here only to show alternative NG solutions (ignoring normal component of NG acceleration, column [4], and/or assuming the time shift of g(r)- function relative to perihelion passage, column [5]). Only for C/2006 K3, C/2006 OF2, C/2006 P1, C/2006 Q1, C/2007 W3 and C/2009 R1 NG models given in the first rows are used as preferred models for studying the dynamical evolution; compare with Table 3. Comet [1] C/2006 K3 C/2006 OF2 C/2006 P1 C/2006 Q1 C/2006 VZ13 C/2007 N3 C/2007 Q3 C/2007 W1 C/2007 W3 C/2008 A1 C/2009 R1 NG parameters defined by Eq. 2 in units of 10−8 au day−2 A3 A1 [2] [4] A2 [3] 15.69 ± 1.67 15.66 ± 1.67 2.384 ± 0.168 2.389 ± 0.166 0.1329 ± 0.0335 33.504 ± 0.700 31.794 ± 0.752 27.519 ± 0.840 1.874 ± 0.804 1.434 ± 0.686 4.576 ± 0.115 3.1277 ± 0.0780 0.09377±0.00962 0.08678±0.00814 0.08650±0.01117 0.156 ± 0.180 0.114 ± 0.239 0.014 ± 0.189 −0.454 ± 0.230 3.9442 ± 0.0125 4.0627 ± 0.0193 4.968 ± 0.572 4.500 ± 0.614 5.458 ± 0.473 5.5964 ± 0.0570 5.1495 ± 0.0320 5.9732 ± 0.0362 5.798 ± 0.490 2.25 ± 2.45 2.40 ± 2.41 −1.370 ± 0.131 −1.372 ± 0.131 0.03138 ±0.00397 1.916 ± 0.550 −2.710 ± 0.728 −11.592 ± 0.632 −0.866 ± 0.483 −0.547 ± 0.376 −3.041 ± 0.135 −1.1912 ± 0.9796 −0.00739±0.00611 −0.02141±0.00696 −0.01535±0.00961 2.675 ± 0.103 2.086 ± 0.136 2.589 ± 0.089 2.080 ± 0.118 −0.6133 ± 0.0175 −0.9758 ± 0.0170 2.248 ± 0.581 1.822 ± 0.643 0.957 ± 0.279 0.7136 ± 0.0384 0.9915 ± 0.0201 0.3252 ± 0.0216 −1.418 ± 0.359 −0.209 ± 0.576 0.0 −0.0059 ± 0.0347 0.0 0.0 10.604 ± 0.189 10.199 ± 0.189 0.0 0.528 ± 0.404 0.0 1.220 ± 0.074 0.0 −0.12700±0.00145 −0.13334±0.00190 0.0 1.657 ± 0.037 0.0 1.592 ± 0.037 0.0 −0.06023±0.00361 −0.05387±0.00433 −1.084 ± 0.316 −0.655 ± 0.503 0.0 0.16800 ±0.00815 0.19393 ±0.00763 0.0 0.776 ± 0.207 t [ days] rms [arcsec] [5] -- -- -- -- -- -- 37.6 ±4.2 50 -- -- -- -- -- 11.3 ±1.9 11.3 -- -- −25.2±5.1 −25 22.32 ±4.6 -- -- −22 ± 21 −22 5.76 ±0.6 -- 10 -- [6] 0.54 0.55 0.36 0.36 0.25 0.37 0.37 0.461 0.392 0.392 0.51 0.54 0.35 0.35 0.491 0.39 0.48 0.39 0.481 0.67 0.96 0.52 0.52 0.521 0.44 0.45 0.491 0.51 1 Best fitting asymmetric NG models (in the sense of minimal value of rms) with assumed A3=0.0), 2 NG models based on pre-perihelion data only (see also Table 3). 1/aori [10−6 au−1] [7] 61.02 ±4.63 61.74 ±4.25 21.21 ±0.49 21.21 ±0.49 57.17 ±4.03 51.08 ±0.48 49.69 ±0.47 44.28 13.96 ±4.80 15.86 ±4.46 −18.39 ±3.87 −23.09 ±4.12 32.77 ±0.18 32.39 ±0.17 32.59 39.13 ±0.49 36.61 ±0.64 40.78 ±0.56 36.90 −36.56 ±1.86 −82.30 ±2.61 31.38 ±3.85 30.70 ±6.72 25.71 123.07 ±1.50 120.14 ±1.56 113.73 12.16 ±3.29 weighted) during the determination of a pure gravitational orbit (GR) or NG orbit, independently. 2.1 The non-gravitational acceleration in the comet's motion To determine the NG cometary orbit we used the standard formal- ism proposed by Marsden et al. (1973, hereafter MSY) where the three orbital components of the NG acceleration acting on a comet are scaled with a function g(r) symmetric relative to perihelion: Fi = Ai· , Ai = const for g(r) g(r) = a (r/r0)−2.15h1 + (r/r0)5.093i−4.614 i = 1,2,3, (1) (2) , where F1, F2, F3 represent the radial, transverse and normal com- ponents of the NG acceleration, respectively and the radial acceler- ation is defined outward along the Sun-comet line. The normal- ization constant a = 0.1113 gives g(1 AU) = 1; the scale dis- tance r0 = 2.808 AU. From orbital calculations, the NG param- eters A1, A2, and A3 were derived together with six orbital ele- ments within a given time interval (numerical details are described in Kr´olikowska 2006). The standard NG model assumes that wa- ter sublimates from the whole surface of an isothermal cometary nucleus. The asymmetric model of NG acceleration is derived by using g(r(t − t )) instead of g(r(t)). Thus, this model introduces an additional NG parameter t -- the time displacement of the maxi- mum of the g(r) relative to the moment of perihelion passage. Typically, the radial component, A1, derived in a symmetric model is positive (reflecting, in average, the stronger sublimation of this part of cometary surface that is directed to the Sun) and dominate in magnitude over the transverse and normal components. This model, however, does not include the possibility of location of an active region(s) on cometary surface. The negative radial com- ponent, A1 < 0, derived in this model, would give a first indica- tion for the asymmetric model of g(r) function (with rather large time displacement of a maximum of the g(r) relative to perihe- lion) or for the existence of active region(s) on comet's surface. Described model is very successful in representing the astrometric data, thus also in allowing the realistic dynamical evolution predic- tions. However, this NG force model does not represent an accurate representation of the actual processes taking place in the cometary nucleus (e.g. see Yeomans 1994). Thus, using this standard model of NG acceleration we can only go as far as a very general and a very qualitative discussion in this field. Therefore, in this paper we place a strong emphasis on orbital dynamics of near-parabolic comets examined here, accounting only for evident physical events, like disruption or fragmentation. For example, in all such cases reg- istered, we notice the coincidence between bursts (or disruptions) and anomalies occurring in O-C-diagram. Therefore, we decided to exclude such data intervals in the process of osculating orbit deter- mination. c(cid:13) 2013 RAS, MNRAS 000, 1 -- 20 In the present investigation we decided to use an NG force model with the smallest number of parameters needed. We tested asymmetric model for several comets from the sample studied here but no improvement was observed. Moreover, in two cases, C/2006 HW51 and C/2006 P1, just two NG parameters were deter- minable with reasonable accuracy. For comets with long time sequences of astrometric data (e.g. belonging to comet group A -- see column 11 of Table 1) we also tested a more general form of the dependence of NG acceleration on a heliocentric distance: Fi = A∗ i · h(r), A∗ i = const for i = 1,2,3. (3) where we adopted two different form for a dependence of accelera- tion on a heliocentric distance, h(r), namely: more general g(r)- like function, g∗(r) (hereafter GEN model type), and Yabushita function, f (r), based on the carbon monoxide sublimation rate (Yabushita 1996, hereafter YAB model type). Thus, we consider here also the following two types of NG models: • GEN: h(r) = g∗(r) = a (r/r0)−m [1 + (r/r0)n]−k • YAB: h(r) = f (r) = 1.0006 r2 × 10−0.07395(r−1) ·(cid:16)1 + 0.0006r5(cid:17)−1 (4) (5) In a GEN type of NG model we have additional four free param- eters: scale distance, r0, and exponents n, m, k. The function g∗(r) is normalized similarly as standard g(r) function, thus a is calcu- lated from the condition: g∗(1au) = 1, also the f (r) function was normalized to unity at r = 1 au. In contrast to comets C/2002 T2 and C/2001 Q4 investigated by Kr´olikowska et al. (2012, hereafter Paper 3), we found that GEN and YAB types of NG model did not improve the orbital data fitting for investigated here comets with more than 3 yr interval of data covering the wide range of heliocentric distances (C/2006 OF2, C/2006 Q1, C/2007 N3 and C/2007 Q3). We were able to determine the NG effects for 11 of 22 comets discovered in the period 2006 -- 2010 (see next section). As far as we know this is the richest sample of LPCs with NG effects of cur- rently (in February 2013) published in periodicals and on the Web Pages. Many sources of osculating orbits of comets are available only in the Web. From these sources we noticed that Nakano (2013) and Rocher (2013) published NG orbits for largest per cents of comets in comparison to other Web sources. In February 2013 both sources presented NG osculating orbits for six comets from the pe- riod of 2006 -- 2010, whereas we determined NG orbits for eleven comets discovered in this period. Additionally, only at Nakano page (2013) and at IAU Minor Planet Center (2013) values of original and future 1/a are given; in the second Web source for three comets with NG orbits: C/2007 W1, C/2008 A1 and C/2008 T2. More de- tails about NG models derived in the present studies are given in the next two sections. In Section 5 we discuss the change of the 1/aori due to incorporation of the NG acceleration in the process of osculating orbit determination using the positional data. Near-parabolic comets observed in 2006 -- 2010 5 Figure 1. The O-C diagrams for comet C/2009 R1 McNaught (weighted data). The upper figure shows O-C based on an NG solution whereas the lower figure presents O-C based on a pure GR orbit. Residuals in right as- cension are shown as magenta dots and in declination -- as blue open circles; the moment of perihelion passage is shown by dashed vertical line. Figure 2. The O-C diagrams for comet C/2008 A1 McNaught (weighted data). Two lower panels show O-C based on the entire data interval for an NG orbit and a pure GR orbit, respectively. Two upper panels present the O-C diagrams for orbits (NG or GR) determined individually for pre- perihelion and post-perihelion orbital branches. Residuals in right ascension are shown as magenta or plum dots and in declination -- as blue or light blue open circles; the moment of perihelion passage is shown by dashed vertical line. 2.2 Osculating orbit determination from the full data interval We found that NG accelerations are well-detectable in the motion of eleven comets during their periods covered by positional obser- vations. These are comets C/2006 K3, C/2006 OF2, C/2006 P1, C/2006 Q1, C/2006 VZ13, C/2007 N3, C/2007 Q3, C/2007 W1, C/2007 W3, C/2008 A1, C/2009 R1. These models are shown as coloured light grey rows in Ta- ble 1, whereas the values of original and future 1/a for these mod- els are given in Table 3, except comet C/2006 VZ13 where only c(cid:13) 2013 RAS, MNRAS 000, 1 -- 20 6 M. Kr´olikowska and P.A. Dybczy´nski the model based on pre-perihelion data are shown for the reason discussed in Section 4.4. The NG parameters of these symmetric NG models are given in Table 2 in the first row of each individual comet. Additionally, in this table we presented some alternate models that we found as less certain in the sense of orbital fitting to data (three creteria are given just below). Asymmetric NG solution (with t -parameter, see previous section) was possible to determine for six comets: C/2006 Q1, C/2007 N3, C/2007 Q3 C/2007 W1, C/2007 W3 and C/2008 A1 (see Table 2). However, only in the case of C/2007 W1 we no- ticed substantial improvement of orbital fitting to data and in the case of C/2008 A1 -- the infinitesimal adjustment (in the sense of at least one of three criteria given in section 2.2). For that reason both asymmetric model are also given in Table 3 as the best NG solution derived from the entire data set. It was surprising that in some cases a normal component od NG acceleration improves the orbit's fitting to data to the much higher degree than the t , see the NG solutions for C/2006 Q1, C/2007 N3 and C/2007 Q3. In another words, neglecting the nor- mal component A3 and determining the A1, A2 and t we get the orbital solution with significantly worse data fitting. One can noticed that almost all these alternate models give original 1/a in a very good agreement with original 1/a-values based on NG models chosen by us to dynamical evolution investigation. The only exception is C/2007 W1 where symmetric model give more hiperbolic orbit that the asymmetric model with t , however both original 1/a-values are certainly negative. However, this comet exhibits erratic behaviour and its more appropriate solutions (given in the Tables 3 and 6) are discussed later. Comet C/2006 VZ13 is quite a different case because of the almost exclusively pre- perihelion data and some possibility of disintegrating processes close to perihelion. By 'well-detectable' NG effects we mean that assuming stan- dard NG model of motion (see Section 2.1) we noticed (for each of these eleven comets) the better orbit fitting to data in comparison to a pure GR orbit measured by three criteria: • decrease in rms, • overcoming or reducing the improper trends in O-C time vari- ations, • increasing the similarity of the O-C distribution to a normal distribution. More details and examples how this analysis works are given in Papers1 -- 3, therefore only two examples are given below. Figs 1 -- 2 show the comparison between the O-C diagrams of NG orbit and GR orbit for C/2009 R1 with moderately manifest- ing NG effects in the motion, and C/2008 A1 with spectacularly visible NG effects in the motion, respectively. We additionally no- ticed the decrease in rms from 0.′′63 (GR orbit) to 0.′′51 (NG orbit) for C/2009 R1 and from 1.′′44 (GR orbit) to 0.′′44 (NG orbit) for C/2008 A1. One can see in Fig. 1 that trends easily visible in the O-C diagram based on GR orbit disappear for NG orbit. Moreover at the beginning of the data set, there are four observations taken on 2009 07 20 and four in 2009 08 01. According to our selec- tion procedure all these measurements in right ascension are not used for GR orbit determination due to unacceptable large residu- als whereas in the case of NG orbit all are well-fitted as one can see in the upper panel of Fig. 1. Such a data recovery, in particu- lar at the edges of observational period (what we noticed in many cases of the NG orbit determination) is an additional argument for a prevalence of NG orbit (and NG model). Comet C/2008 A1 is a very special case because we still can see in the O-C diagram well-visible trends in residuals in right ascension and declination for NG orbit determined from a whole data set (third panel from the top in Fig 2). Also, the distributions of residuals in a and d are not fitted well to normal distributions. Thus, for this comet it was necessary to divide the data into two parts: data before and after perihelion passage. It was very surprising that for both orbital legs the NG orbits were perfectly determinable (both NG solutions are discussed later in this paper). The O-C diagrams for pre-perihelion orbits and post-perihelion orbits are presented in the two upper panels of Fig 2 for NG model and pure GR model, respectively. One can see the great improvements of NG orbit fitting for pre- perihelion data and only slight improvements for post-perihelion data. One can even get an impression that the NG orbit determined from the entire data set (third panel in Fig. 2) gives a better fitting to post-perihelion data that orbit determined from the post-perihelion subset of data. Unfortunately, some trends, in particular in declina- tion, are still noticeable. On the other hand, however, for NG orbit based on post-perihelion data we also noticed the recovery of some measurements taken at the end of observational arc. Thus, we con- clude that NG orbit based on the post-perihelion data arc seems to be more adequate to predict the future of this comet as well as the uncertainty of this prediction. The most manifesting NG accelerations we found in C/2007 W1 case where a similar approach was necessary to be used as for C/2008 A1. The same method was applied earlier for C/2007 W1 by Nakano (2009a; 2009b). We found that the orbit obtained on the basis of the whole data set proved to be inade- quate to describe the actual motion of both comets (C/2007 W1 and C/2008 A1). We noticed this fact by subscript 'un' (i.e. un- certain) in column [10] of Table 1. Among comets with very long data intervals (> 3 yr) there are two other objects (C/2007 N3 and C/2007 Q3) with well visible trends in O-C diagrams taken for NG orbit and based on the full observational interval. Therefore, for these two comets it was also necessary to determine the osculating orbit from pre- and post-perihelion subsets of data, independently. Further, in the comet C/2006 VZ13 case, an adequate NG orbit for past dynamical evolution were determined only for some part of pre-perihelion data (ending 40 days before perihelion passage). Analysing orbits of eleven comets with well-detectable NG ef- fects in their orbital motion we concluded that the motion of six of them can be model by one set of NG parameters during the period covered by data. In the case of remaining five comets the NG be- haviour is more complicated. Thus, we proposed in these cases to model their motion separately before and after perihelion passage. It turn out that NG effects are easily visible in the time inter- val covered by positional observations also in the motion of comet C/2010 X1. However, this is very special case because the nature of this NG behaviour is likely to be violent -- this comet started to disintegrate at least about one months before perihelion passage (in August 2011). Therefore, the NG model used here is completely inadequate (see also Section 4.2). Thus, we modelled the motion of this comet based on the shorter interval of data, e.g. carried out be- fore the disintegration process was started. It turn out, that the best orbit (GR orbit), for the past evolution was determined from an arc of data limited to the observations taken before 1 May 2011 what can suggest that a disruption started even before August 2011. For this reason we also marked the GR solution based on almost entire data interval by subscript 'un' in Table 1. For other three comets, C/2006 HW51, C/2006 L2 and C/2008 T2, the NG effects are worse detectable than in previous cases. Additionally, the radial component of standard NG acceler- ation, A1 (see Eq. 2), is negative for these objects (see also Sec- c(cid:13) 2013 RAS, MNRAS 000, 1 -- 20 Near-parabolic comets observed in 2006 -- 2010 7 Table 3. Characteristics of models taken in this investigation for previous and next perihelion dynamical evolution. Columns [3] and [4] provide a qualitative assessment of the O-C distribution and O-C diagram, respectively for the types of models given in column [2], where symbol ++ means distributions well- fitted to a normal distribution or O-C diagrams without any meaningful trends in right ascension or/and declination, + denotes distributions rather similar to a Gaussian or O-C diagram with slight trends only, - means non-Gaussian distribution and significant trends in O-C diagram, - - denotes the worse characteristics of O-C distribution or/and O-C diagram. NG models with negative radial component of NG acceleration, A1 < 0, are marked as NGA1. The rms and a qualitative assessment of O-C distribution are additionally given in columns [6] -- [7] for GR model based on the same interval of data as NG model given in column [2]. A quality class obtained in each individual case as well as the 1/aori and 1/afut are given in columns [8] -- [10], respectively. In the case of PRE-type of model described in columns [2] -- [9], the type of model used for future 1/a-determination is listed in column [11]. Comet [1] 2006 HW51 2006 K3 2006 L2 2006 OF2 2006 P1 2006 Q1 2006 VZ13 2007 N3 2007 O1 2007 Q1 2007 Q3 2007 W1 2007 W3 2008 A1 2008 C1 2008 J6 2008 T2 2009 K5 2009 O4 2009 R1 2010 H1 2010 X1 Model type [2] GR NGA1 PRE GR NG GR NGA1 PRE GR NG PRE GR NG NG PRE GR PRE NG PRE GR NG GR GR PRE GR NG PRE NG NG+t 2 NG PRE NG NG+t 2 NG GR NG GR GR NGA1 PRE GR PRE GR GR DIST NGA1 GR NG GR GRMay GRApr Fit to gauss [3] + ++ + ++ ++ + ++ ++ ++ + + + ++ - ++ ++ + + + + - ++ ++ + + ++ ++ ++ ++ ++ ++ - - + + ++ + + ++ O-C [4] + ++ ++ ++ + ++ ++ + + + ++ ++ + + + + + + - + - + ++ - - ++ ++ + + ++ ++ + - ++ ++ ++ + + ++ rms ′′ [5] 0.29 0.28 0.29 0.54 0.49 0.46 0.37 0.36 0.36 0.25 0.37 0.34 0.39 0.33 0.35 0.47 0.58 0.39 0.39 0.49 0.67 0.52 0.28 0.44 0.45 0.36 0.36 0.47 0.38 0.39 0.39 0.33 0.47 0.39 0.39 0.51 0.83 0.46 0.37 rmsGR ′′ [6] GR fit to gauss [7] 0.29 0.69 0.49 0.38 0.25 0.50 0.40 0.50 0.49 0.61 2.96 0.54 0.47 1.44 1.44 0.36 0.38 0.38 0.63 + + ++ - + - + - - - - - - ++ - - - - ++ ++ - - - - + Orbit class [8] 1a 1a 2a 1a 1a 1a 1b 1a+ 1a 1b 1a+ 1a 2a 1a 1a+ 1a 3a 1a+ 1a+ 1b 1a 1b 1b 1a 1a 2a 2a 1b 1b 1b 1b 1a 1a+ 1a 1b 1b 2b 1b 1b 1/aori 10−6 au−1 [9] 1/afut 10−6 au−1 [10] Model type for 1/afut if different from [2] [11] 90.12 ± 3.37 37.45 ± 8.34 −131.28±± 4.67 −95.28 ± 1.43 −170.88± 7.73 −658.82±± 0.23 467.65 ±± 3.56 707.44 ±± 0.31 47.31 ± 3.37 30.32 ± 4.40 24.84 ± 6.19 61.02 ±± 4.63 12.89 ± 1.43 43.56 ± 4.21 63.56 ± 4.52 21.21 ±± 0.49 16.42 ± 0.62 57.17 ±± 4.03 51.08 ±± 0.48 49.44 ± 0.45 491.21 ± 20.10 13.96 ± 4.80 823.61 ± 2.06 29.31 ± 0.59 828.64 ±± 0.59 32.77 ±± 0.18 23.36 ± 4.70 −496.83± 4.69 54.95 ±799.09 −449.88±741.43 41.91 ± 0.53 131.77 ± 3.63 118.96 ±± 0.96 39.13 ±± 0.49 554.38 ± 7.09 −42.71 ± 2.34 549.97 ±± 5.79 −36.56 ±± 1.86 343.89 ±± 18.10 31.38 ±± 3.85 120.84 ± 2.03 246.52 ± 2.82 256.41 ±± 2.24 123.07 ±± 1.50 247.21 ± 1.89 120.14 ± 1.56 502.56 ± 11.77 38.57 ± 11.77 115.95 ± 59.16 648.87 ±221.41 −479.69± 3.99 25.35 ± 4.00 275.92 ± 1.06 12.22 ± 1.06 19.19 ± 1.14 218.90 ± 7.83 11.47 ± 1.08 45.50 ± 0.55 49.42 ± 0.22 44.22 ± 2.75 55.96 ± 4.91 12.16 ±± 3.29 240.15 ± 75.41 24.10 ± 2.20 27.24 ± 3.95 552.91 ± 0.41 554.68 ± 0.22 550.33 ± 1.60 −55.96 ± 4.91 170.43 ±±197.25 784.70 ± 75.46 disintegration NG, class: 1b POST GR 1, class: 1a POST GR, class: 1a POST NG, class: 2a POST NG, class: 1b POST GR, class: 1a 1 Model based on data started from heliocentric distance of 2.0 au from the Sun 2 Asymmetric models relative to perihelion passage based on entire data sets tion 2.1). However, the GR orbit is quite acceptable since only slight trends are visible in the O-C diagrams of these comets. Therefore, we discussed for them the GR orbit (Table 1) as at least similarly likely solution as NG orbit (Table 3) and included to the group of comets with weak NG effects (Section 4.3). Comet C/2009 K5 also have marginally detectable NG effects with a negative A1 but it differs from the previous three comets in visible trends in the O-C diagrams for GR orbit as well NG or- bit. Thus, we included this comet to special objects (marked by subscript 'un' in column [10] of Table 1) where it was necessary to divide the data into pre-perihelion and post-perihelion orbital branches (see Section 4.2). c(cid:13) 2013 RAS, MNRAS 000, 1 -- 20 An interesting example is comet C/2008 C1 that has marginally detectable NG effects. This is surprising because of short interval of data (0.3 yr). We decided to use the GR solution as more reliable for this comet of second quality orbit (Table 1). NG effects are completely not detectable in the orbital motion of five comets, C/2007 O1, C/2007 Q1, C/2008 J6, C/2009 O4, C/2010 H1, and for these objects we have just pure GR osculat- ing orbits (Table 1). All these comets have osculating perihelion distance qosc > 2 au and two of them have poor quality orbits (see Table 1). 8 M. Kr´olikowska and P.A. Dybczy´nski Table 4. Quantities for establishing accuracy of orbit. Slightly modified new version of Table II from MSE L & M Mean error of 1/aosc in units of 10−6 au−1 < 1 [1 , 5[ [5 , 20[ [20 , 100[ [100 , 500[ [500 , 2500[ [2 500 , 12 500[ > 12 500 8 7 6 5 4 3 2 1 0 -1 -2 time span of observations in months or days > 48 months [24 , 48[ [12 , 24[ [ 6 , 12[ [ 3 , 6[ [1.5 , 3[ [23 days , 1.5 months[ [12 , 23[ days [7 , 12[ [3 , 7[ [1 , 3[ 2.3 Osculating orbit determination from some part of data According to the brief overview of orbital determination given in Section 2.2 we have seven special comets, C/2006 VZ13, C/2007 N3, C/2007 Q3, C/2007 W1, C/2008 A1, C/2009 K5 and C/2010 X1, where we have to apply a non-standard approach and determine two individual osculating orbits (GR or NG if possi- ble) for two orbital legs. These solutions were marked as PRE and POST models in column [2] of Table 3. For C/2009 K5 we addi- tionally presented DIST model based on data taken from the helio- centric distances larger than rh > 2.5 au. For the remaining comets we noticed that osculating orbits (GR or NG, see Table 1) based on the entire data interval well rep- resent their orbital motion in the period covered by observations. However, for some of these comets (wherever we though it relevant and meaningful) we also determined the osculating orbits based on only a subset of pre-perihelion measurements (i.e. for C/2006 OF2, C/2006 Q1 and C/2008 C1). These additional orbital solutions are presented here just for comparison and wider discussion of the dy- namical evolution of each comet carried out in Part II. Moreover, as was discussed in Section 2.2, for comets C/2006 HW51, C/2006 L2 and C/2008 T2 the NG models based on entire set of data (where radial component of NG acceleration is negative) are also given just for comparison. Generally, in Table 3 are presented all models (of osculat- ing orbit) taken further in the second part of this investigation for previous and next perihelion dynamical evolution. The best model derived for each comet is given always in the first row and just these best models are used in Section 5 for statistical analysis. The NG models based on the entire data sets are shown in coloured light grey rows, similarly as in Table 1. 3 ACCURACY OF THE COMETARY ORBIT In 1978 MSE formulated the recipe to evaluate the accuracy of the osculating cometary orbits obtained from the positional data. They proposed to measure this accuracy by the quantity Q defined as Q = 1/2 · (L + M + N) + d , where L denotes a small integer number which depends on the mean error of the determination of the osculating 1/a, M -- a small integer number which depends on the time interval covered by the observations, N -- a small integer number that reflects the number of planets, whose perturbation were taken into account, and d equals 0.5 or 1 to make Q an integer number. (6) Values of L, M and N are obtained following the scheme presented in their original Table II (MSE). According to MSE the integer Q- value calculated from Eq. 6 should be next replaced with the orbit quality class as follows: value of Q = 9,8 means orbit of 1A orbital class, Q = 7 -- 1B class, Q = 6 -- 2A class, Q = 5 -- 2B class, and Q < 5 means the worse than second class orbit. There are three reasons for which we found that some slight, but important adjustment of the above orbital accuracy assessment should be done: (i) In the modern orbit determination all Solar system planets are taken into account, therefore we always have N = 3. Thus, Eq. 6 can be rewritten for our purpose in a form: Q = Q∗ + d , where Q∗ = 0.5 · (L + M + 3) and d = 1 or 0.5 (7) (ii) Current cometary observations are generally of a high preci- sion and the resulting osculating 1/a-uncertainties are often smaller than 10−6 au−1. It should be reflected in a higher L number than is allowed in the original MSE table. Similarly, a higher M number should be attributed to very long time intervals covered by con- temporary positional observations of LPCs. Thus, the mean error of 1/aosc smaller than 1 unit (i.e. 1 × 10−6 au−1) gives now L = 7 and a time span of data can be greater than 48 months (M = 8, see Table 4). (iii) Almost all orbits of currently discovered LPCs would be classified as the 1A quality class. For example, we found that using the original d definition (to be 0.5 or 1) we obtained 15 comets of orbital class 1A and just two comets (C/2007 Q1 and C/2010 H1) of an obvious cases of second or worse orbital class (extremely short orbital arcs) for the sample studied here. Thus, additional modifi- cation seems to be necessary to obtain a better diversification of orbit accuracy classes. We proposed below, the new d definition and division the first quality class into three new classes (1a+, 1a, 1b) instead former 1A and 1B. Thus, the more relevant definition of Q for currently discov- ered LPCs takes the form: Q = Q∗ + d , where Q∗ = 0.5 · (L + M + 3) and d = 0 or 0.5 (8) How to calculate the quantities L and M is described in Table 4 that is a simpler form of original Table II given by MSE. Value of d equals 0.5 or 0 to make Q an integer number. When Q∗ is an intermediate between two integers (that define two consecutive orbital classes) the proposed new recipe gives the final Q-values exactly the same as originally proposed by MSE. To distinguish the proposed quality system from MSE system, we use lower-case letter 'a' and 'b' in quality class descriptions in- stead of original 'A' and 'B' in the following way: Q = 9 -- 1a+ class, Q = 8 -- 1a class, Q = 7 -- 1b class, Q = 6 -- 2a class, Q = 5 -- 2b class, Q = 4 -- 3a class, Q = 3 -- 3b class, and Q 6 2 -- 4 class, where Q is calculated according to eq. 8. We additionally propose to introduce a special 1a+ quality class in case of Q = 9. The quality classes 3a, 3b and 4 were not defined by MSE, but we adopted here the idea published by Minor Planet Center (at http://www.minorplanetcenter.net/iau/info/UValue.html) as 'a log- ical extension to the MSE scheme'. This new orbit quality scheme separates the orbits of a very good quality in MSE system, 1A, into three quality classes in the new system, where the worst of orbits in 1A class (Q∗=7) in MSE are classified as 1b in the new scheme. The reader can check how c(cid:13) 2013 RAS, MNRAS 000, 1 -- 20 it works using the Q∗ value given in last column of Table 1 to cal- culate the MSE quality class. In the MWC 08 only pure GR orbits are classified according to MSE method. It seems, however, that there are no contraindi- cations to use such a procedure to qualify also NG orbits. These NG orbits are often of a lower quality than GR orbit obtained on the basis of the same time interval. This is an obvious consequence of higher uncertainties of NG orbital elements as a result of addi- tional NG parameters to determine simultaneously with six orbital elements. In our opinion, the quality comparison between GR orbit and NG orbit should not be carried out exclusively on the basis of an orbital quality because NG orbit is always closer to reality by its physical assumptions, and therefore is more appropriate to describe the cometary motion than GR orbit. On the other hand, a quality as- sessment for NG orbit gives an easy indication of the uncertainty of orbital elements. It is worth to stress, that any effective orbit quality assess- ment system (including the MSE method and this proposed here) describes a particular orbital solution based on a given data set (or subset), procedure used for this data processing as well as the adopted force model. Therefore, orbit of each individual comet can be classified differently, depending on the preferred orbital solu- tion. 3.1 Application of the modified method of orbital quality assessment to comets considered so far The new modified MSE procedure described above, was applied for the determination of orbital classes of all osculating orbits of the investigated here comets. The results are given in columns 12 and 9 of Table 1 and in column 8 of Table 3. As was mentioned above, one can notice the higher diversification in orbital classes between investigated comets than using the original MSE recipe. This can easily be checked using Eq. 7 and values given in col- umn 12 of Table 1. According to the proposed method we have a lower fraction of comets of the best orbital classes: 11 (1a+ and 1a classes) instead of 15 (1A class). Additionally, these 11 comets are now divided into 5 comets of 1a+ class and 6 comets of 1a class when using entire data sets shown in Table 1. Moreover, the orbit of one comet, C/2008 C1, is reclassified as a second class orbit. We noticed that at the IAU Minor Planet Center web pages or- bital quality codes are given also for newly discovered comets. In February 2013, for five comets investigated in this paper, the qual- ity codes are not shown, namely for C/2007 Q1 (extremely short arc of data) C/2007 W1 (NG orbit), C/2008 A1 (NG orbit), C/2008 T2 (NG orbit with negative radial component of NG acceleration) and C/2010 X1. The remaining 17 objects are classified as follows: 13 comets have 1A class, 2 comets -- class 1B (C/2006 VZ13, C/2010 H1), and 2 objects -- of class 2A (C/2008 C1, C/2009 O4). We also estimated the orbit of C/2008 C1 as a 2a quality class, but orbit of C/2009 O4 as 1b class. However, 2A quality code for C/2009 O4 is there a natural consequence of shorter intervals of data taken for orbit determination than in the present investigation. Similarly, the JPL Small-Body Database Browser (2013) shows (in February 2013) the osculating orbits on the basis of full interval of positional data. We also applied the modified method of orbital quality deter- mination to pure GR as well as NG orbits of comets from Papers 1 & 2, where all those Oort spike comets were chosen from MWC 08 as objects with highest quality orbits (classes 1A or 1B). The full list of these new quality class assessment for the entire sample of near-parabolic comets is given in Table 5. An extended table, for c(cid:13) 2013 RAS, MNRAS 000, 1 -- 20 Near-parabolic comets observed in 2006 -- 2010 9 Table 5. New quality class assessment for 86 near-parabolic comets inves- tigated in Paper1&2 C/ Class C/ Class C/ Class C/ Class 37 comets of NG orbits 1885 X1 1946 U1 1974 F1 1989 X1 1993 Q1 2001 Q4 2004 B1 1980 E1 1999 H3 2005 B1 2a 1b 1a 1b 1b 1a+ 1a+ 1a+ 1a 1a+ 1892 Q1 1952 W1 1978 H1 1990 K1 1996 E1 2002 E2 qosc < 3.1 au 2a 2a 1b 1a 1b 1b 1913 Y1 1956 R1 1986 P1 1991 F2 1997 J2 2002 T7 1b 1b 1a 1b 1a+ 1a+ 1940 R2 1959 Y1 1989 Q1 1993 A1 1999 Y1 2003 K4 2a 2b 2a 1a 1a+ 1a+ qosc > 3.1 au 1983 O1 1a 2000 CT54 1a+ 2005 EL173 1a+ 1984 W2 1a 2000 SV74 1a+ 2005 K1 1a 1997 BA6 1a+ 1a 2002 R3 2006 S2 1b 49 comets of GR orbits 1992 J1 1a+ 2001 K3 1a 1972 L1 1b 1976 U1 1a 1987 F1 1a 1993 F1 1a 1999 F2 1a 1999 S2 1a 2000 K1 1a 2001 G1 2002 J5 1a+ 2003 WT42 1a+ 1a 2005 G1 2006 YC 2a 1973 W1 1978 A1 1987 H1 1993 K1 1999 J2 1999 U1 2000 O1 2001 K5 2002 L9 2004 P1 2005 Q1 2007 JA21 qosc < 3.1 au 2a qosc > 3.1 au 1b 1a 1a+ 1a 1a+ 1a 1a 1a+ 1a+ 1a 1a 1a 1974 V1 1978 G2 1987 W3 1997 A1 1999 K5 1999 U4 2000 Y1 2002 A3 2003 G1 2004 T3 2006 E1 2007 Y1 1976 D2 1a 1979 M3 1b 1a 1988 B1 1a+ 1999 F1 1999 N4 1a 1a 2000 A1 1a 2001 C1 1a 2002 J4 1a 2003 S3 2004 X3 1a 1a+ 2006 K1 1b 1b 1a 1b 1a+ 1a+ 1a 1a 1a 1a 1a 2a 108 LPCs investigated by us that includes details about the obser- vational material used for orbit determination, type of models used, as well as the orbit quality estimation, is available in our web page (Dybczy´nski & Kr´olikowska 2013b). It turned out that among 37 comets with NG orbits, six (C/1885 X1, C/1892 Q1, C/1940 R2, C/1952 W1, C/1959 Y1 and C/1989 Q1) should be classified as second quality orbit ac- cording to a modified, more restrictive method proposed here (Ta- ble 5). Next three with pure GR orbit are now also class 2 objects (C/2001 K3, C/2006 YC and C/2007 Y1). Summarizing, according to a more restrictive orbital quality assessment proposed here we obtained 23 comets of 1a+ class, 38 comets of 1a class, 16 -- 1b class, 8 -- 2a class, and 1 object of 2b class in the sample of 86 comets analysed in Papers 1 & 2 that were chosen from MWC 08 as Oort spike comets having pure GR orbit of the highest quality (class 1) or NG orbit (then the quality class is not specified in the catalogue). 4 IMPLICATION OF DATA STRUCTURE ON ORBIT DETERMINATION In this section we show how the data structure influence the proper choice of the tailored orbit determination method, how it affects the quality of osculating orbits and the possibility of the determination of NG effects in the motion of near-parabolic comet. We performed an extensive review and tests of various models for each of inves- 10 M. Kr´olikowska and P.A. Dybczy´nski tigated comets, including asymmetric ones (i.e. NG models with t -shift of maximum of g(r)-function relative to perihelion passage, see section 2.1) and orbital modelling based only on one branch of their orbits (pre- or post-perihelion, see section 2.3). From the en- tire range of solutions obtained for these comets we included here only the best models, in the sense of three criteria mentioned in Section 2.2: the decrease of rms, the similarity of O-C distribution to the Gaussian distribution and the absence of trends in O-C dia- grams, and keeping a minimum number of necessary NG parame- ters in the case of NG models (Table 3). We found useful, for this aspects of further discussion, to divide the investigated sample of comets into four groups (see also column 11 of Table 1): group A: four comets with more than 3 yrs of data and heliocen- tric distance span at least from 6 au before perihelion to 6 au after perihelion, group B: eight peculiar comets, e.g. extremely bright or and/or with detected split event and/or strong/variable NG effects in their motion, group C: six comets of non-detectable or very weak NG effects in their motions, group D: four comets of a second or worse quality of osculating orbit. Below we discuss the differences in data sets between these groups and whether and how these differences affect the precise determination of NG orbit. The reference table to individual solu- tions discussed in this section is implicitly Table 3. 4.1 Group A: comets with long time-intervals of observations One can see in Fig. 3 that all four comets of this group have been observed more than three years in a broad range of heliocentric dis- tances from at least 6 au before the perihelion passage to over 6 au after perihelion. Thus, long time sequences of data should allow us to model the NG orbital motion in great details. In particular, these should allow for examining various forms of g(r)-like function (Pa- per 3). In fact, long time series of data allow us to determine the NG effects of all these comets basing on the entire ranges of data, and all their NG orbits are of the highest accuracy (1a+ class, see Table 1). In all cases where we were able to apply an asymmet- ric NG model relative to the moment of perihelion passage (Sec- tion 2.1), such a model however did not show any decrease in rms in comparison to a symmetric NG model (Table 2) and did not give a better similarity of O-C distribution to a normal distribu- tion and/or better O-C diagram. Moreover, it was not possible to derive any dedicated form of g(r)-like functions for these comets. We were also not able to formulate any concrete conclusions about the potential deviation of value of the exponent m in Eq. 4 from the standard value mSTD = −2.15, or about the different value of scale distance than the standard r0 = 2.808 au (Eq. 2). C/2006 OF2 Broughton NG models of this comet provide O-C distributions very good ap- proximated by Gaussian distribution and give very reasonable val- ues of NG parameters with dominant and positive radial component of NG acceleration and negligible normal component in compari- son to the remaining two A1 and A2 components (Table 2). How- ever, even assuming A3 = 0, the tau-shift can not be determined (its value oscillates with large amplitude around more than 100 days Figure 3. Comets with long time intervals of data arranged in order of de- creasing qosc (from top to bottom). Time distribution of positional obser- vations with corresponding heliocentric (red curve) and geocentric (green curve) distance at which they were taken. Horizontal dotted lines show the perihelion distance for a given comet whereas vertical dotted lines -- the moment of perihelion passage. The horizontal axis for all panels is exactly the same. For Comet C/2007 Q3 time interval, where the O-C diagrams give significant trends in right ascension and declination (remains visible even in a NG model) is shown in light gray -- this period correlates with the moment when a secondary fragment broke away from the nucleus. before perihelion). Due to some slight trends in the O-C diagram we added the gravitational PRE model as an alternative for study- ing past dynamics of this object. C/2006 Q1 McNaught Here, NG models result in O-C distributions good approximated by Gaussian distribution and no trends in O-C-diagram was noticed. The normal component of NG motion in the standard MSY model seems to be important in orbital fitting and gives a significantly decrease of rms in comparison to model with two NG parameters (radial and transverse components) as well as in model including also t -shift of g(r)-function and ignoring normal component (Ta- ble 2). We added the gravitational PRE model as an alternative for studying past dynamics of this object -- just for comparison. C/2007 N3 Lulin This comet has the smallest perihelion distance in this group, qosc = 1.21 au. Therefore, starting this investigation we suspected that we would get some interesting information about g(r)-like function for this comet. Unfortunately, models based on individually adjusted g(r)-like function did not give noticeably better fitting to observa- tions. For NG model with three components of NG accelerations the significant decrease of rms was noticed from 0.′′50 to 0.′′35. It can be seen (Table 2) that values of three NG parameters, A1, A2, A3, c(cid:13) 2013 RAS, MNRAS 000, 1 -- 20 although small, are well-defined, however A3 component slightly dominates over A1 and A2. Asymmetric model with two compo- nents of NG accelerations and tau (A1 + A2 + t ) gives the rms of 0.′′49 (significantly greater than symmetric model with A3) Asym- metric model with four parameters (A1 + A2 + A3 + t ) gives rms of 0.′′35, thus indistinguishable from the symmetric model with three parameters (A1 + A2 + A3, see Table 2). Moreover, all considered NG models based on the entire in- terval of observations give the O-C distribution that substantially differs from a normal distribution. For that reason we would rec- ommend gravitational models PRE and POST instead, especially for studying past and future dynamics of this object. NG solution based on entire data set as well PRE (POST) type of solution give very similar values of 1/aori (1/afut). Thus, past and future dynamics of this comet seems to be very well defined. C/2007 Q3 Siding Spring This comet must be examined in a special way. In the middle of March 2010, Nick Howes reported a small secondary piece of C/2007 Q3 on the picture taken using Faulkes Telescope North. The existence of this secondary component was later confirmed dur- ing the follow-up observations taken from Mar. 17 up to Apr. 9 by many other observers (Colas et al. 2010). Indeed, this period cor- relates with a time interval where significant trends in both right ascension and in declination appear in the O-C diagrams even in the NG model of motion (light grey part of data in the third panel from the top of Fig. 3). GR orbit of C/2007 Q3 based only on pre- perihelion data is still 1a+ quality class. Maybe due to this additional event, the NG model of C/2007 Q3 based on entire data interval gives radial component, A1, smaller than the other two components A2 and A3 (Table 2). Similarly as in C/2007 N3 we noticed here important role of a nor- mal component of NG acceleration in rms decreasing. Similarly as in remaining comets in this group, the NG solu- tion based on entire data set as well PRE (POST) type of solution give similar values of 1/aori (1/afut). Thus, past and future dynam- ics also for this comet seems be well-defined. 4.2 Group B: peculiar comets We found that eight comets in the studied sample exhibit some kind of unusual activity or are troublesome in determining the osculating orbit. All of them we classified as peculiar objects and discussed in this section. In the bottom part of Fig. 4 the positional measurements for four comets with the smallest perihelion distances in our sample are shown: C/2006 P1, C/2009 R1, C/2010 X1 and C/2007 W1. Among the remaining four comets in this group we have C/2008 A1 with strong and variable NG effects in its motion, two more comets with NG effects clearly seen in the motion and eas- ily determinable from the entire interval of data (C/2007 W3 and C/2006 K3), and C/2009 K5, whose osculating orbit was especially difficult to determine (see below). Comets C/2007 W1 (1.2 yr of data) and C/2008 A1 (2 yr) exhibit the most manifesting NG effects in their motion among comets studied in this paper. NG models based on the entire set of positional data proved to be completely inappropriate for both these objects; the detailed discussion of NG models based on full data sets was given in Section 2.2. Moreover, NG effects appear to be variable inside the interval covered by positional data. Such c(cid:13) 2013 RAS, MNRAS 000, 1 -- 20 Near-parabolic comets observed in 2006 -- 2010 11 Figure 4. The same as in Fig. 3 for peculiar comets. The horizontal axis is common for all comets and is the same as in Fig. 3. The vertical scale unit is also the same as in the previous figure. Four comets with perihelion distances, qosc, below 1.0 au are shown in the bottom part of this figure. an erratic behaviour can be detected by using data taken before perihelion passage and after perihelion to determine the set of NG parameters, separately for both orbital branches. The results are given in Table 6, where also the NG parameters derived by Nakano (2009a,b,c) are shown. One can see that our values of NG param- eters are in very good agreement with Nakano though he assumed that A3 = 0 and used slightly different sets of data. Unfortunately, Nakano analysed NG effects for pre-perihelion and post-perihelion data separately only for C/2007 W1, so solely solutions for this comet could be compared in Table 6. C/2006 K3 McNaught NG effects easily determinable from the entire interval of data de- spite a moderate perihelion distance of 2.5 au. The radial compo- nent of NG acceleration dominates and is well-determined, we de- cided to include the normal component to the model since this gives slight improvements in rms and O-C-diagram and generalize the NG solution; no other tailored model is needed for this object. C/2006 P1 McNaught This comet (qosc = 0.17 au) was the second brightest comet ob- served by ground-based observers since 1935 and demonstrated a spectacularly structured huge dust tail (e.g. Jones et al. 2008). It might be surprising that in case of such an active comet with extremely small perihelion distance it was possible to obtain a very well determined, standard NG orbit from the whole data set. Best NG solution is based on radial and transverse components of NG acceleration. 12 M. Kr´olikowska and P.A. Dybczy´nski Table 6. NG parameters derived in the present investigation and by Nakano (Nakano 2009a,b,c) for two comets with strongly manifesting NG effects in the motion. Source NG model [1] [2] A1 [3] NG parameters (Eq. 2) in units of 10−8 A2 A3 [4] [5] present NK1731A NK1731B present NK1807 PRE POST PRE POST NG PRE POST NG C o m e t C/2007 W1 B o a t t i n i 1.002 ± 0.139 5.866 ± 0.272 1.905 ± 0.072 5.753 ± 0.111 -0.7253 ± 0.0032 -0.783 ± 0.172 -0.5243 ± 0.0302 -0.705 ± 0.150 -0.4916 ± 0.0703 0.138 ± 0.250 -- -- 5.150 ± 0.032 4.608 ± 0.136 10.094 ± 0.282 5.099 ± 0.047 C o m e t C/2008 A1 M c N a u g h t 0.1939 ± 0.0076 1.844 ± 0.203 -4.431 ± 0.306 0.9915 ± 0.0201 1.894 ± 0.233 6.142 ± 0.291 0.7641 ± 0.0272 -- No of obs. [6] 926 777 804 733 997 393 544 869 rms ′′ [7] 0.49 0.59 0.57 0.63 0.44 0.28 0.54 0.71 interval of data [yyyymmdd] [8] 20071120 -- 20080612 20080630 -- 20081217 20071120 -- 20080612 20080630 -- 20081107 20080110 -- 20100117 20080110 -- 20080928 20081001 -- 20100117 20080110 -- 20090714 C/2007 W1 Boattini C/2008 A1 McNaught This comet is also discussed at the beginning of this section to- gether with comet C/2008 A1. We detected strong and variable NG effects in its motion. As a result we recommend separate, non- gravitational PRE and POST models for studies of its past and future dynamics. The similar approach was proposed by Nakano, see Table 6 for more details. Our all osculating orbit solutions (Section 2.1, Tables 2-3) show that C/2007 W1 (qosc = 0.85 au) , having a negative value of 1/aori is an excellent candidate to be an interstellar comet. It is therefore important at this moment to refer to two quite different publications on this comet. In the first, Villanueva et al. (2011) measured a chemical composition of C/2007 W1 using NIRSPEC at Keck-2. They derived the abun- dance ratios of eleven volatile species relative to the water and concluded that almost all these ratios are among the highest ever detected in comets (see figure 8 therein). Thus, this comet seems to be very peculiar also in the light of chemical composition. interesting from our point of view, In the second paper Wiegert et al. (2011) reported a new daytime meteor shower de- tected using Canadian Meteor Orbit Radar. They analysed the data in the 2002-2009 interval and detected Daytime Craterid shower in two years: in 2003 and 2008. Next, they concluded that this shower can be connected with C/2007 W1 because of the similarity of both sets of orbital elements, excluding eccentricities. They argued that the eccentricity of C/2007 W1 is known with so great uncertainty that this comet can be a short-period comet giving two showers. Ac- cording to the authors, the second shower detected in 2008 would be after C/2007 W1 perihelion passage in 2007, the first one -- at the previous perihelion passage of this comet. In our opinion, how- ever, the orbit of this comet is known much more accurately than Wiegert et al. (2011) argued. Thus, only the shower in 2008 could be related to C/2007 W1. C/2007 W3 LINEAR NG effects clearly seen in the motion of this comet and the standard NG model is easily determinable from the entire interval of data. The asymmetric model is marginally determinable, however with large uncertainties of t -shift and with no improvements of orbital fitting (Table 2). No other tailored model is necessary to represent the positional data of this object. This comet is also discussed at the beginning of this section to- gether with comet C/2007 W1. Due to the nature of the detected NG forces (strong and variable) we recommend separate, nongravi- tational PRE and POST models for studies of its past and future dy- namics. For comparison we show also two models based on the en- tire data set: an asymmetric one and a standard, symmetric model. C/2009 K5 McNaught This comet also seems to be a peculiar object. Considering rather long time interval of observations (2.5 yr) it could be option- ally included into the group of comets with long sequences of data since its observations cover quite large heliocentric distances from 4.35 au before perihelion to 6.25 au after perihelion (orbit 1a+ class). This fact, together with small perihelion distance, should create a perfect opportunity to determine the NG effects either from the whole data set (as for all comets of long data sets in this paper, previous subsection) or from pre-perihelion and post-perihelion data individually, as in the case of C/2007 W1 or C/2008 A1. In contrast to the expectation, the NG effects cannot be reliably deter- mined from the entire data set of C/2009 K5 as well as individu- ally from pre- or post- perihelion orbit branches. Thus, we decided to include this comet to peculiar objects. We recommend separate gravitational PRE and POST models for this comet and present two other models for comparison. C/2009 R1 McNaught This comet of a very small perihelion distance (qosc = 0.405 au) was observed only prior to its perihelion passage and was lost soon after it. There is no information about this comet after perihelion and we can speculate that this comet has disintegrated. Among comets observed only before perihelion passage C/2009 R1 ex- hibits strong and well-determinable NG effects during interval cov- ered by positional measurements and belongs to comets with good quality NG orbit (see also Fig. 1). C/2010 X1 Elenin Another peculiar comet of a very small perihelion distance (qosc = 0.482 au). It starts to disintegrate about one month before peri- helion and it turn out that only a pure gravitational orbit can be c(cid:13) 2013 RAS, MNRAS 000, 1 -- 20 Near-parabolic comets observed in 2006 -- 2010 13 not succeed in determining NG effects in their orbital motion. We can expect this from quick inspection of Fig. 5. In the remaining three cases, the situation is not as clear as above. From a review of Fig. 5 (notice that the scale of horizontal and vertical axes in both panels are the same) it is not obvious that NG effects are not detectable within time intervals covered by data. Of course, in this type of a qualitative discussion we assume some physical similarities between considered comets, mainly in their global activity. In fact, for three comets described below, C/2006 HW51, C/2006 L2 and C/2008 T2, we detected some traces of NG ef- fects in positional data with a negative radial component of the NG acceleration (models marked as NGA1 in column [2] of Table 3; see also discussion in Section 2.1). However, in all these cases we noticed only slight improvements in data fitting in comparison to GR orbit. Therefore, the interpretation of these NG models can be twofold. These models reflect the actual NG acceleration of these comets (for example giving some indication of the existence of ac- tive sources on the nucleus as was mention above), or the observed slight improvements in data fitting are only the result of a larger number of parameters taken into consideration when determining the NG orbit. We decided, however, to include these NG models to Table 3 solely as alternative models for dynamical status discus- sions based on the previous perihelion calculations, see Part II. Figure 5. The same as in Fig. 3 for comets of non-detectable or very weak NG effects. The horizontal axes covers exactly the same time-interval in both panels and the same time-scale was used as in Fig. 3. C/2006 HW51 Siding Spring well determined from the shorter interval of data -- the part of data not included in the orbit determination is shown in light grey in Fig. 4. On the occasion of this comet, it is worth mentioning that even when a cometary disintegration was observed, some au- thors derived NG orbit with standard (constant!) NG parameters A1, A2, A3, although they describe the systematic acceleration act- ing on a comet, which is a function of the heliocentric distance from the Sun. These standard NG parameters cannot correctly ac- count for a sudden change in orbital motion due to comet's par- tial disruption. Therefore, the interpretation of results obtained in such a case should be restricted to the statement that some NG ef- fects are clearly seen in the cometary motion but nothing more. It seems to us that the values of orbital elements in such a NG case also should be treated with great caution. Comet C/2010 X1 just may be an good example of such a case of misusing the standard MSY method. 4.3 Group C: comets of non-detectable or very weak NG effects Whether NG effects are noticeable in comet's motion or not de- pends on many factors such as quality and structure of data, the general level of activity and physical properties of comet (the nu- cleus structure, chemical composition, its shape and mass). Thus, generally each case should be individualized. However, it turns out that we often can pretty well predict whether it is possible to de- termine the NG orbit from the inspection of structure of the data, where by 'structure' we mean here all that can be seen in the plot of the heliocentric and geocentric distances of all positional mea- surements. It is rather not surprising that for three of six comets to be described in this section, namely for C/2007 O1 (qosc = 2.88 au), C/2009 O4 (qosc = 2.56 au) and C/2008 J6 (qosc = 2.00 au) we do c(cid:13) 2013 RAS, MNRAS 000, 1 -- 20 The data structure of this comet is qualitatively quite similar to that of comet C/2006 K3, a number of measurements is also very similar (about 200 observations in both cases). Furthermore, both comets passed perihelion rather far from the Sun (more than 2.2 au). It seems, however, that the longer time interval of data (1.7 years), resulting in a wider range of observed heliocentric distances before perihelion in the case of C/2006 K3 (3.95 au at the moment of dis- covery in comparison to 2.87 au at for C/2006 HW51), causes that the NG acceleration in C/2006 K3 is clearly visible in its motion, while in the motion of comet C/2006 HW51 it is not so easily dis- cernible. The NG effects can be firmly detected in the motion of C/2006 K3 also due to the fact that this comet was significantly more active than C/2006 HW51. One can speculate that the activ- ity of the comet C/2006 HW51 is limited only to some active areas somehow specifically located on the surface of the comet causing that standard MSY model gives some traces of NG effects with neg- ative radial component of NG acceleration. We recommend pure gravitational model for this object, but we also present NG and gravitational PRE models for comparison. C/2006 L2 McNaught Similar arguments to the presented above seems to be correct in the case of comet C/2006 L2, that also passed through perihelion not very close to the Sun (qosc = 1.994 au) and displays some traces of NG effects with negative A1. We recommend pure gravitational model for this object, while NG and gravitational PRE models are presented for comparison. C/2007 O1 LINEAR In the data of C/2007 O1, we have more than one-year gap in po- sitional measurements. This is due to the fact that the object was initially discovered as an asteroid, and more than a year later it 14 M. Kr´olikowska and P.A. Dybczy´nski was rediscovered as a comet. Data of such an unusual structure, where 12 observations were collected far before perihelion when the object was more than 4 au from the Sun and the rest of data were taken after the perihelion passage, together with the rather large perihelion distance, can effectively prevent the determination of NG effects despite quite a long period formally covered by mea- surements, so we present only pure gravitational solution for this comet. C/2008 J6 Hill Comet C/2008 J6 was followed from a heliocentric distance of 2.04 au to 3.43 au, thus in a wider range of heliocentric distances what is more promising. However, it was discovered after passing through the perihelion. In this case, the chance to detect the NG ac- celeration in the motion of even moderately active comet is low, only GR model is presented. C/2008 T2 Cardinal Comet C/2008 T2 seems to be a more unusual object. In terms of the structure of data we have a situation quite similar to that of comet C/2007 W3 (Fig. 4), but here we have much more mea- surements as well as the comet came closer to the Sun at peri- helion (1.20 au compared to 1.78 au for C/2007 W3). Both facts should allow us to determine the NG effects easier for C/2008 T2 than for C/2007 W3. However, in this comet the NG effects are loosely detectable (almost at a noise level. One can suppose that comet C/2008 T2 probably exhibits another character of activity than C/2007 W3 or/and physically differs from C/2007 W3. We recommend GR model obtained from the entire dataset for this ob- ject, but we also present NG and gravitational PRE models for com- parison. C/2009 O4 Hill Comet C/2009 O4 was observed only before perihelion passage in the narrow range of heliocentric distance from 3.04 au to 2.57 au and only during 4.5 months period, therefore only the gravitational model can be obtained. 4.4 Group D: comets of weak quality of osculating orbits Due to a generally poor quality of osculating orbits of these four comets in comparison to others and their asymmetric distribution of observations relative to perihelion we should be very careful when making statements about their past and future motion. Moreover, we should admit that one of them (C/2006 VZ13), split into separate fragments or disintegrated near perihelion. C/2006 VZ13 LINEAR The observations of C/2006 VZ13 (qosc = 1.01 au, the smallest per- ihelion distance in this group) were stopped very soon after its per- ihelion passage. C/2006 VZ13 (Fig. 6), was observed longer than remaining objects in this group, about eight months, while three others less than five months. However, it belongs to this group be- cause in fact the only adequate orbit for its past dynamical evolution can be determined from the pre-perihelion data. We decided to cut the pre-perihelion string of data on July 1, 2007 (1.23 au from the Sun), i.e. 40 days prior to the perihelion passage because with such Figure 6. The same as in Fig. 3 for comets of weak quality of osculating orbits; also the same time-scale was used in horizontal axis. The time inter- val not taken for the determination of PRE type of osculating orbit of comet C/2006 VZ13 is shown in light grey ink. a restriction we derived the NG osculating orbit that gives O-C dia- gram free from any trends in right ascension or declination. For this reason the orbit based on this time interval is the most appropriate as starting orbit for the past dynamical evolution (see Part II of this investigation). The range of data that was not used for PRE type of model determination is shown in light grey ink in Fig. 6. Thus, the data time interval taken for the past evolutionary studies was only ∼7.5 months in this case. Shortening the time interval of data by almost 20 per cent resulted here in a more than four-fold reduction of a precision of 1/aosc-determination, and resulted in a decrease in its orbital class from 1b (NG orbit determined from the entire data set) to class 2a (NG orbit based on pre-perihelion data). For a future dynamical evolution we have no choice and for this purpose the NG orbit based on the entire data set was used. It is worth not- ing that the future orbit should be treated with great care because of our ignorance of the fate of this comet shortly after perihelion passage (the last observation was taken four days after perihelion). C/2007 Q1 McNaught This comet have the greatest perihelion distance (qosc = 3.01 au) and the worst quality orbit (3a) in this group (and in the whole sample of comets examined in this paper) is determined for this comet.Its poor quality is a direct consequence of the shortest data arc throughout the sample (only 24 days) and also of an un- usual moment of discovery, more than eight months after it passed through perihelion. Usually, when the astrometric observations in- clude perihelion then the orbit have a chance to be more precisely determined. Only the pure gravitational orbit can be determined in this case. C/2008 C1 Chen-Gao Despite a very short time interval of data, the NG effects are de- tectable in the motion of this comet (qosc = 1.26 au). However the NG parameters are not well determined and the decrease in rms is not observed, so we present the NG model for this comet only for comparison and recommend the GR model. c(cid:13) 2013 RAS, MNRAS 000, 1 -- 20 C/2010 H1 Garradd It is impossible to detect the NG effects from the set of data for this comet due to very narrow data range -- observations span a short time period. Additionally, we can notice that it passed the perihelion at the moderately large distances from the Sun (qosc = 2.745 au), and only a small number of measurements were taken. As a result only a GR model can be obtained. 5 ORIGINAL AND FUTURE ORBITS In the present numerical calculations, a dynamical evolution in- vestigation of a given object starts from the swarm of VCs con- structed using the osculating orbit (so-called nominal osculating orbit) determined in the respective model shown in Table 3. We performed dynamical calculations for each model presented in this table. Of course, for models based on PRE data we follow only the past evolution, whereas for models based on POST data -- only the future evolution. Each individual swarm of starting osculating or- bits is constructed according to a Monte Carlo method proposed by Sitarski (1998), where the entire swarm fulfil the Gaussian statistics of fitting to positional data used for a given osculating orbit deter- mination. Similarly to our previous investigations (see for example Paper 1), each swarm consists of 5 001 VCs including the nominal orbit; we checked that the number of 5 000 orbital clones gives a sufficient sample for obtaining reliable statistics at each step of our study, including the end of our numerical calculations, i.e. at the previous and next perihelion passage (see Part II of this investiga- tion). Therefore, we are able to determine the uncertainties of origi- nal and future reciprocal of semimajor axis (1/aori and 1/afut), that are here taken at 250 au from the Sun, i.e. where planetary pertur- bations are already completely negligible (Todorovic-Juchniewicz 1981). Values of 1/aori and 1/afut and their uncertainties derived by fitting the 1/a-distribution of original and future swarm of VCs to Gaussian distribution are given in columns [9] -- [10] of Table 3. All 1/a-distributions as well as distributions of other orbital elements of analysed comets were still perfectly Gaussian at 250 au from the Sun. However, further evolution under the Galactic tides and stellar perturbation can potentially introduce significant deforma- tions in the initially 6D-normal distribution of orbital elements in the swarms of VCs (up to 10D-normal distribution in the NG case), what will be shown in the Part II of this investigation. 5.1 Original semimajor axes of NG orbits First, it is important to notice that an osculating NG orbit deter- mined from a given set of data is not the same orbit as one de- termined from these same observations but under the assumption of purely gravitational motion (Kr´olikowska 2001). Therefore, the change of original 1/a-value due to incorporating the NG acceler- ation in the process of orbit determination can be even very large though the NG effects provide only modest changes in the origi- nal 1/a-value along an individual orbit. Typically, the original ec- centricity of NG orbit are smaller than eccentricity of GR orbit for a given comet with determinable NG acceleration (where both GR and NG orbits are derived from the same set of positional data). We confirmed these general findings in the present studies. However, we found some interesting atypical behaviour for comet C/2007 W1. The differences between the inverse original semimajor axes c(cid:13) 2013 RAS, MNRAS 000, 1 -- 20 Near-parabolic comets observed in 2006 -- 2010 15 Figure 7. Shifts of 1/aori due to the NG acceleration for eleven comets from the period 2006-2010 (red symbols) with well-determined NG effects and 37 previously analysed comets (black symbols; Paper 1 & 2). Three largest uncertainties of 1/aori,NG belong to comets C/1959 Y1, C/1952 W1 and C/1892 Q1. derived in NG model of motion and less realistic pure gravitational model of motion are presented for 11 comets examined here (red symbols) and 37 comets studied in the Paper1&2 (black symbols) in Fig. 7. In the bottom panel the dependence of the strength of NG forces on the osculating perihelion distance, qosc is clearly vis- ible for the whole set of 48 comets with determined NG effects. To maintain consistency with previous our studies, the values of 1/aori and their uncertainties presented in this plot are based on osculating NG orbits determined from the entire data sets except of the comet C/2007 W1 where we took the value of 1/aori determined from the NG orbit based on pre-perihelion subset of data as signifi- cantly more adequate for this comet (see also Section 4.2). Nakano (2009a,b) also considered the pre-perihelion and post-perihelion or- bital branches independently for this particular comet. He derived a similar value of 1/aori = −0.000059 au−1 on the basis of pre- perihelion data only (the detailed comparison of both NG models is given in Section 4.2. However, for the NG orbit based on the en- tire observational interval we obtained even a more negative value of 1/aori = −0.000082 ± 0.000003 au−1. Thus, C/2007 W1 Boat- tini is the first serious candidate for interstellar comet among LPCs examined by us so far. In the case of comet C/2008 A1 we show in Fig. 7 the results based on the entire data interval though for this comet the NG or- bit determined from the full interval is also highly unsatisfactory (as we discuss earlier). The largest differences in D Eori was de- rived just for this comet (the rightmost red symbol in the bottom panel of Fig.7), whereas based on a pre-perihelion subset of data we derived 1/aori = 0.000120 ± 0.000002 au−1 for NG orbit and D Eori = −0.000011 au−1. However, we would like to stress that C/2008 A1 as well as C/2007 W1 are comets that both have ex- tremely strongly manifesting and variable NG effects inside their observational intervals. 16 M. Kr´olikowska and P.A. Dybczy´nski Table 7. Three samples of comets with 1/aori < 10−4 au−1: qS, qM and qL in comparison to MWC 08. The number of comets are given for first quality class orbits in MWC 08, except the sample of comets investigated in this paper, where 3 objects with quality class 2 and 3 are included; we indicate this by '+ 3' in columns [3] & [9]. Period qosc < 3.1 au N u m b e r o f c o m e t s 3.1 6 qosc < 5.2 au i n t h e d i f f e r e n t s a m p l e s qosc > 5.2 au All comets of discovery MWC 08 Sample qS MWC 08 Sample qM MWC 08 Sample qL MWC 08 analysed by us [1] before 1900 1900 -- 1949 1950 -- 1999 2000 -- 2005 2006 -- 2010 All [2] 10 26 24 10 12 82 [3] 2 3 16 6 19 + 3 46 + 3 [4] -- 6 22 14 3 45 [5] -- -- 17 14 4 35 [6] -- -- 11 12 3 26 [7] -- -- 11 11 2 24 [8] 10 32 57 36 18 153 [9] 2 3 44 31 25 105 + 3 Sets of original and future 1/a-values taken for statistical anal- ysis (see next section) are based on the preferred osculating orbits. It is obvious that NG orbits -- if determinable in the motion of a given comet -- are always more realistic than pure gravitational or- bit derived from the same data set. 5.2 Observed distributions of original and future semimajor axes In the discussion below we do not pretend to describe planetary per- turbations on LPCs in general. Instead, we aim to present really ob- served changes in 1/a for a significant percentage of the observed LPCs, additionally for the first time fully accounting for their un- certainties. We also discuss the contribution of the NG forces into the 1/a change during a cometary flyby through the planetary sys- tem. To describe the original and future 1/a-distributions we di- vided 108 comets investigated by us so far (in the present investi- gation and Papers 1 & 2) into three subsamples of a different com- pleteness according to their osculating perihelion distances: qS -- this sample consists of 49 LPCs with qosc < 3.1 au and in- cludes comets with strongly manifesting NG effects in their mo- tion discovered before 2006 (Papers 1 & 2) and a complete sample of comets discovered in the period 2006-2010 investigated in this work. Thus, we have here 46 of 89 (52 per cent) comets discovered before 2011 with first quality class orbit by using original MSE method (Section 3) and additionally three comets of poorer quality class orbits. qM -- complete sample of 35 LPCs with 3.1 au 6 qosc < 5.2 au discovered in the period 1970-2006 (first quality class orbits, Pa- per 2). qL -- almost complete sample of 24 LPCs with qosc > 5.2 au dis- covered before 2008 (first quality class orbits, Paper 2); only two objects, C/2005 L3 and C/2007 D1, were not taken into account in Paper 2 because they were still observable at the beginning of 2012. More detailed descriptions of these three samples are given in Table 7 where the completeness of these samples in comparison to MWC 08 is shown. MWC 08 is complete to the end of the 2007. In the period of 2006 -- 2010, considerable number of 48 LPCs with 1/aori < 10−4 au−1 were discovered (eight of them are still observ- able), and all of them, having qosc < 3.1 au are investigated in this paper and are shown by red points in Fig. 8. In total, our sample of 108 comets discussed here constitutes more than 60 per cent of all first class Oort spike comets discovered after 1800 (78 per cent of those discovered after 1950), for which observations are finished. For the present discussion only the preferred models (shown in the first row for each comet in Table 3) were taken into consideration. In a pure GR model of cometary motion, the values of d (1/a) = 1/afut − 1/aori directly inform about planetary pertur- bation which a comet suffered passing through the planetary sys- tem. When the NG osculating orbit is derived we should expect that NG effects contribute in some extent to the value of d (1/a). It is not so easy to measure these contributions, in particular for these comets where two separate NG osculating orbits are preferred to describe their actual orbital motion. We have made such an attempt in the case of comets with NG orbits investigated here. Table 8 presents three values of d (1/a) for all comets with NG models chosen by us as the preferred solution in Table 3. For each comet, the first row gives d (1/a)-value for the recom- mended NG model, the second row -- d (1/a)-value for the best GR model, whereas the third row shows d (1/a)-value derived for the same NG osculating orbit taken as a starting orbit, however NG accelerations were not included during the integration back- ward and forward to 250 au from the Sun for original and future 1/a-determination. Of course, the last orbit does not represent data (see column [5]) since the best GR orbit that gives good data fit- ting (second row, column [4]) is substantially different than the best NG orbit. However, the difference between first row and second (and third) row may give some estimation for the NG contribution to the d (1/a)-value. For comets with separate NG orbits derived for pre-perihelion branch and post-perihelion branch, we calculate the d (1/a) from the equation: d (1/a) = (1/afut − 1/aosc,post) − (1/aori − 1/aosc,pre), where 1/aosc,post and 1/aosc,pre are values of 1/aosc for post- perihelion and pre-perihelion orbit, respectively. One can notice using Table 8 that for comets with perihelion distance greater than 2.0 au the d (1/a)-value can be interpreted as planetary perturbations even in the NG models of motion. We have also confirmed this conclusion for all comets studied in Pa- per 1 & 2 where differences in d (1/a)-value between NG and GR models do not exceed 10 per cent, except C/1997 J2 where we obtained this difference at the level of 40 per cent due to a very small value of d (1/a) (−41 and −30 for GR and NG orbit, respec- tively). However, for smaller perihelion distances (qosc < 2.0 au) we can obtain even a change of sign of d (1/a), like for C/2009 R1 (qosc = 0.405 au) and C/2008 A1 (qosc = 1.07 au) with the highest magnitude of NG acceleration (columns [10] -- [13] in Table 8). Dif- ferences of more than 30 per cent in d (1/a) between NG and GR models are found only for three comets investigated here (two pecu- liar comets, C/2008 A1 and C/2009 R1, and one of weak quality of osculating orbit, C/2006 VZ13) and five comets of 37 with NG or- c(cid:13) 2013 RAS, MNRAS 000, 1 -- 20 Near-parabolic comets observed in 2006 -- 2010 17 Table 8. Contribution of planetary perturbations and NG effects to d (1/a) = 1/afut − 1/aori for nine comets with NG orbits given as the preferred models of their motion in Table 3. Comets are arranged in order of increasing qosc. In the first row of individual comet d (1/a) for the main NG model (given in the first rows in Table 3) is presented, the second row shows d (1/a) obtained using the best GR model and third row displays the d (1/a) derived using the same NG orbit as in the first row, however during integration backward and forward to the 250 au from the Sun the NG acceleration was excluded. In the second row (column [9]) the percentage change in d (1/a) relative to first row is also given. Values of 1/aori, 1/afut, and d (1/a) are given in units of 10−6 au−1 used in this paper, and the magnitude of the NG acceleration, A = qA2 3, in units of 10−8 au·day−2. 1 + A2 2 + A2 Comet [1] C/2006 P1 qosc [au] [2] 0.171 C/2009 R1 0.405 C/2007 W1 0.850 C/2006 VZ13 1.015 C/2008 A1 1.073 C/2007 W3 1.776 C/2006 OF2 2.431 C/2006 K3 2.501 C/2006 Q1 2.764 [3] NG GR GR NG GR GR NG GR GR NG GR GR NG GR GR NG GR GR NG GR GR NG GR GR NG GR GR type of Model of motion orbit for 1/aori [4] NG GR NG NG GR NG rms for [4] ′′ 1/aori Model of orbit for 1/afut 1/afut d (1/a) [5] 0.25 0.25 0.79 0.51 0.63 4.87 0.49 0.61 4.50 0.39 0.40 5.31 0.28 0.47 14.11 0.52 0.54 3.70 0.36 0.38 0.49 0.54 0.69 1.19 0.37 0.50 2.05 [6] 57.2 31.4 37.4 12.2 -89.6 12.2 -42.7 -14.1 71.8 14.0 13.6 132.8 120.8 132.4 137.6 31.4 14.2 21.2 21.2 23.7 24.9 61.0 61.9 62.0 51.1 38.3 52.7 [7] NG GR NG NG GR NG POST,NG POST,GR POST,NG NG GR NG POST,NG POST,GR POST,NG NG GR NG NG GR NG NG GR NG NG GR NG [8] 467.6 490.7 496.7 170.4 -723.0 -592.4 554.4 892.0 716.5 491.2 442.2 371.2 246.5 258.4 1065.05 343.9 408.8 415.8 -658.8 -662.4 -661.1 -131.3 -123.0 -122.9 707.4 696.6 711.0 [9] 410.5 459.3 (11.4%) 459.3 158.3 -812.6 -551.7 586.5 632.9 (7.9%) 634.0 269.3 435.6 (62%) 30.4 -534.1 268.9 267.6 312.5 394.6 (26.3%) 394.6 -680.0 -686.1 (0.9%) -686.1 -192.3 -184.9 (3.9%) -184.9 656.4 658.3 (0.3%) 658.3 N G p a r a m e t e r s pre-perihelion A2/A1 [11] 0.23 A [10] 1.33 post-perihelion A2/A1 [13] 0.23 A [12] 1.33 6.02 0.24 6.02 0.24 1.33 0.72 5.92 0.13 2.13 0.46 5.63 0.66 4.98 0.41 12.62 0.61 5.56 0.45 5.56 0.45 2.75 0.58 2.75 0.58 15.85 0.14 15.85 0.14 35.19 0.06 35.19 0.06 PRE,NG PRE,GR PRE,NG PRE,NG GR PRE,NG PRE,NG PRE,GR PRE,NG NG GR NG NG GR NG NG GR NG NG GR NG bits determined in Papers 1 & 2. In the following discussion, when we use the term 'planetary perturbation', we keep in mind that iden- tification of planetary perturbation with d (1/a) = 1/afut − 1/aori in the case of NG orbit is subject to an additional contribution, that exceed 30 per cent of d (1/a)-value just for a few of them. This contribution is obviously caused by slightly different evolution of orbital elements due to NG acceleration along a given orbit. In Fig. 8 we show d (1/a) = 1/afut − 1/aori, in a function of osculating perihelion distance for the sample of 22 comets investi- gated here (red points) in comparison with all remaining 86 comets investigated by us so far. The uncertainties of d (1/a) were derived by calculating the d (1/ai) where i = 1, ...,5001 for each VC in the swarms, and then fitting the derived d (1/a)-distribution of in- dividual swarm of VCs to Gaussian distribution. The uncertainties of perihelion distance and inclination are significantly below the size of points in these figures. It is interesting to notice that the largest d (1/a)-uncertainty in the group of comets investigated in the present paper and in the entire sample of 108 Oort spike comets, were obtained for C/2009 R1 McNaught (orbital class: 1b), easy recognizable red point with largest vertical error bar. For C/2009 R1 we have d (1/a) = 154 ± 199 in units used in this paper. At a first glance, this comet with large inclination to ecliptic (∼ 77◦) seems to suffer a relatively moderate planetary perturbations (NG model). However, we found that the nominal VC have passed about 1.3 au from Jupiter in 2009 (on August 25), and 0.15 au from Mercury in c(cid:13) 2013 RAS, MNRAS 000, 1 -- 20 July 2010 (six days after perihelion passage, the last observation of this comet was taken 10 days earlier). Among the sample of comets from the 2006 -- 2010 period, the second comet with large error of d (1/a) is C/2007 Q1 Garradd (d (1/a) = −510 ± 62), however, this resulted from the extremely poor quality of its orbit (class 3a). Apart from three comets (C/1940 R2, C/1980 E1 and C/2002 A3) lying outside Fig. 8 all other suffer only moderate per- turbations in 1/a, with a significant per cent of very small values (points inside the horizontal band around zero in the figure), con- versely to widely spread opinion, that comets coming closer to the Sun than Jupiter almost certainly suffer strong planetary perturba- tions. It means, that in the sample of analysed Oort spike comets the number of objects that have a chance for returning in the next perihelion passage as an Oort spike comet is remarkable (see mid- dle panels in Fig. 9) and accounts to about 14 per cent of all in- vestigated by us near-parabolic comets (108 objects). This shows the observed transparency of the Solar system (see Dybczy´nski 2004). This transparency seems to be significantly different for large perihelion near-parabolic comets than for small perihelion comets and, taking into account small number statistics fluctua- tions it is consistent with the results of the Monte Carlo simulations of this phenomenon, recently published by Fouchard et al. (2013). The detailed percentages of comets with 1/afut still smaller than 100 × 10−6 au−1 within three discussed samples are as follows: 18 M. Kr´olikowska and P.A. Dybczy´nski Figure 8. Change of reciprocals of semimajor axis during passage through the planetary zone, d (1/a) = 1/afut −1/aori, as a function of osculating per- ihelion distance, where the sample of 22 comets investigated in this paper are shown by red dots. The sample of 86 comets studied in Paper1&2 were divided into three subsamples qS, qM and qL (see Table 7) and are given as black, blue and green dots, respectively. The vertical line marks the position of Jupiter semimajor axis at 5.2 au. The horizontal grey band between two horizontal dashed lines shows the area where planetary perturbations are smaller than 10−4 au−1. qS sample qM sample qL sample 10 per cent 6 per cent 33 per cent. This feature of the 1/afut distribution is also clearly visible in Fig. 8, where the distributions of d (1/a) as well as the distri- butions of 1/aori and 1/afut are presented in Fig. 9 in the bottom, top and middle panels, respectively. The filled light steel-blue his- tograms show the distributions of 22 comets investigated in this paper. These steel-blue histograms, are overprinted on the blue his- tograms that represent qS sample (49 comets). Similarly, the distri- butions of comets of qosc > 5.2 au (filled light turquoise histograms, 24 objects) are overprinted on the distributions of qM+qL sample of comets (filled green histograms, 59 objects of qosc > 3.1 au). It should be stressed here that the uncertainties of 1/aori, 1/afut and d (1/a) are taken into account in these histograms by consid- ering full cloud of 5001 virtual orbits for each comet. For exam- ple, the global distribution of 1/aori in both upper panels (marked by thick black ink) is based on all clouds of VCs of 108 comets, thus in total of 540 108 VCs, whereas the global distributions of 1/afut and d (1/a) are based on 107 cometary swarms (in total of 535 107 VCs) where C/2010 X1 was not taken into account for fu- ture distribution because of its disruption during perihelion pas- sage. One can see that a prominent maximum of the observed plan- etary perturbation is visible only for comets with qosc > 3.1 au whereas for qS sample of comets the distribution of planetary per- turbation is broad with no sharp peak. There is another interesting feature visible in Fig. 9. The sec- ondary peak between 0.0005 au−1 and 0.0006 au−1 is present in the distributions of 1/afut that is visible in the qS sample as well as in the sample of comets with qosc > 3.1. This corresponds to Figure 9. Distribution of 1/aori (top panels), 1/afut (middle panels) and d (1/a) for small perihelion (qosc < 3.1 au, left-side filled histograms) and large perihelion (qosc > 3.1 au, right-side filled histograms) Oort spike comets. The histogram given in each panel in thick black ink always repre- sents the distribution of the whole sample of 108 comets. The uncertainties of 1/a-determinations were incorporated into these 1/a-histograms by tak- ing the full cloud of VCs for each comet. The light steel-blue histograms in the left-side panels show the sample of 22 comets investigated in this pa- per, where the light turquoise histograms in the right-side panels represent the sample of comets with qosc > 5.2 au. Completeness of all samples are shown in Table 7. the local maximum between 1 700 au and 2 000 au for the future semimajor axes. Outside the value of 0.0006 au−1 we observe sig- nificant and fast decrease in the 1/afut-distribution. It is interesting that the similar trend of decrease is observed around 0.0006 au−1 in the 1/aori-distribution for the sample of near-parabolic comets in MWC 08 (of first quality orbits). We noticed that the investigated here sample of small perihe- lion Oort spike comets differs from the sample of small perihelion comets with strongly manifesting NG effects investigated in Pa- per 1 in the context of future history of these objects. Then, we derived that 60 per cent of them will be lost on hyperbolic orbits in the future, here we have only 33 per cent of such comets (seven objects escaping on hyperbolic trajectories and one object that dis- integrated at perihelion). Taking into account all Oort spike comets with small perihelion distances investigated by us so far (49 ob- jects) we have now 49 per cent of comets escaping in the future from Solar system on hyperbolas whereas for the whole sample of comets -- the similar value of 53 per cent. Thus, the small perihe- lion cometary sample and the large perihelion sample appear to be quite similar in this regard. However, more comets, especially these with qosc < 3.1 au need to be analysed with the NG effects taken into account. In the Part II we will discuss the differences between both samples in the context of dynamical status of near-parabolic comets. c(cid:13) 2013 RAS, MNRAS 000, 1 -- 20 6 SUMMARY We showed that the individualized approach to the osculating or- bit determination is advisable, especially for the purpose of past and future dynamical evolution of near-parabolic comets. Develop- ing such an attempt we have determined the osculating orbits for a complete sample of so-called Oort spike comets with a small peri- helion distance (qosc < 3.1 au) discovered in five year period from 2006 to 2010. For eleven of these comets (50 per cent) we detected NG effects in their orbital behaviour using entire sets of data (Ta- ble 1) and discussed various NG models (Table 2). The detailed in- vestigation shows, however, that in five of them even more individ- ualized approach is necessary (Table 3). Thus, we determined oscu- lating orbits for pre-perihelion and post-perihelion orbital branch, separately. We argued that the separate NG solutions for each or- bital leg for two of them, C/2007 W1 and C/2008 A1, are more adequate for past and future orbital evolution. Our solutions for both comets are very similar to solutions previously obtained by Nakano (Section 4.2). Additionally, for C/2006 VZ13 the NG so- lution based on pre-perihelion orbital leg is presented. For the re- maining two comets with NG effects detectable in the entire (very long) data sets, C/2007 N3 and C/2007 Q3, we proposed GR or- bits derived from a pre-perihelion and post-perihelion branch as the best osculating orbit for past and future evolutionary calculations, respectively. Unfortunately, the NG effects were indeterminable for the pre-perihelion and post-perihelion leg separately in both comets despite long-time series of data. Next, we discuss the possibility of NG acceleration detection in the motion of near-parabolic comets in the light of data location along their orbital tracks. We also proposed a modified orbital accuracy assessment on the basis of classical MSE method. Using 22 Oort spike comets in- vestigated in this paper and 86 comets examined in Papers 1 & 2 we showed that the proposed modifications provide a better diver- sification between orbital quality classes of currently discovered comets. Next, we have analysed the original and future inverse semi- major axes of these 22 near-parabolic comets taken at the distance of 250 au before their entrance to the inner Solar system and at the distance of the 250 au after perihelion passage, respectively. We discussed these results in the context of the whole sample of 108, so called, Oort spike comets investigated by us so far. Our conclu- sions are as follows: • We noticed a different shape and slightly different median value of 1/aori-distributions for small perihelion (qosc < 3.1 au) and large perihelion (qosc > 3.1 au) samples of near-parabolic comets. In the upper panels in Fig. 9, the 1/aori-distribution for small per- ihelion comets is more broad than for large perihelion comets. However, in these distributions included are both dynamically new comets as well as dynamically old. Thus, a more detailed discus- sion is necessary and this will be given in Part II where the results for previous perihelion distance are presented. Having the value of qprev we can construct 1/aori-distribution for dynamically new and dynamically old comets separately, and then discuss the shape of actual Oort spike, i.e. comets first time coming from the Oort spike in the light of dynamical investigation covering three consecutive perihelion passages (thus in a scale of tens of millions of years). to have an interstellar origin. This, • Among investigated near-parabolic comets only C/2007 W1 together with seems Villanueva et al. (2011) findings that the abundance ratios of al- most all important volatiles in C/2007 W1 are one of the highest ever detected in comets, makes this comet particularly unique. c(cid:13) 2013 RAS, MNRAS 000, 1 -- 20 Near-parabolic comets observed in 2006 -- 2010 19 • Prominent maximum of d (1/a) is visible only for comets with qosc > 3.1 au whereas for small perihelion comets the distribution of d (1/a) is broad with no sharp peak (Fig. 9). • The number of comets leaving the Solar system as so called Oort spike comets is 14 per cent in the investigated sample (middle panels in Fig. 9) Moreover, about half of them (7 per cent of all, 33 per cent of qL sample) are comets with large perihelion distance (qosc > 5.2 au). To say, that these comets will be Oort spike comets in the future is rather premature and the analysis of their future motion and next perihelion distance is necessary -- this will be given in Part II of this study. • In the sample of near-parabolic comets with small perihelion distances investigated by us so far (49 objects, among them 22 in- vestigated in this paper) we have now 49 per cent of comets escap- ing in the future from the Solar system on barycentric, hyperbolic orbits. Since the similar value of 53 per cent is derived for the entire sample, we conclude that the small perihelion and large perihelion near-parabolic samples appear to be similar in this regard. How- ever, more comets, especially these with qosc < 3.1 au need to be analysed by taking into account NG effects. • We noticed a secondary peak between 0.0005 au−1 and 0.0006 au−1 in the distribution of 1/afut (middle panels in Fig. 9). This corresponds to a local maximum between ∼1 700 au and 2 000 au for the future semimajor axes. Above the value of 0.0006 au−1 we observe a significant and fast decrease in the 1/afut-distribution. A similar trend of decrease is observed around 0.0006 au−1 in 1/aori-distribution for the sample of near-parabolic comets in MWC 08 (of first quality orbits). ACKNOWLEDGEMENTS We thank the anonymous referee for the review that improved this paper. The main part of the orbital calculation was performed us- ing the numerical orbital package developed by Professor Grzegorz Sitarski and the Solar System Dynamics and Planetology Group at SRC PAS. References Colas F., Manzini F., Howes N., Brysinck, E., 2010, IAU Circ., 9135, 2 Dones L., Weissman P. R., Levison H. F., Duncan M. J., 2004, in Comets II. Festou M. C., Keller H. U., Weaver H. A., eds, Universily of Arizona Press, Tuscon, pp 153 -- 174 Dybczy´nski P. A., 2001, A&A, 375, 643 Dybczy´nski P. A., 2004, A&A, 428, 247 Dybczy´nski P. A., 2005, in Knezevi´c Z., Milani A., eds, IAU Col- loq. 197: Dynamics of Populations of Planetary Systems Long term dynamical evolution of the Oort cloud comets: galactic and planetary perturbations. pp 335 -- 340 Dybczy´nski P. A., Kr´olikowska M., 2011, MNRAS, 416, 51, Pa- per 2 Dybczy´nski P. A., Kr´olikowska M., 2013a, in preparation, Part II Dybczy´nski 2013b, (http://apollo.astro.amu.edu. pl/WCP) Fern´andez J. A., ed. 2005, Comets - Nature, Dynamics, Origin and their Cosmological Relevance Vol. 328 of Astrophysics and Space Science Library, Springer, Dordrecht Fouchard M., Froeschl´e C., Matese J. J., Valsecchi G., 2005, Ce- lestial Mechanics and Dynamical Astronomy, 93, 229 Kr´olikowska M., P. A., 20 M. Kr´olikowska and P.A. Dybczy´nski Fouchard M., Rickman H., Froeschl´e C., Valsecchi G. B., 2011, A&A, 535, A86 Fouchard M., Rickman H., Froeschl´e C., Valsecchi G. B., 2013, Icarus, 222, 20 IAU Minor Planet Center, 2013, MPC Database Search (http://www.minorplanetcenter.net/db search/) Jones G. H., Morrill J. S., Battams K., Owens M. J., Howard R. A., Stenborg G. A., 2008, in European Planetary Science Congress 2008 Comet C/2006 P1 (McNaught): Observations by STEREO and SOHO. p. 832 (http://meetings.copernicus.org/epsc2008) JPL Small-Body Database Browser, 2013, JPL Database Search (http://ssd.jpl.nasa.gov/sbdb.cgi) Kr´olikowska M., 2001, A&A, 376, 316 Kr´olikowska M., in Barbara W., ed., ESA SP-500: Dynami- cal evolution of five long-period comets with evident non- gravitational effects. pp 629 -- 631 Kr´olikowska M., 2004, A&A, 427, 1117 Kr´olikowska M., 2006, Acta Astronomica, 56, 385 Kr´olikowska M., Dybczy´nski P. A., 2010, MNRAS, 404, 1886, Paper 1 Kr´olikowska M., Dybczy´nski P. A., Sitarski G., 2012, A&A, 544, A119, Paper 3 Levison H. F., 1996, in Rettig T. W., Hahn J. M., eds, ASP Conf. Ser. 107: Completing the Inventory of the Solar System Comet Taxonomy. pp 173 -- 191 Levison H. F., Dones L., Duncan M. J., 2001, AJ, 121, 2253 Marsden B. G., Sekanina Z., Everhart E., 1978, AJ, 83, 64 Marsden B. G., Sekanina Z., Yeomans D. K., 1973, AJ, 78, 211 Morbidelli A., 2005, arXiv:astro-ph/0512256 Nakano (http://www.oaa.gr.jp/oaacs/nk.htm) Nakano (http://www.oaa.gr.jp/oaacs/ nk1731a.htm) Nakano (http://www.oaa.gr.jp/oaacs/ nk1731b.htm) Nakano S., 2009, Nakano Note 1807 (http://www.oaa.gr.jp/oaacs/ nk1807.htm) Oort J. H., 1950, Bull.Astron.Inst.Nether., 11, 91 Rocher (http://www.imcce.fr/en/ephemerides/donnees/comets/trinum.php) Sitarski G., 1998, Acta Astronomica, 48, 547 Todorovic-Juchniewicz B., 1981, Acta Astronomica, 31, 192 Villanueva G. L., Mumma M. J., Disanti M. A., Bonev B. P., Gibb E. L., Magee-Sauer K., Blake G. A., Salyk C., 2011, Icarus, 216, 227 Wiegert P. A., Brown P. G., Weryk R. J., Wong D. K., 2011, MN- RAS, 414, 668 Yabushita S., 1989, AJ, 97, 262 Yabushita S., 1996, MNRAS, 283, 347 Yeomans D. K., 1994, in Milani A., Martino M. D., Cellino A., eds, IAU Symp. 160: Asteroids, Comets, Meteors 1993 Vol. 160, A Review of Comets and Non Gravitational Forces. Kluwer Aca- demic Publishers, pp 241 -- 254 Cometary Nakano S., S., Nakano Nakano Note 1731b Note 1731a 2009, 2009, P., 2013, Notes S., 2013, Notes c(cid:13) 2013 RAS, MNRAS 000, 1 -- 20
1907.00189	2	1907	2019-09-19T07:32:17	Unveiling Dust Aggregate Structure in Protoplanetary Disks by Millimeter-wave Scattering Polarization	[ "astro-ph.EP" ]	Dust coagulation in a protoplanetary disk is the first step of planetesimal formation. However, a pathway from dust aggregates to planetesimals remains unclear. Both numerical simulations and laboratory experiments have suggested the importance of dust structure in planetesimal formation, but it is not well constrained by observations. We study how dust structure and porosity alters polarimetric images at millimeter wavelength by performing 3D radiative transfer simulations. Aggregates with different porosity and fractal dimension are considered. As a result, we find that dust aggregates with lower porosity and/or higher fractal dimension are favorable to explain observed millimeter-wave scattering polarization of disks. Aggregates with extremely high porosity fail to explain the observations. In addition, we also show that particles with moderate porosity show weak wavelength dependence of scattering polarization, indicating that multi-wavelength polarimetry is useful to constrain dust porosity. Finally, we discuss implications for dust evolution and planetesimal formation in disks.	astro-ph.EP	astro-ph	Draft version September 20, 2019 Typeset using LATEX twocolumn style in AASTeX62 9 1 0 2 p e S 9 1 . ] P E h p - o r t s a [ 2 v 9 8 1 0 0 . 7 0 9 1 : v i X r a Unveiling Dust Aggregate Structure in Protoplanetary Disks by Millimeter-wave Scattering Polarization Ryo Tazaki,1 Hidekazu Tanaka,1 Akimasa Kataoka,2 Satoshi Okuzumi,3 and Takayuki Muto4 -- 1Astronomical Institute, Graduate School of Science Tohoku University, 6-3 Aramaki, Aoba-ku, Sendai 980-8578, Japan 2National Astronomical Observatory of Japan, Mitaka, Tokyo 181-8588, Japan 3Department of Earth and Planetary Sciences, Tokyo Institute of Technology, Meguro-ku, Tokyo, 152-8551, Japan 4Division of Liberal Arts, Kogakuin University, 1-24-2 Nishi-Shinjuku, Shinjuku-ku, Tokyo 163-8677, Japan (Accepted September 17, 2019) ABSTRACT Dust coagulation in a protoplanetary disk is the first step of planetesimal formation. However, the pathway from dust aggregates to planetesimals remains unclear. Both numerical simulations and labo- ratory experiments have suggested the importance of dust structure in planetesimal formation, but it is not well constrained by observations. We study how the dust structure and porosity alters polarimet- ric images at millimeter wavelength by performing 3D radiative transfer simulations. Aggregates with different porosity and fractal dimension are considered. As a result, we find that dust aggregates with lower porosity and/or higher fractal dimension are favorable to explain the observed millimeter-wave scattering polarization of disks. Although we cannot rule out the presence of aggregates with extremely high porosity, a population of dust particles with relatively compact structure is at least necessary to explain polarized-scattered waves. In addition, we show that particles with moderate porosity show weak wavelength dependence of scattering polarization, indicating that multi-wavelength polarimetry is useful to constrain dust porosity. Finally, we discuss implications for dust evolution and planetesimal formation in disks. 1. INTRODUCTION Planetesimal formation, or the first step of planet for- mation, begins with coagulation of micron-sized dust particles in protoplanetary disks. Initially, micron-sized dust particles coagulate to form fluffy dust aggregates whose fractal dimension is about 2 (Weidenschilling & Cuzzi 1993; Ossenkopf 1993; Wurm & Blum 1998; Kempf et al. 1999; Krause & Blum 2004). However, it is still a matter of debate how fluffy dust aggregates grow to form planetesimals in disks. One possibility is that they can maintain their fluffy structure during growth, and the volume filling factor becomes as low as 10−4. Then, coagulation and subsequent compaction of the aggregates form planetesimals (Okuzumi et al. 2012; Kataoka et al. 2013a). Another possibility is that they are compressed at some initial moment during growth, and the volume filling factor is increased to higher than 0.01 (Zsom et al. 2010; Lorek et al. 2018). These rel- atively compact dust aggregates are thought to evolve Corresponding author: Ryo Tazaki [email protected] into planetesimals via the gravitational collapse of dust clouds led by the streaming instability (Youdin & Good- man 2005; Johansen et al. 2007; Bai & Stone 2010a,b; Dr¸azkowska & Dullemond 2014) or by direct coagu- lation with mass transfer (Windmark et al. 2012a,b). Therefore, constraining dust porosity or structure by observations seems to be a very helpful way to discrimi- nate which planetesimal formation process dominates in disks. Recently, the Atacama Large Millimeter/submillimeter Array (ALMA) opened a new window to observe sub- millimeter-wave polarization of protoplanetary disks (Kataoka et al. 2016b, 2017; Stephens et al. 2017; Cox et al. 2018; Sadavoy et al. 2018; Hull et al. 2018; Lee et al. 2018; Alves et al. 2018; Bacciotti et al. 2018; Girart et al. 2018; Ohashi et al. 2018; Dent et al. 2019; Taka- hashi et al. 2019; Harrison et al. 2019). Several mech- anisms to explain polarized millimeter-wave radiation have been proposed: scattering (Kataoka et al. 2015; Pohl et al. 2016; Yang et al. 2016a,b, 2017; Okuzumi & Tazaki 2019), grain alignment with magnetic field (Cho & Lazarian 2007; Bertrang et al. 2017), alignment with the radiation field (Lazarian & Hoang 2007a; Tazaki 2 Tazaki et al. et al. 2017), the gold alignment (Yang et al. 2019), and mechanical alignment (Lazarian & Hoang 2007b; Kataoka et al. 2019). One possible origin is a scattering scenario. In the scattering model, polarized millimeter-waves are pro- duced by scattering of thermal emission from ambient dust particles. Observed polarization patterns and po- larization fractions are consistent with the predictions of the scattering model (Kataoka et al. 2016b; Stephens et al. 2017; Hull et al. 2018; Lee et al. 2018; Bacciotti et al. 2018; Girart et al. 2018; Ohashi et al. 2018; Dent et al. 2019; Harrison et al. 2019). In addition, the scat- tering model may also explain the wavelength depen- dencies of the observed polarization fraction, such as for HL Tau disk (Stephens et al. 2017) and for DG Tau disk (Bacciotti et al. 2018; Harrison et al. 2019). Interestingly, if scattering occurs in disks, it can con- strain the maximum dust particle radius (Kataoka et al. 2015). Previous studies assume solid spherical parti- cles, and the porosity or structure of dust particles are ignored for the sake of simplicity. However, the light scattering process sensitively depends on dust particle size and structure (Kimura et al. 2006; Shen et al. 2008, 2009; Tazaki et al. 2016; Tazaki & Tanaka 2018; Ysard et al. 2018). Thus, current constraints on dust proper- ties could be affected if dust porosity and structure are taken into account. The role of dust porosity and struc- ture on millimeter-wave scattering polarization is not yet clarified. Therefore, this paper studies how these dust particle properties affect millimeter-wave-scattering po- larization. This paper is organized as follows. Section 2 describes our dust particle models and methods. Section 3 sum- marizes optical properties of dust particles. In Section 4, using the optical properties obtained in Section 3, we perform radiative transfer simulations of disks, and the results are presented. By comparing our results with polarimetric observations by ALMA, we discuss impli- cations for planetesimal formation in Section 5. Section 6 presents a summary. 2. DUST MODEL AND METHOD 2.1. Dust particle structure during growth Prior to presenting our dust particle models in Sec- tion 2.3, we briefly review the expected dust structure in protoplanetary disks. The structure of dust aggregates depends on how they coagulate. In disks, coagulation begins with low- collision velocity: no restructuring of dust aggregates occurs upon impact at this stage. Two limiting cases for aggregation are often used: Ballistic Cluster-Cluster Aggregation (BCCA) and Ballistic Particle-Cluster Ag- gregation (BPCA). Aggregates consist of unit particles (monomer) of the same radius for the sake of simplicity. BCCA is successive collisions between similar-sized aggregates. This type of coagulation is expected to oc- cur if the aggregate-size distribution is narrow. Initial dust coagulation in disks is thought to take place in this manner (Ormel et al. 2007; Okuzumi et al. 2009, 2012). Aggregates formed by collisions of similar-sized aggregates tend to show fractal dimension 1.4 (cid:46) df (cid:46) 2 (Weidenschilling & Cuzzi 1993; Ossenkopf 1993; Wurm & Blum 1998; Kempf et al. 1999; Krause & Blum 2004). Since the volume filling factor f and radius of an aggre- gate a obey f ∝ adf−3, their volume filling factor de- creases with increasing radius. For 100 µm-sized BCCA aggregates consisting 0.1 µm monomers, the volume fill- ing factor is f = 3 × 10−4; thus, they are extremely fluffy. BPCA is successive collisions of a monomer and an aggregate. This type of collision is realized when small and large aggregates coexist in disks, such as due to fragmentation of large aggregates. It has been suggested that fragmentation of dust aggregates seems to be nec- essary to explain disk infrared observations (Dullemond & Dominik 2005). The BPCA aggregates tend to have df ≈ 3, and hence, their volume filling factor is con- stant and is about f = 0.15 (Kozasa et al. 1992). Do- minik et al. (2016) studied hierarchical coagulation (an aggregate of aggregates), where an aggregate consist- ing of some monomers is used as a projectile in BPCA instead of using a monomer particle. Hierarchical coag- ulation produces an approximately 10-times lower filling factor f ≈ 0.01, yet still shows df ≈ 3 at the core of the aggregate. In disks, dust structure might also be affected by some compaction events. For high-speed collisions such that the impact energy is high enough to rearrange aggre- gate structure, but still insufficient to disrupt an aggre- gate, collisional compaction occurs (Dominik & Tielens 1997; Blum & Wurm 2000; Wada et al. 2007, 2008, 2009; Paszun & Dominik 2009). Wada et al. (2008) showed that the fractal dimension of BCCA aggregates can be increased by a factor of up to 2.5 upon such an impact. Laboratory measurements have often observed bounc- ing of two colliding dust aggregates (Blum & Munch 1993; Weidling et al. 2012; Kothe et al. 2013; Brisset et al. 2017). Weidling et al. (2009) suggested that se- quential bouncing collisions gradually increase the vol- ume filling factor up to f ≈ 0.36. However, the con- ditions for bouncing collisions are still being debated (Wada et al. 2011; Seizinger & Kley 2013; Kothe et al. 2013). Kataoka et al. (2013b,a) proposed gas and grav- itational compaction of dust aggregates which increases Unveiling Dust Aggregate Structure in Protoplanetary Disks 3 (cid:112) the volume filling factor. Dust structure under these compressions is characterized by a bi-fractal structure, where df = 3 on the larger scale and 2 on smaller scale. To summarize, dust aggregates in disks might be cat- egorized into two limiting groups: compact dust aggre- gates showing fractal dimension df ≈ 3 with 0.01 (cid:46) f (cid:46) 0.4 and fluffy dust aggregates with df ≈ 2 with f (cid:28) 1. 2.2. Dust composition At (sub-)millimeter wavelengths, disk emission comes mostly from the outer cold regions where water ice con- denses onto dust particles. Thus, we assume that a dust particle is a mixture of silicate, water ice, troilite, and amorphous carbon. Although silicate, troilite, and amorphous carbon are probably separate grain popula- tions, we assume that these components are mixed into a single dust particle for the sake of simplicity. The mass fraction of each component is determined by a recipe described in Min et al. (2011), where we adopt the carbon partition parameter w = 0.5. The derived mass fractions (material density) of silicate, water ice, troilite, and amorphous carbon are 32% (3.30 g cm−3), 45% (0.92 g cm−3), 10% (4.83 g cm−3), and 13% (1.80 g cm−3), respectively. Water ice dominates the mass and volume of dust particles. The resulting mean ma- terial density is 1.48 g cm−3. Optical constants of sili- cate, water ice, troilite, and amorphous carbon are taken from Draine (2003), Warren & Brandt (2008), Henning & Stognienko (1996), and Zubko et al. (1996), respec- tively. Their optical constants are mixed by using the Bruggeman mixing rule (Bruggeman 1935). At λ = 1 mm, the mixed refractive index, m, is m = 2.6 + 0.074i. 2.3. Dust particle models adopted in this study We consider three types of dust structure as illustrated in Figure 1. Basic properties of each dust model are as follows. • Solid spheres: Solid spheres have homogeneous structure, and the filling factor is unity. Since solid spheres are commonly used in previous studies, it is valuable to compare how their optical properties are different from those of lower density dust par- ticles resulting from structure and porosity. The important parameter of this model is its radius. • Fluffy dust aggregates: We define fluffy dust ag- gregates as aggregates with df ≈ 2. We use the characteristic radius ac to describe the aggregate radius (Mukai et al. 1992; Kozasa et al. 1992). We adopt a fractal dimension df = 1.9 and fractal prefactor k0 = 1.03, which are the typical values for BCCA aggregates formed with oblique colli- sions (Tazaki et al. 2016). The fractal prefactor is defined by N = k0(ag/a0)df ; N is the num- ber of monomers and ag = 3/5ac is the radius of gyration. Since we fix the monomer radius as a0 = 0.1 µm, the free parameter of the model is the aggregate radius ac only. The volume filling factor is given by f = 3 × 10−4(ac/100 µm)−1.1. In this study, we only consider ac > 4 µm so that the volume filling factor of fluffy aggregates is al- ways less than 0.01. • Compact dust aggregates: Compact dust aggre- gates are defined as aggregates with df = 3 and 0.01 ≤ f < 1 in this paper. This model is intended to mimic BPCA aggregates (with/without hierar- chical effects). Aggregate radius and the volume filling factor are the important parameters. We use the term "compact" because these grains are relatively compact compared to the fluffy dust ag- gregate model with f < 0.01. In this paper, we use the term dust particles, or simply dust, in more general contexts, i.e., when we intend to mention more general cases rather than to indicate a specific model given above. We mainly focus on dust particle radii from sub- millimeter (∼ 100 µm) to centimeter size (∼ 104 µm). Since this study aims to understand the scattering po- larization of disks at (sub)millimeter wavelengths, dust particle radii should be comparable to or larger than the observing wavelength in order to attain high scattering albedo. 2.4. Computation of optical properties Computing optical properties of compact/fluffy dust aggregates is not an easy task, whereas those of solid spheres can be readily computed by the Mie theory (Bohren & Huffman 1983). Powerful numerical tech- niques for directly solving their optical properties have been developed, such as the T-Matrix Method (Mack- owski & Mishchenko 1996) and the Discrete Dipole Approximation (Draine & Flatau 1994). However, millimeter-sized BPCA aggregates of 0.1 µm monomers contain about 1011 monomers, and solving for electro- magnetic interactions between every pair of monomers is very time-consuming. Therefore, computing those properties with numerical techniques is not a realistic choice given current computer capabilities. In order to reduce computational cost, approximate methods are useful. For fluffy dust aggregates, we can approximate their optical properties by using the mod- ified mean field theory (Tazaki et al. 2016; Tazaki & Tanaka 2018). On the other hand, the approximation to the optical properties of compact aggregates is some- what more difficult because multiple scattering comes 4 Tazaki et al. of amax and amin, respectively. n(a)da represents the number of dust particles in a size range [a, a + da]. We set p = −3.5 and amin = 0.01 µm, although our results are insensitive to amin as long as amin (cid:28) amax. 3. MILLIMETER-WAVE SCATTERING PROPERTIES OF DUST PARTICLES We present optical properties of solid spheres, com- pact dust aggregates and fluffy dust aggregates at mil- limeter wavelength. It is useful to introduce a quantity to assess an abil- ity of dust particles for producing millimeter-wave scat- tering polarization. Dust particles efficiently produce a polarized scattered light when their single scattering albedo ω is high and the degree of linear polarization P for 90◦ scattered waves is high. Therefore, the product P ω can be used as a diagnostic of efficient scattering polarization (Kataoka et al. 2015). However, from the point of view of radiative transfer, the single scatter- ing albedo ω becomes a bad approximation to apparent scattering efficiency because forward scattering domi- nates the value of the albedo when a particle is larger than the wavelength (e.g., Mulders et al. 2013; Min et al. 2016; Tazaki et al. 2019). Since forward scattering does not seem to change the direction of incident light upon scattering, it is effectively not scattering. In addition, forward scattered light is unpolarized. Thus, it is not so important for polarized radiative transfer. A simple way to mimic apparent scattering efficiency is to define the effective scattering opacity κeff sca = (1− g)κsca, where κsca is the scattering opacity and g is the asymmetry parameter (e.g., Birnstiel et al. 2018). The asymme- try parameter diminishes for isotropic scattering and is unity for perfect forward scattering. Thus, if scattered light is highly concentrated into the forward direction (g ∼ 1), κeff sca ≈ 0, indicating that apparently no scat- tering occurs. By using the effective scattering opacity, we can also define the effective single scattering albedo ωeff = κeff sca + κabs). Therefore, we use P ωeff in- stead of P ω in order to quantify millimeter-wave scat- tering polarization. sca/(κeff 3.1. Solid spheres and Compact dust aggregates Figure 2 shows P ω (left panel) andP ωeff (right panel) as a function of the maximum dust radius amax for solid spheres (f = 1) and compact dust aggregates (f = 0.1, 0.01) at wavelength λ = 1 mm. In order to understand behavior of P ωeff against amax, we start to discuss P ω, and then P ωeff for each dust model. It is also useful to define the size parameter x = 2πamax/λ. Since we adopt λ = 1 mm, the size parameter of unity corresponds to amax = 160 µm. Figure 1. Cartoon illustrating dust particle models used in this paper. Dust models are characterized by the volume filling factor f and fractal dimension df . Solid spheres have df = 3 and f = 1, whereas compact dust aggregates have df = 3 and f < 1. We also consider fluffy dust aggregates whose fractal dimension is df ≈ 2 consisting of unit particles (monomers) of radius a0 = 0.1 µm. The volume filling factor of fluffy aggregates is size dependent and it is typically much smaller than 0.01. into play. Hence, as our current best, conservative esti- mate, we adopt the effective medium theory (EMT). It was recently found that the optical properties of com- pact dust aggregates, in particular for scattering opac- ity, could be (not perfectly but roughly) approximated by EMT (see Figure 6 in Tazaki & Tanaka 2018). This is partly because compact aggregates have df = 3, and hence, their scattering behaviors, such as interference of scattered waves, have some similarities to those of the Mie theory, which also deals with spherical particles (df = 3). Thus, we anticipate that qualitatively sim- ilar results can be obtained even if we use EMT. We adopt the Maxwell Garnett mixing rule (Maxwell Gar- nett 1904) to obtain an effective refractive index, meff , of a mixture of vacuum (the matrix component) and dust composition (the inclusion component) described in Section 2.2. Since the Maxwell Garnett mixing rule works well when the matrix component dominates, we only consider f ≤ 0.1 (Kolokolova & Gustafson 2001). For solid spheres and compact aggregates, optical properties are averaged over the dust size distribution in order to suppress strong resonances of scattering properties. We assume a dust size distribution obey- ing n(a)da ∝ apda with maximum and minimum radii Fluffy dust aggregateaca0=0.1μmMonomerSolid spheredf=3,f=1Compact dust aggregatedf=3,0.01≤f<1df≈2,f<0.01Unveiling Dust Aggregate Structure in Protoplanetary Disks 5 Figure 2. P ω and P ωeff for solid spheres (f = 1) and compact dust aggregates (f = 0.1, 0.01) at λ = 1 mm. (Left) Solid lines represent P ω. Red, blue, and green lines show the results for the volume filling factor of f = 1, 0.1, and 0.01, respectively. Dotted and dot-dashed lines indicate the single scattering albedo ω and the degree of linear polarization P at scattering angle of 90◦, respectively. Filled circles indicate P ω at the position of amaxf = 160 µm. (Right) Solid and dashed lines indicate results for P ωeff and P ω, respectively. Filled circles indicate P ωeff at amaxf = 160 µm. The P ωeff -values at x > 1 is significantly attenuated by considering the effective single scattering albedo. For solid spheres (f = 1), the maximum value of P ω appears at around x ≈ 1. The single scattering albedo ω is usually much less than unity when x (cid:28) 1 and becomes of order unity when x (cid:38) 1. On the other hand, the de- gree of linear polarization P is high as long as x (cid:28) 1, but it suddenly decreases for x (cid:38) 1. As a result, P ω is maximized at x ≈ 1. P ωeff for solid spheres is almost the same as P ω because, as the maximum dust radius increases, the degree of polarization drops much faster than the term 1 − g. Thus, considering P ωeff instead of P ω does not significantly change the previous inter- pretation by Kataoka et al. (2015), where solid spherical particles are assumed. Next, we focus on compact dust aggregates (f = 0.1 and 0.01). First of all, decreasing the volume filling fac- tor leads to a reduction of the single scattering albedo at x ≤ 1 because the scattering opacity is proportional to the aggregate mass, whereas the absorption opacity is not. A notable impact of decreasing f is that the degree of polarization becomes high even if x > 1. This is due to the fact that aggregates with f (cid:28) 1 have an effective re- fractive index close to that of vacuum, and are regarded as optically thin particles. Since multiple scattering is suppressed inside optically thin particles, a high degree of polarization is obtained. P ω increases with increasing amax until xmeff − 1 exceeds unity at which aggregates become optically thick. Since xmeff − 1 is proportional to amaxf (Kataoka et al. 2014), the maximum value of P ω approximately occurs at amaxf ≈ λ/2π ≈ 160 µm. For example, when f = 0.1, P ω is maximized at ap- proximately amax = 1.6 mm. The maximum P ω-value is larger for a lower filling factor because lower filling fac- tor makes aggregates less absorbing and also suppresses depolarization due to multiple scattering. Unlike the case of solid spheres, the effective single scattering albedo plays a significant role for compact aggregates. As shown in Figure 2 (right), the maximum value of P ωeff is significantly attenuated at x > 1, in particular for the case of f = 0.01. As a result, as the volume filling factor decreases, the maximum value of P ωeff decreases. This suggests that compact aggregates with lower filling factor pro- duce fainter polarized-scattered light than solid spheres. We will confirm this by performing radiative transfer simulations in Section 4. 3.2. Fluffy dust aggregates The scattering properties of fluffy dust aggregates are significantly different from those of spherical particles. We use the MMF theory to compute optical properties of fluffy aggregates. In this section, we use df = 2.0 and k0 = 1.0 in order to attain clear physical insights into the optical properties of fluffy dust aggregates. First of all, the degree of linear polarization of fluffy dust aggregates is as high as 100% because multiple scat- tering is suppressed when x0 (cid:28) 1 and df ≤ 2 (Tazaki et al. 2016), where x0 = 2πa0/λ is the size parameter of the monomer particle. Here, the wavelength of interest is (sub-)millimeter wavelength and the monomer radius is of sub-micron size, and therefore, this condition is satisfied. Even if ac (cid:29) λ/2π, the highest degree of po- larization remains 100%. Thus, we obtain P ωeff = ωeff . 00.20.40.60.8110−310−210−1100101102Pωamax(cm)f=1.00f=0.10f=0.0100.20.40.60.8110−310−210−1100101102Pωeff,Pωamax(cm)f=1.00f=0.10f=0.016 Tazaki et al. In Figure 3, we show the opacities of fluffy dust ag- gregates as a function of the characteristic radius ac. The absorption opacity does not depend on ac because fluffy aggregates with different ac have the same mass- to-area ratio, and the absorption opacity remains the same (Kataoka et al. 2014) 1. At ac (cid:46) λ/2π, the scat- tering opacity increases with increasing aggregate radius (mass). On the other hand, once the aggregate radius exceeds λ/2π, the effective scattering opacity saturates. An analytical solution to the upper limit on the effective scattering opacity can be found under the conditions of x0 (cid:28) 1 and df = 2 (see Appendix A for derivation). The analytical solution to the effective scattering opacity of fluffy dust aggregates at ac > λ/2π is κeff sca = κsca,mono 2x2 0 , (1) where κsca,mono is the scattering opacity of the single monomer. It is clear from Equation (1) that the effective scattering opacity of fluffy aggregates with df = 2 does not depend on the aggregate radius or the number of monomers. The physical interpretation of Equation (1) can be captured when the differential scattering cross-section per unit mass Z11 is plotted as a function of scatter- ing angle (Figure 4). For an aggregate smaller than the wavelength, Rayleigh scattering occurs. This means that scattering is coherent for all scattering angles. Once the aggregate radius becomes larger than the wave- length, Z11 saturates at intermediate and backward scattering angles. This is because scattered waves from a pair of monomers separated by a distance larger than (approximately) the wavelength are out-of-phase. On the other hand, scattered waves from particles separated by smaller distance can be in phase for such scattering angles. Therefore, scattered light at these scattering an- gles is dominated by coherent scattered light from the small-scale structure of the aggregate rather than the large-scale structure. As a consequence, Z11 becomes irrelevant to ac at these scattering angles, and then κeff sca also becomes independent on how large the aggregate is. The above explanation is also illustrated schematically in Figure 7 in Tazaki et al. (2019). By using Equation (1) and formulae for opacities in the Rayleigh limit (Bohren & Huffman 1983), the effec- tive single scattering albedo of fluffy aggregates, whose 1 Strictly speaking, in the Rayleigh limit, the absorption opacity depends on the number of monomers (Stognienko et al. 1995; Hen- ning & Stognienko 1996; Tazaki & Tanaka 2018) due to monomer- monomer interaction. However, this effect is saturated when the number of monomers exceeds ∼ 100, and then the size dependence becomes negligible for further large aggregates. Figure 3. Opacities of fluffy dust aggregates with fractal dimension df = 2 and fractal prefactor k0 = 1.0 at λ = 1 mm. Red and blue lines represent the effective scattering opacity and the absorption opacity, respectively. The dashed line indicates the scattering opacity. The gray horizontal line is an analytical upper limit (Equation 1). characteristic radius is larger than the wavelength, is (cid:26) m2 − 1 m2 + 2 (cid:27)(cid:30)(cid:12)(cid:12)(cid:12)(cid:12) m2 − 1 m2 + 2 (cid:12)(cid:12)(cid:12)(cid:12)2 . ωeff = x0 x0 + 3R , R = Im (2) By substituting values into Equation (2), we obtain P ωeff = 6 × 10−3, where we have assumed m = 2.6 + 0.074i and P = 1. As a result, it is found that fluffy aggregates are inefficient scatterers of millimeter-wave radiation. Even if the monomer radius is a few microns, the effective albedo is still (cid:46) 0.1. Therefore, fluffy aggre- gates of (sub-)micron-sized monomers are not likely to contribute millimeter-wave scattering in protoplanetary disks. Finally, we address how a different choice of frac- tal dimension affects scattering properties, since the fractal dimension of dust aggregates in disks might be 2 (cid:46) df (cid:46) 3. Although the detailed angular dependence of scattering matrix elements is not easy to compute, the qualitative wavelength dependence of opacities might be obtained by using the MMF theory (Tazaki & Tanaka 2018). In Figure 5, we show the effective scattering albedo of dust aggregates with various fractal dimen- sion having ac = 160 µm and a0 = 0.1 µm. Since the MMF theory is incapable of predicting the degree of po- larization P for df > 2 due to the importance of multiple scattering, we only show ωeff in Figure 5. At λ > 2πac = 1 mm (Rayleigh limit), the effective albedo increases with fractal dimension. Since we have fixed both ac and a0, increasing fractal dimension results in increasing the aggregate mass. Therefore, the effec- 10−410−310−210−110010110−310−210−1100Eq.(1)κ(abs,sca),κeffsca(cm2g−1)characteristicradiusofaggregate[cm]κscaκeffscaκabsUnveiling Dust Aggregate Structure in Protoplanetary Disks 7 tions (Mulders et al. 2013; Min et al. 2016; Tazaki et al. 2019). As a result, at (sub-)millimeter wavelengths, increas- ing the fractal dimension of aggregates gradually in- creases the effective scattering albedo. Thus, aggregates with higher fractal dimension are more likely to scatter at millimeter wavelength. 4. RADIATIVE TRANSFER SIMULATIONS In Section 3 we discussed the optical properties of solid spheres and compact/fluffy dust aggregates. In this sec- tion we perform radiative transfer simulations of disks containing these dust particles. 4.1. Model and Method In order to simulate millimeter-wave scattering polar- ization of a protoplanetary disk, we use a publicly avail- able 3D Monte Carlo radiative transfer code RADMC- 3D (Dullemond et al. 2012). We assume a vertically isothermal disk, and the radial temperature profile ap- proximately obeys T (r) = 86 K(r/10 au)−0.5, where T is the dust temperature and r is the distance from the cen- tral star. The inner and outer disk radii are 10 au and 100 au, respectively, and the dust surface density is set as Σd = 0.157 g cm−2 (r/10 au)−1, leading to 10−4M(cid:12) total disk dust mass. We assume the dust model is the same everywhere within the disk. In addition, we ignore polarized thermal emission from aligned grains. The number of photon packets used in Monte Carlo scatter- ing simulations is 109. Since our primary focus is on how dust properties affect polarization, we fix both the disk model and disk inclination angle. Since moderate disk inclination angles are favored for scattering polarization (Yang et al. 2016a), we assume the inclination angle is 45◦. The imaging wavelength is set as λ = 1 mm. Dust models used in simulations are solid spheres and compact/fluffy dust aggregates. In Table 1, we summa- rize the optical properties used in simulations for solid spheres and compact dust aggregates (amaxf = 160 µm with f = 1, 0.1, 0.01) as well as for fluffy dust aggre- gates (df = 1.9, k0 = 1.03, ac = 160 µm). It is worth reminding the reader that the value of 160 µm is chosen so that the dust particle radius is equal to λ/2π, where λ = 1 mm. We will also discuss how different choices of amax affect the results. The quantity amaxf is a useful quantity to charac- terize particle properties because the mass-to-area ratio of dust particles is proportional to af , and thus, both absorption and dynamical properties are characterized by af (Kataoka et al. 2014). Indeed, as pointed out by Kataoka et al. (2014), the mass absorption opacity of aggregates with the same amaxf -value become very similar (Table 1). Figure 4. Differential scattering cross-section per unit mass of fluffy dust aggregates with df = 2 and k0 = 1.0 at λ = 1 mm. As the aggregate radius increases, the differential cross- section saturates at intermediate and backward scattering angles as a natural consequence of single scattering. Figure 5. The effective single scattering albedo ωeff of dust aggregates with ac = 160 µm for various df estimated by the MMF theory. Dashed line represent analytic solution applicable when λ < 2πac and df = 2 (Equation 2). The vertical line indicates the position of λ = 2πac. Refractive index is set as m = 2.6 + 0.074i. tive albedo increases with df . At λ (cid:28) 2πac, the effec- tive albedo for higher fractal dimension becomes smaller due to forward scattering. Figure 5 predicts that aggre- gates with higher fractal dimension produce faint and reddish scattered light in the short wavelength domain, e.g., infrared wavelengths, whereas those with df = 2 produce bright and blue scattered light. This tendency has already been confirmed by rigorous computations of optical properties as well as radiative transfer simula- 10−410−310−210−1100101102103104105020406080100120140160180Z11[cm2/g/str]Scatteringangle[deg]ac=10µmac=100µmac=1mmac=1cmac=10cm10−410−310−210−1100101101102103104Eq.(2)2πac=λωeffλ(µm)df=3.0df=2.8df=2.6df=2.4df=2.2df=2.08 Dust model Tazaki et al. Table 1. Optical Properties of Dust models at Wavelength λ = 1 mm Solid Spheres (amax = 160 µm) Compact aggregates (amax = 1.6 mm, f = 0.1) Compact aggregates (amax = 1.6 cm, f = 0.01) Fluffy aggregate (ac = 160 µm) sca (cm2 g−1) κabs (cm2 g−1) κeff 5.54 2.09 0.11 4.29 × 10−3 3.37 2.17 1.95 1.94 g 0.30 0.89 0.99 0.14 ωeff 0.62 0.49 0.05 2.21 × 10−3 P (θ = 90◦) P ωeff 0.51 0.82 0.45 0.92 0.05 1.00 2.21 × 10−3 1.00 4.2. Results of radiative transfer simulations Figure 6 shows a polarization map of a disk con- taining solid spheres (f = 1) and compact aggregates (f = 0.1, 0.01). In each image, polarization angles (E- vector direction) are indicated by bars whose length is proportional to the polarization fraction. As the volume filling factor decreases, the polarized intensity from the disk diminishes. This can be inter- preted as a consequence of low effective scattering albedo as shown in Figure 2. For the cases of f = 1 and 0.1, polarization angles (E-vector direction) tend to be ori- ented parallel to the disk minor axis at the inner regions of disks. This is characteristic of scattering polarization from an optically thin disk (Kataoka et al. 2016a; Yang et al. 2016a). For f = 0.1, near-and-far side asymmetry of the po- larization pattern can be seen, whereas, for f = 1, the pattern is symmetric. For f = 0.1, the polarization ori- entation at the disk near side tends to be orientated in the azimuthal direction, whereas that of the far-side is orientated parallel to the minor axis. The dip in po- larized intensity at the near side is the location where the polarization angle changes its direction from radial at the inner disk to azimuthal at the outer disk. This asymmetric polarization pattern is caused by anisotropic scattering. For f = 1, the assumed particle radius is comparable to λ/2π, and hence, scattering is close to isotropic. However, for f = 0.1, the assumed particle radius (amax = 1.6 mm for the f = 0.1 model) is larger than λ/2π ≈ 160 µm, and thus, forward scattering oc- curs. Since forward scattered light amplifies a scattered- light component with azimuthal polarization at the disk near-side, near-far side asymmetry of the polarization pattern appears. Next, we study the dust particle radius dependence of the average polarization fraction obtained by radiative transfer simulations. Figure 7 shows the average polar- ization fraction (cid:104)Pdisk(cid:105), which is a ratio of the polarized flux to the total flux for the entire disk, for various amax. We compare the particle radius dependence of P ωeff as well as P ω, where we use (cid:104)Pdisk(cid:105) = CP ωeff to compare simulation results; the C-value is a numerical factor cal- ibrated at amax = 160 µm for each model. As shown in Figure 7, the solid sphere model (f = 1) shows strong dependence on amax. This means that a high polarization fraction occurs at the vicinity of amax ≈ λ/2π. On the other hand, for the case of f = 0.1, dependence on dust radius becomes much weaker. Therefore, when dust porosity is taken into account, scattering polarization can be detected for a wider range of dust particle radius compared to the solid sphere model, although if the porosity is too high (f = 0.01) insufficient polarized intensity is produced. At amax = 160 µm, decreasing the volume filling factor makes the polarization fraction of scattered light small as a consequence of lower albedo for lower filling factor particles. This is consistent with the results in Figure 2. As shown in Figure 7, P ωeff can reproduce the over- all dependence of polarization fraction on amax observed in simulations. It is also shown that P ω fails to repro- duce the dust particle radius dependence, indicating the importance of considering ωeff instead of ω. As a result, we conclude that dust aggregates with lower filling factor produce smaller polarization fractions due to the effective reduction of their scattering opacity caused by strong forward scattering. In other words, higher polarization fraction is more likely to be produced by relatively compact dust particles (f (cid:38) 0.1). Next, we perform radiative transfer simulations with fluffy dust aggregates with ac = 160 µm, and the result is shown in Figure 8. It is found that fluffy aggregates show very faint polarized intensity. This is mainly due to low scattering albedo as can be seen in Table 1. What happens when the aggregate radius is further increased? We perform additional radiative transfer simulations for fluffy aggregates with ac = 1.6 mm and 1.6 cm, and we find that these larger aggregates give rise to almost the same results as found with ac = 160 µm. This is due to saturation of the effective scattering opacity as can be seen in Figure 3. As a result, it is found that a fluffy ag- gregate of sub-micron monomers with any characteristic radius is unlikely to produce millimeter-wave scattering polarization. As a result, we conclude that higher fractal dimension and higher filling factor are more favorable to producing millimeter-wave scattering polarization. 5. DISCUSSION Unveiling Dust Aggregate Structure in Protoplanetary Disks 9 Figure 6. Polarized intensity [mJy/beam] at λ = 1 mm for solid spheres (f = 1) and compact dust aggregates (f = 0.1, 0.01) with amaxf = 160 µm. From left to right panel, the volume filling factor is decreased. Red bars and their length represent polarization orientations and polarization fraction, where the reference length for 1% polarization fraction is shown in right bottom in each panel. Total intensity contours in each panel are shown for [3, 12, 25, 35, 50, 100, 200, 300] × 0.03 [mJy/beam]. FWHM of the beam size is 0".1 as shown in the left bottom circle. Inclination angle of the disk is 45 degrees and the distance between the observer and the disk is assumed to be 100 pc. Bottom side of each image corresponds to the near side of the disk. Figure 7. Disk polarization fraction (cid:104)Pdisk(cid:105) versus amax for f = 1 (red), 0.1 (blue), and 0.01 (green). The square, triangle, and circle symbols show the values of (cid:104)Pdisk(cid:105) directly obtained from radiative transfer simulations, while the solid and dot-dashed lines show the estimates (cid:104)Pdisk(cid:105) = CP ωeff and (cid:104)Pdisk(cid:105) = CP ω, respectively. By comparing our results (Sections 3 and 4) and ALMA polarimetric observations, Section 5.1 discusses constraints on dust size and porosity in protoplanetary disks. We also discuss the wavelength dependence of scattering polarization and implications for future multi- wavelength polarimetric observations. In Section 5.2, we discuss implications for planetesimal formation. In Sec- tion 5.3, we also compare our results with the cometary dust particles of 67P/Churyumov-Gerasimenko. Figure 8. Same as Figure 6, but for fluffy dust aggregates with ac = 160 µm. Polarized intensity is much fainter than that of solid sphere and compact aggregate models. 5.1. Comparison with disk observations 5.1.1. Dust structure and porosity estimate Recent polarimetric observations of disks by ALMA have shown that scattering polarization can be com- monly seen in various disks. In particular, observed scattering-like polarization (polarization angles parallel to the disk minor axis) often shows a polarization frac- tion of about 1% (Kataoka et al. 2016b; Stephens et al. 2017; Hull et al. 2018; Lee et al. 2018; Bacciotti et al. 2018; Girart et al. 2018; Ohashi et al. 2018; Dent et al. 2019; Harrison et al. 2019). 0.010.1110−310−210−1100⟨Pdisk⟩(%)amax(cm)f=1.00f=0.10f=0.0110 Tazaki et al. In Section 4, we found that compact aggregates with f = 0.01 and fluffy aggregates (f (cid:28) 0.01) do not produce polarized-scattered waves at millimeter wave- length. In other words, aggregates with extremely high porosity (f (cid:46) 0.01 or porosity higher than 99%) do not explain observations whatever the value of their frac- tal dimension is. Dust particles with higher fractal di- mension (df (cid:38) 2) and/or higher volume filling factor (f (cid:38) 0.1) seem to be necessary to explain the observed scattering polarization. In order to constrain f and df in more detail, further radiative transfer modeling for each object is necessary, although this is beyond the scope of this paper. It is important to keep in mind that we cannot rule out the presence of dust particles with lower fractal di- mension (df (cid:46) 2) and/or lower volume filling factor (f (cid:46) 0.1) in disks because these particles are just invis- ible in millimeter-wave scattering. Mixed populations of fluffy aggregates and compact aggregates might be another solution to explain observations. However, in this case, increasing the mass abundance of fluffy ag- gregates with respect to compact aggregates will reduce the polarization fraction because fluffy aggregates only contribute to the total flux via thermal emission. Hence, a large mass abundance of fluffy dust aggregates might not be favored. We present a rule-of-thumb estimate of the upper limit of the mass abundance of fluffy dust aggregates. Suppose a disk consists of two dust populations: solid spheres and fluffy dust aggregates, and denote Ms and Mf by the total mass of solid spheres and fluffy dust aggregates in the disk. If their absorption opacities are similar and the disk is optically thin, the polar- ized intensity is proportional to Ms, while the total intensity is proportional to Mf + Ms. The polariza- tion fraction of the disk, Pdisk, may be approximated by Pdisk ≈ P0Ms/(Ms + Mf ), where P0 is the polarization fraction of the disk consisting of solid spheres only. In our simulation, P0 (cid:39) 2.4 % (Figure 6). Since observed polarization fraction is about 1%, the mass abundance of fluffy aggregates should be Mf /(Mf + Ms) (cid:46) 0.6. Al- though this upper limit depends on a disk model used, such as temperature structure and optical depth, more detailed analysis is necessary. However, this is beyond the scope of this paper. In any case, we at least need a population of dust particles with relatively compact structure to explain polarized-scattered waves. The presence of fluffy aggregates could be tested by investigating polarized thermal emission from aligned grains in disks. In disks, in addition to the scattering po- larization studied in this paper, polarized thermal emis- sion due to grain alignment has also been proposed (Cho & Lazarian 2007; Tazaki et al. 2017; Bertrang et al. 2017; Yang et al. 2019; Kataoka et al. 2019). Recently, Kirch- schlager et al. (2019) investigated the intrinsic polariza- tion properties of porous dust particles. They found that a porosity higher than 70% is not favorable to produce polarized thermal emission. Their conclusion is similar to those obtained in this study. 5.1.2. Dust size estimate Hereafter, we assume that dust particles in disks have df = 3, since higher fractal dimension is favored. For dust particles with df = 3, the polarization fraction de- pends on dust radius as we showed in Figure 7. Con- versely, the observed scattering polarization fraction should contain information from which we can constrain the dust radius in disks. Since previous studies have relied on the assumption of f = 1 (solid sphere), scattering polarization occurs effi- ciently at the vicinity of amax ≈ λ/2π (Figure 7). Thus, the derived maximum dust radius is about a few times 100 µm because the observing wavelengths of ALMA are in the (sub-)millimeter domain. However, the maximum dust radius is thought to depend on the physical prop- erties of disks, such as age, gas and dust surface density (Birnstiel et al. 2012). Hence, if the derived dust con- straints are true, we need to explain how amax of various disks is fine-tuned at about this radius. By considering particle porosity, this fine-tuning issue may be relaxed. Figure 9 shows P ωeff as a function of amaxf for various values of the volume filling fac- tor. For the case of f = 1, P ωeff is a sharp function of amax, where the peak appears at amax = λ/2π. As the volume filling factor decreases, the width of P ωeff broadens, while the peak value is attenuated (see also Figure 7). Therefore, for particles with moderate poros- ity (0.1 (cid:46) f (cid:46) 1), scattering polarization can be ex- pected for a wider range of amax, while for extremely high porosity (f (cid:46) 0.01) it is hard to produce scattering polarization. Hence, moderate porosity dust particles have a role to relax the tight constraint on amax. There is another issue concerning dust size con- straints, that is, the inconsistency between opacity index and scattering polarization (Kataoka et al. 2016a; Yang et al. 2016a). If the disk is optically thin at observ- ing wavelengths, the spectral slope of the observed flux density depends on the opacity index β, where β is the spectral slope of the absorption opacity, i.e., κabs ∝ λ−β. In Figure 9, the opacity index is shown as a function of amaxf , where the opacity index is defined at wavelengths between 870 µm and 1.3 mm. As shown in Figure 9, the opacity index at which scattering polarization can be anticipated is typically about β (cid:38) 1. However, the Unveiling Dust Aggregate Structure in Protoplanetary Disks 11 models seem to be favored as the origin of scattering polarization rather than the solid sphere model. Stephens et al. (2017) found that the polarization pattern of HL Tau disk is wavelength dependent. At λ = 0.87 mm, the polarization pattern of HL Tau's disk is consistent with a scattering origin, whereas at λ = 3 mm, the polarization pattern becomes circular symmetric, indicating that another origin for the polar- ization is important, such as grain alignment (Tazaki et al. 2017; Kataoka et al. 2017; Yang et al. 2019). DG Tau's disk also shows a wavelength dependent polariza- tion pattern between λ = 0.87 mm (Bacciotti et al. 2018) and λ = 3 mm (Harrison et al. 2019), where polariza- tion at λ = 0.87 mm is partially explained by scattering polarization and polarization at λ = 3 mm is perhaps caused by grain alignment (Harrison et al. 2019). The observed wavelength dependence of disks around HL Tau and DG Tau might be reproduced either with or without dust porosity (e.g., see the blue solid line and the dashed line in Figure 10). Although wave- length dependences between ALMA bands is similar to each other, they have completely different wavelength dependence at far-infrared wavelengths. Therefore, far- infrared wavelength polarimetry, such as by the SOFIA telescope and also the future SPICA telescope, is essen- tial to distinguish between these models. The SOFIA and SPICA telescopes are insufficient to spatially re- solve most disks, and hence, axisymmetric polarization patterns will be canceled out. However, scattering polar- ization tends to have a unidirectional polarization pat- tern for inclined disks, and therefore, we may expect to detect polarized waves even if the polarization pattern is integrated over the entire disk. It is worth keeping in mind that at far-infrared wavelengths, disks may be- come optically thick. Yang et al. (2017) pointed out that large optical depths of disks may reduce the scattering polarization fraction. 5.2. Implications for planetesimal formation Porosity evolution of dust aggregates in disks has been simulated by a number of authors (Ormel et al. 2007; Okuzumi et al. 2009; Zsom et al. 2010; Okuzumi et al. 2012; Kataoka et al. 2013a; Krijt et al. 2015, 2016; Lorek et al. 2018). However, how dust aggregates grow to form planetesimals is a matter of debate. Here, we discuss implications for dust evolution and planetesimal forma- tion by assuming that particles responsible for polarized scattered light dominate the entire population in disks. A possible scenario to form a planetesimal is direct coagulation via fluffy dust aggregates. Dust aggregates consisting of sub-micron-sized icy monomers are found to be very sticky, and they are resistant to compaction Figure 9. Comparison of scattering polarization efficiency P ωeff (top) and opacity index β (bottom) as a function of amaxf . Wavelength is set as 1 mm. Different lines represent different volume filling factor f = 1, 0.1, 0.05, and 0.01. opacity index of disks is typically equal to or less than unity (Testi et al. 2014), implying the presence of mm to cm-sized dust particles in the disk (Draine 2006). Re- cently, Dent et al. (2019) clearly showed that scattering polarization is detected at disk regions where β (cid:46) 1. This inconsistency also seems to occur for another disk (e.g., Stephens et al. 2017; Hull et al. 2018). Dust porosity has been suggested as a solution for this inconsistency. However, the opacity index of mod- erately porous particles is almost at the Rayleigh-limit value, that is, β ≈ βISM ≈ 1.6 (Planck Collaboration et al. 2014). Thus, this inconsistency may not be simply solved by considering particle porosity. Perhaps, other parameters, such as the functional shape of the dust size distribution and the optical depth of the disks, may be important to solve this issue. Recently, it has been pointed out that the spectral index could be affected by scattering if the disk is optically thick (Liu 2019; Zhu et al. 2019). 5.1.3. Wavelength dependence The wavelength dependence of scattering polarization is another important point. In Figure 10, we show P ωeff as a function of wavelength. For the case of f = 1, scattering polarization efficiently occurs at λ ≈ 2πamax (Kataoka et al. 2015). If dust porosity is taken into ac- count, the wavelength dependence becomes weaker than that predicted from the solid sphere model as already expected from Figure 7. Therefore, the wavelength de- pendence of scattering polarization is useful to constrain dust porosity. For example, if we detect scattering po- larization both at Band 3 and 7 of ALMA, porous dust 00.10.20.30.40.50.6Pωeff01230.0010.010.11βBand6−7amaxf(cm)f=1f=0.1f=0.05f=0.0112 Tazaki et al. tions; however, the onset of bouncing collisions is still a matter of debate (Wada et al. 2011; Seizinger & Kley 2013; Kothe et al. 2013; Brisset et al. 2017), and hence, further studies are necessary to draw more robust con- clusions. A second possibility is the increase of the internal den- sity of dust aggregates caused by dust collisions with high-mass ratio (Okuzumi et al. 2009; Dominik et al. 2016). For example, monomer-aggregate collisions with- out restructuring produces aggregates with df = 3 with f ≈ 0.15 which is known as BPCA. Such collisions are expected when fragmentation of dust aggregates oc- curs efficiently, and produces tiny fragments as small as monomer particle (Wada et al. 2008; Paszun & Dominik 2009). It was suggested that fragmentation of dust ag- gregates seems to be necessary to explain disk infrared observations (Dullemond & Dominik 2005). In addition, recent laboratory measurements sug- gested that the adhesion energy of icy dust might be smaller than previous estimates, implying importance of dust fragmentation in disks. Musiolik et al. (2016a,b) showed that CO2-ice, which is also expected to con- dense onto dust particles in outer disk regions, does not show a high adhesion energy. Hence, dust aggregates of CO2-ice-coated particles are thought to be much more fragile than those of H2O-ice. It is worth mentioning that recent experiments also question a high adhesion energy of H2O-ice at low temperature (T (cid:46) 150 − 200 K) (Gundlach et al. 2018; Musiolik & Wurm 2019), although earlier works by Gundlach & Blum (2015) confirmed high adhesion energy even at temperatures down to 100 K. Motivated by these laboratory experiments, Okuzumi & Tazaki (2019) have performed a dust coagulation sim- ulation taking the non-sticky properties of the CO2-ice- coated particles. Since lower adhesion energy predicts lower critical fragmentation velocity (Dominik & Tielens 1997), the maximum dust radius is also reduced if the fragmentation limits coagulation (Birnstiel et al. 2011, 2012). As a result, Okuzumi & Tazaki (2019) showed that scattering-like polarization of HL Tau's disk can be successfully explained by considering fragmentation of CO2-ice-coated particles. Although laboratory experiments are still inconclu- sive, efficient dust fragmentation and subsequent dust collisions with high-mass ratio may potentially explain the origin of the relatively compact dust aggregates sug- gested by millimeter-wave scattering polarization. A third possibility is that they are produced as frag- ments of differentiated planetesimals. If planetesimals are large enough to be molten, that is, differentiated, its fragments can be very compact particles (f ≈ 1). Figure 10. Wavelength dependence of scattering polariza- tion. Solid lines indicate compact aggregate models with amax = 100, 250, 500 µm and f = 0.1. Dashed line is the results for the solid sphere model with amax = 100 µm. Wavelengths of some observational bands of the SOFIA tele- scope and ALMA are indicated by black arrows. and fragmentation upon high-speed collisions (Suyama et al. 2008; Wada et al. 2008, 2009). By assuming perfect sticking, fluffy dust aggregates (f ≈ 10−4) form in disks and they finally become planetesimals by direct coagula- tion (Okuzumi et al. 2012; Kataoka et al. 2013a). These results have also been confirmed by more recent studies (Krijt et al. 2015, 2016; Lorek et al. 2018). However, as we mentioned in Section 5.1.1, these aggregates are not favored from observations of millimeter-wave scattering polarization. Although the presence of these aggregates cannot be ruled out only by scattering-polarization ob- servations, the presence of relatively compact particles are not anticipated in this model. Therefore, a dust evolution model should answer how relatively compact particles could be formed in disks in order to explain millimeter-wave scattering polarization. The first possibility is dust compaction due to bounc- ing collisions. Laboratory experiments suggest that se- quential bouncing collisions lead to gradual compaction of a dust aggregate (Weidling et al. 2009). If bouncing collisions occur, Zsom et al. (2010) suggested that the volume filling factor is increased up to 0.36 within 104 or- bital timescales, which is consistent with the scattering polarization constraints. Windmark et al. (2012a) show that dust coagulation stalls at about 100 µm, which is also similar to the dust size expected from scattering polarization, although the dust size where coagulation stalls depends on disk properties, such as gas density, temperature, and turbulent strength. The dust evolu- tion from a bouncing scenario seems to be consistent with that inferred from scattering polarization observa- 00.10.20.30.40.50.60.7102103SOFIABandCSOFIABandDSOFIABandEALMABand7AMLABand6ALMABand3Pωeffλ(µm)f=0.1,amax=100µmf=0.1,amax=250µmf=0.1,amax=500µmf=1.0,amax=100µmUnveiling Dust Aggregate Structure in Protoplanetary Disks 13 Melting events of icy dust particles, such as more gentle version of chondrule forming events, may also explain compact particles. However, in these case, it is neces- sary to explain why amax is adjusted to λ/2π. Although the origin of compact (sub-)millimeter-size particles is an open question, these particles are ex- pected to form planetesimals via either coagulation with mass transfer (Windmark et al. 2012a,b) or the Stream- ing instability (Youdin & Goodman 2005; Johansen et al. 2007; Bai & Stone 2010a,b). Windmark et al. (2012a) shows that if dust coagulation stalls at sub- millimeter size and if a small amount of larger seed par- ticles are present, i.e., centimeter size, the seed particles grow to form planetesimals via mass transfer. Mean- while, if compact dust particles have St > 10−2, where St is the Stokes number and describes dynamical cou- pling between a dust particle and gas, they are subjected to the Streaming instability (Dr¸azkowska & Dullemond 2014). Thus, the gravitational collapse of dust clumps led by the streaming instability is also another feasible pathway toward planetesimals. 5.3. Comparison with cometary dust in the Solar System Comets in our Solar System are thought to be prim- itive objects, and hence, they are regarded as a living fossil of icy planetesimals. Cometary dust particles pro- vide useful insights into how they form in the early solar nebula. Recently, the Rosetta orbiter followed the comet 67P/Churyumov-Gerasimenko (hereafter 67P) and con- ducted in-situ measurements of cometary dust particles. Comet 67P is a km-sized comet, and it is considered to be a primordial rubble pile object (Davidsson et al. 2016) with bulk density of 0.533±0.006 g cm−3 (Patzold et al. 2016). Three instruments onboard Rosetta are dedicated to analyze dust particles : MIDAS (Micro- Imaging Dust Analysis System), GIADA (Grain Im- pact Analyser and Dust Accumulator), and COSIMA (Cometary Secondly Ion Mass Analyzer). These instruments revealed that cometary dust seems to have two different families in terms of morphol- ogy. MIDAS (Micro-Imaging Dust Analysis System) uses an the atomic force microscope and provides 3D tomographic images of cometary dust particles. MI- DAS found that cometary dust particles seem to have two morphological populations: compact aggregate and fluffy aggregates (Bentley et al. 2016; Mannel et al. 2016, 2019). The dust aggregates analyzed by the MIDAS have sizes from a few to 10 µm, and their subunit diam- eter is typically about from 0.1 µm (Mannel et al. 2019) to 1 µm (Bentley et al. 2016; Mannel et al. 2016). In addition, Bentley et al. (2016) measured fractal dimen- sions of fluffy aggregates and found df = 1.7 ± 0.1. GI- ADA measures the cross-section, momentum, and mass of each dust particle and is mainly sensitive to (sub- )millimeter-sized dust particles. GIADA also measured two populations of dust particles (Della Corte et al. 2015; Fulle et al. 2015, 2016a,b). Fulle et al. (2016a,b) show that fluffy aggregates detected by GIADA seem to have fractal dimension df = 1.87. Also, Fulle et al. (2016b) derived the volume filling factor of compact dust particles as f = 0.48± 0.08. Measurements by COSIMA also support two morphological populations of dust par- ticles (Langevin et al. 2016; Lasue et al. 2019). Fluffy dust aggregates (df (cid:46) 2) detected by Rosetta seem to be primordial aggregates formed in the early solar nebula. However, Fulle et al. (2016b) estimated a mass fraction of fluffy aggregates contained in the nu- cleus of comet 67P and it is only about 0.015%. Hence, comet 67P likely consists of compact dust aggregates with f = 0.48±0.08. If compact aggregates with f ≈ 0.5 are distributed in disks, they are sufficient to produce millimeter-wave scattering in disks as we showed in Sec- tions 3 and 4. Hence, dust particles seen in scattering polarization might be building blocks of cometary ob- jects. In addition to three dust analyzers, OSIRIS (Opti- cal, Spectroscopic, and Infrared Remote Imaging Sys- tem) also provide images of the nucleus as well as dust particles. Based on the OSIRIS images, the tensile strength of comet 67P is estimated (Groussin et al. 2015; Basilevsky et al. 2016). Recently, Tatsuuma et al. (2019) studied the tensile strength of dust aggregates and found that the tensile strength of comet 67P is reproduced when the monomer radius is between 3.3 − 220 µm. Hence, sub-millimeter-sized solid spheres are candidates to explain the measured tensile strength. To summarize, dust particles seen in scattering polar- ization might be precursors of cometary objects. How- ever, it is worth keeping in mind that the present-day cometary dust particles are biased to more compact dust structure because cometary particles are the end prod- uct of a long series of compaction events. Also, dust outflow from a cometary coma may also selectively re- move fluffy dust aggregates. 6. SUMMARY We have studied how dust structure and porosity af- fect scattering polarization at millimeter wavelength be- cause these quantities are important to understand how planetesimals form in protoplanetary disks. First of all, we have computed the optical properties of solid spheres and compact/fluffy dust aggregates at millime- 14 Tazaki et al. ter wavelengths, and then radiative transfer simulations were performed in order to assess their influence on millimeter-wave scattering polarization. Our primary findings are as follows: Thus, multi-wavelength polarimetric observations by ALMA as well as far-infrared disk polarimery seems to be useful to constrain dust porosity in disks (Figure 10). 1. The effective single scattering albedo ωeff of com- pact aggregates with f = 0.01 and fluffy aggre- gates with f (cid:28) 0.01 is shown to be very small. As a result, dust particles with higher fractal di- mension and/or lower porosity are more favorable to explain scattering polarization observations (see Section 3 and also Figure 2). This is confirmed by performing 3D radiative transfer simulations in disks (Section 4). 2. The polarization pattern of a disk containing mod- erately porous particles shows near-and-far side asymmetry. Polarization angles at the disk near side tend to show azimuthal directions, whereas those of the far side show the direction parallel to the minor axis (Figure 6). 3. Although a high porosity is not preferred, for mod- erately porous particles, the width of P ωeff be- comes broad, indicating that scattering polariza- tion can be expected for a wider range of amax. This may relax the tight constraints on amax for the solid sphere model (Figure 9). 4. The wavelength dependence of scattering polar- ization becomes weaker for moderately porous particles compared to solid spheres (Figure 7). 5. Detection of scattering polarization from a disk requires the presence of relatively compact dust aggregates. Aggregates with higher fractal dimen- sion and lower porosity are favored. Although we cannot rule out the presence of fluffy aggregates, at least some amount of compact dust particles should be formed in disks (Section 5.1.1). 6. Dust particles, which can cause millimeter-wave scattering, are similar to those contained in the comet 67P. Thus, icy planetesimals, like comets, might be formed from these dust particles via ei- ther the streaming instability or mass transfer. R.T. would like to thank Cornelis P. Dullemond for making the RADMC-3D code public. R.T. was sup- ported by a Research Fellowship for Young Scientists from the Japan Society for the Promotion of Science (JSPS) (JP17J02411). This work was also supported by JSPS KAKENHI Grant Numbers JP19H05068 (R.T.), JP17H01103 (H.T.), JP18K13590 and JP19H05088 (A.K.), and JP19K03926 and JP18H05438 (S.O.). Software: RADMC-3D (Dullemond et al. 2012) A. DERIVATION OF THE UPPER BOUND ON THE EFFECTIVE SCATTERING OPACITY OF FLUFFY APPENDIX The effective scattering opacity of dust aggregates can be written by AGGREGATES (cid:90) 1 −1 κsca(1 − g) = 1 m 2π k2 (1 − µ)S11,agg(µ)dµ, (A1) where m is the mass of the dust aggregate, k is the wave number, S11,agg is the (1,1) element of the scattering matrix of dust aggregates, and µ = cos θ, where θ is the scattering angle. Using the single scattering assumption, a scattering matrix element of dust aggregates can be written by S11,agg(µ) = N 2S11,mono(µ)S(q), (A2) where N is the number of monomers, S(q) is the static structure factor and q = 2k sin(θ/2) is the magnitude of the scattering vector (Tazaki et al. 2016). When qRg (cid:29) 1 and df = 2, we can approximately decompose the scattering phase function of fluffy dust aggregates by the sum of coherent and incoherent contribution: 11,agg + Sincoherent S11,agg(µ) = Scoherent , Scoherent 11,agg (cid:39) N 2S11,mono(µ)δ(µ − 1), Sincoherent (cid:39) N S11,mono(µ)[(qR0) 11,agg 11,agg −2 + 1], (A3) (A4) (A5) Unveiling Dust Aggregate Structure in Protoplanetary Disks 15 where we have used Equation (29) of Tazaki et al. (2016). Using Equations (A1 and A3) and qR0 (cid:28) 1, we obtain κsca(1 − g) (cid:39) 1 m 2πN k2 (1 − µ)(qR0) −2S11,mono(µ)dµ, (A6) (cid:90) 1 −1 It is worth noting that the coherent component (forward scattering) does not contribute to the effective albedo. As a result, we obtain . (A7) κeff sca ≡ κsca(1 − g) = κsca,mono 2x2 0 REFERENCES Alves, F. O., Girart, J. M., Padovani, M., et al. 2018, A&A, Draine, B. T., & Flatau, P. J. 1994, Journal of the Optical 616, A56 Bacciotti, F., Girart, J. M., Padovani, M., et al. 2018, ApJL, 865, L12 Bai, X.-N., & Stone, J. M. 2010a, ApJ, 722, 1437 -- . 2010b, ApJL, 722, L220 Basilevsky, A. T., Krasil'nikov, S. S., Shiryaev, A. A., et al. Society of America A, 11, 1491 Dr¸azkowska, J., & Dullemond, C. P. 2014, A&A, 572, A78 Dullemond, C. P., & Dominik, C. 2005, A&A, 434, 971 Dullemond, C. P., Juhasz, A., Pohl, A., et al. 2012, RADMC-3D: A multi-purpose radiative transfer tool, Astrophysics Source Code Library, , , ascl:1202.015 2016, Solar System Research, 50, 225 Fulle, M., Altobelli, N., Buratti, B., et al. 2016a, MNRAS, Bentley, M. S., Schmied, R., Mannel, T., et al. 2016, 462, S2 Nature, 537, 73 Fulle, M., Della Corte, V., Rotundi, A., et al. 2015, ApJ, Bertrang, G. H. M., Flock, M., & Wolf, S. 2017, MNRAS, 802, L12 464, L61 Birnstiel, T., Klahr, H., & Ercolano, B. 2012, A&A, 539, -- . 2016b, MNRAS, 462, S132 Girart, J. M., Fern´andez-L´opez, M., Li, Z.-Y., et al. 2018, A148 ApJL, 856, L27 Birnstiel, T., Ormel, C. W., & Dullemond, C. P. 2011, Groussin, O., Jorda, L., Auger, A.-T., et al. 2015, A&A, A&A, 525, A11 583, A32 Birnstiel, T., Dullemond, C. P., Zhu, Z., et al. 2018, ApJL, 869, L45 Blum, J., & Munch, M. 1993, Icarus, 106, 151 Blum, J., & Wurm, G. 2000, Icarus, 143, 138 Bohren, C. F., & Huffman, D. R. 1983, Absorption and scattering of light by small particles Brisset, J., Heisselmann, D., Kothe, S., Weidling, R., & Blum, J. 2017, A&A, 603, A66 Bruggeman, D. A. G. 1935, Annalen der Physik, 416, 636 Cho, J., & Lazarian, A. 2007, ApJ, 669, 1085 Cox, E. G., Harris, R. J., Looney, L. W., et al. 2018, ApJ, 855, 92 Gundlach, B., & Blum, J. 2015, ApJ, 798, 34 Gundlach, B., Schmidt, K. P., Kreuzig, C., et al. 2018, MNRAS, 479, 1273 Harrison, R. E., Looney, L. W., Stephens, I. W., et al. 2019, arXiv e-prints, arXiv:1905.06266 Henning, T., & Stognienko, R. 1996, A&A, 311, 291 Hull, C. L. H., Yang, H., Li, Z.-Y., et al. 2018, ApJ, 860, 82 Johansen, A., Oishi, J. S., Mac Low, M.-M., et al. 2007, Nature, 448, 1022 Kataoka, A., Muto, T., Momose, M., Tsukagoshi, T., & Dullemond, C. P. 2016a, ApJ, 820, 54 Kataoka, A., Okuzumi, S., Tanaka, H., & Nomura, H. 2014, Davidsson, B. J. R., Sierks, H., Guttler, C., et al. 2016, A&A, 568, A42 A&A, 592, A63 Della Corte, V., Rotundi, A., Fulle, M., et al. 2015, A&A, Kataoka, A., Okuzumi, S., & Tazaki, R. 2019, ApJ, 874, L6 Kataoka, A., Tanaka, H., Okuzumi, S., & Wada, K. 2013a, 583, A13 A&A, 557, L4 Dent, W. R. F., Pinte, C., Cortes, P. C., et al. 2019, MNRAS, 482, L29 -- . 2013b, A&A, 554, A4 Kataoka, A., Tsukagoshi, T., Pohl, A., et al. 2017, ApJL, Dominik, C., Paszun, D., & Borel, H. 2016, arXiv e-prints, 844, L5 arXiv:1611.00167 Kataoka, A., Muto, T., Momose, M., et al. 2015, ApJ, 809, Dominik, C., & Tielens, A. G. G. M. 1997, ApJ, 480, 647 Draine, B. T. 2003, ApJ, 598, 1026 -- . 2006, ApJ, 636, 1114 78 Kataoka, A., Tsukagoshi, T., Momose, M., et al. 2016b, ApJL, 831, L12 16 Tazaki et al. Kempf, S., Pfalzner, S., & Henning, T. K. 1999, Icarus, 141, Okuzumi, S., & Tazaki, R. 2019, arXiv e-prints, 388 arXiv:1904.03869 Kimura, H., Kolokolova, L., & Mann, I. 2006, A&A, 449, Ormel, C. W., Spaans, M., & Tielens, A. G. G. M. 2007, 1243 Kirchschlager, F., Bertrang, G. H. M., & Flock, M. 2019, arXiv e-prints, arXiv:1906.10699 Kolokolova, L., & Gustafson, B. A. S. 2001, JQSRT, 70, 611 Kothe, S., Blum, J., Weidling, R., & Guttler, C. 2013, Icarus, 225, 75 Kozasa, T., Blum, J., & Mukai, T. 1992, A&A, 263, 423 Krause, M., & Blum, J. 2004, PhRvL, 93, 021103 Krijt, S., Ormel, C. W., Dominik, C., & Tielens, A. G. G. M. 2015, A&A, 574, A83 -- . 2016, A&A, 586, A20 Langevin, Y., Hilchenbach, M., Ligier, N., et al. 2016, Icarus, 271, 76 Lasue, J., Maroger, I., Botet, R., et al. 2019, arXiv e-prints, arXiv:1903.01206 Lazarian, A., & Hoang, T. 2007a, MNRAS, 378, 910 -- . 2007b, ApJL, 669, L77 Lee, C.-F., Li, Z.-Y., Ching, T.-C., Lai, S.-P., & Yang, H. 2018, ApJ, 854, 56 Liu, H. B. 2019, arXiv e-prints, arXiv:1904.00333 Lorek, S., Lacerda, P., & Blum, J. 2018, A&A, 611, A18 Mackowski, D. W., & Mishchenko, M. I. 1996, Journal of the Optical Society of America A, 13, 2266 Mannel, T., Bentley, M. S., Schmied, R., et al. 2016, MNRAS, 462, S304 Mannel, T., Bentley, M. S., Boakes, P. D., et al. 2019, arXiv e-prints, arXiv:1907.01266 Maxwell Garnett, J. C. 1904, Philosophical Transactions of the Royal Society of London Series A, 203, 385 Min, M., Dullemond, C. P., Kama, M., & Dominik, C. 2011, Icarus, 212, 416 Min, M., Rab, C., Woitke, P., Dominik, C., & M´enard, F. 2016, A&A, 585, A13 Mukai, T., Ishimoto, H., Kozasa, T., Blum, J., & Greenberg, J. M. 1992, A&A, 262, 315 Mulders, G. D., Min, M., Dominik, C., Debes, J. H., & Schneider, G. 2013, A&A, 549, A112 Musiolik, G., Teiser, J., Jankowski, T., & Wurm, G. 2016a, ApJ, 818, 16 -- . 2016b, ApJ, 827, 63 Musiolik, G., & Wurm, G. 2019, ApJ, 873, 58 Ohashi, S., Kataoka, A., Nagai, H., et al. 2018, ApJ, 864, 81 Okuzumi, S., Tanaka, H., Kobayashi, H., & Wada, K. 2012, A&A, 461, 215 Ossenkopf, V. 1993, A&A, 280, 617 Paszun, D., & Dominik, C. 2009, A&A, 507, 1023 Patzold, M., Andert, T., Hahn, M., et al. 2016, Nature, 530, 63 Planck Collaboration, Ade, P. A. R., Aghanim, N., et al. 2014, A&A, 564, A45 Pohl, A., Kataoka, A., Pinilla, P., et al. 2016, A&A, 593, A12 Sadavoy, S. I., Myers, P. C., Stephens, I. W., et al. 2018, ApJ, 869, 115 Seizinger, A., & Kley, W. 2013, A&A, 551, A65 Shen, Y., Draine, B. T., & Johnson, E. T. 2008, ApJ, 689, 260 -- . 2009, ApJ, 696, 2126 Stephens, I. W., Yang, H., Li, Z.-Y., et al. 2017, ApJ, 851, 55 Stognienko, R., Henning, T., & Ossenkopf, V. 1995, A&A, 296, 797 Suyama, T., Wada, K., & Tanaka, H. 2008, ApJ, 684, 1310 Takahashi, S., Machida, M. N., Tomisaka, K., et al. 2019, ApJ, 872, 70 Tatsuuma, M., Kataoka, A., & Tanaka, H. 2019, ApJ, 874, 159 Tazaki, R., Lazarian, A., & Nomura, H. 2017, ApJ, 839, 56 Tazaki, R., & Tanaka, H. 2018, ApJ, 860, 79 Tazaki, R., Tanaka, H., Muto, T., Kataoka, A., & Okuzumi, S. 2019, MNRAS, 485, 4951 Tazaki, R., Tanaka, H., Okuzumi, S., Kataoka, A., & Nomura, H. 2016, ApJ, 823, 70 Testi, L., Birnstiel, T., Ricci, L., et al. 2014, in Protostars and Planets VI, ed. H. Beuther, R. S. Klessen, C. P. Dullemond, & T. Henning, 339 Wada, K., Tanaka, H., Suyama, T., Kimura, H., & Yamamoto, T. 2007, ApJ, 661, 320 -- . 2008, ApJ, 677, 1296 -- . 2009, ApJ, 702, 1490 -- . 2011, ApJ, 737, 36 Warren, S. G., & Brandt, R. E. 2008, Journal of Geophysical Research (Atmospheres), 113, D14220 Weidenschilling, S. J., & Cuzzi, J. N. 1993, in Protostars and Planets III, ed. E. H. Levy & J. I. Lunine, 1031 Weidling, R., Guttler, C., & Blum, J. 2012, Icarus, 218, 688 Weidling, R., Guttler, C., Blum, J., & Brauer, F. 2009, ApJ, 752, 106 ApJ, 696, 2036 Okuzumi, S., Tanaka, H., & Sakagami, M.-a. 2009, ApJ, Windmark, F., Birnstiel, T., Guttler, C., et al. 2012a, 707, 1247 A&A, 540, A73 Unveiling Dust Aggregate Structure in Protoplanetary Disks 17 Windmark, F., Birnstiel, T., Ormel, C. W., & Dullemond, Yang, H., Li, Z.-Y., Stephens, I. W., Kataoka, A., & C. P. 2012b, A&A, 544, L16 Wurm, G., & Blum, J. 1998, Icarus, 132, 125 Looney, L. 2019, MNRAS, 483, 2371 Youdin, A. N., & Goodman, J. 2005, ApJ, 620, 459 Ysard, N., Jones, A. P., Demyk, K., Bout´eraon, T., & Yang, H., Li, Z.-Y., Looney, L., & Stephens, I. 2016a, Koehler, M. 2018, A&A, 617, A124 MNRAS, 456, 2794 Yang, H., Li, Z.-Y., Looney, L. W., et al. 2016b, MNRAS, 460, 4109 Zhu, Z., Zhang, S., Jiang, Y.-F., et al. 2019, arXiv e-prints, arXiv:1904.02127 Zsom, A., Ormel, C. W., Guttler, C., Blum, J., & Dullemond, C. P. 2010, A&A, 513, A57 Yang, H., Li, Z.-Y., Looney, L. W., Girart, J. M., & Zubko, V. G., Mennella, V., Colangeli, L., & Bussoletti, E. Stephens, I. W. 2017, MNRAS, 472, 373 1996, MNRAS, 282, 1321
1711.10119	1	1711	2017-11-28T04:33:47	Golden Elliptical Orbits in Newtonian Gravitation	[ "astro-ph.EP" ]	In spherical symmetry with radial coordinate $r$, classical Newtonian gravitation supports circular orbits and, for $-1/r$ and $r^2$ potentials only, closed elliptical orbits [1]. Various families of elliptical orbits can be thought of as arising from the action of perturbations on corresponding circular orbits. We show that one elliptical orbit in each family is singled out because its focal length is equal to the radius of the corresponding unperturbed circular orbit. The eccentricity of this special orbit is related to the famous irrational number known as the golden ratio. So inanimate Newtonian gravitation appears to exhibit (but not prefer) the golden ratio which has been previously identified mostly in settings within the animate world.	astro-ph.EP	astro-ph	Golden Elliptical Orbits in Newtonian Gravitation Dimitris M. Christodoulou ABSTRACT In spherical symmetry with radial coordinate r, classical Newtonian gravita- tion supports circular orbits and, for −1/r and r2 potentials only, closed elliptical orbits [1]. Various families of elliptical orbits can be thought of as arising from the action of perturbations on corresponding circular orbits. We show that one elliptical orbit in each family is singled out because its focal length is equal to the radius of the corresponding unperturbed circular orbit. The eccentricity of this special orbit is related to the famous irrational number known as the golden ratio. So inanimate Newtonian gavitation appears to exhibit (but not prefer) the golden ratio which has been previously identified mostly in settings within the animate world. 1. Introduction In 1873, J. Bertrand [1] proved that the only spherically symmetric gravitational poten- tials that can support bound closed noncircular orbits are the Newton-Kepler −1/r potential [12] and the isotropic Hooke r2 potential [8], where r is the spherical radial coordinate with respect to the central mass that generates the potential. The elliptical orbits in these two potentials were already known to I. Newton [12]. Both potentials also support circular orbits and the elliptical orbits can be thought of as arising from such circular orbits perturbed by disturbances of any arbitrary amplitude. Given a circular orbit of radius ro, an infinite family of elliptical orbits can thus be obtained with eccentricities in the range 0 < e < 1, where e is related to the ratio of semiaxes b/a of the ellipses by e = r1 − b2 a2 . (1) In recent work [3], we found a new geometric property that characterizes each family of eliptical orbits and this property switches between the two potentials in a highly symmetric fashion: the circular radius ro is the harmonic mean of the radii of the turning points rmax = a(1 + e) and rmin = a(1 − e) of the ellipses in a −1/r potential; whereas ro is the geometric mean of of the turning points rmax = a and rmin = b of the ellipses in an – 2 – r2 potential. For the reader's convenience, we summarize in Section 2 the derivation of these properties and we proceed in Section 3 to search in each family for special orbits with additional geometric properties. Interestingly, we find that Newtonian gravitation singles out some elliptical orbits whose eccentricities are related to the golden ratio conjugate [14] √5 − 1 2 ϕ∗ ≡ ≈ 0.618 . (2) The importance of the result lies in the fact that this famous irrational number is not introduced by the geometry of space as in the ubiquitous case of π in Euclidean spaces; the golden ratio conjugate is instead singled out by the dynamics of the noncircular orbits. We discuss our conclusions further in Section 4. 2. Known Geometric Properties of Elliptical Orbits We apply the laws of energy and angular momentum conservation to a family of elliptical orbits that arise from disturbances acting on a given circular orbit of radius ro. We distinguish two cases, a Newton-Kepler potential and an isotropic Hooke potential. 2.1. Newton-Kepler −1/r Potential Consider an equilibrium orbit with radius r = ro in a −1/r potential and assume that this orbit is perturbed to an elliptical shape with turning points rmin = ro − A1 and rmax = ro + A2, where 0 < A1 < ro and A2 > A1 (Figure 1). At the turning points T1 and T2 shown in Figure 1, the radial velocity is zero (dr/dt = 0) and the total energy per unit mass can then be written as [7] E = L2 2r2 − GM r , (3) where G is the gravitatonal constant, M is the mass that generates the potential, and the specific angular momentum satisfies L2 = GMro = const., hence eq. (3) can be written in the form = const. , if dr/dt = 0 . (4) (5) = E GM ro 2r2 − 1 r Applied to the turning points, this equation yields 1 A1 − 1 A2 = 2 ro , – 3 – Fig. 1.- Schematic diagram of a circular orbit O(ro) and the associated golden elliptical orbit in a Newtonian gravitational potential of the form −1/r due to a star located at focus F1. This focus coincides with the center O of the circular orbit. Point K on ⊙O is the center of the ellipse; T1 and T2 are the turning points; and L1 and L2 are the endpoints of the latus rectum. The golden ellipse is singled out because of the equality c = p = ro and its eccentricity e = φ∗ ≈ 0.618. or equivalently – 4 – 1 rmin + 1 rmax = 2 ro . (6) This last equation shows that, in a −1/r potential, the circular equilibrium radius ro is the harmonic mean of the radii of the turning points rmin and rmax of the elliptical orbits [3]. 2.2. Isotropic Hooke r2 Potential Consider an equilibrium orbit with radius r = ro in a Ω2r2/2 potential (Ω =const.) and assume that this orbit is perturbed to an elliptical shape with turning points rmin = b and rmax = a, where a and b < a are the semiaxes of the ellipse (Figure 2). Here the radii of the turning points take these special values because in the r2 potential the gravitational force always points toward the center of the ellipse where the central mass is located. At the turning points E1 and E2 shown in Figure 2, the radial velocity is zero (dr/dt = 0) and the total energy per unit mass can then be written as E Ω2/2 = r4 o r2 + r2 = const. , if dr/dt = 0 , where the constant specific angular momentum is given by L = Ωr2 o in this case. Applied to the turning points, this equation yields rmaxrmin = ab = r2 o . (7) (8) This last equation shows that, in an isotropic r2 potential, the circular equilibrium radius ro is the geometric mean of the semiaxes a and b of the elliptical orbits [3]. 3. New Geometric Properties of Elliptical Orbits The fundamental difference between the elliptical orbits in the above two cases is im- posed by the dynamics: the gravitational force always points toward focus F1 in Figure 1, whereas it always points toward center O in Figure 2. This difference is responsible for producing equations (6) and (8) and it also leads to several more geometric properties of the ellipses in each case, as we describe below. – 5 – Fig. 2.- Schematic diagram of a circular orbit O(ro) and the associated golden elliptical orbit in an isotropic Hooke gravitational potential of the form r2 due to a star located at center O. Unlike in the Newtonian case, the gravitational force always points toward O in this case. Points E1 and E2 are the endpoints of the semiaxes a and b of the ellipse. The golden ellipse is singled out because of the equality c = ro, where c is the focal length, and its eccentricity e = √φ∗ ≈ 0.786. – 6 – 3.1. Newtonian Ellipses Using the well-known equations of the ellipse with semiaxes a and b, viz. and equation (6) takes the form a = (rmax + rmin)/2 , b = √rmaxrmin , aro = b2 , (9) (10) (11) which indicates that, in Newtonian ellipses, the semiminor axis b is the geometric mean of a and ro. More importantly, the semiaxes cannot be chosen independently in a particular family corresponding to a circular radius ro. Since the semiaxes are also related to the eccentricity e (eq. (1)), each family of ellipses and the values of a and b are fully determined by the choice of e for fixed ro. Written in the equivalent form ro = b2 a ≡ p , equation (11) indicates that ro is also equal to the semilatus rectum p. Combining equations (12) and (1), we find that ro = a(1 − e2) = b√1 − e2 , and using the definition of the focal length c, i.e., c = ae, this equation takes the form ro = c(cid:18)1 − e2 e (cid:19) . (12) (13) (14) Equation (13) shows that there are no ellipses in which one or the other semiaxis equals ro. But equation (14) indicates that there exists a special ellipse in the family for which the focal length c equals ro (Figure 1). The eccentricity of the special ellipse is found by setting c = ro in equation (14), in which case we find that e2 + e − 1 = 0 , (15) whose solution is the golden ratio conjugate shown in equation (2) above. To summarize, the only special ellipse that is singled out by the Newton-Kepler dynamics for a given ro is the golden ellipse with e = ϕ∗ and c = p = ro, whereas all other ellipses in the family obey only p = ro. – 7 – 3.2. Hookean Ellipses We have already seen (eq. (8)) that in a family of Hookean ellipses, the radius ro of the corresponding circular orbit is the geometric mean of the semiaxes a and b, which also implies that the areas of these ellipses are all equal to πr2 o. Therefore, the semiaxes cannot be chosen independently within a particular family. Since the semiaxes are also related to the eccentricity e (eq. (1)), each family of ellipses and the values of a and b are fully determined by the choice of e for fixed ro. Using equations (8) and (1), we find that and the semilatus rectum (eq. (12)) can be written as ro = a(1 − e2)1/4 = b(1 − e2)−1/4 , p = ro(1 − e2)3/4 . Hence, these ellipses are very different than the Newtonian ellipses. Using the definition of the focal length c = ae, equation (16) takes the form ro = c(cid:20)(1 − e2)1/4 e (cid:21) , (16) (17) (18) which indicates that there exists a special ellipse in the family for which the focal length c equals ro. The eccentricity of the special ellipse is found by setting c = ro in equation (18), in which case we find that (19) whose solution is e = √ϕ∗. This ellipse is plotted in Figure 2. Since c = ro, an additional property is that its focal length is the geometric mean of its semiaxes, viz. e4 + e2 − 1 = 0 , c2 = ab . (20) 4. Discussion The main result of this work is the appearance of the golden ratio conjugate ϕ∗ (eq. (2)) in Newtonian dynamics that predicts the existence of families of elliptical orbits only in −1/r and r2 spherically symmetric gravitational potentials [1]. In each case, the eccentricity e of the golden elliptical orbit is related to ϕ∗ (e = ϕ∗ and e = √ϕ∗, respectively); these relations appear in the special case in which the focal length of the ellipse is equal to the radius of the corresponding circular orbit of the family (c = ro in Section 3 above). – 8 – A literature search shows that various authors use two different definitions of the golden ellipse: some authors [9, 10] define as golden the ellipse that is inscribed in a golden rectangle; others [13] construct the golden ellipse from golden right triangles in which case it is the eccentricity that is equal to the golden ratio conjugate. This latter definition is applicable to our Newtonian ellipses in which b/a = √ϕ∗. On the other hand, the former definition is applicable to our Hookean ellipses that have b/a = ϕ∗ and they can be inscribed in a golden rectangle. Most of the appearances of ϕ∗ in the animate world are well-known if not famous. The golden ratio has been identified in phyllotaxis of plants [4]; in human constructions, including the Parthenon and the Egyptian pyramids [11]; in the honeycombs of bees [6]; and recently in the shapes of red blood cells [15]. Lately, the golden ratio has also appeared in the inanimate world and our work concerning Newtonian dynamics falls in this category. Two more examples concern black holes [5] and cosmological theories [2], although the interested reader may easily track down more cases in solar neutrino mixing and quantum mechanics. In the inanimate world, the golden ratio is usually introduced by the geometry of space In our case, Newtonian or spacetime, but this is not the case in Newtonian dynamics. gravitation allows for infinite families of elliptical orbits for a given circular equilibrium orbit, so the theory does not show a preference for ellipses with eccentricities e = ϕ∗. The golden ellipses are special only because they exhibit additional geometric properties, as we described in Sections 2.1 and 3.1. In our solar system, no planetary orbit has orbital eccentricity anywhere near the golden value of 0.618; our neighboring planets and the dwarf planets all move in fairly circular orbits with e < 0.442. Furthermore, we searched the exoplanet database (//exoplanets.org) that currently lists nearly 3,000 exoplanets with confirmed orbits and we found only three planets with e = 0.610, two planets with e = 0.630, and no other planets with intermediate eccen- tricities. So exoplanetary orbits appear to support the theoretical picture of no preference for the golden Newtonian elliptical orbits. REFERENCES [1] Bertrand, J., , M´ecanique analytique, C. R. Acad. Sci. Paris, 77, 849-853, 1873. [2] Bryant, M. S., & Hobill, D. W., Bianchi IX Cosmologies and the Golden Ratio, Classical and Quantum Gravity, 34, 125010 (23 pages), 2017. [3] Christodoulou, D. M., & Kazanas, D., A Physical Interpretation of the Titius-Bode – 9 – Rule and its Connection to the Closed Orbits of Bertrand's Theorem, Research in Astronomy & Astrophysics, 17, 129 (6 pages), 2017. [4] Coxeter, H. S. M., Introduction to Geometry, Wiley, New York, 1969. [5] Cruz, N., Olivares, M., & Villanueva, J. R., The Golden Ratio in Schwarzschild-Kottler Black Holes, European Physics Journal C, 77, 123 (6 pages), 2017. [6] Favre, D., Golden Ratio in the Elliptical Honeycomb, Journal of Nature and Science, 2, e173 (9 pages), 2016. [7] Goldstein, H., Classical Mechanics, Addison-Wesley, Reading, MA, 1950. [8] Hooke, R., De Potentia Restitutiva, or of Spring. Explaining the Power of Springing Bodies, J. Martyn, London, 1678. [9] Huntley, H. E., The golden Ellipse, Fibonacci Quarterly, 12, 38-40, 1974. [10] Kapur, J. N., The Golden Ellipse, International Journal of Mathematical Education in Science and Technology, 18, 205-214, 1987. [11] Livio, M., The Golden Ratio, Broadway Books, New York, 2002. [12] Newton, I., Philosophae Naturalis Principia Mathematica, S. Pepys, Reg. Soc. Praesses, London, 1687. [13] Stakhov, A. P., The Mathematics of Harmony, World Scientific, Singapore, 2009. [14] Weisstein, E. W., Golden Ratio Conjugate, From MathWorld-A Wolfram Web Resource, http://mathworld.wolfram.com/GoldenRatioConjugate.html . [15] Zhang, X.-J., & Ou-Yang, Z.-C., Mechanism Behind the Beauty: Golden Ratio Appears in Red Blood Cell Shape, Communications in Computational Physics, 21, 559-569, 2017. Dimitris M. Christodoulou: Department of Mathematical Sciences, University of Massachusetts Lowell, Lowell, MA 01854, USA Email address: dimitris [email protected] This preprint was prepared with the AAS LATEX macros v5.2.
1506.01084	1	1506	2015-06-02T23:18:59	WHFast: A fast and unbiased implementation of a symplectic Wisdom-Holman integrator for long term gravitational simulations	[ "astro-ph.EP", "astro-ph.IM", "math.NA", "nlin.CD", "physics.comp-ph" ]	We present WHFast, a fast and accurate implementation of a Wisdom-Holman symplectic integrator for long-term orbit integrations of planetary systems. WHFast is significantly faster and conserves energy better than all other Wisdom-Holman integrators tested. We achieve this by significantly improving the Kepler-solver and ensuring numerical stability of coordinate transformations to and from Jacobi coordinates. These refinements allow us to remove the linear secular trend in the energy error that is present in other implementations. For small enough timesteps we achieve Brouwer's law, i.e. the energy error is dominated by an unbiased random walk due to floating-point round-off errors. We implement symplectic correctors up to order eleven that significantly reduce the energy error. We also implement a symplectic tangent map for the variational equations. This allows us to efficiently calculate two widely used chaos indicators the Lyapunov characteristic number (LCN) and the Mean Exponential Growth factor of Nearby Orbits (MEGNO). WHFast is freely available as a flexible C package, as a shared library, and as an easy-to-use python module.	astro-ph.EP	astro-ph	Mon. Not. R. Astron. Soc. 000, 000–000 (0000) Printed 4 June 2015 (MN LATEX style file v2.2) WHFast: A fast and unbiased implementation of a symplectic Wisdom-Holman integrator for long term gravitational simulations Hanno Rein1,2 and Daniel Tamayo1,3,4 1 Department of Physical and Environmental Sciences, University of Toronto at Scarborough, Toronto, Ontario M1C 1A4, Canada 2 Department of Astronomy and Astrophysics, University of Toronto, Toronto, Ontario, M5S 3H4, Canada 3 Canadian Institute for Theoretical Astrophysics, 60 St. George St, University of Toronto, Toronto, Ontario M5S 3H8, Canada 4 Centre for Planetary Sciences Fellow Submitted: 8th May 2015. Accepted: 2nd June 2015. ABSTRACT We present WHFast, a fast and accurate implementation of a Wisdom-Holman symplectic integrator for long-term orbit integrations of planetary systems. WHFast is significantly faster and conserves energy better than all other Wisdom-Holman integrators tested. We achieve this by significantly improving the Kepler-solver and ensuring numerical sta- bility of coordinate transformations to and from Jacobi coordinates. These refinements allow us to remove the linear secular trend in the energy error that is present in other implementa- tions. For small enough timesteps we achieve Brouwer’s law, i.e. the energy error is dominated by an unbiased random walk due to floating-point round-off errors. We implement symplectic correctors up to order eleven that significantly reduce the en- ergy error. We also implement a symplectic tangent map for the variational equations. This allows us to efficiently calculate two widely used chaos indicators the Lyapunov characteristic number (LCN) and the Mean Exponential Growth factor of Nearby Orbits (MEGNO). WHFast is freely available as a flexible C package, as a shared library, and as an easy-to- use python module. Key words: methods: numerical — gravitation — planets and satellites: dynamical evolution and stability 1 INTRODUCTION Celestial mechanics, the field that deals with the motion of celes- tial objects, has been an active field of research since the days of Newton and Kepler. Analytic solutions only exist for a few special cases. Historically, the main driver for the development of pertur- bation theory has been the problem of planets orbiting the Sun. Be- cause the central body is so much more massive than the planets, it is profitable to ask how the small mutual tugs between the planets modify the Keplerian orbits they would each individually follow around the Sun in the absence of the other bodies. This analytical approach has been, and continues to be, successful in explaining many important features of planetary orbits. However, the Solar System is chaotic, and the rise of computing power has yielded many important insights. There is therefore considerable interest in developing fast and accurate numerical integrators. A large number of such integrators have been developed over the years to perform this task. For many long term integrations, symplectic integrators have proven to be a favourable choice. Sym- plectic schemes incorporate the symmetries of Hamiltonian sys- tems, and therefore typically conserve quantities like the energy and angular momentum better than non-symplectic integrators. c(cid:13) 0000 RAS For integrations of planetary systems, Wisdom & Holman (1991), and independently Kinoshita et al. (1991), developed a widely used class of symplectic integrators. The ideas of Wisdom & Holman (1991) developed from the original ideas of the mapping method of Wisdom (1981). We refer to these as a Wisdom-Holman mapping or a Wisdom-Holman integrator. Since then, many authors have modified and built upon this method, and several have made their integrators publicly available to the astrophysics community (e.g. Chambers & Migliorini 1997; Duncan et al. 1998). The Wisdom-Holman integrator exploits the intuition from perturbation theory that one can separate the problem into a system of Keplerian orbits about the Sun, modified by small perturbations among the planets. The nuisance is that while Newton provided us the solution to the two-body problem, Poincar´e showed that the remaining superimposed perturbations are not integrable. Analyt- ically, the traditional way forward is to average over the short- period oscillations in the problem to yield approximate solutions. The great insight of Wisdom and Holman was that, at the same level of approximation, one can add high frequency terms. By judicious choice of these additional frequencies, the perturbations among the planets can be transformed into trivially integrated delta functions. 2 Hanno Rein and Daniel Tamayo The result is an exceedingly efficient integrator that has proven an indispensable tool for modern studies in celestial mechanics. In this paper, we present results from a complete reimple- mentation of the Wisdom-Holman integrator. We show how to speed up the algorithm in several ways and dramatically increase its accuracy. Many of the improvements are related to finite dou- ble floating-point precision on modern computers (IEEE754, ISO 2011). The fact that almost all real numbers cannot be represented exactly in floating-point precision leads to important consequences for the numerical stability of any algorithm and the growth of nu- merical round-off error. To our knowledge, we present the first publicly available Wisdom-Holman integrator that is unbiased, i.e. the errors are ran- dom and uncorrelated. This leads to a very slow error growth. For sufficiently small timesteps, we achieve Brouwer’s law, i.e., the en- ergy error grows as time to the power of one half. We have also sped up the integrator through various improve- ments to the integrator’s Kepler-solver. Our implementation allows for the evolution of variational equations (to determine whether or- bits are chaotic) at almost no additional cost. Additionally, we im- plement so-called symplectic correctors up to order eleven to in- crease the accurary (Wisdom et al. 1996), allow for arbitrary unit choices, and do not tie the integration to a particular frame of ref- erence. We make our integrator, which we call WHFast, publicly avail- able in its native C99 implementation and as an easy-to-use python module. The remainder of this paper is structured as follows. We first summarize the concepts and algorithms used in this paper, includ- ing Jacobi Coordinates, our choice of Hamiltonian splitting, the symplectic Wisdom-Holman map, symplectic correctors and the variational equations in Sect. 2. We then go into detail discussing the improvements we have made to these algorithms in Sect. 3. Nu- merical tests are presented in Sect. 4 before we conclude in Sect. 5. 2 BACKGROUND The Hamiltonian H of the gravitational N-body system can be writ- ten as the sum of kinetic and potential terms in Cartesian coordi- nates N−1(cid:88) i=0 p2 i 2mi − N−1(cid:88) N−1(cid:88) i=0 j=i+1 H = 2.1 Jacobi Coordinates Rather than reference planet positions to the central star, a planet’s Jacobi coordinates are measured relative to the centre-of-mass of all bodies with lower indices. For concreteness, consider a system of N particles with masses mi, i = 0, . . . , N − 1. Let ri be the posi- tion vector of the i-th particle with respect to an arbitrary origin that is fixed in an inertial frame. Here we assume that the particles are ordered such that i = 0 corresponds to the central object, i = 1 to the innermost object orbiting the central object and so on. The ex- istence of such an ordering does not restrict the architecture of the system. For example, the coordinates of an equal-mass binary with a circumbinary particle can be expressed in Jacobi coordinates. But note that the ordering might in general be non-unique and that it can change during an integration. This can have important implications for a numerical scheme using Jacobi coordinates. The Jacobi coordinate r(cid:48) i of the i-th particles is the position relative to Ri−1, the centre-of-mass of all the particles interior to the i-th particle: i(cid:88) i = ri − Ri−1, r(cid:48) 1 Mi where Ri = for i = 1, . . . , N − 1 i(cid:88) (2) (3) m jr j and Mi = m j. j=0 j=0 The momenta, however, transform differently1. The momen- i and the corresponding Jacobi mass are given Other quantities such as the velocity and acceleration (also the co- ordinates in the variational equations, see below) transform in the same way. This is because the Jacobi coordinates are a linear func- tion of the Cartesian coordinates, and the velocity is the time deriva- tive of the position in both coordinates systems. tum conjugate to r(cid:48) by p(cid:48) i = m(cid:48) i r(cid:48) and Note that the Jacobi mass m(cid:48) i is the reduced mass of mi and Mi−1. Explicit expressions for the momenta can be found by evaluating the time derivative rEq. 4. The Jacobi coordinates above are relative coordinates for i = 1, . . . , N − 1. For the 0-th coordinate, a different convention is used, mi Mi−1 mi + Mi−1 i = m(cid:48) iv(cid:48) m(cid:48) i = mi Mi−1 Mi (4) = . i r(cid:48) 0 = RN−1, m(cid:48) 0 = MN−1, p(cid:48) 0 = p j. (5) Thus, r(cid:48) tem, p(cid:48) 0 points towards the centre-of-mass of the entire sys- 0 is the total momentum and m(cid:48) 0 is the total mass. j=0 N−1(cid:88) Gmim j ri − r j . (1) 2.2 Hamiltonian Splitting One way forward toward separating out the two-body Keplerian Hamiltonians is to transform to heliocentric coordinates involving the centre-of-mass and the ri−r0. However, rewriting the Cartesian momenta in terms of heliocentric momenta (which have an addi- tional component along the centre-of-mass momentum), leads to several cross-terms. Alternatively, Jacobi worked out a coordinate system in which the kinetic terms are particularly clean, and the ki- netic energy remains a sum of squares. For readers that may not be familiar, and because our improved accuracy is largely due to mod- ifications of the manner in which we transform between Cartesian and Jacobi coordinates, we briefly review them (see also Plummer 1918; Sussman & Wisdom 2001; Murray & Dermott 2000). After some algebra, we can rewrite the Hamiltonian in Eq. 1 in terms of the conjugate momenta of the Jacobi coordinates (e.g. Murray & Dermott 2000; Sussman & Wisdom 2001). We only rewrite the kinetic term and keep the potential term expressed as a function of the Cartesian coordinates: H = − N−1(cid:88) N−1(cid:88) N−1(cid:88) i (6) Gmim j ri − r j . 2 p(cid:48) 2m(cid:48) i i=0 i=0 j=i+1 1 But note that we do not need to calculate the momenta explicitly in our algorithm. c(cid:13) 0000 RAS, MNRAS 000, 000–000 WHFast: A fast and unbiased Wisdom-Holman integrator 3 After grouping terms in the Hamiltonian, we arrive at H = + 0 2 i 2 i=1 p(cid:48) 2m(cid:48) p(cid:48) 2m(cid:48) Gm(cid:48) i Mi r(cid:48) i − N−1(cid:88) N−1(cid:88) (cid:124)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:123)(cid:122)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:125) 0(cid:124)(cid:123)(cid:122)(cid:125)H0 N−1(cid:88) N−1(cid:88) i − N−1(cid:88) (cid:124)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:123)(cid:122)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:125) Gm(cid:48) i Mi r(cid:48) Gmim j ri − r j i=1 HKepler j=i+1 i=1 i=0 + i HInteraction . (8) Note that the kinetic term is still diagonal, i.e. there are no cross terms involving pip j with i (cid:44) j. Next, we add and subtract the term H± = N−1(cid:88) (7) . Gm(cid:48) i Mi r(cid:48) i i=1 The first term, H0, simply describes the motion of the centre-of- mass r(cid:48) 0 along a straight line. For that reason this term is often ig- nored. However, we keep it which will allow us to integrate parti- cles without any restriction to a particular frame of reference. The terms HKepler can be split up further into a sum of (cid:17) 2 (cid:16)HKepler (9) = i p(cid:48) 2m(cid:48) i i . − Gm(cid:48) i Mi (cid:16)HKepler r(cid:48) i (cid:17) i Each of the Hamiltonians describes the Keplerian motion of the i-th particle with mass mi around the centre-of-mass of all interior particles with total mass Mi−1. After some more algebra, the interaction term can be simpli- fied and split into two parts, one of which can be easily computed in Jacobi coordinates and the other in Cartesian coordinates of the inertial frame H Interaction = − N−1(cid:88) N−1(cid:88) N−1(cid:88) (10) Gmim j ri − r j . Gm(cid:48) iMi r(cid:48) i i=2 i=0 j=i+1 j(cid:44)1 N−1(cid:88) i Mi Gm(cid:48) r(cid:48) i One important point to note is that our choice of H± is slightly different from that used by Murray & Dermott (2000) and Wisdom & Holman (1991). These authors use (H±)WH1991 = where Mi = m0 . Their choice leads to the usual disturbing function in perturbation theory. We conducted various tests but found no significant difference between these mass choices. We therefore chose our prescription, Eq. 7, which has a simpler phys- ical interpretation: the mass entering Kepler’s third law is simply the interior mass. Mi Mi−1 (11) i=1 . 2.3 Wisdom-Holman Mapping Our goal is to find a solution to the equations of motion for par- ticles governed by the Hamiltonian in Eq. 1. No analytic solution exists to the full Hamiltonian and we thus need to find an approxi- mate solution. There are many different ways to do that. Here, we describe the idea of constructing a symplectic integrator by means of splitting the Hamiltonian into smaller parts, each of which can be easily integrated. The introduction of Jacobi coordinates led us to the Hamil- tonian splitting described in Sect. 2.2. Analytic solutions can be found for the evolution of the system under each of the individual c(cid:13) 0000 RAS, MNRAS 000, 000–000 Hamiltonians H 0 and H Interaction. The solution to H 0 simply corre- sponds to motion along a straight line. The solution to H Interaction is a kick step where the velocities change due the inter-particle accel- erations but the position remain constant. The solution to H Kepler is a set of two-body Kepler orbits, which can also be easily solved with an iterative algorithm. We discuss the details related to the Kepler problem in Sect. 2.6. Now that we have broken down the full Hamiltonian into in- dividual Hamiltonians, to all of which we know the solution (or can easily calculate it), we can construct a symplectic integrator for the total Hamiltonian using an operator split method (e.g. Saha & Tremaine 1992). Let us describe the evolution of particles under a Hamiltonian H for a time dt using the operator notation H(dt). The notation H 2(dt) ◦ H 1(dt) means applying operator H 1 first, then applying operator H 2. It is easy to see that many of the oper- ators commute with each other, i.e. (cid:104) H 0, H Kepler (cid:104) H 0, H Interaction (cid:17) (cid:16) H Kepler (cid:17) (cid:105) (cid:105) (cid:21) (cid:20)(cid:16) H Kepler = 0 = 0 = 0 ∀i, j, i j , (14) where [ H 1, H 2] = H 1 ◦ H 2 − H 2 ◦ H 1. This leads to the following Drift-Kick-Drift (DKD) operator splitting scheme, which we refer to as the Wisdom-Holman map: (Drift) Evolve the system under H Kepler(dt/2) ◦ H 0(dt/2). (Kick) Evolve the system under H Interaction(dt). (Drift) Evolve the system under H Kepler(dt/2) ◦ H 0(dt/2). H Kepler and H 0 in the first and last step doesn’t The ordering of matter as they commute. The first and last steps can be combined if the system is evolved for multiple timesteps. Note that the evolution of H Kepler and H 0 is most easily accomplished in Jacobi coordinates. The interaction Hamiltonian H Interaction, however, contains terms that depend on both the Carte- sian and Jacobi coordinates. The simplest way to calculate these terms is to convert to Cartesian coordinates, evaluate the ri − r j term, convert the accelerations back to Jacobi accelerations, and calculate the remaining terms. (12) (13) 2.4 Symplectic Correctors The operator splitting method used in the symplectic integrator dis- cussed above effectively adds high frequency terms to the Hamil- tonian. An argument often used in favour of symplectic integrators is that, although these high-frequency terms alter the Hamiltonian, they do not change the long term evolution as they average out. However, they do lead to relatively large short term oscillations, for example in the energy error. The idea of a symplectic corrector, first used by Tittemore & Wisdom (1989) and fully developed by Wisdom et al. (1996), is to remove some of these high frequency terms using perturbation theory. The basic procedure is as follows. Before the start of an in- tegration, we convert from real coordinates to so-called mapping coordinates. Then we perform the integration using our standard symplectic map. After the simulation has finished (or whenever we need an output) we convert back from mapping to real coordinates. The symplectic corrector operator that we use is a combination of H Interaction(dt) and H Kepler(dt) ◦ H 0(dt) operators applied several for different (positive and negative) intervals dt. If is the order 4 Hanno Rein and Daniel Tamayo of the perturbations, i.e. the mass ratio and therefore the relative H Interaction compared to H Kepler, then one can show magnitude of that the use of symplectic correctors can lead to a scheme of order O(dtK) + O(2dt2) where K is the order of the symplectic corrector (Mikkola & Palmer 2000). A second order Wisdom-Holman map without symplectic correctors has an energy error of order O(2dt2). Because this coordinate transformation for the symplectic corrector is only performed for outputs and at the beginning and end of the simulation, its effect on the speed of the algorithm is negligible for sparse output. A full derivation of the symplectic correctors would go beyond the scope of this paper and we refer the reader to Wisdom et al. (1996) and Mikkola & Palmer (2000). The corrector coefficients are listed in a compact form in Wisdom (2006). We implement a third, fifth, seventh and eleventh order sym- plectic corrector for WHFast. Whether the high order-symplectic correctors provide any improvement over the low-order ones de- pends on the mass ratios in the system. For Jupiter-mass planets, a symplectic corrector of fifth order is no less accurate than a higher order one. If in doubt, there is no harm done in using a higher- order corrector as the speed implications are minimal. Thus, we implement the eleventh-order symplectic corrector by default. 2.5 Chaos Indicators A powerful tool for studying the long term evolution of Hamil- tonian systems is the Lyapunov characteristic number (LCN). The inverse of the LCN is the Lyapunov timescale and gives an estimate of how fast two nearby particle trajectories diverge. If the system is chaotic, the divergence is exponential in time and the Lyapunov timescale is finite. Thus, measuring the LCN gives us an estimate of whether the system is chaotic and, if so, on what timescale. A more recent approach with similar informative value is the Mean Exponential Growth factor of Nearby Orbits, or MEGNO for short (Cincotta et al. 2003). The MEGNO, Y(t), is a scalar function of time, and provides a clear picture of resonant structures and of the locations of stable and unstable periodic orbits. There are two ways to calculate the LCN or the MEGNO. Conceptually the simplest is to integrate an additional shadow par- ticle for each body in the simulation, i.e. a particle with slightly per- turbed initial conditions. One can then directly measure the diver- gence of each particle’s path from its shadow. The second approach is to consider each body’s six-dimensional displacement vector δiδiδi from its shadow (in both position and velocity) as a dynamical vari- able. One can then obtain differential equations for each δiδiδi vector by applying a variational principle to the trajectories of the origi- nal bodies. We choose to follow the latter approach, as it is both faster and numerically more robust (Tancredi et al. 2001). In this scheme, one can imagine shadow particles with phase-space coor- = ξξξi +δδδi, where ξξξi = (ri, vi) is the phase-space coordinate dinates ξξξs i of the i-th original particle. Initially, we set each component of δδδi to a small value. We follow the work of Mikkola & Innanen (1999) who de- scribe how to efficiently couple the variational equations to the original equations of motion. This allows us to construct a sym- plectic integrator for the variational equations (a symplectic tangent map). An important advantage of this method is that we only solve Kepler’s equation once for each particle/shadow-particle pair (one of the most time-consuming steps in a Wisdom-Holman integrator for small particle numbers). The MEGNO is then straightforwardly computed from the (cid:90) t (cid:80)N−1 variations as (Cincotta et al. 2003) (cid:80)N−1 δδδi(t(cid:48)) · δδδi(t(cid:48)) i=0 δδδ2 2 t t(cid:48) i=0 dt(cid:48) . 0 i (t(cid:48)) Y(t) = (15) If Y(t) → ∞, then the system is chaotic. For quasi-periodic orbits, the MEGNO converges to a finite value, Y(t) → 2 (e.g. Hinse et al. 2010). One can obtain the Lyapunov characteristic number (LCN), the inverse of the Lyapunov timescale, from the time evolution of the MEGNO via a linear least square fit to Y(t). 2.6 Kepler Problem with Variations In this section, we summarize how to solve the two-body Kepler problem numerically, including the variational equations. Although the solution has been known since the days of Newton, the tran- scendental nature of Kepler’s equation does not admit a closed- form mathematical expression. We closely follow the work of (Mikkola & Innanen 1999) where the reader can find additional information that we have left out. The equivalent one-body Hamiltonian for the Kepler problem is HKepler = 1 2 v2 − M r , (16) i from p(cid:48) where M is the total mass of the two bodies. For consistency with Mikkola & Innanen (1999), we have dropped the primes, have scaled out m(cid:48) i, and rewritten Eq. 9 in non-dimensional form, i.e. the gravitational constant G = 1 for the remainder of this pa- per. However, we have taken care to remove any dependence on the choice of units from our implementation, so G can be freely set by the user in our implementation of the algorithm. Our task is to find the final positions and velocities r and v of a particle evolving under this Hamiltonian for some time dt, given the initial conditions r0 and v0. Thus, we seek the effect of the operator HKepler(dt). It is advantageous to solve the Kepler problem numerically using the Gauss f and g functions, which express the relevant quan- tities in terms of r0 and v0 (Wisdom & Holman 1991). This avoids the computationally expensive conversion between Cartesian and classical orbital elements, and avoids coordinate singularities asso- ciated with circular orbits. We find that it is advantageous to use universal variables in this solution (Stumpff 1962). This approach provides greater speed and numerical stability compared to a solu- tion using elliptic elements. It also avoids the singularity associated with the transition from elliptic to hyperbolic motion. To solve the analogue of Kepler’s equation for the particle’s position in time, we make use of several special functions. Let us begin by defining the c-functions (Stumpff 1962) as a series expan- sion: cn(z) ≡ ∞(cid:88) (17) (−z) j (n + 2 j)! , j=0 which satisfy the recursion relation cn(z) = − z cn+2. 1 n! (18) c(cid:13) 0000 RAS, MNRAS 000, 000–000 WHFast: A fast and unbiased Wisdom-Holman integrator 5 The c-functions are related to trigonometric functions, for example √ √ z√ sin z c0(z) = cos z and c1(z) = , (19) and thus satisfy the following relationships (Mikkola 1997), which are related to the half-angle formula for trigonometric functions: c5(z) = c4(z) = [c5(z/4) + c4(z/4) + c3(z/4)c2(z/4)] 1 16 1 c3(z/4) [1 + c1(z/4).] 8 (20) (21) Values for c0 through c3 are then readily computed from Eq. 18. Next, we introduce the so called G-functions (Stiefel & Scheifele 1971) which in turn depend on the c-functions: Gn(β, X) ≡ Xncn(βX2). The G-functions also satisfy recursion relationships similar to those mentioned above for the c-functions. We can easily calcu- late derivatives of Gn by looking at the series expansion of cn (see Mikkola & Innanen 1999, for details). With this framework, we can now write down the steps needed to find the solution to the Kepler Hamiltonian in compact form. (22) First, we need to calculate the following three quantities from the initial conditions r0, v0: 0 − v2 2M β = r0 η0 = r0 · v0 (24) ζ0 = M − βr0 (25) where r0 = r0 and v0 = v0. Note that the semi-major axis a can be written as a = M/β. (23) Second, we need to solve Kepler’s equation which, using the (26) above notation, takes the form r0X + η0G2(β, X) + ζ0G3(β, X) − dt = 0. We solve this equation for X. This is a non-algebraic (i.e. transcen- dental) equation that we need to solve iteratively, for example using Newton’s method. In Sect. 3.2, we describe our algorithm in detail. Third, having solved Kepler’s Equation, we can calculate the so called Gauss f and g-functions as well as their time derivatives via G2 f = 1 − M r0 g = dt − MG3 where r = r0 + η0G1 + ζ0G2. Note that all the G-functions depend on β and the X value found in the second step. f = − M G1 r0r g = 1 − M G2 r (27) (28) , Fourth, we write the final positions and velocities as a linear transformation of the initial conditions using the Gauss f and g- functions: r = f r0 + gv0 This completes the solution of the Kepler problem. v = f r0 + gv0. (29) To solve for the variational equations, we also make use of the G-functions. Fortunately, we only need to solve Kepler’s equation once (to solve for X). We then get the solution for the variational equations without solving another transcendental equation and thus have only one iteration loop per timestep for both the particle and its variational counterpart. The position and velocity components c(cid:13) 0000 RAS, MNRAS 000, 000–000 of δδδ at the end of the timestep, δr and δv, can be written as δr = f δr0 + g δv0 + r0 δ f + v0 δg (30) f δr0 + g δv0 + r0 δ f + v0 δg, δv = (31) where the variations δ f , δg, δ f and δg can be derived from Eqs. 27- 28 (see Mikkola & Innanen 1999 for the explicit expressions). 2.7 Types of Numerical Errors There are three distinct effects contributing to the energy error of a symplectic integrator (see e.g., Quinn & Tremaine 1990). See also Rein & Spiegel (2015) for a similar discussion for non-symplectic integrators. First, there is an error term associated with the integrator it- self because we are not solving the equations of motion for the Hamiltonian H exactly. For symplectic integrators such as those discussed here, this error term is bound and we call it Ebound. If the mass ratio of the planets to the star is , then the order of this error term is roughly O( dt2) for integrators without symplectic corec- tors and O( dtK) + O(2 dt2) for those with symplectic correctors (see Sect. 2.4). Note that Ebound is independent of time t. Second, there is an error term associated with the finite preci- sion of numbers represented on a computer. We can only represent a small subset of all real numbers exactly in floating-point precision. Thus after every operation such as an addition or multiplication, the computer rounds to a nearby floating-point number. For CPUs and compilers that follow the IEEE754 standard (ISO 2011), we are guaranteed to round to the nearest floating-point number. Thus, if all operations follow the IEEE754 standard, then as long as the al- gorithm itself is unbiased, we expect the error to grow as the square t, where N is the number of timesteps. This is the best behaviour achievable; to do better we would have to move to extended precision or use fewer operations. This fundamental limit is known as Brouwer’s law (Newcomb 1899; Brouwer 1937). root of the number of operations, i.e. Erand ∼ √ N ∼ √ Third, if any parts of the integration algorithm are biased, the errors will be correlated. This leads to a faster long-term energy- error growth than if errors are uncorrelated; it grows linearly with time, i.e. Ebias ∼ N ∼ t. For a given integrator, which of these three error terms domi- nates depends on the nature of the simulation, the timestep, and the total integration time (number of timesteps). 3 IMPROVEMENTS The algorithms we describe in Sect. 2 have been used successfully for many years. In the following, we show how to significantly im- prove the speed and accuracy of the algorithms by taking special care in the implementation of several details, many of which are related to finite floating-point precision on modern computers. For the remainder of this paper, we will assume that we work with a CPU that follows the IEEE 754 standard for floating-point arithmetic. Most importantly, we assume that all floating-point op- erations follow the rounding to nearest, ties to even rule (ISO 2011). What follows is in principle applicable to any precision. However, we work exclusively in double floating-point precision (64 bit) which is used on almost all modern CPUs. 6 Hanno Rein and Daniel Tamayo 3.1 Jacobi Coordinate Transformations The evolution under the effect of the interaction Hamiltonian is most efficiently done in Cartesian coordinates. On the other hand, the evolution of the Kepler Hamiltonian is easier in Jacobi coordi- nates. We thus need an efficient way to convert to and from Jacobi coordinates. Luckily, the conversion from Cartesian to Jacobi coordinates and back can be done efficiently in O(N). We construct the algo- rithms from the definitions above and list them here in pseudo code. As before, primes denote Jacobi coordinates, Note that these algo- rithms work even if some of the bodies are test particles with mi = 0 (for i (cid:44) 0). To convert from Cartesian to Jacobi coordinates: R ← m0 · r0 for i ← 1, N − 1 do i ← ri − R/Mi−1 r(cid:48) R ← R · (1 + mi/Mi−1) + mi · r(cid:48) 0 ← R/MN−1 r(cid:48) i (cid:46) This is the centre-of-mass. Similarly, we construct the algorithm to convert back from Jacobi to Cartesian coordinates as follows: (cid:46) Centre of mass. (cid:46) Loop is in reverse order. 0 · MN−1 R ← r(cid:48) for i ← N − 1, 1 do R ← (R − mi · r(cid:48) ri ← r(cid:48) + R R ← R · Mi−1 r0 ← R/m0 i i)/Mi (cid:46) Setting the coordinate of the 0-th particle. We thoroughly tested the conversions to and from Jacobi coordi- nates to ensure they are unbiased. This task turns out to be much harder than we naıvely expected. As an example, consider the fol- lowing algorithm which is formally equivalent to the above but nu- merically much less stable. (cid:46) Loop is in reverse order. (cid:46) r(cid:48) 0 is the centre-of-mass. R ← 0 for i ← N − 1, 1 do 0 + Mi−1/Mi · r(cid:48) ri ← r(cid:48) R ← R + mi/Mi · r(cid:48) 0 − R i r0 ← r(cid:48) i − R (cid:46) Setting the coordinate of the 0-th particle. In the above algorithm, we access r(cid:48) 0 multiple times and have to do a subtraction in the last step. This significantly promotes error propagation and leads to floating-point errors that can be orders of magnitudes higher than in the other implementation. After many timesteps, this leads to a linear secular growth in the energy error. 3.2 Implementation of Newton’s Method To solve Kepler’s equation (Eq. 26) for X, we need to use an iter- ative scheme. We now describe our implementation of Newton’s method in floating-point arithmetic. The straightforward imple- mentation is an iteration loop that terminates when the change to X is small, e.g., X ← initial guess repeat dX ← − f (X)/ f (cid:48)(X) X ← X + dX until dX/X < eps. Here, eps is a small number just above machine precision, typi- cally eps ∼ 10−15. We use a different implementation of Newton’s method that is both faster and more accurate, despite the fact that it is algebraically equivalent to the above implementation. (cid:46) Any number different from X works. X ← initial guess Xprev1 ← NaN repeat Xprev2 ← Xprev1 Xprev1 ← X X ← (X · f (cid:48)(X) − f (X))/ f (cid:48)(X) until X = Xprev1 or X = Xprev2 Note that the equal sign in the above breakout condition is evaluated in floating-point precision. In comparison to the first algorithm, at each iteration step we test whether the iteration has converged by a simple comparison rather than by a slow division and absolute- value operation. We keep track of two previous values instead of just one be- cause for certain initial conditions, the iteration can cycle indefi- nitely between two nearby floating-point numbers and not converge to a single floating-point number. Our implementation thus ensures that the value of X is more accurately calculated than in the straightforward implementation using a heuristic value of eps. A further advantage of rewriting Newton’s method in the above form is that the term on the right- hand-side of the last line can be simplified significantly for the Ke- pler problem, giving: X ← X(η0G1 + ζ0G2) − η0G2 − ζ0G3 + dt r0 + η0G1 + ζ0G2 (32) where the G’s on the right-hand-side all depend on X and β (see Eq. 22). We also experimented with higher-order generalizations of Newton’s method (Householder’s methods). For typical cases where the orbits are not extremely elliptical (e (cid:46) 0.99) and the timestep is much smaller than the shortest orbital period, we found Newton’s method to always be fastest. This is because when the value and derivatives of the function are easily evaluated, the pre- cision gain from these higher-order methods does not compensate for the increased computation cost of each iteration. In other words, while higher-order methods will converge in fewer iterations than Newton’s method, the overall computation time is longer. At large eccentricities and long timesteps, the G-function evaluations be- come expensive (one must recursively apply the quarter-angle for- mulas described in Sect. 3.5), and higher-order methods are help- ful. For large eccentricities we use a higher order method described in detail in Sect. 3.4. To safeguard against rare cases where New- ton’s method might fail, we also implemented a failsafe bisection method. We find that the bisection method is only triggered when the timestep is comparable to the orbital period. dt(cid:48) r = dt · (cid:104)r−1(cid:105) 3.3 The Initial Guess for Kepler’s Equation: Short Timesteps The quantity X in Eq. 32 can also be expressed as (cid:90) t0+dt X = (33) where t0 is the time at the beginning of the timestep, and (cid:104)r−1(cid:105) is the time-averaged value of r−1 over the interval [t0, t0 + dt]. Thus, if the orbit’s eccentricity e is low, or more generally if the timestep c(cid:13) 0000 RAS, MNRAS 000, 000–000 t0 WHFast: A fast and unbiased Wisdom-Holman integrator 7 is short enough that the orbital radius does not vary much, then X ≈ dt/r0. The troublesome cases are highly eccentric orbits near pericentre where the radius changes rapidly. For such cases, the radius varies by a factor of 1 + e ≈ 2 from pericentre to a true anomaly of 90◦. We can therefore estimate the timescale over which the orbital radius varies near pericentre as (cid:32) 1 − e (cid:33)1/2 ∼ (1 − e)3/2 na 1 + e n a(1 − e) Tchar = q vq = , (34) (cid:33) (cid:32) 1 − 1 2 η X = · dt r0 where q is the pericentre distance, vq is the speed at pericentre and n is the mean motion. Thus, if one does not resolve pericentre pas- sages (i.e., n dt = ∆M (cid:38) (1 − e)3/2), X will differ from dt/r0 near pericentre (but may nevertheless conform to the simple approxima- tion at apocentre where the body moves slowly). More quantitatively, one can non-dimensionalize Eq. 32, set- ting X = r0X/dt. One can then solve the equation perturbatively, assuming the deviations from X = 1 are small. This procedure re- quires that the following three non-dimensional parameters in the equation also be much smaller than unity, χ ≡ βdt2 (35) r2 0 One can show that when our heuristic estimate ∆M (cid:28) (1 − e)3/2 is satisfied, χ, η, ζ (cid:28) 1. In this case, one can extend the solution of Eq. 32 to higher order. For the initial guess in our algorithm, we go up to second order ζ ≡ ζ0dt2 r3 0 η ≡ η0dt r2 0 . . (36) We experimented with higher-order initial guesses (see Danby 1987 for explicit expressions), but found these to be slower, even for small eccentricities and timesteps. This can again be attributed to the computational efficiency of each iteration of Newton’s method. 3.4 Large Eccentricities and Timesteps The previous two sections describe an optimized algorithm for solving Kepler’s equation when the timestep and eccentricities are low. We have also developed an improved handling of high- eccentricity/long-timestep cases. In this regime, both the solver and initial guess should be modified. Like previous authors (Conway 1986; Danby 1987), we found the root-finding method of Laguerre-Conway to be most stable. However, unlike Danby (1987), who finds the method to always converge (presumably using comparatively small timesteps), we of- ten have to resort to bisection when the timestep is comparable to the orbital period. Of course, such long timesteps should not be chosen anyway, since they poorly sample inter-planet interactions, and are more susceptible to timestep resonances (Wisdom & Hol- man 1992; Touma & Wisdom 1993; Rauch & Holman 1999). We also had to modify the breakout condition used for New- ton’s method. While the Laguerre-Conway algorithm sometimes also bounces between two floating-point values once it has con- verged, in this regime the method often executes larger-period cy- cles (e.g., it will periodically repeat the last eight floating-point numbers). We therefore chose to store the values from each iter- ation and exit the loop whenever a result was repeated. One way to determine which solver should be used is to check whether dt is smaller than Tchar (Eq. 34). However, because Tchar c(cid:13) 0000 RAS, MNRAS 000, 000–000 is expensive to compute from r0 and v0, we instead check how much the first iteration of Newton’s method deviates from the ini- tial guess, as a fraction of 2πβ−1/2. The latter is a natural quantity to compare against since it is the value of X when the timestep is equal to the orbital period. We found a threshold of ∼ 1% to strike a good balance over a wide parameter range in timestep/eccentricity space, though the algorithm’s speed is not particularly sensitive to the exact value adopted. Finally, the method can be sped up in this regime with an improved initial guess for X, since dt/r0 in Eq. 36 blows up near pericentre as the eccentricity gets large. Danby & Burkardt (1983) provide a widely used initial guess using classical orbital elements but, to our knowledge, no comparably simple initial guess has been found for universal variables. In this high-eccentricity / long timestep regime, most existing methods using universal variables choose to make the expensive conversion to orbital elements and use Danby’s guess. We instead observed that because (cid:104)r−1(cid:105) = a−1 over one orbital period, X = dt/a for a timestep of one orbit. We find that over a relevant parameter range with timesteps logarithmically spaced between 0.03 and 1 orbital periods, and eccentricities between 0.999 and 0.9999, our improved guess is faster than converting to orbital elements and us- ing Danby’s by ≈ 30%. In a manner analogous to that described in the previous section, we also solved Eq. 32 perturbatively around X = dt/a = β dt/M in the regime χ (cid:29) 1, but we found the second-order solution to be a slower initial guess than the simple X = β dt/M. 3.5 Implementation of c-functions Finding a solution to Kepler’s equation is done iteratively and is thus the most expensive step in solving the Kepler problem. The iteration itself involves the calculation of multiple G-functions, which in turn require the calculation of c-functions. Thus, it is par- ticularly important to optimize these functions for both speed and accuracy. When calculating chaos indicators, we need c0, c1, c2, c3, c4 and c5. If we are not integrating the variational equations, we only need c0, c1, c2 and c3. We first ensure that z is smaller than 0.1 to guarantee that the series expansion of c in Eq. 17 converges. We do this by dividing z repeatedly by 4. Note that divisions by powers of 2 are fast and ex- act in floating-point arithmetic. To calculate the series expansion, we need an inverse factorial for every term. Calculating this in- verse factorial by multiplying floating-point numbers and then im- plementing a floating-point division would be very slow. We found that the fastest way to calculate the inverse factorial is to use a sim- ple lookup table. We checked that the series expansions of the c- functions converge very quickly for small z and thus we only store inverse factorials up to 1/34! in the lookup table. Any larger facto- rial would contribute less than one part in 1016 to the sum and can thus be neglected (as we work in double floating-point precision). We always calculate the first two terms in the series expan- sion. We then enter a loop and add more terms until the result no longer changes. Because z is small and the inverse factorials de- crease quickly, we are assured that the series will converge to a sin- gle floating-point number. This allows us to simply check whether the value changes from one iteration to the next, which is much faster than evaluating relative changes (cf. Sect. 3.2). Once the c-functions are calculated for the small z value, we 8 Hanno Rein and Daniel Tamayo use the relations in Eqs. 20-21 with Eq. 18 to calculate the c- functions for the original z value. Because this algorithm is an integral part of the integrator, we list the function to calculate c(z) in pseudo code: (cid:46) Counter for quarter-angle formula. (cid:46) Ensure that z is small. (cid:46) Hard coded first two terms for c4. (cid:46) p will the (−z) j factor in the loop. (cid:46) Third term in c4 contains factor 1 8! . (cid:46) Keep old value to check for convergence. (cid:46) 1/k! comes from lookup table. (cid:46) Converged? (cid:46) Use Eq. 18 to get c3, c2 and c1. (cid:46) Apply quarter angle formula n times. n ← 0 while z > 0.1 do z ← z/4 n ← n + 1 4! − z · 1 5! − z · 1 6! 7! c4 ← 1 c5 ← 1 ¯z ← −z p ← ¯z k ← 8 repeat k! k! c4,prev ← c4 p ← p · ¯z c4 ← c4 + p · 1 k ← k + 1 c5 ← c5 + p · 1 k ← k + 1 until c4 = c4,prev 6 − z · c5 c3 ← 1 2 − z · c4 c2 ← 1 c1 ← 1 − z · c3 while n > 0 do z ← 4 · z 16 · (c5 + c4 + c3 + c2) c5 ← 1 8 · c3 · (1 − c1) c4 ← 1 c3 ← 1 6 − z · c5 2 − z · c4 c2 ← 1 c1 ← 1 − z · c3 n ← n − 1 c0 ← 1 − z · c2 3.7 A full integration in Jacobi coordinates The algorithms to convert to and from Jacobi coordinates that we describe in Sect. 3.1 are unbiased and fast. Nevertheless, we aim to avoid as many conversion as possible. As it turns out, we can reduce the number of conversions per timestep to two, one for the positions from Jacobi coordinates to the inertial frame, and one for the accelerations from the inertial frame to Jacobi accelerations. But note that this is only possible under the following assumptions: 1) the particle position and velocities are not changed in-between timesteps, e.g. manually by the user or by collisions, 2) outputs are not required at every timestep, 3) variational equations are not integrated, 4) no additional velocity- dependent forces are present. In such a case, an integration starting from an arbitrary inertial frame is achieved as follows: calculate Jacobi coordinates drift all particles under HKepler for half a timestep, dt/2 while t < tmax do calculate 1st part of HInteraction in Jacobi coordinates update positions in the inertial frame calculate 2nd part of HInteraction in inertial frame convert accelerations from 2nd part to Jacobi accelerations apply kick from Jacobi accelerations to Jacobi velocities if not last timestep then drift all particles under HKepler for a full timestep dt drift all particles under HKepler for half a timestep dt/2 update both positions and velocities in the inertial frame. Note that we never update the velocities in the inertial frame until the end of the simulation (or when an output is needed). We only convert the positions and velocities to Jacobi coordinates at the very beginning and not at every timestep. Besides the obvious speed- up, avoiding to go back and fourth between different coordinate systems reduces the build-up of round-off errors and thus makes the integrator more robust. 3.6 Implementation of Gauss f and g-functions 3.8 LCN calculation The precise implementation of Gauss f and g functions matters for long term integrations. The straightforward implementation follow- ing (Mikkola & Innanen 1999) leads to the f and g-functions in Eq. 27. Note that for timesteps smaller than half an orbital period, the term MG2/r0 in f is small compared to the first term (which is just 1). The same argument holds true for g. We can define new f and g-functions f = −M G2 g = dt − MG3 (cid:17) This allows us to rewrite the last step in solving the Kepler problem as f = − M G1 r0r g = − M G2 r (37) (38) (cid:17) r0 (cid:16) f r0 + gv0 r = (cid:16) f r0 + gv0 v = + v0. (39) + r0 . Although this step is algebraically equivalent to the original Eq. 29, we achieve higher precision. The reason is that we can now ensure that the small quantities in brackets are summed before they are added to the larger quantity (the initial value). We implement the same trick for the variational equations, Eqs. 30 and 31 To calculate the Lyapunov characteristic number and the Lyapunov timescale we need to perform a linear least square fit to the function Y(t). Thus we need the mean and the covariance of Y(t). Storing all previous values of Y(t) just to calculate its mean and covariance is inefficient. We therefore implement an efficient one-pass method described by P´ebay (2008). This method lets us calculate the LCN at every timestep in O(1) and has the further advantage of being numerically more robust than the standard implementation. 4 NUMERICAL RESULTS In this section, we test the speed, accuracy and numerical stability of WHFast and compare it to other publicly available and widely used integrators. We begin by briefly defining our nomenclature for these other integrators and summarizing their properties. MERCURY is a mixed-variable symplectic integrator imple- mented in fortran and provided by the MERCURY package (Cham- bers & Migliorini 1997). This Wisdom-Holman style integrator uses high-order symplectic correctors. We directly call the fortran code without any modifications. c(cid:13) 0000 RAS, MNRAS 000, 000–000 WHFast: A fast and unbiased Wisdom-Holman integrator 9 Figure 1. Tests of the Kepler-solver. Simulation with two bodies, integrated for 100 orbits with varying eccentricity and timestep. Left column: results using the standard WH integrator. Right column: results using our new WHFast integrator. Top row: relative energy error at the end of the simulation. Middle row: sign of the energy error at the end of the simulation. Bottom row: average runtime for one timestep. c(cid:13) 0000 RAS, MNRAS 000, 000–000 876543210eccentricity, log10(1−e)WHWHFAST-NOCOR876543210eccentricity, log10(1−e)3.02.52.01.51.00.5timestep, log10(dt/torb)876543210eccentricity, log10(1−e)3.02.52.01.51.00.5timestep, log10(dt/torb)10-1610-1510-1410-1310-1210-1110-1010-910-810-710-6relative energy error101sign of energy error0.000.150.300.450.600.750.901.05runtime per timestep [µs]10 Hanno Rein and Daniel Tamayo SWIFTER-WHM is again a classical 2nd-order Wisdom-Holman integrator without symplectic correctors (Wisdom & Holman 1991). We use the integrator provided by the SWIFTER package. It is implemented in fortran and we directly call the SWIFTER exe- cutable without any modifications. SWIFTER-HELIO is also 2nd-order symplectic integrator with- out symplectic correctors (Duncan et al. 1998). It uses democratic heliocentric coordinates. We again use the integrator provided by the SWIFTER package. It is implemented in fortran and we directly call the SWIFTER executable without any modifications. SWIFTER-TU4 is a 4th-order symplectic integrator. It is not a Wisdom-Holman integrator but splits the Hamiltonian in kinetic and potential terms (Gladman et al. 1991). We also use the integra- tor provided by the SWIFTER package. It is implemented in fortran and we directly call the SWIFTER executable without any modifica- tions. For a more direct comparison, we also make use of an integra- tor that we simply refer to as WH. It is based on the SWIFTER-WHM integrator in SWIFTER but ported to C and available in the REBOUND (Rein & Liu 2012) package. Like the SWIFTER-WHM integrator, it is a symplectic integrator that works in the heliocentric frame, and does not implement any symplectic correctors. Note that this is not the original integrator used by Wisdom & Holman (1991), which is not publicly available. WHFast is C99 compliant. The C99 standard guarantees that floating point operations are not re-ordered by the compiler (un- less one of the fast-math options is turned on). Because of that, the final positions and velocities of particles agree down to the last bit across different platforms. This makes WHFast platform inde- pendent and the simulation results reproducible. We verified this on different architectures (Linux, MacOSX), different CPUs (Intel Core i5-3427U, Intel Xeon E5-2697 v2, Intel Xeon E5-2620 v3) and different compilers (Apple LLVM 6.1.0, gcc 4.4.7). Figure 2. Relative energy error in simulations of the outer Solar System after 1000 Jupiter orbits as a function of the number of steps per orbit. is biased over large regions of the parameter space (there are large blue and red areas in the second row). On the other hand, WHFast has a random energy error throughout the parameter space. Having a biased energy error will lead to a long-term linear growth of the energy error (see below). In the entire parameter space explored, WHFast requires less time to complete a timestep than WH. The speed-up is typically be- tween 20% and 100%. For the integrations performed in this sec- tion, we convert to and from Jacobi coordinates at every timestep to provide a fair comparison. Thus, the speed-up and the energy- conservation properties of WHFast are in fact even better than shown here in any actual production run (see Sect. 3.7). 4.1 Two-body Kepler Solver 4.2 Short Term Energy Conservation The kernel of every Wisdom-Holman integrator is the Kepler solver. We describe our implementation in detail in Sections 3.2- 3.6. Here, we test the Kepler solver using a two-body problem. The two body problem is invariant with respect to rescaling of the total mass, the mass ratio, the value of the gravitational con- stant and the orbital period. What does matter is the eccentric- ity of the orbit and the ratio of the timestep to the orbital period. We thus scan the parameter space in those two dimensions by in- tegrating two bodies for 100 orbital periods. We explore an ex- tremely wide parameter space. The eccentricities range from zero to 0.999 999 99 = 1− 10−8. The range of timesteps goes from 0.1% of the orbital period all the way up to one orbital period. Fig. 1 shows the performance of WHFast (right column) com- pared to WH (left column). The top row shows the absolute value of the relative energy error at the end of the simulation. The middle row shows the sign of the energy error. The bottom row shows the average runtime for a single timestep. The vertical lines visible in the top row correspond to timestep resonances (Wisdom & Holman 1992; Touma & Wisdom 1993; Rauch & Holman 1999). One can see that WHFast is significantly more accurate than the standard WH integrator for the most important parts of param- eter space (eccentricities less than ∼ 0.99). The relative energy is conserved better by two to three orders of magnitude. Most im- portantly, note that the energy error in the standard WH integrator To compare the accuracy of the different integrators in a realistic test case, we run simulations of the outer Solar System for one thousand Jupiter orbits (12 000 years). We include the Sun and four massive bodies with approximate initial conditions corresponding to those of Jupiter, Saturn, Uranus and Neptune. In each simulation the initial conditions and masses are randomly perturbed by 0.1%. In Fig. 2, we plot the relative energy errors at the end of the simu- lation as a function of the number of timesteps imposed per Jupiter orbit. One can see that all the integrators except SWIFTER-TU4 are second-order schemes. For timesteps between 20% and 0.1% of the orbital period of Jupiter (50 to 1000 timesteps per orbit), their error decreases quadratically with decreasing timesteps. This is the error term Ebound introduced in Sect. 2.7. However, decreasing the timestep also increases the number of floating point operations. There will therefore be a timestep value at which the numerical round-off error dominates over the error as- sociated with the symplectic method itself Ebound. For that reason we find that for small timesteps, less than 0.1% of the shortest or- bital period, the errors of all integrators rise instead of decreasing further. Thus there is an optimum timestep dtopt that yields the min- imum energy error. This optimum timestep depends on the length of the integration and will be larger for longer simulations. In Fig. 2 one can see that the errors of WH, SWIFTER-WHM, c(cid:13) 0000 RAS, MNRAS 000, 000–000 102103104105106steps per orbit10-1610-1510-1410-1310-1210-1110-1010-910-810-710-610-5relative energy errorWHSWIFTER-WHMSWIFTER-TU4SWIFTER-HELIOMERCURYWHFAST-NOCORWHFASTWHFast: A fast and unbiased Wisdom-Holman integrator 11 SWIFTER-HELIO and MERCURY rise very rapidly after reaching dtopt, scaling as at least dt−2 for the first decade. The optimum timestep for WHFast is roughly 0.1% of the shortest orbital period. However, WHFast’s error grows much more slowly with decreasing timestep than that of the other second-order integrators. In fact, the error is dominated by Erand and thus fol- lows dt−1/2 as the number of timesteps Nsteps increases as ∼ dt−1 if we keep the total integration time constant. Thus the behaviour of WHFast in Fig. 2 for small timesteps can be seen as the first indica- tion that WHFast follows Brouwer’s law (see Sects. 2.7 and 4.4). The SWIFTER-TU4 integrator is the only other integrator we tested that seems to follow Brouwer’s law, but it performs poorly at large timesteps. This is expected, since unlike the other integra- tors, SWIFTER-TU4 does not assume a Keplerian splitting and must therefore take smaller timesteps to accurately reproduce the orbital motions. Integrators with symplectic correctors, MERCURY and WHFast, perform significantly better for long timesteps. Their energy con- servation is three orders of magnitude better (Ebound is three orders of magnitude smaller) compared to integrators without symplec- tic correctors. This is due to the mass ratio of Jupiter and the Sun being roughly 10−3. The order of the symplectic corrector is not very important for relatively high mass ratios such as these, i.e. a fifth-order symplectic corrector performs as well as an 11th-order one. For much smaller mass ratios (when the mass ratio is less than the timestep ratio), higher-order symplectic correctors are advanta- geous. Note that dtopt for almost all of the integrators is 10−3 orbital periods of Jupiter, i.e. 4 days. This is significant because Mercury’s orbital period is 88 days. Thus if we included Mercury in our sim- ulation, we would be very restricted in our timestep choice. We need more than 20 timesteps (dt ≈ 4 days) to resolve Mercury’s orbit accurately. However, if we choose choose a timestep smaller than 4 days, we start to accumulate errors in the outer Solar Sys- tem. It is worth reiterating that the simulations shown in Fig. 2 all ran for only 1000 orbits. If we ran a longer simulation with the same timestep, we would have more timesteps and thus accumu- late more round-off errors by the end of the simulation. One can therefore reach better energy conservation with a longer timestep. In other words, dtopt is larger for longer integration times. 4.3 Speed Comparison We run the same simulations as in Sect. 4.2 to compare the speed of the different integrators. Fig. 3 shows the relative energy error as a function of runtime. The results show that no matter what the desired energy error is, WHFast is the fastest integrator. In the large timestep limit, the speed-up compared to MERCURY is roughly a fac- tor of 5. In the small timestep limit, dt < dtopt, we can only compare WHFast to SWIFTER-TU4, as all the other integrators’ errors are significantly larger (by 4 to 5 orders of magnitude) due to numer- ical roundoff errors (see below). SWIFTER-TU4 is as fast for small timesteps as WHFast but, as noted above, is unsuitable for large timesteps since it is not a Wisdom-Holman integrator. It is only shown here as a comparison. c(cid:13) 0000 RAS, MNRAS 000, 000–000 Figure 3. Relative energy error in simulations of the outer Solar System after 1000 Jupiter orbits as a function of run time. 4.4 Long Term Energy Conservation Let us finally address the most important benchmark, the long term energy conservation properties of WHFast compared to other inte- grators in a real world test case. In this section we only study the energy error, but other conserved properties like the angular mo- mentum behave the same way. In Fig. 4, we show the time evolu- tion of the relative energy error in a simulation of the outer Solar System. As in Sect. 4.2, we include the Sun and four massive bod- ies with approximate initial conditions corresponding to those of Jupiter, Saturn, Uranus and Neptune. The timestep for all simula- tions is 1.5 days. Note that this timestep is smaller than what one would typically choose for this kind of integration. However, with a 1.5 day timestep we reach machine precision for integrators that use symplectic correctors, allowing us to better quantify the long- term behaviour of WHFast. All the effects we discuss here are also present in simulations with longer timesteps, but they would man- ifest themselves in the relative energy error at a later time. We run four simulations for each integrator and randomly perturb the ini- tial conditions and masses by 0.1% in each simulation. We plot the individual simulations as thin lines, and the average error as a bold line. The lower relative energy bound is set by machine precision for all integrators, roughly 10−16. WHFast and MERCURY, the inte- grators in our sample that have symplectic correctors, almost reach this limit early on in the simulation. The bound energy error Ebound is approximately 10−14. The integrators without symplectic correc- tors, SWIFTER-WHM and WH have an energy error roughly three order of magnitudes higher Ebound ≈ 10−10.5. From Fig. 4 it is clear that the integrators MERCURY, WH and SWIFTER-WHM show a linear behaviour in the energy error at late times. This is due to the term Ebias. The Ebias term already domi- nates at early times (after 100 Jupiter orbits) for MERCURY because the symplectic correctors lower the value of Ebound. For WH and SWIFTER-WHM the Ebias term dominates after 10 000 Jupiter orbits. This result shows that one or more steps in these integration algo- rithms are biased. We found that the two main contributions were the inaccurate implementation of the rootfinder for Kepler’s equa- tion and the conversions to and from Jacobi coordinates. In WHFast, 10-1100101102103runtime [s]10-1610-1510-1410-1310-1210-1110-1010-910-810-710-610-5relative energy errorfasterWHSWIFTER-WHMSWIFTER-TU4SWIFTER-HELIOMERCURYWHFAST-NOCORWHFAST12 Hanno Rein and Daniel Tamayo Figure 4. Relative energy error in simulations of the outer Solar System as a function of time for different symplectic integrators. The timestep for all simulations is 1.5 days. Ebias is absent, showing that its implementation is completely unbi- ased. Since all integrators are implemented in double floating-point precision and use the same timestep, they all have roughly the same error term Erand. However, it is only visible in Fig. 4 for the WHFast integrator. For all other integrators the linearly growing term Ebias dominates over Erand. If we increase the timestep, the linear error growth will show up at a later time because Ebound will be larger. However, it is still present at all times. Let us think of a symplectic integrator as an ex- act integrator for a perturbed Hamiltonian H with high frequency terms added compared to H in Eq. 1, see e.g. Wisdom et al. (1996). H, let us call this Then the quantity related to the energy error for E, should be conserved exactly at all times (that is the idea of a symplectic integrator). However, if the implementation is biased, E will undergo a linear growth at all times. With WHFast, we im- prove the conservation of E by many orders of magnitude in any integration, regardless of timestep. This difference could have important implication for the dy- namical evolution of the system and could for example push it from a stable to an unstable region of parameter space. We plan to study the effect of different integrators on systems near a chaotic/non- chaotic separatrix in a follow up paper. 5 CONCLUSIONS In this paper, we presented WHFast, a new implementation of a symplectic Wisdom-Holman integrator. Key advantages and im- provements over other publicly available implementations of sym- plectic integrators are: WHFast is faster by a factor of 1.5 to 5. Of that, a 50% speedup comes from the improved Kepler solver, where we use a fast con- vergence criteria for Newton’s method and an efficient implemen- tation of c and G-functions. The remainder of the speedup is due to combining drift steps at the end and beginning of each timestep and to only converting to and from Jacobi coordinates when needed. The Kepler solver is more accurate and unbiased. We achieve this thanks to improvements to the convergence criteria in Newton’s method, a Laguerre-Conway solver for highly eccentric orbits with long timesteps, the high accuracy implementations of the c and G- functions and a careful ordering of floating-point operations. We remove the secular energy error that grows linearly with integration time. This is due to two improvements. First, the un- biased Kepler solver. Second, the improved and also unbiased co- ordinate transformations to and from Jacobi coordinates. To our knowledge, WHFast is the first publicly available implementation of a Wisdom-Holman integrator that follows Brouwer’s law over long timescales for small enough timesteps and does not show a linear growth in the energy error. We implement variational equations that allow us to compute the Lyapunov timescale and the MEGNO. Our algorithm to calcu- late the Lyapunov timescale uses a numerically stable algorithm that is based on a one-pass covariance filter. The variational equa- tions do not require us to solve Kepler’s equation and are thus very inexpensive to calculate. Symplectic correctors of order 3, 5, 7, and 11 are imple- mented. These symplectic corrector allow for high-accuracy simu- lations of systems with small mass ratios. Even for relatively mas- sive planets like those in the Solar System, symplectic correctors achieve an improvement of three orders of magnitude. For long in- tegrations, the performance cost of symplectic correctors is negli- gible and so our default setting uses an 11th-order corrector. WHFast lets the centre-of-mass move freely during an integra- tion. We integrate an additional degree of freedom in order for our integrator to work in any inertial frame, i.e. one is not restricted to the heliocentric or barycentric frame. Additionally, we do not tie our implementation to a specific choice of units. The integrator is available as an easy to use python module. The module works on both python 2 and 3. It can be installed on most Unix and MacOS systems with a single command: pip install rebound The following python script imports the rebound module, adds par- c(cid:13) 0000 RAS, MNRAS 000, 000–000 100101102103104105106107time [orbits]10-1610-1510-1410-1310-1210-1110-1010-910-810-710-610-5relative energy errorMERCURYWHFAST-NOCORWHWHFASTSWIFTER-WHMWHFast: A fast and unbiased Wisdom-Holman integrator 13 Mikkola, S. & Palmer, P. 2000, Celestial Mechanics and Dynam- ical Astronomy, 77, 305 Murray, C. D. & Dermott, S. F. 2000, Solar System Dynamics (Cambridge University Press) Newcomb, S. 1899, Astronomische Nachrichten, 148, 321 P´ebay, P. 2008, Sandia Report SAND2008-6212, Sandia National Laboratories Plummer, H. C. K. 1918, An introductory treatise on dynamical astronomy (Cambridge, University Press) Quinn, T. & Tremaine, S. 1990, AJ, 99, 1016 Rauch, K. P. & Holman, M. 1999, The Astronomical Journal, 117, 1087 Rein, H. & Liu, S.-F. 2012, A&A, 537, A128 Rein, H. & Spiegel, D. S. 2015, MNRAS, 446, 1424 Saha, P. & Tremaine, S. 1992, AJ, 104, 1633 Stiefel, E. L. & Scheifele, G. 1971, Linear and Regular Celes- tial Mechanics. (Die Grundlehren d. math. Wissenschaften, Band 174) (Berlin/Heidelberg/New York: Springer) Stumpff, K. 1962, in Himmelsmechanik, Band I (Berlin: VEB Deutscher Verlag der Wissenschaften) Sussman, G. J. & Wisdom, J. 2001, Structure and interpretation of classical mechanics (MIT Press) Tancredi, G., S´anchez, A., & Roig, F. 2001, AJ, 121, 1171 Tittemore, W. C. & Wisdom, J. 1989, Icarus, 78, 63 Touma, J. & Wisdom, J. 1993, Science, 259, 1294 Wisdom, J. 2006, AJ, 131, 2294 Wisdom, J. & Holman, M. 1991, AJ, 102, 1528 —. 1992, AJ, 104, 2022 Wisdom, J., Holman, M., & Touma, J. 1996, Fields Institute Com- munications, Vol. 10, p. 217, 10, 217 Wisdom, J. L. 1981, PhD thesis, California Institute of Technol- ogy, Pasadena. ticles to the simulation, selects an integrator and timestep and runs the integration. import rebound rebound . add (m =1) rebound . add (m =0.001 , a =1.) rebound . add (m =0.001 , a=2. , e =0.1) rebound . integrator = ’whfast ’ rebound .dt = 0.01 rebound . integrate (6.2831) More complicated examples and the source code of WHFast (writ- ten in C, compliant with the C99 standard) can be found in the REBOUND package. REBOUND includes several other integrators, col- lision detection algorithms, a gravity tree code and much more. The REBOUND git repository is hosted at https://github.com/ hannorein/rebound. We also provide an experimental hybrid integrator for sim- ulations in which close encounters occur. The hybrid integrator switches over to a high-order non-symplectic integrator (IAS15, Rein & Spiegel 2015) during a close encounter. A detailed discus- sion of this integrator and its properties will be given in a follow-up paper. We hope that with the speed and accuracy improvements, WHFast will become the go-to integrator package for short and long-term orbit simulations of planetary systems. ACKNOWLEDGMENTS This research has been supported by the NSERC Discovery Grant RGPIN-2014-04553. We thank Wayne Enright, Philip Sharp and Scott Tremaine for stimulating discussions and Jack Wisdom for a helpful referee report. REFERENCES Brouwer, D. 1937, AJ, 46, 149 Chambers, J. E. & Migliorini, F. 1997, in Bulletin of the Ameri- can Astronomical Society, Vol. 29, AAS/Division for Planetary Sciences Meeting Abstracts #29, 1024 Cincotta, P., Giordano, C., & Sim´o, C. 2003, Physica D: Nonlinear Phenomena, 182, 151 Conway, B. A. 1986, Celestial mechanics, 39, 199 Danby, J. & Burkardt, T. 1983, Celestial Mechanics, 31, 95 Danby, J. M. A. 1987, Celestial Mechanics, 40, 303 Duncan, M. J., Levison, H. F., & Lee, M. H. 1998, AJ, 116, 2067 Gladman, B., Duncan, M., & Candy, J. 1991, Celestial Mechanics and Dynamical Astronomy, 52, 221 Hinse, T. C., Christou, A. A., Alvarellos, J. L. A., & Go´zdziewski, K. 2010, MNRAS, 404, 837 ISO. 2011, ISO/IEC/IEEE 60559:2011 Information technology — Microprocessor Systems — Floating-Point arithmetic (New York, NY, USA: IEEE), 58 Kinoshita, H., Yoshida, H., & Nakai, H. 1991, Celestial Mechan- ics and Dynamical Astronomy, 50, 59 Mikkola, S. 1997, Celestial Mechanics and Dynamical Astron- omy, 67, 145 Mikkola, S. & Innanen, K. 1999, Celestial Mechanics and Dy- namical Astronomy, 74, 59 c(cid:13) 0000 RAS, MNRAS 000, 000–000
1108.4836	1	1108	2011-08-24T13:26:38	The Canada-France Ecliptic Plane Survey - Full Data Release: The orbital structure of the Kuiper belt	[ "astro-ph.EP" ]	We report the orbital distribution of the trans-neptunian objects (TNOs) discovered during the Canada-France Ecliptic Plane Survey, whose discovery phase ran from early 2003 until early 2007. The follow-up observations started just after the first discoveries and extended until late 2009. We obtained characterized observations of 321 sq.deg. of sky to depths in the range g ~ 23.5--24.4 AB mag. We provide a database of 169 TNOs with high-precision dynamical classification and known discovery efficiency. Using this database, we find that the classical belt is a complex region with sub-structures that go beyond the usual splitting of inner (interior to 3:2 mean-motion resonance [MMR]), outer (exterior to 2:1 MMR), and main (in between). The main classical belt (a=40--47 AU) needs to be modeled with at least three components: the `hot' component with a wide inclination distribution and two `cold' components (stirred and kernel) with much narrower inclination distributions. The hot component must have a significantly shallower absolute magnitude (Hg) distribution than the other two components. With 95% confidence, there are 8000+1800-1600 objects in the main belt with Hg <= 8.0, of which 50% are from the hot component, 40% from the stirred component and 10% from the kernel; the hot component's fraction drops rapidly with increasing Hg. Because of this, the apparent population fractions depend on the depth and ecliptic latitude of a trans-neptunian survey. The stirred and kernel components are limited to only a portion of the main belt, while we find that the hot component is consistent with a smooth extension throughout the inner, main and outer regions of the classical belt; the inner and outer belts are consistent with containing only hot-component objects. The Hg <= 8.0 TNO population estimates are 400 for the inner belt and 10,000 for the outer belt within a factor of two.	astro-ph.EP	astro-ph	1 1 : v i X r a to appear in the Astronomical Journal (submitted 25 May 2011) The Canada-France Ecliptic Plane Survey - Full Data Release: The orbital structure of the Kuiper belt1 J-M. Petit2,4, J.J. Kavelaars3, B.J. Gladman4, R.L. Jones3,4, J.Wm. Parker5, C. Van Laerhoven4,6, P. Nicholson7, G. Mars8, P. Rousselot2, O. Mousis2, B. Marsden9,12 A. Bieryla5, M. Taylor10, M.L.N. Ashby9, P. Benavidez11, A. Campo Bagatin11, G. Bernabeu11 ABSTRACT We report the orbital distribution of the trans-neptunian objects (TNOs) discovered during the Canada-France Ecliptic Plane Survey (CFEPS), whose discovery phase ran from early 2003 until early 2007. The follow-up observations started just after the first discoveries and extended until late 2009. We obtained characterized observations of 321 sq.deg. of sky to depths in the range g ∼23.5 - 24.4 AB mag. We provide a database of 169 TNOs with high-precision dynamical classification and known discovery efficiency. Using this database, we find that the classical belt is a complex region with sub-structures that go beyond the usual splitting of inner (interior to 3:2 mean-motion resonance [MMR]), main (between 3:2 and 2:1 MMR), and outer (exterior to 2:1 MMR). The main classical belt (a=40–47 AU) needs to be modeled with 1Based on observations obtained with MegaPrime/MegaCam, a joint project of CFHT and CEA/DAPNIA, at the Canada- France-Hawaii Telescope (CFHT) which is operated by the National Research Council (NRC) of Canada, the Institute National des Sciences de l'Universe of the Centre National de la Recherche Scientifique (CNRS) of France, and the University of Hawaii. This work is based in part on data products produced at the Canadian Astronomy Data Centre as part of the Canada-France-Hawaii Telescope Legacy Survey, a collaborative project of NRC and CNRS. 2Institut UTINAM, CNRS-UMR 6213, Observatoire de Besanc¸on, BP 1615, 25010 Besanc¸on Cedex, France 3Herzberg Institute of Astrophysics, National Research Council of Canada, Victoria, BC V9E 2E7, Canada 4Department of Physics and Astronomy, 6224 Agricultural Road, University of British Columbia, Vancouver, BC, Canada 5Planetary Science Directorate, Southwest Research Institute, 1050 Walnut Street, Suite 300, Boulder, CO 80302, USA 6Department of Planetary Sciences, University of Arizona, 1629 E. University Blvd, Tucson, AZ, 85721-0092, USA 7Cornell University, Space Sciences Building, Ithaca, New York 14853, USA 8Observatoire de la Cote d'Azur, BP 4229, Boulevard de l'Observatoire, F-06304 Nice Cedex 4, France 9Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138 10Department of Physics and Astronomy, University of Victoria, Victoria, BC V8W 2Y2, Canada 11Departamento de Fisica, Ingenieria de Sistemas y Teoria de la Senal, E.P.S.A., Universidad de Alicante, Apartado de Correos 99, Alicante 03080, Spain 12Deceased – 2 – at least three components: the 'hot' component with a wide inclination distribution and two 'cold' components (stirred and kernel) with much narrower inclination distributions. The hot component must have a significantly shallower absolute magnitude (Hg) distribution than the other two components. With 95% confidence, there are 8000+1800−1600 objects in the main belt with Hg ≤ 8.0, of which 50% are from the hot component, 40% from the stirred component and 10% from the kernel; the hot component's fraction drops rapidly with increasing Hg. Because of this, the apparent population fractions depend on the depth and ecliptic latitude of a trans- neptunian survey. The stirred and kernel components are limited to only a portion of the main belt, while we find that the hot component is consistent with a smooth extension throughout the inner, main and outer regions of the classical belt; in fact, the inner and outer belts are consistent with containing only hot-component objects. The Hg ≤ 8.0 TNO population estimates are 400 for the inner belt and 10,000 for the outer belt to within a factor of two (95% confidence). We show how the CFEPS Survey Simulator can be used to compare a cosmogonic model for the the orbital element distribution to the real Kuiper belt. Subject headings: Kuiper Belt, surveys; PACS 96.30.Xa 8 9 – 3 – 1. Introduction The minor body populations of the solar system provide, via their orbital and physical properties, win- dows into the dynamical and chemical history of the Solar System. Recognition of the structural complexity in the trans-neptunian region has lead to models that describe possible dynamical evolutionary paths, such as a smooth migration phase for Neptune (Malhotra 1993), the large scale re-ordering of the outer solar system (Tsiganis et al. 2005; Thommes et al. 1999), the scattering of now-gone rogue planets (Gladman & Chan 2006), or the close passage of a star (Ida et al. 2000). Evaluating these models is fraught with dangers due to observational biases affecting our knowledge of the intrinsic populations of the trans-neptunian region (see Kavelaars et al. 2008; Jones et al. 2010, for discussion of these issues). Over the past twenty years, many different Kuiper Belt surveys (those with more than 10 detections include Jewitt et al. (1996); Larsen et al. (2001); Trujillo et al. (2001); Gladman et al. (2001); Allen et al. (2002); Millis et al. (2002); Elliot et al. (2005); Petit et al. (2006); Jones et al. (2006); Schwamb et al. (2010)) have been slowly building up a sample, albeit with differing flux and pointing biases. Jones et al. (2006) enumerates the aspects of surveys that must be carefully recorded and made public if quantitative comparisons with models are to be made. The primary goal of the Canada-France Ecliptic Plane Survey (CFEPS) is the production of a catalogue of trans-neptunian objects (TNOs) combined with a precise account of the observational biases inherent to that catalog. The description of the biases, combined with provisioning of a 'survey simulator', enables researchers to quantitatively compare the outcome of their model simulations to the observed TNO popu- lations. In Jones et al. (2006) we described our initial 'pre-survey' and general motivation for this project, and Kavelaars et al. (2009) (P1 hereafter) describes the first year of operation of this survey (the L3 data re- lease). This manuscript describes the observations that make up the integrated seven years of the project and provide our complete catalog (the L7 release) of near-ecliptic detections and characterizations along with fully-linked high-quality orbits. In summary, the 'products' of the CFEPS survey consists of four items: 1. A list of detected CFEPS TNOs, associated with the block of discovery, 2. a characterization of each survey block, 3. a Survey Simulator that takes the a proposed Kuiper Belt model, exposes it to the known detection biases of the CFEPS blocks and produces simulated detections to be compared with the real detections, and 4. the CFEPS-L7 model population. In Sections 2 and 3, we describe the observation and characterization of the CFEPS TNO sample. The dynamical classification of all tracked TNOs in our sample is given in Section 4. In Section 5, we update our parametrized model of the main and inner classical Kuiper Belt (P1) and give an improved estimate of the total number of objects in each of these dynamical subpopulations. We also extend our model to the non-resonant, non-scattering part of the belt beyond the 2:1 MMR with Neptune. Section 6 gives an order of magntitude estimate of the scattering disk's population. Section 7 demonstrates the use of our Survey 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 – 4 – 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 Simulator to compare the results of a cosmogonic model to the CFEPS detections. Finally in Section 8, we present our conclusions and put our findings in perspective. 2. Observations and Initial reductions The discovery component of the CFEPS project imaged ∼320 square degrees of sky, almost all of which was within a few degrees of the ecliptic plane. Discovery observations occurred in blocks of ≈ 16 fields acquired using the Canada-France-Hawaii Telescope (CFHT) MegaPrime camera which delivered discovery image quality (FWHM) of 0.7 - 0.9 arc-seconds in queue-mode operations. The 0.96◦ × 0.94◦ MegaPrime FOV is paved by 36 individual 4600x2048 CCDs, each pixel having a scale of 0.187(cid:48)(cid:48). The CFEPS designation of a 'block' of discovery fields was: a leading 'L' followed by the year of observations (3,4,5 and 7) and then a letter representing the two week period of the year in which the discovery observations were acquired (example: L3f occured in the second half of March 2003). Discovery observations occurred between March 2003 and July 2005 plus one block of fields (L7a) observed in January 2007. The CFEPS presurvey block (Jones et al. 2006) in 2002 also consisted of a single contiguous sky patch. To enhance our sensitivity to the latitude distribution of the Kuiper belt we also acquired two survey blocks of 11 square degrees each, at ∼10◦ ecliptic latitude (L5r) and ∼20◦ ecliptic latitude (L5s). Each of the discovery blocks was searched for TNOs using our Moving Object Pipeline (MOP; see Petit et al. 2004). Table 1 provides a summary of the survey fields, imaging circumstances and detection thresholds, both for CFEPS and for the presurvey. Figure 1 presents the sky coverage of our discovery blocks. For a detailed description of the initial CFEPS observing plan, field sequencing and follow-up strategy see Jones et al. (2006) and P1. 3. Sample Characterization The photometric calibration of the discovery triplets to a common reference frame and determination of our detection efficiency is required for our survey simulator analysis. It is presented in Appendix B and the photometric measurements of all CFEPS TNOs acquired in photometric conditions are given in Table 7. We characterized the magnitude-dependent detection probability of each discovery block by inserting artificial sources in the images and running these images through our detection pipeline to recover these artificial sources. We used the g filter at CFHT for all our discovery observations, except for block L3h which was acquired using the r filter. For that block's fields, we shifted the limits to a nominal g value using a color of (g − r) = 0.70, corresponding to the mean (g − r) color of our full CFEPS sample. The TNOs in each block that have a magnitude brighter than that block's 40% detection probability are considered to be part of the CFEPS characterized sample. Because detection efficiencies below ∼ 40% determined by human operators and our MOP diverge - MOP accepts more faint objects, at the expense of false detections - (Petit et al. 2004), and since characterization is critical to the CFEPS goals, we chose not to utilize the sample faint-ward of the measured 40% detection-efficiency level for quantitative science (although we report these Block L3f L3h L3q L3s L3w L3y L4h L4j L4k L4m L4n L4o L4p L4q L4v L5c L5i L5j L5r L5s L7a DECa RAa HRS DEG 12:42 −04:33 13:03 −06:48 22:01 −12:04 19:43 −01:20 22:21 04:33 07:30 21:48 Total 13:35 −09:00 15:12 −16:51 15:12 −18:47 19:14 −22:47 19:23 −21:33 19:15 −23:46 20:53 −18:27 21:26 −16:05 15:10 02:35 Total 17:13 09:11 16:18 −22:18 16:09 −19:59 03:55 22:36 14:35 22:28 08:43 18:30 Total Grand Total 0.80 0.81 0.89 0.87 0.87 0.85 0.89 0.89 0.90 0.89 0.90 0.90 0.85 0.85 0.78 0.84 0.90 0.89 0.90 0.90 0.89 3 14 9 5 13 10 54 20 10 19 4 4 2 9 14 18 100 21 7 3 1 1 9 42 196 13 4x4 4x4 4x4 14x1 16x1 4x4 2003-03-24 2003-04-26 2003-08-31 2003-09-23 2003-12-16 2003-12-24 G.MP9401 R.MP9601 G.MP9401 G.MP9401 G.MP9401 G.MP9401 limitd gAB 23.75 24.43† 24.08 23.95 24.25 24.08 Detection limitse rate ("/h) 1.7 to 5.1 0.8 to 6.2 1.2 to 6.2 0.8 to 8.0 0.8 to 6.0 1.7 to 5.1 direction (DEG) −10.0 to 50.0 5.6 to 41.6 −38.0 to −2.0 −42.6 to −5.0 −29.0 to 11.0 −6.0 to 24.0 94 sqr. deg. 7x2; 1x1 8x2 8x2 12x1 14x1 13x1 8x2 8x2 2x2; 1x1; 5x2 133 sqr. deg. 7x2; 1x1 8x2 8x2 3x2; 1x1; 2x2 3x2; 1x1; 2x2 patchy 94 sqr. deg. 2004-04-26 2004-04-25 2004-05-24 2004-06-25 2004-07-22 2004-07-24 2004-08-15 2004-08-19 2004-11-09 G.MP9401 G.MP9401 G.MP9401 G.MP9401 G.MP9401 G.MP9401 G.MP9401 G.MP9401 G.MP9401 2005-02-10 2005-05-12 2005-06-10 2005-09-03 2005-09-03 2007-01-19 G.MP9401 G.MP9401 G.MP9401 G.MP9401 G.MP9401 G.MP9401 24.06 24.00 24.35 23.76 23.74 23.53 24.00 24.21 24.40 24.30 23.84 23.49 23.89 24.00 23.98 0.8 to 6.0 0.8 to 5.6 0.8 to 5.7 0.8 to 5.6 0.8 to 6.0 0.8 to 6.0 1.0 to 5.7 1.2 to 6.1 0.8 to 6.3 0.8 to 6.4 0.4 to 7.3 0.4 to 7.0 0.7 to 7.5 0.7 to 7.5 0.8 to 7.7 2.0 to 42.0 −3.6 to 36.4 −1.0 to 35.0 −25.0 to 15.0 −27.7 to 12.3 −24.7 to 11.3 −30.0 to 0.0 −35.5 to −0.5 −34.0 to −2.0 −1.0 to 31.0 −9.4 to 32.2 −9.9 to 33.9 −42.1 to −1.9 −41.8 to −2.0 −4.1 to 34.9 2 11 7 5 11 10 46 16 10 16 4 4 1 9 10 14 84 19 7 3 1 1 8 39 – 5 – Table 1. Summary of Field positions and Detections. Fillb Factor Charact. Det.c Track. Disc. Geometry DEG x DEG Discovery date filter Pre 22:00 -13:00 0.90 169 321 sqr. deg. 10 3.5x2 2002-08-05 R 24.85(cid:63) 0.8 to 8.0 −35.0 to −5.0 Note. - (a) RA/DEC is the approximate center of the field. (b) Fill Factor is the fraction of the rectangle covered by the mosaic and useful for TNO searching. (c) The number of objects in columns 5 and 6 correspond to those detected and tracked in the characterized sample, as defined in Sect. 3. (d) The limiting magnitude of the survey, gAB, is in the SDSS photometric system and corresponding to a 40% efficiency of detection. (e)Detection limits give the limits on the sky motion in rate ("/hr) and direction ("zero degrees" is due West, and positive to the North). † Although the L3h block was acquired in r filter, the reported limiting magnitude has been translated to g band by applying an offset of g − r = 0.7, which is the average g − r color of our full sample (see Table 7). (cid:63) The Presurvey block was acquired in R filter with the CFH12K camera (Jones et al. 2006). The limiting magnitude has been translated to g band by applying an offset of g − R = 0.8. – 6 – 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 discoveries, many of which were tracked to precise orbits). The characterized CFEPS sample consists of 196 objects of the 231 discovered (see Table 7 for a list of these TNOs). The fraction of objects detected bright-ward of our cutoff is consistent with the shape of the TNO luminosity function (Petit et al. 2008) and typical decay in detection efficiency due to gradually increasing stellar confusion and the rapid fall-off at the SNR limit. Our discovery and tracking observations were made using short exposures designed to maximize the efficiency of detection and tracking of the TNOs in the field. These observations do not provide the high- precision flux measurements necessary for possible classification based on broadband colors of TNOs and we do not comment here on this aspect of the CFEPS sample. 4. Tracking and Lost Objects Tracking during the first opposition was done using the built-in followup of the CFEPS project. Sub- sequent tracking, over the next 3 oppositions, occurred at a variety of facilities, including CFHT. The obser- vational efforts outside CFHT are summarized in Table 2. In spring 2006 the CFEPS project made an initial data release of the complete observing record for the L3 objects (objects discovered in 2003; before all the refinement observations for all objects were complete). The L3 release was reported to the Minor Planet Center (MPC) (Gladman et al. 2006; Kavelaars et al. 2006a,b) and additional followup that has occurred since the 2006 release has also been reported to the MPC. The final release of the complete observing record for all remaining CFEPS objects is available from the MPC (Kavelaars et al. 2011). Detailed astrometric and photometric data for the CFEPS objects can be found on the CFEPS specific databases1. The corre- spondence between CFEPS internal designations and MPC designations can be determined using Tables 3 and 4, or from electronic tables on the cfeps.net site. All characterized and tracked objects are prefixed by L and are used with the survey simulator for our modeling below. The tracking observations provide sufficient information to allow reliable orbits to be determined such that unambiguous dynamical classification can be achieved in nearly all cases. Ephemeris errors are smaller than a few tens of arc-seconds over the next 5 years. Our standard was to pursue tracking observations until the semimajor axis uncertainty was < 0.1%; in Tables 3 and 4, orbital elements are shown to the precision with which they are known, with typical frac- tional accuracies on the order of 10−4 or better. In the cases of resonant objects even this precision may not be enough to determine the amplitude of the resonant argument. 1http://www.cfeps.net/tnodb/, http://www.obs-besancon.fr/bdp/ – 7 – Table 2. Follow-up/Tracking Observations. UT Date Telescope No. Obs. ESO 2.2m 2002 Aug 05 CFHT + 12k 2002 Sep 03 NOT 2.56m Calar-Alto 2.2-m 2002 Sep 02 2002 Sep 30 CFHT 3.5-m 2002 Nov 28 CFHT 3.5-m 2003 Jul 26 2004 Feb 19 WIYN 3.5-m 2004 Apr 15 Hale 5-m 2004 May 24 Mayall 3.8-m 2004 Aug 12 CFHT 3.5m 2004 Sep 06 KPNO 2m 2004 Sep 11 Mayall 3.8-m 2004 Sep 16 Hale 5-m 2004 Sep 21 CFHT 3.5m 2005 Jul 08 Gemini-North 8-m 2005 Jul 09 Hale 5-m 2005 Jul 11 ESO 2.2m 2005 Aug 01 VLT UT-1 2005 Sep 24 WIYN 3.5-m 2005 Oct 03 2005 Nov 04 Mayall 3.8m 2005 Dec 04 MDM 2.4-m 2006 Jan 28 2006 May 01 CFHT 3.5m 2006 May 02 WIYN 3.5-m 2006 May 26 CFHT 3.5m 2006 Jun 25 Mayall 3.8-m CFHT 3.5m 2006 Jul 03 Hale 5-m 2006 Jul 26 CFHT 3.5m 2006 Sep 18 2006 Sep 26 MMT 6.5m 2006 Oct 22 Hale 5-m 2006 Oct 21 WHT 4m Hale 5-m Hale 5-m 6 6 9 6 10 6 4 73 6 15 15 25 20 4 45 47 25 53 9 72 31 10 50 23 32 20 2 18 15 7 11 29 17 – 8 – Fig. 1.- Geometry of the CFEPS discovery-blocks. The RA and DEC grid is indicated with dotted lines. The black solid curves show constant ecliptic latitudes of -60◦, -30◦, 0◦, 30◦, 60◦, from bottom to top. The – 9 – Of the 196 TNOs in our CFEPS characterized sample 169 have been tracked through 3 oppositions or more (ie. not lost) and their orbits are now known to a precision of ∆a/a < 0.1% and can be reliably classified into orbital sub-populations (see below). The very high fraction of our characterized sample for which classification is possible (86%) is by far the largest 'tracking fraction' among large scale TNO surveys to-date and is due to the strong emphasis on followup observations in our observing strategy, made possible thanks to the time allocation committees of the many observatories listed in Table 2. The initial tracking of TNOs discovered by CFEPS is through blind return to the discovery fields to ensure that there is no orbital bias in the tracked fraction. We do find, however, that the tracked fraction is a function of the magnitude of the TNO and have characterized this bias. For the full CFEPS fields we find the same magnitude dependance as for the L3 fields for objects brighter than the limit of the characterized sample, which we model as ft,L7(g) = 1.0 − 0.25(g − 22.8) (g (cid:54) 22.8) (g > 22.8) (cid:26) 1.0 (cid:26) 1.0 where ft,L7 is the tracked fraction. The tracked fraction remains well above 50% down to the characterized limit of the survey blocks. We have also re-examined the magnitude dependence of the tracked fraction of our Pre-survey discoveries (Jones et al. 2006) and find ft,L7(g) = 1.0 − 2.5(g − 24.1) (g (cid:54) 24.1) (g > 24.1) . The Pre-survey observations used much longer exposure times than for CFEPS, hence the deeper limiting magnitude reached. We also had a smaller survey area and were able to perform a more thorough follow-up campaign, resulting in a tracking efficiency that essentially was 100% up to the limiting magnitude of the discoveries. Our Pre-survey discovery observations where reported on the Landolt-R system and we have transformed our Pre-survey limits to g, for use in our survey simulator, using a constant color offset of (g - R) = 0.8 (Hainaut & Delsanti 2002). 4.1. Orbit Classification We adopt the convention that, based on orbital elements and dynamical behavior, the Kuiper Belt can be divided into three broad orbital classes. An object is checked against each dynamical class in the order below to decide whether or not it belongs to that class, each object can belong to only one class. A schematic representation of this dynamical classifaction is shown in Figure 1 of Gladman et al. (2008). • resonant (objects currently in a mean-motion resonance with Neptune) • scattering (objects that over 10 Myr forward in time integrations experience encounters with Neptune resulting in variation of semimajor-axis of more than 1.5 AU) 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 – 10 – Table 2-Continued UT Date Telescope No. Obs. 2.1-m reflector 2006 Nov 23 WIYN 3.5-m 2007 Feb 14 2007 Feb 21 Hale 5-m 2007 May 15 Hale 5-m 2007 May 15 KPNO 2m 2007 Jun 02 MMT 6.5m 2007 Sep 11 WIYN 3.5-m 2007 Sep 16 Hale 5-m 2007 Nov 08 WIYN 3.5-m 2008 May 03 WIYN 3.5-m 2008 Jun 07 2008 Oct 23 WIYN 3.5-m 2008 Dec 06 Hale 5-m CFHT 3.5m 2009 Jan 26 2009 Apr 17 MMT 6.5m 2009 Apr 23 Subaru 8-m 2009 Jun 20 WIYN 3.5-m CTIO 4-m 41 3 22 45 23 3 32 27 30 52 28 3 9 19 3 1 22 Note. - UT Date is the start of the ob- serving run; No. Obs. is the number of as- trometric measures reported from the observ- ing run. Only observations not part of the Very Wide component of CFHT-LS are re- ported here. Runs with low numbers of as- trometric measures were either wiped out by poor weather, or not meant for CFEPS objects follow-up originally. – 11 – • classical or detached belt (everything that remains). One further sub-divides the classical belt into: – inner classical belt (objects with semi-major axis interior to the 3:2 MMR) – main classical belt (objects whose semi-major axis is between the 3:2 and 2:1 MMRs) – outer classical belt (objects with semi-major axis exterior to the 2:1 MMR with e < 0.24) – detached (those objects with semi-major axis beyond the 2:1 MMR that have e > 0.24) The classical belt is often also divided into high-inclination and low-inclination objects. For the L7 model, we work from the hypothesis described in Brown (2001) that there exist two distinct populations, one with a wide inclination distribution (the 'hot' population), and the other one with a narrow inclination distribution (the 'cold' population), with both populations overlapping with each other in inclination space (thus some "cold" objects may have large inclination, and some "hot" objects may have low inclination). In the literature, the separation between hot and cold populations is sometimes presented as a sharp cut in inclination, often around 5◦, under the assumption that an object with inclination less (greater) than that threshold has a very high likelihood to be a member of the cold (hot) population. As will be seen in Section 5.1.1, the veracity of this assumption depends on the physical size of the objects being sorted, larger objects (H <7) having a much higher probability of being from the hot population, regardless of their inclination, while the objects from the cold population dominate at smaller (H>8) sizes. A strict inclination cut does not isolate the two mixed populations. Following the procedure in Gladman et al. (2008) (similar to Chiang et al. 2003b), we extend the L3 sample classification given in P1 to our full CFEPS sample as of November 2009 (including all refinement observations to that date). Using this classification procedure, 15 of our objects remain insecure (even though these have observational arcs extending across 5 oppositions!); all of these are due to their proximity to a resonance border where the remaining astrometric uncertainty makes it unclear if the object is actually resonant. We list these 'insecure' objects in the category shown by 2 of the 3 clones. Table 3 gives the clas- sification of all characterized objects used for comparison with the Survey Simulator's artificial detections. Several objects had been independently discovered before we submitted our observations to the MPC and are marked with a PD suffix. Although we do not claim 'discoverer credit' for these objects, they have just as much scientficially-exploitable value because they were detected during our characterized observations and hence can to be included when running our survey simulator. Table 4 gives the classification of the tracked objects below the 40% efficiency threshold, hence deemed non-characterized and not used in our Survey Simulator comparisons. 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 – 12 – Table 3. Characterized Object Classification. DESIGNATIONS CFEPS MPC a AU e i ◦ dist AU Comment L3y11 (131697) 2001 XH255 34.925 0.0736 2.856 34.0 5:4 MPCW Resonant Objects L4h14 L3s06 L5c23 L7a10 L4k11 L4h15 L5c11 L4h06 L4v18 L4m02 L3s02 L4h09PD L3h19 L3w07 L4h07 L3h11 L3w01 L4j11 L4v09 L3h14 L3s05 L4v13 L4k01 L3h01 L5i06PD L4h10PD L4v12 L4h08 L5c08 L3y06 L5c13PD L4v05 L3y12PD L4k10 L3q08PD L4n03 L3w03 2004 HM79 (143685) 2003 SS317 2005 CF81 2005 GH228 2004 KC19 2004 HB79 2005 CD81 2004 HY78 2004 VY130 2004 MS8 2003 SO317 (47932) 2000 GN171 2003 HF57 2003 TH58 2004 HA79 2003 HA57 2005 TV189 2004 HX78 2004 VX130 2003 HD57 2003 SR317 2004 VV130 2004 KB19 2004 FW164 2001 KQ77 1995 HM5 2004 VZ130 2004 HZ78 2006 CJ69 2003 YW179 1999 CX131 2004 VE131 (126154) 2001 YH140 2004 KK19 (135742) 2002 PB171 2004 OQ15 2003 YJ179 36.441 36.456 36.473 36.663 39.258 39.260 39.262 39.302 39.342 39.344 39.346 39.352 39.36 39.36 39.378 39.399 39.41 39.420 39.430 39.44 39.44 39.454 39.484 39.492 39.505 39.521 39.551 39.580 42.183 42.193 42.240 42.297 42.332 42.410 43.63 43.646 43.66 0.07943 0.2360 0.06353 0.18814 0.23605 0.22862 0.15158 0.19571 0.27616 0.29677 0.2750 0.28120 0.194 0.0911 0.24697 0.1710 0.1884 0.15270 0.20696 0.179 0.1667 0.18827 0.21859 0.1575 0.15619 0.25197 0.28159 0.15095 0.22866 0.1537 0.23387 0.25889 0.14043 0.14391 0.125 0.12472 0.0794 1.172 5.905 0.405 17.151 5.637 2.661 21.344 12.584 10.203 12.249 6.563 10.815 1.423 27.935 22.700 27.626 34.390 16.272 5.745 5.621 8.348 23.924 17.156 9.114 15.617 4.814 11.581 13.310 17.916 2.384 9.757 5.198 11.078 4.485 5.450 9.727 1.446 38.0 28.2 34.4 30.6 30.2 32.0 45.2 31.8 28.5 27.8 32.3 28.5 32.4 35.8 38.4 32.7 32.0 33.6 34.8 32.9 35.5 32.8 39.5 33.3 36.2 31.1 29.2 34.8 35.5 35.7 41.8 39.6 36.4 46.0 40.7 40.5 40.3 I I 4:3 4:3 4:3 4:3 3:2 3:2 3:2 3:2 3:2 3:2 3:2 3:2 3:2 3:2 3:2 3:2 3:2 3:2 3:2 3:2 3:2 3:2 3:2 3:2 3:2 3:2 3:2 3:2 5:3 5:3 5:3 5:3 5:3 5:3 7:4 7:4 7:4 – 13 – Table 3-Continued DESIGNATIONS CFEPS MPC a AU e i ◦ L4v10 K02O03 2004 VF131 2000 OP67 43.672 43.72 0.21492 0.19 1 0.816 0.751 dist AU 42.0 39.3 Comment 7:4 7:4 L4h11 2004 HN79 45.736 0.22936 11.669 37.4 15:8 I L4h18 L4k16 L4k20 K02O12 L4v06 2004 HP79 2004 KL19 2004 KM19 2002 PU170 2004 VK78 47.567 47.660 47.720 47.75 47.764 0.18250 0.32262 0.29180 0.2213 0.33029 L3y07 L5c19PD (131696) 2001 XT254 2002 CZ248 52.92 53.039 0.3221 0.38913 2.253 5.732 1.686 1.918 1.467 0.518 5.466 39.5 32.3 33.8 47.2 32.5 36.6 36.2 2:1 2:1 2:1 2:1 2:1 7:3 MPCW 7:3 L5c12 2002 CY224 53.892 0.34651 15.733 36.3 12:5 L4j08 L3f04PD L4j06PD L4k14 L4h02PD 2004 HO79 (60621) 2000 FE8 2002 GP32 2004 KZ18 2004 EG96 55.206 55.29 55.387 55.419 55.550 0.41166 0.4020 0.42195 0.38191 0.42291 5.624 5.869 1.559 22.645 16.213 37.3 36.0 32.1 34.4 32.2 L4v08 2004 VU130 62.194 0.42806 8.024 49.7 5:2 5:2 5:2 5:2 5:2 3:1 L3y02 2003 YQ179 88.38 0.5785 20.873 39.3 5:1 I I (11:8) Inner Classical Belt L3y14PD L4q12PD L4q10 L4k18 L4o01 L3w06 (131695) 2001 XS254 2000 OB51 1999 OJ4 2004 KD19 2004 OP15 2003 YL179 37.220 37.820 38.017 38.257 38.584 38.82 0.05211 0.03501 0.02539 0.01707 0.05532 0.002 L4k12 L4q05 L4k19 L3w05 L4h16 L5s01PD L3s01 L4q15 2004 KH19 2004 QE29 2005 JB186 2003 YK179 2004 HL79 (120347) 2004 SB60 2003 SN317 1999 ON4 Main Classical Belt 40.772 40.878 41.471 41.67 42.126 42.028 42.50 42.571 0.11721 0.08372 0.10588 0.146 0.07520 0.10667 0.0421 0.03995 4.262 4.458 4.000 2.126 22.946 2.525 35.230 24.125 20.220 19.605 16.759 23.931 1.497 3.187 35.3 36.6 38.1 38.9 38.7 38.7 43.6 37.5 38.0 42.7 40.0 43.7 41.5 40.9 Comment I (12:7) – 14 – Table 3-Continued DESIGNATIONS CFEPS MPC a AU e i ◦ L3h05 L3q02PD L3s03 K02O20 K02P32 L5c03 L5i01 L4p02 L4p01 K02O40 L4q03 L3w11 L4m03 L4h05PD L4j10 L4j02 L4k04 L5c07PD L7a06 L3w10 L3y01 L3y05 L3h18 L4p05 L7a05 L5j04 L4h01PD L4p06PD L4h12 L5i03PD L4h13 L4v03 L5c22 L3h13 L3h09 L5i05 L7a07 L3q06PD L4k03 L5c21PD L3q09PD L5c18 L3h20 L5c24PD L4v02 42.604 2003 HY56 2001 QB298 42.618 42.63 2003 SQ317 2002 PV170 42.643 42.65 2002 PX170 2005 CE81 42.715 42.778 2006 HA123 2004 PU117 42.817 42.983 2004 PT117 43.015 2002 PY170 43.020 2004 QD29 43.078 2003 TK58 43.120 2004 MT8 43.255 2001 FK185 43.259 2004 HH79 43.269 2004 HF79 43.272 2004 KG19 43.398 2005 XU100 43.500 2006 WF206 43.542 2003 TL58 43.582 2003 YX179 43.585 2003 YS179 43.612 2003 HG57 43.620 2004 PW117 43.684 2005 BV49 2005 LB54 43.690 (181708) 1993 FW 43.717 2001 QY297 43.835 43.888 2004 HK79 43.898 2001 KO77 43.947 2004 HJ79 43.951 2004 VC131 2007 DS101 43.991 44.04 2003 HH57 44.05 2003 HC57 44.077 2005 JY185 2005 BW49 44.097 44.10 2001 QJ298 44.123 2004 KF19 44.126 2005 EE296 44.15 2001 QX297 2007 CS79 44.159 44.17 2003 HE57 44.172 1999 CU153 2004 VB131 44.189 0.037 0.0962 0.0795 0.016 0.041 0.04666 0.04615 0.01461 0.04115 0.030 0.11388 0.0647 0.04195 0.03994 0.06010 0.02547 0.02164 0.10283 0.04246 0.0456 0.044 0.022 0.0323 0.06023 0.04575 0.04752 0.04807 0.08332 0.07800 0.14569 0.04419 0.07395 0.08474 0.088 0.072 0.06848 0.07959 0.0388 0.06348 0.06804 0.0275 0.03582 0.100 0.06520 0.07267 2.578 1.800 28.568 1.271 1.570 3.084 3.303 1.874 1.238 3.016 23.862 3.355 2.239 1.171 8.610 1.484 0.963 7.869 2.056 7.738 4.850 3.727 2.098 1.862 7.981 3.006 7.750 1.547 1.946 20.726 3.317 0.490 1.389 1.436 1.038 2.139 2.102 2.151 0.108 3.296 0.911 1.540 8.863 2.698 1.747 dist AU 42.5 39.1 39.3 42.2 42.8 40.8 41.0 42.4 43.6 43.0 40.6 45.6 44.9 41.7 43.2 42.4 42.4 41.7 44.4 42.2 42.5 43.8 43.0 46.0 41.8 41.8 41.9 42.8 41.3 37.7 45.0 40.7 44.6 40.2 43.4 44.6 41.9 45.2 41.4 46.2 43.5 42.8 40.0 42.7 46.5 – 15 – Table 3-Continued Comment I (9:5) I (11:6) DESIGNATIONS CFEPS MPC L4j03 L4p09 L4p08PD L5j03 L4j01 K02P41 L4k02 L3w08 L3w02 L3w04 L5i08 K02O43 L4n04 K02O32 L4q16 L4j12 L5c20PD L3w09 L5c10PD L5c06 L4v01 L4k15PD L5c02 L4q11 L4j07 L5i02PD L4p03 L7a04PD L4k13 L3q04PD L4v14 L4j05 L4q09 L3y03 L4k17 L7a11PD L3y09 L5c14 L3h04 L4m04 2004 HG79 2004 PX117 2001 QZ297 2005 LA54 2004 HE79 2002 PA171 2004 KE19 2003 TJ58 2003 TG58 (143991) 2003 YO179 2005 JJ186 2002 PC171 2004 MU8 2002 PW170 (66452) 1999 OF4 2006 JV58 2002 CZ224 2004 XX190 1999 CJ119 2007 CQ79 2004 VA131 2003 LB7 2006 CH69 1999 OM4 2004 HD79 2001 KW76 2004 PV117 2002 CY248 2006 JU58 2002 PT170 2004 VD131 2004 HC79 2000 PD30 2003 YU179 2004 KJ19 2000 CO105 2003 YV179 2007 CR79 2003 HX56 2004 MV8 a AU 44.200 44.261 44.283 44.314 44.316 44.34 44.360 44.40 44.54 44.602 44.636 44.706 44.856 44.88 44.933 44.961 44.980 45.171 45.325 45.441 45.538 45.580 45.735 45.924 45.941 46.013 46.069 46.191 46.239 46.24 46.324 46.399 46.519 46.75 46.967 47.046 47.10 47.149 47.196 47.234 e i ◦ 0.02298 0.09965 0.06442 0.06719 0.09805 0.076 0.04981 0.0864 0.103 0.1370 0.09431 0.059 0.08180 0.074 0.06380 0.06094 0.06304 0.1042 0.06651 0.07721 0.09613 0.13130 0.03535 0.11643 0.03205 0.21613 0.15343 0.14635 0.12464 0.143 0.12253 0.16064 0.02232 0.1597 0.23543 0.14750 0.222 0.21876 0.2239 0.17503 3.595 3.747 1.856 7.919 3.089 2.511 1.178 0.954 1.660 19.393 4.141 3.574 3.580 3.933 2.660 0.317 1.687 1.577 3.205 1.185 0.767 2.294 1.791 2.088 1.305 10.460 4.324 7.038 7.035 3.703 3.646 1.446 4.594 4.855 24.421 19.270 15.569 21.869 29.525 27.205 dist AU 43.2 46.1 42.0 41.6 40.0 47.7 42.6 40.8 43.7 41.3 41.8 42.7 48.2 47.4 45.2 42.2 47.7 40.9 42.3 45.8 41.2 40.1 44.2 44.0 47.3 39.6 39.5 51.8 46.5 50.5 41.5 39.0 45.7 39.6 38.5 49.3 41.1 36.9 45.5 39.1 Outer Classical Belt L4q06 L4q14 2004 QG29 2004 QH29 48.480 50.859 0.23517 0.22922 27.134 12.010 37.8 39.9 – 16 – 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 121 (64%) of the tracked sample are in the classical belt, split into 6 (3%) inner, 101 (54%) main, 3 (2%) outer and 11 (6%) detached belt objects. Orbital integration shows that 58 (31%) objects are in a mean-motion resonance with Neptune, 25 (13%) of which are plutinos. The remaining sample consists of 9 (5%) objects on scattering orbits. The apparent motion of TNOs in our opposition discovery fields is approximately θ((cid:48)(cid:48)/hr) (cid:39) (147 AU)/r, where r is the heliocentric distance in AU. With a typical seeing of 0.7 - 0.9 arc-second and a timebase of 70 - 90 minutes between first and third frames, we were sensitive to objects as distant as r (cid:39)125 AU, provided they are large enough to be above our flux limit. The furthest object discovered in CFEPS lies at 58.3 AU from the Sun (L3q03 = 2003 QX113, a detached object with a = 49.55 AU). The short exposure times used (70 - 90 seconds) allowed us to detect objects as close as 15 AU without trailing. We elected to use a rate of motion cut corresponding to objects further than 20 AU from Earth. In the following sections we present a parameterization of the intrinsic classical Kuiper belt and scat- tering disk population implied by our observations. The differing detectability of these populations, in a flux-limited survey, implies that the intrinsic population ratios will be different from the observed ones. We present the more complex analysis of the resonant populations in a companion paper (Gladman et al. 2011). 5. The classical belt's orbital distribution This section presents the results of our search for an empirical parameterized orbit distribution for the various components of the so-called 'classical' belt. For each sub-component we start with a simple parameterization of the intrinsic orbit and absolute magnitude distributions. We then use the CFEPS Survey Simulator2 to determine which members of the intrinsic population would have been detected by the survey. The orbital-element distributions of the simulated detections are then compared to our characterized sample. This process is iterated with models of increasing complexity until arriving at a model that provides a statistically-acceptable match; no cosmogonic considerations are invoked. Our model search process provided acceptable parameterizations of the main classical belt, the inner classical belt and the outer+detached population. Our goal is to discover the main features of the orbital distribution and provide a population estimate for each orbital sub-component. While our success in finding acceptable models is not a proof of model uniqueness, we were surprised, in many cases, by the restricted range of acceptable models. To evaluate a model's quality, we extend the method defined in P1 to more variables. We compute the Anderson-Darling (AD) (Information Technology Laboratory 2011) statistic for the distributions of the orbital elements a, e, i, q, and for r (heliocentric distance at discovery) and g magnitude. We use Kuiper's modified Kolmogorov-Smirnov (KKS) statistic for the mean anomaly M. We follow the same procedure as 2The survey simulator is available on-line, with all informations needed to use it, at http://www.cfeps.net as a stand-alone package, or at http://CFEPSSim.obs-besancon.fr/, as an on-line service. – 17 – Table 3-Continued DESIGNATIONS CFEPS MPC a AU e i ◦ dist AU Comment L5c16 2005 CG81 53.834 0.23684 26.154 44.6 Detached Classical Belt L5i04 L3q03 L7a02 L4p04PD L5c15 L4n06 L4n05 L3f01 L4h21 L5j02 L5r01 2005 JK186 2003 QX113 2006 WG206 2000 PE30 2005 CH81 2004 OS15 2004 OR15 2003 FZ129 2004 HQ79 2005 LC54 2005 RH52 47.264 49.55 50.416 54.318 55.156 55.760 56.248 61.71 63.299 67.354 153.800 0.24363 0.252 0.29111 0.34216 0.31812 0.31667 0.33882 0.3840 0.42264 0.46279 0.74644 Scattering Disk 2004 KV18 2004 MW8 2004 PY117 2003 QW113 2006 BS284 2004 VH131 2004 VG131 2003 HB57 30.192 33.479 39.953 50.99 59.613 60.036 64.100 159.6 0.18517 0.33308 0.28088 0.484 0.43949 0.62928 0.50638 0.7613 L4k09 L4m01 L4p07 L3q01 L7a03 L4v11 L4v04 L3h08 27.252 6.753 14.297 18.416 5.136 4.248 6.919 5.793 6.473 22.443 20.447 13.586 8.205 23.545 6.922 4.575 11.972 13.642 15.499 38.1 58.3 38.9 37.6 37.6 39.5 37.3 38.0 36.6 43.1 39.0 26.6 31.4 29.6 38.2 47.0 26.8 31.8 38.4 I (10:3) I scattering Note. - M:N: object in the M:N resonance; I: indicates that the orbit classification is insecure (see Gladman et al. (2008) for an explanation of the exact meaning); (M:N): object may be in the M:N resonance; MPCW : indicates object was in MPC database but found +1◦ from predicted location. Objects prefixed with L are the characterized, tracked objects discovered during CFEPS; objects prefixed with K02 were discovered in our pre-survey (Jones et al. 2006); The full orbital elements are available in electronic form from either http://www.cfeps.net/tnodb/ or the MPC. – 18 – Table 4. Non Characterized Object Classification. DESIGNATIONS CFEPS MPC a AU e i ◦ dist AU Comment Resonant Objects U5j06 39.369 0.22055 13.525 31.2 3:2 U3s04 2003 SP317 45.961 0.1694 5.080 44.9 17:9 I U7a08 47.702 0.19600 7.020 38.4 U5j01PD (136120) 2003 LG7 62.157 0.47825 20.104 33.1 2:1 3:1 2003 YM179 2003 FA130 2002 WL21 2003 YR179 2003 YT179 2003 HZ56 1999 CN119 2003 YP179 2003 YN179 2000 JF81 U3w13 U3f02 U4j09 U7a09 U3w17 U3y16 U3y04 U3h06 U5c17PD U4n01 U3y08 U4n02 U3w16 U4j04PD U7a01 Main Classical Belt 40.960 42.602 42.642 42.701 43.103 43.421 43.542 43.63 43.733 43.915 44.03 44.056 44.272 46.117 0.056 0.031 0.00775 0.09231 0.0415 0.0523 0.028 0.010 0.04043 0.13500 0.079 0.06176 0.006 0.10218 23.414 0.288 3.044 2.931 2.552 9.823 1.684 2.550 0.999 0.271 0.947 2.943 2.768 1.742 40.2 41.3 42.3 44.8 41.6 41.3 44.4 43.5 44.5 43.7 41.3 46.8 44.4 44.9 Scattering Disk 42.621 0.16444 4.742 38.9 I (5:3) Note. - M : N: object in the M:N resonance; I: indicates that the orbit classification is insecure (see Gladman et al. (2008) for an explanation of the exact meaning). – 19 – 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 used in P1 to determine the significance of the computed statistics. For each model parameterization we use the Survey Simulator to draw a large 'parent' population from the model. We then draw sub-samples with the same total number as in our L7 characterized sample. Using this simulated 'observed' population we compute the various statistics that result from comparing to our large 'parent' population. This re-sampling is repeated 5000 times providing a distribution of statistic values for the given parameterization, ie. 'boot- strapping' the statistic. The probability of statistic measured for the L7 sample is determined by comparing that statistic value to the range of statistic values returned by the bootstrap process. We reject a model if the minimum statistical probability determined in this way is returned by fewer than 5% of the model bootstraps. 5.1. The main classical belt In P1 we presented a model that matched the orbital distribution of the main classical belt objects detected in the L3 sample; due to the smaller number of objects in the L3 sample, we restricted ourselves to fit only selected orbital elements and considered a constrained range of the phase-space volume available to main-classical belt objects. In addition, P1 did not attempt to determine the absolute magnitude distribution using our detections but instead utilized values available in the literature. Here we restrict our main classical belt model to the 40 AU ≤ a ≤ 47 AU range, to avoid the complex borders of the 3:2 and 2:1 MMR regions, which includes 88 characterized CFEPS TNOs. This sample size allowed us to remove external constraints on the magnitude distribution and explore a more complete model of the available phase-space. Figures 2 and 3 present (a, i), (a, q) and (i, q) projections of the main-belt TNO orbital elements for characterized CFEPS detections and multi-opposition orbits in the MPC. These figures make it clear that objects with q < 39 AU are dominantly from the high-inclination population, as was already apparent in the L3 model. The distribution of low-i objects, which span a narrower range of semimajor axis than their high-i cousins, exhibit considerable phase space structure. In an effort to find a parameterization that yielded these interesting sub-structures we investigated a substantial range of empirical representations. We were, however, unable to find a two-component model (like that in P1) that sufficiently reproduced structure observed in the current sample. A more complex representation is required. After much effort we arrived at our 'L7 model' (based on CFEPS discoveries up to mid-2007). The L7 model is composed of three components (Fig. 4), the fine details of which are presented in Appendix A. These components are a population with a wide inclination distribution (the hot population) superposed on top of a population with narrow inclination component with two semi-major axis / eccentricity distributions (the stirred and kernel populations). The hot population is defined as a band in perihelion distance q es- sentially confined to the range 35 to 40 AU, with soft exponential decay outside this range. Using a 'core' (Elliot et al. 2005) definition based only on inclination does not take into account the transition in the e/i distribution beyond a (cid:39)44.4 AU clearly visible in both Figs. 2 and 3. With the qualifier that there will be mixing from the low-i tail from the hot component, we thus split the 'cold' population of the main classical belt into two sub-components. The stirred population have orbits drawn from a narrow-inclination distribu- tion with semi-major axes starting at a=42.5 AU and extending to a (cid:39) 47 AU, with a range of eccentricities – 20 – 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 that increases as one goes to larger a. The stirred component does not contain the sharp density change at a (cid:39) 44.5 AU. There are more low-i and moderate-e TNOs per unit semimajor axis at a ∼ 44 − 44.5 AU than at smaller and larger semi-major axis, indicating that a third component is required. To model this com- ponent we insert a dense low-inclination concentration, which we call the kernel, near a=44 AU to account for this intrinsic population. The kernel may be the same as the clustering in the a =42–44 region seen as far back as Jewitt & Luu (1995) and Jewitt et al. (1996). This also appears to be the same structure that Chiang (2002) and Chiang et al. (2003a) posited (with rightful skepticism) as a possible collisional family. Although we share the concern that normally the relative speeds from a large parent-body breakup should be larger than this clump's observed dispersion, we find that regardless of interpretation, there is considerable observational support for a tightly-confined structure in orbital element space near the location Chiang et al. pointed to. Recent collisional modeling studies (eg. Leinhardt et al. 2010) raise the possibility of grazing impacts forming low-speed families in the Kuiper Belt, motivated by the Haumea family (Brown et al. 2007). The large number of D ≥ 170 km (absolute magnitude3 Hg ≤ 8) objects in the kernel implies that the parent body would have been a dwarf planet at least as large as Pluto, an unlikely possibility. The kernel thus appears to be the longest-recognized dynamical sub-structure in the classical Kuiper Belt, a structure which requires confinement in all of a, e, and i. There may be other possible representations of the orbital distribution that are consistent with the CFEPS detections, with different boundaries or divisions of the phase space. We have found, however, the generic necessity of a 3-component model can not be avoided. The main characteristics of our model must be similar to reality, because a considerable amount of tuning was needed to achieve an acceptable model. From this 3-component model, we can then provide robust measurements of the sizes of the subpopulations in the Kuiper belt and generate a synthetic 'de-biased' model of the orbital distribution of the main belt which can be used for various modeling purposes, such as collisional dust production (Stark & Kuchner 2010). 5.1.1. The luminosity function The absolute magnitude Hg distribution can be represented by an exponential function N (H) ∝ 10αH with 'slope' α. Hg is converted into apparent magnitude g by g = Hg + 2.5 log (r2∆2Φ(µ)), where ∆ is the geocentric and r the heliocentric distance, µ the phase angle (Sun-TNO-observer) and Φ(µ) the phase function defined by Bowell et al. (1989). We find that two different values of α, one for the hot and one for the cold distributions, are required by our observations. Allowing the stirred and kernel components to have differing values of α did not provide an improved match to the observations and is not required. 3The g-band apparent magnitude of a TNO at heliocentric and geocentric distance of 1 AU if viewed at 0◦ phase angle – 21 – 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 We have run a series of model cases using the orbital element distributions described previously while varying the luminosity functions slopes for the hot component, αh, and for the cold (kernel + stirred) compo- nents αc. For each case, we varied the other orbit model parameters to find the best possible match between the cumulative distribution functions of the Survey Simulator observed Kuiper belt and the L7 sample for each of the selected values of αh and αc. In this way we determined the range of allowed power-law slopes for the limited range of TNO sizes, 7 (cid:46) Hg (cid:46) 8, probed by our observations. Our best fit values are αc = 1.2+0.2−0.3 and αh = 0.8+0.3−0.2, with Fig. 5 presenting the joint 95% confidence region for these slopes. A single value of α for all sub-components in our model is rejected at >99% confidence. The αc determined here is in good agreement with the range derived by Bernstein et al. (2004) for the low-inclination objects and somewhat steeper than that reported in Elliot et al. (2005) while our value for αh overlaps the ranges proposed by both Bernstein et al. (2004) and Elliot et al. (2005) for what they call the excited population. Fraser et al. (2010) also found markedly different values for the slope of the cold component, 0.59–1.05, and the hot component, 0.14–0.56. While those slopes are consistent with Elliot et al. (2005) they are shallower than Bernstein et al. (2004) and our own estimates. The Fraser et al. (2010) results, however, probed smaller-size objects than our observations and the difference in slopes may be reflective of a change in size distribution around H ∼ 8.5 where the CFEPS detections dwindle. Thus, in the limited size ranges probed by these surveys, there appears to be reasonable agreement on the slope of luminosity function for these components of the Kuiper belt with the hot and cold components exhibiting slopes that are significantly different. The value of size distribution slopes reported here range from 0.8–1.2 and are considerably larger than the best-fit slopes discussed in many previous analyses that attempted to determine a global luminosity func- tion for the Kuiper belt. For example, Petit et al. (2008) reviewed estimates of α ranging from 0.5–0.8 for surveys that cover the range Hg (cid:39)5–10. Fraser & Kavelaars (2009) and Fuentes et al. (2009) demonstrated that a slope of ∼ α = 0.75 is a decent representation of the 'average' belt down to magnitude ∼ mr=25, but that there is a gradual flattening of the apparent luminosity-function slope at fainter magnitudes, continuing to a slope which may become extremely flat somewhere beyond H > 10 according to the Bernstein et al. (2004) analysis of a deep HST search. The quest for a single 'master' luminosity function, however, is misguided: 1. Because there are different slopes for the hot and cold main-belt components, the slope should be α (cid:39)0.8 at large sizes (where the hot component dominates) and become steeper (if looking in the ecliptic where the cold population is visible) when the depth of the survey results begins to probe the size range at which the cold-population surface density becomes comparable to the hot population. 2. The on-sky density of the (essentially non-resonant) cold population is essentially dependent only on the ecliptic latitude. The hot population's sky density varies with both latitude and longitude due to the fact that the resonant populations are hot. Thus, the H magnitude at which the steeper cold component power-law takes over will also depend on the latitude and longitude of the survey. Interestingly, extrapolating from the ∼4000 objects in the cold belt with Hg ≤8 (see sect. 5.1.3) to – 22 – larger objects, one finds that there should be only ∼1 TNO with Hg <5. This is consistent with the current census of large objects in the cold belt, which should be close to complete (Trujillo & Brown 2003). Simi- larly, one would expect to have only ∼1 TNO with Hg <3.5 in the non-resonant hot population, which again corresponds to our knowledge of the Kuiper belt (Brown 2008). Currently, the MPC report 6 objects with absolute magnitude <3.5 in the classical belt region as defined for our population estimate. 5 of them are clearly part of the hot population, with inclinations between 20 and 30 degrees, the last one being Quaoar with an intermediate inclination of 8 degrees. The realization that the hot component has a low-i tail means that caution must be exercised because one simply cannot isolate the 'cold' cosmogonic population with the commonly-used i < 5◦ cut. For example, in the ecliptic at bright (say roughly mr ∼ 22) magnitudes, the low-i tail of the hot component can be numerically comparable to the sky density of 'cold' objects. Thus, it is not possible to isolate the cold component at bright magnitudes based simply on orbital inclination. 5.1.2. Acceptable range for main parameters. In this section we fix the slopes just determined, i.e. αh = 0.8 and αc = 1.2 and examine the range of model parameters allowed by the L7 detections. Due to the large number of orbital parameters to adjust and the time required by each survey simulation (10–50 minutes on the fastest available computers), we did not run an automated minimum-finding algorithm, but rather did a manual search on a multidimensional parameter grid. Acceptable values (rejectable at less than 95% confidence) for the inclination width (see Appendix A) of the hot component σh range from 14◦–29◦. A hot-component width σh = 16◦ is acceptable not just for the main-belt population but also reproduces the observed inner and outer classical populations (see Sections 5.2 and 5.3) and thus we adopt this value as the width of hot component. The acceptable range for σc is 2.3◦-3.5◦, with a peak of the probability near 2.6◦, which we adopt. Brown (2001) analysed the MPC database at the time and concluded the existence of the cold component to the inclination distribution; with σc = 2.2+0.2−0.6 degrees (one-sigma uncertainties), consistent with our results. Elliot et al. (2005) in their initial analysis of the Deep Ecliptic Survey estimated a 1.94 ± 0.19-degree width for the cold component. Gulbis et al. (2010), however, recently re-analysed the detections from the Deep Ecliptic Survey, and found a 2.0+0.6−0.5-degree width (one-sigma uncertainties) for the cold component. Thus, the DES is also in reasonable agreement with our results, given the uncertainties. Brown & Pan (2004) found a much narrower width of 1.3◦ (no uncertainty given) for the cold component, with respect to a locally-determined Laplace plane for each semimajor axis. We have not repeated a similar analysis. The L7 distribution contains an excess of intermediate-inclination objects (i in range 6◦–10◦) when compared to models with σh ≥ 16◦ and σc = 2.2◦. This 'bump' in the cumulative inclination distribution can also be seen in the DES sample, Millis et al. (2002, Fig. 13) and Elliot et al. (2005, Fig. 17), between inclinations of 8◦ and 10◦. The Survey Simulator approach accounts for the distributions of all orbital elements simultaneously and thus the L7 model makes the inclination bump part of the cold component 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 – 23 – 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 because the objects in this inclination range have e and a distributions that make them part of the cold component, hence increasing its width. Although it was possible to keep a cold width of 2.2◦ or lower by introducing a third inclination component the observations do not currently demand this increase in complexity. The observed cumulative inclination distribution has two steep increases corresponding to the cold and hot component, both of which are steeper than for our model. This indicates that the actual differential distribution of each component is probably more confined than sin(i) times a Gaussian centered on zero. It is remarkable that the hot component of the main classical belt extends up to 35◦ and stops abruptly. This limit is seen not only in the CFEPS, but also in the MPC databases (see Figures 2 and 3). We experimented with sin2(i) times a Gaussian centered on zero, but this did not result in a significant improvement to our fit. Note that Elliot et al. (2005) find that sin(i) times a Gaussian plus Lorenzian give their best fit to the classical belt inclination distribution. More recently, Gulbis et al. (2010) find that sin(i) times a Gaussian of width ∼ 7◦ and centered around ∼ 20◦ best fits what they call the 'Scattered Object' inclination distribution. We did not test this functional form as this introduces an extra parameter, which is not demanded by the current sample. The Brown (2001) functional form may not be an exact representation of every sub-component's inclination distributions; we can, however, obtain an acceptable match to the CFEPS survey with this functional form. The fraction of each component (hot versus cold inclination components) varies with the Hg-magnitude limit, due to their differing values of α. We report here the acceptable range for the fractions of each sub- population at the Hg ≤8.0 limit. We find that the fraction of the hot component, fh, cannot exceed 62% and is at least 33%, with a best match to the observations at fh (cid:39) 0.51. This hot-component fraction and widths are close to the nominal L3 model from P1. We find that the fraction of the kernel fk has to be larger than 0.05, but less than 0.30 at 95% confidence and adopt fk = 0.11. The fraction in the stirred is then fs=0.38, when considering Hg < 8.0. Figure 6 presents the comparison of our nominal model with a, e, i and g apparent-magnitude distri- butions. When biased by the CFEPS survey simulator, the L7-model reproduces the detections extremely well. Our hot/cold population fractions differ from those reported in some other works, but details are impor- tant in the comparison. Brown (2001) report a hot fraction of 81%. This fraction listed must be treated with the caution engendered by the realization that the MPC sample has a non-uniform H-magnitude limit, mak- ing interpretation of a fractional population (given the different luminosity functions) difficult. The Gulbis et al. (2010) estimate is even more difficult to compare, because the classification scheme used explicitly separates out many of the highest-inclination main-belt TNOs into portions of the 'scattered' population (even though many of these TNOs are very decoupled from Neptune) and thus the relatively small 'hot' width of 8+3−2 has been forced down; a direct comparison of the relative populations is thus not possible. Tru- jillo et al. (2001) has a H-magnitude limit that is more uniform than the MPC sample but they mix together the various orbital classes when reporting the relative fraction of hot and cold component objects. 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 – 24 – 5.1.3. Population Estimates The procedure in Section 4.3 of P1 was used to derive a population estimate for the main classical belt. Unlike much of the literature, which gives population estimates for objects larger than an estimated diameter, CFEPS gives population estimates for absolute magnitude smaller than a given value of Hg, and thus an unknown albedo is not introduced into the estimate4. These estimates and their uncertainties are given assuming our orbital model. They would change if we were to change our parameterization. In particular, increasing the width of the inclination distributions 'hides' more of the population far from the ecliptic. Alternately, decreasing the cold component's width to 1.3◦ requires changing the hot/cold fraction and results in a decrease of the total main-belt population by 20%. In principle, the very deepest blocks in our survey are sensitive to a limit of Hg (cid:39) 9.5 for a perihelion detection on the most eccentric orbits in our main-belt model. The Survey Simulator shows, however, that based on our model orbit distribution, the vast majority of our detections should have Hg ≤ 8.0, consistent with our largest-H classical-belter detection, Hg = 8.1. Thus our population estimate is given to the limit to which the survey has reasonable sensitivity: Nclassical(Hg ≤ 8.0) = (8000+1800−1600) where the uncertainties reflect a 95% confidence limit assuming the underlying orbital model and its param- eter values are correct. Our measured value for α essentially is only for the range Hg = 7–8 which dominate our detections. The formula N (Hg ≤ H1) = 10α×∆H N (Hg ≤ H0) , where ∆H = H1 − H0, allows one to scale population estimates of P1 to Hg = 8.0, and also compare with other populations like the inner belt or the plutinos (which can come closer to Earth than the main classical belt). Here care must be taken to distinguish between the hot components and the others because they have different H-magnitude slopes, hence the extrapolation factor to any particular H-limit is different for each sub-component. Figure 7 shows a schematic representation of the fractional population sizes of all the dynamical classes measured in the L7 model. This figure demonstrates that one must be careful when comparing the relative sizes of various sub-populations whose size-distributions are different because the relative populations will vary with the H-magnitude limit being considered. Due to the lack of phase relations with Neptune and good statistics due to large numbers of main-belt detections in ecliptic surveys, the main classical-belt population estimates should be the most certain in the literature of all the populatoin estimates. Comparing the L7 main-belt estimate with the literature yields satisfactory agreement (details are given in Appendix C). Table 5 provides our current population estimates, after accounting for the size distribution scalings and using the same assumptions as in P1, i.e. an albedo 4More subtly, surveys at different latitudes and longitudes probe different average distances as they look into the trans-neptunian region due to the different distance distibutions of resonant and non-resonanat populations; thus a given apparent magnitude depth actually probes at different average size limit. Stating a population limit to a stated H-magnitude limit is thus more meaningful. – 25 – 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 of pg=0.05, hence Hg(Dp=100 km) = 9.16. Hahn & Malhotra (2005) give an essentially-identical estimate of 130,000 TNOs with 40.1 < a < 47.2 AU and D > 100 km (with no error estimate), which is certainly within our 95% confidence region even with the small differences in albedo and phase-space boundaries used. If we use constant α values for the two components and extrapolate to D >100 km, we find the same population estimate as Hahn & Malhotra (2005) and are a factor of a few higher than Trujillo et al. (2001) (see Appendix C). Our current estimates agree with P1 when scaled to the Hg < 8 limit where CFEPS is sensitive. Thus, it appears that the main belt's population for Hg < 8 is secure, where we have provided the first detailed breakdown of the hot and cold component's individual populations and detailed sub-structure. 5.1.4. Discussion Some characteristics of the main classical belt that require explanation are the bimodal nature of the in- clination distribution, the relative importance of the so-called hot component, and the marked sub-structures in (a, e) space for the low-inclination objects. Jewitt et al. (1998), Trujillo et al. (2001), Allen et al. (2001), Trujillo & Brown (2001) and Kavelaars et al. (2008), reported the existence of an edge of the Kuiper Belt at 47–50 AU. Because the samples on which they based their estimate were heavily biased towards low-inclination objects, they were really detecting an edge of the cold component of the classical belt. In addition, Figure 14 of Trujillo et al. (2001), Figures 2 and 3 of Trujillo & Brown (2001), and Figure 3 of Kavelaars et al. (2008) all show a marked peak at around 44 AU followed by a very fast decrease in the number of objects past 44.5–45 AU, with perhaps a low density tail past 50 AU. The above papers vary in how sharp they consider the "cut off" to be. In hindsight it is clear that what they were reporting as an edge is in fact due the presence of the low- inclination kernel and stirred components, which dominate the low-latitude detections and fall off quickly beyond 45 AU. As Kavelaars et al. (2008) point-out, the peaked nature of the distribution is absent in the 'hot' component and entirely absent from the 'scattering disk' population. The stirred component's density is a rapidly-decreasing function of semimajor axis that becomes very small by the time the 2:1 resonance is reached. This hints at a possible connection between the kernel, the stirred component and the migration of the 2:1 to its current location; this outer edge appears only in the low-inclination component. We will show below that a scenario with the hot component continuous across the 2:1 resonance is in agreement with the data. The L7 sample contains a cluster of 6 objects with large e and i just interior to the 2:1 MMR. Amongst these, only the one with a < 47 AU (L4k17, a = 46.967) was included in our analysis of models of the classical belt, the other five being in the region where the exact limit of the resonance is not easy to analytically define. This cluster could very well be a group of objects "dropped out" when the 2:1 MMR shrank at the end of Neptune's evolution (Sec. 7). Gulbis et al. (2006) reported a difference between the B-R color of the 'Core' and the 'Halo', the former being redder than the latter, from photometric measurement they later acquired on the DES sample. – 26 – Table 5. Model dependent population estimates. N (Hg ≤ 8) N (D ≥ 100km) Inner Classical Belt All 400+400−200 3, 000+3,500 −2,000 Main Classical Belt Hot Stirred Kernel All 4, 100+900−800 3, 000+700−600 900+200−200 8, 000+2,000 −2,000 35, 000+8,000 −7,000 75, 000+17,000 −15,000 20, 000+5,000 −5,000 130, 000+30,000 −27,000 Outer/Detached Classical Belt All (a >48) 10, 000+7,000 −5,000 80, 000+60,000 −40,000 Note. - Our model estimates are given for each sub-population within the Kuiper belt. The un- certainties reflect 95% confidence intervals for the model-dependent population estimate. Values for N (D>100 km) are derived assuming an albedo of pg=0.05, hence Hg = 9.16. Remember that the rela- tive importance of each population will vary with the upper Hg limit. – 27 – Our orbital survey was also not designed to yield precision photometry, and the g − i and g − r colors that we can obtain from Table 7 are too uncertain to address this point. In Sec. 8 we discuss some consmogonic implications of these features, review how well the current models reproduce them, and propose future directions. 5.2. The inner classical belt The 'inner' classical belt is the non-resonant and non-scattering population between Neptune and the 3:2 resonance. Paper P1 contained only two such TNOs, preventing us from deducing a detailed description of this region of the Kuiper belt. There are six inner classical belt objects in the L7 sample (see Table 3), providing the opportunity to start constraining an orbital distribution. The phase space is cut by the ν8 secular resonance which eliminates almost all inner-belt TNOs with 7◦ < i < 20◦ making the intrinsic inclination distribution difficult to interpret. If one uses a definition of 'cold' belt as those objects with i < 5◦ (eg., Lykawka & Mukai 2007), one concludes that a large fraction of the inner belt is cold. Such an analysis, however, neglects the bias towards detecting the lowest-i TNOs from the hot population in ecliptic surveys and the removal of moderate inclination objects via the ν8. Determining the intrinsic orbital distribution of the inner belt is precisely the sort of problem in which a simulator approach provides a clearer understanding. 5.2.1. Parametric Model For the inner-belt population we utilized the same form of semi-major axis and perihelion distance distributions as for the hot component of the main classical belt (see Appendix A), changing the range of semi-major axis to be 37 < a < 39 AU and fixing the size distributions for the hot and cold components to be the same as those found for the main belt populations. We then attempted to find a model that included both a hot and a cold component using the same inclination widths and fractions as for the main belt, these models were rejected at >95% confidence. Using the same (a, q) model but with a single-component inclination distribution width of σh = 16◦, like the main belt's hot component (cutting away 7◦ < i < 20◦ orbits as they were proposed) provides a perfectly-acceptable match to the L7 inner-belt detections. In fact, inclination widths of 5◦ < σh < 20.0◦ were found to be acceptable. Even restricting one's attention only to the inner-belt TNOs with i < 7◦ (inclinations below ν8 instability region) still requires an inclination distribution wider than the cold component of the main belt, indicating that the evidence against an inner- belt cold component comes from not just the largest-i detections. 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 – 28 – 5.2.2. Population Estimates Using a single component model with σh = 16◦ and α = 0.8 we determine Ninner(Hg ≤ 8.0) = 400+400−200 (Table 5). This estimate is in good agreement with the L3-sample's estimate of 290+690−250. As before, the uncertainties reflect 95% confidence limits given the intrinsic model distribution and does not reflect our uncertainty in the model. These random uncertainties are a factor of two, due to the small number of inner belt detection. 5.2.3. Discussion Romanishin et al. (2010) compared photometric colors of inner-belt TNOs to other categories and found a good match between the inner belt and the high-inclination objects from the main belt, while a marked difference from the low inclination objects from the main belt, supporting the "hot-only" hypothesis. To attempt to duplicate the conclusion, we compared our photometric data for each population. Unfortunately, but also unsurprisingly, the quality of our photometric data is insufficient for such a comparison. The median uncertainty on our g − i and g − r colors is ∼ 0.25, which is about five times more than for the Romanishin et al. (2010) data. We are thus unable to provide additional verification from our current photometric colors. The successful use of the same orbital distribution for the inner belt and the main-belt's hot component suggests that the entire inner belt may be the low-a tail of the hot main belt. This would be a cosmogonically appealing unification of the sub-populations of the Kuiper Belt. If true, then (at least to order of magnitude) the TNO linear number density at the boundary (we chose 40 AU) extracted from each model should be comparable. Denoting P (Hg < 8.0) as the number of objects per AU with Hg < 8.0, we find Pinner(Hg < 8.0, 40 AU) = 270+180−100 AU−1. For the hot main belt, Pmain(Hg < 8.0, 40 AU) = 670+160−140 AU−1. At this interface, the hot main-belt number density is ∼ 3 times that of the extrapolated inner belt. Given the very uncertain nature of these estimates and the fact that they are anywhere close leads us to postulate that the inner-belt and hot-main TNOs were emplaced by a single cosmogonic process. In this hypothesis, the reduced inner-belt density would be due to the smaller volume of stable phase space in the inner belt region (because there is a smaller available stable range of e) as well as the significant range of inclinations from 7◦ to 20◦ destablized by the ν8 . Scaling the inner-belt population density, to account for this reduced inclination range, results in Pinner(Hg ≤ 8.0, 40 AU) = 500+300−200 AU−1, consistent with the value from the main belt estimate at the 2σ level. Figure 8 presents the linear number density versus a for the scaled inner belt5 compared to those of hot main belt and outer+detached populations. In this picture, the lack of a cold inner-belt component is significant. Assuming that the cold component originally existed in this region, the plausible mechanism for the cold component's destruction is the ν8 resonance sweeping out through the inner belt at some time, eliminating all low-i TNOs. The nearby 3:2 mean-motion resonance also lacks a cold component (Brown 2001; Kavelaars et al. 2008), which argues 5The 7-20◦ portion of a sin i exp(−0.5i2/(16◦2) inclination distribution accounts for 46% of the sin i-weighted phase space; to correct a population in the remaining phase space back to the original needs to be multiplied by 1/(1-0.46) = 1.85. – 29 – 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 that if it swept slowly through the 36–39 AU region the cold component must have already been removed; otherwise, Hahn & Malhotra (2005) show that the low-i objects should have been captured into the 3:2 and preserved (because the 3:2 shields its members from the effects of the ν8). One possible interpretation is that the 3:2 only obtained particles from a scattering hot population (as in the Levison et al. (2008) model) and ended with a large jump to its current location, but a reason for the lack of a cold population inside 39 AU would need to be provided. Because the 3:2 location depends on only the semimajor axis of Neptune, one might expect that the 3:2 resonance's arrival at its current value would occur before the ν8 reaches its current location due to the latter's dependence on the orbital elements of multiple planets and the existence of other remaining mass in the system (Nagasawa & Ida 2000). 5.3. The outer edge of the hot belt We successfully construct a model of the non-resonant, non-scattering TNOs with semimajor axis beyond the 2:1 resonance by simply extending the L7 main-belt model out into this region. Using the classification system from Gladman et al. (2008), our current sample contains 3 outer-belt TNOs and 11 detached TNOs; the distinction between them is set by an arbitrary cut in eccentricity at e = 0.24. For our current analysis, we group these two populations, under the hypothesis that they share a smoothly-varying orbital distribution.6. In order to avoid problems with the exact border of the 2:1, we start our modelling at a = 48 AU; this eliminated 1 detached TNO, reducing our sample to 13. The outer/detached objects share the same (q, i) distributions as the hot main classical belt. This suggests that again (as for the inner belt) the outer population may be a smooth extension of the main-belt hot component. To model the outer/detached TNOs, we thus use the same prescription as for the hot-main classical belt, with α = 0.8 and an a range from 48 AU to a value amax, with density varying as a−β, with β = 2.5. We tried varying the exponent β of the a distribution. For shallow distributions, i.e. β ≤ 1.5, the model is rejected when amax exceeds ∼100 AU, because it creates too many simulated detections close to amax. The range 2.0 ≤ β ≤ 3.0 produces acceptable models with no constraint on amax. Models with larger values of β exhibit a very steep decrease of number density at large a and fail to produce enough detections with a >60 AU. We thus adopt β = 2.5, as for the main classical belt. The number of objects needed to reproduce our 13 outer/detached detections is insensitive to our choice of amax due to the strong detection biases. Hence we formulate our population estimate for a population with no outer edge, finding a population beyond 48 AU of Nouter/detached(Hg ≤ 8.0) = 10, 000+7,000 −5,000 (see Table 5). Of these, only a small number Nouter(Hg ≤ 8.0) = 500+350−250 have e < 0.24, thus belonging to the outer belt defined by Gladman et al. (2008). As for our analysis of the inner belt, we computed the number density of TNOs per unit a at a main/outer interface at 47 AU, Pouter/detached(Hg ≤ 8.0, 47 AU) = 340+230−150 AU−1 and compare it to 6Although it remains to be seen if the very large-inclination objects like Buffy (Allen et al. 2006) or Drac (Gladman et al. 2009) are part of such a distribution. – 30 – the value from the outer edge of the main belt Pmain,outer(Hg ≤ 8.0) = 490+110−100 AU−1. Hence the TNO number density per unit a in the outer/detached belt is the same as that of the hot main belt, within uncer- tainties. There was absolutely no coupling in the debiasing procedure of these two TNO populations; this matching result was not tuned in any way. Figure 8 demonstrates that an initially-uniform semimajor axis distribution for all three of these Kuiper belt sub-components as a single dynamical population is a plausible scenario. Given the number of papers discussing a noticable edge to the distribution (Jewitt et al. 1998; Allen et al. 2001; Trujillo & Brown 2001) this continuity may be surprising. Realize that the continuity is in the hot component, which our analysis indicates is actually present throughout the region from Neptune to at least several hundred AU. This population has a pericenter distribution with very few q's above 40 AU, and may very well have been emplaced as a sort of fossilized scattered disk (Gladman et al. 2002) as illustrated in Morbidelli & Levison (2004). This same process, however, does not emplace the kernel and stirred components which dominate the main-belt region for Hg > 8.0, nor produce the dramatic fall-off beyond 45 AU in these cold populations. 6. The scattering disk If the Centaurs and then JFCs do indeed come from one of the Kuiper Belt's sub-populations, then their penultimate meta-stable source will be the set of TNOs currently scattering off Neptune, as defined by Morbidelli et al. (2004) and Gladman et al. (2008). Hence we wish to give a population estimate for this 'actively scattering' population. Unfortunately the region occupied by the scattering objects is not a simply- connected region definable by a simple parameter-space cut; they are intimately mixed with stable resonant and non-resonant objects and providing a full dynamical model of this region is well beyond the scope of the current manuscript. Here we examine available models of the scattering disk using the CFEPS Survey Simulator to provide an order of magnitude population estimate for this important transient population. The definition of the scattering population has evolved over the last 15 years. A cosmogonic perspective is easily adopted by workers doing numerical simulations. In such simulations the 'scattered' disk is taken to be comprised of TNOs that currently do not have encounters with Neptune but were delivered onto those orbits via an encounter. Morbidelli et al. (2004) quantified this definition by requiring that scattered TNO needs to have it semi-major axis change by more than 1.5 AU over the life of the Solar System. In this process, knowledge of orbital history is required for classification and, clearly, this information is not available for a given real TNO. More problematically, if the Levison et al. (2008) model is correct then the entire Kuiper Belt would qualify as having scattered off Neptune, making the term scattered disk object a meaningless distinction. Gladman et al. (2008) proposed a practical definition for classification based on the orbit of known objects at the current epoch, in which the 'scattering objects' are those currently (in the next 10 Myr) undergoing scattering encounters with Neptune in a forward simulation. In the current manuscript we consider two definitions of the scattering disk, one based on a parameterized region of phase space and one based on numerical modelling of a particular scattering process, to derive an estimate of the scattering 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 – 31 – 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 population. Trujillo et al. (2000) and Hahn & Malhotra (2005) both give population estimates of the scattered Kuiper Belt, but based on different definitions of this population. The former called the scattered Kuiper Belt the region of phase space 50 AU ≤ a ≤ 200 AU and 34 AU ≤ q ≤ 36 AU. Based on their detection of 4 objects with preliminary orbits in this region, they provide a population estimate of 18,000–50,000 objects (1-σ range) with D > 100 km, assuming an Hg distribution slope of 0.8. Using CFEPS and the same orbit and Hg distributions as Trujillo et al. (2000) we estimate the population in that region of the phase space to be 2,100–17,500 objects (95% confidence range), about a factor of 4 less than Trujillo et al. (2000)'s estimate. This estimate is based on scaling the Survey Simulator's detections to match all the L7 detections in this range of q. Awkwardly, none of the L7 detections with orbits in this q range are actually members of the scattering class, thus this estimate is more correctly an estimate of some restricted portion of the Detached population. Two of the objects (1999 CV118 and 1999 CF119) used by Trujillo et al. (2000) for their population estimate were later found to not have orbits in the region they termed the scattered disk, hence their population estimate of this region should be divided by 2, making it more compatible with our estimate7. The choice of the q=34–36 AU region was motivated by the candidate 'scattering' orbits known to Trujillo et al. (2000), intending this to be a source region for the Centaurs and JFCs, as postulated by Duncan & Levison (1997). However, the majority of the known TNOs in that phase space cut are not currently interacting with Neptune and are on resonant or detached orbits (Gladman et al. 2008). A simple phase-space cut is not appropriate for the scattering disk population. To obtain an order-of-magnitude population estimate via a dynamical model, we used the result of nu- merical integrations by Gladman & Chan (2006). This model attempted to produce the detached population via secular interaction with rogue planets, where the additional planet persists for the first 200 Myr of the simulation. Gladman & Chan (2006) find that the scattering particles in their simulations that survive to the end of the a 4 Gyr integration largely forget their initial state. To obtain a scattering disk model, we selected the orbital elements of actively scattering test particles during the last 500 Myr of one 4.5-Gyr integration. We then slightly smeared the orbital elements and applied an H-magnitude distribution with slope α=0.8. We found a reasonable match between the orbital elements of the L7 scattering sub-population and our input scattering model, as observed by the Survey Simulator, althought the inclination distribution was somewhat too cold, yielding a confidence level of only 8%. The apparent magnitude distribution was rejected at more than 99% whatever the slope of the Hg distribution we used; it is plausible that this rejection is due to a change in luminosity-function slope in the size range probed by our observations, as the faintest absolute magnitude of our detections is Hg = 10 because scattering TNOs include many q < 30 AU members. Three of the CFEPS active scatterers were inside 30 AU at the time of detection. The match between the orbital model and the observations allow us to be reasonably confident our population estimate is good to a factor of ten and we do not feel this order of magnitude estimate warrants further tuning until a larger sample of scattering objects is in hand. While there is clearly future room to better test models, we give here the first published estimate of the active scattering population. The results are given in Table 6 for Hg < 10 and for 7The other two sources (1999 TL66 and 1999 CY118) are found to be on scattering obits (Gladman et al. 2008). – 32 – 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 diameter D >100 km (Hg < 9.16 assuming an albedo of pg=0.05). The quoted factor-of-three uncertainty accounts only for the Poisson variation. We estimate an actively-scattering population that is about 2-3% that of the sum of the classical belts. Interpretation of this number is problematic. A very large actively-scattering population would require that the current disk could not be the steady state intermediary between the Centaurs and a longer-lived source in the trans-neptunian region, in which case the currently actively-scattering population is more likely to be the long-lived tail of a roughly 100× more populous primordial population (Duncan & Levison 1997). For the 2-3% figure, the active scatterers could conceivably be now dominated by objects that have left the resonant, detached, or classical populations in the last Gyr. 7. Testing cosmogonic Kuiper belt models The CFEPS-L7 model is an empirical parametric model that properly reproduces the observed orbital distribution of the Kuiper belt, once passed through our survey simulator. The purpose of this parametric model is to provide absolutely-calibrated population estimates of the various sub-populations of the Kuiper belt. The model also exhibits important features of the intrinsic Kuiper belt that a cosmogonic model should reproduce. For example, one needs to produce a cluster of objects at low inclination and low eccentricity near 44 AU, that we call the kernel. There is also a low-i component extending from the outer edge of the ν8 secular resonance at 42.4 AU out to the 2:1 MMR with Neptune. Finally, there is a hot component with a confined q range that extends in semimajor-axis from the inner belt at ∼35 AU out to several hundred AU with a decreasing surface density. The synthetic L7 model is also useful for observational modeling of our Kuiper belt, with Stark & Kuchner (2010) as an example for the outer Solar System dust distribution based on the L3 model. The ability to provide a detailed quantitive comparison with a cosmogonic model is, however, the true power of the CFEPS survey. This is done by passing a proposed model of the current Kuiper Belt distribution through the CFEPS survey simulator and then comparing this detection-biased model with the real CFEPS detections. Through this procedure one can choose between models in a statistically robust way. Both the CFEPS L7 synthetic model and the CFEPS survey simulator are available from the project web site www.cfeps.net. Several models have been proposed to explain the dynamical structure of the Kuiper belt (Malhotra 1993; Ida et al. 2000; Hahn & Malhotra 2005; Levison et al. 2008, to name a few). Since the primary Table 6. Scattering disk population estimates. N (Hg ≤ 10) N (D > 100km) Scattering disk 25, 000+20,000 −15,000 5, 000+5,000 −3,000 – 33 – 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 purpose of CFEPS was to validate or refute cosmogonic models, we present an example of this process. Because we had available both an orbital element distribution and a resonance-occupation analysis (Levison, 2010, private communication), we have chosen to use Run B of Levison et al. (2008) as an example of how one uses the CFEPS Survey Simulator to compare a model to the observed Kuiper belt. The simulation in question (motivated by the Nice model of the re-arrangment of the outer Solar System) has already some known problems pointed out by its authors, but the model's intriguing aspects make it a good example of the comparison process. All the fictitious TNOs in the Run B model were dynamically classified following the Gladman et al. (2008) procedure (Van Laerhoven, 2010, private communication). The final planetary configuration in the Nice model was intentionally made different from that of the Solar System to avoid the ν8 secular resonance inadvertently sweeping through and destroying the belt and thus has many objects in the 40 < a < 42.4 AU range at low inclination, where the real ν8 resonance would eliminate them. To avoid this complication, we restricted the comparison to the range 42.4 < a < 47 AU, yielding 128 non-resonant model TNOs from run B. We use the following procedure to generate the large number of TNOs required as input to the survey simulator. First we select a model object at random, and vary the orbital elements uniformly by ±0.2 AU in both a and q and ±0.5◦ in inclination and then randomize the elements Ω, ω, and M, There was no size distribution given for the model objects, so the Hg magnitude of each object was drawn from an exponential distribution (eq. 5.1.1). Given the orbital elements and Hg magnitude, we then use our Survey Simulator to determine if the model object would have been detected by the CFEPS-L7 obersvations. We repeat the procedure until we have a set number of simulated detections and then compare the a, e, i, q and r one-dimensional cumulative distributions of the simulated detections to those of the L7 sample, using the Anderson-Darling test. For the Hg distribution, we tried single slopes of 0.6 ≤ α ≤ 1.3 and also a model with α = 1.2 for the low-i and α = 0.8 for the high-inclination objects, as in our favored model. All the models for the Hg distribution produced acceptable matches to the apparent magnitude distribution, but had no effect on the orbital element distributions, so we do not show the magnitude distribution. Fig. 9 compares the distribution of a, e, i and r of the L7 sample and simulated detections from the model. There is remarkably-good agreement for the a distribution. Although the Nice model does not exhibit a clustering around 44 AU as strong as the L7 sample, the difference between the two distribution is not statistically significant. On the other hand, the model's e distribution is too excited compared to the observed one, as already noted by Levison et al. (2008). This then results in detection distances that are overly dominated by small- distance detections. For both cumulative distributions, the Anderson-Darling test says that the hypothesis that the observed objects could be drawn from the model can be rejected at >99.9% confidence. The model's i distribution is not a good match either, again as already noted by Levison et al. (2008); the AD test rejects the i distribution at more than 99.9% confidence. This is mostly because the L7 distribution has two components, while the Run B input model gives an essentially unimodal distribution. The simulated – 34 – detections from the model appear roughly consistent up to i (cid:39) 4◦, but there is a lack of high-i TNOs, to which the Anderson-Darling test is sensitive. Run B lacks the hot component that peaks between 15◦ and 20◦ and extends past 30◦. Looking only at the i < 6◦ region, here too the AD test rejects the model at more than 99.9% confidence. Most of the Run B classical belt comes from the low-i outer part of the planetesimal disk, which acquires inclinations similar to that of the inner initial disk, with a width 6◦. The true cold component, on the contrary, has a width certainly less than 3◦. We conclude that one needs a strongly bi- modal input population to the Nice model in order to produce the desired bi-modal inclination distribution we see in the real Kuiper belt. More worrying is the claim by Levison et al. (2008) that increasing the width of the initial population produced the same final inclination distribution, meaning there is a missing ingredient in this model to explain the dynamical structure of the Kuiper belt. Levison et al. (2008) mention another simulation, Run E, which generated too many hot objects compared to the cold population; so there may be an intermediate parameter set that could match the observations. 8. Conclusions This paper's modeling concerns the non-resonant Kuiper Belt, although the L7 release lists all detec- tions from the CFEPS survey fields from 2002–2007 for the sake of completeness. Due to the complexity of the modelling required because of the phase relations with Neptune, the resonant populations are presented in a separate paper (Gladman et al. 2011). We find that the debiased orbital and H-magnitude distributions show that there is considerable sub-structure in the main Kuiper Belt. We quantitatively measured the size of the various sub-populations, create an empirical model of these sub-populations in the L7 synthetic model, and provide an algorithm (the CFEPS survey simulator) to quantitatively compare cosmogonic models to the intrinsic Kuiper Belt. Here we summarize the results and offer a synthesis and interpretation. A plausible hypothesis is that the hot population permeates the entire Kuiper Belt region from 30 AU up to at least 200 AU, albeit with a projected surface density (onto the invariable plane) that decreases with semimajor axis. Even the resonant populations are consistent with the idea that the entire hot component is a vestigal "fossilized" scattered disk from an epoch when TNOs with perihelia up to ∼40 AU were being weakly scattered by a massive object at the inner edge of the Kuiper Belt (whether this object was Neptune or something else is unclear from the present data). The inclination distribution of this hot population can be represented by sin (i) times a gaussian of width ∼ 16◦. Note however that, due to the strong bias against detection of large-i objects in an ecliptic survey, our current sample does not provide a strong constraint on the width or the functional form of the hot component. A scenario in which the inner belt, hot main-belt, outer belt and detached populations, along with the resonant populations were all emplaced simultaneously from a population scattered outward during the final stages of planet formation, with a single size distri- bution, initial inclination distribution, colour distribution and binary fraction, is an attractive hypothesis. The plausibly-continuous initial number density across the inner/main and main/outer boundary (see Fig. 8) supports this idea. The Kuiper Belt's (surviving) 'cold' population is entirely confined between semimajor axes of 42.4 AU 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 – 35 – 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 and the 2:1 resonance with Neptune, and its inclination distribution (measured relative to the J2000 ecliptic) is adequately represented via sin (i) times a gaussian of width 2.6◦, with an acceptable range from 2.3◦ up to 3.5◦. There are indeed i < 5◦ TNOs in the low-i tail of the hot population all over the Kuiper belt and even in the 42.4 < a < 47 AU region, so an inclination cut does not provide a clean separation between the hot and cold components of the main belt. In the current belt we claim that all i < 5◦ TNOs with semi-major axis outside the above range are the hot-component objects that happen to have lower inclinations. The cold population exhibits a particularly-strong grouping in band of about 1-AU a thickness, centered at 44 AU (which we call the kernel). The linear number density (#/AU) of 'cold' belt objects increases from the inner edge at 42.4 AU up to a maximum at ∼ 44.4 AU, all with rather low eccentricities. Past 44.4 AU, the linear number density drops noticeably, and classical TNOs tend to have higher eccentricities; the CFEPS- L7 model uses a 'stirred' population that covers the 42.4–47 AU range with a single parameterization. We favor the idea that this cold component is primordial (the objects formed at roughly their current heliocentric distances), although this is not required. The primordial distance range of the cold population is difficult to constrain. The inner boundary at a=42.4 AU may have been eroded via scattering by massive bodies and resonance migration; an important condition is that any sequence of events cannot allow either the inner belt, or the mean-motion and secular resonances that probably migrated through it, to have preserved a cold component today. The coincidence of the stirred population's outer edge with the 2:1 resonance suggests to us that the kernel marks the original outer edge and that the larger-a cold objects have either (i) been dragged out of the a < 44.4 AU region via trapping and then drop-off in the 2:1 as it went past (in the fashion studied by Hahn & Malhotra (2005)) or (ii) due to weak scattering out of the 40 < a < 44.4 AU region. Perhaps the edge of the original cold population around 45 AU may be explained by the global evolution of solid matter in turbulent protoplanetary disks (Stepinski & Valageas 1996, 1997), although an even-more extreme density contrast may be needed at ∼30 AU to prevent Neptune's continued migration outward (Gomes et al. 2004). Sharp drops in surface density are commonly observed in protoplanetary disks at about this 30–50 AU scale (Johnstone et al. 1998; Mann & Williams 2009, 2010). There is an issue with a primordial origin of the cold population at this location. The on-ecliptic mass density of this population is extremely low and it would be difficult to form multi-hundred km TNOs in a low surface density environment. This may not be impossible due to recent work on forming planetesimals big (Morbidelli et al. 2009a; Youdin 2011), which can be favored by external photoevaporation (Throop & Bally 2005), and may be supported by the fact that it appears that there are simply no cold objects larger than H ∼ 5; all the larger objects are in the other populations which may come from closer to the Sun where the mass density was higher. The kernel around 44 AU is an intriguing feature. A collisional family explanation would eliminate the idea that the 44.4-AU edge is a primordial edge, but would instead simply be where the very-low velocity dispersion breakup occured. This velocity dispersion is even lower than for the putative Haumea collisional family (Brown et al. 2007). An additional puzzle is the unclear significance that the kernel is bounded between the 7:4 and the 9:5 MMRs . One possible, very ad-hoc, explanation would be that the 2:1 MMR started its migration interior to 43.5 AU while being wide (due to a large Neptune eccentricity), and then – 36 – 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 had a stochastic jump by a few tenth of AU while near 44 AU, leaving behind a pile of objects that we see as the kernel today. The hot population poses other strong constraints. Models by Gomes (2003); Hahn & Malhotra (2005); Levison et al. (2008) all succeed in creating a hot population that has a similar radial extent to what is cur- rently observed, but have varying success in matching the inclination distribution. When slowly migrating Neptune over long distances (> 8 AU) into an initially-cold disk, Gomes (2003) and Hahn & Malhotra (2005) generated a reasonable TNO fraction with inclinations up to 35◦. When migrating Neptune over a shorter distance (Gomes 2003), or in a hot disk (Hahn & Malhotra 2005), the fraction of high-i TNOs is noticeably reduced, while still reaching the same maximum i. Levison et al. (2008), on the contrary, mi- grate Neptune over a short distance (2–3 AU) into a warm scattered disk (with (cid:104)i(cid:105) = 6◦) and essentially maintain the input inclination distribution. They report that increasing the initial i distribution resulted in the same final population, which lacks TNOs with i > 30◦ and which we have confirmed is colder than the actual belt. These facts appear to indicates that Neptune had to slowly migrate over a long distance in a cold disk in order to obtain the observed inclination distribution of the hot population. However, Morbidelli et al. (2009b) showed that a long and slow migration of Neptune, coupled with a similar migration of the other giant planets, does not correctly reproduce the secular architecture of the solar system, in particular the amplitudes of the eigenmodes characterising the current secular evolution of the eccentricities of Jupiter and Saturn. They conclude that only the Nice model can reproduce the current dynamics of the inner solar system and the giant planets. Unfortunately this scenario does not produce Kuiper-Belt components with orbital properties that agree with the L7 orbit catalog. The idea that the hot population originated from a planetesimal population scattered outward by Nep- tune, whose resonant and largest-q members are preserved, is extremely attractive. Thus much of Levison et al. (2008)'s general scenario has many pleasing aspects and one is tempted to think of the hot population as the transplanted population, even if our results show that the inclination distribution is a stumbling block. Contrary to some statements (eg Fraser et al. 2010), we find that the Nice model is not good at producing the the hot population's inclination distribution, but suprisingly produces rather well the cold population's fine structure in the semimajor axis distribution. The large-i TNOs which do appear could instead be interpreted as coming from the 'evader' mechanism of Gomes (2003). In this conception the Fraser et al. (2010) finding, that the luminosity distribution of the Jovian Trojans is more similar to the cold than hot TNO populations, makes perfect sense in a scenario in which the injection of bodies into the Jovian Trojan region occurs from the same source region as the implantation of the Kuiper Belt's cold population. In the Nice model this seems unlikely because jovian Trojan capture occurs just after the Jupiter-Saturn mutual 1:2 resonance crossing (Morbidelli et al. 2005) and involves small bodies closer to the planets than the cold outer disk that is the main source of the cold Kuiper belt. If so, the hot component cannot be generated from the Nice model's inclined inner disk, as this would have the same size distribution as the Jovian Trojans. One needs another source for the hot population, one that is not too perturbed by the initial instability in Neptune's motion. A final caveat concerns the existence of wide binaries in the cold belt; Parker & Kavelaars (2010) showed that the Neptune scattering occuring in the Nice model would disrupt nearly all such wide binaries, thus requiring a more gentle mechanism to move the cold-belt to its current location if that population did – 37 – not form in-situ. Our current understanding of the trans-neptunian region is not likely to advance rapidly for time scales of order a decade unless new surveys begin to efficiently probe TNOs that were rare in the ecliptic surveys. The most likely approach that would result in an advance are moderate-depth (24th magnitude) wide-field surveys (many hundreds of square degrees) at higher ecliptic latitudes, or deeper (25th magnitude) surveys covering ∼ 100 sq. deg. targeting regions of sky that attempt to isolate cosmogonically-interesting sub- populations. We hope that CFEPS will serve as a standard for the need for well-characterized discovery and tracking. The Large Synoptic Survey Telescope (LSST) (Ivezic et al. 2008) should certainly firm up the main-belt dynamical sub-structure along with the colour and size distributions for those components. A. Appendix A In this Appendix, we give details of the algorithm used to generate the CFEPS-L7 model of the main classical belt. The main classical belt objects are constrained in what is essential 3-dimensional phase-space due to the (confirmed a posteriori) fact that the mean anomaly and longitudes of ascending node and perihelion are all uniformly distributed in the intrinsic population. Thus the L7 model consists of 3 sub-populations constrained by 3 orbital-element distributions to determine for each sub-population. The inclination distribution of each subcomponent is well represented by a probability distribution proportional to sin(i) times a gaussian exp[i2/(2σ2)], where past results indicate a 'cold'-component width of ∼ 2.5◦ and a 'hot'-component width of ∼ 15◦ (Brown 2001; Kavelaars et al. 2008). The hot component occupies the semimajor axis range from 40.0 to 47.0 AU and is defined by: • an a distribution with a Probability Density Function (PDF) proportional to a−5/2, corresponding to a surface density proportional to a−7/2; • an inclination distribution proportional to sin(i) × exp[i2/(2σ2 • we eliminate objects from the region unstable due to the ν8 secular resonance: a < 42.4 AU and h)], with width σh = 16◦; i < 12◦; • a perihelion distance q distribution that is mostly uniform between 35 and 40 AU, with soft shoulders at both ends extending over ∼1 AU; the PDF is proportional to 1/([1 + exp ((35 − q)/0.5)][1 + exp ((q − 40)/0.5)]); any object with q <34 AU is rejected; • finally, we reject objects with q < 38 − 0.2i (deg) to account for weaker stability of low-q orbits at low inclination. 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 We have found that the exact form of the truncation at low perihelion distance is unimportant, as long as the – 38 – limiting value of q is a decreasing function of the inclination; this is justified dynamically as low-inclination orbits cannot have q < 38 AU and remain stable (Duncan et al. 1995). The stirred component covers only the range of stable semimajor axis at low inclinations: • an a distribution with PDF proportional to a−5/2 between 42.4 (limit of the ν8 resonance) and 47 AU; • a uniform e distribution between 0.01 and a maximum value depending on the semimajor axis, emax = 0.04 + (a − 42) × 0.032, to reproduce the structure seen in figs. 2 and 3; • randomly keep objects with probability 1/(1 + exp [(e − 0.6 + 19.2/a)/0.01]), which corresponds to a soft cut at q = 38 + 0.4 ∗ (a − 47); • an inclination distribution proportional to sin(i) times a gaussian of width σc = 2.6◦; • again, we reject objects with q < 38 − 0.2 × i(deg) as for the hot component. Finally, the kernel provides the group of objects with low inclination in the middle of the main classical belt as seen in fig. 2: • a uniform a distribution between 43.8 and 44.4 AU; • a uniform e distribution between 0.03 and 0.08; • an inclination distribution proportional to sin(i) times a gaussian of width σc = 2.6◦, identical to the stirred population. For all components, the remaining orbital elements (longitude of node, argument of perihelion and mean anomaly) are drawn at random uniformly between 0◦ and 360◦. All elements are generated in the invariable plane reference frame (inclination 1◦ 35' 13.86" with respect to J2000 ecliptic plane with direction of ascending node at 107◦ 36' 30.8"). In particular, we state widths of the inclination distribution with respect to the invariable plane. Elliot et al. (2005), Brown & Pan (2004) and Gulbis et al. (2010) studied the distribution of inclinations with respect to their self-determined Kuiper Belt Plane, which differ from the invariable plane. To evaluate the acceptability of each model we evaluate our parameterization in distinct portions of phase-space. • i ≥ 10 deg • i < 10 deg • a > 44.4 AU • a ≤ 44.4 AU 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 • the entire main-belt region – 39 – We computed the probability of the AD or KKS statistics in each region separately and consider the min- imum on all element distributions and all sub-regions when determining if a particular parameterization is rejected. The variable parameters are the i-width of the hot component σh, the cold component's i width σc, the H-magnitude distribution of these two components (with slopes αh, αc), the hot population's fraction of the main belt fh, and the kernel fraction fk, with the stirred component forming the remainder: fs = 1−fh−fk. The CFEPS-L7 model has the following known weaknesses: 1. Resonant orbits will be generated by chance in the main-belt region (especially for the 5:3, 7:4, and 9:5 resonances). 2. The ν8 resonance cut is done in osculating, rather than proper, orbital elements space and thus some L7 model objects objects near the resonance will be unstable. 3. There are four tiny semimajor axis gaps in our model: small regions (∼ 0.3 AU in a) on both sides of the 3:2 and 2:1 resonannces. B. Appendix B The CFEPS project is built on the observations acquired as the 'Very Wide' component of the CFHT Legacy Survey (CFHTLS-VW). All discovery imaging data is publicly available from the Canadian Astron- omy Data Centre (CADC8). These images were acquired using the CFHT Queue Service Observing (QSO) system. For each field observed on a photometric night the CFHT QSO provides calibrated images using their ELIXIR processing software (Magnier & Cuillandre 2004). Our photometry below is reported in the Sloan system (Fukugita et al. 1996) with the calibrations contained in the header of each image as provided by ELIXIR. Color corrections were computed using the average color for Kuiper belt objects (g − r) ∼ 0.7. Differential aperture photometry was determined for each of our detected objects observed on photometric nights and these fluxes are reported in Table 7. All CFEPS discovery observations were acquired in pho- tometric conditions in a relatively narrow range of seeing conditions due to queue-mode acquisition. The photometry below supercedes information that may be in the Minor Planet Center's observational database. 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 8http://www.cadc.hia.nrc.gc.ca – 40 – Table 7. Object Fluxes. Object g L3f01 L3f04PD L3h01 L3h04 L3h05 L3h08 L3h09 L3h11 L3h13 L3h14 L3h18 L3h19 L3h20 L3q01 L3q02PD L3q03 L3q04PD L3q06PD L3q08PD L3q09PD L3s01 L3s02 L3s03 L3s05 L3s06 L3w01 L3w02 L3w03 L3w04 L3w05 L3w06 L3w07 L3w08 L3w09 L3w10 L3w11 L3y01 L3y02 L3y03 L3y05 L3y06 L3y07 L3y09 L3y11 L3y12PD 23.66 22.74 23.83 24.32 24.36 · · · 22.73 23.44 23.73 23.27 23.42 · · · · · · 23.89 23.50 23.19 24.15 23.58 23.67 23.50 23.54 23.81 22.90 23.67 22.82 22.89 23.56 23.76 22.44 24.20 23.65 22.95 23.96 23.53 23.95 24.03 24.04 23.38 23.41 23.89 23.37 23.42 23.65 23.82 21.73 σg 0.41 0.33 0.27 0.17 0.07 · · · 0.04 0.20 0.10 0.15 0.09 · · · · · · 0.21 0.10 0.17 0.43 0.30 0.18 0.24 0.12 0.28 0.20 0.30 0.03 0.61 0.11 0.07 0.03 0.31 0.25 0.09 0.13 0.10 0.20 0.12 0.26 0.09 0.09 0.03 0.18 0.09 0.17 0.32 0.03 Ng r σr Nr 4 4 4 4 2 .. 4 7 4 3 3 .. .. 3 3 4 4 4 3 4 6 6 5 5 5 5 4 5 5 4 4 5 4 3 5 4 4 6 4 4 3 4 4 4 4 23.12 · · · 23.02 23.82 23.56 23.15 22.29 23.13 23.29 22.81 22.53 23.67 23.15 22.96 22.49 22.36 23.31 · · · · · · · · · · · · 23.40 · · · · · · · · · · · · 22.80 22.50 21.65 23.70 23.13 · · · · · · 22.72 23.04 23.49 22.62 22.69 22.79 22.99 · · · 22.87 23.01 23.51 20.81 0.15 · · · 0.11 0.33 0.30 0.75 0.23 0.10 0.27 1.42 0.17 0.26 0.30 0.17 0.07 0.25 0.14 · · · · · · · · · · · · 0.23 · · · · · · · · · · · · 0.08 0.08 0.04 0.17 0.19 · · · · · · 0.15 0.13 0.15 0.55 0.19 0.11 0.12 · · · 0.23 0.14 0.21 0.04 3 .. 3 7 7 7 10 3 7 8 8 9 7 3 4 4 4 .. .. .. .. 4 .. .. .. .. 4 4 4 3 4 .. .. 4 4 4 3 4 4 3 .. 5 5 4 4 i · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 22.76 22.23 22.37 23.10 · · · · · · · · · 22.54 23.18 22.65 22.88 21.89 · · · 22.54 · · · 21.53 23.77 · · · 22.46 22.79 · · · 22.00 23.34 · · · · · · · · · · · · · · · · · · · · · · · · · · · σi · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 0.36 0.02 0.00 0.16 · · · · · · · · · 0.12 0.11 0.23 0.13 0.07 · · · 0.07 · · · 0.02 0.07 · · · 0.09 0.14 · · · 0.63 0.13 · · · · · · · · · · · · · · · · · · · · · · · · · · · Ni .. .. .. .. .. .. .. .. .. .. .. .. .. 3 3 1 3 .. .. .. 2 2 3 2 3 .. 3 .. 3 3 .. 3 3 .. 3 3 .. .. .. .. .. .. .. .. .. – 41 – 36 40 44 48q (AU) 5 15 25 35i (deg) 40 42 44 46 48a (AU)0 10 20 30 i (deg)– 42 – 36 40 44 48q (AU) 5 15 25 35i (deg) 40 42 44 46 48a (AU)0 10 20 30 i (deg)– 43 – 0.000.050.100.150.20eHot component0 10 20 30 i (deg)Stirred component0.000.050.100.150.20eKernel component404142434445464748a (AU)0 10 20 30 i (deg)All components404142434445464748a (AU)– 44 – Fig. 5.- Contour plots of the 'minimum probability' statistic for a range of main classical belt models. Each model has a different slope of the H distribution for both the hot component (αh) and the other components (αc). Contour levels for 1% and 5% probabilities are shown. Acceptable models are interior to the 5% level curve. The dashed line indicates the locus with identical slopes for all components. Th e plus sign indicates the adopted model (which gave the best match). 0.80.91.01.11.21.31.41.5Slope of cold component c0.50.60.70.80.91.01.11.21.3Slope of hot component h1%5%Best – 45 – Fig. 6.- CFEPS+Pre objects (red solid squares) compared to our main-belt model's distribution in a (upper left), e (upper right), i (lower left) and g magnitude (lower right) distributions when the intrinsic (dashed line) are observed through the CFEPS Survey Simulator (resulting thin solid lines). The model used here is the one described in Section 5.1, with values of the parameters corresponding to our nominal case (see Section 5.1.2). – 46 – Fig. 7.- A representation of the fractional populations of the various dynamical classes measured in the L7 model. The surface area of each population shown is proportional to the relative population for objects with Hg ≤ 8 (left) and Hg ≤ 9.16 (right), corresponding to D ≥ 100 km, assuming an albedo p = 0.05. The wedge label "Other reson." refers to resonant populations measured other than those individually labelled (4:3, 7:3, 5:4, 3:1, 5:1; see Gladman et al. 2011) . The outer annulus is comprised entirely of "hot" population objects while the "cold" populations, of which there is only the Kernel and Stirred components, are represented by the inner circle. The white area corresponds to the main belt. – 47 – Fig. 8.- The linear number density (/AU) for three Kuiper Belt components: the inner belt (a < 39 AU), the hot main belt (40 < a < 47) and the outer plus detached belts (a > 48 AU). Each region's total population is scaled to the number with Hg ≤ 8, as determined by our model population estimates. The inner belt's population has been scaled up by a factor of 1.85 to account for the ν8 resonance (see footnote 5). The solid lines represent the model population determined independently for each zone while the grey dashed line indicates the smooth extension of the hot main belt model to the semi-major axis range occupied by the inner belt and the outer+detached populations, where the inner-belt decay at lower a occurs because of the rapidly-shrinking stable (a, q) phase-space volume available. A continuous primordial a−2.5 hot population could, within uncertaties, account for all three populations. This suggests that these three Kuiper Belt components are a single dynamical population. – 48 – Fig. 9.- CFEPS+Pre objects (red solid squares) compared to the Nice model's distribution in a (upper left), e (upper right), i (lower left) and r (lower right) distributions when the intrinsic (black dashed line) distri- bution is biased via the CFEPS Survey Simulator (resulting blue solid lines). This shown case corresponds to a single-slope α = 1.1 Hg distribution. – 49 – Table 7-Continued Ng r σr Nr 4 3 .. .. .. 4 7 3 3 4 .. 6 4 4 4 4 4 4 .. .. .. .. .. 3 3 3 3 3 3 3 3 3 3 4 4 4 4 4 3 4 3 3 4 3 5 5 22.73 · · · 23.03 23.74 23.53 · · · · · · · · · · · · 23.41 23.96 · · · 23.96 · · · · · · 22.57 23.43 23.45 23.96 24.01 24.35 23.79 24.17 · · · · · · · · · · · · · · · · · · · · · 23.48 22.77 23.35 22.72 · · · · · · 22.52 22.92 23.34 23.65 23.74 23.38 23.54 23.43 23.46 22.59 0.10 · · · 0.28 0.27 0.64 · · · · · · · · · · · · 0.31 0.35 · · · 0.21 · · · · · · 2.16 0.25 0.16 0.25 0.09 0.43 0.23 0.33 · · · · · · · · · · · · · · · · · · · · · 0.31 0.75 0.37 0.13 · · · · · · 0.10 0.35 0.13 0.31 0.63 0.25 1.00 0.32 0.40 0.13 3 .. 7 7 7 .. .. .. .. 3 7 .. 3 .. .. 4 3 3 4 3 4 7 4 .. .. .. .. .. .. .. 4 2 6 3 .. .. 4 4 3 3 7 3 3 3 7 7 Object g L3y14PD l3f05 l3h10 l3h15 l3h16 l3q05 l3q07 l3w14 l3w19 U3f02 U3h06 U3s04 U3w13 U3w16 U3w17 U3y04 U3y08 U3y16 u3h02 u3h03 u3h07 u3h12 u3h17 u3w12 u3w15 u3w18 u3y10 u3y13 L4h01PD L4h02PD L4h05PD L4h06 L4h07 L4h08 L4h09PD L4h10PD L4h11 L4h12 L4h13 L4h14 L4h15 L4h16 L4h18 L4h21 L4j01 L4j02 23.68 23.71 · · · · · · · · · 23.68 24.19 23.98 24.00 24.07 · · · 24.10 24.39 24.07 24.45 24.25 24.25 23.93 · · · · · · · · · · · · · · · 24.43 24.07 24.49 24.00 24.28 23.77 23.50 23.96 23.83 23.70 23.01 21.34 22.96 22.98 24.11 23.64 23.68 24.05 24.04 23.34 23.75 23.81 23.35 σg 0.19 0.18 · · · · · · · · · 0.16 0.21 0.20 0.11 0.21 · · · 0.45 0.21 0.14 0.09 0.13 0.13 0.37 · · · · · · · · · · · · · · · 0.15 0.21 0.19 0.09 0.16 0.40 0.18 0.16 0.12 0.08 0.03 0.18 0.09 0.09 0.34 0.23 0.30 0.05 0.45 0.38 0.05 0.18 0.09 i · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 23.14 23.27 · · · · · · · · · · · · · · · · · · · · · · · · 23.48 · · · · · · · · · · · · · · · · · · · · · · · · · · · 22.41 · · · 22.62 22.08 22.60 22.53 · · · · · · · · · 22.01 23.00 · · · 22.37 σi · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 0.25 0.15 · · · · · · · · · · · · · · · · · · · · · · · · 0.26 · · · · · · · · · · · · · · · · · · · · · · · · · · · 0.14 · · · 0.16 0.09 0.01 0.15 · · · · · · · · · 0.25 0.12 · · · 0.12 Ni .. .. .. .. .. .. .. .. .. .. .. .. .. 3 3 .. .. .. .. .. .. .. .. 3 .. .. .. .. .. .. .. .. .. 3 .. 3 3 3 2 .. .. .. 3 3 .. 3 – 50 – Table 7-Continued Ng r σr Nr i σi Ni 5 6 4 5 5 3 8 4 3 4 4 4 4 4 4 4 3 4 4 4 4 4 4 4 5 8 5 5 4 4 4 4 4 3 4 4 4 4 4 4 4 4 3 4 4 4 23.06 22.66 21.70 22.10 22.79 23.06 23.04 23.14 23.11 22.59 22.97 23.22 22.53 23.44 22.99 22.72 23.15 23.32 22.48 22.93 22.55 23.08 23.40 23.19 · · · 22.72 23.33 · · · · · · 22.45 · · · · · · 22.21 23.48 23.07 22.45 22.33 22.74 21.81 · · · · · · 22.88 23.01 23.04 23.54 23.11 0.18 0.12 0.03 0.09 0.20 0.12 0.30 0.11 0.17 0.04 0.34 0.03 0.29 0.28 0.10 0.11 0.12 0.19 0.17 1.27 0.12 0.34 0.17 0.29 · · · 0.10 0.21 · · · · · · 0.11 · · · · · · 0.39 0.41 0.14 0.14 1.43 0.08 0.23 · · · · · · 0.17 0.09 0.10 0.19 0.01 7 6 4 3 7 2 5 3 2 2 9 3 3 4 3 4 3 4 3 5 5 3 6 3 .. 3 3 .. .. 4 .. .. 3 3 2 4 4 6 4 .. .. 7 7 3 3 3 22.91 22.46 21.74 21.91 22.43 22.82 22.77 · · · 23.03 22.18 22.25 22.96 22.34 23.19 22.58 · · · 23.03 22.93 22.12 23.32 22.33 · · · · · · 22.47 · · · · · · · · · 22.14 22.72 22.50 23.05 23.49 21.95 · · · · · · 22.09 21.58 22.38 21.56 24.08 22.87 22.62 · · · 22.84 23.83 22.72 0.15 0.14 0.04 0.04 0.32 0.06 0.26 · · · 0.11 0.15 0.05 0.12 0.19 0.15 0.07 · · · 0.17 0.95 0.02 0.12 0.05 · · · · · · 0.46 · · · · · · · · · 0.00 0.49 0.00 0.15 0.79 0.76 · · · · · · 0.11 0.09 0.16 0.30 0.37 0.14 0.14 · · · 0.41 0.43 0.23 3 3 2 2 3 3 3 .. 3 2 3 3 4 3 2 .. 3 3 2 2 3 .. .. 2 .. .. .. 1 3 1 2 3 3 .. .. 2 3 3 3 3 3 2 .. 3 3 3 Object g L4j03 L4j05 L4j06PD L4j07 L4j08 L4j10 L4j11 L4j12 L4k01 L4k02 L4k03 L4k04 L4k09 L4k10 L4k11 L4k12 L4k13 L4k14 L4k15PD L4k16 L4k17 L4k18 L4k19 L4k20 L4m01 L4m02 L4m03 L4m04 L4n03 L4n04 L4n05 L4n06 L4o01 L4p01 L4p02 L4p03 L4p04PD L4p05 L4p06PD L4p07 L4p08PD L4p09 L4q03 L4q05 L4q06 L4q09 23.87 23.55 22.13 22.96 23.51 23.75 23.55 23.61 24.01 23.14 23.54 24.16 23.69 24.43 23.32 23.20 23.99 24.08 23.22 23.99 23.07 23.61 23.66 23.80 23.83 23.43 23.57 23.58 23.72 23.60 23.69 23.65 22.98 23.90 23.83 23.17 21.96 23.57 22.34 22.32 23.81 23.61 23.57 23.56 23.97 24.11 σg 0.13 0.10 0.04 0.19 0.16 0.21 0.32 0.08 0.12 0.20 0.28 0.15 0.22 0.23 0.15 0.20 0.16 0.14 0.07 0.15 0.17 0.09 0.29 0.27 0.14 0.25 0.39 0.53 0.11 0.17 0.24 0.09 0.09 0.13 0.20 0.11 0.14 0.08 0.10 0.34 0.21 0.21 0.07 0.11 0.24 0.12 – 51 – Table 7-Continued Object g L4q10 L4q11 L4q12PD L4q14 L4q15 L4q16 L4v01 L4v02 L4v03 L4v04 L4v05 L4v06 L4v08 L4v09 L4v10 L4v11 L4v12 L4v13 L4v14 L4v18 l4h03 l4h04 l4h17 l4h19 l4k05 l4k06 l4k08 l4o02 l4q01 l4q02 l4q04 l4q08 l4v07 l4v16 l4v17 l4v19 U4j04PD U4j09 U4n01 U4n02 u4h20 u4k07 u4q13 L5c02 L5c03 L5c06 23.54 23.92 24.11 23.59 24.04 23.71 24.18 23.80 22.83 24.13 24.11 23.66 23.95 23.52 23.87 24.16 24.00 22.72 23.25 22.95 23.46 23.89 23.59 23.91 23.74 24.16 23.94 23.79 23.81 23.77 23.72 23.30 23.39 24.30 24.19 23.99 24.10 24.19 24.20 24.13 24.14 24.61 24.31 23.59 23.96 24.27 σg 0.27 0.14 0.09 0.15 0.24 0.17 0.19 0.02 0.12 0.09 0.17 0.15 0.16 0.08 0.21 0.20 0.08 0.09 0.16 0.22 0.17 0.17 0.09 0.39 0.23 0.23 0.70 0.37 0.06 0.20 0.08 0.11 0.47 0.43 0.20 0.43 0.21 0.30 0.25 0.32 0.20 0.25 0.11 0.13 0.27 0.22 Ng 4 4 3 7 4 4 4 4 4 4 4 4 4 5 4 4 4 4 4 3 4 3 4 4 4 4 4 5 3 4 5 3 4 3 3 3 5 2 4 4 3 3 3 2 3 5 r · · · · · · · · · · · · · · · · · · 24.00 23.01 21.95 23.32 23.32 23.03 23.17 23.03 22.93 23.64 23.44 22.68 22.57 22.66 · · · · · · · · · · · · · · · · · · · · · 23.53 · · · · · · · · · · · · · · · · · · · · · · · · 23.25 23.29 · · · · · · · · · · · · · · · 22.65 23.27 · · · σr · · · · · · · · · · · · · · · · · · 1.07 0.27 0.09 0.22 0.12 0.11 0.27 0.08 0.20 0.77 0.20 0.70 0.17 0.15 · · · · · · · · · · · · · · · · · · · · · 0.38 · · · · · · · · · · · · · · · · · · · · · · · · 0.25 0.17 · · · · · · · · · · · · · · · 0.07 0.05 · · · Nr i σi Ni .. .. .. .. .. .. 2 3 3 3 2 3 2 4 3 4 3 5 5 4 .. .. .. .. .. .. .. 3 .. .. .. .. .. .. .. .. 4 6 .. .. .. .. .. 4 3 .. 22.44 22.81 · · · 22.89 · · · 20.17 · · · · · · · · · 23.45 23.04 22.51 23.16 22.83 22.85 23.58 23.60 22.32 · · · 22.57 · · · 23.48 · · · · · · 23.20 23.96 · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 22.98 22.83 22.80 21.50 23.13 · · · 22.85 22.42 23.01 22.68 0.03 0.16 · · · 0.34 · · · 0.00 · · · · · · · · · 0.23 0.07 0.12 0.02 0.09 0.05 0.17 0.13 0.02 · · · 0.27 · · · 0.21 · · · · · · 0.14 0.32 · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 0.12 0.23 0.00 1.11 0.13 · · · 0.19 0.06 0.15 0.73 3 3 .. 3 .. 1 .. .. .. 2 2 2 2 3 3 3 4 3 .. 2 .. 3 .. .. 4 2 .. .. .. .. .. .. .. .. .. .. 3 3 1 3 3 .. 3 3 3 3 – 52 – Table 7-Continued Ng r σr Nr i σi Ni 5 6 3 4 4 4 3 4 4 3 3 4 3 4 3 3 4 4 4 7 4 4 4 5 3 5 4 3 5 6 3 5 5 2 4 5 3 4 4 4 4 4 4 5 6 4 22.17 22.66 · · · 23.16 · · · · · · · · · 23.27 22.67 · · · 23.24 23.69 22.84 22.83 23.37 · · · · · · · · · 22.97 · · · · · · · · · · · · 22.20 22.50 21.64 · · · · · · · · · · · · · · · 23.37 23.11 · · · · · · · · · · · · 23.26 · · · 22.85 · · · 22.75 23.35 22.64 · · · 23.29 0.12 0.20 · · · 0.14 · · · · · · · · · 0.57 0.05 · · · 0.28 0.59 0.10 0.07 0.07 · · · · · · · · · 0.58 · · · · · · · · · · · · 0.18 0.10 0.18 · · · · · · · · · · · · · · · 0.24 0.25 · · · · · · · · · · · · 0.12 · · · 0.00 · · · 0.11 0.22 0.10 · · · 0.26 2 4 .. 3 .. .. .. 4 3 .. 4 4 4 3 4 .. .. .. 3 .. .. .. .. 3 3 3 .. .. .. .. .. 3 4 .. .. .. .. 4 .. 1 .. 3 4 4 .. 4 22.01 22.67 23.10 23.10 21.19 23.16 22.61 23.27 22.63 · · · 23.17 22.52 22.79 22.48 23.27 22.86 · · · · · · · · · · · · · · · 22.41 22.41 · · · 22.23 21.52 · · · · · · · · · · · · 23.46 · · · 23.06 · · · 22.92 22.83 22.74 23.27 21.92 22.82 22.92 22.55 23.26 22.67 · · · 22.94 0.08 0.07 0.11 0.18 0.02 0.12 0.07 0.15 0.11 · · · 0.08 0.97 0.14 0.03 0.16 0.90 · · · · · · · · · · · · · · · 0.20 0.08 · · · 0.05 0.15 · · · · · · · · · · · · 0.14 · · · 0.16 · · · 0.21 0.09 0.13 0.15 0.09 0.38 0.12 0.11 0.11 0.10 · · · 0.27 3 3 3 3 3 2 3 3 2 .. 3 3 3 3 4 4 .. .. .. .. .. 3 3 .. 4 2 .. .. .. .. 3 .. 3 .. 3 2 4 3 3 3 4 4 3 4 .. 2 Object g L5c07PD L5c08 L5c10PD L5c11 L5c12 L5c13PD L5c14 L5c15 L5c16 L5c18 L5c19PD L5c20PD L5c21PD L5c22 L5c23 L5c24PD L5i01 L5i02PD L5i03PD L5i04 L5i05 L5i06PD L5i08 L5j02 L5j03 L5j04 L5r01 L5s01PD l5c01 l5c04 U5c17PD U5j01PD U5j06 u5c09 u5i07 u5j05 L7a02 L7a03 L7a04PD L7a05 L7a06 L7a07 L7a10 L7a11PD l7a12 U7a01 22.94 23.59 24.02 23.66 22.28 23.80 23.41 24.16 23.08 23.84 23.76 24.04 23.75 23.59 24.19 23.84 23.66 23.85 23.77 23.07 23.81 23.12 23.20 23.29 23.16 22.51 23.65 20.84 24.05 24.21 24.18 23.83 23.67 24.35 24.29 23.50 23.50 23.80 23.29 23.68 23.78 23.43 23.63 23.34 23.86 24.13 σg 0.11 0.24 0.05 0.47 0.04 0.22 0.09 0.16 0.16 0.14 0.55 0.20 0.20 0.19 0.09 0.40 0.27 0.16 0.21 0.21 0.21 0.05 0.61 0.11 0.07 0.11 0.22 0.02 0.55 0.33 0.30 0.16 0.23 0.21 0.41 0.06 0.33 0.07 0.11 0.20 0.13 0.20 0.04 0.16 0.13 0.16 872 873 874 875 876 – 53 – C. Appendix C Comparing previously-published population estimates of the main classical belt, either as a whole or for the various components, with our present values must ensure that the same limiting Hg magnitude and the same region of the phase space are adopted. The main difference between P1 and the present L7 model is the higher q cut-off that was applied to the P1 sample. Restricting our current sample to the same region of phase space as was modeled in P1 gives very similar population estimates for the main belt. Secondly, P1's cold component was restricted in extend to a ≤ 45 AU and hence had a smaller population than in our current model, and conversely, the hot population was slightly overestimated compared to our current value, for the region of phase space. Lastly, because we use widely different H-magnitude slopes, the population estimates should be compatible for the detected Hg=7–8 range, but diverge for smaller TNOs (larger H). Scaling P1 to the Hg=8 limit, we find NP1(Hg ≤ 8.0) = 4400+1800−1100, while restricting our currennt model to the same phase space gives NAll(Hg ≤ 8.0) = 5800+1300−1200, in reasonable agreement. The latest independant population estimate of the main classical belt was done by Trujillo et al. (2001), who estimate 38,000+5400−2700 objects bigger than D = 100 km, with uncertainties being 3σ confidence. This number is more than 3 times smaller than our Hg=9.16 estimate. Can the two numbers be reconciled ? First, one must match the size ranges of the population being estimated. Trujillo et al. used a red albedo pR=0.04 and a solar red magnitude of -27.1. In this case a TNO of D = 100 km has an absolute magnitude HR = 8.8. Assuming the same g - R=0.8 color as we used in Kavelaars et al. (2009), this corresponds to Hg = 9.6. When looking at Fig. 9 of Trujillo et al. (2001) one clearly sees that either the assumption of an exponential luminosity function breaks at around mR ∼ 24, or the debiasing is incorrect faintward of that value. In particular, the lack of debiased objects fainter than 24 would push the population estimate down. For the main belt model used by Trujillo et al., mR = 24 corresponds to HR ∼ 7.7 or Hg ∼ 8.5. Hence their population estimate is probably more applicable to that limit but not to smaller sizes. With q = 4, or α = 0.6, the population estimates of Trujillo et al. (2001) is NTrujillo(Hg ≤ 8.5) = 8300+1200−600 .[3σ] Restricting our model to the same phase space and extrapolating our population estimate out to Hg ≤ 8.5, we obtain NAll(Hg ≤ 8.5) = 19, 000+4100−3700, [95%conf idence] a factor of 2 larger than Trujillo et al. (2001) However, this last number is an extrapolation beyond the limit to which CFEPS really measured the population. A more secure comparison from our point of view is done for Hg ≤ 8.0. The numbers become NTrujillo(Hg ≤ 8.0) = 4200+600−300, – 54 – and NAll(Hg ≤ 8.0) = 5500+1300−1100.[95%conf idence] Hence we marginally agree with Trujillo et al. (2001) at Hg ≤ 8.0. At Hg > 8.0, an extrapolation of our result (using our two H slopes determined at larger sizes) rapidly diverges from the Trujillo et al. (2001) estimate; if the slope does indeed drop near H ∼8 (or R (cid:39)24) to a shallower slope (Fuentes & Holman 2008) then the two estimates are less discrepant. Acknowledgements This research was supported by funding from the Natural Sciences and Engineering Research Council of Canada, the Canadian Foundation for Innovation, the National Research Council of Canada, and NASA Planetary Astronomy Program NNG04GI29G. This project could not have been a success without the dedi- cated staff of the Canada-France-Hawaii telescope as well as the assistance of the skilled telescope operators at KPNO and Mount Palomar. We dedicate this paper to the memory of Brian G. Marsden, for his devotion to orbital determination and passionate encouragement to observational planetary astronomers. Facilities: In addition to CFHT (MegaPrime) this work was made possible thanks to access to facilities listed in Table 2. 877 878 879 880 881 882 883 884 885 886 887 888 889 890 – 55 – REFERENCES Allen, R. L., Bernstein, G. M., & Malhotra, R. 2001, ApJ, 549, L241 -. 2002, AJ, 124, 2949 Allen, R. L., Gladman, B., Kavelaars, J. J., Petit, J.-M., Parker, J. W., & Nicholson, P. 2006, ApJ, 640, L83 Bernstein, G. M., Trilling, D. E., Allen, R. L., Brown, M. E., Holman, M., & Malhotra, R. 2004, AJ, 128, 1364 Bowell, E., Hapke, B., Domingue, D., Lumme, K., Peltoniemi, J., & Harris, A. W. 1989, in Asteroids II, ed. R. P. Binzel, T. Gehrels, & M. S. Matthews, 524–556 Brown, M. E. 2001, AJ, 121, 2804 -. 2008, in The Solar System Beyond Neptune, ed. Barucci, M. A., Boehnhardt, H., Cruikshank, D. P., & Morbidelli, A. , LPI (Tuscon: University of Arizona Press), 335–344 Brown, M. E., Barkume, K. M., Ragozzine, D., & Schaller, E. L. 2007, Nature, 446, 294 Brown, M. E., & Pan, M. 2004, AJ, 127, 2418 Chiang, E. I. 2002, ApJ, 573, L65 Chiang, E. I., Lovering, J. R., Millis, R. L., Buie, M. W., Wasserman, L. H., & Meech, K. J. 2003a, Earth Moon and Planets, 92, 49 Chiang, E. I., et al. 2003b, AJ, 126, 430 Duncan, M. J., & Levison, H. F. 1997, Science, 276, 1670 Duncan, M. J., Levison, H. F., & Budd, S. M. 1995, AJ, 110, 3073 Elliot, J. L., et al. 2005, AJ, 129, 1117 Fraser, W. C., Brown, M. E., & Schwamb, M. E. 2010, Icarus, 210, 944 Fraser, W. C., & Kavelaars, J. J. 2009, AJ, 137, 72 Fuentes, C. I., George, M. R., & Holman, M. J. 2009, ApJ, 696, 91 Fuentes, C. I., & Holman, M. J. 2008, AJ, 136, 83 Fukugita, M., Ichikawa, T., Gunn, J. E., Doi, M., Shimasaku, K., & Schneider, D. P. 1996, AJ, 111, 1748 Gladman, B., & Chan, C. 2006, ApJ, 643, L135 Gladman, B., Holman, M., Grav, T., Kavelaars, J., Nicholson, P., Aksnes, K., & Petit, J.-M. 2002, Icarus, 157, 269 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 – 56 – Gladman, B., Kavelaars, J. J., Petit, J.-M., Morbidelli, A., Holman, M. J., & Loredo, T. 2001, AJ, 122, 1051 Gladman, B., et al. 2006, Minor Planet Electronic Circulars, 2006-H29 -. 2009, ApJ, 697, L91 -. 2011, AJ, submitted, 1, 1 Gladman, B. J., Marsden, B. G., & van Laerhoven, C. 2008, in The Solar System Beyond Neptune, ed. A. Barucci, H. Boehnhardt, D. Cruikshank, & A. Morbidelli, LPI (Tucson: University of Arizona Press), 43–57 Gomes, R. S. 2003, Icarus, 161, 404 Gomes, R. S., Morbidelli, A., & Levison, H. F. 2004, Icarus, 170, 492 Gulbis, A. A. S., Elliot, J. L., Adams, E. R., Benecchi, S. D., Buie, M. W., Trilling, D. E., & Wasserman, L. H. 2010, AJ, 140, 350 Gulbis, A. A. S., Elliot, J. L., & Kane, J. F. 2006, Icarus, 183, 168 Hahn, J. M., & Malhotra, R. 2005, AJ, 130, 2392 Hainaut, O. R., & Delsanti, A. C. 2002, A&A, 389, 641 Ida, S., Larwood, J., & Burkert, A. 2000, ApJ, 528, 351 Information Technology Laboratory, N. 2011, NIST/SEMATECH e-Handbook of Statistical Methods, Tech. rep., National Institute of Standards and Technology, U.S. Commerce Department Ivezic, Z., et al. 2008, Serbian Astronomical Journal, 176, 1 Jewitt, D., Luu, J., & Chen, J. 1996, AJ, 112, 1225 Jewitt, D., Luu, J., & Trujillo, C. 1998, AJ, 115, 2125 Jewitt, D. C., & Luu, J. X. 1995, AJ, 109, 1867 Johnstone, D., Hollenbach, D., & Bally, J. 1998, ApJ, 499, 758 Jones, R. L., Parker, J. W., Bieryla, A., Marsden, B. G., Gladman, B., Kavelaars, J., & Petit, J. 2010, AJ, 139, 2249 Jones, R. L., et al. 2006, Icarus, 185, 508 Kavelaars, J., Allen, L., Gladman, B., Petit, J.-M., van Laerhoven, C., Nicholson, P., Margot, J.-L., & Marsden, B. G. 2006a, Minor Planet Electronic Circulars, 2006-H35 -. 2011, Minor Planet Electronic Circulars, 2011- 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 – 57 – Kavelaars, J., Jones, L., Gladman, B., Parker, J. W., & Petit, J.-M. 2008, in The Solar System Beyond Neptune, ed. M. A. Barucci, H. Boehnhardt, D. P. Cruikshank, & A. Morbidelli (Tucson: University of Arizona Press), 59–69 Kavelaars, J., et al. 2006b, Minor Planet Electronic Circulars, 2006-H30 Kavelaars, J. J., et al. 2009, AJ, 137, 4917 Larsen, J. A., et al. 2001, AJ, 121, 562 Leinhardt, Z. M., Marcus, R. A., & Stewart, S. T. 2010, ApJ, 714, 1789 Levison, H. F., Morbidelli, A., Vanlaerhoven, C., Gomes, R., & Tsiganis, K. 2008, Icarus, 196, 258 Lykawka, P. S., & Mukai, T. 2007, Icarus, 189, 213 Magnier, E. A., & Cuillandre, J.-C. 2004, PASP, 116, 449 Malhotra, R. 1993, Nature, 365, 819 Mann, R. K., & Williams, J. P. 2009, ApJ, 694, L36 -. 2010, ApJ, 725, 430 Millis, R. L., Buie, M. W., Wasserman, L. H., Elliot, J. L., Kern, S. D., & Wagner, R. M. 2002, AJ, 123, 2083 Morbidelli, A., Bottke, W., Nesvorn´y, D., & Levison, H. 2009a, Icarus, 204, 558 Morbidelli, A., Brasser, R., Tsiganis, K., Gomes, R., & Levison, H. F. 2009b, A&A, 507, 1041 Morbidelli, A., Emel'yanenko, V. V., & Levison, H. F. 2004, MNRAS, 355, 935 Morbidelli, A., & Levison, H. F. 2004, AJ, 128, 2564 Morbidelli, A., Levison, H. F., Tsiganis, K., & Gomes, R. 2005, Nature, 435, 462 Nagasawa, M., & Ida, S. 2000, AJ, 120, 3311 Parker, A. H., & Kavelaars, J. J. 2010, ApJ, 722, L204 Petit, J.-M., Holman, M., Scholl, H., Kavelaars, J., & Gladman, B. 2004, MNRAS, 347, 471 Petit, J.-M., Holman, M. J., Gladman, B. J., Kavelaars, J. J., Scholl, H., & Loredo, T. J. 2006, MNRAS, 365, 429 Petit, J.-M., Kavelaars, J. J., Gladman, B., & Loredo, T. 2008, in The Solar System Beyond Neptune, ed. M. A. Barucci, H. Boehnhardt, D. P. Cruikshank, & A. Morbidelli (Tucson: University of Arizona Press), 71–87 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 – 58 – Romanishin, W., Tegler, S. C., & Consolmagno, G. J. 2010, AJ, 140, 29 Schwamb, M. E., Brown, M. E., Rabinowitz, D. L., & Ragozzine, D. 2010, ApJ, 720, 1691 Stark, C. C., & Kuchner, M. J. 2010, in Bulletin of the American Astronomical Society, Vol. 42, Bulletin of the American Astronomical Society, 527–+ Stepinski, T. F., & Valageas, P. 1996, A&A, 309, 301 -. 1997, A&A, 319, 1007 Thommes, E. W., Duncan, M. J., & Levison, H. F. 1999, Nature, 402, 635 Throop, H. B., & Bally, J. 2005, ApJ, 623, L149 Trujillo, C. A., & Brown, M. E. 2001, ApJ, 554, L95 -. 2003, Earth Moon and Planets, 92, 99 Trujillo, C. A., Jewitt, D. C., & Luu, J. X. 2000, ApJ, 529, L103 -. 2001, AJ, 122, 457 Tsiganis, K., Gomes, R., Morbidelli, A., & Levison, H. F. 2005, Nature, 435, 459 Youdin, A. N. 2011, ArXiv e-prints 975 976 977 978 979 980 981 982 983 984 985 986 987 988 This preprint was prepared with the AAS LATEX macros v5.2. – 59 – Table 7-Continued Object g U7a08 U7a09 24.00 24.11 σg 0.18 0.21 Ng r σr Nr i σi Ni 4 3 23.62 23.45 0.14 0.16 4 4 23.25 23.41 0.32 0.28 3 3 Note. - 'L' objects are the tracked, characterized (i.e. with flux above the 40% detection efficiency level) objects of CFEPS. 'l' objects are the non-tracked, char- acterized objects of CFEPS. There is no ephemeris-based bias in those losses. Most of them were not recovered at checkup (either too faint or sheared out of field cover- age westward). 'U' objects are the tracked, non-characterized (i.e. with flux below 40% detection efficiency level) objects of CFEPS. 'u' objects are the non-tracked, non-characterized objects of CFEPS. Magnitudes listed for photometric observa- tions from CFHT. Some numbers are missing because the corresponding object was not re-observed in a particular filter from CFHT in photometric conditions. This is the case for lost objects, some of the PD objects which we did not try to track, or objects tracked solely at other facilities. g, r, i columns give the apparent magnitude of the object in the correspondant filter. σx is the uncertainty on the magnitude in filter x. Nx is the number of measurements in filter x used to derive the apparent magnitude and its uncertainty.
1504.04625	2	1504	2016-04-21T01:59:01	Kepler-539: a young extrasolar system with two giant planets on wide orbits and in gravitational interaction	[ "astro-ph.EP" ]	We confirm the planetary nature of Kepler-539b (aka Kepler object of interest K00372.01), a giant transiting exoplanet orbiting a solar-analogue G2 V star. The mass of Kepler-539b was accurately derived thanks to a series of precise radial velocity measurements obtained with the CAFE spectrograph mounted on the CAHA 2.2m telescope. A simultaneous fit of the radial-velocity data and Kepler photometry revealed that Kepler-539b is a dense Jupiter-like planet with a mass of Mp = 0.97 Mjup and a radius of Rp = 0.747 Rjup, making a complete circular revolution around its parent star in 125.6 days. The semi-major axis of the orbit is roughly 0.5 au, implying that the planet is at roughly 0.45 au from the habitable zone. By analysing the mid-transit times of the 12 transit events of Kepler-539b recorded by the Kepler spacecraft, we found a clear modulated transit time variation (TTV), which is attributable to the presence of a planet c in a wider orbit. The few timings available do not allow us to precisely estimate the properties of Kepler-539c and our analysis suggests that it has a mass between 1.2 and 3.6 Mjup, revolving on a very eccentric orbit (0.4<e<0.6) with a period larger than 1000 days. The high eccentricity of planet c is the probable cause of the TTV modulation of planet b. The analysis of the CAFE spectra revealed a relatively high photospheric lithium content, A(Li)=2.48 dex, which, together with both a gyrochronological and isochronal analysis, suggests that the parent star is relatively young.	astro-ph.EP	astro-ph	Astronomy & Astrophysics manuscript no. Kepler-539 May 3, 2018 c(cid:13) ESO 2018 6 1 0 2 r p A 1 2 . ] P E h p - o r t s a [ 2 v 5 2 6 4 0 . 4 0 5 1 : v i X r a Kepler-539: a young extrasolar system with two giant planets on wide orbits and in gravitational interaction L. Mancini1,2, J. Lillo-Box3,4, J. Southworth5, L. Borsato6, D. Gandolfi7,8, S. Ciceri1, D. Barrado3, R. Brahm9,10, and Th. Henning1 1 Max Planck Institute for Astronomy, Konigstuhl 17, 69117 – Heidelberg, Germany e-mail: [email protected] 2 INAF – Osservatorio Astrofisico di Torino, via Osservatorio 20, 10025 – Pino Torinese, Italy 3 Depto. de Astrof´ısica, Centro de Astrobiolog`ıa (CSIC-INTA), ESAC campus 28691 – Villanueva de la Canada, Spain 4 European Southern Observatory, Alonso de Cordova 3107, Vitacura, Casilla 19001, Santiago de Chile, Chile 5 Astrophysics Group, Keele University, Keele ST5 5BG, UK 6 Dip. di Fisica e Astronomia "Galileo Galilei", Universit`a di Padova, Vicolo dell'Osservatorio 2, 35122 – Padova, Italy 7 Dip. di Fisica, Universit`a di Torino, via P. Giuria 1, 10125 – Torino, Italy 8 Landessternwarte Konigstuhl, Zentrum fur Astronomie der Universitat Heidelberg, Konigstuhl 12, 69117 – Heidelberg, Germany 9 Instituto de Astrof´ısica, Pontificia Universidad Cat´olica de Chile, Av. Vicuna Mackenna 4860, 7820436 – Macul, Santiago, Chile 10 Millennium Institute of Astrophysics, Av. Vicuna Mackenna 4860, 7820436 – Macul, Santiago, Chile Preprint online version: May 3, 2018 ABSTRACT We confirm the planetary nature of Kepler-539 b (aka Kepler object of interest K00372.01), a giant transiting exoplanet orbiting a solar-analogue G2 V star. The mass of Kepler-539 b was accurately derived thanks to a series of precise radial velocity measurements obtained with the CAFE spectrograph mounted on the CAHA 2.2-m telescope. A simultaneous fit of the radial-velocity data and Kepler photometry revealed that Kepler-539 b is a dense Jupiter-like planet with a mass of Mp = 0.97 ± 0.29 MJup and a radius of Rp = 0.747 ± 0.018 RJup, making a complete circular revolution around its parent star in 125.6 days. The semi-major axis of the orbit is roughly 0.5 au, implying that the planet is at ≈ 0.45 au from the habitable zone. By analysing the mid-transit times of the 12 transit events of Kepler-539 b recorded by the Kepler spacecraft, we found a clear modulated transit time variation (TTV), which is attributable to the presence of a planet c in a wider orbit. The few timings available do not allow us to precisely estimate the properties of Kepler-539 c and our analysis suggests that it has a mass between 1.2 and 3.6 MJup, revolving on a very eccentric orbit (0.4 < e ≤ 0.6) with a period larger than 1000 days. The high eccentricity of planet c is the probable cause of the TTV modulation of planet b. The analysis of the CAFE spectra revealed a relatively high photospheric lithium content, A(Li) = 2.48 ± 0.12 dex, which, together with both a gyrochronological and isochronal analysis, suggests that the parent star is relatively young. Key words. stars: planetary systems – stars: fundamental parameters – stars: individual: Kepler-539 1. Introduction The transiting extrasolar planet (TEP) population turns out to be the best atout in the hands of exoplanetary scien- tists. Thanks to the early ground-based systematic surveys (e.g., Bakos et al. 2004; Alonso et al. 2004; McCullough et al. 2005; Pollacco et al. 2006; Pepper et al. 2007; Alsubai et al. 2013; Bakos et al. 2013) and then to those from the space (CoRoT: Barge et al. 2008, Kepler : Borucki et al. 2011), more than 1290 transiting planetary systems have now been found. The possibility to easily derive most of their physical and orbital parameters, and to investigate even the composition of their atmosphere, make the TEPs the most suitable targets to detect and study in detail. They currently represent the best statistical sample to constrain the theoretical models of planetary formation and evolu- tion. Moreover, transit time variation (TTV) studies al- lowed the detection of additional non-transiting bodies in many TEP systems, highlighting the great efficacy of high- precision and high-cadence photometry in the search for extrasolar planets. The majority of TEPs have been discovered by the Kepler space telescope, which has revealed how varied they are, in terms of mass and size, and how diverse their ar- chitectures can be, confirming science-fiction pictures in several cases or going beyond human imagination in oth- ers. The numerous TEP discoveries achieved by Kepler are based on time-series photometric monitoring of the bright- ness of over 145 000 main sequence stars, in which the periodic-dimming signal caused by a TEP can be found. However, such a signal can be mimic by other astrophysical objects in particular configurations. Radial velocity (RV) follow-up observations of the 'potential' parent stars are therefore a fundamental step for distinguishing real TEPs from false positive cases. Here we focus our attention on the Kepler system Kepler-539. Thanks to precise RV measurements, presented 1 L. Mancini et al.: The planetary system Kepler-539 in Sect. 2, we confirm the planetary nature of Kepler-539 b (aka KOI-372 b, K00372.01, KIC 6471021), a dense gas- giant planet moving on a wide orbit around a young and active G2 V star (V = 12.6 mag). The analysis of the phys- ical parameters of Kepler-539 b and its parent star are de- scribed in Sect. 3. Moreover, the clear variation observed in the mid-times transits of Kepler-539 b strongly support the existence of an additional massive planet, Kepler-539 c, in the system, moving on a larger and very eccentric orbit, as discussed in Sect. 4. We summarise our results in Sect. 5. 2. Observations and data analysis 2.1. Kepler photometry The Kepler spacecraft monitored Kepler-539 from quar- ters 0 to 17 (i.e. four years; from May 2009 to May 2013). It was labelled as a Kepler object of interest (KOI) due to a ∼ 0.2% dimming in its light curve with a period of ∼ 125 days (Borucki et al. 2011). This periodic dim- ming is actually caused by the transit of a Jupiter-like planet candidate, Kepler-539 b, moving on a quite wide or- bit around the star. 12 transits of Kepler-539 b are present in the Kepler long cadence (LC) light curves. We have labelled them from cycle -5 to cycle 6 (see Fig. 1). Two of the transits are incomplete (cycles -2 and -1); two are most likely contaminated by starspot crossing anomalies (cycles -5 and 2); three were also covered in short cadence (SC) (cycles 4, 5 and 6; see Fig. 2). The complete Kepler light curve is shown in Fig. 3, which highlights a signifi- cant stellar variability (0.0470± 0.0002 mag peak-to-peak). McQuillan et al. (2013) found a periodic photometric mod- ulation in the light curve and, by assuming that it is in- duced by a star-spot activity, estimated a stellar rotation period of 11.769 ± 0.016 days. This value is in good agree- ment with that found by Walkowicz & Basri (2013a,b), i.e. 11.90 ± 3.45 days. 2.2. Radial velocity follow-up observations We monitored Kepler-539 between July 2012 and July 2015 with the Calar Alto Fibre-fed Echelle spectrograph (CAFE; Aceituno et al. 2013) mounted on the 2.2 m tele- scope at the Calar Alto Observatory (Almer´ıa, Spain) as part of our follow-up programme of Kepler candidates. This programme has already confirmed the planetary na- ture of Kepler-91 b (Lillo-Box et al. 2014a,b), Kepler-432 b (Ciceri et al. 2015), Kepler-447 b (Lillo-Box et al. 2015b), and has identified and characterised some false positives in the sample of Kepler planet candidates (Lillo-Box et al. 2015a). CAFE has a spectral coverage from 4000 A to 9500 A, divided into 84 orders with a mean resolving power of R = 63 000. We acquired 28 spectra, which were re- duced using the dedicated pipeline provided by the ob- servatory (Aceituno et al. 2013). We used thorium-argon (ThAr) exposures obtained after each science spectrum to wavelength-calibrate the corresponding data. The final spectra have signal-to-noise ratios in the range S/N = 7– 16. A radial velocity (RV) was obtained from each spectrum by using the cross-correlation technique through a weighted binary mask (Baranne et al. 1996). The mask is composed of more than 2000 sharp and isolated spectral lines in the CAFE wavelength range. The cross-correlation was per- formed in a ±30 km s−1 range around the expected RV of 2 the star. The peak of the cross-correlation function (CCF) was measured by fitting a four-term Gaussian profile. This velocity was then corrected for the barycentric Earth radial velocity at mid-exposure time. Since we took several con- secutive spectra on the same nights, we decided to combine the RV values of the corresponding pairs in the cases where their individual signal-to-noise was low (i.e., S/N < 10) and mutual discrepancies were larger than 50 m s−1. This procedure can also diminish the effect of short term vari- ability, such as the granulation noise, on the radial velocity (although here the expected amplitudes are of the order of few tens of m s−1). Some of the spectra were neglected due to low quality, mainly caused by weather conditions or problems with the stability of the ThAr lamp. We ob- tained 20 RV values in total which are reported, together with their observing times, in Table 1 and are compatible with the presence of a ∼ 1 MJup planet in the system (see Fig. 7 and Sect. 4). Table 1: RV and BVS measurements of Kepler-539 from CAHA/CAFE. Date of observation BJD(TDB)-2450000 6116.5291378 6124.3823828 6138.4327769 6523.4275256 6598.2879800 6804.5209537 6811.4391381 6821.6225006 6834.4194255 6859.5818459 7136.6259467 7152.5198831 7160.5096750 7169.5836757 7193.4502198 7254.4473639 7258.3855966 7263.4322174 7264.4220946 7266.3737337 RV (km s−1) 10.408 10.383 10.387 10.402 10.448 10.439 10.475 10.432 10.437 10.409 10.285 10.431 10.450 10.500 10.425 10.361 10.400 10.402 10.413 10.449 errRV (km s−1) 0.026 0.029 0.018 0.024 0.033 0.029 0.035 0.049 0.025 0.038 0.039 0.042 0.021 0.032 0.045 0.035 0.045 0.028 0.025 0.034 FWHM (km s−1) -0.041 -0.065 -0.117 -0.100 0.081 -0.146 0.159 -0.118 0.086 -0.391 -0.266 0.087 -0.267 0.038 -0.088 -0.021 -0.106 0.014 0.018 0.169 BVS (km s−1) 9.696 9.517 9.656 9.572 9.518 9.250 9.407 9.198 9.342 10.036 9.541 9.422 9.312 9.153 9.556 9.727 9.559 9.519 9.599 9.995 2.3. Spectral analysis and the age of Kepler-539 We derived the spectroscopic parameters of the host star Kepler-539 from the co-added CAFE spectra, which has a S/N ratio of about 40 per pixel at 5500 A. Following the procedures described in Gandolfi et al. (2013, 2015), we used a customised IDL1 software suite to fit the composite CAFE spectrum to a grid of synthetic theoretical spectra. The latter were calculated with the stellar spectral synthe- sis program SPECTRUM (Gray & Corbally 1994) using ATLAS9 plane-parallel model atmospheres (Kurucz 1979), under the assumptions of local thermodynamic equilibrium (LTE) and solar atomic abundances as given in Grevesse & Sauval (1998). We fitted spectral features that are sensitive to different photospheric parameters. Briefly, we used the wings of the Balmer lines to estimate the effective temper- ature Teff of the star, and the Mg i 5167, 5173, and 5184 A, 1 The acronym IDL stands for Interactive Data Language and is a trademark of ITT Visual Information Solutions. L. Mancini et al.: The planetary system Kepler-539 Fig. 1: The twelve transit events of Kepler-539 b observed by Kepler in long-cadence mode. The transits at epoch -5 and 2 are clearly affected by star-spot-crossing events. Times are in BKJD (Barycentric Kepler Julian Date – equivalent to BJD(TDB) minus 2454833.0). Fig. 2: The three transit events of Kepler-539 b observed by Kepler in short-cadence mode. Times are in BKJD (Barycentric Kepler Julian Date – equivalent to BJD(TDB) minus 2454833.0). the Ca i 6162 and 6439 A, and the Na i D lines to deter- mine the surface gravity log g(cid:63). The iron abundance [Fe/H] and microturbulent velocity vmicro was derived by applying the method described in Blackwell & Shallis (1979) on iso- lated Fe i and Fe ii lines. To determine the macroturbulent velocity vmacro, we adopted the calibration equations for solar like stars from Doyle et al. (2014). The projected ro- tational velocity v sin i(cid:63) was measured by fitting the profile of several clean and unblended metal lines2. We found that Kepler-539 has an effective temperature of Teff = 5820 ± 80 K, log g(cid:63) = 4.4 ± 0.1 (cgs), [Fe/H]= −0.01 ± 0.07 dex, vmicro = 1.1 ± 0.1 km s−1, vmacro = 3.2 ± 0.6 km s−1, and 2 Here i(cid:63) refers to the inclination of the stellar rotation axis with respect to the line of sight. 3 127.0127.5128.0128.5-0.04-0.03-0.02-0.010.000.01NormalisedFlux-5252.5253.0253.5254.0-4378.0378.5379.0379.5380.0-3504.0504.5505.0505.5-2629.5630.0630.5631.0-1755.0755.5756.0756.5757.00880.5881.0881.5882.0882.5-0.04-0.03-0.02-0.010.000.01BKJDNormalisedFlux11006.51007.01007.51008.0BKJD21132.01132.51133.01133.5BKJD31257.51258.01258.51259.0BKJD41383.01383.51384.01384.51385.0BKJD51509.01509.51510.01510.5BKJD6L. Mancini et al.: The planetary system Kepler-539 Fig. 3: The entire Kepler light-curve data of Kepler-539. The large stellar variability can be reasonable interpreted as induced by a star-spot activity. Times are in BKJD (Barycentric Kepler Julian Date – equivalent to BJD(TDB) minus 2454833.0). The red arrows mark the mid-times of the twelve transits of Kepler-539 b. v sin i(cid:63) = 4.4 ± 0.5 km s−1(Table 3). According to the Straizys & Kuriliene (1981) calibration scale for dwarf stars, the effective temperature of Kepler-539 translates to a G2 V spectral type. Fig. 4: CAFE co-added spectrum of Kepler-539 (black line) encompassing the Li i 6707.8 A absorption doublet. The best fitting ATLAS9 spectrum is overplotted with a thick red line. The vertical dashed line marks the position of the Li doublet. The lowest part of the plot displays the residuals to the fit. The CAFE co-added spectrum of Kepler-539 reveals the presence of a moderate Li i 6707.8 A absorption dou- blet (Fig. 4). We estimated the photospheric lithium abun- dance of the star by fitting the Li doublet using ATLAS9 LTE model atmospheres. We fixed the stellar parameters to the values given in Table 3 and allowed our code to fit the lithium content. Adopting a correction for non-LTE 4 effects of +0.006 dex (Lind et al. 2009), we measured a lithium abundance of A(Li) = log (n(Li)/n(H)) + 12 = 2.48 ± 0.12 dex. The photospheric lithium content and the moderate ro- tation period of Kepler-539 suggest that the star is rela- tively young. Fig. 5 shows the lithium abundance of Kepler- 539 compared to the A(Li) of the dwarf stars of the Pleiades, Hyades, and NGC 752 open clusters, as listed in Soderblom et al. (1993), Pace et al. (2012), and Sestito et al. (2004), respectively. Kepler-539 has to be older than ∼0.1 Gyr, as the star falls below the envelope of the Pleiades. Although lithium depletion becomes ineffective beyond an age of 1–2 Gyr (Sestito & Randich 2005), Kepler- 539 lies between the envelopes of the other two clusters, sug- gesting an age intermediate between the age of the Hyades (∼0.6 Gyr) and the age of NGC 752 (∼2.0 Gyr). This is further confirmed by the fact that the lithium content of Kepler-539 is intermediate between the average lithium abundance measured in early G-type stars of 0.6-Gyr-old open clusters (A(Li) = 2.58 ± 0.15 dex) and that of 2- Gyr-old open clusters (A(Li) = 2.33 ± 0.17 dex; Sestito & Randich 2005). We used Eq. (32) from Barnes (2010) and the rotation period of Kepler-539 to infer its gyrochronological age, as- suming a convective turnover time-scale of τc = 34 days (Barnes & Kim 2010) and a zero-age main sequence rota- tion period of P0 = 1.1 days (Barnes 2010). We found a gyrochronological age of 1.0 ± 0.3 Gyr, which supports the relatively young scenario. Our estimation is in good agree- ment with the 1.15 Gyr gyrochronological age predicted by Walkowicz & Basri (2013a,b) and with that estimated using theoretical models (see Sect. 3 and Table 3). By having short-cadence data over a quite large times- pan, it is in principle possible to perform an asteroseismic analysis and try to precisely estimate the age of a star. However, in the case of Kepler-539, the SC data are only available for less than half the quarters. In particular, the 2004006008001000120014001600-0.04-0.03-0.02-0.010.000.010.02BKJDNormalisedFluxL. Mancini et al.: The planetary system Kepler-539 continuous time baseline of SC data is not more than 125 d in the best case (BJD 2456015 − 2456139). Furthermore the activity variation is predominant in the data. We anal- ysed such SC data by filtering the variability due to activ- ity with moving averages, but we did not find evidence of a clear frequency comb in the power spectrum and there- fore we did not get particular clues about patterns due to solar-like oscillations. We concluded that no reliable aster- oseismic analysis could be performed for Kepler-539 based on the Kepler data. Table 2: Nearby visual companions around Kepler-539 (from Adams et al. 2012) and their dilution effect on the depth of the transit events. Companion Distance ((cid:48)(cid:48)) Ks (mag) ∆Ks Dilution B C D Ea 2.49 3.56 4.99 5.94 23.2 22.4 22.7 17.1 8.6 8.0 8.2 4.0 0.005 % 0.01 % 0.008 % 1.31 % Notes. aAlso known as KIC 6471028. bution, see Figueira et al. 2016). This value indicates only a weak correlation between the two parameters, which is only significant at a 2σ level for the median value and still at less than 3σ for the upper boundary. Consequently, we assume no correlation between both parameters, although we warn about the weak probability of correlation. 3. Physical properties of the system Fig. 5: Lithium abundances of F- and G-type dwarf stars in the Pleiades (black dots; Soderblom et al. 1993), Hyades (green squares; Pace et al. 2012), and NGC 752 (blue trian- gles; Sestito et al. 2004) open clusters. The red dot marks the position of Kepler-539. 2.4. Excluding false-positive scenarios A faint pulsating variable star or a binary system in the background/foreground can mimic a planetary transit sig- nal on the target star. High-resolution images are extremely useful to exclude this possibility (e.g., Lillo-Box et al. 2014c). A Ks-band, high-resolution, adaptive optics im- age of Kepler-539 was obtained with ARIES on the 6.5 m MMT telescope by Adams et al. (2012), who found four stars within 6(cid:48)(cid:48) of the target, see Table 2. Following Lillo- Box et al. (2012), we have estimated the dilution effect caused by each companion, finding that it is very small for the three closest targets, roughly 0.023% in total and thus negligible. Concerning the very bright companion "E", this is another object targeted as KIC 6471028. As it lies outside the Kepler aperture, this star does not contami- nate Kepler-539. KIC 6471028 was also detected with the AstraLux North instrument mounted on the CAHA 2.2 m telescope by Lillo-Box et al. (2012), who estimated that this is a K2-K4 background dwarf. An intense stellar activity could mimic the presence of a planetary body in the RV signal, thus causing a false pos- itive case. We have also analysed such a possibility by de- termining the bisector velocity span (BVS) from the CAFE spectra (the values are reported in Table 1). The bisector analysis provides a Pearson correlation coefficient between the RV and the BVS of 0.28+0.31−0.42 (median value and 95% confidence intervals from the posterior probability distri- Fig. 6: Phased Kepler long-cadence (top light curve) and short-cadence (bottom light curve) data zoomed around transit phase. The TTVs (see Sect. 4) were removed from the data before plotting. The jktebop best fits are shown using solid lines. The residuals of the fits are plotted at the base of the figure. For the determination of the physical parameters of the Kepler-539 system, we proceeded as in Ciceri et al. (2015). We first selected for analysis all data within two transit durations of a transit, ignoring the two with only partial coverage in the Kepler light curve, and converted from flux to magnitude units. Each transit was then detrended by fitting a polynomial versus time. High polynomial orders of 3 to 5 were required to account for the brightness variations of the host star caused by spot activity. We then fitted the photometry and the CAFE RV measurements simultaneously using the jktebop code (Southworth 2013), after modifying it to allow fitting for individual times of mid-transit. The Kepler LC and SC 5 L. Mancini et al.: The planetary system Kepler-539 Table 3: Final parameters of the planetary system KOI-0372. Parameter Stellar parameters R.A. (J2000) . . . . . . . . . . . . . . . . . . . . . Dec. (J2000) . . . . . . . . . . . . . . . . . . . . . . Spectral typeb . . . . . . . . . . . . . . . . . . . . Kepler magnitude . . . . . . . . . . . . . . . . Effective temperature . . . . . . . . . . . . . Iron abundance . . . . . . . . . . . . . . . . . . . Lithium abundance . . . . . . . . . . . . . . . Microturb. velocity . . . . . . . . . . . . . . . Macroturb. velocityc . . . . . . . . . . . . . . Proj. rotat. velocity . . . . . . . . . . . . . . Stellar rotation periodd . . . . . . . . . . . Stellar age (from gyrochronology) . Stellar age (from isochrones) . . . . . . Stellar mass . . . . . . . . . . . . . . . . . . . . . . Stellar radius . . . . . . . . . . . . . . . . . . . . . Stellar mean density . . . . . . . . . . . . . . Stellar surface gravity . . . . . . . . . . . . Planetary parameters (Kepler-539 b) Planetary mass . . . . . . . . . . . . . . . . . . . Planetary radius . . . . . . . . . . . . . . . . . . Planetary mean density . . . . . . . . . . . Planetary surface gravity . . . . . . . . . Planetary equilibrium temperature Safronov number . . . . . . . . . . . . . . . . . Orbital parameters Time of mid-transit . . . . . . . . . . . . . . . Orbital period . . . . . . . . . . . . . . . . . . . . Semi-major axis . . . . . . . . . . . . . . . . . . Orbital inclination . . . . . . . . . . . . . . . . Fractional star radius . . . . . . . . . . . . . Fractional planet radius . . . . . . . . . . RV semi-amplitude . . . . . . . . . . . . . . . Barycentric RV . . . . . . . . . . . . . . . . . . . Eccentricity . . . . . . . . . . . . . . . . . . . . . . Nomen. Unit Valuea . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19h56m29:39s 41◦52(cid:48)00.3(cid:48)(cid:48) . . . . . . . . . . . . . . . . . . . . . . Kp . . . . . . . . . . . . Teff . . . . . . . . . . [Fe/H] . . . . . . . . . A(Li) . . . . . . . . . . vmicro . . . . . . . . . vmacro . . . . . . . . v sin i(cid:63) . . . . . . . . . .Prot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . M(cid:63) . . . . . . . . . . . . R(cid:63) . . . . . . . . . . . . ρ(cid:63) . . . . . . . . . . . log g(cid:63) . . . . . . . . . . . . . . . . . . . . . . . . . mag . . . . . . . . . . . K . . . . . . . . . . . dex . . . . . . . . . . dex . . . . . . . . km s−1 . . . . . . km s−1 . . . . . . km s−1 . . . . . . . . day . . . . . . . . . . Gyr . . . . . . . . . . Gyr . . . . . . . . . . M(cid:12) . . . . . . . . . . R(cid:12) . . . . . . . . . . ρ(cid:12) . . . . . . . . . . cgs . . . . . . . . . . . Mb . . . . . . . . . . . . Rb . . . . . . . . . . . . ρb . . . . . . . . . . . . gb . . . . . . . . . . . . Teq . . . . . . . . . . . . . Θ . . . . . . . . . . . MJup . . . . . . . . RJup . . . . . . . . . ρJup . . . . . . . . . m s−2 . . . . . . . . . . K . . . . . . . . . . . . . . . . . . . . . . . . . . . T0 . . . . . . . . . . . Porb . . . . . . . . . . . . a . . . . . . . . . . . . . . i . . . . . . . . . . . . . rA . . . . . . . . . . . . rb . . . . . . . . . . . . KA . . . . . . . . . . . . . γ . . . . . . . . . . . . . . e . . . . . . . . . BJDTDB . . . . . . days . . . . . . . . . au . . . . . . . . .degree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . m s−1 . . . . . . . km s−1 . . . . . . . . . . . . . . . . . . G2 V 12.39 5820 ± 80 −0.01 ± 0.07 2.5 ± 0.1 1.1 ± 0.1 3.2 ± 0.6 4.4 ± 0.5 1.0 ± 0.3 1.1+0.6 +0.3 −0.0 −0.1 11.769 ± 0.016 1.048 ± 0.034 ± 0.025 0.952 ± 0.017 ± 0.007 4.502 ± 0.014 ± 0.003 1.216 ± 0.054 0.97 ± 0.29 ± 0.02 0.747 ± 0.016 ± 0.006 2.18 ± 0.66 ± 0.02 43 ± 12 387.6 ± 6.0 1.24 ± 0.37 ± 0.01 2 455 588.8710 ± 0.0030 125.63243 ± 0.00071 0.4988 ± 0.0054 ± 0.0039 89.845 ± 0.086 0.01057 ± 0.00078 0.000854 ± 0.000085 132.3 ± 6.3 9.959 ± 0.007 < 0.39e Notes. Where there are two error bars, the first is a statistical error, coming from the measured spectroscopic and photometric parameters, while the second is a systematic error and is given only for those parameters which have a dependence on theoretical stellar models. a The adopted parameters assume a circular orbit. b With an accuracy of ± 1 sub-class. c Using the calibration equations of Doyle et al. (2014). d From McQuillan et al. (2013). e The 99% confidence upper limit on the eccentricity from a model in which the eccentricity is allowed to vary in the fit. data were fitted separately, accounting for the long effec- tive exposure times in the LC data by oversampling the fitted model by a factor of five. The following parameters were fitted: the fractional radii of the two objects (r(cid:63) = R(cid:63)/a and rp = Rp/a, where a is the orbital semi-major axis), the orbital inclination i, the time of midpoint of each transit, the velocity ampli- tude K(cid:63), the systemic velocity Vγ, and the coefficients of the polynomial for each transit. The orbital period, Porb, and reference transit midpoint, T0, were fixed to the values measured from the transit times; these were used only for phasing the RV measurements. As in the other papers of our series (e.g. Mancini et al. 2013a,b), a quadratic limb darkening law was used, with the linear term fitted and the quadratic term fixed to 0.27 (Sing 2010). We rescaled the error bars of the LC, SC and RV data to give a reduced χ2 of χ2 ν = 1.0 for each versus the fitted model. Based on our experience with similar studies (e.g. Mancini et al. 2014a,b), this procedure is necessary for taking into account underes- timated data errors and additional sources of white noise; for example an RV jitter due to the activity of the host star. The best fits to the photometry and the RVs are shown in Figs. 6 and 7. The uncertainties in the fitted parameters were derived by running both Monte Carlo and residual- permutation simulations (Southworth 2008) and choosing the larger of the two error bars for each parameter. The results obtained using the SC data are more precise than 6 L. Mancini et al.: The planetary system Kepler-539 Table 4: Kepler times of transit midpoint of Kepler-539 b and their residuals. Time of minimum BJD(TDB)−2400000 54960.74947 ± 0.00109 55086.35000 ± 0.00040 55212.96694 ± 0.00042 55337.57977 ± 0.00043 55463.23346 ± 0.00118 55588.87808 ± 0.00041 55714.51794 ± 0.00043 55840.13340 ± 0.00096 55965.77212 ± 0.00043 56091.39778 ± 0.00030 56217.00030 ± 0.00030 56342.66701 ± 0.00030 Cycle no. -5 -4 -3 -2 -1 0 1 2 3 4 5 6 Residual (JD) 0.040612 0.008712 -0.006781 -0.026380 -0.005126 0.007068 0.014495 -0.002478 0.003803 -0.002963 -0.001233 0.001403 Table 5: The next mid-transit times of Kepler-539 b. Cycle BJD(TDB) Date Time (UT) 16 17 18 19 20 2457598.9899 2457724.6223 2457850.2547 2457975.8872 2458101.5196 2016 Jul 2016 Dec 2017 Apr 2017 Aug 2017 Dec 29 02 06 10 14 11:40 02:56 18:07 09:12 00:29 of mid-transit, T0 = 2 455 588.8710± 0.0030 BJD (TDB). A plot of the residuals around the fit (see Fig. 8) shows a clear and modulated deviation from the predicted transit times, meaning that the orbital period of Kepler-539 b is variable (the maximum TTV is 58.5± 1.6 min and was measured for the transit at cycle −5). The uncertainties of Porb and T0 have been increased to account for this, but we stress that these values should be used with caution as they are based on only 12 timings, of which two are incomplete and oth- ers are affected by starspot anomalies that can introduce offsets in the transit timing measurements – several studies have quantified this effect (e.g. Barros et al. 2013; Oshagh et al. 2013; Mazeh et al. 2015; Ioannidis et al. 2016). We have reported the predicted mid-transit times in Table 5 for the next year and a half. In order to identify the origin of this modulated TTV, we used the TRADES (TRAnsits and Dynamics of Exoplanetary Systems; Borsato et al. 2014) code, which al- lows modelling of the dynamics of multiple-planet systems by reproducing the observed mid-transit times and RVs, with the possibility to choose among four different algo- rithms. We thus investigated the possibility that Kepler- 539 is a planetary system formed by a star and two plan- ets, b and c. We ran two series of simulations, the first by selecting the PIKAIA algorithm (Charbonneau 1995), the second with the Particle Swarm Optimization (PSO) algo- rithm (Tada 2007). In all the cases, we fixed tight bound- aries for the parameters of planet b (centred on the values in Table 3) and assumed the zero eccentricity case, which means that we fixed the argument of pericentre (ωb) to 90◦, and let only the mean anomaly (νb) of planet b to freely float between 0 and 360 degrees. We use a time of refer- ence (or epoch, tepoch = 2455463.244) that is very close to the transit cycle 0, forcing the planet b to be initially at the pericentre. Because of this, all the solutions of both the algorithms returned zero values for νb. Furthermore, 7 Fig. 7: Upper panel: phased RVs for Kepler-539 and the best fit from jktebop. Lower panel: residuals of RVs versus best fit. those from the LC data, so were adopted as the final set of photometric parameters. We find that the observations are fully consistent with Kepler-539 b moving on a circular orbit. However, we per- formed the fit both fixing the orbital eccentricity e to zero and fitting for it via the combination terms e cos ω and e sin ω, where ω is the argument of periastron. In the latter case, we obtained e = 0.18+0.21−0.18. We then used the Bayesian information criterion (BIC) to evaluate the preferred sce- nario, finding that the eccentric case is not favoured over the zero eccentricity model. Finally, we estimated the age and the physical prop- erties of the system from the SC results and the Teff and [Fe/H] measured from our spectra. Constraints from the- oretical stellar models were used to make the solution de- terminate. We followed the HSTEP approach (Southworth 2012, and references therein), which yielded measurements of the properties of the star and planet accompanied by both statistical and systematic errorbars. The final results are reported in Table 3. 4. Planet Kepler-539 c from transit time variation We have analysed the 12 mid-transit times of Kepler-539 b to characterise the ephemeris of the transit and check if there is a possible transit time variation (TTV), which could be a sign of the presence of additional bodies in this system. Since the transits at cycles -5 and 2 are quite af- fected by starspots, we used the prism3 and gemc4 codes (Tregloan-Reed et al. 2013, 2015) that allowed us to fit both the full transit events and the shorter starspot-occultation events contemporaneously. The parameters of the starspots coming from the fits are reported in Appendix A. The mid- transit times for the other epochs were estimated using jk- tebop. The resulting timings are tabulated in Table 4 and were fitted with a straight line to obtain the orbital period, Porb = 125.63243 ± 0.00071 days, and the reference time 3 Planetary Retrospective Integrated Star-spot Model. 4 Genetic Evolution Markov Chain. L. Mancini et al.: The planetary system Kepler-539 Fig. 8: O–C diagram for the timings of Kepler-539 b at mid-transit versus a linear ephemeris. The timings in blue refer to those coming from the Kepler LC data, while those marked with green are from SC data. The red points refer to the two LC transits affected by star spots. The two incomplete LC transits are marked with orange points. we assumed for simplicity the two planets to be coplanar. We investigated different configurations of two planet sys- tem, with the period of planet c, Pc, varying from 150 to 2500 days. We fitted the pair (ec cos ωc, ec sin ωc) instead of (ec, ωc) to avoid correlation between the parameters. We set 0.9 as the highest value allowed for the eccentricity, while we let ωc and νc freely vary as for νb. The solutions obtained from the TRADES fitting processes were further analysed using the Frequency Map Analysis (FMA; Laskar et al. 1992) to check if they are stable. We found that the best solutions are those related to configurations with Pc > 1000 days and a mass of the planet c between 1.2 and 3.6 MJup on a very eccentric or- bit (0.4 < ec ≤ 0.6). The final parameters, with confidence intervals estimated with a bootstrapping method, are re- ported in Table 6 for the five best solutions, based on the χ2, and displayed in figures shown in Appendix B. The es- timated confidence intervals are very tight to the fitted values because the algorithms found solutions that corre- spond to minimum surrounded by very high peaks in the χ2-space. We also exhaustively investigated narrow region (200 − 800 days) of Pc (in steps of about 300 days), but the solutions that we obtained are not favoured because the corresponding χ2 is much higher than those reported in Table 6. Knowing the orbital period of Kepler-539 c, we can eas- ily estimate the semi-major axis and the equilibrium tem- perature of the planet c for each solutions. These values are shown in right-hand columns of Table 6. Moreover, we can use Eq. (7.29) from Haswell (2010) for estimating the maximum TTV signal expected for Kepler-539 b, i.e. TTVmax = (MA + Mb)ec(ab/ac)3Pc. (1) The last column of Table 6 reports the values of TTVmax for each of the five solutions. The best TRADES solution is also that for which the maximum TTV signal is higher than the others and in better agreement with the observations. Very interestingly, the high eccentricity of planet c gives a direct explanation of the modulation of the TTV of planet b: when planet c moves close to planet b, it gravitationally kicks its smaller sibling, causing the maximum in the TTV signal; after the conjunction, planet b re-circularizes own orbit very slowly, because of tidal interactions with the star and planet c moving far from it. This causes the decreasing of amplitude of its TTV that we see in Fig 8. We again stress that all the values presented in Table 6 should be taken with caution, because the existence of planet c can be definitively constrained only by new tran- sit observations, able to completely cover the TTV phase of planet b. In particular, one should verify if there is a periodic repetition of the peak and the modulation. 5. Discussion and conclusions Thanks to precise RV measurements obtained with the high-resolution spectrograph CAFE, we confirmed the planetary nature of Kepler-539 b, a dense Jupiter-like planet with a mass of 0.97 ± 0.29 MJup and a radius of 0.747 ± 0.018 RJup, revolving with a period of 125.6 days on a circular orbit around a G2 V star, similar to the Sun. The parameters of the parent star and the planet b were ob- tained by analysing the CAFE spectra and through a joint fit to the RV data and Kepler transit light curves. Orbiting at ∼ 0.5 au from its host, Kepler-539 b is located not so far (≈ 0.45 au) from its habitable zone (HZ), which has a width of ∼ 0.72 au, as we estimated it based on the stel- lar parameters reported in Table 3 and the HZ calculator by Kopparapu et al. (2013, 2014). The parent star is quite active, as is shown by the 0.047 mag peak-to-peak modu- lation present in the long time-series photometry and from the star-spot anomalies clearly visible in two of the transit events monitored by Kepler. Even though it has physical characteristics that resemble those of the Sun, Kepler-539 is much younger: 1.0±0.3 Gyr as estimated from its lithium abundance and gyrochronology. An analysis of the Kepler mid-transit times of Kepler- 539 b revealed a clear modulated TTV signal. We studied this with TRADES, a code able to make a simultaneous fit of RV and mid-transit times and compare the results with simulated data of various multi-planetary configura- tions. We found that the TTV can be explained through the presence of a 1.2–3.6 MJup Jovian-like planet c on a very eccentric (e = 0.43–0.61) and wider orbit than that of Kepler-539 b, with a period larger than 1000 days. Each of the five solutions that we have presented are related to a planetary system which is stable according to the results of an FMA analysis. Since the orbit of Kepler-539 c is highly eccentric, then the modulation of the TTV signal of Kepler-539 b can be easily explained. As the distance of planet c from the parent star varies with time, there is a periodic change of the mid- transit time of planet b. In particular, when the planet c is in conjunction with planet b, the gravitational interaction 8 aeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeae-5-4-3-2-10123456-40-200204060CyclenumberO-CminL. Mancini et al.: The planetary system Kepler-539 Table 6: Left-hand columns: parameters of Kepler-539 c from the best fit of the Kepler mid-transit times and CAFE-RV measurements. The best five solutions obtained with TRADES are shown. Left-hand columns: parameters determined from the previous. PSO PIKAIA Algortihm Mc ( MJup) 3.60+0.30−0.23 1.27+0.05−0.05 1.20+0.11−0.11 2.72+0.15−0.19 2.28+0.15−0.16 PIKAIA PIKAIA PSO Pc (day) 1040+30−19 1705+70−49 945+7−6 1784+66−59 963+29−27 νc (◦) 33.95+0.76−1.02 23.13+0.84−0.72 18.00+0.57−0.60 349.14+0.50−0.57 355.50+0.34−0.38 ec 0.432+0.010 −0.006 0.605+0.009 −0.005 0.448+0.010 −0.010 0.563+0.010 −0.010 0.457+0.009 −0.009 ωc (◦) 270.67+0.05−0.06 166.27+0.83−0.86 329.27+2.69−2.54 296.56+2.17−1.70 343.54+1.34−1.05 χ2 ν 34.2 39.5 57.2 62.9 66.6 a (au) 2.04 ± 0.04 2.84 ± 0.08 1.91 ± 0.02 2.92 ± 0.08 1.94 ± 0.05 Teq (K) TTVmax (min) 271 ± 4 230 ± 4 280 ± 3 226 ± 4 278 ± 4 31.0 ± 3.6 9.3 ± 1.1 11.8 ± 1.4 17.8 ± 2.2 22.4 ± 2.7 Notes. We used two different algorithm: the PIKAIA algorithm (Charbonneau 1995) and the Particle Swarm Optimization (PSO) algorithm (Tada 2007), respectively. The first five columns refer to the values of the parameters determined from the fitting processes, while the others values have been determined using the previous ones. In particular, δt represents the maximum TTV for the inner planet b. Fig. 9: The period–planetary mass diagram for transiting planets in the mass range 0.3 MJup < Mp < 11 MJup. The size of each circle is proportional to the corresponding planetary radius, while the colour indicates equilibrium temperature (data taken from TEPCat). The dashed line demarcates the high-mass regime (Mp > 2MJup). The positions of Kepler- 539 b and Kepler-539 c are highlighted. Kepler-539 c is marked with a box since we do not know its radius. The error bars have been suppressed for clarity. between the two planets is maximum and we see a large amplitude of the TTV signal. Then, when planet c moves away, planet b re-circularizes its orbit very slowly and we see the gradual decreasing of the amplitude of its TTV. Given its relatively young age, the physical parame- ters of the Kepler-539 planetary system can be very pre- cious for astrophysicists working on theories of planet for- mation and evolution. If we visualise the known transit- ing planets, reported in TEPCat5, in a period–mass dia- gram, we can see that Kepler-539 b and Kepler-539 c occupy sparsely-populated regions of this plot (here for the mass of Kepler-539 c we have considered the best value found by TRADES). We also underline that the transits of Kepler-539 b are quite deep (0.8%) and that the host star is relatively bright (V = 12.5 mag) and therefore amenable for exoplanet at- mosphere studies. The star-planet distance is large enough to consider Kepler-539 b as a non-inflated planet with a low irradiation level. Since there are only a handful of transiting 5 TEPcat (Transiting Extrasolar Planet Catalogue) is avail- able at www.astro.keele.ac.uk/jkt/tepcat/ (Southworth 2011). Jupiter-like planets with a low irradiation level, this plane- tary system could be a precious target for future studies of exoplanet atmospheres of normal giant gas planets. Unfortunately, the parameters of Kepler-539 c are not well constrained as they are based on a poorly-sampled TTV (only 12 transit timings). More transits of Kepler- 539 b and more precise RV observations with a larger- aperture telescope are required to properly model its TTV signal and accurately determine the existence and the phys- ical properties of Kepler-539 c on a quite solid ground. While there are now several high-resolution spectrographs with RV performance better than CAFE, the observations of new transits of Kepler-539 b are not straightforward with ground-based facilities. The 125.6-day orbital period and the long transit duration (9.31 hr) put serious limitations for the successful of a such long-term monitoring and re- quest a large observational effort with various telescopes located at different Earth's longitudes. Furthermore, the use of different telescopes could compromise the precision of the measurements of T0 at the end. On the other hand, the observations of new mid-transit times of Kepler-539 b 9 L. Mancini et al.: The planetary system Kepler-539 McCullough, P. R., Stys, J. E., Valenti, J. A., et al. 2005, PASP, 117, 783 McQuillan, A., Mazeh, T., Aigrain, S. 2013, ApJ, 775, L11 Oshagh, M., Santos, N. C., Boisse, I., et al. 2013, A&A, 556, A19 Pace, G., Castro, M., Mel´endez, J., et al. 2012, A&A, 541, A150 Pepper, J., Pogge, R. W., DePoy, D. L., et al. 2007, PASP, 119, 923 Pollacco, D. L., Skillen, I., Collier Cameron, A., et al. 2006, PASP, 118, 1407 Sestito, P., Randich, S., Pallavicini, R. 2004, A&A, 426, 809 Sestito, P., Randich, S. 2005, A&A, 442, 615 Sing, D. 2010, A&A, 510, A21 Soderblom, D. R., Jones, B. F., Balachandran, S., et al. 1993, AJ, 106, 1059 Southworth, J. 2008, MNRAS, 386, 1644 Southworth, J. 2011, MNRAS, 417, 2166 Southworth, J. 2012, MNRAS, 426, 1291 Southworth, J. 2013, A&A, 557, A119 Straizys, V. & Kuriliene, G. 1981, Ap&SS, 80, 353 Tada, T. 2007, Journal of Japan Society of Hydrology & Water Resources, 20, 450 Takeda, Y., Honda S., Ohnishi, T., et al. 2013, PASJ, 65, 53 Tregloan-Reed, J., Southworth, J., Tappert, C. 2013, MNRAS, 428, 3671 Tregloan-Reed, J., Southworth, J., Burgdorf, M. 2015, MNRAS, 450, 1760 Walkowicz, L. M., & Basri, G. S. 2013a, MNRAS, 436, 1883 Walkowicz, L. M., & Basri, G. S. 2013b, VizieR On-line Data Catalog: J/MNRAS/436/1883 Appendix A: Starspot parameters In this appendix we report a Table containing the starspot parameters that were determined from the prism+gemc fitting of the transit light curves at cycles −5 and 2, as defined in Sect. 2.1. can be easily performed by the incoming space telescope CHEOPS. This facility is therefore highly recommended for a new detailed follow-up study of the Kepler-539 planetary system. Acknowledgements. This paper is based on observations collected with the 2.2 m Telescope at the Centro Astron´omico Hispano Alem´an (CAHA) in Calar Alto (Spain) and the publicly available data ob- tained with the NASA space satellite Kepler. Operations at the Calar Alto telescopes are jointly performed by the Max-Planck- Institut fur Astronomie (MPIA) and the Instituto de Astrof´ısica de Andaluc´ıa (CSIC). This research has been partially funded by Spanish grant AYA2012-38897-C02-01. J.L.-B. thanks the CSIC JAE-predoc program for Ph.D. fellowship. R.B. is supported by CONICYT- PCHA/Doctorado Nacional. R.B. acknowledges additional support from project IC120009 "Millenium Institute of Astrophysics (MAS)" of the Millennium Science Initiative, Chilean Ministry of Economy. We wish to thank Ennio Poretti for very helpful comments. We ac- knowledge the use of the following internet-based resources: the ESO Digitized Sky Survey; the TEPCat catalogue; the SIMBAD data base operated at CDS, Strasbourg, France; and the arXiv scientific paper preprint service operated by Cornell University. References Aceituno, J., Sanchez, S. F., Grupp, F., et al. 2013, A&A, 552, A31 Adams, E. R., Ciardi, D. R., Dupree, A. K., et al. 2012, AJ, 144, 42 Alonso, R., Brown, T. M., Torres, G., et al. 2004, ApJ, 613, 153 Alsubai, K. A., Parley, N. R., Bramich, D. M., et al. 2013, AcA, 63, 465 Bakos, G. ´A., Noyes, R. W., Kov´acs, G., et al. 2004, PASP, 116, 266 Bakos, G. ´A., Csubry, Z., Penev, K., et al. 2013, PASP, 125, 154 Baranne, A., Queloz, D., Mayor, M., et al. 1996, A&AS, 119, 373 Barge, P., Baglin, A., Auvergne, M., et al. 2008, A&A, 482, L17 Barnes, S. A. 2010, ApJ, 722, 222 Barnes, S. A. & Kim, Y.-C. 2010, ApJ, 721, 675 Barros, S. C. C., Bou´e, G., Gibson, N. P., et al. 2013, MNRAS, 430, 3032 Blackwell, D. E., Shallis M. J. 1979, MNRAS, 186, 673 Borsato, L., Marzari, F., Nascimbeni, V., et al. 2014, A&A, 571, A38 Borucki, W. J., Koch, D., Basri, G., et al. 2011, ApJ, 736, 19 Charbonneau, P. 1995, ApJS, 101, 309 Ciceri, S., Lillo-Box, J., Southworth, J., et al. 2015, A&A, 573, L5 Doyle, A. P., Davies, G. R., Smalley, B., et al. 2014, MNRAS, 444, 3592 Figueira, P., Faria, J. P., Adibekyan, V. Z, et al. 2016, proceedings of the conference Habitability in the Universe: From the Early Earth to Exoplanets, arXiv:1601.05107 Gandolfi, D., Parviainen, H., Fridlund, M. et al. 2013, A&A, 557, AA74 Gandolfi, D., Parviainen, H., Deeg, H. J., et al. 2015, A&A, 576, A11 Gray R. O., Corbally C. J. 1994, AJ, 107, 742 Grevesse, N., Sauval, A. J. 1998, SSRv, 85, 161 Haswell C. A. 2010, Transiting Exoplanets, Cambridge University Press Ioannidis, P., Huber, K. F., Schmitt, J. H. M. M. 2016, A&A, 585, A72 Kopparapu, R. K., Ramirez, R., Kasting, J. F., et al. 2013, ApJ, 765, 131 Kopparapu, R. K., Ramirez, R., SchottelKotte, J., et al. 2014, ApJ, 787, L29 Kurucz R. L. 1979, ApJS, 40, 1 Laskar, J., Froeschl´e, C., Celletti, A. 1992, Physica D, 56, 253 Lillo-Box, J., Barrado, D., Bouy, H. 2012, A&A, 546, A10 Lillo-Box, J., Barrado, D., Moya, A., et al. 2014a, A&A, 562, A109 Lillo-Box, J., Barrado, D., Henning, Th., et al. 2014b, A&A, 568, L1 Lillo-Box, J., Barrado, D., Bouy, H. 2014c, A&A, 566, A103 Lillo-Box, J., Barrado, D., Mancini, L., et al. 2015a, A&A, 576, A88 Lillo-Box, J., Barrado, D., Santos, N. C., et al. 2015b, A&A, 577, A105 Lind, K., Asplund, M., Barklem, P. S. 2009, A&A, 503, 541 Mancini, L., Southworth, J., Ciceri, S., et al. 2013a, A&A, 551, A11 Mancini, L., Ciceri, S., Chen, G., et al. 2013b, MNRAS, 436, 2 Mancini, L., Southworth, J., Ciceri, S., et al. 2014a, A&A, 562, A126 Mancini, L., Southworth, J., Ciceri, S., et al. 2014b, A&A, 568, A127 Mazeh, T., Holczer, T., Shporer, A. 2015, ApJ, 800, 142 10 L. Mancini et al.: The planetary system Kepler-539 Table A.1: Starspot parameters derived from the prism+gemc fitting of the transit light curves at cycles −5 and 2. aThe longitude of the centre of the spot is defined to be 0◦ at the centre of the stellar disc and can vary from −90◦ to 90◦. bThe co-latitude of the centre of the spot is defined to be 0◦ at the north pole and 180◦ at the south pole. cAngular radius of the starspot (note that an angular radius of 90◦ covers half of stellar surface). dSpot contrast (note that 1.0 equals the brightness of the surrounding photosphere). Cycle −5 2 θ(◦) a −26.18 ± 3.73 18.88 ± 11.79 φ(◦) b 28.31 ± 10.14 34.27 ± 8.21 rspot(◦) c 36.43 ± 6.80 18.60 ± 6.46 d ρspot 0.69 ± 0.11 0.70 ± 0.28 11 L. Mancini et al.: The planetary system Kepler-539 Appendix B: TRADES simulations In this appendix we report plots based on the results of the TRADES simulations for modelling the TTV detected in the transit timings of KOI-327 b. In particular, the figures show RV plots (bottom panels) and O–C diagrams from lin- ear ephemeris for planet Kepler-539 b (top panels). In both cases, the observed data are compared with those obtained from the simulations (see Sect. 4) The figures are ordered as in Table 6. Figures 1, 3 and 4 were obtained by using the PIKAIA algorithm (Charbonneau 1995), while figures 2 and 5 with the Particle Swarm Optimization (PSO) algo- rithm (Tada 2007). Fig. B.2: Same as for Fig B.1, but showing the 2nd best so- lution found by trades. Fig. B.3: Same as for Fig B.1, but showing the 3rd best so- lution found by trades. Fig. B.1: O–C diagram (with residuals) from linear ephemeris for planet Kepler-539 b (top panel); observations plotted as black points (with error bars), simulations plot- ted as open blue circles. Bottom panel shows the RV ob- servations as solid black circles, simulations at the same BJD(UTC) as open blue circles, and the dotted gray line is the RV model for the whole simulation. This the best solution found by trades. 12 O-C [m]-40.0-20.00.020.040.060.0observationssimulationsO-C [d]-0.0200.0000.0200.040res [m]-20.0-10.00.010.020.0res [d]-0.015-0.010-0.0050.0000.0050.0100.015N (planet b)−4−3−2−101234567RV modelRV [m/s]-200.00-100.000.00100.00res [m/s]-100.000.00100.00BJD - 2454000.200022502500275030003250O-C [m]-40.0-20.00.020.040.060.0observationssimulationsO-C [d]-0.0200.0000.0200.040res [m]-15.0-10.0-5.00.05.010.015.0res [d]-0.010-0.0050.0000.0050.010N (planet b)−5−4−3−2−1012345678RV modelRV [m/s]-200.00-100.000.00100.00res [m/s]-100.000.00100.00BJD - 2454000.200022502500275030003250O-C [m]-40.0-20.00.020.040.060.0observationssimulationsO-C [d]-0.0200.0000.0200.040res [m]-30.0-20.0-10.00.010.0res [d]-0.020-0.0100.0000.010N (planet b)−4−3−2−101234567RV modelRV [m/s]-200.00-100.000.00100.00res [m/s]-100.000.00100.00BJD - 2454000.200022502500275030003250L. Mancini et al.: The planetary system Kepler-539 Fig. B.4: Same as for Fig B.1, but showing the 4th best so- lution found by trades. Fig. B.5: Same as for Fig B.1, but showing the 5th best so- lution found by trades. 13 O-C [m]-50.0-25.00.025.050.0observationssimulationsO-C [d]-0.0200.0000.0200.040res [m]-10.00.010.020.030.040.0res [d]-0.0100.0000.0100.0200.030N (planet b)−5−4−3−2−1012345678RV modelRV [m/s]-200.00-100.000.00100.00res [m/s]-100.000.00100.00BJD - 2454000.200022502500275030003250O-C [m]-40.0-20.00.020.040.060.0observationssimulationsO-C [d]-0.0200.0000.0200.040res [m]-20.00.020.040.0res [d]-0.0100.0000.0100.020N (planet b)−4−3−2−101234567RV modelRV [m/s]-200.00-100.000.00100.00res [m/s]-200.00-100.000.00100.00BJD - 2454000.200022502500275030003250

Subsets and Splits